I'm running OS4.2 dev on a Sun SPARCstation 10 with sm71 supersparc-II and 512 megs ram. I have several times run in to problems that causes the system to freeze on heavy cpu load. This happens when doing multiple instances of heavy compiling and when launching other applications simultaneously. Sometimes the system just hangs and sometimes the display shows some info about a process receiving an invalid interupt and something about zalloc. It gives me the option to either press m for monitor or r for reboot. Is this happening to anyone else or do I have a hardware problem?
Jesus ! SS10 + SM71 + 512 MB, you pretty much maxed out the old lady ! :D
About your problem, did you setup a swap ? It's my belief it's not automatic in the install process, and the way to setup one is pretty much black magic if you believe the doc (use disk utility, and if the free partition you're targeting is small 'enough' then it is made to be a swap, or so I read - 'enough' not being clearly defined). While NS/OS can fare pretty well without swap or so it seems, it very well may help in your situation.
Otherwise, I've read that you're not supposed to have more than 256 MB Ram with NS 3.3, I don't know if that limit applies to OS 4.2 as well.
Yeah, I've squeezed in as much as possible into the ols ss10, it actually holds dual sm71 not that two cpus do any good in Openstep. I've tried to unplug one of the cpus but same thing happens with only one cpu present. The ss10 has a ROSS 2.25R PROM which is said to "sometimes behave diffrent" from sun PROM. Noone seems to know what "different" means any maybe there is nothing to it, but then again... :)
When the machine boots I see something about swapspace but not on a separate partition, on the / filesystem. Maybe this is something I should look into. Thanks
Quote from: "linmag7"Yeah, I've squeezed in as much as possible into the ols ss10, it actually holds dual sm71 not that two cpus do any good in Openstep. I've tried to unplug one of the cpus but same thing happens with only one cpu present. The ss10 has a ROSS 2.25R PROM which is said to "sometimes behave diffrent" from sun PROM. Noone seems to know what "different" means any maybe there is nothing to it, but then again... :)
Hum... 2xSM71 is rated as a "fairly hot" configuration ; I'd have vouched for an overheating problem but if you say that a mono-proc config yields the same problems... Did you try swapping your cpus just in case your #0 is a bit weaker than the other ? Anyway, I wouldn't leave the 2nd SM71 in place (unless you just run OS 4.2 from time to time but mainly use Solaris or Linux that can both make sense of the 2nd module). I'd rather keep it as a spare.
As of the 2.25r prom, there was a talk about it just recently, and I was considering buying one on ePay out of despair for not finding a plain 2.25 in France, but ultimately, a kindered soul took pity on me and burned the Sun-annointed 2.25 one for me. I don't plan to switch to a 2.25r, because to do so means I would be running high end Ross Hypersparcs, which are not NeXTstep friendly. And I can't see any rationale behind that, because a SS10 doesn't really qualify as a number cruncher anymore, whereas it makes a fine slab substitute, much easier to come by in Europe, for us NS addicts (If I understand well, most of NeXT original parts are channeled to the collectible market via Universities offloading their old gear ; an this is prohibited in France, a public-funded uni must auction every item of its inventory via a public-approved channel, oversighted by the ministry of finance).
Hi,
The system isn't running very hot, when touching the CPU heasink it hardly feels warm. But I guess the core of the CPU could still be hot and #0 cpu which is located beneath the 2nd cpu could be alot warmer... I only use the system for openstep so I don't really need two cpu. I'll try swaping between the two and maybe it solves the problem... I'll let you know if it does...
IMHO, even if CPU #2 feels "barely" warm, because it's not even initiated, cpu #0 must be fairly hot after a couple hours of compile time. Compiling for a long time is very cpu-intensive.
But your story reminds me of a x86 bug that appeared on some K6-2 config a long time ago while building linux kernels. My memories are sketchy here, but it might well be a gcc bug in the 2.x serie. Maybe mutatis mutandis, there are some bugs still lying in that old code that are triggered under certain conditions ; that would be bad luck indeed.
Hi,
I've recently swaped cpus in my SS10, I'm now running a different SM71 cpu. I alse switched to some more homogeneous memory dimms, I now have two sets of four identical modules. The system is running stable now, but it could go for days/weeks before so you never really know. I let you know if I think it solved the problem.
I guees my efforts was in vain. Today the kernel crashed, this time it died screaming: "panic: ialloc: dup alloc". Has anyone seen this error message before. It gave me the option of rebooting or enter a small kernel debugger "monitor"
The kernel panic "panic: ialloc: dup alloc". turns out to be due to accessing a dirty filesystem. I had a filesystem mounted manually, that is, it was not in my fstab and hence no automatic fsck after a powerfailure. I ran an fsck and now things work fine. Seems like NeXTSTEP aborts with a kernel panic when accesing dirty filesystems.
Quote from: "linmag7"The kernel panic "panic: ialloc: dup alloc". turns out to be due to accessing a dirty filesystem. I had a filesystem mounted manually, that is, it was not in my fstab and hence no automatic fsck after a powerfailure. I ran an fsck and now things work fine. Seems like NeXTSTEP aborts with a kernel panic when accesing dirty filesystems.
Finally, you got it ! Congrats, that makes sense.
I've removed some memory from the SS10 so it now "only" has 256 meggs instead of 512 of RAM. Since then I've had no problems what so ever with unexplained system lock-ups. I don't know if having more than 256 meggs of ram is known to cause problems with OpenSTEP on sparc or if there something wrong with my memory. I've also heard that the SS10 can act a bit strange on some revisions of the motherboard. The memory itself is polled from a sun Ultra 30 and I know that the Ultra 30 has different timings on their memory modules so maybe it's just that...