Ye Olde temporary freezes are back, same old symptoms. Any suggestions?
I'm running a SuSE-7.1 kernel, with noapic option, have 3 disks 2 on the
HPT366. Presumabaly one CPU is trying to tell the other that the TLB cache
is invalid. Is there any point in increasing the 'stuck' value any further?
I'm keeping the box up for much longer now, as I don't have to boot windows
so much any more : )
Jun 11 15:55:59 linux kernel: stuck on TLB IPI wait (CPU#1)
Jun 11 15:56:23 linux last message repeated 3 times
Jun 11 23:53:49 linux kernel: stuck on TLB IPI wait (CPU#0)
Jun 11 23:53:53 linux pppd[9593]: sent [LCP EchoReq id=0x2a magic=0x87feac36]
Jun 11 23:53:53 linux pppd[9593]: rcvd [LCP EchoRep id=0x2a magic=0x8e143f2c]
Jun 11 23:54:07 linux pppd[9593]: Modem hangup
Jun 11 23:54:07 linux kernel: stuck on TLB IPI wait (CPU#0)
Jun 11 23:54:07 linux kernel: stuck on TLB IPI wait (CPU#0)
Seems to happen mostly (but not always) when ppp is running, the code has
changed a bit., from when this problem was first reported
(marc.theaimsgroup.com archives this list).
/usr/src/linux/arch/i386/kernel/smp.c
/*
* Spin waiting for completion
*
* It turns out Intel seem to have been tuning their chips
* a little. The PIII-500+ seem to execute this little bit
* way faster than their older silicon. We now add on a factor
* guessed from the TSC calibration.
*/
stuck = 50000000 + cpu_data[cpu].loops_per_jiffy*(HZ/2);
while (smp_invalidate_needed) {
/*
* Take care of "crossing" invalidates
*/
if (test_bit(cpu, &smp_invalidate_needed))
clear_bit(cpu, &smp_invalidate_needed);
--stuck;
if (!stuck) {
printk("stuck on TLB IPI wait
(CPU#%d)\n",cpu);
break;
}
}
__restore_flags(flags);
Rob
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
