[linux-cifs-client] recent regression in mainline CIFS code

[Steve French]

On Wed, Apr 29, 2009 at 10:27 AM, Jeff Layton <jlayton at redhat.com> wrote:
> On Wed, 29 Apr 2009 10:20:31 -0400
> Jeff Layton <jlayton at redhat.com> wrote:
>
>> I've just started seeing a recent regression in the mainline cifs code
>> when doing the general cthon04 tests. The small compile test does this:
>>
>> $ gcc -w -o stat stat.c -lm
>> /usr/bin/ld: final link failed: Bad file descriptor
>> collect2: ld returned 1 exit status
>>
>> Unix extensions are enabled here. I've not tested this w/o them yet.
>> stracing shows this:
>>
>> 4866  open("stat", O_RDWR)              = 13
>> 4866  fcntl(13, F_GETFD)                = 0
>> 4866  fcntl(13, F_SETFD, FD_CLOEXEC)    = 0
>> 4866  fstat(13, {st_mode=S_IFREG|0644, st_size=0, ...}) = 0
>> 4866  mmap(NULL, 16384, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fd4e7075000
>> 4866  lseek(13, 0, SEEK_SET)            = 0
>> 4866  read(13, 0x7fd4e7075000, 64)      = -1 EBADF (Bad file descriptor)
>> 4866  lseek(13, 64, SEEK_CUR)           = 64
>> 4866  write(13, "\6\0\0\0\5\0\0\0@\0\0\0\0\0\0\0@\0@\0\0\0\0\0@\0@\0\0\0\0\0\300"..., 448) = -1 EBADF (Bad file descriptor)
>>
>> I ran these tests a month or two ago and didn't see this, so I think
>> it's a recent regression.
>>
>
> On a hunch, I backed out these commits that added the lookup intent
> open patches:
>
> a6ce4932fbdbcd8f8e8c6df76812014351c32892
> bc8cd4390c9129fbd286b10fa99972dfb68cd069
> 88dd47fff4891545bfcfdf39146dde8380771766
>
> ...and that seems to have fixed the problem. Steve, it might be best to
> back out those patches for now

I think backing them out will cause more git problems (corrupt trees),
than simply giving Shirish some time to fix it - we still have about a
month or so until 2.6.30 goes out.  If we can't fix it by rc6 we need
to back it out.

Shirish,
If you get stuck debugging this, let me know ASAP

-- 
Thanks,

Steve