Messages in this thread | | | Date | Thu, 31 Dec 1998 10:47:12 -0800 (PST) | From | John Kennedy <> | Subject | Re: good 2.1.x SMP kernel is? [possible smoking gun?] |
| |
12/31/98 @ 10:32:18 AM (Thursday)
Some more datapoints, and a possible smoking gun.
After 2.2.0-pre1 came out, I put the 2nd processor back into the machine and crashed it about 20 times trying to catch it in the act.
The machine is stable, in general, if:
o Only one of the two CPUs is used o You mount the partitions as sync (not good test; can't stress it)
Same (SMP) kernel, all other things held constant.
I spent most of yesterday trying to get it up on a serial console on the theory that if it would spit out any messages on that, I wouldn't have them disappear when the machine rebooted and/or otherwise froze. I never did get it to output it to the console, but I did get one message on the VGA console during one lockup (vs reboot):
swap_duplicate: entry 0b190000, offset exceeds max
Hmm. Swapping problems would happen while compiling but not under idle use, so I tried to abuse that. I did repeated kernel builds under things like -j32 and -j64 (both worked) where just -j failed (ran out of virtual memory, didn't write down exact error; typical fork issue).
Since this tends to happen an hour+ into my build, perhaps memory needs to be more scrambled by use to trigger the problem. I have never seen my new box use much swap, but that isn't surprising since I now have more memory than I used to have combined memory & swap.
As I said, my ASUS P2B-DS is an upgrade over my previous system. The previous system had 64MB of RAM, then was later upgraded to 96MB. The new system has 256MB. I did keep the old disk drive, and was using the old swap partition:
Disk /dev/sdb: 255 heads, 63 sectors, 1106 cylinders Units = cylinders of 16065 * 512 bytes
Device Boot Start End Blocks Id System /dev/sdb6 326 338 104391 82 Linux swap
No swap on /dev/sda. Sized Ok for 64MB and still more than 96, but less than 256. The only reason I'm using that is because it was there already and it seemed better than having no swap at all (at the time).
That was ~2AM last night, so I did a swapoff -a and restarted the big build. Got up this morning and it had finished it, no crashes (but only one test; I rebuilt 6-9 times on one CPU before I decided that was probably Ok).
Am I getting crashes because I have less swap than physical memory?
--- john
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |