freebsd crash under I/O - got error messages part 4

[John Baldwin]

On Sunday, December 07, 2014 3:57:25 pm Wojciech Puchar wrote:
> even without UFS_DIRHASH there are still errors under heavier I/O load.
> 
> 
> 
> lock order reversal:
>   1st 0xfffff800182c9068 ufs (ufs) @ kern/vfs_subr.c:2137
>   2nd 0xfffffe00612443d0 bufwait (bufwait) @ ufs/ffs/ffs_vnops.c:262
>   3rd 0xfffff8004a877418 ufs (ufs) @ kern/vfs_subr.c:2137
> KDB: stack backtrace:
> db_trace_self_wrapper() at db_trace_self_wrapper+0x2b/frame 
> 0xfffffe007870ae20
> kdb_backtrace() at kdb_backtrace+0x39/frame 0xfffffe007870aed0
> witness_checkorder() at witness_checkorder+0xdc2/frame 0xfffffe007870af60
> __lockmgr_args() at __lockmgr_args+0x9ea/frame 0xfffffe007870b0a0
> ffs_lock() at ffs_lock+0x84/frame 0xfffffe007870b0f0
> VOP_LOCK1_APV() at VOP_LOCK1_APV+0xd9/frame 0xfffffe007870b120
> _vn_lock() at _vn_lock+0xaa/frame 0xfffffe007870b190
> vget() at vget+0x67/frame 0xfffffe007870b1d0
> vfs_hash_get() at vfs_hash_get+0xe1/frame 0xfffffe007870b220
> ffs_vgetf() at ffs_vgetf+0x40/frame 0xfffffe007870b2b0
> softdep_sync_buf() at softdep_sync_buf+0x3b3/frame 0xfffffe007870b390
> ffs_syncvnode() at ffs_syncvnode+0x286/frame 0xfffffe007870b410
> ffs_truncate() at ffs_truncate+0x614/frame 0xfffffe007870b600
> ufs_direnter() at ufs_direnter+0x722/frame 0xfffffe007870b6c0
> ufs_mkdir() at ufs_mkdir+0x4d0/frame 0xfffffe007870b850
> VOP_MKDIR_APV() at VOP_MKDIR_APV+0xd1/frame 0xfffffe007870b880
> kern_mkdirat() at kern_mkdirat+0x1be/frame 0xfffffe007870baa0
> amd64_syscall() at amd64_syscall+0x216/frame 0xfffffe007870bbb0
> Xfast_syscall() at Xfast_syscall+0xfb/frame 0xfffffe007870bbb0
> --- syscall (136, FreeBSD ELF64, sys_mkdir), rip = 0x800c0e93a, rsp = 
> 0x7fffffffd3f8, rbp = 0x7fffffffd520 ---

The LORs are known false positives (certainly the dirhash one is and is
documented as such in the source).  They are unrelated to the hard I/O
errors you had in your first mail.  I'm not sure how to track those down
further however. :(

-- 
John Baldwin