Bug#463808: kernel crash (Cause : 0000041c)

[tsbogend@alpha.franken.de (Thomas Bogendoerfer)]

On Tue, Feb 05, 2008 at 11:04:28AM -0700, Martin Michlmayr wrote:
> CCing Thomas Bogendoerfer

this looks like a userspace program accessing some mmap-ped data,
which is triggering a dbe. The first question is which application is it ?
My first thought was an X server, but I don't even know, if there is
one for the O2. What program are running, when the machine crashes ?

> > Got dbe at 0x2ac2bffc
> > [...]

> > Index: 20 pgmask=4kb va=0007fd06000 asid=6b
> >         [pa=0000c67b000 c=3 d=1 v=1 g=0] [pa=000032bf000 c=3 d=0 v=0 g=0]
> > Index: 21 pgmask=4kb va=0002acf0000 asid=6b
> >         [pa=0000d1c1000 c=3 d=0 v=1 g=0] [pa=0000c188000 c=3 d=0 v=0 g=0]
> > Index: 25 pgmask=4kb va=0002acf0000 asid=6b
> >         [pa=0000d1c1000 c=3 d=0 v=1 g=0] [pa=0000c188000 c=3 d=0 v=0 g=0]
> > Index: 41 pgmask=4kb va=0002aac8000 asid=6b
> >         [pa=0000368e000 c=3 d=0 v=0 g=0] [pa=000506c0000 c=3 d=1 v=1 g=0]

these are the only TLBs which are user space. What looks strange to me
is Index 21 and Index 25. Both are mapping the same page. Looking at
the R4600/4700 manual (didn't have a R5k manual handy) indicates no
problem with duplicate TLB entries. I can't check, if pyhsical addresses
are correct, because I don't know how memory is mapped. I need to see the
CRIME MC messages from bootup. Even better is a complete boot log.

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessary a
good idea.                                                [ RFC1925, 2.3 ]