There are two variants of stat functions - ones prefixed with double
underbars which don't care about preemption and ones without which
disable preemption before manipulating per-cpu counters. It's unclear
whether the underbarred ones assume that preemtion is disabled on
entry as some callers don't do that.
This patch unifies diskstats access by implementing disk_stat_lock()
and disk_stat_unlock() which take care of both RCU (for partition
access) and preemption (for per-cpu counter access). diskstats access
should always be enclosed between the two functions. As such, there's
no need for the versions which disables preemption. They're removed
and double underbars ones are renamed to drop the underbars. As an
extra argument is added, there's no danger of using the old version
unconverted.
disk_stat_lock() uses get_cpu() and returns the cpu index and all
diskstat functions which access per-cpu counters now has @cpu
argument to help RT.
This change adds RCU or preemption operations at some places but also
collapses several preemption ops into one at others. Overall, the
performance difference should be negligible as all involved ops are
very lightweight per-cpu ones.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
disk->part[] is protected by its matching bdev's lock. However,
non-critical accesses like collecting stats and printing out sysfs and
proc information used to be performed without any locking. As
partitions can come and go dynamically, partitions can go away
underneath those non-critical accesses. As some of those accesses are
writes, this theoretically can lead to silent corruption.
This patch fixes the race by using RCU for the partition array and dev
reference counter to hold partitions.
* Rename disk->part[] to disk->__part[] to make sure no one outside
genhd layer proper accesses it directly.
* Use RCU for disk->__part[] dereferencing.
* Implement disk_{get|put}_part() which can be used to get and put
partitions from gendisk respectively.
* Iterators are implemented to help iterate through all partitions
safely.
* Functions which require RCU readlock are marked with _rcu suffix.
* Use disk_put_part() in __blkdev_put() instead of directly putting
the contained kobject.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
* Implement disk_devt() and part_devt() and use them to directly
access devt instead of computing it from ->major and ->first_minor.
Note that all references to ->major and ->first_minor outside of
block layer is used to determine devt of the disk (the part0) and as
->major and ->first_minor will continue to represent devt for the
disk, converting these users aren't strictly necessary. However,
convert them for consistency.
* Implement disk_max_parts() to avoid directly deferencing
genhd->minors.
* Update bdget_disk() such that it doesn't assume consecutive minor
space.
* Move devt computation from register_disk() to add_disk() and make it
the only one (all other usages use the initially determined value).
These changes clean up the code and will help disk->part dereference
fix and extended block device numbers.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
In hd_struct, @partno is used to denote partition number and a number
of other places use @part to denote hd_struct. Functions use @part
and @index instead. This causes confusion and makes it difficult to
use consistent variable names for hd_struct. Always use @partno if a
variable represents partition number.
Also, print out functions use @f or @part for seq_file argument. Use
@seqf uniformly instead.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
This patch makes the following misc updates in preparation for
disk->part dereference fix and extended block devt support.
* implment part_to_disk()
* fix comment about gendisk->part indexing
* rename get_part() to disk_map_sector()
* don't use n which is always zero while printing disk information in
diskstats_show()
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
d805dda4 tried to fix error case handling in add_partition() but had a
few problems.
* disk->part[] entry is set early and left dangling if operation
fails.
* Once device initialized, the last put_device() is responsible for
freeing all the resources. The failure path freed part_stats and p
regardless of put_device() causing double free.
* holders subdir holds reference to the disk device, so failure path
should remove it to release resources properly which was missing.
This patch fixes the above problems and while at it move partition
slot busy check into add_partition() for completeness and inlines
holders subdirectory creation. Using separate function for it just
obfuscates the code.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Abdel Benamrouche <draconux@gmail.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
delete_partition() was noop for zero length partition. As the
addition code allows creating zero lenght partition and deletion is
assumed to always succeed, this causes memory leak for zero length
partitions. Allow zero length partitions to end their meaningless
lives.
While at it, allow deleting zero lenght partition via
BLKPG_DEL_PARTITION ioctl too.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Recent block_class iteration updates 5c6f35c5..27f3025 converted all
class device iteration to class_for_each_device() and
class_find_device(), which are correct but pain in the ass to use.
This pach converts them to newly introduced class_dev_iterator so that
they can use more natural control structures instead of separate
callbacks and struct to pass parameters to them.
This results in smaller and easier code.
This patch also restores the original behavior of not printing header
in /proc/partitions if there's no partition to print. This is trivial
but still user-visible behavior.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
block_class_lock protects major_names array and bdev_map and doesn't
have anything to do with block class devices. Don't grab them while
iterating over block class devices.
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Recent block_class iteration updates 5c6f35c5..27f3025 broke partition
info printouts.
* printk_all_partitions(): Partition print out stops when it meets a
partition hole. Partition printing inner loop should continue
instead of exiting on empty partition slot.
* /proc/partitions and /proc/diskstats: If all information can't be
read in single read(), the information is truncated. This is
because find_start() doesn't actually update the counter containing
the initial seek. It runs to the end and ends up always reporting
EOF on the second read.
This patch fixes both problems.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Iterating over entries using callback usually isn't too fun especially
when the entry being iterated over can't be manipulated freely. This
patch converts class->p->class_devices to klist and implements class
device iterator so that the users can freely build their own control
structure. The users are also free to call back into class code
without worrying about locking.
class_for_each_device() and class_find_device() are converted to use
the new iterators, so their users don't have to worry about locking
anymore either.
Note: This depends on klist-dont-iterate-over-deleted-entries patch
because class_intf->add/remove_dev() depends on proper synchronization
with device removal.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
A klist entry is kept on the list till all its current iterations are
finished; however, a new iteration after deletion also iterates over
deleted entries as long as their reference count stays above zero.
This causes problems for cases where there are users which iterate
over the list while synchronized against list manipulations and
natuarally expect already deleted entries to not show up during
iteration.
This patch implements dead flag which gets set on deletion so that
iteration can skip already deleted entries. The dead flag piggy backs
on the lowest bit of knode->n_klist and only visible to klist
implementation proper.
While at it, drop klist_iter->i_head as it's redundant and doesn't
offer anything in semantics or performance wise as klist_iter->i_klist
is dereferenced on every iteration anyway.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
raid5 can overflow with more than 255 stripes, and we can increase it
to an int for free on both 32 and 64-bit archs due to the padding.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Remove hw_segments field from struct bio and struct request. Without virtual
merge accounting they have no purpose.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Update the description of fifo_batch to match the current implementation,
and include a description of how to tune it.
Signed-off-by: Aaron Carroll <aaronc@gelato.unsw.edu.au>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Deadline currently only batches sector-contiguous requests, so except
for a few circumstances (e.g. requests in a single direction), it is
essentially first come first served. This is bad for throughput, so
change it to CSCAN, which means requests in a batch do not need to be
sequential and are issued in increasing sector order.
Signed-off-by: Aaron Carroll <aaronc@gelato.unsw.edu.au>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
struct request has an ioprio member but it is never updated because
currently bios do not hold io context information. The implication of
this is that virtio_blk ends up passing useless information to the
backend driver.
That said, some IO schedulers such as CFQ do store io context
information in struct request, but use private members for that, which
means that that information cannot be directly accessed in a IO
scheduler-independent way.
This patch adds a function to obtain the ioprio of a request. We should
avoid accessing ioprio directly and use this function instead, so that
its users do not have to care about future changes in block layer
structures or what the currently active IO controller is.
This patch does not introduce any functional changes but paves the way
for future clean-ups and enhancements.
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
It was only used by ps3disk, and it should probably have been
REQ_TYPE_LINUX_BLOCK + REQ_LB_OP_FLUSH.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
But blkdev_issue_discard() still emits requests which are interpreted as
soft barriers, because naïve callers might otherwise issue subsequent
writes to those same sectors, which might cross on the queue (if they're
reallocated quickly enough).
Callers still _can_ issue non-barrier discard requests, but they have to
take care of queue ordering for themselves.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
We may well want mkfs tools to use this to mark the whole device as
unwanted before they format it, for example.
The ioctl takes a pair of uint64_ts, which are start offset and length
in _bytes_. Although at the moment it might make sense for them both to
be in 512-byte sectors, I don't want to limit the ABI to that.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Barriers should be submitted with the WRITE flag set.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Let the compiler see what's going on, and it can all get a lot simpler.
On PPC64 this reduces the size of the code calculating these bits by
about 60%. On x86_64 it's less of a win -- only 40%.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
We can benefit from knowing that the file system no longer cares about
the contents of certain sectors, by throwing them away immediately and
then never having to garbage collect them, and using the extra free
space to make our operations more efficient. Do so.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Some block devices benefit from a hint that they can forget the contents
of certain sectors. Add basic support for this to the block core, along
with a 'blkdev_issue_discard()' helper function which issues such
requests.
The caller doesn't get to provide an end_io functio, since
blkdev_issue_discard() will automatically split the request up into
multiple bios if appropriate. Neither does the function wait for
completion -- it's expected that callers won't care about when, or even
_if_, the request completes. It's only a hint to the device anyway. By
definition, the file system doesn't _care_ about these sectors any more.
[With feedback from OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> and
Jens Axboe <jens.axboe@oracle.com]
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
I have another request for the block filter SG_IO command whitelist,
specifically the MMC streaming command set SET READ AHEAD command.
The command applies only to MMC CDROM/DVDROM drives with the streaming
optional feature set. The command is useful to cdparanoia in that it
allows explicit cache control side effects that are, on many drives,
cdparanoia's most efficient way to flush/disable the media cache on
cdrom drives. I am aware of no reason why it should not be accessible
from usespace.
Also note that the command is already fully accessible through the
SCSI-native version of the SG_IO ioctl as well as the traditional SG
interface. The command is only being refused on block devices. That
means that on a typical stock distro, the command is available through
/dev/sg* but not /dev/scd* although both are typically available and
accessible. Filtering the command is not providing any protection,
only a confusing inconsistency.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Set RLKEY bit in the HW context for kernel QPs so that kernel QPs can
use the reserved L_Key for memory reference.
Signed-off-by: Vladimir Sokolovsky <vlad@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6:
tcp: Fix tcp_hybla zero congestion window growth with small rho and large cwnd.
net: Fix netdev_run_todo dead-lock
tcp: Fix possible double-ack w/ user dma
net: only invoke dev->change_rx_flags when device is UP
netrom: Fix sock_orphan() use in nr_release
ax25: Quick fix for making sure unaccepted sockets get destroyed.
Revert "ax25: Fix std timer socket destroy handling."
[Bluetooth] Add reset quirk for A-Link BlueUSB21 dongle
[Bluetooth] Add reset quirk for new Targus and Belkin dongles
[Bluetooth] Fix double frees on error paths of btusb and bpa10x drivers
Symbol name spaghetti which is too complicated to cleanup on this stage
of the release cycle breaks the build on BCM1480 platforms.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Because of rounding, in certain conditions, i.e. when in congestion
avoidance state rho is smaller than 1/128 of the current cwnd, TCP
Hybla congestion control starves and the cwnd is kept constant
forever.
This patch forces an increment by one segment after #send_cwnd calls
without increments(newreno behavior).
Signed-off-by: Daniele Lacamera <root@danielinux.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Benjamin Thery tracked down a bug that explains many instances
of the error
unregister_netdevice: waiting for %s to become free. Usage count = %d
It turns out that netdev_run_todo can dead-lock with itself if
a second instance of it is run in a thread that will then free
a reference to the device waited on by the first instance.
The problem is really quite silly. We were trying to create
parallelism where none was required. As netdev_run_todo always
follows a RTNL section, and that todo tasks can only be added
with the RTNL held, by definition you should only need to wait
for the very ones that you've added and be done with it.
There is no need for a second mutex or spinlock.
This is exactly what the following patch does.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Ali Saidi <saidi@engin.umich.edu>
When TCP receive copy offload is enabled it's possible that
tcp_rcv_established() will cause two acks to be sent for a single
packet. In the case that a tcp_dma_early_copy() is successful,
copied_early is set to true which causes tcp_cleanup_rbuf() to be
called early which can send an ack. Further along in
tcp_rcv_established(), __tcp_ack_snd_check() is called and will
schedule a delayed ACK. If no packets are processed before the delayed
ack timer expires the packet will be acked twice.
Signed-off-by: David S. Miller <davem@davemloft.net>
Jesper Dangaard Brouer <hawk@comx.dk> reported a bug when setting a VLAN
device down that is in promiscous mode:
When the VLAN device is set down, the promiscous count on the real
device is decremented by one by vlan_dev_stop(). When removing the
promiscous flag from the VLAN device afterwards, the promiscous
count on the real device is decremented a second time by the
vlan_change_rx_flags() callback.
The root cause for this is that the ->change_rx_flags() callback is
invoked while the device is down. The synchronization is meant to mirror
the behaviour of the ->set_rx_mode callbacks, meaning the ->open function
is responsible for doing a full sync on open, the ->close() function is
responsible for doing full cleanup on ->stop() and ->change_rx_flags()
is meant to do incremental changes while the device is UP.
Only invoke ->change_rx_flags() while the device is UP to provide the
intended behaviour.
Tested-by: Jesper Dangaard Brouer <jdb@comx.dk>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
SLOB's ksize calculation was braindamaged and generally harmlessly
underreported the allocation size. But for very small buffers, it could
in fact overreport them, leading code depending on krealloc to overrun
the allocation and trample other data.
Signed-off-by: Matt Mackall <mpm@selenic.com>
Tested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This reverts commit 135aedc38e, as
requested by Hans Verkuil.
It was a patch for 2.6.28 where the BKL was pushed down from v4l core to
the drivers, not for 2.6.27!
Requested-by: Hans Verkuil <hverkuil@xs4all.nl>
Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-of-by: Linus Torvalds <torvalds@linux-foundation.org>
