Greetings everyone, Since FreeBSD 5.3-BETA7 I think it was, I have experienced solid lockups on 2 SMP machines. It only seems to affect SMP machines, an Athlon XP 2000+ and a PIV 2.8 (with HTT disabled) do not experience these issues. Load seems unimportant, but high load seems to trigger the hangs quicker. The rather odd part is, that one of the boxes had no problems whatsoever before we added a SCSI controller to it (sym(4)), which worked perfectly in the previous Linux installation. Since this change we are experiencing these hangs. The other affected box has an ahc(4) SCSI card, so the problem may lie within the SCSI subsystem. The console of one of the boxes seemed to lag seriously behind, with delays up to one second (when using the console) while top(1) indicated the box was 99% idle. Both systems use the BSD scheduler, no preemption and such. You can find dmesg(8) output at http://rink.nu/tmp/nexus.txt and http://rink.nu/tmp/charm.txt. The latter has been compiled with debugging flags, and a serial line is hooked to it to see if something can be found. Alas, even a simple kernel build seems to be enough to solidly hang Charm. I am unable to send a break over the console to try to break into the debugger. Disabling SMP may avoid the problem, I have tried this now in a new kernel. This seems to make the box stable. Has someone else experienced such hangs? Since they seem to appear on 2 boxes, I doubt they are hardware-related. Anyone have any idea how to pinpoint these problems? There are very serious, at least in my opinion. Regards, Rink
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:20 UTC