[vdr] Re: vdr[3330]: ERROR: can't record MPEG1! - possible cause

[Usenet-372111@zocki.toppoint.de (Rainer Zocholl)]

Usenet-372111@zocki.toppoint.de(Rainer Zocholl)  25.09.01 17:47


Together with the broken records i came almost always
to a kernel oops. (2.4.9/2.4.10 )



The kernel oops always in this code:

f3 a5 a8 02 74 02 66 a5 a8 01 74 01 a4 01 ab 5c 10 00 00 8b

0100 F3            REPZ
0101 A5            MOVSW
0102 A802          TEST    AL,02
0104 7402          JZ      0108

0106 66            DB      66
0107 A5            MOVSW

0108 A801          TEST    AL,01
010A 7401          JZ      010D

010C A4            MOVSB

010D 01AB5C10      ADD     [BP+DI+105C],BP = disaassembler bug (8086 only)
0111 0000          ADD     [BX+SI],AL
0113 8Bxxxx        MOV     SP,[BP+DI+6F]


that's "memcpy()" or similar.
(That might not be the actual cause, because that routine is
used a lot)

eax: ffffffc5 ecx:3ffbe8db(wow...lot to copy) edx: 194 (404 dez)
esi: d8ec08fb edi:d948dfff 
                  ~~~~~~~~ -> end of 4k-page
                  
"Kernel panic: Aiee, killing interrupt handler!"
"In interrupt handler - not syncing"

Process swapper: (pid: 0, stackpage=c02c700)


So i assume:
The kernel should swap in a new memory page (at 376MB physical ram...)
and can't.
Because he is alsready in an interrupt handler he panics.



Two questions:

a) Is it allowed that an interrupt *handler* is using swapped memory?

b) Could the parameter for the memcpy be wrong and it is copyin too much?