lockdep: fix delayacct locking bug
PATCH 1/1 lockdep: fix delayacct locking bug
======================================================
[ INFO: soft-safe -> soft-unsafe lock order detected ]
2.6.19-rc4 #1
------------------------------------------------------
mm_tester/1875 [HC0[0]:SC0[0]:HE0:SE1] is trying to acquire:
(&tsk->delays->lock){--..}, at: [<c0156652>] __delayacct_add_tsk+0x131/0x1d3
and this task is already holding:
(&sighand->siglock){.+..}, at: [<c0156b69>] taskstats_exit_send+0x52/0x3b2
which would create a new lock dependency:
(&sighand->siglock){.+..} -> (&tsk->delays->lock){--..}
but this new dependency connects a soft-irq-safe lock:
(&sighand->siglock){.+..}
... which became soft-irq-safe at:
[<c013d0da>] mark_lock+0x5c/0x38c [<c013de51>] __lock_acquire+0x348/0x971 [<c013ea35>] lock_acquire+0x5c/0x77 [<c032132d>] _spin_lock_irqsave+0x3c/0x4b [<c012fda9>] lock_task_sighand+0x1f/0x3c [<c01309fd>] group_send_sig_info+0x25/0x56 [<c0130a4c>] send_group_sig_info+0x1e/0x30 [<c012a7c6>] it_real_fn+0x1d/0x5f [<c013a453>] hrtimer_run_queues+0xf2/0x14a [<c012e4fa>] run_timer_softirq+0x20/0x166 [<c012b307>] __do_softirq+0x6b/0xe6 [<c0106279>] do_softirq+0x61/0xd1 [<ffffffff>] 0xffffffff
to a soft-irq-unsafe lock:
(&tsk->delays->lock){--..}
... which became soft-irq-unsafe at:
... [<c013d0da>] mark_lock+0x5c/0x38c
[<c013deec>] __lock_acquire+0x3e3/0x971 [<c013ea35>] lock_acquire+0x5c/0x77 [<c032108a>] _spin_lock+0x33/0x3e [<c015642b>] delayacct_end+0x58/0x7b [<c015651e>] __delayacct_blkio_end+0x37/0x3a [<c0157c0f>] sync_page+0x38/0x3b [<c031f920>] __wait_on_bit_lock+0x2a/0x52 [<c0157bc9>] __lock_page+0x5b/0x61 [<c0159bc9>] read_cache_page+0xea/0x13a [<c01abc05>] read_dev_sector+0x2e/0x84 [<c01ad364>] read_lba+0x5c/0xc0 [<c01ad602>] efi_partition+0x8f/0x6ed [<c01abf3f>] rescan_partitions+0x106/0x20b [<c019655b>] do_open+0x2aa/0x3ae [<c01966b4>] blkdev_get+0x55/0x60 [<c01abdd7>] register_disk+0x17c/0x1de [<c01e1810>] add_disk+0x36/0x41 [<c02775a7>] ide_disk_probe+0x8ef/0x918 [<c026b40f>] generic_ide_probe+0x1f/0x20 [<c02586a3>] really_probe+0x39/0xda [<c02588cb>] __driver_attach+0x69/0xa1 [<c0257dbe>] bus_for_each_dev+0x3a/0x5c [<c02585c3>] driver_attach+0x16/0x18 [<c0258092>] bus_add_driver+0x61/0x165 [<c010049b>] init+0x123/0x2d1 [<c0104be7>] kernel_thread_helper+0x7/0x10 [<ffffffff>] 0xffffffff
other info that might help us debug this:
<snip 607 lines of other output>
stack backtrace: [<c0104eb6>] dump_trace+0x69/0x1b6 [<c010501b>] show_trace_log_lvl+0x18/0x2c [<c010562e>] show_trace+0xf/0x11 [<c01056de>] dump_stack+0x15/0x17 [<c013d7f1>] check_usage+0x262/0x26c [<c013e331>] __lock_acquire+0x828/0x971 [<c013ea35>] lock_acquire+0x5c/0x77 [<c032108a>] _spin_lock+0x33/0x3e [<c0156652>] __delayacct_add_tsk+0x131/0x1d3 [<c0156b94>] taskstats_exit_send+0x7d/0x3b2 [<c01294ad>] do_exit+0x1f6/0x7f0 [<c0129b37>] complete_and_exit+0x0/0x13 [<c0103f5b>] syscall_call+0x7/0xb DWARF2 unwinder stuck at syscall_call+0x7/0xb
Leftover inexact backtrace:
=======================
Make the delayacct lock irqsave; this avoids the possible deadlock where an interrupt is taken while holding the delayacct lock which needs to take the delayacct lock.
--- kernel/delayacct.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-)
kernel/delayacct.c +9 -6
@@ -66,6 +66,7 @@ static void delayacct_end(struct timespe {struct timespec ts;s64 ns; +unsigned long flags;do_posix_clock_monotonic_gettime (end); ts =timespec_sub (*end, *start);
@@ -73,10 +74,10 @@ static void delayacct_end(struct timespeif (ns <0 )return ; -spin_lock (¤t->delays->lock); +spin_lock_irqsave (¤t->delays->lock, flags); *total += ns; (*count)++; -spin_unlock (¤t->delays->lock); +spin_unlock_irqrestore (¤t->delays->lock, flags); }void __delayacct_blkio_start (void )
@@ -104,6 +105,7 @@ int __delayacct_add_tsk(struct taskstatss64 tmp;struct timespec ts;unsigned long t1,t2,t3; +unsigned long flags;/* Though tsk->delays accessed later, early exit avoids * unnecessary returning of other data
@@ -136,14 +138,14 @@ int __delayacct_add_tsk(struct taskstats/* zero XXX_total, non-zero XXX_count implies XXX stat overflowed */ -spin_lock (&tsk->delays->lock); +spin_lock_irqsave (&tsk->delays->lock, flags); tmp = d->blkio_delay_total + tsk->delays->blkio_delay; d->blkio_delay_total = (tmp < d->blkio_delay_total) ?0 : tmp; tmp = d->swapin_delay_total + tsk->delays->swapin_delay; d->swapin_delay_total = (tmp < d->swapin_delay_total) ?0 : tmp; d->blkio_count += tsk->delays->blkio_count; d->swapin_count += tsk->delays->swapin_count; -spin_unlock (&tsk->delays->lock); +spin_unlock_irqrestore (&tsk->delays->lock, flags); done:return 0 ;
@@ -152,11 +154,12 @@ done:__u64 __delayacct_blkio_ticks (struct task_struct *tsk) {__u64 ret; +unsigned long flags; -spin_lock (&tsk->delays->lock); +spin_lock_irqsave (&tsk->delays->lock, flags); ret =nsec_to_clock_t (tsk->delays->blkio_delay + tsk->delays->swapin_delay); -spin_unlock (&tsk->delays->lock); +spin_unlock_irqrestore (&tsk->delays->lock, flags);return ret; }
Quoted 8 lines · Peter Zijlstra
======================================================
[ INFO: soft-safe -> soft-unsafe lock order detected ]
2.6.19-rc4 #1
------------------------------------------------------
mm_tester/1875 [HC0[0]:SC0[0]:HE0:SE1] is trying to acquire:
(&tsk->delays->lock){--..}, at: [<c0156652>]__delayacct_add_tsk+0x131/0x1d3
Quoted 4 lines
and this task is already holding:
(&sighand->siglock){.+..}, at: [<c0156b69>]taskstats_exit_send+0x52/0x3b2
which would create a new lock dependency:
(&sighand->siglock){.+..} -> (&tsk->delays->lock){--..}
Thanks.
I introduced this bug in
"[PATCH 5/6] taskstats: kill ->taskstats_lock in favor of ->siglock"
Oleg.