[linux-cifs-client] fsx-linux failing with latest cifs-2.6 git tree

[Jeff Layton]

I've found a regression, but haven't started tracking it down yet. The fsx-linux program in the LTP suite is now failing with latest cifs-2.6 git tree:

# uname -r
2.6.28-rc5
# /usr/src/ltp-full-20081031/testcases/bin/fsx-linux /mnt/cifs/testfile
truncating to largest ever: 0x13e76
truncating to largest ever: 0x2e52c
READ BAD DATA: offset = 0x2335c, size = 0xb54e
OFFSET	GOOD	BAD	RANGE
0x2e327	000000	0x0703	  0x107
operation# (mod 256) for the bad datamay be 3
LOG DUMP (17 total operations):
1: 1227282708.080905 TRUNCATE UP	from 0x0 to 0x13e76
2: 1227282708.082120 WRITE    0x17098 thru 0x26857 (0xf7c0 bytes) HOLE
3: 1227282708.096950 READ     0xc73e thru 0x1b801 (0xf0c4 bytes)
4: 1227282708.105306 MAPWRITE 0x32e00 thru 0x331fc (0x3fd bytes)
5: 1227282708.122103 MAPWRITE 0x7ac1 thru 0x11029 (0x9569 bytes)
6: 1227282708.126967 READ     0x1f62e thru 0x2177f (0x2152 bytes)
7: 1227282708.127011 WRITE    0x756 thru 0xede (0x789 bytes)
8: 1227282708.127934 READ     0x18f13 thru 0x27d18 (0xee06 bytes)
9: 1227282708.128163 READ     0xf369 thru 0x1b0af (0xbd47 bytes)
10: 1227282708.128337 MAPWRITE 0x17461 thru 0x19892 (0x2432 bytes)
11: 1227282708.128488 TRUNCATE DOWN	from 0x331fd to 0x2e52c
12: 1227282708.141938 READ     0x23922 thru 0x2e52b (0xac0a bytes)	***RRRR***
13: 1227282708.142138 MAPREAD  0x1fddb thru 0x283d7 (0x85fd bytes)
14: 1227282708.142423 MAPREAD  0x6741 thru 0xf6cb (0x8f8b bytes)
15: 1227282708.142954 TRUNCATE DOWN	from 0x2e52c to 0x2858c	******WWWW
16: 1227282708.143876 WRITE    0x2e42f thru 0x3bdce (0xd9a0 bytes) HOLE	***WWWW
17: 1227282708.156140 READ     0x2335c thru 0x2e8a9 (0xb54e bytes)	***RRRR***
Correct content saved for comparison
(maybe hexdump "/mnt/cifs/testfile" vs "/mnt/cifs/testfile.fsxgood")


...this Fedora rawhide kernel, however seems to be fine:

kernel-debug-2.6.27.5-117.fc10.x86_64

...the funny thing is that I spent time yesterday backporting most of the more recent patches to RHEL4/5. Those kernels are also not failing. It's possible that there's something wrong with the build I've done. Can you try replicating this when you get the chance?

Thanks,
-- 
Jeff Layton <jlayton at redhat.com>