I have to admit that was pretty awesome.
Alex
On Jun 14, 2011, at 2:46 PM, Josh Marshall <joshua.r.marshall.1...@gmail.com>
wrote:
Is there a resource whereby I could pickup this kind of debug skill, or is
this experience, and assembly knowledge?
On Mon, Jun 13, 2011 at 3:01 AM, <cinap_len...@gmx.de> wrote:
> ok, i think i can explain the "unlock of ilock: pc f0140fc5, held by ..."
> message.
>
> i extracted that 9pcflop kernel with muzgos cdfpart from the current iso
> and
> disassembled with acid arround.
>
> the routine that calls unlock looks like this:
>
> the routine that calls the unlock (shutdown from ether79c970.c)
> 0xf0140f79 0xf0140f79 SUBL $0x14,SP
> 0xf0140f7c 0xf0140f7c MOVL 0x18(SP),AX
> # ctlr = arg->foobar
> 0xf0140f80 0xf0140f80 MOVL 0x78(AX),CX
> 0xf0140f83 0xf0140f83 MOVL CX,0x10(SP)
> 0xf0140f87 0xf0140f87 MOVL CX,0x0(SP)
> # ilock(ctlr)
> 0xf0140f8a 0xf0140f8a CALL 0xf019af31
> 0xf0140f8f 0xf0140f8f MOVL 0x10(SP),AX
> 0xf0140f93 0xf0140f93 MOVL AX,0x0(SP)
> 0xf0140f96 0xf0140f96 MOVL $0x18,AX
> 0xf0140f9b 0xf0140f9b MOVL AX,0x4(SP)
> # io32r(ctlr, 0x18) ???
> 0xf0140f9f 0xf0140f9f CALL 0xf01409b4
> 0xf0140fa4 0xf0140fa4 MOVL 0x10(SP),AX
> 0xf0140fa8 0xf0140fa8 MOVL AX,0x0(SP)
> 0xf0140fab 0xf0140fab MOVL $0x18,AX
> 0xf0140fb0 0xf0140fb0 MOVL AX,0x4(SP)
> # io16r(ctlr, 0x18);
> 0xf0140fb4 0xf0140fb4 CALL 0xf014093f
> 0xf0140fb9 0xf0140fb9 MOVL 0x10(SP),AX
> 0xf0140fbd 0xf0140fbd MOVL AX,0x0(SP)
> # unlock(ctlr)
> 0xf0140fc0 0xf0140fc0 CALL 0xf019b0be
> 0xf0140fc5 0xf0140fc5 ADDL $0x14,SP
> 0xf0140fc8 0xf0140fc8 RET
>
> it have no debug symbols so i looked for magic numbers
> arround (in this case 0x18)
>
> wich turned out to be:
>
> ether79c970.c:31: Sreset = 0x0018, /* software reset
> */
>
> # this looks like io32r() from ether79c970.c
> acid: asm(0xf01409b4)
> 0xf01409b4 0xf01409b4 SUBL $0x8,SP
> 0xf01409b7 0xf01409b7 MOVL 0xc(SP),CX
> # port = arg2
> 0xf01409bb 0xf01409bb MOVL 0x1c(CX),CX
> # port += arg1
> 0xf01409be 0xf01409be ADDL 0x10(SP),CX
> 0xf01409c2 0xf01409c2 MOVL CX,0x0(SP)
> # inl(port)
> 0xf01409c5 0xf01409c5 CALL 0xf010036e
> 0xf01409ca 0xf01409ca ADDL $0x8,SP
> 0xf01409cd 0xf01409cd RET
>
> # this is inl()
> acid: asm(0xf010036e)
> 0xf010036e 0xf010036e MOVL 0x4(SP),DX
> 0xf0100372 0xf0100372 INL DX,AX
> 0xf0100373 0xf0100373 RET
>
> so there is your bug!
>
> static void
> shutdown(Ether *ether)
> {
> Ctlr *ctlr;
>
> ctlr = ether->ctlr;
> ilock(ctlr);
> io32r(ctlr, Sreset);
> io16r(ctlr, Sreset);
> unlock(ctlr);
> }
>
> it was commented out before...
>
> the file just changed... i think they try to fix it right now...
>
> --
> cinap
>
>
