CentOS 6.4在运行XFS时系统crash的bug分析

最近有一台CentOS 6.4的服务器发生多次crash，kernel version 是Linux 2.6.32-431.29.2.el6.x86_64。从vmcore-dmesg日志内容及crash工具解析vmcore文件来看，与已知bug比较匹配。

【分析过程】

1、查看vmcore-dmesg日志

<3>CIFS VFS: Error -11 sending data on socket to server

<4>NOHZ: local_softirq_pending 100

<3>CIFS VFS: Error connecting to socket. Aborting operation

<3>CIFS VFS: cifs_mount failed w/return code = -512

<3>BUG: scheduling while atomic: mysqld/23447/0xffff8801

<4>Modules linked in: des_generic ecb md4 nls_utf8 cifs xfs exportfs tcp_diag inet_diag bluetooth rfkill sunrpc bonding 8021q garp stp llc ipv6 microcode power_meter iTCO_wdt iTCO_vendor_support i2c_i801 lpc_ich mfd_core igb dca i2c_algo_bit i2c_core ptp pps_core sg ses enclosure ext4 jbd2 mbcache sd_mod crc_t10dif ahci megaraid_sas(U) dm_mirror dm_region_hash dm_log dm_mod [last unloaded: speedstep_lib]

<4>Pid: 23447, comm: mysqld Not tainted 2.6.32-431.29.2.el6.x86_64 #1

<4>Call Trace:

<4> [<ffffffff8105da66>] ? __schedule_bug+0x66/0x70

<4> [<ffffffff81529440>] ? thread_return+0x640/0x770

<4> [<ffffffffa000439c>] ? dm_table_unplug_all+0x5c/0x100 [dm_mod]

<4> [<ffffffff810a6d31>] ? ktime_get_ts+0xb1/0xf0

<4> [<ffffffff815295e3>] ? io_schedule+0x73/0xc0

<4> [<ffffffff811c933d>] ? __blockdev_direct_IO_newtrunc+0xb7d/0x1270

<4> [<ffffffff8116e2da>] ? kmem_getpages+0xba/0x170

<4> [<ffffffff81142029>] ? zone_statistics+0x99/0xc0

<4> [<ffffffffa0352180>] ? xfs_get_blocks_direct+0x0/0x20 [xfs]

<4> [<ffffffff81121eb5>] ? mempool_alloc_slab+0x15/0x20

<4> [<ffffffff811c9aa7>] ? __blockdev_direct_IO+0x77/0xe0

<4> [<ffffffffa0352180>] ? xfs_get_blocks_direct+0x0/0x20 [xfs]

<4> [<ffffffffa0351b00>] ? xfs_end_io_direct_write+0x0/0x80 [xfs]

<4> [<ffffffffa03518dc>] ? xfs_vm_direct_IO+0xec/0x170 [xfs]

<4> [<ffffffffa0352180>] ? xfs_get_blocks_direct+0x0/0x20 [xfs]

<4> [<ffffffffa0351b00>] ? xfs_end_io_direct_write+0x0/0x80 [xfs]

<4> [<ffffffff81226cd6>] ? security_inode_need_killpriv+0x16/0x20

<4> [<ffffffff81120322>] ? generic_file_direct_write+0xc2/0x190

<4> [<ffffffffa03583ba>] ? xfs_file_dio_aio_write+0x12a/0x280 [xfs]

<4> [<ffffffffa035883f>] ? xfs_file_aio_write+0x17f/0x190 [xfs]

<4> [<ffffffff81188b4a>] ? do_sync_write+0xfa/0x140

<4> [<ffffffff8109afa0>] ? autoremove_wake_function+0x0/0x40

<4> [<ffffffff811a7d74>] ? notify_change+0x234/0x340

<4> [<ffffffff81226d36>] ? security_file_permission+0x16/0x20

<4> [<ffffffff81188e48>] ? vfs_write+0xb8/0x1a0

<4> [<ffffffff811898d2>] ? sys_pwrite64+0x82/0xa0

<1>BUG: unable to handle kernel paging request at 00000001a1d9d1a0

2、用crash工具解析vmcore文件，可以看到系统在执行task_tick_fair调用update_curr函数时发生异常。

关于update_curr函数的描述可参考：

https://blog.csdn.net/kickxxx/article/details/9704575

https://blog.csdn.net/arm7star/article/details/78648718

crash> bt

PID: 23447  TASK: ffff881063e38040  CPU: 10  COMMAND: "mysqld"

#0 [ffff8810b8803980] machine_kexec at ffffffff81038f3b

#1 [ffff8810b88039e0] crash_kexec at ffffffff810c5af2

#2 [ffff8810b8803ab0] oops_end at ffffffff8152ca50

#3 [ffff8810b8803ae0] no_context at ffffffff8104a00b

#4 [ffff8810b8803b30] __bad_area_nosemaphore at ffffffff8104a295

#5 [ffff8810b8803b80] bad_area_nosemaphore at ffffffff8104a363

#6 [ffff8810b8803b90] __do_page_fault at ffffffff8104aabf

#7 [ffff8810b8803cb0] do_page_fault at ffffffff8152e99e

#8 [ffff8810b8803ce0] page_fault at ffffffff8152bd55

[exception RIP: update_curr+324]

RIP: ffffffff8105a6e4  RSP: ffff8810b8803d98  RFLAGS: 00010082

RAX: ffff881063e38040  RBX: 0000000044034048  RCX: ffff88107fef3140

RDX: 0000000000019410  RSI: 0000000000000000  RDI: ffff881063e38078

RBP: ffff8810b8803dc8   R8: ffffffff8160c325   R9: 0000000000000421

R10: 0000000000000010  R11: 0000000000000421  R12: ffff8810b88168a8

R13: 00000000000d8be5  R14: 0000000000000001  R15: ffff881063e38040

ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018

#9 [ffff8810b8803dd0] task_tick_fair at ffffffff8105ab2b

#10 [ffff8810b8803e00] scheduler_tick at ffffffff8105dc21

#11 [ffff8810b8803e40] update_process_times at ffffffff8108404e

#12 [ffff8810b8803e70] tick_sched_timer at ffffffff810acad6

#13 [ffff8810b8803ea0] __run_hrtimer at ffffffff8109f6be

#14 [ffff8810b8803ef0] hrtimer_interrupt at ffffffff8109fa26

#15 [ffff8810b8803f70] local_apic_timer_interrupt at ffffffff81031f1d

#16 [ffff8810b8803f90] smp_apic_timer_interrupt at ffffffff815325e5

#17 [ffff8810b8803fb0] apic_timer_interrupt at ffffffff8100bb93

--- <IRQ stack> ---

#18 [ffff880144035528] apic_timer_interrupt at ffffffff8100bb93

[exception RIP: vprintk+593]

RIP: ffffffff81072a91  RSP: ffff8801440355d8  RFLAGS: 00000246

RAX: 00000000000114c0  RBX: ffff880144035668  RCX: 0000000000007c8c

RDX: ffff8810b8800000  RSI: 0000000000000046  RDI: 0000000000000246

RBP: ffffffff8100bb8e   R8: 000000000001bdf6   R9: 00000000fffffffb

R10: 0000000000000001  R11: 000000000000000a  R12: 81eb2000812991a9

R13: 0000000000000001  R14: ffffffffa003f1fe  R15: ffffffffffffffff

ORIG_RAX: ffffffffffffff10  CS: 0010  SS: 0018

#19 [ffff880144035670] printk at ffffffff81528845

#20 [ffff8801440356d0] printk_address at ffffffff81010da1

#21 [ffff8801440356e0] print_trace_address at ffffffff81010f2c

#22 [ffff880144035710] print_context_stack at ffffffff81011141

#23 [ffff880144035780] dump_trace at ffffffff81010110

#24 [ffff880144035820] show_trace_log_lvl at ffffffff81010eb5

#25 [ffff880144035850] show_trace at ffffffff81010ee5

#26 [ffff880144035860] dump_stack at ffffffff8152868e

#27 [ffff8801440358a0] __schedule_bug at ffffffff8105da66

...skipping...

#10 [ffff8810b8803e00] scheduler_tick at ffffffff8105dc21

#11 [ffff8810b8803e40] update_process_times at ffffffff8108404e

#12 [ffff8810b8803e70] tick_sched_timer at ffffffff810acad6

#13 [ffff8810b8803ea0] __run_hrtimer at ffffffff8109f6be

#14 [ffff8810b8803ef0] hrtimer_interrupt at ffffffff8109fa26

#15 [ffff8810b8803f70] local_apic_timer_interrupt at ffffffff81031f1d

#16 [ffff8810b8803f90] smp_apic_timer_interrupt at ffffffff815325e5

#17 [ffff8810b8803fb0] apic_timer_interrupt at ffffffff8100bb93

--- <IRQ stack> ---

#18 [ffff880144035528] apic_timer_interrupt at ffffffff8100bb93

[exception RIP: vprintk+593]

RIP: ffffffff81072a91  RSP: ffff8801440355d8  RFLAGS: 00000246

RAX: 00000000000114c0  RBX: ffff880144035668  RCX: 0000000000007c8c

RDX: ffff8810b8800000  RSI: 0000000000000046  RDI: 0000000000000246

RBP: ffffffff8100bb8e   R8: 000000000001bdf6   R9: 00000000fffffffb

R10: 0000000000000001  R11: 000000000000000a  R12: 81eb2000812991a9

R13: 0000000000000001  R14: ffffffffa003f1fe  R15: ffffffffffffffff

ORIG_RAX: ffffffffffffff10  CS: 0010  SS: 0018

#19 [ffff880144035670] printk at ffffffff81528845

#20 [ffff8801440356d0] printk_address at ffffffff81010da1

#21 [ffff8801440356e0] print_trace_address at ffffffff81010f2c

#22 [ffff880144035710] print_context_stack at ffffffff81011141

#23 [ffff880144035780] dump_trace at ffffffff81010110

#24 [ffff880144035820] show_trace_log_lvl at ffffffff81010eb5

#25 [ffff880144035850] show_trace at ffffffff81010ee5

#26 [ffff880144035860] dump_stack at ffffffff8152868e

#27 [ffff8801440358a0] __schedule_bug at ffffffff8105da66

#28 [ffff8801440358c0] thread_return at ffffffff81529440

#29 [ffff880144035980] io_schedule at ffffffff815295e3

#30 [ffff8801440359a0] __blockdev_direct_IO_newtrunc at ffffffff811c933d

#31 [ffff880144035b50] __blockdev_direct_IO at ffffffff811c9aa7

#32 [ffff880144035bd0] xfs_vm_direct_IO at ffffffffa03518dc [xfs]

#33 [ffff880144035c60] generic_file_direct_write at ffffffff81120322

#34 [ffff880144035cd0] xfs_file_dio_aio_write at ffffffffa03583ba [xfs]

#35 [ffff880144035d60] xfs_file_aio_write at ffffffffa035883f [xfs]

#36 [ffff880144035dc0] do_sync_write at ffffffff81188b4a

#37 [ffff880144035ef0] vfs_write at ffffffff81188e48

#38 [ffff880144035f30] sys_pwrite64 at ffffffff811898d2

#39 [ffff880144035f80] system_call_fastpath at ffffffff8100b072

RIP: 00000038fe00f1a3  RSP: 00007f0bd85c3a48  RFLAGS: 00010206

RAX: 0000000000000012  RBX: ffffffff8100b072  RCX: 0000000000000000

RDX: 0000000000100000  RSI: 00007f0b60c40000  RDI: 0000000000000479

RBP: 00007f0bd85c39a0   R8: 0000000000100000   R9: 000000007aa00000

R10: 000000007aa00000  R11: 0000000000000293  R12: 0000000001350c40

R13: 00007f0bad491b98  R14: 000000007aa00000  R15: 00007f0bad491b98

ORIG_RAX: 0000000000000012  CS: 0033  SS: 002b

3、这个问题官方有详细的解决方案描述，Thread overran stack, or stack corrupted while running XFS，可参考
https://access.redhat.com/solutions/54544

【结论】

CentOS6.4 Linux 2.6.32-431.29.2.el6.x86_64这个版本之前有两个相关的bug，693280和918359，如下

https://gist.githubusercontent.com/namikawa/8677d8742994e3c19ee1/raw/06da129ab4483c0c3ede65319c784b5698e49a1c/CentOS_6.5_kernel_CHANGELOG

问题在kernel-2.6.32-431.46.1.el6后修复，建议升级到CentOS 7.4后可以解决。