Re: oops in bk pull (oct 03)

[Alexander Viro]

On Fri, 4 Oct 2002, Linus Torvalds wrote:

> 
> On Fri, 4 Oct 2002, Alexander Viro wrote:
> > 
> > Hell knows.  The only explanation I see (and that's not worth much) is that
> > we somehow confuse the chipset and get crapped on something like next cache
> > miss.
> 
> I don't see any better explanation right now, so I guess we just revert 
> that thing.
> 
> The only other notion I might come up with is stack corruption, ie the
> code in pci_read_bases() might corrupt the return stack subtly (it does
> add another local variable whose address is taken), causing a jump to a
> random address on return. Compiler bug?

I doubt it.  I've read through the objdump output and code looks OK.
Diff between old and new _definitely_ looks sane.  FWIW, chipset is
Via 686A, gcc is from debian-stable (2.95.4-11woody1).  I'll try to
find some RH box and build with the same .config, but I would be
surprised if it changes anything.