stuck on TLB IPI wait - resolved??

Andrew Daviel (andrew@daviel.org)
Tue, 1 Jun 1999 10:33:30 -0700 (PDT)


(also sent - maybe - to linux-smp; not sure if my mailer lost it or
what. sorry if duplicated.)

I have had a couple of "stuck on TLB IPI wait" hangs,
and searching the kernel archives see four other people posting
about this but no resolution (apart from a perhaps-not-serious
suggestion to downgrade the kernel to 2.1.x).
Did the issue in fact get resolved or explained ?

I have RedHat 6.0,
gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)
and compiled both 2.2.5 and 2.3.2 kernels. System is
ASUS P2B-DS m/b (100MHz bus), Intel 440BX etc., dual Celeron (Mendocino,
stepping 5) o/c to 85MHz bus, 450MHz CPU, w/128Mb RAM

with 2.2.5:
May 30 07:50:01 daviel-v kernel: unexpected IRQ vector 208 on CPU#1!
May 30 07:50:01 daviel-v kernel: unexpected IRQ vector 61 on CPU#1!
May 30 07:50:01 daviel-v last message repeated 24 times
May 30 07:50:02 daviel-v kernel: stuck on TLB IPI wait (CPU#0)
..
May 30 08:00:00 daviel-v kernel: unexpected IRQ vector 61 on CPU#1!
May 30 08:00:00 daviel-v kernel: stuck on TLB IPI wait (CPU#1)

with 2.3.2:
May 28 09:24:12 daviel-v kernel: unexpected IRQ vector 61 on CPU#1!
May 28 09:24:34 daviel-v last message repeated 2 times
May 28 09:30:00 daviel-v kernel: stuck on TLB IPI wait (CPU#1)
May 28 09:30:01 daviel-v last message repeated 24 times

Soon after this (the last syslog entry) the system appeared totally locked
- softdog no good, no response to keyboard, no response to reset button,
no response to front panel "soft" powerswitch. I was running setiathome on
1 CPU, basically, and mprime (Mersenne prime numbers) on the other, for 10
hours or so.

Based on what others said, I'm just now trying to use more memory to make
it hang ... 9 copies setiathome, 4 copies xanim, ssh ... load 11, 125M RAM
+ 70M swap .. getting a little sluggish but seems solid so far...

If the problem's related to overclocking, I'll slow it down, but
it doesn't happen often enough to quickly keep trying things.

Andrew Daviel

Vancouver Webpages, TRIUMF, etc.

-
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/