linux-stable 3.11.8 oopses
Greetings,
This is more of just a report than anything else. I run linux-stable, and while btrfs in 3.11.7 seemed fine, I've had oopses ever since installating 3.11.8. The server is not under heavy load when it happens, and it happens fairly shortly after a clean boot.
The oopses all look like the following; here's a representative couple.
INFO: task btrfs-transacti:2103 blocked for more than 120 seconds.
btrfs-transacti D c1089143 0 2103 2 0x00000000d2391e80 00000046 d2391e60 c1089143 00000002 f74a1980 c198a100 9204095a 000002b7 c198a100 f7bd7100 efbfa640 f74a1980 d2391e58 c10113a7 f7bd7148 00000000 00000002 efbfa960 d2391e60 c107a391 00000000 d1009a80 d1009bd0
Call Trace: [<c1089143>] ? idle_balance+0x2a3/0x3b0 [<c10113a7>] ? __switch_to+0xc7/0x340 [<c107a391>] ? finish_task_switch+0x41/0xc0 [<c114c616>] ? kmem_cache_alloc+0xe6/0x100 [<f865d050>] ? btrfs_free_path+0x20/0x30 [btrfs] [<c15e7e13>] schedule+0x23/0x60 [<c15e64b5>] schedule_timeout+0x1d5/0x240 [<c15e8b1d>] ? _raw_spin_lock+0xd/0x10 [<f869dc40>] ? btrfs_run_ordered_operations+0x270/0x2b0 [btrfs] [<c103e628>] ? default_spin_lock_flags+0x8/0x10 [<c15e8cbf>] ? _raw_spin_lock_irqsave+0x2f/0x50 [<c106d418>] ? prepare_to_wait+0x48/0x70 [<f868418c>] btrfs_commit_transaction+0x34c/0xa50 [btrfs] [<f8684cc4>] ? start_transaction+0x84/0x4a0 [btrfs] [<c105aaa0>] ? cascade+0x130/0x130 [<c106d1b0>] ? wake_up_bit+0x30/0x30 [<f867d0d9>] transaction_kthread+0x169/0x1f0 [btrfs] [<f867cf70>] ? btrfs_destroy_marked_extents.clone.96.clone.119+0x1e0/0x1e0 [btrfs] [<c106ca0c>] kthread+0x9c/0xa0 [<c1070000>] ? run_posix_cpu_timers+0x360/0x820 [<c15efef7>] ret_from_kernel_thread+0x1b/0x28 [<c106c970>] ? flush_kthread_worker+0xb0/0xb0 INFO: task btrfs-transacti:2103 blocked for more than 120 seconds.
btrfs-transacti D c1089143 0 2103 2 0x00000000d2391e80 00000046 d2391e60 c1089143 00000002 f74a1980 c198a100 9204095a 000002b7 c198a100 f7bd7100 efbfa640 f74a1980 d2391e58 c10113a7 f7bd7148 00000000 00000002 efbfa960 d2391e60 c107a391 00000000 d1009a80 d1009bd0
Call Trace: [<c1089143>] ? idle_balance+0x2a3/0x3b0 [<c10113a7>] ? __switch_to+0xc7/0x340 [<c107a391>] ? finish_task_switch+0x41/0xc0 [<c114c616>] ? kmem_cache_alloc+0xe6/0x100 [<f865d050>] ? btrfs_free_path+0x20/0x30 [btrfs] [<c15e7e13>] schedule+0x23/0x60 [<c15e64b5>] schedule_timeout+0x1d5/0x240 [<c15e8b1d>] ? _raw_spin_lock+0xd/0x10 [<f869dc40>] ? btrfs_run_ordered_operations+0x270/0x2b0 [btrfs] [<c103e628>] ? default_spin_lock_flags+0x8/0x10 [<c15e8cbf>] ? _raw_spin_lock_irqsave+0x2f/0x50 [<c106d418>] ? prepare_to_wait+0x48/0x70 [<f868418c>] btrfs_commit_transaction+0x34c/0xa50 [btrfs] [<f8684cc4>] ? start_transaction+0x84/0x4a0 [btrfs] [<c105aaa0>] ? cascade+0x130/0x130 [<c106d1b0>] ? wake_up_bit+0x30/0x30 [<f867d0d9>] transaction_kthread+0x169/0x1f0 [btrfs] [<f867cf70>] ? btrfs_destroy_marked_extents.clone.96.clone.119+0x1e0/0x1e0 [btrfs] [<c106ca0c>] kthread+0x9c/0xa0 [<c1070000>] ? run_posix_cpu_timers+0x360/0x820 [<c15efef7>] ret_from_kernel_thread+0x1b/0x28 [<c106c970>] ? flush_kthread_worker+0xb0/0xb0
This is on x86_64, server-class hardware (ECC memory, etc). I ran a scrub overnight that reported no errors, and I'll do it again now that I've reverted to 3.11.7.
Is there any other info that might be useful to the btrfs devs about this?
Charles
Signature · 4 lines
----------------------------------------------------------------------- Charles Cazabon GPL'ed software available at: http://pyropus.ca/software/ -----------------------------------------------------------------------
Quoted 6 lines · Charles Cazabon
This is on
x86_64, server-class hardware (ECC memory, etc). I ran a scrub overnight that reported no errors, and I'll do it again now that I've reverted to 3.11.7.Is there any other info that might be useful to the btrfs devs about this?
sysrq+w would dump the blockage to kernel messages. Depending on your kernel you may need to first enable sysrq functions with
echo 1 > /proc/sys/kernel/sysrq
then
echo 2 > /proc/sysrq-trigger
then
dmesg
Chris Murphy
Quoted 9 lines · Chris Murphy
Quoted 6 lines · Charles Cazabon
This is on
x86_64, server-class hardware (ECC memory, etc). I ran a scrub overnight that reported no errors, and I'll do it again now that I've reverted to 3.11.7.Is there any other info that might be useful to the btrfs devs about this?
Quoted 7 lines
sysrq+w would dump the blockage to kernel messages. Depending on your kernel you may need to first enable sysrq functions with
echo 1 >
/proc/sys/kernel/sysrqthen
echo 2 >
/proc/sysrq-trigger
Cute. That 2 is supposed to be a w, hence sysrq+w.
Chris Murphy