re-add some of the comments that got lost in the refactoring.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Now that we have nice numerical relations for the states, remove the macro
magics.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Now what its only two functions, they again look rather similar.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Generate the state bit definitions from the lockdep_states.h file.
Also, move LOCK_USED to last, so that the
USED_IN
USED_IN_READ
ENABLED
ENABLED_READ
states are nicely bit aligned -- we're going to use that property
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Introduce a header file to generate all the states from.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
s/HELD_OVER/ENABLED/g
so that its similar to the hard and soft-irq names.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
s/\(LOCKF\?_ENABLED_[^ ]*\)S\(_READ\)\?\>/\1\2/g
So that the USED_IN and ENABLED have the same names.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Here is another version, with the incremental patch rolled up, and
added reclaim context annotation to kswapd, and allocation tracing
to slab allocators (which may only ever reach the page allocator
in rare cases, so it is good to put annotations here too).
Haven't tested this version as such, but it should be getting closer
to merge worthy ;)
--
After noticing some code in mm/filemap.c accidentally perform a __GFP_FS
allocation when it should not have been, I thought it might be a good idea to
try to catch this kind of thing with lockdep.
I coded up a little idea that seems to work. Unfortunately the system has to
actually be in __GFP_FS page reclaim, then take the lock, before it will mark
it. But at least that might still be some orders of magnitude more common
(and more debuggable) than an actual deadlock condition, so we have some
improvement I hope (the concept is no less complete than discovery of a lock's
interrupt contexts).
I guess we could even do the same thing with __GFP_IO (normal reclaim), and
even GFP_NOIO locks too... but filesystems will have the most locks and fiddly
code paths, so let's start there and see how it goes.
It *seems* to work. I did a quick test.
=================================
[ INFO: inconsistent lock state ]
2.6.28-rc6-00007-ged31348-dirty #26
---------------------------------
inconsistent {in-reclaim-W} -> {ov-reclaim-W} usage.
modprobe/8526 [HC0[0]:SC0[0]:HE1:SE1] takes:
(testlock){--..}, at: [<ffffffffa0020055>] brd_init+0x55/0x216 [brd]
{in-reclaim-W} state was registered at:
[<ffffffff80267bdb>] __lock_acquire+0x75b/0x1a60
[<ffffffff80268f71>] lock_acquire+0x91/0xc0
[<ffffffff8070f0e1>] mutex_lock_nested+0xb1/0x310
[<ffffffffa002002b>] brd_init+0x2b/0x216 [brd]
[<ffffffff8020903b>] _stext+0x3b/0x170
[<ffffffff80272ebf>] sys_init_module+0xaf/0x1e0
[<ffffffff8020c3fb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
irq event stamp: 3929
hardirqs last enabled at (3929): [<ffffffff8070f2b5>] mutex_lock_nested+0x285/0x310
hardirqs last disabled at (3928): [<ffffffff8070f089>] mutex_lock_nested+0x59/0x310
softirqs last enabled at (3732): [<ffffffff8061f623>] sk_filter+0x83/0xe0
softirqs last disabled at (3730): [<ffffffff8061f5b6>] sk_filter+0x16/0xe0
other info that might help us debug this:
1 lock held by modprobe/8526:
#0: (testlock){--..}, at: [<ffffffffa0020055>] brd_init+0x55/0x216 [brd]
stack backtrace:
Pid: 8526, comm: modprobe Not tainted 2.6.28-rc6-00007-ged31348-dirty #26
Call Trace:
[<ffffffff80265483>] print_usage_bug+0x193/0x1d0
[<ffffffff80266530>] mark_lock+0xaf0/0xca0
[<ffffffff80266735>] mark_held_locks+0x55/0xc0
[<ffffffffa0020000>] ? brd_init+0x0/0x216 [brd]
[<ffffffff802667ca>] trace_reclaim_fs+0x2a/0x60
[<ffffffff80285005>] __alloc_pages_internal+0x475/0x580
[<ffffffff8070f29e>] ? mutex_lock_nested+0x26e/0x310
[<ffffffffa0020000>] ? brd_init+0x0/0x216 [brd]
[<ffffffffa002006a>] brd_init+0x6a/0x216 [brd]
[<ffffffffa0020000>] ? brd_init+0x0/0x216 [brd]
[<ffffffff8020903b>] _stext+0x3b/0x170
[<ffffffff8070f8b9>] ? mutex_unlock+0x9/0x10
[<ffffffff8070f83d>] ? __mutex_unlock_slowpath+0x10d/0x180
[<ffffffff802669ec>] ? trace_hardirqs_on_caller+0x12c/0x190
[<ffffffff80272ebf>] sys_init_module+0xaf/0x1e0
[<ffffffff8020c3fb>] system_call_fastpath+0x16/0x1b
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This modifies the timer code in a way to allow lockdep to detect
deadlocks resulting from a lock being taken in the timer function
as well as around the del_timer_sync() call.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
I happened to forked lots of processes, and hit NULL pointer dereference.
It is because in copy_process() after checking max_threads, 0 is returned
but not -EAGAIN.
The bug is introduced by "CRED: Detach the credentials from task_struct"
(commit f1752eec61).
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: James Morris <jmorris@namei.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With exclusive waiters, every process woken up through the wait queue must
ensure that the next waiter down the line is woken when it has finished.
Interruptible waiters don't do that when aborting due to a signal. And if
an aborting waiter is concurrently woken up through the waitqueue, noone
will ever wake up the next waiter.
This has been observed with __wait_on_bit_lock() used by
lock_page_killable(): the first contender on the queue was aborting when
the actual lock holder woke it up concurrently. The aborted contender
didn't acquire the lock and therefor never did an unlock followed by
waking up the next waiter.
Add abort_exclusive_wait() which removes the process' wait descriptor from
the waitqueue, iff still queued, or wakes up the next waiter otherwise.
It does so under the waitqueue lock. Racing with a wake up means the
aborting process is either already woken (removed from the queue) and will
wake up the next waiter, or it will remove itself from the queue and the
concurrent wake up will apply to the next waiter after it.
Use abort_exclusive_wait() in __wait_event_interruptible_exclusive() and
__wait_on_bit_lock() when they were interrupted by other means than a wake
up through the queue.
[akpm@linux-foundation.org: coding-style fixes]
Reported-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Mentored-by: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Chuck Lever <cel@citi.umich.edu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: <stable@kernel.org> ["after some testing"]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Revert commit 0c2d64fb6c because it causes
(arguably poorly designed) existing userspace to spend interminable
periods closing billions of not-open file descriptors.
We could bring this back, with some sort of opt-in tunable in /proc, which
defaults to "off".
Peter's alanysis follows:
: I spent several hours trying to get to the bottom of a serious
: performance issue that appeared on one of our servers after upgrading to
: 2.6.28. In the end it's what could be considered a userspace bug that
: was triggered by a change in 2.6.28. Since this might also affect other
: people I figured I'd at least document what I found here, and maybe we
: can even do something about it:
:
:
: So, I upgraded some of debian.org's machines to 2.6.28.1 and immediately
: the team maintaining our ftp archive complained that one of their
: scripts that previously ran in a few minutes still hadn't even come
: close to being done after an hour or so. Downgrading to 2.6.27 fixed
: that.
:
: Turns out that script is forking a lot and something in it or python or
: whereever closes all the file descriptors it doesn't want to pass on.
: That is, it starts at zero and goes up to ulimit -n/RLIMIT_NOFILE and
: closes them all with a few exceptions.
:
: Turns out that takes a long time when your limit -n is now 2^20 (1048576).
:
: With 2.6.27.* the ulimit -n was the standard 1024, but with 2.6.28 it is
: now a thousand times that.
:
: 2.6.28 included a patch titled "rlimit: permit setting RLIMIT_NOFILE to
: RLIM_INFINITY" (0c2d64fb6c)[1] that
: allows, as the title implies, to set the limit for number of files to
: infinity.
:
: Closer investigation showed that the broken default ulimit did not apply
: to "system" processes (like stuff started from init). In the end I
: could establish that all processes that passed through pam_limit at one
: point had the bad resource limit.
:
: Apparently the pam library in Debian etch (4.0) initializes the limits
: to some default values when it doesn't have any settings in limit.conf
: to override them. Turns out that for nofiles this is RLIM_INFINITY.
: Commenting out "case RLIMIT_NOFILE" in pam_limit.c:267 of our pam
: package version 0.79-5 fixes that - tho I'm not sure what side effects
: that has.
:
: Debian lenny (the upcoming 5.0 version) doesn't have this issue as it
: uses a different pam (version).
Reported-by: Peter Palfrader <weasel@debian.org>
Cc: Adam Tkac <vonsch@gmail.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Cc: <stable@kernel.org> [2.6.28.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
alpha:
kernel/async.c: In function 'run_one_entry':
kernel/async.c:141: warning: format '%lli' expects type 'long long int', but argument 2 has type 'async_cookie_t'
kernel/async.c:149: warning: format '%lli' expects type 'long long int', but argument 2 has type 'async_cookie_t'
kernel/async.c:149: warning: format '%lld' expects type 'long long int', but argument 4 has type 's64'
kernel/async.c: In function 'async_synchronize_cookie_special':
kernel/async.c:250: warning: format '%lli' expects type 'long long int', but argument 3 has type 's64'
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86: APIC: enable workaround on AMD Fam10h CPUs
xen: disable interrupts before saving in percpu
x86: add x86@kernel.org to MAINTAINERS
x86: push old stack address on irqstack for unwinder
irq, x86: fix lock status with numa_migrate_irq_desc
x86: add cache descriptors for Intel Core i7
x86/Voyager: make it build and boot
"ftrace: use struct pid" commit 978f3a45d9
converted ftrace_pid_trace to "struct pid*".
But we can't use do_each_pid_task() without rcu_read_lock() even if
we know the pid itself can't go away (it was pinned in ftrace_pid_write).
The exiting task can detach itself from this pid at any moment.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
sched_rt: don't use first_cpu on cpumask created with cpumask_and
sched: fix buddie group latency
sched: clear buddies more aggressively
sched: symmetric sync vs avg_overlap
sched: fix sync wakeups
cpuset: fix possible deadlock in async_rebuild_sched_domains
Current refcounting for modules (done if CONFIG_MODULE_UNLOAD=y) is
using a lot of memory.
Each 'struct module' contains an [NR_CPUS] array of full cache lines.
This patch uses existing infrastructure (percpu_modalloc() &
percpu_modfree()) to allocate percpu space for the refcount storage.
Instead of wasting NR_CPUS*128 bytes (on i386), we now use
nr_cpu_ids*sizeof(local_t) bytes.
On a typical distro, where NR_CPUS=8, shiping 2000 modules, we reduce
size of module files by about 2 Mbytes. (1Kb per module)
Instead of having all refcounters in the same memory node - with TLB misses
because of vmalloc() - this new implementation permits to have better
NUMA properties, since each CPU will use storage on its preferred node,
thanks to percpu storage.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Eric Paris reported:
> I have an hp dl785g5 which is unable to successfully run
> 2.6.29-0.66.rc3.fc11.x86_64 or 2.6.29-rc2-next-20090126. During bootup
> (early in userspace daemons starting) I get the below BUG, which quickly
> renders the machine dead. I assume it is because sparse_irq_lock never
> gets released when the BUG kills that task.
Adjust lock sequence when migrating a descriptor with
CONFIG_NUMA_MIGRATE_IRQ_DESC enabled.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
cpumask_and() only initializes nr_cpu_ids bits, so the (deprecated)
first_cpu() might find one of those uninitialized bits if nr_cpu_ids
is less than NR_CPUS (as it can be for CONFIG_CPUMASK_OFFSTACK).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Similar to the previous patch, by not clearing buddies we can select entities
past their run quota, which can increase latency. This means we have to clear
group buddies as well.
Do not use the group clear for pick_next_task(), otherwise that'll get O(n^2).
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
It was noticed that a task could get re-elected past its run quota due to buddy
affinities. This could increase latency a little. Cure it by more aggresively
clearing buddy state.
We do so in two situations:
- when we force preempt
- when we select a buddy to run
Signed-off-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Reinstate the weakening of the sync hint if set. This yields a more
symmetric usage of avg_overlap.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Pawel Dziekonski reported that the openssl benchmark and his
quantum chemistry application both show slowdowns due to the
scheduler under-parallelizing execution.
The reason are pipe wakeups still doing 'sync' wakeups which
overrides the normal buddy wakeup logic - even if waker and
wakee are loosely coupled.
Fix an inversion of logic in the buddy wakeup code.
Reported-by: Pawel Dziekonski <dzieko@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
generic-ipi: use per cpu data for single cpu ipi calls
cpumask: convert lib/smp_processor_id to new cpumask ops
signals, debug: fix BUG: using smp_processor_id() in preemptible code in print_fatal_signal()
* 'timers-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
hrtimer: prevent negative expiry value after clock_was_set()
hrtimers: allow the hot-unplugging of all cpus
hrtimers: increase clock min delta threshold while interrupt hanging
* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, ds, bts: cleanup/fix DS configuration
ring-buffer: reset timestamps when ring buffer is reset
trace: set max latency variable to zero on default
trace: stop all recording to ring buffer on ftrace_dump
trace: print ftrace_dump at KERN_EMERG log level
ring_buffer: reset write when reserve buffer fail
tracing/function-graph-tracer: fix a regression while suspend to disk
ring-buffer: fix alignment problem
Impact: prevent false positive WARN_ON() in clockevents_program_event()
clock_was_set() changes the base->offset of CLOCK_REALTIME and
enforces the reprogramming of the clockevent device to expire timers
which are based on CLOCK_REALTIME. If the clock change is large enough
then the subtraction of the timer expiry value and base->offset can
become negative which triggers the warning in
clockevents_program_event().
Check the subtraction result and set a negative value to 0.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Impact: fix CPU hotplug hang on Power6 testbox
On architectures that support offlining all cpus (at least powerpc/pseries),
hot-unpluging the tick_do_timer_cpu can result in a system hang.
This comes from the fact that if the cpu going down happens to be the
cpu doing the tick, then as the tick_do_timer_cpu handover happens after the
cpu is dead (via the CPU_DEAD notification), we're left without ticks,
jiffies are frozen and any task relying on timers (msleep, ...) is stuck.
That's particularly the case for the cpu looping in __cpu_die() waiting
for the dying cpu to be dead.
This patch addresses this by having the tick_do_timer_cpu handover happen
earlier during the CPU_DYING notification. For this, a new clockevent
notification type is introduced (CLOCK_EVT_NOTIFY_CPU_DYING) which is triggered
in hrtimer_cpu_notify().
Signed-off-by: Sebastien Dugue <sebastien.dugue@bull.net>
Cc: <stable@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: avoid timer IRQ hanging slow systems
While using the function graph tracer on a virtualized system, the
hrtimer_interrupt can hang the system on an infinite loop.
This can be caused in several situations:
- the hardware is very slow and HZ is set too high
- something intrusive is slowing the system down (tracing under emulation)
... and the next clock events to program are always before the current time.
This patch implements a reasonable compromise: if such a situation is
detected, we share the CPUs time in 1/4 to process the hrtimer interrupts.
This is enough to let the system running without serious starvation.
It has been successfully tested under VirtualBox with 1000 HZ and 100 HZ
with function graph tracer launched. On both cases, the clock events were
increased until about 25 ms periodic ticks, which means 40 HZ.
So we change a hard to debug hang into a warning message and a system that
still manages to limp along.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The smp_call_function can be passed a wait parameter telling it to
wait for all the functions running on other CPUs to complete before
returning, or to return without waiting. Unfortunately, this is
currently just a suggestion and not manditory. That is, the
smp_call_function can decide not to return and wait instead.
The reason for this is because it uses kmalloc to allocate storage
to send to the called CPU and that CPU will free it when it is done.
But if we fail to allocate the storage, the stack is used instead.
This means we must wait for the called CPU to finish before
continuing.
Unfortunatly, some callers do no abide by this hint and act as if
the non-wait option is mandatory. The MTRR code for instance will
deadlock if the smp_call_function is set to wait. This is because
the smp_call_function will wait for the other CPUs to finish their
called functions, but those functions are waiting on the caller to
continue.
This patch changes the generic smp_call_function code to use per cpu
variables if the allocation of the data fails for a single CPU call. The
smp_call_function_many will fall back to the smp_call_function_single
if it fails its alloc. The smp_call_function_single is modified
to not force the wait state.
Since we now are using a single data per cpu we must synchronize the
callers to prevent a second caller modifying the data before the
first called IPI functions complete. To do so, I added a flag to
the call_single_data called CSD_FLAG_LOCK. When the single CPU is
called (which can be called when a many call fails an alloc), we
set the LOCK bit on this per cpu data. When the caller finishes
it clears the LOCK bit.
The caller must wait till the LOCK bit is cleared before setting
it. When it is cleared, there is no IPI function using it.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Jens Axboe <jens.axboe@oracle.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
root_count was being incremented in cgroup_get_sb() after all error
checking was complete, but decremented in cgroup_kill_sb(), which can be
called on a superblock that we gave up on due to an error. This patch
changes cgroup_kill_sb() to only decrement root_count if the root was
previously linked into the list of roots.
Signed-off-by: Paul Menage <menage@google.com>
Tested-by: Serge Hallyn <serue@us.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
css_tryget() and cgroup_clear_css_refs() contain polling loops; these
loops should have cpu_relax calls in them to reduce cross-cache traffic.
Signed-off-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I fixed a bug in cgroup_clone() in Linus' tree in commit 7b574b7
("cgroups: fix a race between cgroup_clone and umount") without noticing
there was a cleanup patch in -mm tree that should be rebased (now commit
104cbd5, "cgroups: use task_lock() for access tsk->cgroups safe in
cgroup_clone()"), thus resulted in lock inconsistency.
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now, cgrp->sibling is handled under hierarchy mutex.
error route should do so, too.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Acked-by Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move the initialization of irq_default_affinity to early_irq_init as
core_initcall is too late.
irq_default_affinity can be used in init_IRQ and potentially timer and
SMP init as well. All of these happen before core_initcall. Moving
the initialization to early_irq_init ensures that it is initialized
before it is used.
Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Acked-by: Mike Travis <travis@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Peter Zijlstra