Miguel Mendez wrote:

> > My main problem is that many (most? all?) NetBSD programs tend to cause
> > a segmentation fault. For example, when typing "w" 30 times, I will get
> > 26 successful executions and 4 times a segmentation fault (occurence is
> > total random). This is only an example and probably valid for all
> > programs (segmentation faults seen with "df", "gcc", "ls", "rm", "sed",
> > etc..).
> 
> I have the same hardware as you, and this problem has been there for a
> long time. AFAIK this only happens with dynamic executables and,

Oh yes, definitely! I found out that DSOs are mapped to an address
starting with 0x5ffe0000 on sgimips. And this is exactly the
segfault-address in most cases.

But today I always found segfaults at 0x7fffxxxx and 0x00000000...


> > Inspecting such a core dump with gdb always leads to the same result:
> > the program pointer is in illegal address space (at least I believe
> > so). The address is always 0x5ffe????.
> 
> Others can correct me here, but IIRC the problem is in the cache
> management code. I thought this had been fixed already, at least for the
> R4000 series, R5000 is a bit different and R8000 is a completely
> different beast.

We should make a test-kernel with caches disabled, if this is
possible on the Indy.


> Btw, I always used your phx assembler for my Amiga coding back in the
> day, great piece of software :)

Oh, thanks a lot! Old Amiga-users are everywhere. :)


-- 
   _  Frank Wille (frank@phoenix.owl.de)
_ //  http://devnull.owl.de/~frank/
\X/   Phx @ #AmigaGer