2.4.22 Oops in NFS write

2.4.22 Oops in NFS write

[Hans-Joachim Baader]

[-- Attachment #1: Type: text/plain, Size: 2791 bytes --]

Hi,

while dd'ing a CD to my server, I got the following oops. I had to
reboot because several processes stuck in D state.

System: Dual Athlon MP 2000+, AMD 760 MP chipset, 512 MB ECC RAM,
Intel eepro100 network card with e100 driver. Kernel compiled with
gcc 2.95.4.


Sep 20 13:12:40 kiwi kernel: Unable to handle kernel paging request at virtual address e0000004
Sep 20 13:12:40 kiwi kernel: e88d4070
Sep 20 13:12:40 kiwi kernel: *pde = 00000000
Sep 20 13:12:40 kiwi kernel: Oops: 0000
Sep 20 13:12:40 kiwi kernel: CPU:    1
Sep 20 13:12:40 kiwi kernel: EIP:    0010:[nfs:__insmod_nfs_O/lib/modules/2.4.22/kernel/fs/nfs/nfs.o_M3F5C+-126864/96]    Not tainted
Sep 20 13:12:40 kiwi kernel: EFLAGS: 00010207
Sep 20 13:12:40 kiwi kernel: eax: 02894c59   ebx: cbb59d38   ecx: cbb59d98   edx: cbb59d3c
Sep 20 13:12:40 kiwi kernel: esi: cbb59d50   edi: 470ca3b2   ebp: 00000000   esp: cbb59d00
Sep 20 13:12:40 kiwi kernel: ds: 0018   es: 0018   ss: 0018
Sep 20 13:12:40 kiwi kernel: Process dd (pid: 15449, stackpage=cbb59000)
Sep 20 13:12:40 kiwi kernel: Stack: 00000000 dc1ba178 dc1ba1b4 dc1ba120 cbb59d38 c0000000 cbb58000 00002078 
Sep 20 13:12:40 kiwi kernel:        dc1ba138 ddbba74c 00000000 dc1ba000 dc1ba120 dfae17bc d5254800 00000078 
Sep 20 13:12:40 kiwi kernel:        cd706000 00001000 d8cdc000 00001000 00000001 c1589e00 00000001 c1589e78 
Sep 20 13:12:40 kiwi kernel: Call Trace:    [kmalloc+175/736] [nfs:__insmod_nfs_S.text_L59648+52612/59648] [nfs:__insmod_nfs_O/lib/modules/2.4.22/kernel/fs/nfs/nfs.o_M3F5C+-127238/96] [nfs:__insmod_nfs_O/lib/modules/2.4.22/kernel/fs/nfs/nfs.o_M3F5C+-136045/96] [nfs:__insmod_nfs_O/lib/modules/2.4.22/kernel/fs/nfs/nfs.o_M3F5C+-118394/96]
Sep 20 13:12:40 kiwi kernel: Code: 03 3c c2 40 39 f0 72 f8 8b 5c 24 24 31 c0 f0 0f b3 43 04 57 
Using defaults from ksymoops -t elf32-i386 -a i386


Code;  00000000 Before first symbol
00000000 <_EIP>:
Code;  00000000 Before first symbol
   0:   03 3c c2                  add    (%edx,%eax,8),%edi
Code;  00000003 Before first symbol
   3:   40                        inc    %eax
Code;  00000004 Before first symbol
   4:   39 f0                     cmp    %esi,%eax
Code;  00000006 Before first symbol
   6:   72 f8                     jb     0 <_EIP>
Code;  00000008 Before first symbol
   8:   8b 5c 24 24               mov    0x24(%esp,1),%ebx
Code;  0000000c Before first symbol
   c:   31 c0                     xor    %eax,%eax
Code;  0000000e Before first symbol
   e:   f0 0f b3 43 04            lock btr %eax,0x4(%ebx)
Code;  00000013 Before first symbol
  13:   57                        push   %edi


Regards,
hjb
-- 
Pro-Linux - Germany's largest volunteer Linux support site
http://www.pro-linux.de/          Public Key ID 0x3DDBDDEA

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Attributes returned by NFS Readdirplus not utilized?

[Binny Gill]

Hello Trond,

I am working with the patch to make readdirplus work for NFSv3 client in linux 2.4.18

I noticed that the attributes returned are mostly thrown away as the fattr->timestamp 
that they get during nfs_cached_lookup() is the mtime of the directory (which is normally way in
the past for directories that dont change that often). I was wondering why we cannot store the 
time the readdirplus results were received in the cache and use that as the fattr->timestamp. 

I would like to use '1 readdirplus' to serve a 'ls -la'. Currently it is more like
'1 readdirplus + getattrs for each child'. Am I missing something here?

Regards,
Binny


__________________________________
Do you Yahoo!?
Yahoo! SiteBuilder - Free, easy-to-use web site design software
http://sitebuilder.yahoo.com