Steve Lord <lord@xxxxxxx> writes:
> I cannot speak for 2.95.2,
I can.
The kernel, compiled from CVS 2001-03-04, compiled nicely. Basic testing
( 42 processes doing "cp /tmp/1 /tmp/2" and "sync") for just a few
seconds resulted in the attached crash. With egcs-2.91.66 this kernel
worked flawlessly.
kernel BUG at buffer.c:188!
invalid operand: 0000
CPU: 0
EIP: 0010:[<c0134d53>]
EFLAGS: 00010282
eax: 0000001c ebx: c2088790 ecx: 00000001 edx: 00000001
esi: 00000000 edi: 00000000 ebp: 00000000 esp: f74ebf6c
ds: 0018 es: 0018 ss: 0018
Process sync (pid: 7963, stackpage=f74eb000)
Stack: c0242405 c0242669 000000bc f7549e40 f7549e40 c0134e58 f7549e40 f74ea000
00000000 00000001 bffffb44 00000000 00000000 f7549e40 c0134f7f 00000000
00000000 f74ea000 00000001 c0134ffb 00000000 c0108f4b bffffbac 00000002
Call Trace: [<c0134e58>] [<c0134f7f>] [<c0134ffb>] [<c0108f4b>]
Code: 0f 0b 83 c4 0c 5b 5e c3 90 83 ec 0c 55 57 56 53 8b 7c 24 20
>>EIP; c0134d53 <_write_buffer+df/e8> <=====
Trace; c0134e58 <sync_buffers+fc/1f0>
Trace; c0134f7f <fsync_dev+f/84>
Trace; c0134ffb <sys_sync+7/10>
Trace; c0108f4b <system_call+33/38>
Code; c0134d53 <_write_buffer+df/e8>
00000000 <_EIP>:
Code; c0134d53 <_write_buffer+df/e8> <=====
0: 0f 0b ud2a <=====
Code; c0134d55 <_write_buffer+e1/e8>
2: 83 c4 0c add $0xc,%esp
Code; c0134d58 <_write_buffer+e4/e8>
5: 5b pop %ebx
Code; c0134d59 <_write_buffer+e5/e8>
6: 5e pop %esi
Code; c0134d5a <_write_buffer+e6/e8>
7: c3 ret
Code; c0134d5b <_write_buffer+e7/e8>
8: 90 nop
Code; c0134d5c <sync_buffers+0/1f0>
9: 83 ec 0c sub $0xc,%esp
Code; c0134d5f <sync_buffers+3/1f0>
c: 55 push %ebp
Code; c0134d60 <sync_buffers+4/1f0>
d: 57 push %edi
Code; c0134d61 <sync_buffers+5/1f0>
e: 56 push %esi
Code; c0134d62 <sync_buffers+6/1f0>
f: 53 push %ebx
Code; c0134d63 <sync_buffers+7/1f0>
10: 8b 7c 24 20 mov 0x20(%esp,1),%edi
--
Per Andreas Buer
|