Joe Feise wrote on 03/12/06 11:33:

> In certain cases, when programmatically traversing a directory tree mounted
> through NFS, I get a kernel oops.


Update: this doesn't seem to be limited to NFS.
I just got a kernel oops of the same kind, in prune_dcache, while compiling:

BUG: unable to handle kernel paging request at virtual address 0a1d2030
 printing eip:
c016a8b3
*pde = 00000000
Oops: 0000 [#2]
PREEMPT
last sysfs file: /class/net/eth2/ifindex
Modules linked in: pl2303 usbserial softdog cisco_ipsec snd_pcm_oss
snd_mixer_oss snd_cs46xx gameport snd_rawmidi snd_s
eq_device snd_ac97_codec snd_ac97_bus snd_pcm snd_timer snd soundcore
snd_page_alloc zoran i2c_algo_bit videodev saa711
1 i2c_core pegasus arc4 ppp_mppe ppp_deflate ppp_generic slhc usblp
CPU:    0
EIP:    0060:[<c016a8b3>]    Tainted: P      VLI
EFLAGS: 00010202   (2.6.16-rc5-mm3 #9)
EIP is at iput+0x1d/0x65
eax: 0a1d201c   ebx: d4f10a00   ecx: d4f10a18   edx: d4f10a18
esi: d6f38420   edi: 0000002a   ebp: ffffe000   esp: e36c1d48
ds: 007b   es: 007b   ss: 0068
Process qmail-send (pid: 2678, threadinfo=e36c0000 task=e353cab0)
Stack: <0>d6f38414 c0167e41 e36c0000 e36c0000 00000082 000049d4 c17deac0 c0168230
       c013c911 00000000 000225da 00000000 00127500 00000000 000329cf 000049d4
       00000005 00000000 00000000 000000d0 00000020 e36c1e14 c05a51f0 0000000c
Call Trace:
 <c0167e41> prune_dcache+0x12d/0x162   <c0168230> shrink_dcache_memory+0x14/0x37
 <c013c911> shrink_slab+0x131/0x1f7   <c013dab9> try_to_free_pages+0xee/0x1ab
 <c0138ea8> __alloc_pages+0x113/0x29c   <c014d344> kmem_getpages+0x2f/0x95
 <c014dff8> cache_grow+0xc4/0x171   <c015f385> link_path_walk+0x69/0xc0
 <c014e20c> cache_alloc_refill+0x167/0x22c   <c014e4cb> kmem_cache_alloc+0x3f/0x41
 <c015230a> get_empty_filp+0x41/0x11c   <c015f778> do_path_lookup+0x102/0x1fb
 <c015f8ab> __path_lookup_intent_open+0x22/0x90   <c015f93c>
path_lookup_open+0x23/0x2b
 <c016010d> open_namei+0x6a/0x68f   <c0164938> fifo_open+0x0/0x2a0