On Mon, Jul 17, 2000 at 10:05:34AM -0700, Keith M Wesolowski wrote:
> Responding to my own mail, yeesh. I was obviously suffering a dumbass
> attack when I wrote this.
>
> > Code; 8801eb1c <r4k_flush_cache_page_s128d16i16+74/324>
> > 8: 8ce5003c lw $a1,60($a3)
> > Code; 8801eb20 <r4k_flush_cache_page_s128d16i16+78/324> <=====
> > c: 8c62003c lw $v0,60($v1) <=====
> >
> > The fault address is 0x3c. The offset of mm in current is 0x2c. Thus
> > the immediate cause appears to be that current->mm is 0x10, obviously
> > nonsense.
>
> The interesting bit is not current->mm, but current->mm->context. The
> offset of context is 60 as shown above in the disassembly. 60 = 3c, so
> it's clear that current->mm is in fact NULL.
>
> Hope this makes things a bit clearer.
Indeed, it does. I've commited a patch for this bug to cvs and would like
to hear reports.
Ralf
|