There is really only one SMP bugfix - the one for handling the deadlock
case. That currently corrupts IDE disk very occasionally for a few unlucky
folks. So that won't go in until resolved.
The other possible one is the capslock patch. Ingo ?
In case you haven't heard, I now have a new version of my deadlock patch that
should be safe for IDE systems too. I now have reports from a few people of
operation under heavy load with no IDE problems nor SMP lockups. I think the
capslock patch is also included, but it wouldn't hurt for someone to check in
case there's a different patch than the allow_interrupts one I have.
Leonard
>From lnz@dandelion.com Wed Aug 6 08:59:29 1997
Date: Tue, 5 Aug 1997 20:37:09 -0700
From: "Leonard N. Zubkoff" <lnz@dandelion.com>
To: linux-smp@vger.rutgers.edu
Subject: 2.0.30 Deadlock Patch
Deadlock Patch 5 is now available from my Linux web site at:
http://www.dandelion.com/Linux/
This version entirely removes the previous "interrupt simulation" mechanism for
resolving deadlocks which has caused problems for IDE systems; the IDE driver
cannot handle having its interrupt handler called unless an interrupt has
actually occurred, so the previous deadlock patches were problematic. A few
people with IDE systems who had problems with the previous patch have now
tested this version and found no problems so far, so I'm making it publicly
available for widespread testing. In addition, this version has apparently
resolved at least one problem with deadlocks using NFS.
The patch is against an unmodified 2.0.30 kernel; it can also be applied to
pre-patch-2.0.31-3 with a one line manual correction to buffer.c
Leonard