Hello,

  I tried again with 2.4.8-ac4, figured it would better than ac2 ...

  I get the same trap (see attached ksymoops)
  (I was lazy so the register values are not correct)


Trace; d0979621 <[scsi_mod]scsi_old_done+281/620>
Trace; d0983455 <[ide-scsi]idescsi_end_request+205/260>
Trace; d09837a5 <[ide-scsi]idescsi_transfer_pc+a5/c0>
Trace; d0983536 <[ide-scsi]idescsi_pc_intr+86/250>
Trace; c0186288 <ide_set_handler+58/60>
Trace; c0107b00 <do_segment_not_present+30/a0>
Trace; d09834b0 <[ide-scsi]idescsi_pc_intr+0/250>
Trace; c01084aa <handle_IRQ_event+3a/70>
Code;  00000000 Before first symbol
00000000 <_EIP>:
Code;  00000000 Before first symbol
    0:   f3 ab                     repz stos %eax,%es:(%edi)
Code;  00000002 Before first symbol
    2:   8b 43 0c                  mov    0xc(%ebx),%eax
Code;  00000005 Before first symbol
    5:   80 b8 d9 00 00 00 03      cmpb   $0x3,0xd9(%eax)
Code;  0000000c Before first symbol
    c:   7f 0a                     jg     18 <_EIP+0x18> 00000018 Before 
first symbol
Code;  0000000e Before first symbol
    e:   0f b6 43 4c               movzbl 0x4c(%ebx),%eax
Code;  00000012 Before first symbol
   12:   c0 e0 00                  shl    $0x0,%al

  I'll keep trying ...


  Francois


Alan Cox wrote:

>>  I can trigger this oops from a user program running from a non 
>>priviledged account : dos -A (dosemu). It kills the kernel in an 
>>interrupt routine so I have to transcribe the oops from the dump 
>>on-screen :-)
>>
> 
> Does this still happen on 2.4.8ac2. I backed out a change in the scsi
> code that seemed to make things very unhappy for several ide-scsi users.
> 
> .
> 
>