Subject: Re: port-sparc/14180: random segfault on IPC with -current
To: None <port-sparc@netbsd.org>
From: der Mouse <mouse@Rodents.Montreal.QC.CA>
List: port-sparc
Date: 10/09/2001 04:25:47
>> 	On a sparc IPC, processes get memory corruption under moderate load
>> 	(like a 'make -j2' kernel compile).
>> 	This usually results in process cc1 getting a signal 11, but I've
>> 	also seen spurious syntax error report from cpp or cc.
>> 	Restarting the build make it go a little bit further.

> Does anyone else see these random coredumps?

Yes, though not with 1.5.x (which I don't run).  I've seen it with
various post-1.4 kernels.  It seems to correlate more with the
hardware; I've seen it on the SLC, ELC, SS1, SS1+.  When I posted about
it, someone speculated that a certain class of sun4c machines including
those and the IPC would do it, but that others (including some other
sun4c machines like the IPX) wouldn't.  This appears to have been the
case; I've stopped using SLC, ELC, 1, and 1+ machines and the problems
have completely vanished.

> There were some reports that IPX and SS1 had the same problem.
> (sun4c only problem?)

I've seen it on the 1 but never on the IPX.  Perhaps it does show up on
the IPX with certain kernel versions (which I've never run).

For me, it was most troublesome with a mud I was trying to run.  It
would generally provoke the problem and cause the mud to crash within
minutes of (mud) startup, but yes, I've seen it with "make build" too.

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse@rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B