* Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git)
@ 2009-03-27 5:07 Larry Finger
2009-03-27 12:54 ` Bartlomiej Zolnierkiewicz
0 siblings, 1 reply; 6+ messages in thread
From: Larry Finger @ 2009-03-27 5:07 UTC (permalink / raw)
To: LKML
On my x86_64 dual core system, I get the following messages logged.
=========================================================
[ INFO: possible irq lock inversion dependency detected ]
2.6.29-Linus-03321-gbe0ea69 #55
---------------------------------------------------------
swapper/0 just changed the state of lock:
(fasync_lock){..+.}, at: [<ffffffff802c7073>] kill_fasync+0x2c/0x4e
but this lock took another, hard-irq-unsafe lock in the past:
(&f->f_lock){--..}
and interrupts could create inverse lock ordering between them.
other info that might help us debug this:
4 locks held by swapper/0:
#0: (&serio->lock){+...}, at: [<ffffffff803c47a4>] serio_interrupt+0x2b/0x83
#1: (&dev->event_lock){+...}, at: [<ffffffff803c8bd3>] input_event+0x40/0x7d
#2: (rcu_read_lock){..--}, at: [<ffffffff803c7541>] input_pass_event+0x0/0xcc
#3: (rcu_read_lock){..--}, at: [<ffffffff803cbad3>] evdev_event+0x0/0xfd
the first lock's dependencies:
-> (fasync_lock){..+.} ops: 0 {
initial-use at:
[<ffffffff80262870>] __lock_acquire+0x79e/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff804642bf>] _write_lock_irq+0x37/0x43
[<ffffffff802c6bf5>] fasync_helper+0x57/0x110
[<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
[<ffffffff802ea819>] __posix_lock_file+0x326/0x549
[<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
[<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
[<ffffffff802b9634>] filp_close+0x56/0x68
[<ffffffff802b96f0>] sys_close+0xaa/0xe9
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
in-hardirq-R at:
[<ffffffffffffffff>] 0xffffffffffffffff
}
... key at: [<ffffffff805f7db8>] fasync_lock+0x18/0x30
-> (&f->f_lock){--..} ops: 0 {
initial-use at:
[<ffffffff80262870>] __lock_acquire+0x79e/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c6c74>] fasync_helper+0xd6/0x110
[<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
[<ffffffff802ea819>] __posix_lock_file+0x326/0x549
[<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
[<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
[<ffffffff802b9634>] filp_close+0x56/0x68
[<ffffffff802b96f0>] sys_close+0xaa/0xe9
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
softirq-on-W at:
[<ffffffff80262857>] __lock_acquire+0x785/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
hardirq-on-W at:
[<ffffffff8026282e>] __lock_acquire+0x75c/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
}
... key at: [<ffffffff81096e08>] __key.20386+0x0/0x8
... acquired at:
[<ffffffff80263404>] __lock_acquire+0x1332/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c6c74>] fasync_helper+0xd6/0x110
[<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
[<ffffffff802ea819>] __posix_lock_file+0x326/0x549
[<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
[<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
[<ffffffff802b9634>] filp_close+0x56/0x68
[<ffffffff802b96f0>] sys_close+0xaa/0xe9
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
the second lock's dependencies:
-> (&f->f_lock){--..} ops: 0 {
initial-use at:
[<ffffffff80262870>] __lock_acquire+0x79e/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c6c74>] fasync_helper+0xd6/0x110
[<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
[<ffffffff802ea819>] __posix_lock_file+0x326/0x549
[<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
[<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
[<ffffffff802b9634>] filp_close+0x56/0x68
[<ffffffff802b96f0>] sys_close+0xaa/0xe9
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
softirq-on-W at:
[<ffffffff80262857>] __lock_acquire+0x785/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
hardirq-on-W at:
[<ffffffff8026282e>] __lock_acquire+0x75c/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff80464126>] _spin_lock+0x31/0x3d
[<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
[<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff
}
... key at: [<ffffffff81096e08>] __key.20386+0x0/0x8
stack backtrace:
Pid: 0, comm: swapper Not tainted 2.6.29-Linus-03321-gbe0ea69 #55
Call Trace:
<IRQ> [<ffffffff80260a11>] print_irq_inversion_bug+0x175/0x184
[<ffffffff80260ab8>] check_usage_forwards+0x98/0xa1
[<ffffffff8026113b>] mark_lock+0x5dd/0xa2b
[<ffffffff80262748>] __lock_acquire+0x676/0x1697
[<ffffffff80262870>] ? __lock_acquire+0x79e/0x1697
[<ffffffff802637be>] lock_acquire+0x55/0x71
[<ffffffff802c7073>] ? kill_fasync+0x2c/0x4e
[<ffffffff8046442b>] _read_lock+0x34/0x41
[<ffffffff802c7073>] ? kill_fasync+0x2c/0x4e
[<ffffffff802c7073>] kill_fasync+0x2c/0x4e
[<ffffffff803cb7bd>] evdev_pass_event+0x6e/0x77
[<ffffffff803cbb47>] evdev_event+0x74/0xfd
[<ffffffff803cbad3>] ? evdev_event+0x0/0xfd
[<ffffffff803c75a5>] input_pass_event+0x64/0xcc
[<ffffffff803c7541>] ? input_pass_event+0x0/0xcc
[<ffffffff803c7cdf>] input_handle_event+0x339/0x345
[<ffffffff803c8bf6>] input_event+0x63/0x7d
[<ffffffff803d1275>] alps_process_packet+0x3b7/0x4cc
[<ffffffff803d13e5>] alps_process_byte+0x5b/0x6e
[<ffffffff803ce797>] psmouse_handle_byte+0x17/0x10d
[<ffffffff803cf766>] psmouse_interrupt+0x277/0x285
[<ffffffff803c47a4>] ? serio_interrupt+0x2b/0x83
[<ffffffff803c47c0>] serio_interrupt+0x47/0x83
[<ffffffff803c58b7>] i8042_interrupt+0x200/0x215
[<ffffffff8027b277>] handle_IRQ_event+0x6b/0xa1
[<ffffffff8027c55c>] handle_edge_irq+0xeb/0x131
[<ffffffff80464a25>] __irqentry_text_start+0x6d/0xd8
[<ffffffff8020cd13>] ret_from_intr+0x0/0xf
<EOI> [<ffffffff8025d068>] ? tick_broadcast_oneshot_control+0x1f/0x102
[<ffffffff802130e6>] ? default_idle+0x35/0x4f
[<ffffffff802130e4>] ? default_idle+0x33/0x4f
[<ffffffff802131ed>] ? c1e_idle+0xda/0x101
[<ffffffff8025713e>] ? atomic_notifier_call_chain+0xf/0x11
[<ffffffff8020b1f6>] ? cpu_idle+0x5c/0x9d
[<ffffffff80451797>] ? rest_init+0x6b/0x6d
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git)
2009-03-27 5:07 Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git) Larry Finger
@ 2009-03-27 12:54 ` Bartlomiej Zolnierkiewicz
2009-03-27 18:06 ` Jonathan Corbet
0 siblings, 1 reply; 6+ messages in thread
From: Bartlomiej Zolnierkiewicz @ 2009-03-27 12:54 UTC (permalink / raw)
To: Larry Finger
Cc: LKML, Jonathan Corbet, Christoph Hellwig, Al Viro, Li Zefan,
Wu Fengguang, Ingo Molnar
Hi,
I've seen this reported before against linux-next:
http://lkml.org/lkml/2009/3/2/399
and I also experienced it myself (since at least next-20090213).
I haven't verified it but I believe that the warning is triggered by
"Move FASYNC bit handling to f_op->fasync()" change (which is commit
76398425bb06b07cc3a3b1ce169c67dc9d6874ed in Linus' tree).
I remember looking a bit more closely into the issue and not seeing
the problem with the locking (though I could have missed something):
file->f_lock is never taken in hard-irq or soft-irq context and in
the only place where file->f_lock is taken with fasync_lock hold we're
protected against IRQs by write_lock_irq().
[ Despite not being a problem now I think that changing spin_[un]lock()
to *_irq() variants for file->f_lock could be (given that it really
fixes the warning) more viable long-term solution than adding special
lockdep handling (well, it could be that one day file->f_lock is used
in soft-irq context and then the irq lock inversion issue will become
a real one) and shouldn't incurr performance penalty since we hold it
only for a very brief time. ]
On Friday 27 March 2009, Larry Finger wrote:
> On my x86_64 dual core system, I get the following messages logged.
>
>
> =========================================================
> [ INFO: possible irq lock inversion dependency detected ]
> 2.6.29-Linus-03321-gbe0ea69 #55
> ---------------------------------------------------------
> swapper/0 just changed the state of lock:
> (fasync_lock){..+.}, at: [<ffffffff802c7073>] kill_fasync+0x2c/0x4e
> but this lock took another, hard-irq-unsafe lock in the past:
> (&f->f_lock){--..}
>
> and interrupts could create inverse lock ordering between them.
>
>
> other info that might help us debug this:
> 4 locks held by swapper/0:
> #0: (&serio->lock){+...}, at: [<ffffffff803c47a4>] serio_interrupt+0x2b/0x83
> #1: (&dev->event_lock){+...}, at: [<ffffffff803c8bd3>] input_event+0x40/0x7d
> #2: (rcu_read_lock){..--}, at: [<ffffffff803c7541>] input_pass_event+0x0/0xcc
> #3: (rcu_read_lock){..--}, at: [<ffffffff803cbad3>] evdev_event+0x0/0xfd
>
> the first lock's dependencies:
> -> (fasync_lock){..+.} ops: 0 {
> initial-use at:
> [<ffffffff80262870>] __lock_acquire+0x79e/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff804642bf>] _write_lock_irq+0x37/0x43
> [<ffffffff802c6bf5>] fasync_helper+0x57/0x110
> [<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
> [<ffffffff802ea819>] __posix_lock_file+0x326/0x549
> [<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
> [<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
> [<ffffffff802b9634>] filp_close+0x56/0x68
> [<ffffffff802b96f0>] sys_close+0xaa/0xe9
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> in-hardirq-R at:
> [<ffffffffffffffff>] 0xffffffffffffffff
> }
> ... key at: [<ffffffff805f7db8>] fasync_lock+0x18/0x30
> -> (&f->f_lock){--..} ops: 0 {
> initial-use at:
> [<ffffffff80262870>] __lock_acquire+0x79e/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c6c74>] fasync_helper+0xd6/0x110
> [<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
> [<ffffffff802ea819>] __posix_lock_file+0x326/0x549
> [<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
> [<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
> [<ffffffff802b9634>] filp_close+0x56/0x68
> [<ffffffff802b96f0>] sys_close+0xaa/0xe9
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> softirq-on-W at:
> [<ffffffff80262857>] __lock_acquire+0x785/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> hardirq-on-W at:
> [<ffffffff8026282e>] __lock_acquire+0x75c/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> }
> ... key at: [<ffffffff81096e08>] __key.20386+0x0/0x8
> ... acquired at:
> [<ffffffff80263404>] __lock_acquire+0x1332/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c6c74>] fasync_helper+0xd6/0x110
> [<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
> [<ffffffff802ea819>] __posix_lock_file+0x326/0x549
> [<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
> [<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
> [<ffffffff802b9634>] filp_close+0x56/0x68
> [<ffffffff802b96f0>] sys_close+0xaa/0xe9
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
>
>
> the second lock's dependencies:
> -> (&f->f_lock){--..} ops: 0 {
> initial-use at:
> [<ffffffff80262870>] __lock_acquire+0x79e/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c6c74>] fasync_helper+0xd6/0x110
> [<ffffffff802ea28c>] locks_delete_lock+0x50/0x9f
> [<ffffffff802ea819>] __posix_lock_file+0x326/0x549
> [<ffffffff802eaa8c>] vfs_lock_file+0x38/0x3a
> [<ffffffff802eab1f>] locks_remove_posix+0x91/0xb2
> [<ffffffff802b9634>] filp_close+0x56/0x68
> [<ffffffff802b96f0>] sys_close+0xaa/0xe9
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> softirq-on-W at:
> [<ffffffff80262857>] __lock_acquire+0x785/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> hardirq-on-W at:
> [<ffffffff8026282e>] __lock_acquire+0x75c/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff80464126>] _spin_lock+0x31/0x3d
> [<ffffffff802c7505>] sys_fcntl+0x2aa/0x36c
> [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b
> [<ffffffffffffffff>] 0xffffffffffffffff
> }
> ... key at: [<ffffffff81096e08>] __key.20386+0x0/0x8
>
> stack backtrace:
> Pid: 0, comm: swapper Not tainted 2.6.29-Linus-03321-gbe0ea69 #55
> Call Trace:
> <IRQ> [<ffffffff80260a11>] print_irq_inversion_bug+0x175/0x184
> [<ffffffff80260ab8>] check_usage_forwards+0x98/0xa1
> [<ffffffff8026113b>] mark_lock+0x5dd/0xa2b
> [<ffffffff80262748>] __lock_acquire+0x676/0x1697
> [<ffffffff80262870>] ? __lock_acquire+0x79e/0x1697
> [<ffffffff802637be>] lock_acquire+0x55/0x71
> [<ffffffff802c7073>] ? kill_fasync+0x2c/0x4e
> [<ffffffff8046442b>] _read_lock+0x34/0x41
> [<ffffffff802c7073>] ? kill_fasync+0x2c/0x4e
> [<ffffffff802c7073>] kill_fasync+0x2c/0x4e
> [<ffffffff803cb7bd>] evdev_pass_event+0x6e/0x77
> [<ffffffff803cbb47>] evdev_event+0x74/0xfd
> [<ffffffff803cbad3>] ? evdev_event+0x0/0xfd
> [<ffffffff803c75a5>] input_pass_event+0x64/0xcc
> [<ffffffff803c7541>] ? input_pass_event+0x0/0xcc
> [<ffffffff803c7cdf>] input_handle_event+0x339/0x345
> [<ffffffff803c8bf6>] input_event+0x63/0x7d
> [<ffffffff803d1275>] alps_process_packet+0x3b7/0x4cc
> [<ffffffff803d13e5>] alps_process_byte+0x5b/0x6e
> [<ffffffff803ce797>] psmouse_handle_byte+0x17/0x10d
> [<ffffffff803cf766>] psmouse_interrupt+0x277/0x285
> [<ffffffff803c47a4>] ? serio_interrupt+0x2b/0x83
> [<ffffffff803c47c0>] serio_interrupt+0x47/0x83
> [<ffffffff803c58b7>] i8042_interrupt+0x200/0x215
> [<ffffffff8027b277>] handle_IRQ_event+0x6b/0xa1
> [<ffffffff8027c55c>] handle_edge_irq+0xeb/0x131
> [<ffffffff80464a25>] __irqentry_text_start+0x6d/0xd8
> [<ffffffff8020cd13>] ret_from_intr+0x0/0xf
> <EOI> [<ffffffff8025d068>] ? tick_broadcast_oneshot_control+0x1f/0x102
> [<ffffffff802130e6>] ? default_idle+0x35/0x4f
> [<ffffffff802130e4>] ? default_idle+0x33/0x4f
> [<ffffffff802131ed>] ? c1e_idle+0xda/0x101
> [<ffffffff8025713e>] ? atomic_notifier_call_chain+0xf/0x11
> [<ffffffff8020b1f6>] ? cpu_idle+0x5c/0x9d
> [<ffffffff80451797>] ? rest_init+0x6b/0x6d
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git)
2009-03-27 12:54 ` Bartlomiej Zolnierkiewicz
@ 2009-03-27 18:06 ` Jonathan Corbet
2009-03-27 19:05 ` Larry Finger
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Jonathan Corbet @ 2009-03-27 18:06 UTC (permalink / raw)
To: Bartlomiej Zolnierkiewicz
Cc: Larry Finger, LKML, Christoph Hellwig, Al Viro, Li Zefan,
Wu Fengguang, Ingo Molnar
On Fri, 27 Mar 2009 13:54:35 +0100
Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> wrote:
> I remember looking a bit more closely into the issue and not seeing
> the problem with the locking (though I could have missed something):
>
> file->f_lock is never taken in hard-irq or soft-irq context and in
> the only place where file->f_lock is taken with fasync_lock hold we're
> protected against IRQs by write_lock_irq().
I do think that the warning is spurious at this time.
> [ Despite not being a problem now I think that changing spin_[un]lock()
> to *_irq() variants for file->f_lock could be (given that it really
> fixes the warning) more viable long-term solution than adding special
> lockdep handling (well, it could be that one day file->f_lock is used
> in soft-irq context and then the irq lock inversion issue will become
> a real one) and shouldn't incurr performance penalty since we hold it
> only for a very brief time. ]
We could do that. When I made the change I'd verified that there were
no users in IRQ context, and I couldn't really see why there should
be. I'd rather avoid adding all those IRQ disables if I can avoid it.
How about, instead, just reversing the order of lock acquisition in
fasync_helper()? That would increase the hold time for f_lock, but I
have a hard time seeing that being a real problem. I'm running with
the following now; all seems well. I'll send it up in a bit if nobody
gripes.
Thanks,
jon
diff --git a/fs/fcntl.c b/fs/fcntl.c
index d865ca6..b9c1a4b 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -531,6 +531,7 @@ int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fap
if (!new)
return -ENOMEM;
}
+ spin_lock(&filp->f_lock); /* outside fasync_lock to keep lockdep happy */
write_lock_irq(&fasync_lock);
for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
if (fa->fa_file == filp) {
@@ -555,14 +556,12 @@ int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fap
result = 1;
}
out:
- /* Fix up FASYNC bit while still holding fasync_lock */
- spin_lock(&filp->f_lock);
if (on)
filp->f_flags |= FASYNC;
else
filp->f_flags &= ~FASYNC;
- spin_unlock(&filp->f_lock);
write_unlock_irq(&fasync_lock);
+ spin_unlock(&filp->f_lock);
return result;
}
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git)
2009-03-27 18:06 ` Jonathan Corbet
@ 2009-03-27 19:05 ` Larry Finger
2009-03-28 0:32 ` Peter Zijlstra
2009-03-28 13:36 ` Bartlomiej Zolnierkiewicz
2 siblings, 0 replies; 6+ messages in thread
From: Larry Finger @ 2009-03-27 19:05 UTC (permalink / raw)
To: Jonathan Corbet
Cc: Bartlomiej Zolnierkiewicz, LKML, Christoph Hellwig, Al Viro,
Li Zefan, Wu Fengguang, Ingo Molnar
Jonathan Corbet wrote:
>
> We could do that. When I made the change I'd verified that there were
> no users in IRQ context, and I couldn't really see why there should
> be. I'd rather avoid adding all those IRQ disables if I can avoid it.
>
> How about, instead, just reversing the order of lock acquisition in
> fasync_helper()? That would increase the hold time for f_lock, but I
> have a hard time seeing that being a real problem. I'm running with
> the following now; all seems well. I'll send it up in a bit if nobody
> gripes.
The patch gets rid of the warning for me.
Larry
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git)
2009-03-27 18:06 ` Jonathan Corbet
2009-03-27 19:05 ` Larry Finger
@ 2009-03-28 0:32 ` Peter Zijlstra
2009-03-28 13:36 ` Bartlomiej Zolnierkiewicz
2 siblings, 0 replies; 6+ messages in thread
From: Peter Zijlstra @ 2009-03-28 0:32 UTC (permalink / raw)
To: Jonathan Corbet
Cc: Bartlomiej Zolnierkiewicz, Larry Finger, LKML, Christoph Hellwig,
Al Viro, Li Zefan, Wu Fengguang, Ingo Molnar
On Fri, 2009-03-27 at 12:06 -0600, Jonathan Corbet wrote:
> On Fri, 27 Mar 2009 13:54:35 +0100
> Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> wrote:
>
> > I remember looking a bit more closely into the issue and not seeing
> > the problem with the locking (though I could have missed something):
> >
> > file->f_lock is never taken in hard-irq or soft-irq context and in
> > the only place where file->f_lock is taken with fasync_lock hold we're
> > protected against IRQs by write_lock_irq().
>
> I do think that the warning is spurious at this time.
I think you're right (although at 1:30 am I can't be sure).
It does point to inconsistent (sloppy) lock usage though, because f_lock
is used both with and without irqs disabled -- so on that ground its
correct to complain.
> diff --git a/fs/fcntl.c b/fs/fcntl.c
> index d865ca6..b9c1a4b 100644
> --- a/fs/fcntl.c
> +++ b/fs/fcntl.c
> @@ -531,6 +531,7 @@ int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fap
> if (!new)
> return -ENOMEM;
> }
> + spin_lock(&filp->f_lock); /* outside fasync_lock to keep lockdep happy */
Please don't put in comments like that, they're worse than useless.
Either explain in detail how and why, or don't bother.
> write_lock_irq(&fasync_lock);
> for (fp = fapp; (fa = *fp) != NULL; fp = &fa->fa_next) {
> if (fa->fa_file == filp) {
> @@ -555,14 +556,12 @@ int fasync_helper(int fd, struct file * filp, int on, struct fasync_struct **fap
> result = 1;
> }
> out:
> - /* Fix up FASYNC bit while still holding fasync_lock */
> - spin_lock(&filp->f_lock);
> if (on)
> filp->f_flags |= FASYNC;
> else
> filp->f_flags &= ~FASYNC;
> - spin_unlock(&filp->f_lock);
> write_unlock_irq(&fasync_lock);
> + spin_unlock(&filp->f_lock);
> return result;
> }
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git)
2009-03-27 18:06 ` Jonathan Corbet
2009-03-27 19:05 ` Larry Finger
2009-03-28 0:32 ` Peter Zijlstra
@ 2009-03-28 13:36 ` Bartlomiej Zolnierkiewicz
2 siblings, 0 replies; 6+ messages in thread
From: Bartlomiej Zolnierkiewicz @ 2009-03-28 13:36 UTC (permalink / raw)
To: Jonathan Corbet
Cc: Larry Finger, LKML, Christoph Hellwig, Al Viro, Li Zefan,
Wu Fengguang, Ingo Molnar
On Friday 27 March 2009, Jonathan Corbet wrote:
> On Fri, 27 Mar 2009 13:54:35 +0100
> Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> wrote:
>
> > I remember looking a bit more closely into the issue and not seeing
> > the problem with the locking (though I could have missed something):
> >
> > file->f_lock is never taken in hard-irq or soft-irq context and in
> > the only place where file->f_lock is taken with fasync_lock hold we're
> > protected against IRQs by write_lock_irq().
>
> I do think that the warning is spurious at this time.
>
> > [ Despite not being a problem now I think that changing spin_[un]lock()
> > to *_irq() variants for file->f_lock could be (given that it really
> > fixes the warning) more viable long-term solution than adding special
> > lockdep handling (well, it could be that one day file->f_lock is used
> > in soft-irq context and then the irq lock inversion issue will become
> > a real one) and shouldn't incurr performance penalty since we hold it
> > only for a very brief time. ]
>
> We could do that. When I made the change I'd verified that there were
> no users in IRQ context, and I couldn't really see why there should
> be. I'd rather avoid adding all those IRQ disables if I can avoid it.
>
> How about, instead, just reversing the order of lock acquisition in
> fasync_helper()? That would increase the hold time for f_lock, but I
> have a hard time seeing that being a real problem. I'm running with
> the following now; all seems well. I'll send it up in a bit if nobody
> gripes.
This is even better and works just fine here.
Thanks,
Bart
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2009-03-28 13:48 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-03-27 5:07 Possible IRQ lock inversion from 2.6.29-Linus-03321-gbe0ea69 (2.6.29-git) Larry Finger
2009-03-27 12:54 ` Bartlomiej Zolnierkiewicz
2009-03-27 18:06 ` Jonathan Corbet
2009-03-27 19:05 ` Larry Finger
2009-03-28 0:32 ` Peter Zijlstra
2009-03-28 13:36 ` Bartlomiej Zolnierkiewicz
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox