linux/drivers/md/dm.c

3245 lines
75 KiB
C
Raw Normal View History

/*
* Copyright (C) 2001, 2002 Sistina Software (UK) Limited.
* Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
*
* This file is released under the GPL.
*/
#include "dm-core.h"
#include "dm-rq.h"
#include "dm-uevent.h"
#include <linux/init.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/sched/signal.h>
#include <linux/blkpg.h>
#include <linux/bio.h>
#include <linux/mempool.h>
#include <linux/dax.h>
#include <linux/slab.h>
#include <linux/idr.h>
#include <linux/uio.h>
#include <linux/hdreg.h>
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
#include <linux/delay.h>
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
#include <linux/wait.h>
#include <linux/pr.h>
#include <linux/refcount.h>
tracing/events: convert block trace points to TRACE_EVENT() TRACE_EVENT is a more generic way to define tracepoints. Doing so adds these new capabilities to this tracepoint: - zero-copy and per-cpu splice() tracing - binary tracing without printf overhead - structured logging records exposed under /debug/tracing/events - trace events embedded in function tracer output and other plugins - user-defined, per tracepoint filter expressions ... Cons: - no dev_t info for the output of plug, unplug_timer and unplug_io events. no dev_t info for getrq and sleeprq events if bio == NULL. no dev_t info for rq_abort,...,rq_requeue events if rq->rq_disk == NULL. This is mainly because we can't get the deivce from a request queue. But this may change in the future. - A packet command is converted to a string in TP_assign, not TP_print. While blktrace do the convertion just before output. Since pc requests should be rather rare, this is not a big issue. - In blktrace, an event can have 2 different print formats, but a TRACE_EVENT has a unique format, which means we have some unused data in a trace entry. The overhead is minimized by using __dynamic_array() instead of __array(). I've benchmarked the ioctl blktrace vs the splice based TRACE_EVENT tracing: dd dd + ioctl blktrace dd + TRACE_EVENT (splice) 1 7.36s, 42.7 MB/s 7.50s, 42.0 MB/s 7.41s, 42.5 MB/s 2 7.43s, 42.3 MB/s 7.48s, 42.1 MB/s 7.43s, 42.4 MB/s 3 7.38s, 42.6 MB/s 7.45s, 42.2 MB/s 7.41s, 42.5 MB/s So the overhead of tracing is very small, and no regression when using those trace events vs blktrace. And the binary output of TRACE_EVENT is much smaller than blktrace: # ls -l -h -rw-r--r-- 1 root root 8.8M 06-09 13:24 sda.blktrace.0 -rw-r--r-- 1 root root 195K 06-09 13:24 sda.blktrace.1 -rw-r--r-- 1 root root 2.7M 06-09 13:25 trace_splice.out Following are some comparisons between TRACE_EVENT and blktrace: plug: kjournald-480 [000] 303.084981: block_plug: [kjournald] kjournald-480 [000] 303.084981: 8,0 P N [kjournald] unplug_io: kblockd/0-118 [000] 300.052973: block_unplug_io: [kblockd/0] 1 kblockd/0-118 [000] 300.052974: 8,0 U N [kblockd/0] 1 remap: kjournald-480 [000] 303.085042: block_remap: 8,0 W 102736992 + 8 <- (8,8) 33384 kjournald-480 [000] 303.085043: 8,0 A W 102736992 + 8 <- (8,8) 33384 bio_backmerge: kjournald-480 [000] 303.085086: block_bio_backmerge: 8,0 W 102737032 + 8 [kjournald] kjournald-480 [000] 303.085086: 8,0 M W 102737032 + 8 [kjournald] getrq: kjournald-480 [000] 303.084974: block_getrq: 8,0 W 102736984 + 8 [kjournald] kjournald-480 [000] 303.084975: 8,0 G W 102736984 + 8 [kjournald] bash-2066 [001] 1072.953770: 8,0 G N [bash] bash-2066 [001] 1072.953773: block_getrq: 0,0 N 0 + 0 [bash] rq_complete: konsole-2065 [001] 300.053184: block_rq_complete: 8,0 W () 103669040 + 16 [0] konsole-2065 [001] 300.053191: 8,0 C W 103669040 + 16 [0] ksoftirqd/1-7 [001] 1072.953811: 8,0 C N (5a 00 08 00 00 00 00 00 24 00) [0] ksoftirqd/1-7 [001] 1072.953813: block_rq_complete: 0,0 N (5a 00 08 00 00 00 00 00 24 00) 0 + 0 [0] rq_insert: kjournald-480 [000] 303.084985: block_rq_insert: 8,0 W 0 () 102736984 + 8 [kjournald] kjournald-480 [000] 303.084986: 8,0 I W 102736984 + 8 [kjournald] Changelog from v2 -> v3: - use the newly introduced __dynamic_array(). Changelog from v1 -> v2: - use __string() instead of __array() to minimize the memory required to store hex dump of rq->cmd(). - support large pc requests. - add missing blk_fill_rwbs_rq() in block_rq_requeue TRACE_EVENT. - some cleanups. Signed-off-by: Li Zefan <lizf@cn.fujitsu.com> LKML-Reference: <4A2DF669.5070905@cn.fujitsu.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2009-06-09 05:43:05 +00:00
#define DM_MSG_PREFIX "core"
/*
* Cookies are numeric values sent with CHANGE and REMOVE
* uevents while resuming, removing or renaming the device.
*/
#define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE"
#define DM_COOKIE_LENGTH 24
static const char *_name = DM_NAME;
static unsigned int major = 0;
static unsigned int _major = 0;
static DEFINE_IDR(_minor_idr);
static DEFINE_SPINLOCK(_minor_lock);
static void do_deferred_remove(struct work_struct *w);
static DECLARE_WORK(deferred_remove_work, do_deferred_remove);
static struct workqueue_struct *deferred_remove_workqueue;
atomic_t dm_global_event_nr = ATOMIC_INIT(0);
DECLARE_WAIT_QUEUE_HEAD(dm_global_eventq);
void dm_issue_global_event(void)
{
atomic_inc(&dm_global_event_nr);
wake_up(&dm_global_eventq);
}
/*
* One of these is allocated (on-stack) per original bio.
*/
struct clone_info {
struct dm_table *map;
struct bio *bio;
struct dm_io *io;
sector_t sector;
unsigned sector_count;
};
/*
* One of these is allocated per clone bio.
*/
#define DM_TIO_MAGIC 7282014
struct dm_target_io {
unsigned magic;
struct dm_io *io;
struct dm_target *ti;
unsigned target_bio_nr;
unsigned *len_ptr;
bool inside_dm_io;
struct bio clone;
};
/*
* One of these is allocated per original bio.
* It contains the first clone used for that original.
*/
#define DM_IO_MAGIC 5191977
struct dm_io {
unsigned magic;
struct mapped_device *md;
blk_status_t status;
atomic_t io_count;
struct bio *orig_bio;
unsigned long start_time;
spinlock_t endio_lock;
struct dm_stats_aux stats_aux;
/* last member of dm_target_io is 'struct bio' */
struct dm_target_io tio;
};
void *dm_per_bio_data(struct bio *bio, size_t data_size)
{
struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
if (!tio->inside_dm_io)
return (char *)bio - offsetof(struct dm_target_io, clone) - data_size;
return (char *)bio - offsetof(struct dm_target_io, clone) - offsetof(struct dm_io, tio) - data_size;
}
EXPORT_SYMBOL_GPL(dm_per_bio_data);
struct bio *dm_bio_from_per_bio_data(void *data, size_t data_size)
{
struct dm_io *io = (struct dm_io *)((char *)data + data_size);
if (io->magic == DM_IO_MAGIC)
return (struct bio *)((char *)io + offsetof(struct dm_io, tio) + offsetof(struct dm_target_io, clone));
BUG_ON(io->magic != DM_TIO_MAGIC);
return (struct bio *)((char *)io + offsetof(struct dm_target_io, clone));
}
EXPORT_SYMBOL_GPL(dm_bio_from_per_bio_data);
unsigned dm_bio_get_target_bio_nr(const struct bio *bio)
{
return container_of(bio, struct dm_target_io, clone)->target_bio_nr;
}
EXPORT_SYMBOL_GPL(dm_bio_get_target_bio_nr);
#define MINOR_ALLOCED ((void *)-1)
/*
* Bits for the md->flags field.
*/
#define DMF_BLOCK_IO_FOR_SUSPEND 0
#define DMF_SUSPENDED 1
#define DMF_FROZEN 2
#define DMF_FREEING 3
#define DMF_DELETING 4
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
#define DMF_NOFLUSH_SUSPENDING 5
#define DMF_DEFERRED_REMOVE 6
#define DMF_SUSPENDED_INTERNALLY 7
#define DM_NUMA_NODE NUMA_NO_NODE
static int dm_numa_node = DM_NUMA_NODE;
/*
* For mempools pre-allocation at the table loading time.
*/
struct dm_md_mempools {
struct bio_set bs;
struct bio_set io_bs;
};
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
struct table_device {
struct list_head list;
refcount_t count;
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
struct dm_dev dm_dev;
};
static struct kmem_cache *_rq_tio_cache;
static struct kmem_cache *_rq_cache;
/*
* Bio-based DM's mempools' reserved IOs set by the user.
*/
#define RESERVED_BIO_BASED_IOS 16
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;
static int __dm_get_module_param_int(int *module_param, int min, int max)
{
locking/atomics: COCCINELLE/treewide: Convert trivial ACCESS_ONCE() patterns to READ_ONCE()/WRITE_ONCE() Please do not apply this to mainline directly, instead please re-run the coccinelle script shown below and apply its output. For several reasons, it is desirable to use {READ,WRITE}_ONCE() in preference to ACCESS_ONCE(), and new code is expected to use one of the former. So far, there's been no reason to change most existing uses of ACCESS_ONCE(), as these aren't harmful, and changing them results in churn. However, for some features, the read/write distinction is critical to correct operation. To distinguish these cases, separate read/write accessors must be used. This patch migrates (most) remaining ACCESS_ONCE() instances to {READ,WRITE}_ONCE(), using the following coccinelle script: ---- // Convert trivial ACCESS_ONCE() uses to equivalent READ_ONCE() and // WRITE_ONCE() // $ make coccicheck COCCI=/home/mark/once.cocci SPFLAGS="--include-headers" MODE=patch virtual patch @ depends on patch @ expression E1, E2; @@ - ACCESS_ONCE(E1) = E2 + WRITE_ONCE(E1, E2) @ depends on patch @ expression E; @@ - ACCESS_ONCE(E) + READ_ONCE(E) ---- Signed-off-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: davem@davemloft.net Cc: linux-arch@vger.kernel.org Cc: mpe@ellerman.id.au Cc: shuah@kernel.org Cc: snitzer@redhat.com Cc: thor.thayer@linux.intel.com Cc: tj@kernel.org Cc: viro@zeniv.linux.org.uk Cc: will.deacon@arm.com Link: http://lkml.kernel.org/r/1508792849-3115-19-git-send-email-paulmck@linux.vnet.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-10-23 21:07:29 +00:00
int param = READ_ONCE(*module_param);
int modified_param = 0;
bool modified = true;
if (param < min)
modified_param = min;
else if (param > max)
modified_param = max;
else
modified = false;
if (modified) {
(void)cmpxchg(module_param, param, modified_param);
param = modified_param;
}
return param;
}
unsigned __dm_get_module_param(unsigned *module_param,
unsigned def, unsigned max)
{
locking/atomics: COCCINELLE/treewide: Convert trivial ACCESS_ONCE() patterns to READ_ONCE()/WRITE_ONCE() Please do not apply this to mainline directly, instead please re-run the coccinelle script shown below and apply its output. For several reasons, it is desirable to use {READ,WRITE}_ONCE() in preference to ACCESS_ONCE(), and new code is expected to use one of the former. So far, there's been no reason to change most existing uses of ACCESS_ONCE(), as these aren't harmful, and changing them results in churn. However, for some features, the read/write distinction is critical to correct operation. To distinguish these cases, separate read/write accessors must be used. This patch migrates (most) remaining ACCESS_ONCE() instances to {READ,WRITE}_ONCE(), using the following coccinelle script: ---- // Convert trivial ACCESS_ONCE() uses to equivalent READ_ONCE() and // WRITE_ONCE() // $ make coccicheck COCCI=/home/mark/once.cocci SPFLAGS="--include-headers" MODE=patch virtual patch @ depends on patch @ expression E1, E2; @@ - ACCESS_ONCE(E1) = E2 + WRITE_ONCE(E1, E2) @ depends on patch @ expression E; @@ - ACCESS_ONCE(E) + READ_ONCE(E) ---- Signed-off-by: Mark Rutland <mark.rutland@arm.com> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: davem@davemloft.net Cc: linux-arch@vger.kernel.org Cc: mpe@ellerman.id.au Cc: shuah@kernel.org Cc: snitzer@redhat.com Cc: thor.thayer@linux.intel.com Cc: tj@kernel.org Cc: viro@zeniv.linux.org.uk Cc: will.deacon@arm.com Link: http://lkml.kernel.org/r/1508792849-3115-19-git-send-email-paulmck@linux.vnet.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org>
2017-10-23 21:07:29 +00:00
unsigned param = READ_ONCE(*module_param);
unsigned modified_param = 0;
if (!param)
modified_param = def;
else if (param > max)
modified_param = max;
if (modified_param) {
(void)cmpxchg(module_param, param, modified_param);
param = modified_param;
}
return param;
}
unsigned dm_get_reserved_bio_based_ios(void)
{
return __dm_get_module_param(&reserved_bio_based_ios,
RESERVED_BIO_BASED_IOS, DM_RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);
static unsigned dm_get_numa_node(void)
{
return __dm_get_module_param_int(&dm_numa_node,
DM_NUMA_NODE, num_online_nodes() - 1);
}
static int __init local_init(void)
{
int r = -ENOMEM;
_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
if (!_rq_tio_cache)
return r;
_rq_cache = kmem_cache_create("dm_old_clone_request", sizeof(struct request),
__alignof__(struct request), 0, NULL);
if (!_rq_cache)
goto out_free_rq_tio_cache;
r = dm_uevent_init();
if (r)
goto out_free_rq_cache;
deferred_remove_workqueue = alloc_workqueue("kdmremove", WQ_UNBOUND, 1);
if (!deferred_remove_workqueue) {
r = -ENOMEM;
goto out_uevent_exit;
}
_major = major;
r = register_blkdev(_major, _name);
if (r < 0)
goto out_free_workqueue;
if (!_major)
_major = r;
return 0;
out_free_workqueue:
destroy_workqueue(deferred_remove_workqueue);
out_uevent_exit:
dm_uevent_exit();
out_free_rq_cache:
kmem_cache_destroy(_rq_cache);
out_free_rq_tio_cache:
kmem_cache_destroy(_rq_tio_cache);
return r;
}
static void local_exit(void)
{
flush_scheduled_work();
destroy_workqueue(deferred_remove_workqueue);
kmem_cache_destroy(_rq_cache);
kmem_cache_destroy(_rq_tio_cache);
unregister_blkdev(_major, _name);
dm_uevent_exit();
_major = 0;
DMINFO("cleaned up");
}
static int (*_inits[])(void) __initdata = {
local_init,
dm_target_init,
dm_linear_init,
dm_stripe_init,
dm_io_init,
dm_kcopyd_init,
dm_interface_init,
dm_statistics_init,
};
static void (*_exits[])(void) = {
local_exit,
dm_target_exit,
dm_linear_exit,
dm_stripe_exit,
dm_io_exit,
dm_kcopyd_exit,
dm_interface_exit,
dm_statistics_exit,
};
static int __init dm_init(void)
{
const int count = ARRAY_SIZE(_inits);
int r, i;
for (i = 0; i < count; i++) {
r = _inits[i]();
if (r)
goto bad;
}
return 0;
bad:
while (i--)
_exits[i]();
return r;
}
static void __exit dm_exit(void)
{
int i = ARRAY_SIZE(_exits);
while (i--)
_exits[i]();
/*
* Should be empty by this point.
*/
idr_destroy(&_minor_idr);
}
/*
* Block device functions
*/
int dm_deleting_md(struct mapped_device *md)
{
return test_bit(DMF_DELETING, &md->flags);
}
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
{
struct mapped_device *md;
spin_lock(&_minor_lock);
md = bdev->bd_disk->private_data;
if (!md)
goto out;
if (test_bit(DMF_FREEING, &md->flags) ||
dm_deleting_md(md)) {
md = NULL;
goto out;
}
dm_get(md);
atomic_inc(&md->open_count);
out:
spin_unlock(&_minor_lock);
return md ? 0 : -ENXIO;
}
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
{
struct mapped_device *md;
spin_lock(&_minor_lock);
md = disk->private_data;
if (WARN_ON(!md))
goto out;
if (atomic_dec_and_test(&md->open_count) &&
(test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
queue_work(deferred_remove_workqueue, &deferred_remove_work);
dm_put(md);
out:
spin_unlock(&_minor_lock);
}
int dm_open_count(struct mapped_device *md)
{
return atomic_read(&md->open_count);
}
/*
* Guarantees nothing is using the device before it's deleted.
*/
int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
{
int r = 0;
spin_lock(&_minor_lock);
if (dm_open_count(md)) {
r = -EBUSY;
if (mark_deferred)
set_bit(DMF_DEFERRED_REMOVE, &md->flags);
} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
r = -EEXIST;
else
set_bit(DMF_DELETING, &md->flags);
spin_unlock(&_minor_lock);
return r;
}
int dm_cancel_deferred_remove(struct mapped_device *md)
{
int r = 0;
spin_lock(&_minor_lock);
if (test_bit(DMF_DELETING, &md->flags))
r = -EBUSY;
else
clear_bit(DMF_DEFERRED_REMOVE, &md->flags);
spin_unlock(&_minor_lock);
return r;
}
static void do_deferred_remove(struct work_struct *w)
{
dm_deferred_remove();
}
sector_t dm_get_size(struct mapped_device *md)
{
return get_capacity(md->disk);
}
struct request_queue *dm_get_md_queue(struct mapped_device *md)
{
return md->queue;
}
struct dm_stats *dm_get_stats(struct mapped_device *md)
{
return &md->stats;
}
static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
struct mapped_device *md = bdev->bd_disk->private_data;
return dm_get_geometry(md, geo);
}
static int dm_blk_report_zones(struct gendisk *disk, sector_t sector,
struct blk_zone *zones, unsigned int *nr_zones,
gfp_t gfp_mask)
{
#ifdef CONFIG_BLK_DEV_ZONED
struct mapped_device *md = disk->private_data;
struct dm_target *tgt;
struct dm_table *map;
int srcu_idx, ret;
if (dm_suspended_md(md))
return -EAGAIN;
map = dm_get_live_table(md, &srcu_idx);
if (!map)
return -EIO;
tgt = dm_table_find_target(map, sector);
if (!dm_target_is_valid(tgt)) {
ret = -EIO;
goto out;
}
/*
* If we are executing this, we already know that the block device
* is a zoned device and so each target should have support for that
* type of drive. A missing report_zones method means that the target
* driver has a problem.
*/
if (WARN_ON(!tgt->type->report_zones)) {
ret = -EIO;
goto out;
}
/*
* blkdev_report_zones() will loop and call this again to cover all the
* zones of the target, eventually moving on to the next target.
* So there is no need to loop here trying to fill the entire array
* of zones.
*/
ret = tgt->type->report_zones(tgt, sector, zones,
nr_zones, gfp_mask);
out:
dm_put_live_table(md, srcu_idx);
return ret;
#else
return -ENOTSUPP;
#endif
}
static int dm_prepare_ioctl(struct mapped_device *md, int *srcu_idx,
struct block_device **bdev)
__acquires(md->io_barrier)
{
struct dm_target *tgt;
struct dm_table *map;
int r;
retry:
r = -ENOTTY;
map = dm_get_live_table(md, srcu_idx);
if (!map || !dm_table_get_size(map))
return r;
/* We only support devices that have a single target */
if (dm_table_get_num_targets(map) != 1)
return r;
tgt = dm_table_get_target(map, 0);
if (!tgt->type->prepare_ioctl)
return r;
if (dm_suspended_md(md))
return -EAGAIN;
r = tgt->type->prepare_ioctl(tgt, bdev);
if (r == -ENOTCONN && !fatal_signal_pending(current)) {
dm_put_live_table(md, *srcu_idx);
msleep(10);
goto retry;
}
return r;
}
static void dm_unprepare_ioctl(struct mapped_device *md, int srcu_idx)
__releases(md->io_barrier)
{
dm_put_live_table(md, srcu_idx);
}
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg)
{
struct mapped_device *md = bdev->bd_disk->private_data;
int r, srcu_idx;
r = dm_prepare_ioctl(md, &srcu_idx, &bdev);
if (r < 0)
goto out;
if (r > 0) {
/*
* Target determined this ioctl is being issued against a
* subset of the parent bdev; require extra privileges.
*/
if (!capable(CAP_SYS_RAWIO)) {
DMWARN_LIMIT(
"%s: sending ioctl %x to DM device without required privilege.",
current->comm, cmd);
r = -ENOIOCTLCMD;
goto out;
}
}
r = __blkdev_driver_ioctl(bdev, mode, cmd, arg);
out:
dm_unprepare_ioctl(md, srcu_idx);
return r;
}
static void start_io_acct(struct dm_io *io);
static struct dm_io *alloc_io(struct mapped_device *md, struct bio *bio)
{
struct dm_io *io;
struct dm_target_io *tio;
struct bio *clone;
clone = bio_alloc_bioset(GFP_NOIO, 0, &md->io_bs);
if (!clone)
return NULL;
tio = container_of(clone, struct dm_target_io, clone);
tio->inside_dm_io = true;
tio->io = NULL;
io = container_of(tio, struct dm_io, tio);
io->magic = DM_IO_MAGIC;
io->status = 0;
atomic_set(&io->io_count, 1);
io->orig_bio = bio;
io->md = md;
spin_lock_init(&io->endio_lock);
start_io_acct(io);
return io;
}
static void free_io(struct mapped_device *md, struct dm_io *io)
{
bio_put(&io->tio.clone);
}
static struct dm_target_io *alloc_tio(struct clone_info *ci, struct dm_target *ti,
unsigned target_bio_nr, gfp_t gfp_mask)
{
struct dm_target_io *tio;
if (!ci->io->tio.io) {
/* the dm_target_io embedded in ci->io is available */
tio = &ci->io->tio;
} else {
struct bio *clone = bio_alloc_bioset(gfp_mask, 0, &ci->io->md->bs);
if (!clone)
return NULL;
tio = container_of(clone, struct dm_target_io, clone);
tio->inside_dm_io = false;
}
tio->magic = DM_TIO_MAGIC;
tio->io = ci->io;
tio->ti = ti;
tio->target_bio_nr = target_bio_nr;
return tio;
}
static void free_tio(struct dm_target_io *tio)
{
if (tio->inside_dm_io)
return;
bio_put(&tio->clone);
}
static bool md_in_flight_bios(struct mapped_device *md)
{
int cpu;
struct hd_struct *part = &dm_disk(md)->part0;
long sum = 0;
for_each_possible_cpu(cpu) {
sum += part_stat_local_read_cpu(part, in_flight[0], cpu);
sum += part_stat_local_read_cpu(part, in_flight[1], cpu);
}
return sum != 0;
}
static bool md_in_flight(struct mapped_device *md)
{
if (queue_is_mq(md->queue))
return blk_mq_queue_inflight(md->queue);
else
return md_in_flight_bios(md);
}
static void start_io_acct(struct dm_io *io)
{
struct mapped_device *md = io->md;
struct bio *bio = io->orig_bio;
io->start_time = jiffies;
generic_start_io_acct(md->queue, bio_op(bio), bio_sectors(bio),
&dm_disk(md)->part0);
if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
bio->bi_iter.bi_sector, bio_sectors(bio),
false, 0, &io->stats_aux);
}
static void end_io_acct(struct dm_io *io)
{
struct mapped_device *md = io->md;
struct bio *bio = io->orig_bio;
unsigned long duration = jiffies - io->start_time;
generic_end_io_acct(md->queue, bio_op(bio), &dm_disk(md)->part0,
io->start_time);
if (unlikely(dm_stats_used(&md->stats)))
dm_stats_account_io(&md->stats, bio_data_dir(bio),
bio->bi_iter.bi_sector, bio_sectors(bio),
true, duration, &io->stats_aux);
/* nudge anyone waiting on suspend queue */
if (unlikely(wq_has_sleeper(&md->wait)))
wake_up(&md->wait);
}
/*
* Add the bio to the list of deferred io.
*/
static void queue_io(struct mapped_device *md, struct bio *bio)
{
unsigned long flags;
spin_lock_irqsave(&md->deferred_lock, flags);
bio_list_add(&md->deferred, bio);
spin_unlock_irqrestore(&md->deferred_lock, flags);
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
queue_work(md->wq, &md->work);
}
/*
* Everyone (including functions in this file), should use this
* function to access the md->map field, and make sure they call
* dm_put_live_table() when finished.
*/
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
{
*srcu_idx = srcu_read_lock(&md->io_barrier);
return srcu_dereference(md->map, &md->io_barrier);
}
void dm_put_live_table(struct mapped_device *md, int srcu_idx) __releases(md->io_barrier)
{
srcu_read_unlock(&md->io_barrier, srcu_idx);
}
void dm_sync_table(struct mapped_device *md)
{
synchronize_srcu(&md->io_barrier);
synchronize_rcu_expedited();
}
/*
* A fast alternative to dm_get_live_table/dm_put_live_table.
* The caller must not block between these two functions.
*/
static struct dm_table *dm_get_live_table_fast(struct mapped_device *md) __acquires(RCU)
{
rcu_read_lock();
return rcu_dereference(md->map);
}
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
rcu_read_unlock();
}
static char *_dm_claim_ptr = "I belong to device-mapper";
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
/*
* Open a table device so we can use it as a map destination.
*/
static int open_table_device(struct table_device *td, dev_t dev,
struct mapped_device *md)
{
struct block_device *bdev;
int r;
BUG_ON(td->dm_dev.bdev);
bdev = blkdev_get_by_dev(dev, td->dm_dev.mode | FMODE_EXCL, _dm_claim_ptr);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
if (IS_ERR(bdev))
return PTR_ERR(bdev);
r = bd_link_disk_holder(bdev, dm_disk(md));
if (r) {
blkdev_put(bdev, td->dm_dev.mode | FMODE_EXCL);
return r;
}
td->dm_dev.bdev = bdev;
td->dm_dev.dax_dev = dax_get_by_host(bdev->bd_disk->disk_name);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
return 0;
}
/*
* Close a table device that we've been using.
*/
static void close_table_device(struct table_device *td, struct mapped_device *md)
{
if (!td->dm_dev.bdev)
return;
bd_unlink_disk_holder(td->dm_dev.bdev, dm_disk(md));
blkdev_put(td->dm_dev.bdev, td->dm_dev.mode | FMODE_EXCL);
put_dax(td->dm_dev.dax_dev);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
td->dm_dev.bdev = NULL;
td->dm_dev.dax_dev = NULL;
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
}
static struct table_device *find_table_device(struct list_head *l, dev_t dev,
fmode_t mode) {
struct table_device *td;
list_for_each_entry(td, l, list)
if (td->dm_dev.bdev->bd_dev == dev && td->dm_dev.mode == mode)
return td;
return NULL;
}
int dm_get_table_device(struct mapped_device *md, dev_t dev, fmode_t mode,
struct dm_dev **result) {
int r;
struct table_device *td;
mutex_lock(&md->table_devices_lock);
td = find_table_device(&md->table_devices, dev, mode);
if (!td) {
td = kmalloc_node(sizeof(*td), GFP_KERNEL, md->numa_node_id);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
if (!td) {
mutex_unlock(&md->table_devices_lock);
return -ENOMEM;
}
td->dm_dev.mode = mode;
td->dm_dev.bdev = NULL;
if ((r = open_table_device(td, dev, md))) {
mutex_unlock(&md->table_devices_lock);
kfree(td);
return r;
}
format_dev_t(td->dm_dev.name, dev);
refcount_set(&td->count, 1);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
list_add(&td->list, &md->table_devices);
} else {
refcount_inc(&td->count);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
}
mutex_unlock(&md->table_devices_lock);
*result = &td->dm_dev;
return 0;
}
EXPORT_SYMBOL_GPL(dm_get_table_device);
void dm_put_table_device(struct mapped_device *md, struct dm_dev *d)
{
struct table_device *td = container_of(d, struct table_device, dm_dev);
mutex_lock(&md->table_devices_lock);
if (refcount_dec_and_test(&td->count)) {
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
close_table_device(td, md);
list_del(&td->list);
kfree(td);
}
mutex_unlock(&md->table_devices_lock);
}
EXPORT_SYMBOL(dm_put_table_device);
static void free_table_devices(struct list_head *devices)
{
struct list_head *tmp, *next;
list_for_each_safe(tmp, next, devices) {
struct table_device *td = list_entry(tmp, struct table_device, list);
DMWARN("dm_destroy: %s still exists with %d references",
td->dm_dev.name, refcount_read(&td->count));
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
kfree(td);
}
}
/*
* Get the geometry associated with a dm device
*/
int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo)
{
*geo = md->geometry;
return 0;
}
/*
* Set the geometry of a device.
*/
int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo)
{
sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors;
if (geo->start > sz) {
DMWARN("Start sector is beyond the geometry limits.");
return -EINVAL;
}
md->geometry = *geo;
return 0;
}
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
static int __noflush_suspending(struct mapped_device *md)
{
return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}
/*
* Decrements the number of outstanding ios that a bio has been
* cloned into, completing the original io if necc.
*/
static void dec_pending(struct dm_io *io, blk_status_t error)
{
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
unsigned long flags;
blk_status_t io_error;
struct bio *bio;
struct mapped_device *md = io->md;
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
/* Push-back supersedes any I/O errors */
if (unlikely(error)) {
spin_lock_irqsave(&io->endio_lock, flags);
if (!(io->status == BLK_STS_DM_REQUEUE && __noflush_suspending(md)))
io->status = error;
spin_unlock_irqrestore(&io->endio_lock, flags);
}
if (atomic_dec_and_test(&io->io_count)) {
if (io->status == BLK_STS_DM_REQUEUE) {
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
/*
* Target requested pushing back the I/O.
*/
spin_lock_irqsave(&md->deferred_lock, flags);
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
if (__noflush_suspending(md))
/* NOTE early return due to BLK_STS_DM_REQUEUE below */
bio_list_add_head(&md->deferred, io->orig_bio);
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
else
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
/* noflush suspend was interrupted. */
io->status = BLK_STS_IOERR;
spin_unlock_irqrestore(&md->deferred_lock, flags);
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
}
io_error = io->status;
bio = io->orig_bio;
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
end_io_acct(io);
free_io(md, io);
if (io_error == BLK_STS_DM_REQUEUE)
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
return;
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
if ((bio->bi_opf & REQ_PREFLUSH) && bio->bi_iter.bi_size) {
/*
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
* Preflush done for flush with data, reissue
* without REQ_PREFLUSH.
*/
bio->bi_opf &= ~REQ_PREFLUSH;
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
queue_io(md, bio);
} else {
/* done with normal IO or empty flush */
if (io_error)
bio->bi_status = io_error;
bio_endio(bio);
}
}
}
void disable_write_same(struct mapped_device *md)
{
struct queue_limits *limits = dm_get_queue_limits(md);
/* device doesn't really support WRITE SAME, disable it */
limits->max_write_same_sectors = 0;
}
void disable_write_zeroes(struct mapped_device *md)
{
struct queue_limits *limits = dm_get_queue_limits(md);
/* device doesn't really support WRITE ZEROES, disable it */
limits->max_write_zeroes_sectors = 0;
}
static void clone_endio(struct bio *bio)
{
blk_status_t error = bio->bi_status;
struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
struct dm_io *io = tio->io;
struct mapped_device *md = tio->io->md;
dm_endio_fn endio = tio->ti->type->end_io;
if (unlikely(error == BLK_STS_TARGET) && md->type != DM_TYPE_NVME_BIO_BASED) {
if (bio_op(bio) == REQ_OP_WRITE_SAME &&
!bio->bi_disk->queue->limits.max_write_same_sectors)
disable_write_same(md);
if (bio_op(bio) == REQ_OP_WRITE_ZEROES &&
!bio->bi_disk->queue->limits.max_write_zeroes_sectors)
disable_write_zeroes(md);
}
if (endio) {
int r = endio(tio->ti, bio, &error);
switch (r) {
case DM_ENDIO_REQUEUE:
error = BLK_STS_DM_REQUEUE;
/*FALLTHRU*/
case DM_ENDIO_DONE:
break;
case DM_ENDIO_INCOMPLETE:
/* The target will handle the io */
return;
default:
DMWARN("unimplemented target endio return value: %d", r);
BUG();
}
}
free_tio(tio);
dec_pending(io, error);
}
/*
* Return maximum size of I/O possible at the supplied sector up to the current
* target boundary.
*/
static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti)
{
sector_t target_offset = dm_target_offset(ti, sector);
return ti->len - target_offset;
}
static sector_t max_io_len(sector_t sector, struct dm_target *ti)
{
sector_t len = max_io_len_target_boundary(sector, ti);
sector_t offset, max_len;
/*
* Does the target need to split even further?
*/
if (ti->max_io_len) {
offset = dm_target_offset(ti, sector);
if (unlikely(ti->max_io_len & (ti->max_io_len - 1)))
max_len = sector_div(offset, ti->max_io_len);
else
max_len = offset & (ti->max_io_len - 1);
max_len = ti->max_io_len - max_len;
if (len > max_len)
len = max_len;
}
return len;
}
int dm_set_target_max_io_len(struct dm_target *ti, sector_t len)
{
if (len > UINT_MAX) {
DMERR("Specified maximum size of target IO (%llu) exceeds limit (%u)",
(unsigned long long)len, UINT_MAX);
ti->error = "Maximum size of target IO is too large";
return -EINVAL;
}
/*
* BIO based queue uses its own splitting. When multipage bvecs
* is switched on, size of the incoming bio may be too big to
* be handled in some targets, such as crypt.
*
* When these targets are ready for the big bio, we can remove
* the limit.
*/
ti->max_io_len = min_t(uint32_t, len, BIO_MAX_PAGES * PAGE_SIZE);
return 0;
}
EXPORT_SYMBOL_GPL(dm_set_target_max_io_len);
static struct dm_target *dm_dax_get_live_target(struct mapped_device *md,
sector_t sector, int *srcu_idx)
__acquires(md->io_barrier)
{
struct dm_table *map;
struct dm_target *ti;
map = dm_get_live_table(md, srcu_idx);
if (!map)
return NULL;
ti = dm_table_find_target(map, sector);
if (!dm_target_is_valid(ti))
return NULL;
return ti;
}
static long dm_dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff,
long nr_pages, void **kaddr, pfn_t *pfn)
{
struct mapped_device *md = dax_get_private(dax_dev);
sector_t sector = pgoff * PAGE_SECTORS;
struct dm_target *ti;
long len, ret = -EIO;
int srcu_idx;
ti = dm_dax_get_live_target(md, sector, &srcu_idx);
if (!ti)
goto out;
if (!ti->type->direct_access)
goto out;
len = max_io_len(sector, ti) / PAGE_SECTORS;
if (len < 1)
goto out;
nr_pages = min(len, nr_pages);
ret = ti->type->direct_access(ti, pgoff, nr_pages, kaddr, pfn);
out:
dm_put_live_table(md, srcu_idx);
return ret;
}
static size_t dm_dax_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff,
void *addr, size_t bytes, struct iov_iter *i)
{
struct mapped_device *md = dax_get_private(dax_dev);
sector_t sector = pgoff * PAGE_SECTORS;
struct dm_target *ti;
long ret = 0;
int srcu_idx;
ti = dm_dax_get_live_target(md, sector, &srcu_idx);
if (!ti)
goto out;
if (!ti->type->dax_copy_from_iter) {
ret = copy_from_iter(addr, bytes, i);
goto out;
}
ret = ti->type->dax_copy_from_iter(ti, pgoff, addr, bytes, i);
out:
dm_put_live_table(md, srcu_idx);
return ret;
}
static size_t dm_dax_copy_to_iter(struct dax_device *dax_dev, pgoff_t pgoff,
void *addr, size_t bytes, struct iov_iter *i)
{
struct mapped_device *md = dax_get_private(dax_dev);
sector_t sector = pgoff * PAGE_SECTORS;
struct dm_target *ti;
long ret = 0;
int srcu_idx;
ti = dm_dax_get_live_target(md, sector, &srcu_idx);
if (!ti)
goto out;
if (!ti->type->dax_copy_to_iter) {
ret = copy_to_iter(addr, bytes, i);
goto out;
}
ret = ti->type->dax_copy_to_iter(ti, pgoff, addr, bytes, i);
out:
dm_put_live_table(md, srcu_idx);
return ret;
}
/*
* A target may call dm_accept_partial_bio only from the map routine. It is
* allowed for all bio types except REQ_PREFLUSH and REQ_OP_ZONE_RESET.
*
* dm_accept_partial_bio informs the dm that the target only wants to process
* additional n_sectors sectors of the bio and the rest of the data should be
* sent in a next bio.
*
* A diagram that explains the arithmetics:
* +--------------------+---------------+-------+
* | 1 | 2 | 3 |
* +--------------------+---------------+-------+
*
* <-------------- *tio->len_ptr --------------->
* <------- bi_size ------->
* <-- n_sectors -->
*
* Region 1 was already iterated over with bio_advance or similar function.
* (it may be empty if the target doesn't use bio_advance)
* Region 2 is the remaining bio size that the target wants to process.
* (it may be empty if region 1 is non-empty, although there is no reason
* to make it empty)
* The target requires that region 3 is to be sent in the next bio.
*
* If the target wants to receive multiple copies of the bio (via num_*bios, etc),
* the partially processed part (the sum of regions 1+2) must be the same for all
* copies of the bio.
*/
void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors)
{
struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
unsigned bi_size = bio->bi_iter.bi_size >> SECTOR_SHIFT;
BUG_ON(bio->bi_opf & REQ_PREFLUSH);
BUG_ON(bi_size > *tio->len_ptr);
BUG_ON(n_sectors > bi_size);
*tio->len_ptr -= bi_size - n_sectors;
bio->bi_iter.bi_size = n_sectors << SECTOR_SHIFT;
}
EXPORT_SYMBOL_GPL(dm_accept_partial_bio);
/*
* The zone descriptors obtained with a zone report indicate
* zone positions within the underlying device of the target. The zone
* descriptors must be remapped to match their position within the dm device.
* The caller target should obtain the zones information using
* blkdev_report_zones() to ensure that remapping for partition offset is
* already handled.
*/
void dm_remap_zone_report(struct dm_target *ti, sector_t start,
struct blk_zone *zones, unsigned int *nr_zones)
{
#ifdef CONFIG_BLK_DEV_ZONED
struct blk_zone *zone;
unsigned int nrz = *nr_zones;
int i;
/*
* Remap the start sector and write pointer position of the zones in
* the array. Since we may have obtained from the target underlying
* device more zones that the target size, also adjust the number
* of zones.
*/
for (i = 0; i < nrz; i++) {
zone = zones + i;
if (zone->start >= start + ti->len) {
memset(zone, 0, sizeof(struct blk_zone) * (nrz - i));
break;
}
zone->start = zone->start + ti->begin - start;
if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
continue;
if (zone->cond == BLK_ZONE_COND_FULL)
zone->wp = zone->start + zone->len;
else if (zone->cond == BLK_ZONE_COND_EMPTY)
zone->wp = zone->start;
else
zone->wp = zone->wp + ti->begin - start;
}
*nr_zones = i;
#else /* !CONFIG_BLK_DEV_ZONED */
*nr_zones = 0;
#endif
}
EXPORT_SYMBOL_GPL(dm_remap_zone_report);
static blk_qc_t __map_bio(struct dm_target_io *tio)
{
int r;
sector_t sector;
struct bio *clone = &tio->clone;
struct dm_io *io = tio->io;
struct mapped_device *md = io->md;
struct dm_target *ti = tio->ti;
blk_qc_t ret = BLK_QC_T_NONE;
clone->bi_end_io = clone_endio;
/*
* Map the clone. If r == 0 we don't need to do
* anything, the target has assumed ownership of
* this io.
*/
atomic_inc(&io->io_count);
block: Abstract out bvec iterator Immutable biovecs are going to require an explicit iterator. To implement immutable bvecs, a later patch is going to add a bi_bvec_done member to this struct; for now, this patch effectively just renames things. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Matthew Wilcox <willy@linux.intel.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Alasdair Kergon <agk@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Boaz Harrosh <bharrosh@panasas.com> Cc: Benny Halevy <bhalevy@tonian.com> Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Chris Mason <chris.mason@fusionio.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jaegeuk Kim <jaegeuk.kim@samsung.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@kernel.org> Cc: Joern Engel <joern@logfs.org> Cc: Prasad Joshi <prasadjoshi.linux@gmail.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: KONISHI Ryusuke <konishi.ryusuke@lab.ntt.co.jp> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Ben Myers <bpm@sgi.com> Cc: xfs@oss.sgi.com Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Len Brown <len.brown@intel.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Wei Yongjun <yongjun_wei@trendmicro.com.cn> Cc: "Roger Pau Monné" <roger.pau@citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Cc: Ian Campbell <Ian.Campbell@citrix.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Jiang Liu <jiang.liu@huawei.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchand@redhat.com> Cc: Joe Perches <joe@perches.com> Cc: Peng Tao <tao.peng@emc.com> Cc: Andy Adamson <andros@netapp.com> Cc: fanchaoting <fanchaoting@cn.fujitsu.com> Cc: Jie Liu <jeff.liu@oracle.com> Cc: Sunil Mushran <sunil.mushran@gmail.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Namjae Jeon <namjae.jeon@samsung.com> Cc: Pankaj Kumar <pankaj.km@samsung.com> Cc: Dan Magenheimer <dan.magenheimer@oracle.com> Cc: Mel Gorman <mgorman@suse.de>6
2013-10-11 22:44:27 +00:00
sector = clone->bi_iter.bi_sector;
dm: flush queued bios when process blocks to avoid deadlock Commit df2cb6daa4 ("block: Avoid deadlocks with bio allocation by stacking drivers") created a workqueue for every bio set and code in bio_alloc_bioset() that tries to resolve some low-memory deadlocks by redirecting bios queued on current->bio_list to the workqueue if the system is low on memory. However other deadlocks (see below **) may happen, without any low memory condition, because generic_make_request is queuing bios to current->bio_list (rather than submitting them). ** the related dm-snapshot deadlock is detailed here: https://www.redhat.com/archives/dm-devel/2016-July/msg00065.html Fix this deadlock by redirecting any bios on current->bio_list to the bio_set's rescue workqueue on every schedule() call. Consequently, when the process blocks on a mutex, the bios queued on current->bio_list are dispatched to independent workqueus and they can complete without waiting for the mutex to be available. The structure blk_plug contains an entry cb_list and this list can contain arbitrary callback functions that are called when the process blocks. To implement this fix DM (ab)uses the onstack plug's cb_list interface to get its flush_current_bio_list() called at schedule() time. This fixes the snapshot deadlock - if the map method blocks, flush_current_bio_list() will be called and it redirects bios waiting on current->bio_list to appropriate workqueues. Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1267650 Depends-on: df2cb6daa4 ("block: Avoid deadlocks with bio allocation by stacking drivers") Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-02-15 16:26:10 +00:00
r = ti->type->map(ti, clone);
switch (r) {
case DM_MAPIO_SUBMITTED:
break;
case DM_MAPIO_REMAPPED:
/* the bio has been remapped so dispatch it */
trace_block_bio_remap(clone->bi_disk->queue, clone,
bio_dev(io->orig_bio), sector);
if (md->type == DM_TYPE_NVME_BIO_BASED)
ret = direct_make_request(clone);
else
ret = generic_make_request(clone);
break;
case DM_MAPIO_KILL:
free_tio(tio);
dec_pending(io, BLK_STS_IOERR);
break;
case DM_MAPIO_REQUEUE:
free_tio(tio);
dec_pending(io, BLK_STS_DM_REQUEUE);
break;
default:
2006-12-08 10:41:05 +00:00
DMWARN("unimplemented target map return value: %d", r);
BUG();
}
return ret;
}
static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
{
block: Abstract out bvec iterator Immutable biovecs are going to require an explicit iterator. To implement immutable bvecs, a later patch is going to add a bi_bvec_done member to this struct; for now, this patch effectively just renames things. Signed-off-by: Kent Overstreet <kmo@daterainc.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: "Ed L. Cashin" <ecashin@coraid.com> Cc: Nick Piggin <npiggin@kernel.dk> Cc: Lars Ellenberg <drbd-dev@lists.linbit.com> Cc: Jiri Kosina <jkosina@suse.cz> Cc: Matthew Wilcox <willy@linux.intel.com> Cc: Geoff Levand <geoff@infradead.org> Cc: Yehuda Sadeh <yehuda@inktank.com> Cc: Sage Weil <sage@inktank.com> Cc: Alex Elder <elder@inktank.com> Cc: ceph-devel@vger.kernel.org Cc: Joshua Morris <josh.h.morris@us.ibm.com> Cc: Philip Kelleher <pjk1939@linux.vnet.ibm.com> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Jeremy Fitzhardinge <jeremy@goop.org> Cc: Neil Brown <neilb@suse.de> Cc: Alasdair Kergon <agk@redhat.com> Cc: Mike Snitzer <snitzer@redhat.com> Cc: dm-devel@redhat.com Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: linux390@de.ibm.com Cc: Boaz Harrosh <bharrosh@panasas.com> Cc: Benny Halevy <bhalevy@tonian.com> Cc: "James E.J. Bottomley" <JBottomley@parallels.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: "Nicholas A. Bellinger" <nab@linux-iscsi.org> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Chris Mason <chris.mason@fusionio.com> Cc: "Theodore Ts'o" <tytso@mit.edu> Cc: Andreas Dilger <adilger.kernel@dilger.ca> Cc: Jaegeuk Kim <jaegeuk.kim@samsung.com> Cc: Steven Whitehouse <swhiteho@redhat.com> Cc: Dave Kleikamp <shaggy@kernel.org> Cc: Joern Engel <joern@logfs.org> Cc: Prasad Joshi <prasadjoshi.linux@gmail.com> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: KONISHI Ryusuke <konishi.ryusuke@lab.ntt.co.jp> Cc: Mark Fasheh <mfasheh@suse.com> Cc: Joel Becker <jlbec@evilplan.org> Cc: Ben Myers <bpm@sgi.com> Cc: xfs@oss.sgi.com Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Len Brown <len.brown@intel.com> Cc: Pavel Machek <pavel@ucw.cz> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com> Cc: Ben Hutchings <ben@decadent.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Guo Chao <yan@linux.vnet.ibm.com> Cc: Tejun Heo <tj@kernel.org> Cc: Asai Thambi S P <asamymuthupa@micron.com> Cc: Selvan Mani <smani@micron.com> Cc: Sam Bradshaw <sbradshaw@micron.com> Cc: Wei Yongjun <yongjun_wei@trendmicro.com.cn> Cc: "Roger Pau Monné" <roger.pau@citrix.com> Cc: Jan Beulich <jbeulich@suse.com> Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com> Cc: Ian Campbell <Ian.Campbell@citrix.com> Cc: Sebastian Ott <sebott@linux.vnet.ibm.com> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Jiang Liu <jiang.liu@huawei.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchand@redhat.com> Cc: Joe Perches <joe@perches.com> Cc: Peng Tao <tao.peng@emc.com> Cc: Andy Adamson <andros@netapp.com> Cc: fanchaoting <fanchaoting@cn.fujitsu.com> Cc: Jie Liu <jeff.liu@oracle.com> Cc: Sunil Mushran <sunil.mushran@gmail.com> Cc: "Martin K. Petersen" <martin.petersen@oracle.com> Cc: Namjae Jeon <namjae.jeon@samsung.com> Cc: Pankaj Kumar <pankaj.km@samsung.com> Cc: Dan Magenheimer <dan.magenheimer@oracle.com> Cc: Mel Gorman <mgorman@suse.de>6
2013-10-11 22:44:27 +00:00
bio->bi_iter.bi_sector = sector;
bio->bi_iter.bi_size = to_bytes(len);
}
/*
* Creates a bio that consists of range of complete bvecs.
*/
static int clone_bio(struct dm_target_io *tio, struct bio *bio,
sector_t sector, unsigned len)
{
struct bio *clone = &tio->clone;
__bio_clone_fast(clone, bio);
if (bio_integrity(bio)) {
int r;
if (unlikely(!dm_target_has_integrity(tio->ti->type) &&
!dm_target_passes_integrity(tio->ti->type))) {
DMWARN("%s: the target %s doesn't support integrity data.",
dm_device_name(tio->io->md),
tio->ti->type->name);
return -EIO;
}
r = bio_integrity_clone(clone, bio, GFP_NOIO);
if (r < 0)
return r;
}
bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector));
clone->bi_iter.bi_size = to_bytes(len);
if (bio_integrity(bio))
bio_integrity_trim(clone);
return 0;
}
static void alloc_multiple_bios(struct bio_list *blist, struct clone_info *ci,
struct dm_target *ti, unsigned num_bios)
{
struct dm_target_io *tio;
int try;
if (!num_bios)
return;
if (num_bios == 1) {
tio = alloc_tio(ci, ti, 0, GFP_NOIO);
bio_list_add(blist, &tio->clone);
return;
}
for (try = 0; try < 2; try++) {
int bio_nr;
struct bio *bio;
if (try)
mutex_lock(&ci->io->md->table_devices_lock);
for (bio_nr = 0; bio_nr < num_bios; bio_nr++) {
tio = alloc_tio(ci, ti, bio_nr, try ? GFP_NOIO : GFP_NOWAIT);
if (!tio)
break;
bio_list_add(blist, &tio->clone);
}
if (try)
mutex_unlock(&ci->io->md->table_devices_lock);
if (bio_nr == num_bios)
return;
while ((bio = bio_list_pop(blist))) {
tio = container_of(bio, struct dm_target_io, clone);
free_tio(tio);
}
}
}
static blk_qc_t __clone_and_map_simple_bio(struct clone_info *ci,
struct dm_target_io *tio, unsigned *len)
{
struct bio *clone = &tio->clone;
tio->len_ptr = len;
__bio_clone_fast(clone, ci->bio);
if (len)
bio_setup_sector(clone, ci->sector, *len);
return __map_bio(tio);
}
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
unsigned num_bios, unsigned *len)
{
struct bio_list blist = BIO_EMPTY_LIST;
struct bio *bio;
struct dm_target_io *tio;
alloc_multiple_bios(&blist, ci, ti, num_bios);
while ((bio = bio_list_pop(&blist))) {
tio = container_of(bio, struct dm_target_io, clone);
(void) __clone_and_map_simple_bio(ci, tio, len);
}
}
static int __send_empty_flush(struct clone_info *ci)
{
unsigned target_nr = 0;
struct dm_target *ti;
/*
* Empty flush uses a statically initialized bio, as the base for
* cloning. However, blkg association requires that a bdev is
* associated with a gendisk, which doesn't happen until the bdev is
* opened. So, blkg association is done at issue time of the flush
* rather than when the device is created in alloc_dev().
*/
bio_set_dev(ci->bio, ci->io->md->bdev);
BUG_ON(bio_has_data(ci->bio));
while ((ti = dm_table_get_target(ci->map, target_nr++)))
__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
bio_disassociate_blkg(ci->bio);
return 0;
}
static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
sector_t sector, unsigned *len)
{
struct bio *bio = ci->bio;
struct dm_target_io *tio;
int r;
tio = alloc_tio(ci, ti, 0, GFP_NOIO);
tio->len_ptr = len;
r = clone_bio(tio, bio, sector, *len);
if (r < 0) {
free_tio(tio);
return r;
}
(void) __map_bio(tio);
return 0;
}
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
static unsigned get_num_discard_bios(struct dm_target *ti)
{
return ti->num_discard_bios;
}
static unsigned get_num_secure_erase_bios(struct dm_target *ti)
{
return ti->num_secure_erase_bios;
}
static unsigned get_num_write_same_bios(struct dm_target *ti)
{
return ti->num_write_same_bios;
}
static unsigned get_num_write_zeroes_bios(struct dm_target *ti)
{
return ti->num_write_zeroes_bios;
}
typedef bool (*is_split_required_fn)(struct dm_target *ti);
static bool is_split_required_for_discard(struct dm_target *ti)
{
return ti->split_discard_bios;
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
}
static int __send_changing_extent_only(struct clone_info *ci, struct dm_target *ti,
unsigned num_bios, bool is_split_required)
{
unsigned len;
/*
* Even though the device advertised support for this type of
* request, that does not mean every target supports it, and
* reconfiguration might also have changed that since the
* check was performed.
*/
if (!num_bios)
return -EOPNOTSUPP;
if (!is_split_required)
len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
else
len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
__send_duplicate_bios(ci, ti, num_bios, &len);
ci->sector += len;
ci->sector_count -= len;
return 0;
}
static int __send_discard(struct clone_info *ci, struct dm_target *ti)
{
return __send_changing_extent_only(ci, ti, get_num_discard_bios(ti),
is_split_required_for_discard(ti));
}
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
static int __send_secure_erase(struct clone_info *ci, struct dm_target *ti)
{
return __send_changing_extent_only(ci, ti, get_num_secure_erase_bios(ti), false);
}
static int __send_write_same(struct clone_info *ci, struct dm_target *ti)
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
{
return __send_changing_extent_only(ci, ti, get_num_write_same_bios(ti), false);
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
}
static int __send_write_zeroes(struct clone_info *ci, struct dm_target *ti)
{
return __send_changing_extent_only(ci, ti, get_num_write_zeroes_bios(ti), false);
}
static bool is_abnormal_io(struct bio *bio)
{
bool r = false;
switch (bio_op(bio)) {
case REQ_OP_DISCARD:
case REQ_OP_SECURE_ERASE:
case REQ_OP_WRITE_SAME:
case REQ_OP_WRITE_ZEROES:
r = true;
break;
}
return r;
}
static bool __process_abnormal_io(struct clone_info *ci, struct dm_target *ti,
int *result)
{
struct bio *bio = ci->bio;
if (bio_op(bio) == REQ_OP_DISCARD)
*result = __send_discard(ci, ti);
else if (bio_op(bio) == REQ_OP_SECURE_ERASE)
*result = __send_secure_erase(ci, ti);
else if (bio_op(bio) == REQ_OP_WRITE_SAME)
*result = __send_write_same(ci, ti);
else if (bio_op(bio) == REQ_OP_WRITE_ZEROES)
*result = __send_write_zeroes(ci, ti);
else
return false;
return true;
}
/*
* Select the correct strategy for processing a non-flush bio.
*/
static int __split_and_process_non_flush(struct clone_info *ci)
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
{
struct dm_target *ti;
unsigned len;
int r;
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
ti = dm_table_find_target(ci->map, ci->sector);
if (!dm_target_is_valid(ti))
return -EIO;
if (__process_abnormal_io(ci, ti, &r))
return r;
len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
r = __clone_and_map_data_bio(ci, ti, ci->sector, &len);
if (r < 0)
return r;
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
ci->sector += len;
ci->sector_count -= len;
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
return 0;
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
}
static void init_clone_info(struct clone_info *ci, struct mapped_device *md,
struct dm_table *map, struct bio *bio)
{
ci->map = map;
ci->io = alloc_io(md, bio);
ci->sector = bio->bi_iter.bi_sector;
}
dm: fix redundant IO accounting for bios that need splitting The risk of redundant IO accounting was not taken into consideration when commit 18a25da84354 ("dm: ensure bio submission follows a depth-first tree walk") introduced IO splitting in terms of recursion via generic_make_request(). Fix this by subtracting the split bio's payload from the IO stats that were already accounted for by start_io_acct() upon dm_make_request() entry. This repeat oscillation of the IO accounting, up then down, isn't ideal but refactoring DM core's IO splitting to pre-split bios _before_ they are accounted turned out to be an excessive amount of change that will need a full development cycle to refine and verify. Before this fix: /dev/mapper/stripe_dev is a 4-way stripe using a 32k chunksize, so bios are split on 32k boundaries. # fio --name=16M --filename=/dev/mapper/stripe_dev --rw=write --bs=64k --size=16M \ --iodepth=1 --ioengine=libaio --direct=1 --refill_buffers with debugging added: [103898.310264] device-mapper: core: start_io_acct: dm-2 WRITE bio->bi_iter.bi_sector=0 len=128 [103898.318704] device-mapper: core: __split_and_process_bio: recursing for following split bio: [103898.329136] device-mapper: core: start_io_acct: dm-2 WRITE bio->bi_iter.bi_sector=64 len=64 ... 16M written yet 136M (278528 * 512b) accounted: # cat /sys/block/dm-2/stat | awk '{ print $7 }' 278528 After this fix: 16M written and 16M (32768 * 512b) accounted: # cat /sys/block/dm-2/stat | awk '{ print $7 }' 32768 Fixes: 18a25da84354 ("dm: ensure bio submission follows a depth-first tree walk") Cc: stable@vger.kernel.org # 4.16+ Reported-by: Bryan Gurney <bgurney@redhat.com> Reviewed-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2019-01-17 15:48:01 +00:00
#define __dm_part_stat_sub(part, field, subnd) \
(part_stat_get(part, field) -= (subnd))
/*
* Entry point to split a bio into clones and submit them to the targets.
*/
static blk_qc_t __split_and_process_bio(struct mapped_device *md,
struct dm_table *map, struct bio *bio)
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
{
struct clone_info ci;
blk_qc_t ret = BLK_QC_T_NONE;
int error = 0;
init_clone_info(&ci, md, map, bio);
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
if (bio->bi_opf & REQ_PREFLUSH) {
struct bio flush_bio;
/*
* Use an on-stack bio for this, it's safe since we don't
* need to reference it after submit. It's just used as
* the basis for the clone(s).
*/
bio_init(&flush_bio, NULL, 0);
flush_bio.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC;
ci.bio = &flush_bio;
ci.sector_count = 0;
error = __send_empty_flush(&ci);
/* dec_pending submits any data associated with flush */
} else if (bio_op(bio) == REQ_OP_ZONE_RESET) {
ci.bio = bio;
ci.sector_count = 0;
error = __split_and_process_non_flush(&ci);
} else {
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
ci.bio = bio;
dm: implement REQ_FLUSH/FUA support for bio-based dm This patch converts bio-based dm to support REQ_FLUSH/FUA instead of now deprecated REQ_HARDBARRIER. * -EOPNOTSUPP handling logic dropped. * Preflush is handled as before but postflush is dropped and replaced with passing down REQ_FUA to member request_queues. This replaces one array wide cache flush w/ member specific FUA writes. * __split_and_process_bio() now calls __clone_and_map_flush() directly for flushes and guarantees all FLUSH bio's going to targets are zero ` length. * It's now guaranteed that all FLUSH bio's which are passed onto dm targets are zero length. bio_empty_barrier() tests are replaced with REQ_FLUSH tests. * Empty WRITE_BARRIERs are replaced with WRITE_FLUSHes. * Dropped unlikely() around REQ_FLUSH tests. Flushes are not unlikely enough to be marked with unlikely(). * Block layer now filters out REQ_FLUSH/FUA bio's if the request_queue doesn't support cache flushing. Advertise REQ_FLUSH | REQ_FUA capability. * Request based dm isn't converted yet. dm_init_request_based_queue() resets flush support to 0 for now. To avoid disturbing request based dm code, dm->flush_error is added for bio based dm while requested based dm continues to use dm->barrier_error. Lightly tested linear, stripe, raid1, snap and crypt targets. Please proceed with caution as I'm not familiar with the code base. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: dm-devel@redhat.com Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-03 09:56:19 +00:00
ci.sector_count = bio_sectors(bio);
dm: ensure bio submission follows a depth-first tree walk A dm device can, in general, represent a tree of targets, each of which handles a sub-range of the range of blocks handled by the parent. The bio sequencing managed by generic_make_request() requires that bios are generated and handled in a depth-first manner. Each call to a make_request_fn() may submit bios to a single member device, and may submit bios for a reduced region of the same device as the make_request_fn. In particular, any bios submitted to member devices must be expected to be processed in order, so a later one must never wait for an earlier one. This ordering is usually achieved by using bio_split() to reduce a bio to a size that can be completely handled by one target, and resubmitting the remainder to the originating device. bio_queue_split() shows the canonical approach. dm doesn't follow this approach, largely because it has needed to split bios since long before bio_split() was available. It currently can submit bios to separate targets within the one dm_make_request() call. Dependencies between these targets, as can happen with dm-snap, can cause deadlocks if either bios gets stuck behind the other in the queues managed by generic_make_request(). This requires the 'rescue' functionality provided by dm_offload_{start,end}. Some of this requirement can be removed by changing the order of bio submission to follow the canonical approach. That is, if dm finds that it needs to split a bio, the remainder should be sent to generic_make_request() rather than being handled immediately. This delays the handling until the first part is completely processed, so the deadlock problems do not occur. __split_and_process_bio() can be called both from dm_make_request() and from dm_wq_work(). When called from dm_wq_work() the current approach is perfectly satisfactory as each bio will be processed immediately. When called from dm_make_request(), current->bio_list will be non-NULL, and in this case it is best to create a separate "clone" bio for the remainder. When we use bio_clone_bioset() to split off the front part of a bio and chain the two together and submit the remainder to generic_make_request(), it is important that the newly allocated bio is used as the head to be processed immediately, and the original bio gets "bio_advance()"d and sent to generic_make_request() as the remainder. Otherwise, if the newly allocated bio is used as the remainder, and if it then needs to be split again, then the next bio_clone_bioset() call will be made while holding a reference a bio (result of the first clone) from the same bioset. This can potentially exhaust the bioset mempool and result in a memory allocation deadlock. Note that there is no race caused by reassigning cio.io->bio after already calling __map_bio(). This bio will only be dereferenced again after dec_pending() has found io->io_count to be zero, and this cannot happen before the dec_pending() call at the end of __split_and_process_bio(). To provide the clone bio when splitting, we use q->bio_split. This was previously being freed by bio-based dm to avoid having excess rescuer threads. As bio_split bio sets no longer create rescuer threads, there is little cost and much gain from restoring the q->bio_split bio set. Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-09-05 23:43:28 +00:00
while (ci.sector_count && !error) {
error = __split_and_process_non_flush(&ci);
dm: ensure bio submission follows a depth-first tree walk A dm device can, in general, represent a tree of targets, each of which handles a sub-range of the range of blocks handled by the parent. The bio sequencing managed by generic_make_request() requires that bios are generated and handled in a depth-first manner. Each call to a make_request_fn() may submit bios to a single member device, and may submit bios for a reduced region of the same device as the make_request_fn. In particular, any bios submitted to member devices must be expected to be processed in order, so a later one must never wait for an earlier one. This ordering is usually achieved by using bio_split() to reduce a bio to a size that can be completely handled by one target, and resubmitting the remainder to the originating device. bio_queue_split() shows the canonical approach. dm doesn't follow this approach, largely because it has needed to split bios since long before bio_split() was available. It currently can submit bios to separate targets within the one dm_make_request() call. Dependencies between these targets, as can happen with dm-snap, can cause deadlocks if either bios gets stuck behind the other in the queues managed by generic_make_request(). This requires the 'rescue' functionality provided by dm_offload_{start,end}. Some of this requirement can be removed by changing the order of bio submission to follow the canonical approach. That is, if dm finds that it needs to split a bio, the remainder should be sent to generic_make_request() rather than being handled immediately. This delays the handling until the first part is completely processed, so the deadlock problems do not occur. __split_and_process_bio() can be called both from dm_make_request() and from dm_wq_work(). When called from dm_wq_work() the current approach is perfectly satisfactory as each bio will be processed immediately. When called from dm_make_request(), current->bio_list will be non-NULL, and in this case it is best to create a separate "clone" bio for the remainder. When we use bio_clone_bioset() to split off the front part of a bio and chain the two together and submit the remainder to generic_make_request(), it is important that the newly allocated bio is used as the head to be processed immediately, and the original bio gets "bio_advance()"d and sent to generic_make_request() as the remainder. Otherwise, if the newly allocated bio is used as the remainder, and if it then needs to be split again, then the next bio_clone_bioset() call will be made while holding a reference a bio (result of the first clone) from the same bioset. This can potentially exhaust the bioset mempool and result in a memory allocation deadlock. Note that there is no race caused by reassigning cio.io->bio after already calling __map_bio(). This bio will only be dereferenced again after dec_pending() has found io->io_count to be zero, and this cannot happen before the dec_pending() call at the end of __split_and_process_bio(). To provide the clone bio when splitting, we use q->bio_split. This was previously being freed by bio-based dm to avoid having excess rescuer threads. As bio_split bio sets no longer create rescuer threads, there is little cost and much gain from restoring the q->bio_split bio set. Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-09-05 23:43:28 +00:00
if (current->bio_list && ci.sector_count && !error) {
/*
* Remainder must be passed to generic_make_request()
* so that it gets handled *after* bios already submitted
* have been completely processed.
* We take a clone of the original to store in
* ci.io->orig_bio to be used by end_io_acct() and
dm: ensure bio submission follows a depth-first tree walk A dm device can, in general, represent a tree of targets, each of which handles a sub-range of the range of blocks handled by the parent. The bio sequencing managed by generic_make_request() requires that bios are generated and handled in a depth-first manner. Each call to a make_request_fn() may submit bios to a single member device, and may submit bios for a reduced region of the same device as the make_request_fn. In particular, any bios submitted to member devices must be expected to be processed in order, so a later one must never wait for an earlier one. This ordering is usually achieved by using bio_split() to reduce a bio to a size that can be completely handled by one target, and resubmitting the remainder to the originating device. bio_queue_split() shows the canonical approach. dm doesn't follow this approach, largely because it has needed to split bios since long before bio_split() was available. It currently can submit bios to separate targets within the one dm_make_request() call. Dependencies between these targets, as can happen with dm-snap, can cause deadlocks if either bios gets stuck behind the other in the queues managed by generic_make_request(). This requires the 'rescue' functionality provided by dm_offload_{start,end}. Some of this requirement can be removed by changing the order of bio submission to follow the canonical approach. That is, if dm finds that it needs to split a bio, the remainder should be sent to generic_make_request() rather than being handled immediately. This delays the handling until the first part is completely processed, so the deadlock problems do not occur. __split_and_process_bio() can be called both from dm_make_request() and from dm_wq_work(). When called from dm_wq_work() the current approach is perfectly satisfactory as each bio will be processed immediately. When called from dm_make_request(), current->bio_list will be non-NULL, and in this case it is best to create a separate "clone" bio for the remainder. When we use bio_clone_bioset() to split off the front part of a bio and chain the two together and submit the remainder to generic_make_request(), it is important that the newly allocated bio is used as the head to be processed immediately, and the original bio gets "bio_advance()"d and sent to generic_make_request() as the remainder. Otherwise, if the newly allocated bio is used as the remainder, and if it then needs to be split again, then the next bio_clone_bioset() call will be made while holding a reference a bio (result of the first clone) from the same bioset. This can potentially exhaust the bioset mempool and result in a memory allocation deadlock. Note that there is no race caused by reassigning cio.io->bio after already calling __map_bio(). This bio will only be dereferenced again after dec_pending() has found io->io_count to be zero, and this cannot happen before the dec_pending() call at the end of __split_and_process_bio(). To provide the clone bio when splitting, we use q->bio_split. This was previously being freed by bio-based dm to avoid having excess rescuer threads. As bio_split bio sets no longer create rescuer threads, there is little cost and much gain from restoring the q->bio_split bio set. Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-09-05 23:43:28 +00:00
* for dec_pending to use for completion handling.
*/
struct bio *b = bio_split(bio, bio_sectors(bio) - ci.sector_count,
GFP_NOIO, &md->queue->bio_split);
ci.io->orig_bio = b;
dm: fix redundant IO accounting for bios that need splitting The risk of redundant IO accounting was not taken into consideration when commit 18a25da84354 ("dm: ensure bio submission follows a depth-first tree walk") introduced IO splitting in terms of recursion via generic_make_request(). Fix this by subtracting the split bio's payload from the IO stats that were already accounted for by start_io_acct() upon dm_make_request() entry. This repeat oscillation of the IO accounting, up then down, isn't ideal but refactoring DM core's IO splitting to pre-split bios _before_ they are accounted turned out to be an excessive amount of change that will need a full development cycle to refine and verify. Before this fix: /dev/mapper/stripe_dev is a 4-way stripe using a 32k chunksize, so bios are split on 32k boundaries. # fio --name=16M --filename=/dev/mapper/stripe_dev --rw=write --bs=64k --size=16M \ --iodepth=1 --ioengine=libaio --direct=1 --refill_buffers with debugging added: [103898.310264] device-mapper: core: start_io_acct: dm-2 WRITE bio->bi_iter.bi_sector=0 len=128 [103898.318704] device-mapper: core: __split_and_process_bio: recursing for following split bio: [103898.329136] device-mapper: core: start_io_acct: dm-2 WRITE bio->bi_iter.bi_sector=64 len=64 ... 16M written yet 136M (278528 * 512b) accounted: # cat /sys/block/dm-2/stat | awk '{ print $7 }' 278528 After this fix: 16M written and 16M (32768 * 512b) accounted: # cat /sys/block/dm-2/stat | awk '{ print $7 }' 32768 Fixes: 18a25da84354 ("dm: ensure bio submission follows a depth-first tree walk") Cc: stable@vger.kernel.org # 4.16+ Reported-by: Bryan Gurney <bgurney@redhat.com> Reviewed-by: Ming Lei <ming.lei@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2019-01-17 15:48:01 +00:00
/*
* Adjust IO stats for each split, otherwise upon queue
* reentry there will be redundant IO accounting.
* NOTE: this is a stop-gap fix, a proper fix involves
* significant refactoring of DM core's bio splitting
* (by eliminating DM's splitting and just using bio_split)
*/
part_stat_lock();
__dm_part_stat_sub(&dm_disk(md)->part0,
sectors[op_stat_group(bio_op(bio))], ci.sector_count);
part_stat_unlock();
dm: ensure bio submission follows a depth-first tree walk A dm device can, in general, represent a tree of targets, each of which handles a sub-range of the range of blocks handled by the parent. The bio sequencing managed by generic_make_request() requires that bios are generated and handled in a depth-first manner. Each call to a make_request_fn() may submit bios to a single member device, and may submit bios for a reduced region of the same device as the make_request_fn. In particular, any bios submitted to member devices must be expected to be processed in order, so a later one must never wait for an earlier one. This ordering is usually achieved by using bio_split() to reduce a bio to a size that can be completely handled by one target, and resubmitting the remainder to the originating device. bio_queue_split() shows the canonical approach. dm doesn't follow this approach, largely because it has needed to split bios since long before bio_split() was available. It currently can submit bios to separate targets within the one dm_make_request() call. Dependencies between these targets, as can happen with dm-snap, can cause deadlocks if either bios gets stuck behind the other in the queues managed by generic_make_request(). This requires the 'rescue' functionality provided by dm_offload_{start,end}. Some of this requirement can be removed by changing the order of bio submission to follow the canonical approach. That is, if dm finds that it needs to split a bio, the remainder should be sent to generic_make_request() rather than being handled immediately. This delays the handling until the first part is completely processed, so the deadlock problems do not occur. __split_and_process_bio() can be called both from dm_make_request() and from dm_wq_work(). When called from dm_wq_work() the current approach is perfectly satisfactory as each bio will be processed immediately. When called from dm_make_request(), current->bio_list will be non-NULL, and in this case it is best to create a separate "clone" bio for the remainder. When we use bio_clone_bioset() to split off the front part of a bio and chain the two together and submit the remainder to generic_make_request(), it is important that the newly allocated bio is used as the head to be processed immediately, and the original bio gets "bio_advance()"d and sent to generic_make_request() as the remainder. Otherwise, if the newly allocated bio is used as the remainder, and if it then needs to be split again, then the next bio_clone_bioset() call will be made while holding a reference a bio (result of the first clone) from the same bioset. This can potentially exhaust the bioset mempool and result in a memory allocation deadlock. Note that there is no race caused by reassigning cio.io->bio after already calling __map_bio(). This bio will only be dereferenced again after dec_pending() has found io->io_count to be zero, and this cannot happen before the dec_pending() call at the end of __split_and_process_bio(). To provide the clone bio when splitting, we use q->bio_split. This was previously being freed by bio-based dm to avoid having excess rescuer threads. As bio_split bio sets no longer create rescuer threads, there is little cost and much gain from restoring the q->bio_split bio set. Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-09-05 23:43:28 +00:00
bio_chain(b, bio);
trace_block_split(md->queue, b, bio->bi_iter.bi_sector);
ret = generic_make_request(bio);
dm: ensure bio submission follows a depth-first tree walk A dm device can, in general, represent a tree of targets, each of which handles a sub-range of the range of blocks handled by the parent. The bio sequencing managed by generic_make_request() requires that bios are generated and handled in a depth-first manner. Each call to a make_request_fn() may submit bios to a single member device, and may submit bios for a reduced region of the same device as the make_request_fn. In particular, any bios submitted to member devices must be expected to be processed in order, so a later one must never wait for an earlier one. This ordering is usually achieved by using bio_split() to reduce a bio to a size that can be completely handled by one target, and resubmitting the remainder to the originating device. bio_queue_split() shows the canonical approach. dm doesn't follow this approach, largely because it has needed to split bios since long before bio_split() was available. It currently can submit bios to separate targets within the one dm_make_request() call. Dependencies between these targets, as can happen with dm-snap, can cause deadlocks if either bios gets stuck behind the other in the queues managed by generic_make_request(). This requires the 'rescue' functionality provided by dm_offload_{start,end}. Some of this requirement can be removed by changing the order of bio submission to follow the canonical approach. That is, if dm finds that it needs to split a bio, the remainder should be sent to generic_make_request() rather than being handled immediately. This delays the handling until the first part is completely processed, so the deadlock problems do not occur. __split_and_process_bio() can be called both from dm_make_request() and from dm_wq_work(). When called from dm_wq_work() the current approach is perfectly satisfactory as each bio will be processed immediately. When called from dm_make_request(), current->bio_list will be non-NULL, and in this case it is best to create a separate "clone" bio for the remainder. When we use bio_clone_bioset() to split off the front part of a bio and chain the two together and submit the remainder to generic_make_request(), it is important that the newly allocated bio is used as the head to be processed immediately, and the original bio gets "bio_advance()"d and sent to generic_make_request() as the remainder. Otherwise, if the newly allocated bio is used as the remainder, and if it then needs to be split again, then the next bio_clone_bioset() call will be made while holding a reference a bio (result of the first clone) from the same bioset. This can potentially exhaust the bioset mempool and result in a memory allocation deadlock. Note that there is no race caused by reassigning cio.io->bio after already calling __map_bio(). This bio will only be dereferenced again after dec_pending() has found io->io_count to be zero, and this cannot happen before the dec_pending() call at the end of __split_and_process_bio(). To provide the clone bio when splitting, we use q->bio_split. This was previously being freed by bio-based dm to avoid having excess rescuer threads. As bio_split bio sets no longer create rescuer threads, there is little cost and much gain from restoring the q->bio_split bio set. Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-09-05 23:43:28 +00:00
break;
}
}
dm: implement REQ_FLUSH/FUA support for bio-based dm This patch converts bio-based dm to support REQ_FLUSH/FUA instead of now deprecated REQ_HARDBARRIER. * -EOPNOTSUPP handling logic dropped. * Preflush is handled as before but postflush is dropped and replaced with passing down REQ_FUA to member request_queues. This replaces one array wide cache flush w/ member specific FUA writes. * __split_and_process_bio() now calls __clone_and_map_flush() directly for flushes and guarantees all FLUSH bio's going to targets are zero ` length. * It's now guaranteed that all FLUSH bio's which are passed onto dm targets are zero length. bio_empty_barrier() tests are replaced with REQ_FLUSH tests. * Empty WRITE_BARRIERs are replaced with WRITE_FLUSHes. * Dropped unlikely() around REQ_FLUSH tests. Flushes are not unlikely enough to be marked with unlikely(). * Block layer now filters out REQ_FLUSH/FUA bio's if the request_queue doesn't support cache flushing. Advertise REQ_FLUSH | REQ_FUA capability. * Request based dm isn't converted yet. dm_init_request_based_queue() resets flush support to 0 for now. To avoid disturbing request based dm code, dm->flush_error is added for bio based dm while requested based dm continues to use dm->barrier_error. Lightly tested linear, stripe, raid1, snap and crypt targets. Please proceed with caution as I'm not familiar with the code base. Signed-off-by: Tejun Heo <tj@kernel.org> Cc: dm-devel@redhat.com Cc: Christoph Hellwig <hch@lst.de> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-03 09:56:19 +00:00
}
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
/* drop the extra reference count */
dec_pending(ci.io, errno_to_blk_status(error));
return ret;
dm: impose configurable deadline for dm_request_fn's merge heuristic Otherwise, for sequential workloads, the dm_request_fn can allow excessive request merging at the expense of increased service time. Add a per-device sysfs attribute to allow the user to control how long a request, that is a reasonable merge candidate, can be queued on the request queue. The resolution of this request dispatch deadline is in microseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadline The dm_request_fn's merge heuristic and associated extra accounting is disabled by default (rq_based_seq_io_merge_deadline is 0). This sysfs attribute is not applicable to bio-based DM devices so it will only ever report 0 for them. By allowing a request to remain on the queue it will block others requests on the queue. But introducing a short dequeue delay has proven very effective at enabling certain sequential IO workloads on really fast, yet IOPS constrained, devices to build up slightly larger IOs -- yielding 90+% throughput improvements. Having precise control over the time taken to wait for larger requests to build affords control beyond that of waiting for certain IO sizes to accumulate (which would require a deadline anyway). This knob will only ever make sense with sequential IO workloads and the particular value used is storage configuration specific. Given the expected niche use-case for when this knob is useful it has been deemed acceptable to expose this relatively crude method for crafting optimal IO on specific storage -- especially given the solution is simple yet effective. In the context of DM multipath, it is advisable to tune this sysfs attribute to a value that offers the best performance for the common case (e.g. if 4 paths are expected active, tune for that; if paths fail then performance may be slightly reduced). Alternatives were explored to have request-based DM autotune this value (e.g. if/when paths fail) but they were quickly deemed too fragile and complex to warrant further design and development time. If this problem proves more common as faster storage emerges we'll have to look at elevating a generic solution into the block core. Tested-by: Shiva Krishna Merla <shivakrishna.merla@netapp.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2015-02-26 05:50:28 +00:00
}
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
/*
* Optimized variant of __split_and_process_bio that leverages the
* fact that targets that use it do _not_ have a need to split bios.
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
*/
static blk_qc_t __process_bio(struct mapped_device *md, struct dm_table *map,
struct bio *bio, struct dm_target *ti)
{
struct clone_info ci;
blk_qc_t ret = BLK_QC_T_NONE;
int error = 0;
init_clone_info(&ci, md, map, bio);
if (bio->bi_opf & REQ_PREFLUSH) {
struct bio flush_bio;
/*
* Use an on-stack bio for this, it's safe since we don't
* need to reference it after submit. It's just used as
* the basis for the clone(s).
*/
bio_init(&flush_bio, NULL, 0);
flush_bio.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH | REQ_SYNC;
ci.bio = &flush_bio;
ci.sector_count = 0;
error = __send_empty_flush(&ci);
/* dec_pending submits any data associated with flush */
} else {
struct dm_target_io *tio;
ci.bio = bio;
ci.sector_count = bio_sectors(bio);
if (__process_abnormal_io(&ci, ti, &error))
goto out;
tio = alloc_tio(&ci, ti, 0, GFP_NOIO);
ret = __clone_and_map_simple_bio(&ci, tio, NULL);
}
out:
/* drop the extra reference count */
dec_pending(ci.io, errno_to_blk_status(error));
return ret;
}
static void dm_queue_split(struct mapped_device *md, struct dm_target *ti, struct bio **bio)
{
unsigned len, sector_count;
sector_count = bio_sectors(*bio);
len = min_t(sector_t, max_io_len((*bio)->bi_iter.bi_sector, ti), sector_count);
if (sector_count > len) {
struct bio *split = bio_split(*bio, len, GFP_NOIO, &md->queue->bio_split);
bio_chain(split, *bio);
trace_block_split(md->queue, split, (*bio)->bi_iter.bi_sector);
generic_make_request(*bio);
*bio = split;
}
}
static blk_qc_t dm_process_bio(struct mapped_device *md,
struct dm_table *map, struct bio *bio)
{
blk_qc_t ret = BLK_QC_T_NONE;
struct dm_target *ti = md->immutable_target;
if (unlikely(!map)) {
bio_io_error(bio);
return ret;
}
if (!ti) {
ti = dm_table_find_target(map, bio->bi_iter.bi_sector);
if (unlikely(!ti || !dm_target_is_valid(ti))) {
bio_io_error(bio);
return ret;
}
}
/*
* If in ->make_request_fn we need to use blk_queue_split(), otherwise
* queue_limits for abnormal requests (e.g. discard, writesame, etc)
* won't be imposed.
*/
if (current->bio_list) {
if (is_abnormal_io(bio))
blk_queue_split(md->queue, &bio);
else
dm_queue_split(md, ti, &bio);
}
if (dm_get_md_type(md) == DM_TYPE_NVME_BIO_BASED)
return __process_bio(md, map, bio, ti);
else
return __split_and_process_bio(md, map, bio);
}
static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio)
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
{
struct mapped_device *md = q->queuedata;
blk_qc_t ret = BLK_QC_T_NONE;
int srcu_idx;
struct dm_table *map;
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
map = dm_get_live_table(md, &srcu_idx);
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
/* if we're suspended, we have to queue this io for later */
if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) {
dm_put_live_table(md, srcu_idx);
if (!(bio->bi_opf & REQ_RAHEAD))
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
queue_io(md, bio);
else
bio_io_error(bio);
return ret;
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
}
ret = dm_process_bio(md, map, bio);
dm_put_live_table(md, srcu_idx);
return ret;
}
static int dm_any_congested(void *congested_data, int bdi_bits)
{
int r = bdi_bits;
struct mapped_device *md = congested_data;
struct dm_table *map;
if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
if (dm_request_based(md)) {
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
/*
* With request-based DM we only need to check the
* top-level queue for congestion.
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
*/
r = md->queue->backing_dev_info->wb.state & bdi_bits;
} else {
map = dm_get_live_table_fast(md);
if (map)
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
r = dm_table_any_congested(map, bdi_bits);
dm_put_live_table_fast(md);
}
}
return r;
}
/*-----------------------------------------------------------------
* An IDR is used to keep track of allocated minor numbers.
*---------------------------------------------------------------*/
static void free_minor(int minor)
{
spin_lock(&_minor_lock);
idr_remove(&_minor_idr, minor);
spin_unlock(&_minor_lock);
}
/*
* See if the device with a specific minor # is free.
*/
static int specific_minor(int minor)
{
int r;
if (minor >= (1 << MINORBITS))
return -EINVAL;
idr_preload(GFP_KERNEL);
spin_lock(&_minor_lock);
r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
spin_unlock(&_minor_lock);
idr_preload_end();
if (r < 0)
return r == -ENOSPC ? -EBUSY : r;
return 0;
}
static int next_free_minor(int *minor)
{
int r;
idr_preload(GFP_KERNEL);
spin_lock(&_minor_lock);
r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
spin_unlock(&_minor_lock);
idr_preload_end();
if (r < 0)
return r;
*minor = r;
return 0;
}
static const struct block_device_operations dm_blk_dops;
static const struct dax_operations dm_dax_ops;
static void dm_wq_work(struct work_struct *work);
static void dm_init_normal_md_queue(struct mapped_device *md)
{
/*
* Initialize aspects of queue that aren't relevant for blk-mq
*/
md->queue->backing_dev_info->congested_fn = dm_any_congested;
dm: do not initialise full request queue when bio based Change bio-based mapped devices no longer to have a fully initialized request_queue (request_fn, elevator, etc). This means bio-based DM devices no longer register elevator sysfs attributes ('iosched/' tree or 'scheduler' other than "none"). In contrast, a request-based DM device will continue to have a full request_queue and will register elevator sysfs attributes. Therefore a user can determine a DM device's type by checking if elevator sysfs attributes exist. First allocate a minimalist request_queue structure for a DM device (needed for both bio and request-based DM). Initialization of a full request_queue is deferred until it is known that the DM device is request-based, at the end of the table load sequence. Factor DM device's request_queue initialization: - common to both request-based and bio-based into dm_init_md_queue(). - specific to request-based into dm_init_request_based_queue(). The md->type_lock mutex is used to protect md->queue, in addition to md->type, during table_load(). A DM device's first table_load will establish the immutable md->type. But md->queue initialization, based on md->type, may fail at that time (because blk_init_allocated_queue cannot allocate memory). Therefore any subsequent table_load must (re)try dm_setup_md_queue independently of establishing md->type. Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:14:02 +00:00
}
static void cleanup_mapped_device(struct mapped_device *md)
{
if (md->wq)
destroy_workqueue(md->wq);
bioset_exit(&md->bs);
bioset_exit(&md->io_bs);
if (md->dax_dev) {
kill_dax(md->dax_dev);
put_dax(md->dax_dev);
md->dax_dev = NULL;
}
if (md->disk) {
spin_lock(&_minor_lock);
md->disk->private_data = NULL;
spin_unlock(&_minor_lock);
del_gendisk(md->disk);
put_disk(md->disk);
}
if (md->queue)
blk_cleanup_queue(md->queue);
cleanup_srcu_struct(&md->io_barrier);
if (md->bdev) {
bdput(md->bdev);
md->bdev = NULL;
}
mutex_destroy(&md->suspend_lock);
mutex_destroy(&md->type_lock);
mutex_destroy(&md->table_devices_lock);
dm_mq_cleanup_mapped_device(md);
}
/*
* Allocate and initialise a blank device with a given minor.
*/
static struct mapped_device *alloc_dev(int minor)
{
int r, numa_node_id = dm_get_numa_node();
struct dax_device *dax_dev = NULL;
struct mapped_device *md;
void *old_md;
md = kvzalloc_node(sizeof(*md), GFP_KERNEL, numa_node_id);
if (!md) {
DMWARN("unable to allocate device, out of memory.");
return NULL;
}
if (!try_module_get(THIS_MODULE))
goto bad_module_get;
/* get a minor number for the dev */
if (minor == DM_ANY_MINOR)
r = next_free_minor(&minor);
else
r = specific_minor(minor);
if (r < 0)
goto bad_minor;
r = init_srcu_struct(&md->io_barrier);
if (r < 0)
goto bad_io_barrier;
md->numa_node_id = numa_node_id;
md->init_tio_pdu = false;
md->type = DM_TYPE_NONE;
mutex_init(&md->suspend_lock);
mutex_init(&md->type_lock);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
mutex_init(&md->table_devices_lock);
spin_lock_init(&md->deferred_lock);
atomic_set(&md->holders, 1);
atomic_set(&md->open_count, 0);
atomic_set(&md->event_nr, 0);
atomic_set(&md->uevent_seq, 0);
INIT_LIST_HEAD(&md->uevent_list);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
INIT_LIST_HEAD(&md->table_devices);
spin_lock_init(&md->uevent_lock);
md->queue = blk_alloc_queue_node(GFP_KERNEL, numa_node_id);
if (!md->queue)
goto bad;
md->queue->queuedata = md;
md->queue->backing_dev_info->congested_data = md;
md->disk = alloc_disk_node(1, md->numa_node_id);
if (!md->disk)
goto bad;
init_waitqueue_head(&md->wait);
INIT_WORK(&md->work, dm_wq_work);
init_waitqueue_head(&md->eventq);
init_completion(&md->kobj_holder.completion);
md->disk->major = _major;
md->disk->first_minor = minor;
md->disk->fops = &dm_blk_dops;
md->disk->queue = md->queue;
md->disk->private_data = md;
sprintf(md->disk->disk_name, "dm-%d", minor);
if (IS_ENABLED(CONFIG_DAX_DRIVER)) {
dax_dev = alloc_dax(md, md->disk->disk_name, &dm_dax_ops);
if (!dax_dev)
goto bad;
}
md->dax_dev = dax_dev;
add_disk_no_queue_reg(md->disk);
format_dev_t(md->name, MKDEV(_major, minor));
md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
if (!md->wq)
goto bad;
md->bdev = bdget_disk(md->disk, 0);
if (!md->bdev)
goto bad;
dm_stats_init(&md->stats);
/* Populate the mapping, nobody knows we exist yet */
spin_lock(&_minor_lock);
old_md = idr_replace(&_minor_idr, md, minor);
spin_unlock(&_minor_lock);
BUG_ON(old_md != MINOR_ALLOCED);
return md;
bad:
cleanup_mapped_device(md);
bad_io_barrier:
free_minor(minor);
bad_minor:
module_put(THIS_MODULE);
bad_module_get:
kvfree(md);
return NULL;
}
dm: fix thaw_bdev This patch fixes a bd_mount_sem counter corruption bug in device-mapper. thaw_bdev() should be called only when freeze_bdev() was called for the device. Otherwise, thaw_bdev() will up bd_mount_sem and corrupt the semaphore counter. struct block_device with the corrupted semaphore may remain in slab cache and be reused later. Attached patch will fix it by calling unlock_fs() instead. unlock_fs() will determine whether it should call thaw_bdev() by checking the device is frozen or not. Easy reproducer is: #!/bin/sh while [ 1 ]; do dmsetup --notable create a dmsetup --nolockfs suspend a dmsetup remove a done It's not easy to see the effect of corrupted semaphore. So I have tested with putting printk below in bdev_alloc_inode(): if (atomic_read(&ei->bdev.bd_mount_sem.count) != 1) printk(KERN_DEBUG "Incorrect semaphore count = %d (%p)\n", atomic_read(&ei->bdev.bd_mount_sem.count), &ei->bdev); Without the patch, I saw something like: Incorrect semaphore count = 17 (f2ab91c0) With the patch, the message didn't appear. The bug was introduced in 2.6.16 with this bug fix: commit d9dde59ba03095e526640988c0fedd75e93bc8b7 Date: Fri Feb 24 13:04:24 2006 -0800 [PATCH] dm: missing bdput/thaw_bdev at removal Need to unfreeze and release bdev otherwise the bdev inode with inconsistent state is reused later and cause problem. and backported to 2.6.15.5. It occurs only in free_dev(), which is called only when the dm device is removed. The buggy code is executed only if md->suspended_bdev is non-NULL and that can happen only when the device was suspended without noflush. Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: stable@kernel.org
2007-10-19 21:38:43 +00:00
static void unlock_fs(struct mapped_device *md);
static void free_dev(struct mapped_device *md)
{
int minor = MINOR(disk_devt(md->disk));
unlock_fs(md);
cleanup_mapped_device(md);
dm: allow active and inactive tables to share dm_devs Until this change, when loading a new DM table, DM core would re-open all of the devices in the DM table. Now, DM core will avoid redundant device opens (and closes when destroying the old table) if the old table already has a device open using the same mode. This is achieved by managing reference counts on the table_devices that DM core now stores in the mapped_device structure (rather than in the dm_table structure). So a mapped_device's active and inactive dm_tables' dm_dev lists now just point to the dm_devs stored in the mapped_device's table_devices list. This improvement in DM core's device reference counting has the side-effect of fixing a long-standing limitation of the multipath target: a DM multipath table couldn't include any paths that were unusable (failed). For example: if all paths have failed and you add a new, working, path to the table; you can't use it since the table load would fail due to it still containing failed paths. Now a re-load of a multipath table can include failed devices and when those devices become active again they can be used instantly. The device list code in dm.c isn't a straight copy/paste from the code in dm-table.c, but it's very close (aside from some variable renames). One subtle difference is that find_table_device for the tables_devices list will only match devices with the same name and mode. This is because we don't want to upgrade a device's mode in the active table when an inactive table is loaded. Access to the mapped_device structure's tables_devices list requires a mutex (tables_devices_lock), so that tables cannot be created and destroyed concurrently. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2014-08-13 18:53:43 +00:00
free_table_devices(&md->table_devices);
dm_stats_cleanup(&md->stats);
free_minor(minor);
module_put(THIS_MODULE);
kvfree(md);
}
static int __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
struct dm_md_mempools *p = dm_table_get_md_mempools(t);
int ret = 0;
if (dm_table_bio_based(t)) {
/*
* The md may already have mempools that need changing.
* If so, reload bioset because front_pad may have changed
* because a different table was loaded.
*/
bioset_exit(&md->bs);
bioset_exit(&md->io_bs);
} else if (bioset_initialized(&md->bs)) {
/*
* There's no need to reload with request-based dm
* because the size of front_pad doesn't change.
* Note for future: If you are to reload bioset,
* prep-ed requests in the queue may refer
* to bio from the old bioset, so you must walk
* through the queue to unprep.
*/
goto out;
}
BUG_ON(!p ||
bioset_initialized(&md->bs) ||
bioset_initialized(&md->io_bs));
ret = bioset_init_from_src(&md->bs, &p->bs);
if (ret)
goto out;
ret = bioset_init_from_src(&md->io_bs, &p->io_bs);
if (ret)
bioset_exit(&md->bs);
out:
/* mempool bind completed, no longer need any mempools in the table */
dm_table_free_md_mempools(t);
return ret;
}
/*
* Bind a table to the device.
*/
static void event_callback(void *context)
{
unsigned long flags;
LIST_HEAD(uevents);
struct mapped_device *md = (struct mapped_device *) context;
spin_lock_irqsave(&md->uevent_lock, flags);
list_splice_init(&md->uevent_list, &uevents);
spin_unlock_irqrestore(&md->uevent_lock, flags);
dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
atomic_inc(&md->event_nr);
wake_up(&md->eventq);
dm_issue_global_event();
}
/*
* Protected by md->suspend_lock obtained by dm_swap_table().
*/
static void __set_size(struct mapped_device *md, sector_t size)
{
lockdep_assert_held(&md->suspend_lock);
set_capacity(md->disk, size);
i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
}
/*
* Returns old map, which caller must destroy.
*/
static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t,
struct queue_limits *limits)
{
struct dm_table *old_map;
struct request_queue *q = md->queue;
bool request_based = dm_table_request_based(t);
sector_t size;
int ret;
lockdep_assert_held(&md->suspend_lock);
size = dm_table_get_size(t);
/*
* Wipe any geometry if the size of the table changed.
*/
if (size != dm_get_size(md))
memset(&md->geometry, 0, sizeof(md->geometry));
__set_size(md, size);
dm table: rework reference counting Rework table reference counting. The existing code uses a reference counter. When the last reference is dropped and the counter reaches zero, the table destructor is called. Table reference counters are acquired/released from upcalls from other kernel code (dm_any_congested, dm_merge_bvec, dm_unplug_all). If the reference counter reaches zero in one of the upcalls, the table destructor is called from almost random kernel code. This leads to various problems: * dm_any_congested being called under a spinlock, which calls the destructor, which calls some sleeping function. * the destructor attempting to take a lock that is already taken by the same process. * stale reference from some other kernel code keeps the table constructed, which keeps some devices open, even after successful return from "dmsetup remove". This can confuse lvm and prevent closing of underlying devices or reusing device minor numbers. The patch changes reference counting so that the table destructor can be called only at predetermined places. The table has always exactly one reference from either mapped_device->map or hash_cell->new_map. After this patch, this reference is not counted in table->holders. A pair of dm_create_table/dm_destroy_table functions is used for table creation/destruction. Temporary references from the other code increase table->holders. A pair of dm_table_get/dm_table_put functions is used to manipulate it. When the table is about to be destroyed, we wait for table->holders to reach 0. Then, we call the table destructor. We use active waiting with msleep(1), because the situation happens rarely (to one user in 5 years) and removing the device isn't performance-critical task: the user doesn't care if it takes one tick more or not. This way, the destructor is called only at specific points (dm_table_destroy function) and the above problems associated with lazy destruction can't happen. Finally remove the temporary protection added to dm_any_congested(). Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-01-06 03:05:10 +00:00
dm_table_event_callback(t, event_callback, md);
/*
* The queue hasn't been stopped yet, if the old table type wasn't
* for request-based during suspension. So stop it to prevent
* I/O mapping before resume.
* This must be done before setting the queue restrictions,
* because request-based dm may be run just after the setting.
*/
if (request_based)
dm_stop_queue(q);
if (request_based || md->type == DM_TYPE_NVME_BIO_BASED) {
/*
* Leverage the fact that request-based DM targets and
* NVMe bio based targets are immutable singletons
* - used to optimize both dm_request_fn and dm_mq_queue_rq;
* and __process_bio.
*/
md->immutable_target = dm_table_get_immutable_target(t);
}
ret = __bind_mempools(md, t);
if (ret) {
old_map = ERR_PTR(ret);
goto out;
}
old_map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
rcu_assign_pointer(md->map, (void *)t);
md->immutable_target_type = dm_table_get_immutable_target_type(t);
dm_table_set_restrictions(t, q, limits);
if (old_map)
dm_sync_table(md);
out:
return old_map;
}
/*
* Returns unbound table for the caller to free.
*/
static struct dm_table *__unbind(struct mapped_device *md)
{
struct dm_table *map = rcu_dereference_protected(md->map, 1);
if (!map)
return NULL;
dm_table_event_callback(map, NULL, NULL);
RCU_INIT_POINTER(md->map, NULL);
dm_sync_table(md);
return map;
}
/*
* Constructor for a new device.
*/
int dm_create(int minor, struct mapped_device **result)
{
int r;
struct mapped_device *md;
md = alloc_dev(minor);
if (!md)
return -ENXIO;
r = dm_sysfs_init(md);
if (r) {
free_dev(md);
return r;
}
*result = md;
return 0;
}
/*
* Functions to manage md->type.
* All are required to hold md->type_lock.
*/
void dm_lock_md_type(struct mapped_device *md)
{
mutex_lock(&md->type_lock);
}
void dm_unlock_md_type(struct mapped_device *md)
{
mutex_unlock(&md->type_lock);
}
void dm_set_md_type(struct mapped_device *md, enum dm_queue_mode type)
{
BUG_ON(!mutex_is_locked(&md->type_lock));
md->type = type;
}
enum dm_queue_mode dm_get_md_type(struct mapped_device *md)
{
return md->type;
}
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
return md->immutable_target_type;
}
dm mpath: disable WRITE SAME if it fails Workaround the SCSI layer's problematic WRITE SAME heuristics by disabling WRITE SAME in the DM multipath device's queue_limits if an underlying device disabled it. The WRITE SAME heuristics, with both the original commit 5db44863b6eb ("[SCSI] sd: Implement support for WRITE SAME") and the updated commit 66c28f971 ("[SCSI] sd: Update WRITE SAME heuristics"), default to enabling WRITE SAME(10) even without successfully determining it is supported. After the first failed WRITE SAME the SCSI layer will disable WRITE SAME for the device (by setting sdkp->device->no_write_same which results in 'max_write_same_sectors' in device's queue_limits to be set to 0). When a device is stacked ontop of such a SCSI device any changes to that SCSI device's queue_limits do not automatically propagate up the stack. As such, a DM multipath device will not have its WRITE SAME support disabled. This causes the block layer to continue to issue WRITE SAME requests to the mpath device which causes paths to fail and (if mpath IO isn't configured to queue when no paths are available) it will result in actual IO errors to the upper layers. This fix doesn't help configurations that have additional devices stacked ontop of the mpath device (e.g. LVM created linear DM devices ontop). A proper fix that restacks all the queue_limits from the bottom of the device stack up will need to be explored if SCSI will continue to use this model of optimistically allowing op codes and then disabling them after they fail for the first time. Before this patch: EXT4-fs (dm-6): mounted filesystem with ordered data mode. Opts: (null) device-mapper: multipath: XXX snitm debugging: got -EREMOTEIO (-121) device-mapper: multipath: XXX snitm debugging: failing WRITE SAME IO with error=-121 end_request: critical target error, dev dm-6, sector 528 dm-6: WRITE SAME failed. Manually zeroing. device-mapper: multipath: Failing path 8:112. end_request: I/O error, dev dm-6, sector 4616 dm-6: WRITE SAME failed. Manually zeroing. end_request: I/O error, dev dm-6, sector 4616 end_request: I/O error, dev dm-6, sector 5640 end_request: I/O error, dev dm-6, sector 6664 end_request: I/O error, dev dm-6, sector 7688 end_request: I/O error, dev dm-6, sector 524288 Buffer I/O error on device dm-6, logical block 65536 lost page write due to I/O error on dm-6 JBD2: Error -5 detected when updating journal superblock for dm-6-8. end_request: I/O error, dev dm-6, sector 524296 Aborting journal on device dm-6-8. end_request: I/O error, dev dm-6, sector 524288 Buffer I/O error on device dm-6, logical block 65536 lost page write due to I/O error on dm-6 JBD2: Error -5 detected when updating journal superblock for dm-6-8. # cat /sys/block/sdh/queue/write_same_max_bytes 0 # cat /sys/block/dm-6/queue/write_same_max_bytes 33553920 After this patch: EXT4-fs (dm-6): mounted filesystem with ordered data mode. Opts: (null) device-mapper: multipath: XXX snitm debugging: got -EREMOTEIO (-121) device-mapper: multipath: XXX snitm debugging: WRITE SAME I/O failed with error=-121 end_request: critical target error, dev dm-6, sector 528 dm-6: WRITE SAME failed. Manually zeroing. # cat /sys/block/sdh/queue/write_same_max_bytes 0 # cat /sys/block/dm-6/queue/write_same_max_bytes 0 It should be noted that WRITE SAME support wasn't enabled in DM multipath until v3.10. Signed-off-by: Mike Snitzer <snitzer@redhat.com> Cc: Martin K. Petersen <martin.petersen@oracle.com> Cc: Hannes Reinecke <hare@suse.de> Cc: stable@vger.kernel.org # 3.10+
2013-09-19 16:13:58 +00:00
/*
* The queue_limits are only valid as long as you have a reference
* count on 'md'.
*/
struct queue_limits *dm_get_queue_limits(struct mapped_device *md)
{
BUG_ON(!atomic_read(&md->holders));
return &md->queue->limits;
}
EXPORT_SYMBOL_GPL(dm_get_queue_limits);
dm: do not initialise full request queue when bio based Change bio-based mapped devices no longer to have a fully initialized request_queue (request_fn, elevator, etc). This means bio-based DM devices no longer register elevator sysfs attributes ('iosched/' tree or 'scheduler' other than "none"). In contrast, a request-based DM device will continue to have a full request_queue and will register elevator sysfs attributes. Therefore a user can determine a DM device's type by checking if elevator sysfs attributes exist. First allocate a minimalist request_queue structure for a DM device (needed for both bio and request-based DM). Initialization of a full request_queue is deferred until it is known that the DM device is request-based, at the end of the table load sequence. Factor DM device's request_queue initialization: - common to both request-based and bio-based into dm_init_md_queue(). - specific to request-based into dm_init_request_based_queue(). The md->type_lock mutex is used to protect md->queue, in addition to md->type, during table_load(). A DM device's first table_load will establish the immutable md->type. But md->queue initialization, based on md->type, may fail at that time (because blk_init_allocated_queue cannot allocate memory). Therefore any subsequent table_load must (re)try dm_setup_md_queue independently of establishing md->type. Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:14:02 +00:00
/*
* Setup the DM device's queue based on md's type
*/
int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t)
dm: do not initialise full request queue when bio based Change bio-based mapped devices no longer to have a fully initialized request_queue (request_fn, elevator, etc). This means bio-based DM devices no longer register elevator sysfs attributes ('iosched/' tree or 'scheduler' other than "none"). In contrast, a request-based DM device will continue to have a full request_queue and will register elevator sysfs attributes. Therefore a user can determine a DM device's type by checking if elevator sysfs attributes exist. First allocate a minimalist request_queue structure for a DM device (needed for both bio and request-based DM). Initialization of a full request_queue is deferred until it is known that the DM device is request-based, at the end of the table load sequence. Factor DM device's request_queue initialization: - common to both request-based and bio-based into dm_init_md_queue(). - specific to request-based into dm_init_request_based_queue(). The md->type_lock mutex is used to protect md->queue, in addition to md->type, during table_load(). A DM device's first table_load will establish the immutable md->type. But md->queue initialization, based on md->type, may fail at that time (because blk_init_allocated_queue cannot allocate memory). Therefore any subsequent table_load must (re)try dm_setup_md_queue independently of establishing md->type. Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:14:02 +00:00
{
int r;
struct queue_limits limits;
enum dm_queue_mode type = dm_get_md_type(md);
switch (type) {
case DM_TYPE_REQUEST_BASED:
r = dm_mq_init_request_queue(md, t);
if (r) {
DMERR("Cannot initialize queue for request-based dm-mq mapped device");
return r;
}
break;
case DM_TYPE_BIO_BASED:
case DM_TYPE_DAX_BIO_BASED:
case DM_TYPE_NVME_BIO_BASED:
dm_init_normal_md_queue(md);
blk_queue_make_request(md->queue, dm_make_request);
break;
case DM_TYPE_NONE:
WARN_ON_ONCE(true);
break;
dm: do not initialise full request queue when bio based Change bio-based mapped devices no longer to have a fully initialized request_queue (request_fn, elevator, etc). This means bio-based DM devices no longer register elevator sysfs attributes ('iosched/' tree or 'scheduler' other than "none"). In contrast, a request-based DM device will continue to have a full request_queue and will register elevator sysfs attributes. Therefore a user can determine a DM device's type by checking if elevator sysfs attributes exist. First allocate a minimalist request_queue structure for a DM device (needed for both bio and request-based DM). Initialization of a full request_queue is deferred until it is known that the DM device is request-based, at the end of the table load sequence. Factor DM device's request_queue initialization: - common to both request-based and bio-based into dm_init_md_queue(). - specific to request-based into dm_init_request_based_queue(). The md->type_lock mutex is used to protect md->queue, in addition to md->type, during table_load(). A DM device's first table_load will establish the immutable md->type. But md->queue initialization, based on md->type, may fail at that time (because blk_init_allocated_queue cannot allocate memory). Therefore any subsequent table_load must (re)try dm_setup_md_queue independently of establishing md->type. Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:14:02 +00:00
}
r = dm_calculate_queue_limits(t, &limits);
if (r) {
DMERR("Cannot calculate initial queue limits");
return r;
}
dm_table_set_restrictions(t, md->queue, &limits);
blk_register_queue(md->disk);
dm: do not initialise full request queue when bio based Change bio-based mapped devices no longer to have a fully initialized request_queue (request_fn, elevator, etc). This means bio-based DM devices no longer register elevator sysfs attributes ('iosched/' tree or 'scheduler' other than "none"). In contrast, a request-based DM device will continue to have a full request_queue and will register elevator sysfs attributes. Therefore a user can determine a DM device's type by checking if elevator sysfs attributes exist. First allocate a minimalist request_queue structure for a DM device (needed for both bio and request-based DM). Initialization of a full request_queue is deferred until it is known that the DM device is request-based, at the end of the table load sequence. Factor DM device's request_queue initialization: - common to both request-based and bio-based into dm_init_md_queue(). - specific to request-based into dm_init_request_based_queue(). The md->type_lock mutex is used to protect md->queue, in addition to md->type, during table_load(). A DM device's first table_load will establish the immutable md->type. But md->queue initialization, based on md->type, may fail at that time (because blk_init_allocated_queue cannot allocate memory). Therefore any subsequent table_load must (re)try dm_setup_md_queue independently of establishing md->type. Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:14:02 +00:00
return 0;
}
struct mapped_device *dm_get_md(dev_t dev)
{
struct mapped_device *md;
unsigned minor = MINOR(dev);
if (MAJOR(dev) != _major || minor >= (1 << MINORBITS))
return NULL;
spin_lock(&_minor_lock);
md = idr_find(&_minor_idr, minor);
if (!md || md == MINOR_ALLOCED || (MINOR(disk_devt(dm_disk(md))) != minor) ||
test_bit(DMF_FREEING, &md->flags) || dm_deleting_md(md)) {
md = NULL;
goto out;
}
dm_get(md);
out:
spin_unlock(&_minor_lock);
return md;
}
EXPORT_SYMBOL_GPL(dm_get_md);
void *dm_get_mdptr(struct mapped_device *md)
{
return md->interface_ptr;
}
void dm_set_mdptr(struct mapped_device *md, void *ptr)
{
md->interface_ptr = ptr;
}
void dm_get(struct mapped_device *md)
{
atomic_inc(&md->holders);
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
BUG_ON(test_bit(DMF_FREEING, &md->flags));
}
int dm_hold(struct mapped_device *md)
{
spin_lock(&_minor_lock);
if (test_bit(DMF_FREEING, &md->flags)) {
spin_unlock(&_minor_lock);
return -EBUSY;
}
dm_get(md);
spin_unlock(&_minor_lock);
return 0;
}
EXPORT_SYMBOL_GPL(dm_hold);
const char *dm_device_name(struct mapped_device *md)
{
return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
static void __dm_destroy(struct mapped_device *md, bool wait)
{
struct dm_table *map;
int srcu_idx;
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
might_sleep();
spin_lock(&_minor_lock);
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
set_bit(DMF_FREEING, &md->flags);
spin_unlock(&_minor_lock);
blk_set_queue_dying(md->queue);
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
/*
* Take suspend_lock so that presuspend and postsuspend methods
* do not race with internal suspend.
*/
mutex_lock(&md->suspend_lock);
map = dm_get_live_table(md, &srcu_idx);
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
if (!dm_suspended_md(md)) {
dm_table_presuspend_targets(map);
dm_table_postsuspend_targets(map);
}
/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
dm_put_live_table(md, srcu_idx);
mutex_unlock(&md->suspend_lock);
dm: separate device deletion from dm_put This patch separates the device deletion code from dm_put() to make sure the deletion happens in the process context. By this patch, device deletion always occurs in an ioctl (process) context and dm_put() can be called in interrupt context. As a result, the request-based dm's bad dm_put() usage pointed out by Mikulas below disappears. http://marc.info/?l=dm-devel&m=126699981019735&w=2 Without this patch, I confirmed there is a case to crash the system: dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt()) Some more backgrounds and details: In request-based dm, a device opener can remove a mapped_device while the last request is still completing, because bios in the last request complete first and then the device opener can close and remove the mapped_device before the last request completes: CPU0 CPU1 ================================================================= <<INTERRUPT>> blk_end_request_all(clone_rq) blk_update_request(clone_rq) bio_endio(clone_bio) == end_clone_bio blk_update_request(orig_rq) bio_endio(orig_bio) <<I/O completed>> dm_blk_close() dev_remove() dm_put(md) <<Free md>> blk_finish_request(clone_rq) .... dm_end_request(clone_rq) free_rq_clone(clone_rq) blk_end_request_all(orig_rq) rq_completed(md) So request-based dm used dm_get()/dm_put() to hold md for each I/O until its request completion handling is fully done. However, the final dm_put() can call the device deletion code which must not be run in interrupt context and may cause kernel panic. To solve the problem, this patch moves the device deletion code, dm_destroy(), to predetermined places that is actually deleting the mapped_device in ioctl (process) context, and changes dm_put() just to decrement the reference count of the mapped_device. By this change, dm_put() can be used in any context and the symmetric model below is introduced: dm_create(): create a mapped_device dm_destroy(): destroy a mapped_device dm_get(): increment the reference count of a mapped_device dm_put(): decrement the reference count of a mapped_device dm_destroy() waits for all references of the mapped_device to disappear, then deletes the mapped_device. dm_destroy() uses active waiting with msleep(1), since deleting the mapped_device isn't performance-critical task. And since at this point, nobody opens the mapped_device and no new reference will be taken, the pending counts are just for racing completing activity and will eventually decrease to zero. For the unlikely case of the forced module unload, dm_destroy_immediate(), which doesn't wait and forcibly deletes the mapped_device, is also introduced and used in dm_hash_remove_all(). Otherwise, "rmmod -f" may be stuck and never return. And now, because the mapped_device is deleted at this point, subsequent accesses to the mapped_device may cause NULL pointer references. Cc: stable@kernel.org Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-08-12 03:13:56 +00:00
/*
* Rare, but there may be I/O requests still going to complete,
* for example. Wait for all references to disappear.
* No one should increment the reference count of the mapped_device,
* after the mapped_device state becomes DMF_FREEING.
*/
if (wait)
while (atomic_read(&md->holders))
msleep(1);
else if (atomic_read(&md->holders))
DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)",
dm_device_name(md), atomic_read(&md->holders));
dm_sysfs_exit(md);
dm_table_destroy(__unbind(md));
free_dev(md);
}
void dm_destroy(struct mapped_device *md)
{
__dm_destroy(md, true);
}
void dm_destroy_immediate(struct mapped_device *md)
{
__dm_destroy(md, false);
}
void dm_put(struct mapped_device *md)
{
atomic_dec(&md->holders);
}
EXPORT_SYMBOL_GPL(dm_put);
static int dm_wait_for_completion(struct mapped_device *md, long task_state)
{
int r = 0;
DEFINE_WAIT(wait);
while (1) {
prepare_to_wait(&md->wait, &wait, task_state);
if (!md_in_flight(md))
break;
if (signal_pending_state(task_state, current)) {
r = -EINTR;
break;
}
io_schedule();
}
finish_wait(&md->wait, &wait);
return r;
}
/*
* Process the deferred bios
*/
static void dm_wq_work(struct work_struct *work)
{
struct mapped_device *md = container_of(work, struct mapped_device,
work);
struct bio *c;
int srcu_idx;
struct dm_table *map;
map = dm_get_live_table(md, &srcu_idx);
while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
spin_lock_irq(&md->deferred_lock);
c = bio_list_pop(&md->deferred);
spin_unlock_irq(&md->deferred_lock);
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
if (!c)
break;
if (dm_request_based(md))
(void) generic_make_request(c);
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
else
(void) dm_process_bio(md, map, c);
}
dm_put_live_table(md, srcu_idx);
}
static void dm_queue_flush(struct mapped_device *md)
{
clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
smp_mb__after_atomic();
queue_work(md->wq, &md->work);
}
/*
* Swap in a new table, returning the old one for the caller to destroy.
*/
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
{
struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
struct queue_limits limits;
int r;
mutex_lock(&md->suspend_lock);
/* device must be suspended */
if (!dm_suspended_md(md))
goto out;
/*
* If the new table has no data devices, retain the existing limits.
* This helps multipath with queue_if_no_path if all paths disappear,
* then new I/O is queued based on these limits, and then some paths
* reappear.
*/
if (dm_table_has_no_data_devices(table)) {
live_map = dm_get_live_table_fast(md);
if (live_map)
limits = md->queue->limits;
dm_put_live_table_fast(md);
}
if (!live_map) {
r = dm_calculate_queue_limits(table, &limits);
if (r) {
map = ERR_PTR(r);
goto out;
}
}
map = __bind(md, table, &limits);
dm_issue_global_event();
out:
mutex_unlock(&md->suspend_lock);
return map;
}
/*
* Functions to lock and unlock any filesystem running on the
* device.
*/
static int lock_fs(struct mapped_device *md)
{
int r;
WARN_ON(md->frozen_sb);
md->frozen_sb = freeze_bdev(md->bdev);
if (IS_ERR(md->frozen_sb)) {
r = PTR_ERR(md->frozen_sb);
md->frozen_sb = NULL;
return r;
}
set_bit(DMF_FROZEN, &md->flags);
return 0;
}
static void unlock_fs(struct mapped_device *md)
{
if (!test_bit(DMF_FROZEN, &md->flags))
return;
thaw_bdev(md->bdev, md->frozen_sb);
md->frozen_sb = NULL;
clear_bit(DMF_FROZEN, &md->flags);
}
/*
* @suspend_flags: DM_SUSPEND_LOCKFS_FLAG and/or DM_SUSPEND_NOFLUSH_FLAG
* @task_state: e.g. TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE
* @dmf_suspended_flag: DMF_SUSPENDED or DMF_SUSPENDED_INTERNALLY
*
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
* If __dm_suspend returns 0, the device is completely quiescent
* now. There is no request-processing activity. All new requests
* are being added to md->deferred list.
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
*/
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
static int __dm_suspend(struct mapped_device *md, struct dm_table *map,
unsigned suspend_flags, long task_state,
int dmf_suspended_flag)
{
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
bool do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG;
bool noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG;
int r;
lockdep_assert_held(&md->suspend_lock);
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
/*
* DMF_NOFLUSH_SUSPENDING must be set before presuspend.
* This flag is cleared before dm_suspend returns.
*/
if (noflush)
set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
else
pr_debug("%s: suspending with flush\n", dm_device_name(md));
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
/*
* This gets reverted if there's an error later and the targets
* provide the .presuspend_undo hook.
*/
dm_table_presuspend_targets(map);
/*
dm: simplify request based suspend The semantics of bio-based dm were changed recently in the case of suspend with "--nolockfs" but without "--noflush". Before 2.6.30, I/Os submitted before the suspend invocation were always flushed. From 2.6.30 onwards, I/Os submitted before the suspend invocation might not be flushed. (For details, see http://marc.info/?t=123994433400003&r=1&w=2) This patch brings the behaviour of request-based dm into line with bio-based dm, simplifying the code and preparing for a subsequent patch that will wait for all in_flight I/Os to complete without stopping request_queue and use dm_wait_for_completion() for it. This change in semantics simplifies the suspend code as follows: o Suspend is implemented as stopping request_queue in request-based dm, and all I/Os are queued in the request_queue even after suspend is invoked. o In the old semantics, we had to track whether I/Os were queued before or after the suspend invocation, so a special barrier-like request called 'suspend marker' was introduced. o With the new semantics, we don't need to flush any I/O so we can remove the marker and the code related to the marker handling and I/O flushing. After removing this codes, the suspend sequence is now: 1. Flush all I/Os by lock_fs() if needed. 2. Stop dispatching any I/O by stopping the request_queue. 3. Wait for all in-flight I/Os to be completed or requeued. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-12-10 23:52:16 +00:00
* Flush I/O to the device.
* Any I/O submitted after lock_fs() may not be flushed.
* noflush takes precedence over do_lockfs.
* (lock_fs() flushes I/Os and waits for them to complete.)
*/
if (!noflush && do_lockfs) {
r = lock_fs(md);
if (r) {
dm_table_presuspend_undo_targets(map);
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
return r;
}
}
/*
* Here we must make sure that no processes are submitting requests
* to target drivers i.e. no one may be executing
* __split_and_process_bio. This is called from dm_request and
* dm_wq_work.
*
* To get all processes out of __split_and_process_bio in dm_request,
* we take the write lock. To prevent any process from reentering
dm: relax ordering of bio-based flush implementation Unlike REQ_HARDBARRIER, REQ_FLUSH/FUA doesn't mandate any ordering against other bio's. This patch relaxes ordering around flushes. * A flush bio is no longer deferred to workqueue directly. It's processed like other bio's but __split_and_process_bio() uses md->flush_bio as the clone source. md->flush_bio is initialized to empty flush during md initialization and shared for all flushes. * As a flush bio now travels through the same execution path as other bio's, there's no need for dedicated error handling path either. It can use the same error handling path in dec_pending(). Dedicated error handling removed along with md->flush_error. * When dec_pending() detects that a flush has completed, it checks whether the original bio has data. If so, the bio is queued to the deferred list w/ REQ_FLUSH cleared; otherwise, it's completed. * As flush sequencing is handled in the usual issue/completion path, dm_wq_work() no longer needs to handle flushes differently. Now its only responsibility is re-issuing deferred bio's the same way as _dm_request() would. REQ_FLUSH handling logic including process_flush() is dropped. * There's no reason for queue_io() and dm_wq_work() write lock dm->io_lock. queue_io() now only uses md->deferred_lock and dm_wq_work() read locks dm->io_lock. * bio's no longer need to be queued on the deferred list while a flush is in progress making DMF_QUEUE_IO_TO_THREAD unncessary. Drop it. This avoids stalling the device during flushes and simplifies the implementation. Signed-off-by: Tejun Heo <tj@kernel.org> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2010-09-08 16:07:00 +00:00
* __split_and_process_bio from dm_request and quiesce the thread
* (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call
* flush_workqueue(md->wq).
*/
set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
if (map)
synchronize_srcu(&md->io_barrier);
dm: add request based barrier support This patch adds barrier support for request-based dm. CORE DESIGN The design is basically same as bio-based dm, which emulates barrier by mapping empty barrier bios before/after a barrier I/O. But request-based dm has been using struct request_queue for I/O queueing, so the block-layer's barrier mechanism can be used. o Summary of the block-layer's behavior (which is depended by dm-core) Request-based dm uses QUEUE_ORDERED_DRAIN_FLUSH ordered mode for I/O barrier. It means that when an I/O requiring barrier is found in the request_queue, the block-layer makes pre-flush request and post-flush request just before and just after the I/O respectively. After the ordered sequence starts, the block-layer waits for all in-flight I/Os to complete, then gives drivers the pre-flush request, the barrier I/O and the post-flush request one by one. It means that the request_queue is stopped automatically by the block-layer until drivers complete each sequence. o dm-core For the barrier I/O, treats it as a normal I/O, so no additional code is needed. For the pre/post-flush request, flushes caches by the followings: 1. Make the number of empty barrier requests required by target's num_flush_requests, and map them (dm_rq_barrier()). 2. Waits for the mapped barriers to complete (dm_rq_barrier()). If error has occurred, save the error value to md->barrier_error (dm_end_request()). (*) Basically, the first reported error is taken. But -EOPNOTSUPP supersedes any error and DM_ENDIO_REQUEUE follows. 3. Requeue the pre/post-flush request if the error value is DM_ENDIO_REQUEUE. Otherwise, completes with the error value (dm_rq_barrier_work()). The pre/post-flush work above is done in the kernel thread (kdmflush) context, since memory allocation which might sleep is needed in dm_rq_barrier() but sleep is not allowed in dm_request_fn(), which is an irq-disabled context. Also, clones of the pre/post-flush request share an original, so such clones can't be completed using the softirq context. Instead, complete them in the context of underlying device drivers. It should be safe since there is no I/O dispatching during the completion of such clones. For suspend, the workqueue of kdmflush needs to be flushed after the request_queue has been stopped. Otherwise, the next flush work can be kicked even after the suspend completes. TARGET INTERFACE No new interface is added. Just use the existing num_flush_requests in struct target_type as same as bio-based dm. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-12-10 23:52:18 +00:00
/*
* Stop md->queue before flushing md->wq in case request-based
* dm defers requests to md->wq from md->queue.
dm: add request based barrier support This patch adds barrier support for request-based dm. CORE DESIGN The design is basically same as bio-based dm, which emulates barrier by mapping empty barrier bios before/after a barrier I/O. But request-based dm has been using struct request_queue for I/O queueing, so the block-layer's barrier mechanism can be used. o Summary of the block-layer's behavior (which is depended by dm-core) Request-based dm uses QUEUE_ORDERED_DRAIN_FLUSH ordered mode for I/O barrier. It means that when an I/O requiring barrier is found in the request_queue, the block-layer makes pre-flush request and post-flush request just before and just after the I/O respectively. After the ordered sequence starts, the block-layer waits for all in-flight I/Os to complete, then gives drivers the pre-flush request, the barrier I/O and the post-flush request one by one. It means that the request_queue is stopped automatically by the block-layer until drivers complete each sequence. o dm-core For the barrier I/O, treats it as a normal I/O, so no additional code is needed. For the pre/post-flush request, flushes caches by the followings: 1. Make the number of empty barrier requests required by target's num_flush_requests, and map them (dm_rq_barrier()). 2. Waits for the mapped barriers to complete (dm_rq_barrier()). If error has occurred, save the error value to md->barrier_error (dm_end_request()). (*) Basically, the first reported error is taken. But -EOPNOTSUPP supersedes any error and DM_ENDIO_REQUEUE follows. 3. Requeue the pre/post-flush request if the error value is DM_ENDIO_REQUEUE. Otherwise, completes with the error value (dm_rq_barrier_work()). The pre/post-flush work above is done in the kernel thread (kdmflush) context, since memory allocation which might sleep is needed in dm_rq_barrier() but sleep is not allowed in dm_request_fn(), which is an irq-disabled context. Also, clones of the pre/post-flush request share an original, so such clones can't be completed using the softirq context. Instead, complete them in the context of underlying device drivers. It should be safe since there is no I/O dispatching during the completion of such clones. For suspend, the workqueue of kdmflush needs to be flushed after the request_queue has been stopped. Otherwise, the next flush work can be kicked even after the suspend completes. TARGET INTERFACE No new interface is added. Just use the existing num_flush_requests in struct target_type as same as bio-based dm. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-12-10 23:52:18 +00:00
*/
if (dm_request_based(md))
dm_stop_queue(md->queue);
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
dm: add request based barrier support This patch adds barrier support for request-based dm. CORE DESIGN The design is basically same as bio-based dm, which emulates barrier by mapping empty barrier bios before/after a barrier I/O. But request-based dm has been using struct request_queue for I/O queueing, so the block-layer's barrier mechanism can be used. o Summary of the block-layer's behavior (which is depended by dm-core) Request-based dm uses QUEUE_ORDERED_DRAIN_FLUSH ordered mode for I/O barrier. It means that when an I/O requiring barrier is found in the request_queue, the block-layer makes pre-flush request and post-flush request just before and just after the I/O respectively. After the ordered sequence starts, the block-layer waits for all in-flight I/Os to complete, then gives drivers the pre-flush request, the barrier I/O and the post-flush request one by one. It means that the request_queue is stopped automatically by the block-layer until drivers complete each sequence. o dm-core For the barrier I/O, treats it as a normal I/O, so no additional code is needed. For the pre/post-flush request, flushes caches by the followings: 1. Make the number of empty barrier requests required by target's num_flush_requests, and map them (dm_rq_barrier()). 2. Waits for the mapped barriers to complete (dm_rq_barrier()). If error has occurred, save the error value to md->barrier_error (dm_end_request()). (*) Basically, the first reported error is taken. But -EOPNOTSUPP supersedes any error and DM_ENDIO_REQUEUE follows. 3. Requeue the pre/post-flush request if the error value is DM_ENDIO_REQUEUE. Otherwise, completes with the error value (dm_rq_barrier_work()). The pre/post-flush work above is done in the kernel thread (kdmflush) context, since memory allocation which might sleep is needed in dm_rq_barrier() but sleep is not allowed in dm_request_fn(), which is an irq-disabled context. Also, clones of the pre/post-flush request share an original, so such clones can't be completed using the softirq context. Instead, complete them in the context of underlying device drivers. It should be safe since there is no I/O dispatching during the completion of such clones. For suspend, the workqueue of kdmflush needs to be flushed after the request_queue has been stopped. Otherwise, the next flush work can be kicked even after the suspend completes. TARGET INTERFACE No new interface is added. Just use the existing num_flush_requests in struct target_type as same as bio-based dm. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-12-10 23:52:18 +00:00
flush_workqueue(md->wq);
/*
* At this point no more requests are entering target request routines.
* We call dm_wait_for_completion to wait for all existing requests
* to finish.
*/
r = dm_wait_for_completion(md, task_state);
if (!r)
set_bit(dmf_suspended_flag, &md->flags);
if (noflush)
clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
if (map)
synchronize_srcu(&md->io_barrier);
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
/* were we interrupted ? */
if (r < 0) {
dm_queue_flush(md);
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
if (dm_request_based(md))
dm_start_queue(md->queue);
dm: prepare for request based option This patch adds core functions for request-based dm. When struct mapped device (md) is initialized, md->queue has an I/O scheduler and the following functions are used for request-based dm as the queue functions: make_request_fn: dm_make_request() pref_fn: dm_prep_fn() request_fn: dm_request_fn() softirq_done_fn: dm_softirq_done() lld_busy_fn: dm_lld_busy() Actual initializations are done in another patch (PATCH 2). Below is a brief summary of how request-based dm behaves, including: - making request from bio - cloning, mapping and dispatching request - completing request and bio - suspending md - resuming md bio to request ============== md->queue->make_request_fn() (dm_make_request()) calls __make_request() for a bio submitted to the md. Then, the bio is kept in the queue as a new request or merged into another request in the queue if possible. Cloning and Mapping =================== Cloning and mapping are done in md->queue->request_fn() (dm_request_fn()), when requests are dispatched after they are sorted by the I/O scheduler. dm_request_fn() checks busy state of underlying devices using target's busy() function and stops dispatching requests to keep them on the dm device's queue if busy. It helps better I/O merging, since no merge is done for a request once it is dispatched to underlying devices. Actual cloning and mapping are done in dm_prep_fn() and map_request() called from dm_request_fn(). dm_prep_fn() clones not only request but also bios of the request so that dm can hold bio completion in error cases and prevent the bio submitter from noticing the error. (See the "Completion" section below for details.) After the cloning, the clone is mapped by target's map_rq() function and inserted to underlying device's queue using blk_insert_cloned_request(). Completion ========== Request completion can be hooked by rq->end_io(), but then, all bios in the request will have been completed even error cases, and the bio submitter will have noticed the error. To prevent the bio completion in error cases, request-based dm clones both bio and request and hooks both bio->bi_end_io() and rq->end_io(): bio->bi_end_io(): end_clone_bio() rq->end_io(): end_clone_request() Summary of the request completion flow is below: blk_end_request() for a clone request => blk_update_request() => bio->bi_end_io() == end_clone_bio() for each clone bio => Free the clone bio => Success: Complete the original bio (blk_update_request()) Error: Don't complete the original bio => blk_finish_request() => rq->end_io() == end_clone_request() => blk_complete_request() => dm_softirq_done() => Free the clone request => Success: Complete the original request (blk_end_request()) Error: Requeue the original request end_clone_bio() completes the original request on the size of the original bio in successful cases. Even if all bios in the original request are completed by that completion, the original request must not be completed yet to keep the ordering of request completion for the stacking. So end_clone_bio() uses blk_update_request() instead of blk_end_request(). In error cases, end_clone_bio() doesn't complete the original bio. It just frees the cloned bio and gives over the error handling to end_clone_request(). end_clone_request(), which is called with queue lock held, completes the clone request and the original request in a softirq context (dm_softirq_done()), which has no queue lock, to avoid a deadlock issue on submission of another request during the completion: - The submitted request may be mapped to the same device - Request submission requires queue lock, but the queue lock has been held by itself and it doesn't know that The clone request has no clone bio when dm_softirq_done() is called. So target drivers can't resubmit it again even error cases. Instead, they can ask dm core for requeueing and remapping the original request in that cases. suspend ======= Request-based dm uses stopping md->queue as suspend of the md. For noflush suspend, just stops md->queue. For flush suspend, inserts a marker request to the tail of md->queue. And dispatches all requests in md->queue until the marker comes to the front of md->queue. Then, stops dispatching request and waits for the all dispatched requests to complete. After that, completes the marker request, stops md->queue and wake up the waiter on the suspend queue, md->wait. resume ====== Starts md->queue. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2009-06-22 09:12:35 +00:00
unlock_fs(md);
dm_table_presuspend_undo_targets(map);
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
/* pushback list is already flushed, so skip flush */
}
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
return r;
}
/*
* We need to be able to change a mapping table under a mounted
* filesystem. For example we might want to move some data in
* the background. Before the table can be swapped with
* dm_bind_table, dm_suspend must be called to flush any in
* flight bios and ensure that any further io gets deferred.
*/
/*
* Suspend mechanism in request-based dm.
*
* 1. Flush all I/Os by lock_fs() if needed.
* 2. Stop dispatching any I/O by stopping the request_queue.
* 3. Wait for all in-flight I/Os to be completed or requeued.
*
* To abort suspend, start the request_queue.
*/
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
{
struct dm_table *map = NULL;
int r = 0;
retry:
mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);
if (dm_suspended_md(md)) {
r = -EINVAL;
goto out_unlock;
}
if (dm_suspended_internally_md(md)) {
/* already internally suspended, wait for internal resume */
mutex_unlock(&md->suspend_lock);
r = wait_on_bit(&md->flags, DMF_SUSPENDED_INTERNALLY, TASK_INTERRUPTIBLE);
if (r)
return r;
goto retry;
}
map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
r = __dm_suspend(md, map, suspend_flags, TASK_INTERRUPTIBLE, DMF_SUSPENDED);
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
if (r)
goto out_unlock;
dm_table_postsuspend_targets(map);
out_unlock:
mutex_unlock(&md->suspend_lock);
return r;
}
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
static int __dm_resume(struct mapped_device *md, struct dm_table *map)
{
if (map) {
int r = dm_table_resume_targets(map);
if (r)
return r;
}
dm_queue_flush(md);
/*
* Flushing deferred I/Os must be done after targets are resumed
* so that mapping of targets can work correctly.
* Request-based dm is queueing the deferred I/Os in its request_queue.
*/
if (dm_request_based(md))
dm_start_queue(md->queue);
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
unlock_fs(md);
return 0;
}
int dm_resume(struct mapped_device *md)
{
int r;
struct dm_table *map = NULL;
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
retry:
r = -EINVAL;
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
mutex_lock_nested(&md->suspend_lock, SINGLE_DEPTH_NESTING);
if (!dm_suspended_md(md))
goto out;
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
if (dm_suspended_internally_md(md)) {
/* already internally suspended, wait for internal resume */
mutex_unlock(&md->suspend_lock);
r = wait_on_bit(&md->flags, DMF_SUSPENDED_INTERNALLY, TASK_INTERRUPTIBLE);
if (r)
return r;
goto retry;
}
map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
if (!map || !dm_table_get_size(map))
goto out;
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
r = __dm_resume(md, map);
if (r)
goto out;
clear_bit(DMF_SUSPENDED, &md->flags);
out:
mutex_unlock(&md->suspend_lock);
return r;
}
/*
* Internal suspend/resume works like userspace-driven suspend. It waits
* until all bios finish and prevents issuing new bios to the target drivers.
* It may be used only from the kernel.
*/
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
static void __dm_internal_suspend(struct mapped_device *md, unsigned suspend_flags)
{
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
struct dm_table *map = NULL;
lockdep_assert_held(&md->suspend_lock);
if (md->internal_suspend_count++)
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
return; /* nested internal suspend */
if (dm_suspended_md(md)) {
set_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
return; /* nest suspend */
}
map = rcu_dereference_protected(md->map, lockdep_is_held(&md->suspend_lock));
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
/*
* Using TASK_UNINTERRUPTIBLE because only NOFLUSH internal suspend is
* supported. Properly supporting a TASK_INTERRUPTIBLE internal suspend
* would require changing .presuspend to return an error -- avoid this
* until there is a need for more elaborate variants of internal suspend.
*/
(void) __dm_suspend(md, map, suspend_flags, TASK_UNINTERRUPTIBLE,
DMF_SUSPENDED_INTERNALLY);
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
dm_table_postsuspend_targets(map);
}
static void __dm_internal_resume(struct mapped_device *md)
{
BUG_ON(!md->internal_suspend_count);
if (--md->internal_suspend_count)
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
return; /* resume from nested internal suspend */
if (dm_suspended_md(md))
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
goto done; /* resume from nested suspend */
/*
* NOTE: existing callers don't need to call dm_table_resume_targets
* (which may fail -- so best to avoid it for now by passing NULL map)
*/
(void) __dm_resume(md, NULL);
done:
clear_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
smp_mb__after_atomic();
wake_up_bit(&md->flags, DMF_SUSPENDED_INTERNALLY);
}
void dm_internal_suspend_noflush(struct mapped_device *md)
{
mutex_lock(&md->suspend_lock);
__dm_internal_suspend(md, DM_SUSPEND_NOFLUSH_FLAG);
mutex_unlock(&md->suspend_lock);
}
EXPORT_SYMBOL_GPL(dm_internal_suspend_noflush);
void dm_internal_resume(struct mapped_device *md)
{
mutex_lock(&md->suspend_lock);
__dm_internal_resume(md);
mutex_unlock(&md->suspend_lock);
}
EXPORT_SYMBOL_GPL(dm_internal_resume);
/*
* Fast variants of internal suspend/resume hold md->suspend_lock,
* which prevents interaction with userspace-driven suspend.
*/
void dm_internal_suspend_fast(struct mapped_device *md)
{
mutex_lock(&md->suspend_lock);
if (dm_suspended_md(md) || dm_suspended_internally_md(md))
return;
set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
synchronize_srcu(&md->io_barrier);
flush_workqueue(md->wq);
dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
}
dm snapshot: suspend origin when doing exception handover In the function snapshot_resume we perform exception store handover. If there is another active snapshot target, the exception store is moved from this target to the target that is being resumed. The problem is that if there is some pending exception, it will point to an incorrect exception store after that handover, causing a crash due to dm-snap-persistent.c:get_exception()'s BUG_ON. This bug can be triggered by repeatedly changing snapshot permissions with "lvchange -p r" and "lvchange -p rw" while there are writes on the associated origin device. To fix this bug, we must suspend the origin device when doing the exception store handover to make sure that there are no pending exceptions: - introduce _origin_hash that keeps track of dm_origin structures. - introduce functions __lookup_dm_origin, __insert_dm_origin and __remove_dm_origin that manipulate the origin hash. - modify snapshot_resume so that it calls dm_internal_suspend_fast() and dm_internal_resume_fast() on the origin device. NOTE to stable@ people: When backporting to kernels 3.12-3.18, use dm_internal_suspend and dm_internal_resume instead of dm_internal_suspend_fast and dm_internal_resume_fast. When backporting to kernels older than 3.12, you need to pick functions dm_internal_suspend and dm_internal_resume from the commit fd2ed4d252701d3bbed4cd3e3d267ad469bb832a. Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Cc: stable@vger.kernel.org
2015-02-26 16:40:35 +00:00
EXPORT_SYMBOL_GPL(dm_internal_suspend_fast);
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
void dm_internal_resume_fast(struct mapped_device *md)
{
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
if (dm_suspended_md(md) || dm_suspended_internally_md(md))
goto done;
dm_queue_flush(md);
done:
mutex_unlock(&md->suspend_lock);
}
dm snapshot: suspend origin when doing exception handover In the function snapshot_resume we perform exception store handover. If there is another active snapshot target, the exception store is moved from this target to the target that is being resumed. The problem is that if there is some pending exception, it will point to an incorrect exception store after that handover, causing a crash due to dm-snap-persistent.c:get_exception()'s BUG_ON. This bug can be triggered by repeatedly changing snapshot permissions with "lvchange -p r" and "lvchange -p rw" while there are writes on the associated origin device. To fix this bug, we must suspend the origin device when doing the exception store handover to make sure that there are no pending exceptions: - introduce _origin_hash that keeps track of dm_origin structures. - introduce functions __lookup_dm_origin, __insert_dm_origin and __remove_dm_origin that manipulate the origin hash. - modify snapshot_resume so that it calls dm_internal_suspend_fast() and dm_internal_resume_fast() on the origin device. NOTE to stable@ people: When backporting to kernels 3.12-3.18, use dm_internal_suspend and dm_internal_resume instead of dm_internal_suspend_fast and dm_internal_resume_fast. When backporting to kernels older than 3.12, you need to pick functions dm_internal_suspend and dm_internal_resume from the commit fd2ed4d252701d3bbed4cd3e3d267ad469bb832a. Signed-off-by: Mikulas Patocka <mpatocka@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Cc: stable@vger.kernel.org
2015-02-26 16:40:35 +00:00
EXPORT_SYMBOL_GPL(dm_internal_resume_fast);
/*-----------------------------------------------------------------
* Event notification.
*---------------------------------------------------------------*/
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
unsigned cookie)
{
char udev_cookie[DM_COOKIE_LENGTH];
char *envp[] = { udev_cookie, NULL };
if (!cookie)
return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
else {
snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
DM_COOKIE_ENV_VAR_NAME, cookie);
return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
action, envp);
}
}
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
return atomic_add_return(1, &md->uevent_seq);
}
uint32_t dm_get_event_nr(struct mapped_device *md)
{
return atomic_read(&md->event_nr);
}
int dm_wait_event(struct mapped_device *md, int event_nr)
{
return wait_event_interruptible(md->eventq,
(event_nr != atomic_read(&md->event_nr)));
}
void dm_uevent_add(struct mapped_device *md, struct list_head *elist)
{
unsigned long flags;
spin_lock_irqsave(&md->uevent_lock, flags);
list_add(elist, &md->uevent_list);
spin_unlock_irqrestore(&md->uevent_lock, flags);
}
/*
* The gendisk is only valid as long as you have a reference
* count on 'md'.
*/
struct gendisk *dm_disk(struct mapped_device *md)
{
return md->disk;
}
EXPORT_SYMBOL_GPL(dm_disk);
struct kobject *dm_kobject(struct mapped_device *md)
{
return &md->kobj_holder.kobj;
}
struct mapped_device *dm_get_from_kobject(struct kobject *kobj)
{
struct mapped_device *md;
md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
dm: fix race between dm_get_from_kobject() and __dm_destroy() The following BUG_ON was hit when testing repeat creation and removal of DM devices: kernel BUG at drivers/md/dm.c:2919! CPU: 7 PID: 750 Comm: systemd-udevd Not tainted 4.1.44 Call Trace: [<ffffffff81649e8b>] dm_get_from_kobject+0x34/0x3a [<ffffffff81650ef1>] dm_attr_show+0x2b/0x5e [<ffffffff817b46d1>] ? mutex_lock+0x26/0x44 [<ffffffff811df7f5>] sysfs_kf_seq_show+0x83/0xcf [<ffffffff811de257>] kernfs_seq_show+0x23/0x25 [<ffffffff81199118>] seq_read+0x16f/0x325 [<ffffffff811de994>] kernfs_fop_read+0x3a/0x13f [<ffffffff8117b625>] __vfs_read+0x26/0x9d [<ffffffff8130eb59>] ? security_file_permission+0x3c/0x44 [<ffffffff8117bdb8>] ? rw_verify_area+0x83/0xd9 [<ffffffff8117be9d>] vfs_read+0x8f/0xcf [<ffffffff81193e34>] ? __fdget_pos+0x12/0x41 [<ffffffff8117c686>] SyS_read+0x4b/0x76 [<ffffffff817b606e>] system_call_fastpath+0x12/0x71 The bug can be easily triggered, if an extra delay (e.g. 10ms) is added between the test of DMF_FREEING & DMF_DELETING and dm_get() in dm_get_from_kobject(). To fix it, we need to ensure the test of DMF_FREEING & DMF_DELETING and dm_get() are done in an atomic way, so _minor_lock is used. The other callers of dm_get() have also been checked to be OK: some callers invoke dm_get() under _minor_lock, some callers invoke it under _hash_lock, and dm_start_request() invoke it after increasing md->open_count. Cc: stable@vger.kernel.org Signed-off-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-11-01 07:42:36 +00:00
spin_lock(&_minor_lock);
if (test_bit(DMF_FREEING, &md->flags) || dm_deleting_md(md)) {
md = NULL;
goto out;
}
dm_get(md);
dm: fix race between dm_get_from_kobject() and __dm_destroy() The following BUG_ON was hit when testing repeat creation and removal of DM devices: kernel BUG at drivers/md/dm.c:2919! CPU: 7 PID: 750 Comm: systemd-udevd Not tainted 4.1.44 Call Trace: [<ffffffff81649e8b>] dm_get_from_kobject+0x34/0x3a [<ffffffff81650ef1>] dm_attr_show+0x2b/0x5e [<ffffffff817b46d1>] ? mutex_lock+0x26/0x44 [<ffffffff811df7f5>] sysfs_kf_seq_show+0x83/0xcf [<ffffffff811de257>] kernfs_seq_show+0x23/0x25 [<ffffffff81199118>] seq_read+0x16f/0x325 [<ffffffff811de994>] kernfs_fop_read+0x3a/0x13f [<ffffffff8117b625>] __vfs_read+0x26/0x9d [<ffffffff8130eb59>] ? security_file_permission+0x3c/0x44 [<ffffffff8117bdb8>] ? rw_verify_area+0x83/0xd9 [<ffffffff8117be9d>] vfs_read+0x8f/0xcf [<ffffffff81193e34>] ? __fdget_pos+0x12/0x41 [<ffffffff8117c686>] SyS_read+0x4b/0x76 [<ffffffff817b606e>] system_call_fastpath+0x12/0x71 The bug can be easily triggered, if an extra delay (e.g. 10ms) is added between the test of DMF_FREEING & DMF_DELETING and dm_get() in dm_get_from_kobject(). To fix it, we need to ensure the test of DMF_FREEING & DMF_DELETING and dm_get() are done in an atomic way, so _minor_lock is used. The other callers of dm_get() have also been checked to be OK: some callers invoke dm_get() under _minor_lock, some callers invoke it under _hash_lock, and dm_start_request() invoke it after increasing md->open_count. Cc: stable@vger.kernel.org Signed-off-by: Hou Tao <houtao1@huawei.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com>
2017-11-01 07:42:36 +00:00
out:
spin_unlock(&_minor_lock);
return md;
}
int dm_suspended_md(struct mapped_device *md)
{
return test_bit(DMF_SUSPENDED, &md->flags);
}
dm: enhance internal suspend and resume interface Rename dm_internal_{suspend,resume} to dm_internal_{suspend,resume}_fast -- dm-stats will continue using these methods to avoid all the extra suspend/resume logic that is not needed in order to quickly flush IO. Introduce dm_internal_suspend_noflush() variant that actually calls the mapped_device's target callbacks -- otherwise target-specific hooks are avoided (e.g. dm-thin's thin_presuspend and thin_postsuspend). Common code between dm_internal_{suspend_noflush,resume} and dm_{suspend,resume} was factored out as __dm_{suspend,resume}. Update dm_internal_{suspend_noflush,resume} to always take and release the mapped_device's suspend_lock. Also update dm_{suspend,resume} to be aware of potential for DM_INTERNAL_SUSPEND_FLAG to be set and respond accordingly by interruptibly waiting for the DM_INTERNAL_SUSPEND_FLAG to be cleared. Add lockdep annotation to dm_suspend() and dm_resume(). The existing DM_SUSPEND_FLAG remains unchanged. DM_INTERNAL_SUSPEND_FLAG is set by dm_internal_suspend_noflush() and cleared by dm_internal_resume(). Both DM_SUSPEND_FLAG and DM_INTERNAL_SUSPEND_FLAG may be set if a device was already suspended when dm_internal_suspend_noflush() was called -- this can be thought of as a "nested suspend". A "nested suspend" can occur with legacy userspace dm-thin code that might suspend all active thin volumes before suspending the pool for resize. But otherwise, in the normal dm-thin-pool suspend case moving forward: the thin-pool will have DM_SUSPEND_FLAG set and all active thins from that thin-pool will have DM_INTERNAL_SUSPEND_FLAG set. Also add DM_INTERNAL_SUSPEND_FLAG to status report. This new DM_INTERNAL_SUSPEND_FLAG state is being reported to assist with debugging (e.g. 'dmsetup info' will report an internally suspended device accordingly). Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Joe Thornber <ejt@redhat.com>
2014-10-28 22:34:52 +00:00
int dm_suspended_internally_md(struct mapped_device *md)
{
return test_bit(DMF_SUSPENDED_INTERNALLY, &md->flags);
}
int dm_test_deferred_remove_flag(struct mapped_device *md)
{
return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
}
int dm_suspended(struct dm_target *ti)
{
dm table: remove dm_get from dm_table_get_md Remove the dm_get() in dm_table_get_md() because dm_table_get_md() could be called from presuspend/postsuspend, which are called while mapped_device is in DMF_FREEING state, where dm_get() is not allowed. Justification for that is the lifetime of both objects: As far as the current dm design/implementation, mapped_device is never freed while targets are doing something, because dm core waits for targets to become quiet in dm_put() using presuspend/postsuspend. So targets should be able to touch mapped_device without holding reference count of the mapped_device, and we should allow targets to touch mapped_device even if it is in DMF_FREEING state. Backgrounds: I'm trying to remove the multipath internal queue, since dm core now has a generic queue for request-based dm. In the patch-set, the multipath target wants to request dm core to start/stop queue. One of such start/stop requests can happen during postsuspend() while the target waits for pg-init to complete, because the target stops queue when starting pg-init and tries to restart it when completing pg-init. Since queue belongs to mapped_device, it involves calling dm_table_get_md() and dm_put(). On the other hand, postsuspend() is called in dm_put() for mapped_device which is in DMF_FREEING state, and that triggers BUG_ON(DMF_FREEING) in the 2nd dm_put(). I had tried to solve this problem by changing only multipath not to touch mapped_device which is in DMF_FREEING state, but I couldn't and I came up with a question why we need dm_get() in dm_table_get_md(). Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-03-06 02:29:52 +00:00
return dm_suspended_md(dm_table_get_md(ti->table));
}
EXPORT_SYMBOL_GPL(dm_suspended);
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
int dm_noflush_suspending(struct dm_target *ti)
{
dm table: remove dm_get from dm_table_get_md Remove the dm_get() in dm_table_get_md() because dm_table_get_md() could be called from presuspend/postsuspend, which are called while mapped_device is in DMF_FREEING state, where dm_get() is not allowed. Justification for that is the lifetime of both objects: As far as the current dm design/implementation, mapped_device is never freed while targets are doing something, because dm core waits for targets to become quiet in dm_put() using presuspend/postsuspend. So targets should be able to touch mapped_device without holding reference count of the mapped_device, and we should allow targets to touch mapped_device even if it is in DMF_FREEING state. Backgrounds: I'm trying to remove the multipath internal queue, since dm core now has a generic queue for request-based dm. In the patch-set, the multipath target wants to request dm core to start/stop queue. One of such start/stop requests can happen during postsuspend() while the target waits for pg-init to complete, because the target stops queue when starting pg-init and tries to restart it when completing pg-init. Since queue belongs to mapped_device, it involves calling dm_table_get_md() and dm_put(). On the other hand, postsuspend() is called in dm_put() for mapped_device which is in DMF_FREEING state, and that triggers BUG_ON(DMF_FREEING) in the 2nd dm_put(). I had tried to solve this problem by changing only multipath not to touch mapped_device which is in DMF_FREEING state, but I couldn't and I came up with a question why we need dm_get() in dm_table_get_md(). Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com>
2010-03-06 02:29:52 +00:00
return __noflush_suspending(dm_table_get_md(ti->table));
[PATCH] dm: suspend: add noflush pushback In device-mapper I/O is sometimes queued within targets for later processing. For example the multipath target can be configured to store I/O when no paths are available instead of returning it -EIO. This patch allows the device-mapper core to instruct a target to transfer the contents of any such in-target queue back into the core. This frees up the resources used by the target so the core can replace that target with an alternative one and then resend the I/O to it. Without this patch the only way to change the target in such circumstances involves returning the I/O with an error back to the filesystem/application. In the multipath case, this patch will let us add new paths for existing I/O to try after all the existing paths have failed. DMF_NOFLUSH_SUSPENDING ---------------------- If the DM_NOFLUSH_FLAG ioctl option is specified at suspend time, the DMF_NOFLUSH_SUSPENDING flag is set in md->flags during dm_suspend(). It is always cleared before dm_suspend() returns. The flag must be visible while the target is flushing pending I/Os so it is set before presuspend where the flush starts and unset after the wait for md->pending where the flush ends. Target drivers can check this flag by calling dm_noflush_suspending(). DM_MAPIO_REQUEUE / DM_ENDIO_REQUEUE ----------------------------------- A target's map() function can now return DM_MAPIO_REQUEUE to request the device mapper core queue the bio. Similarly, a target's end_io() function can return DM_ENDIO_REQUEUE to request the same. This has been labelled 'pushback'. The __map_bio() and clone_endio() functions in the core treat these return values as errors and call dec_pending() to end the I/O. dec_pending ----------- dec_pending() saves the pushback request in struct dm_io->error. Once all the split clones have ended, dec_pending() will put the original bio on the md->pushback list. Note that this supercedes any I/O errors. It is possible for the suspend with DM_NOFLUSH_FLAG to be aborted while in progress (e.g. by user interrupt). dec_pending() checks for this and returns -EIO if it happened. pushdback list and pushback_lock -------------------------------- The bio is queued on md->pushback temporarily in dec_pending(), and after all pending I/Os return, md->pushback is merged into md->deferred in dm_suspend() for re-issuing at resume time. md->pushback_lock protects md->pushback. The lock should be held with irq disabled because dec_pending() can be called from interrupt context. Queueing bios to md->pushback in dec_pending() must be done atomically with the check for DMF_NOFLUSH_SUSPENDING flag. So md->pushback_lock is held when checking the flag. Otherwise dec_pending() may queue a bio to md->pushback after the interrupted dm_suspend() flushes md->pushback. Then the bio would be left in md->pushback. Flag setting in dm_suspend() can be done without md->pushback_lock because the flag is checked only after presuspend and the set value is already made visible via the target's presuspend function. The flag can be checked without md->pushback_lock (e.g. the first part of the dec_pending() or target drivers), because the flag is checked again with md->pushback_lock held when the bio is really queued to md->pushback as described above. So even if the flag is cleared after the lockless checkings, the bio isn't left in md->pushback but returned to applications with -EIO. Other notes on the current patch -------------------------------- - md->pushback is added to the struct mapped_device instead of using md->deferred directly because md->io_lock which protects md->deferred is rw_semaphore and can't be used in interrupt context like dec_pending(), and md->io_lock protects the DMF_BLOCK_IO flag of md->flags too. - Don't issue lock_fs() in dm_suspend() if the DM_NOFLUSH_FLAG ioctl option is specified, because I/Os generated by lock_fs() would be pushed back and never return if there were no valid devices. - If an error occurs in dm_suspend() after the DMF_NOFLUSH_SUSPENDING flag is set, md->pushback must be flushed because I/Os may be queued to the list already. (flush_and_out label in dm_suspend()) Test results ------------ I have tested using multipath target with the next patch. The following tests are for regression/compatibility: - I/Os succeed when valid paths exist; - I/Os fail when there are no valid paths and queue_if_no_path is not set; - I/Os are queued in the multipath target when there are no valid paths and queue_if_no_path is set; - The queued I/Os above fail when suspend is issued without the DM_NOFLUSH_FLAG ioctl option. I/Os spanning 2 multipath targets also fail. The following tests are for the normal code path of new pushback feature: - Queued I/Os in the multipath target are flushed from the target but don't return when suspend is issued with the DM_NOFLUSH_FLAG ioctl option; - The I/Os above are queued in the multipath target again when resume is issued without path recovery; - The I/Os above succeed when resume is issued after path recovery or table load; - Queued I/Os in the multipath target succeed when resume is issued with the DM_NOFLUSH_FLAG ioctl option after table load. I/Os spanning 2 multipath targets also succeed. The following tests are for the error paths of the new pushback feature: - When the bdget_disk() fails in dm_suspend(), the DMF_NOFLUSH_SUSPENDING flag is cleared and I/Os already queued to the pushback list are flushed properly. - When suspend with the DM_NOFLUSH_FLAG ioctl option is interrupted, o I/Os which had already been queued to the pushback list at the time don't return, and are re-issued at resume time; o I/Os which hadn't been returned at the time return with EIO. Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Cc: dm-devel@redhat.com Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-08 10:41:09 +00:00
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);
struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, enum dm_queue_mode type,
unsigned integrity, unsigned per_io_data_size,
unsigned min_pool_size)
{
struct dm_md_mempools *pools = kzalloc_node(sizeof(*pools), GFP_KERNEL, md->numa_node_id);
unsigned int pool_size = 0;
unsigned int front_pad, io_front_pad;
int ret;
if (!pools)
return NULL;
switch (type) {
case DM_TYPE_BIO_BASED:
case DM_TYPE_DAX_BIO_BASED:
case DM_TYPE_NVME_BIO_BASED:
pool_size = max(dm_get_reserved_bio_based_ios(), min_pool_size);
front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
io_front_pad = roundup(front_pad, __alignof__(struct dm_io)) + offsetof(struct dm_io, tio);
ret = bioset_init(&pools->io_bs, pool_size, io_front_pad, 0);
if (ret)
goto out;
if (integrity && bioset_integrity_create(&pools->io_bs, pool_size))
goto out;
break;
case DM_TYPE_REQUEST_BASED:
pool_size = max(dm_get_reserved_rq_based_ios(), min_pool_size);
front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
/* per_io_data_size is used for blk-mq pdu at queue allocation */
break;
default:
BUG();
}
ret = bioset_init(&pools->bs, pool_size, front_pad, 0);
if (ret)
goto out;
if (integrity && bioset_integrity_create(&pools->bs, pool_size))
goto out;
return pools;
out:
dm_free_md_mempools(pools);
return NULL;
}
void dm_free_md_mempools(struct dm_md_mempools *pools)
{
if (!pools)
return;
bioset_exit(&pools->bs);
bioset_exit(&pools->io_bs);
kfree(pools);
}
struct dm_pr {
u64 old_key;
u64 new_key;
u32 flags;
bool fail_early;
};
static int dm_call_pr(struct block_device *bdev, iterate_devices_callout_fn fn,
void *data)
{
struct mapped_device *md = bdev->bd_disk->private_data;
struct dm_table *table;
struct dm_target *ti;
int ret = -ENOTTY, srcu_idx;
table = dm_get_live_table(md, &srcu_idx);
if (!table || !dm_table_get_size(table))
goto out;
/* We only support devices that have a single target */
if (dm_table_get_num_targets(table) != 1)
goto out;
ti = dm_table_get_target(table, 0);
ret = -EINVAL;
if (!ti->type->iterate_devices)
goto out;
ret = ti->type->iterate_devices(ti, fn, data);
out:
dm_put_live_table(md, srcu_idx);
return ret;
}
/*
* For register / unregister we need to manually call out to every path.
*/
static int __dm_pr_register(struct dm_target *ti, struct dm_dev *dev,
sector_t start, sector_t len, void *data)
{
struct dm_pr *pr = data;
const struct pr_ops *ops = dev->bdev->bd_disk->fops->pr_ops;
if (!ops || !ops->pr_register)
return -EOPNOTSUPP;
return ops->pr_register(dev->bdev, pr->old_key, pr->new_key, pr->flags);
}
static int dm_pr_register(struct block_device *bdev, u64 old_key, u64 new_key,
u32 flags)
{
struct dm_pr pr = {
.old_key = old_key,
.new_key = new_key,
.flags = flags,
.fail_early = true,
};
int ret;
ret = dm_call_pr(bdev, __dm_pr_register, &pr);
if (ret && new_key) {
/* unregister all paths if we failed to register any path */
pr.old_key = new_key;
pr.new_key = 0;
pr.flags = 0;
pr.fail_early = false;
dm_call_pr(bdev, __dm_pr_register, &pr);
}
return ret;
}
static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type,
u32 flags)
{
struct mapped_device *md = bdev->bd_disk->private_data;
const struct pr_ops *ops;
int r, srcu_idx;
r = dm_prepare_ioctl(md, &srcu_idx, &bdev);
if (r < 0)
goto out;
ops = bdev->bd_disk->fops->pr_ops;
if (ops && ops->pr_reserve)
r = ops->pr_reserve(bdev, key, type, flags);
else
r = -EOPNOTSUPP;
out:
dm_unprepare_ioctl(md, srcu_idx);
return r;
}
static int dm_pr_release(struct block_device *bdev, u64 key, enum pr_type type)
{
struct mapped_device *md = bdev->bd_disk->private_data;
const struct pr_ops *ops;
int r, srcu_idx;
r = dm_prepare_ioctl(md, &srcu_idx, &bdev);
if (r < 0)
goto out;
ops = bdev->bd_disk->fops->pr_ops;
if (ops && ops->pr_release)
r = ops->pr_release(bdev, key, type);
else
r = -EOPNOTSUPP;
out:
dm_unprepare_ioctl(md, srcu_idx);
return r;
}
static int dm_pr_preempt(struct block_device *bdev, u64 old_key, u64 new_key,
enum pr_type type, bool abort)
{
struct mapped_device *md = bdev->bd_disk->private_data;
const struct pr_ops *ops;
int r, srcu_idx;
r = dm_prepare_ioctl(md, &srcu_idx, &bdev);
if (r < 0)
goto out;
ops = bdev->bd_disk->fops->pr_ops;
if (ops && ops->pr_preempt)
r = ops->pr_preempt(bdev, old_key, new_key, type, abort);
else
r = -EOPNOTSUPP;
out:
dm_unprepare_ioctl(md, srcu_idx);
return r;
}
static int dm_pr_clear(struct block_device *bdev, u64 key)
{
struct mapped_device *md = bdev->bd_disk->private_data;
const struct pr_ops *ops;
int r, srcu_idx;
r = dm_prepare_ioctl(md, &srcu_idx, &bdev);
if (r < 0)
goto out;
ops = bdev->bd_disk->fops->pr_ops;
if (ops && ops->pr_clear)
r = ops->pr_clear(bdev, key);
else
r = -EOPNOTSUPP;
out:
dm_unprepare_ioctl(md, srcu_idx);
return r;
}
static const struct pr_ops dm_pr_ops = {
.pr_register = dm_pr_register,
.pr_reserve = dm_pr_reserve,
.pr_release = dm_pr_release,
.pr_preempt = dm_pr_preempt,
.pr_clear = dm_pr_clear,
};
static const struct block_device_operations dm_blk_dops = {
.open = dm_blk_open,
.release = dm_blk_close,
.ioctl = dm_blk_ioctl,
.getgeo = dm_blk_getgeo,
.report_zones = dm_blk_report_zones,
.pr_ops = &dm_pr_ops,
.owner = THIS_MODULE
};
static const struct dax_operations dm_dax_ops = {
.direct_access = dm_dax_direct_access,
.copy_from_iter = dm_dax_copy_from_iter,
.copy_to_iter = dm_dax_copy_to_iter,
};
/*
* module hooks
*/
module_init(dm_init);
module_exit(dm_exit);
module_param(major, uint, 0);
MODULE_PARM_DESC(major, "The major number of the device mapper");
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");
module_param(dm_numa_node, int, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(dm_numa_node, "NUMA node for DM device memory allocations");
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");