CPU related dmesg badness with Hypersparc

[vom513 <vom513%gmail.com@localhost>]

Hello all,

I’ve been sniffing around this issue on my SS10 and SS20s.  It seems like with Ross Hypersparcs (in my case - always 2 CPUS / MP) - I get occasional stray interrupt and ping cpu errors:

[ 23509.812904] stray interrupt cpu0 ipl 0x7 pc=0xf0280938 npc=0xf028093c psr=0x1e8000c5<S,PS>
[ 35972.020078] stray interrupt cpu0 ipl 0x7 pc=0xf0280970 npc=0xf0280974 psr=0x1e9000c5<S,PS>
[ 37759.698875] autoconfiguration error: xcall(cpu0,0xf0008d54) from 0xf000b87c: couldn't ping cpus: cpu1

There doesn’t seem to be any other negative impact though - system doesn’t crash, keeps on ticking etc.

This doesn’t seem to happen with my Supersparc CPUs (I have tested using SM51 + SM61).

I have also tried multiple machines and multiple CPUs (different types) - same behavior.

I currently have 2 x Ross 125s in the SS20.

I found these bits through some googling:

http://gnats.netbsd.org/20113

https://marc.info/?l=netbsd-port-sparc&m=111593476615219&w=2

The mention of hme caught my eye as I have and am using hme in both my machines that experience this.  I took the hme out of the SS20, reconfigured to use le and I couldn’t get this to manifest in ~ 20 hours of uptime - whereas the dmesg logs above were from ~ 6 - 10 hours of uptime.

I wrote a script that just ftp’s a couple hundred meg file over, md5’s it, deletes it, and repeats.  Just trying to put some work on the machine.

Anyone else seeing this ?  I will also note that Solaris 9 doesn’t seem to do this.  I know this is somewhat apples and oranges, and perhaps if something is going on Solaris isn’t aware or doesn’t log it.  Just wanted to offer that as a datapoint as well as to this not seeming to be hardware issues.

Thanks.