2.5.44: Strange oopses triggered by pipe_write?

Hi,
I just left my 2.5.44 box unattended for hour and half, and when
I came back, I saw very strange things on screen:

Debug: sleeping function called from illegal context at include/asm/semaphore.h:119

Call Trace:
  [<c011a0f3>] __might_sleep+0x43/0x47
  [..........] pipe_write+0x7f/0x230
  ...........  vfs_write+0xc1/0x160
  ...........  sys_write+0x2a/0x3c
  [<c0107437>] syscall_call+0x7/0xb

bad: scheduling while atomic!

  [<c0117ea2>] schedule+0x2e/0x480
  ............ sys_write+0x33/0x3c
  [<c010745e>] work_resched+0x5/0x16
  
Unable to handle kernel paging request at virtual address 401202b8
 printing eip:
4004cb65
*pde = 10f6a067
*pte = 00000000
Oops: 0004
parport_pc parport tvaudio bttv tuner video-buf videodev i810_audio
 ac97_codec soundcore af_packet nls_cp852 nls_iso8859-2 ipx p8022 psnap llc
 e100
CPU: 0

EIP: 0023:[<4004cb65>] Not tainted
EFLAGS: 00010246
eax: 00000004 ebx: 40131704 ecx: 4012d54c edx: 401320a8
esi: 00001000 edi: 00000000 ebp: bffffc88 esp: bffffc70
ds: 002b es: 002b ss: 002b
Process cat (pid: 26569, threadinfo=c586c000, task=da9ac100)
<0>Kernel panic: Aiee, killing interrupt handler!
In interrupt handler - not syncing

It looks to me like that system first complained a bit about some spinlock being held, panicked in schedule, and then it printed very strange oopses: they look like userspace CPU context.

Kernel is 2.5.44, compiled for SMP, running on machine with 1 CPU. Before this incident machine was up for about 60 hours. After reboot fsck found 31 deleted inodes with zero dtime.
Thanks,
Petr Vandrovec [email protected]