Fix problem that deleting multiple logical drives could cause a panic.
It fixes a panic which can be easily reproduced in the following way: Just
create several "arrays," each with multiple logical drives via hpacucli,
then delete the first array, and it will blow up in deregister_disk(), in
the call to get_host() when it tries to dig the hba pointer out of a NULL
queue pointer.
The problem has been present since my code to make rebuild_lun_table
behave better went in.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cca.cpqcorp.net>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
USB is going to switch the signature of the callbacks to
void callback(struct urb *urb, int status)
This patch will ease the transition.
Signed-off-by: Oliver Neukum <oneukum@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
We could use DEFINE_IDR for statically allocated idr
that allow us to save a few lines of code.
And spell fix.
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The accelerated iSCSI traffic could use a private IP address unknown to the OS:
- The IP address is required in both drivers to manage ARP requests and connection set up.
- Added an control call to retrieve the ip address.
- Reply to ARP requests dedicated to the private IP address.
Signed-off-by: Divy Le Ray <divy@chelsio.com>
Signed-off-by: Karen Xie <kxie@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The code appears to be dead: nobody call these functions, plus build
breaks when UGETH_FILTERING is enabled:
ucc_geth.c:1848: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c:1848: warning: its scope is only this definition or declaration, which is probably not what you want
ucc_geth.c: In function 'ugeth_82xx_filtering_get_match_addr_in_hash':
ucc_geth.c:1856: error: dereferencing pointer to incomplete type
ucc_geth.c:1874: error: dereferencing pointer to incomplete type
ucc_geth.c:1877: warning: return from incompatible pointer type
ucc_geth.c: At top level:
ucc_geth.c:1885: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c: In function 'ugeth_82xx_filtering_add_addr_in_hash':
ucc_geth.c:1894: error: dereferencing pointer to incomplete type
ucc_geth.c:1909: warning: passing argument 2 of 'ugeth_82xx_filtering_get_match_addr_in_hash' from incompatible pointer type
ucc_geth.c:1909: warning: assignment from incompatible pointer type
ucc_geth.c:1918: error: dereferencing pointer to incomplete type
ucc_geth.c: At top level:
ucc_geth.c:1928: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c: In function 'ugeth_82xx_filtering_clear_addr_in_hash':
ucc_geth.c:1947: warning: passing argument 2 of 'ugeth_82xx_filtering_get_match_addr_in_hash' from incompatible pointer type
ucc_geth.c:1947: warning: assignment from incompatible pointer type
ucc_geth.c:1954: error: dereferencing pointer to incomplete type
ucc_geth.c: At top level:
ucc_geth.c:2060: warning: 'struct enet_addr' declared inside parameter list
ucc_geth.c: In function 'ugeth_82xx_filtering_add_addr_in_paddr':
ucc_geth.c:2064: error: dereferencing pointer to incomplete type
ucc_geth.c:2073: error: dereferencing pointer to incomplete type
ucc_geth.c:2075: warning: passing argument 2 of 'hw_add_addr_in_paddr' from incompatible pointer type
make[2]: *** [ucc_geth.o] Error 1
The code is there since the driver was merged, and nobody seem to be
interested in fixing or actually using it. If we ever want the
filtering support, we can always revert the patch and fix it, but so
far it just draws reader's attention.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The driver doesn't check ioremap() return value, and doesn't free the
remapped memory.
This patch fixes it.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
No need to call ucc_geth_memclean() so many times, just check for
errors in ucc_geth_open(), and call ucc_geth_stop() in case of errors.
The ucc_geth_stop() may be called anytime and will do the right thing.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
open() routine calls stop() in case of errors, the function will try
to free the requested IRQ. But we don't know if it was actually
requested, so the code might issue bogus free_irq(0, dev) call.
Fix this by rearranging the code so that now request_irq() is the last
call in the open() routine, and move free_irq() into the close().
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The timeout handling code is currently broken in several ways:
- It calls stop() (which frees all the memory and IRQ), and then
calls startup() (which won't re-request IRQ, neither it will
re-init the Fast UCC structure).
- It calls these routines from the softirq context, which is wrong,
since stop() calls free_irq() (which might sleep) and startup()
allocates things with GFP_KERNEL.
- It won't soft-reset the PHY. We need the PHY reset for at least
MPC8360E-MDS boards with Marvell 88E1111 PHY, the PHY won't recover
from timeouts w/o the reset.
So the patch fixes these problems by implementing the workqueue for the
timeout handling, and there we fully re-open the device via close() and
open() calls. The close/open paths do the right things, and I can see
that the driver actually survive the timeouts.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently the routines wait for the various bits w/o an assumption that
bits may never get set. When timeouts happen I see that these bits never
get set and so the routines hang the kernel.
With this patch we'll wait the graceful stop for 100 ms, and then will
simply exit. There is nothing* we can do about that, but it's OK since
we'll do full reset later.
* Well, actually, there is also not-graceful variant for the TX stop,
but specs says that we never should use it.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original message was unhelpful and extremely alarming to our poor
users, despite its charm. Make it less frightening.
Signed-off-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Optimize the lightly loaded case, by only synchronizing discards stats
when qlen > 10 indicate potential for drops.
Notice Robert Olsson might disagree with this patch.
Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Discard packet counter debug statements that can be turned on
at runtime by users to assist debugging of the driver code.
Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implementing discard counters for the NIU driver turned out to be more
complicated than first assumed.
The discard counters for the NIU neptune chip are only 16-bit (even
though this is a 64-bit chip). These 16-bit counters can overflow
quickly, especially considering this is a 10Gbit/s ethernet card.
The overflow indication bit is, unfortunatly, not usable as the
counter value does not wrap, but remains at max value 0xFFFF.
Resulting in lost counts until the counter is reset.
The read and reset scheme also poses a problem. Both in theory and in
practice counters can be lost in between reading nr64() and clearing
the counter nw64(). For this reason, the number of counter clearings
nw64() is limited/reduced. On the fast-path the counters are only
syncronized once it exceeds 0x7FFF. When read by userspace, its
syncronized fully.
Signed-off-by: Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
A major source of overhead in the enc28j60 driver is the SPI transfers. Each
SPI transfer entails two kernel thread context switches. One major source of
SPI transfers is the enc28j60_set_bank() functions which runs before every
register access. This patch reduces the number of SPI transfers that
enc28j60_set_bank() performs in two ways:
1. removes unnecessary bank switch for the registers that are present in all
banks
2. when switching from banks 0 or 3 to banks 1 or 2 (i.e. only one bit
changes) enc28j60_set_bank() does only one SPI transfer instead of two
According to my tests these changes reduce the number of SPI transfers in
about 25%.
Signed-off-by: Baruch Siach <baruch@tkos.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>
| commit 3d29b0c33d
| Author: John W. Linville <linville@tuxdriver.com>
| Date: Fri Oct 31 14:13:12 2008 -0400
|
| netdevice zd1201: Convert directly reference of netdev->priv to netdev_priv()
|
| We have some reasons to kill netdev->priv:
| 1. netdev->priv is equal to netdev_priv().
| 2. netdev_priv() wraps the calculation of netdev->priv's offset, obviously
| netdev_priv() is more flexible than netdev->priv.
| But we cann't kill netdev->priv, because so many drivers reference to it
| directly.
|
| OK, becasue Dave S. Miller said, "every direct netdev->priv usage is a bug",
| and I want to kill netdev->priv later, I decided to convert all the direct
| reference of netdev->priv first.
|
| (Original patch posted by Wang Chen <wangchen@cn.fujitsu.com> w/ above
| changelog but using dev->ml_priv. That doesn't seem appropriate
| to me for this driver, so I've revamped it to use netdev_priv()
| instead. -- JWL)
This commit changed the allocation of netdev, but didn't change
the free method of it.
This causes "zd" be used after the memory, which is pointed by "zd", being
freed by free_netdev().
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Impact: New kerneldoc comments
Additional documentation added to all the alloc_cpumask and free_cpumask
functions.
Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (minor additions)
Impact: New API
This will be needed in x86 code to allocate the domain and old_domain
cpumasks on the same node as where the containing irq_cfg struct is
allocated.
(Also fixes double-dump_stack on rare CONFIG_DEBUG_PER_CPU_MAPS case)
Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (re-impl alloc_cpumask_var)
On PAE systems, GEM allocates pages using shmem, and passes these
pages to be bound into AGP, however the AGP interfaces + the x86
set_memory interfaces all take unsigned long not dma_addr_t.
The initial fix for this was a mess, so we need to do this correctly
for 2.6.29.
Signed-off-by: Dave Airlie <airlied@redhat.com>
These buffers don't have active rendering still occurring to them, they just
need either a flush to be emitted or a retire_requests to occur so that we
notice they're done. Return unbusy so that one of the two occurs. The two
expected consumers of this interface (OpenGL and libdrm_intel BO cache) both
want this behavior.
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When we read the write-intent-bitmap off the device, we currently
read a whole number of pages.
When PAGE_SIZE is 4K, this works due to the alignment we enforce
on the superblock and bitmap.
When PAGE_SIZE is 64K, this case read past the end-of-device
which causes an error.
When we write the superblock, we ensure to clip the last page
to just be the required size. Copy that code into the read path
to just read the required number of sectors.
Signed-off-by: Neil Brown <neilb@suse.de>
Cc: stable@kernel.org
This patch fixes a segfault in ppp_shutdown_interface() and
ppp_destroy_interface() when a PPP connection is closed. I bisected
the problem to the following commit:
commit c8019bf3af
Author: Wang Chen <wangchen@cn.fujitsu.com>
Date: Thu Nov 20 04:24:17 2008 -0800
netdevice ppp: Convert directly reference of netdev->priv
1. Use netdev_priv(dev) to replace dev->priv.
2. Alloc netdev's private data by alloc_netdev().
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The original ppp_generic code treated the netdev and struct ppp as
independent data structures which were freed separately. In moving the
ppp struct into the netdev, it is now possible for the private data to
be freed before the call to ppp_shutdown_interface(), which is bad.
The kfree(ppp) in ppp_destroy_interface() is also wrong; presumably
ppp hasn't worked since the above commit.
The following patch fixes both problems.
Signed-off-by: James Chapman <jchapman@katalix.com>
Reviewed-by: Wang Chen <wangchen@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The kernel_accept() does not hold the module refcount of newsock->ops->owner,
so we need __module_get(newsock->ops->owner) code after call kernel_accept()
by hand.
In sunrpc, the module refcount is missing to hold. So this cause kernel panic.
Used following script to reproduct:
while [ 1 ];
do
mount -t nfs4 192.168.0.19:/ /mnt
touch /mnt/file
umount /mnt
lsmod | grep ipv6
done
This patch fixed the problem by add __module_get(newsock->ops->owner) to
kernel_accept(). So we do not need to used __module_get(newsock->ops->owner)
in every place when used kernel_accept().
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
this warning:
arch/x86/kernel/microcode_amd.c: In function ‘apply_microcode_amd’:
arch/x86/kernel/microcode_amd.c:163: warning: cast from pointer to integer of different size
arch/x86/kernel/microcode_amd.c:163: warning: cast from pointer to integer of different size
triggers because we want to pass the address to the microcode MSR,
which is 64-bit even on 32-bit. Cast it explicitly to express this.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
these warnings:
kernel/trace/trace_sched_switch.c: In function ‘tracing_sched_register’:
kernel/trace/trace_sched_switch.c:96: warning: passing argument 1 of ‘register_trace_sched_wakeup_new’ from incompatible pointer type
kernel/trace/trace_sched_switch.c:112: warning: passing argument 1 of ‘unregister_trace_sched_wakeup_new’ from incompatible pointer type
kernel/trace/trace_sched_switch.c: In function ‘tracing_sched_unregister’:
kernel/trace/trace_sched_switch.c:121: warning: passing argument 1 of ‘unregister_trace_sched_wakeup_new’ from incompatible pointer type
Trigger because sched_wakeup_new tracepoints need the same trace
signature as sched_wakeup - which was changed recently.
Fix it.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
this warning:
kernel/trace/trace.c: In function ‘print_lat_fmt’:
kernel/trace/trace.c:1826: warning: unused variable ‘state’
Triggers because 'state' has become unused - remove it.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
this warning:
kernel/hrtimer.c: In function ‘hrtimer_cpu_notify’:
kernel/hrtimer.c:1574: warning: unused variable ‘dcpu’
is caused because 'dcpu' is only used in the CONFIG_HOTPLUG_CPU case.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: cleanup
Remove struct sigfram32 and rt_sigframe32 because there is no user.
Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: cleanup
Use rt_sigframe_ia32 instead of rt_sigframe32.
Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: cleanup
Include following headers for dependency.
asm/sigcontext.h
asm/siginfo.h
asm/ucontext.h
Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: cleanup
In asm/traps.h :-
do_double_fault : added under X86_64
sync_regs : added under X86_64
math_error : moved out from X86_32 as it is common for both 32 and 64 bit
math_emulate : moved from X86_32 as it is common for both 32 and 64 bit
smp_thermal_interrupt : added under X86_64
mce_threshold_interrupt : added under X86_64
Signed-off-by: Jaswinder Singh <jaswinder@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: New mm functionality.
Add pgprot_writecombine. pgprot_writecombine will be aliased to
pgprot_noncached when not supported by the architecture.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: mm behavior change.
Make pgprot_noncached uc_minus instead of strong UC. This will make
pgprot_noncached to be in line with ioremap_nocache() and all the other
APIs that map page uc_minus on uc request.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: New mm functionality.
Hookup remap_pfn_range and vm_insert_pfn and corresponding copy and free
routines with reserve and free tracking.
reserve and free here only takes care of non RAM region mapping. For RAM
region, driver should use set_memory_[uc|wc|wb] to set the cache type and
then setup the mapping for user pte. We can bypass below
reserve/free in that case.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: Introduces new hooks, which are currently null.
Introduce generic hooks in remap_pfn_range and vm_insert_pfn and
corresponding copy and free routines with reserve and free tracking.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: New currently unused interface.
Add a generic interface to follow pfn in a pfnmap vma range. This is used by
one of the subsequent x86 PAT related patch to keep track of memory types
for vma regions across vma copy and free.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Impact: Code transformation, new functions added should have no effect.
Drivers use mmap followed by pgprot_* and remap_pfn_range or vm_insert_pfn,
in order to export reserved memory to userspace. Currently, such mappings are
not tracked and hence not kept consistent with other mappings (/dev/mem,
pci resource, ioremap) for the sme memory, that may exist in the system.
The following patchset adds x86 PAT attribute tracking and untracking for
pfnmap related APIs.
First three patches in the patchset are changing the generic mm code to fit
in this tracking. Last four patches are x86 specific to make things work
with x86 PAT code. The patchset aso introduces pgprot_writecombine interface,
which gives writecombine mapping when enabled, falling back to
pgprot_noncached otherwise.
This patch:
While working on x86 PAT, we faced some hurdles with trackking
remap_pfn_range() regions, as we do not have any information to say
whether that PFNMAP mapping is linear for the entire vma range or
it is smaller granularity regions within the vma.
A simple solution to this is to use vm_pgoff as an indicator for
linear mapping over the vma region. Currently, remap_pfn_range
only sets vm_pgoff for COW mappings. Below patch changes the
logic and sets the vm_pgoff irrespective of COW. This will still not
be enough for the case where pfn is zero (vma region mapped to
physical address zero). But, for all the other cases, we can look at
pfnmap VMAs and say whether the mappng is for the entire vma region
or not.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
When CONFIG_PM is selected, the VFP code does not have any handler
installed to deal with either saving the VFP state of the current
task, nor does it do anything to try and restore the VFP after a
resume.
On resume, the VFP will have been reset and the co-processor access
control registers are in an indeterminate state (very probably the
CP10 and CP11 the VFP uses will have been disabled by the ARM core
reset). When this happens, resume will break as soon as it tries to
unfreeze the tasks and restart scheduling.
Add a sys device to allow us to hook the suspend call to save the
current thread state if the thread is using VFP and a resume hook
which restores the CP10/CP11 access and ensures the VFP is disabled
so that the lazy swapping will take place on next access.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Impact: cleanup, avoid sparse warning
Included asm/idle.h for c1e_remove_cpu() declaration. Fixes this
sparse warning:
CHECK arch/x86/kernel/process.c
arch/x86/kernel/process.c:284:6: warning: symbol 'c1e_remove_cpu' was not declared. Should it be static?
Signed-off-by: Jaswinder Singh <jaswinder@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This patch fixes a long-standing performance bug in classic RCU that
results in massive internal-to-RCU lock contention on systems with
more than a few hundred CPUs. Although this patch creates a separate
flavor of RCU for ease of review and patch maintenance, it is intended
to replace classic RCU.
This patch still handles stress better than does mainline, so I am still
calling it ready for inclusion. This patch is against the -tip tree.
Nevertheless, experience on an actual 1000+ CPU machine would still be
most welcome.
Most of the changes noted below were found while creating an rcutiny
(which should permit ejecting the current rcuclassic) and while doing
detailed line-by-line documentation.
Updates from v9 (http://lkml.org/lkml/2008/12/2/334):
o Fixes from remainder of line-by-line code walkthrough,
including comment spelling, initialization, undesirable
narrowing due to type conversion, removing redundant memory
barriers, removing redundant local-variable initialization,
and removing redundant local variables.
I do not believe that any of these fixes address the CPU-hotplug
issues that Andi Kleen was seeing, but please do give it a whirl
in case the machine is smarter than I am.
A writeup from the walkthrough may be found at the following
URL, in case you are suffering from terminal insomnia or
masochism:
http://www.kernel.org/pub/linux/kernel/people/paulmck/tmp/rcutree-walkthrough.2008.12.16a.pdf
o Made rcutree tracing use seq_file, as suggested some time
ago by Lai Jiangshan.
o Added a .csv variant of the rcudata debugfs trace file, to allow
people having thousands of CPUs to drop the data into
a spreadsheet. Tested with oocalc and gnumeric. Updated
documentation to suit.
Updates from v8 (http://lkml.org/lkml/2008/11/15/139):
o Fix a theoretical race between grace-period initialization and
force_quiescent_state() that could occur if more than three
jiffies were required to carry out the grace-period
initialization. Which it might, if you had enough CPUs.
o Apply Ingo's printk-standardization patch.
o Substitute local variables for repeated accesses to global
variables.
o Fix comment misspellings and redundant (but harmless) increments
of ->n_rcu_pending (this latter after having explicitly added it).
o Apply checkpatch fixes.
Updates from v7 (http://lkml.org/lkml/2008/10/10/291):
o Fixed a number of problems noted by Gautham Shenoy, including
the cpu-stall-detection bug that he was having difficulty
convincing me was real. ;-)
o Changed cpu-stall detection to wait for ten seconds rather than
three in order to reduce false positive, as suggested by Ingo
Molnar.
o Produced a design document (http://lwn.net/Articles/305782/).
The act of writing this document uncovered a number of both
theoretical and "here and now" bugs as noted below.
o Fix dynticks_nesting accounting confusion, simplify WARN_ON()
condition, fix kerneldoc comments, and add memory barriers
in dynticks interface functions.
o Add more data to tracing.
o Remove unused "rcu_barrier" field from rcu_data structure.
o Count calls to rcu_pending() from scheduling-clock interrupt
to use as a surrogate timebase should jiffies stop counting.
o Fix a theoretical race between force_quiescent_state() and
grace-period initialization. Yes, initialization does have to
go on for some jiffies for this race to occur, but given enough
CPUs...
Updates from v6 (http://lkml.org/lkml/2008/9/23/448):
o Fix a number of checkpatch.pl complaints.
o Apply review comments from Ingo Molnar and Lai Jiangshan
on the stall-detection code.
o Fix several bugs in !CONFIG_SMP builds.
o Fix a misspelled config-parameter name so that RCU now announces
at boot time if stall detection is configured.
o Run tests on numerous combinations of configurations parameters,
which after the fixes above, now build and run correctly.
Updates from v5 (http://lkml.org/lkml/2008/9/15/92, bad subject line):
o Fix a compiler error in the !CONFIG_FANOUT_EXACT case (blew a
changeset some time ago, and finally got around to retesting
this option).
o Fix some tracing bugs in rcupreempt that caused incorrect
totals to be printed.
o I now test with a more brutal random-selection online/offline
script (attached). Probably more brutal than it needs to be
on the people reading it as well, but so it goes.
o A number of optimizations and usability improvements:
o Make rcu_pending() ignore the grace-period timeout when
there is no grace period in progress.
o Make force_quiescent_state() avoid going for a global
lock in the case where there is no grace period in
progress.
o Rearrange struct fields to improve struct layout.
o Make call_rcu() initiate a grace period if RCU was
idle, rather than waiting for the next scheduling
clock interrupt.
o Invoke rcu_irq_enter() and rcu_irq_exit() only when
idle, as suggested by Andi Kleen. I still don't
completely trust this change, and might back it out.
o Make CONFIG_RCU_TRACE be the single config variable
manipulated for all forms of RCU, instead of the prior
confusion.
o Document tracing files and formats for both rcupreempt
and rcutree.
Updates from v4 for those missing v5 given its bad subject line:
o Separated dynticks interface so that NMIs and irqs call separate
functions, greatly simplifying it. In particular, this code
no longer requires a proof of correctness. ;-)
o Separated dynticks state out into its own per-CPU structure,
avoiding the duplicated accounting.
o The case where a dynticks-idle CPU runs an irq handler that
invokes call_rcu() is now correctly handled, forcing that CPU
out of dynticks-idle mode.
o Review comments have been applied (thank you all!!!).
For but one example, fixed the dynticks-ordering issue that
Manfred pointed out, saving me much debugging. ;-)
o Adjusted rcuclassic and rcupreempt to handle dynticks changes.
Attached is an updated patch to Classic RCU that applies a hierarchy,
greatly reducing the contention on the top-level lock for large machines.
This passes 10-hour concurrent rcutorture and online-offline testing on
128-CPU ppc64 without dynticks enabled, and exposes some timekeeping
bugs in presence of dynticks (exciting working on a system where
"sleep 1" hangs until interrupted...), which were fixed in the
2.6.27 kernel. It is getting more reliable than mainline by some
measures, so the next version will be against -tip for inclusion.
See also Manfred Spraul's recent patches (or his earlier work from
2004 at http://marc.info/?l=linux-kernel&m=108546384711797&w=2).
We will converge onto a common patch in the fullness of time, but are
currently exploring different regions of the design space. That said,
I have already gratefully stolen quite a few of Manfred's ideas.
This patch provides CONFIG_RCU_FANOUT, which controls the bushiness
of the RCU hierarchy. Defaults to 32 on 32-bit machines and 64 on
64-bit machines. If CONFIG_NR_CPUS is less than CONFIG_RCU_FANOUT,
there is no hierarchy. By default, the RCU initialization code will
adjust CONFIG_RCU_FANOUT to balance the hierarchy, so strongly NUMA
architectures may choose to set CONFIG_RCU_FANOUT_EXACT to disable
this balancing, allowing the hierarchy to be exactly aligned to the
underlying hardware. Up to two levels of hierarchy are permitted
(in addition to the root node), allowing up to 16,384 CPUs on 32-bit
systems and up to 262,144 CPUs on 64-bit systems. I just know that I
am going to regret saying this, but this seems more than sufficient
for the foreseeable future. (Some architectures might wish to set
CONFIG_RCU_FANOUT=4, which would limit such architectures to 64 CPUs.
If this becomes a real problem, additional levels can be added, but I
doubt that it will make a significant difference on real hardware.)
In the common case, a given CPU will manipulate its private rcu_data
structure and the rcu_node structure that it shares with its immediate
neighbors. This can reduce both lock and memory contention by multiple
orders of magnitude, which should eliminate the need for the strange
manipulations that are reported to be required when running Linux on
very large systems.
Some shortcomings:
o More bugs will probably surface as a result of an ongoing
line-by-line code inspection.
Patches will be provided as required.
o There are probably hangs, rcutorture failures, &c. Seems
quite stable on a 128-CPU machine, but that is kind of small
compared to 4096 CPUs. However, seems to do better than
mainline.
Patches will be provided as required.
o The memory footprint of this version is several KB larger
than rcuclassic.
A separate UP-only rcutiny patch will be provided, which will
reduce the memory footprint significantly, even compared
to the old rcuclassic. One such patch passes light testing,
and has a memory footprint smaller even than rcuclassic.
Initial reaction from various embedded guys was "it is not
worth it", so am putting it aside.
Credits:
o Manfred Spraul for ideas, review comments, and bugs spotted,
as well as some good friendly competition. ;-)
o Josh Triplett, Ingo Molnar, Peter Zijlstra, Mathieu Desnoyers,
Lai Jiangshan, Andi Kleen, Andy Whitcroft, and Andrew Morton
for reviews and comments.
o Thomas Gleixner for much-needed help with some timer issues
(see patches below).
o Jon M. Tollefson, Tim Pepper, Andrew Theurer, Jose R. Santos,
Andy Whitcroft, Darrick Wong, Nishanth Aravamudan, Anton
Blanchard, Dave Kleikamp, and Nathan Lynch for keeping machines
alive despite my heavy abuse^Wtesting.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
