Skip to contentSkip to conversation

LTP NFS data corruption in 2.6.15-git9

DISCUSSION Andi Kleen 1 participant 2 messages Jan 13, 2006

Subsystems (2)
lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

FWIW - i was just running LTP- for 2.6.15-git9 on a x86-64 nfs root machine talking to 2.6.15 machine with NFSv3/UDP. During the IO tests I got this. Export was with rw,async.

The network was 100 GBit/s so I wouldn't assume it was caused by fragment reordering [Why UDP? That is what the kernel nfs root mounter selected]

To reproduce just run LTP 20051103 runltp over NFS.

-Andi

doio(rwtest03) (21516) 04:19:53
---------------------
*** DATA COMPARISON ERROR *** check_file(/tmp/ltp-16542/mm-buff-21500, 6133913, 37155, B:21516:linux:doio*, 19 , 0) failed

Comparison fd is 3, with open flags 0
Corrupt regions follow - unprintable chars are represented as '.'
-----------------------------------------------------------------
corrupt bytes starting at file offset 6160384

    1st 32 expected bytes:  516:linux:doio*B:21516:linux:doi
    1st 32 actual bytes:    ................................

Request number 12

          fd 4 is file /tmp/ltp-16542/mm-buff-21500 - open flags are 02 O_RDWR,
          write done at file offset 6133913 - pattern is B (0102)
          number of requests is 1, strides per request is 1
          i/o byte count = 37155
          memory alignment is unaligned

syscall: mmap-write(NULL, 12800000, PROT_WRITE, MAP_SHARED, 4, 0)

        file is mmaped to: 0xf71f2000
        file-mem=0xf77cb899, length=37155, buffer=0x8073368

doio(rwtest03) (21514) 04:19:53
---------------------
(parent) pid 21516 exited because of data compare errors

rwtest(rwtest03) : doio reported errors (r=4)
rwtest03    1  FAIL  :  doio reported errors (r=4)
rwtest03    1  FAIL  :  Test failed

doio(rwtest04) (21540) 04:21:03
---------------------
*** DATA COMPARISON ERROR *** check_file(/tmp/ltp-16542/mm-sync-21525, 12522037, 83920, Q:21540:linux:doio*, 1 9, 0) failed

Comparison fd is 5, with open flags 0
Corrupt regions follow - unprintable chars are represented as '.'
-----------------------------------------------------------------
corrupt bytes starting at file offset 12582913

    1st 32 expected bytes:  Q:21540:linux:doio*Q:21540:linux
    1st 32 actual bytes:    B:21541:linux:doio*B:21541:linux

Request number 126

          fd 4 is file /tmp/ltp-16542/mm-sync-21525 - open flags are 010002 O_RD

WR,O_SYNC,
write done at file offset 12522037 - pattern is Q (0121) number of requests is 1, strides per request is 1 i/o byte count = 83920
memory alignment is unaligned

syscall: mmap-write(NULL, 12800000, PROT_WRITE, MAP_SHARED, 4, 0)

        file is mmaped to: 0xf7257000
        file-mem=0xf7e48235, length=83920, buffer=0x8073369

doio(rwtest04) (21539) 04:21:03
---------------------
(parent) pid 21540 exited because of data compare errors

-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs

author
#1 lore raw

Re: LTP NFS data corruption in 2.6.15-git9 II

Quoted 10 lines · Andi Kleen

FWIW - i was just running LTP- for 2.6.15-git9 on a x86-64 nfs root machine talking to 2.6.15 machine with NFSv3/UDP. During the IO tests I got this. Export was with rw,async.

The network was 100 GBit/s so I wouldn't assume it was caused by fragment reordering [Why UDP? That is what the kernel nfs root mounter selected]

To reproduce just run LTP 20051103 runltp over NFS.

As a followup the test machine got a lot of

do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!
do_vfs_lock: VFS is out of sync with lock manager!

messages after the test.

-Andi

-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files for problems? Stop! Download the new AJAX search engine that makes searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs