本文共 2736 字，大约阅读时间需要 9 分钟。

环境

• Red Hat Enterprise Linux 5.7 or later
• Red Hat Enterprise Linux 6
• Red Hat Enterprise Linux 7
• XFS file system

问题

• In Red Hat Enterprise Linux 6, an xfs filesystem hung and running xfs_check gave the following error:

# xfs_check /dev/sdm1ERROR: The filesystem has valuable metadata changes in a log which needs tobe replayed. Mount the filesystem to replay the log, and unmount it beforere-running xfs_check. If you are unable to mount the filesystem, then usethe xfs_repair -L option to destroy the log and attempt a repair.Note that destroying the log may cause corruption -- please attempt a mountof the filesystem before doing this.

NOTE: Older xfsprogs packages contained the xfs_check tool. This tool is very slow and does not scale well for large filesystems. As such, it has been deprecated in favor of xfs_repair -n.

• In the above scenario, the filesystem would also not mount:

# mount /dev/sdm1 /new/mount: Structure needs cleaning# mount -f /dev/sdm1 /new//dev/sdm1 100G 49G 46G 52% /new

• The last resort was running xfs_repair -L which was able to cause the filesystem to mount and work.

kernel: XFS internal error XFS_WANT_CORRUPTED_GOTO at line 1572 of file fs/xfs/xfs_alloc.c.  Caller 0xffffffff88871969kernel: kernel: Call Trace:kernel:  [
   
    ] :xfs:xfs_free_ag_extent+0x19e/0x67ekernel:  [
    
     ] :xfs:xfs_free_extent+0xa9/0xc9kernel:  [
     
      ] :xfs:xfs_bmap_finish+0xee/0x15fkernel:  [
      
       ] :xfs:xfs_itruncate_finish+0x185/0x2afkernel:  [
       
        ] :xfs:xfs_inactive+0x1de/0x40fkernel:  [
        
         ] :xfs:xfs_fs_clear_inode+0xa4/0xebkernel: [
         
          ] clear_inode+0xd2/0x123kernel: [
          
           ] generic_delete_inode+0xde/0x143kernel: [
           
            ] do_unlinkat+0xd5/0x141kernel: [
            
             ] tracesys+0x71/0xe0kernel: [
             
              ] tracesys+0xd5/0xe0kernel: kernel: xfs_force_shutdown(sdm1,0x8) called from line 4269 of file fs/xfs/xfs_bmap.c. Return address = 0xffffffff8887a78akernel: Filesystem "sdm1": Corruption of in-memory data detected. Shutting down filesystem: sdm1kernel: Please umount the filesystem, and rectify the problem(s)kernel: nfsd: non-standard errno: 5
             
            
           
          
         
        
       
      
     
    
   

决议

• The remedy is to xfs_repair -L to discard the log and repair the filesystem, which also reconstructs the free space trees on-disk.
• Then, remount the filesystem.

根源

• Corruption has occurred in a free space btree (some unknown time ago), but this has only been detected now when you tried to delete a file and free an extent.

• Also check your storage integrity (though no scsi errors seems to be present in the logs) - this can be caused by all sorts of conditions outside of XFS itself (storage errors, or memory corruption, etc).

诊断步骤

• Examine syslog for the signature messages, as per Issue section: XFS_WANT_CORRUPTED_GOTO at line 1572