dual processor instability

[the_shrubber]

Hi all,

I currently have a dual 800 mhz machine (i think it's a quicksilver), and not matter what i do reformatting/reinstallation-wise, i cannot seem to make it stop kernel-panicking.

What i eventually did for a bearable amount of stability was to disable my second processor using some command in Open Firmware. This has increased my maximum uptime from one day to five days (what i mean to say was that my system was so stable that i thought disabling the processor actually did the trick, but after five days of tranquility, paf! "walked more than nbuf in core"... at least it's a different kernel panic message than what i'm used to)

Would anyone know what else i can do?

The most majorly naughty thing that i think i'm doing is mounting my home directory over NFS, but seeing as to how i'm at work, i don't really see any way to NOT do that for a prolonged period of time in order to troubleshoot. Okay, i guess it's possible but horribly inconvenient. Please tell me there's something i can try besides that!

Also, if it helps any, it seems that the times it crashes are when it's working hard (oh yea, real helpful, there), like compiling something, but i don't have any more precise language than that.

Cheers,

(-S-)

[crazy_canuck]

get your G4 checked out. If the actual G4 is silver, it a quicksilver, if its blue, its not.

[cybergoober]

Well, since it's a dual processor 800, then yes it is a quicksilver. This is not normal. I have a DP 800 and the only time I've had to restart was after an update that required it. You should get your G4 looked at.

[the_shrubber]

Hmm... that would be nice, but the Apple tech support are stuck thinking that i should try running my machine off the network for a week or something, which i guess might be a sensible troubleshooting step, but erm, using my machine off the network isn't using my machine. Is there any way to push this sort of thing through, or do my company have to go to outside sources?

[griffman]

Does your machine have any third-party RAM in it? Some instability in the past with OS X has been traced to flakey RAM (even though the RAM worked fine in OS 9).

Also, as much as I hate to agree with Apple's tech support, it seems like a quick way to maybe help pinpoint the problem. If you really get a kernel panic in under a day, then it shouldn't require a week - just try loading your home directory locally for a day or so, and see if the machine is any more stable.

Make sure you enable the Crash Reporter in the Console application (in Utilities; enable the reporter in Prefs). Then after a KP, open up Console and see if the error logs show anything. Also take a look at the system log and see if you can see what it was logging just before the KP.

Finally, any hardware attached other than keyboard and mouse? Try disconnecting everything other than the basics and run for a while.

-rob.

[the_shrubber]

Hmmm... nope, it's pretty vanilla.

I guess part of what makes this difficult is that it doesn't crash consistently, i mean, yah i say once a day, but sometimes it's once every two days, or once every three. Well, you guys are right, i guess i'll have to find a not busy week and go play. Will keep y'all posted.

Do let me know if you have ideas. Kernel panics get logged by the crash reporter? Hmm... they didn't seem to be for me, but maybe i wasn't checking. What does happen though is that i keep getting these messages:

WindowServer[18297]: CGXSetWindowProperty: Operation on a window not owned by you

Ideas?

Thanks a million

[griffman]

The KP itself won't be logged by the crash reporter, but there might be some indication just prior to the crash of what happened (in either the crash log or the system log).

I haven't seen that window server error message before, sorry.

Also, Apple likes to get the text of KPs sent to them via the feedback page, although it's a pain in the butt to do...

-rob.

[the_shrubber]

Okay, here's an update on my quest for stability...

I've still been too busy to do REAL troubleshooting, like unplugging my mac from the network or not mounting my home directory via NFS (sigh), but...

I've interestingly enough discovered that you can reset nvram to its factory defaults. Note, i ONLY JUST DID THIS, like, three seconds ago, so i don't know if it actually will work, but running sort and diff reveals that there are some number of differences (where they came from, i don't know). I'll report in a fortnight if mac turns out to be more stable.

(Resetting nvram - command-option-p-r on reboot... i think you wait till it chimes a second time)

(Diff results:
1d0
< ASVP 010717000600
3,5c2
< betty-cpu0 passed
< betty-cpu1 passed
< boot-args cpus=1
---
> boot-args
7c4
< boot-device mac-io/ata-4@1f000/@0:9,\\:tbxi
---
> boot-device hd:,\\:tbxi
12d8
< cpus
22c18
< fcode-debug? true
---
> fcode-debug? false
24c20
< input-device-1 /ipc
---
> input-device-1 scca
34c30
< output-device-1 /ipc
---
> output-device-1 scca

This represents a change TO the factory defaults.)

Anyone know where these diffs would have come from? What does YOUR nvram say, oh quicksilver-owners?

[the_shrubber]

welp... THAT didn't do it... sigh... /me sets boot-args=cpus=1 again

[jonesy]

What filesystem are you using? HFS or UFS.

I went through a couple of weeks of troubleshooting before I realized the FS was the problem.

[the_shrubber]

i'm using the default (HFS+)... though i would LOVE to switch to UFS just to have case-sensitivity back

[jonesy]

Quote: Originally posted by the_shrubber i'm using the default (HFS+)... though i would LOVE to switch to UFS just to have case-sensitivity back

I would love that too.

Sorry. I was using UFS for a time and I figured that's what caused my KPs. Long story short: OS X wants to make changes to directories/files that UFS is too strict about. (I haven't confirmed this, its only a theory, but I don't have the time or desire to truly test.) After switching back to HFS the panics disappeared.

I did have a client w/ a new dual processor G4 that experienced similar problems to what you discribe. (Funny they never happened while I was there. grrr!) I've also noted a few peop;e on Apple's disscussion boards telling a similar tale with dual processor woes.

So, mebbby Apple had a bad release of the G4 but I haven't seen them fess to any such thing. I dunno.

I'd like to hear about any solution. (Apple's solution for my client was to reinstall. I don't think it worked. The machine works but is buggy and slow.)

Good luck.

[the_shrubber]

Hmm... well, if you haven't already told them to try it, disabling my second processor does improve uptime like anything. Of course, it's not perfect, and you lose a processor, but better than crashing all the time.

One thing i'm tempted to do is physically switch my processors around (still leaving the cpus=1 switch) and see if its starts crashing. If it does, then that would hint that one of the G4s is just bad.

[diprotic]

Hi all.
My research group is running an 800MHz Dual Quicksilver. We are having the same problem here. The problem occurs EXACTLY as described above with UNIX and Mac OS X apps. We initally thought that the problem was temperature dependent but I am almost sure that it is load-dependent issue. I am doing some pretty heavy image rendering and it seems to trigger the problem pretty consistently. I have taken pictures of the screen dump here:
http://www.geneseo.edu/~spb1/sc_dump1.jpg
and here:

http://www.geneseo.edu/~spb1/sc_dump2.jpg

Disabling the second processor delayed, not fixed the problem. (Instead of it happening on the first frame, it happened on the fifth).
Here are my questions (as someone new to the Mac community):
a) Does Apple read these posts?
b) Is it possible they are already working on a solution?
c) Does anyone know of a utility that will allow me to track processor loads? It would be nice to pinpoint the "kill" level.

Thanks!
Sean

[jonesy]

There might be a few Apple employees reading these but not in any official capacity.

If you want then to see anything go to their forums on apple.com and gripe there. The respone isn't stellar though. They seem to like reinstalling a lot. (I guess it keeps the customer busy.) This is BSD fer christssakes! You shouldn't have to reinstall unless the damage to your system is beyond control!

Shrubber-

I'll have my client's try that, mebby they can get enough uptime to get their work done until I find a real solution. Where'd you go to do that, BTW?

Thanks

[the_shrubber]

diprotic, thanks for putting it in much clearer terms than "well, it seems to die when it's working hard", and also for the reassurance that it's not just me

As for tracking processor loads, i take it you mean something other than top and the little cpu monitor docklet? Ever tried looking at http://www.freshmeat.net?

jonesy,

open Terminal
sudo nvram boot-args = cpus=1

(perhaps cpus=1,v if you want it in verbose mode as well, but just a guess)

The uptime is rather bearable. I can go for a week, maybe two, sometimes more.

[dave@mmu]

It's to do with the screw that holds down the heatsink on your processors. The local apple store or qualified apple repair shop will change it. I had a friend with exactly the same problem and his was changed; never happened since. Apparently it expands when it gets hot and moves the cpu's connections.
Dave