Re: A fatal bug in traps.c in all versions >=2.0.11

Leonard N. Zubkoff (
Sat, 5 Oct 1996 09:32:23 -0700

Date: Sat, 5 Oct 1996 14:38:26 +0300 (EET DST)
From: Linus Torvalds <torvalds@cs.Helsinki.FI>

The line was replaces because soem Pentium Pro chipsets seem to do bad
things with reserved opcodes, and it was a debugging aid. I've restored
it to the old behaviour now in my own sources.

Please don't do that. That is not a debugging aid, if you recall my last
message on the subject. It is now a workaround for an acknowledged erratum in
the Pentium Pro CPU's local APIC. Spurious interrupts may be delivered as
exception 15 rather than interrupt 15. I'm afraid the current code or
something similar to ignore exception 15 is going to be permanent. The BIOS
can disable the local APIC for single processor operation, but in a multiple
processor system I believe there's no workaround except to ignore the bogus

Furthermore, I cannot see how changing that could cause the person's problem.
Without the workaround, the old code would kill the offending process with a
SIGSEGV. If the old code is somehow working, this must be explained.