On Tue, 2002-07-30 at 00:24, Karl M. Hegbloom wrote:
> I'm complaining that when I compile large software systems, and
> there's a lot of disk activity, my computer locks up. It behaves,
> IMO, as though there is a deadlock in the file system code. Programs
> that are already running keep going, but I cannot launch new ones or
> do anything that requires access to the filesystem. I have no
> recourse but to press the reset button, and at reboot, the log replay
> error happens. No wonder; it cannot write the log entry and then gets
> interrupted by a hardware reset while it's waiting for the deadlock to
> I wonder if it's an SMP + XFS + EVMS (or any combination involving
> SMP) issue, so I wrote to these lists to see if anyone knows what's
> the matter.
> It occurs to me that you may need more information... I'm running a
> Debian 2.4.18 kernel image with (can I get the XFS and EVMS version
> numbers from the running kernel somehow? If not, please make that
> possible... Ah, yes, dmesg... Alright, Mr. XFS? Will you PLEASE make
> it print a version number at boot?) patches supplied by Debian
> packages. IIRC, it was more than two months ago that I built the
> kernel I am running now, which displays the buggy behavior. (The last
> try with fresh patches would not boot due to the unaligned access
> bug you already know about.)
You say current code will not boot? The problem is you need a much
more recent XFS than ones which work in 2.4.18 to function with EVMS.
Can you restate the unaligned access error?
As for an XFS version, unfortunately the way we release code does
not fit well with a version number - I sometimes wonder if we should
change this - but it would mean a bunch of work on this end.