xfs
[Top] [All Lists]

Re: XFS and ACPI sleep states compat in 2.5

To: Eric Sandeen <sandeen@xxxxxxx>
Subject: Re: XFS and ACPI sleep states compat in 2.5
From: Karol Kozimor <sziwan@xxxxxxxxxxx>
Date: Wed, 19 Feb 2003 17:04:04 +0100
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <Pine.LNX.4.44.0302190841341.20624-100000@stout.americas.sgi.com>
Mail-followup-to: Eric Sandeen <sandeen@xxxxxxx>, linux-xfs@xxxxxxxxxxx
References: <20030219100207.GA15374@hell.org.pl> <Pine.LNX.4.44.0302190841341.20624-100000@stout.americas.sgi.com>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/1.4i
Thus wrote Eric Sandeen:
> > There seems, however, to be an issue loosely connected with suspend.
> Does this behave differently with the patch vs. without?

No. As I stated earlier, the issue was present with 2.5.59 and 2.5.60, so
without your patch applied.

> > Suspending and resuming almost universally seems to trigger the strange
> > ,,Unknown error 990'' problems upon any file access.
> Are there any messages in the system logs before/after this?  990
> is "EFSCORRUPTED" out of xfs, perhaps the filesystem has shut down.
> There should be some error messages in the logs, if EFSCORRUPTED
> is being generated.

Well, nothing here. The logs are silent. Is it possible for the filesystem
to shut down selectively? I.e. I have two XFS filesystems mounted at / and
/home. When the problem occurs (at random circumstances, also before any
suspends), I am able to access one part of the filesystem (e.g. /bin,
/usr/local/bin), while on the same time it is impossible to access the
other (/usr/bin).

> I can try to take a look at this, but my problem with ACPI in the past
> has always been finding a machine which implements ACPI well enough
> in the bios that it has -any- hope of working correctly....

ACPI development has been quite active over the last months. ACPI included
in vanilla 2.4.x is about a year (or more) old, and is buggy. The new code
for 2.4.x is on acpi.sf.net and is synchronized with recent 2.5.x releases.
Anyway, either you've seen a very old (pre-2002) version of ACPI code, or
you must have been extremely unlucky.

> If the patch I posted doesn't make this any -worse- then I'll go ahead
> and check it in; if it is causing this problem, perhaps I should hold
> off.

As I said, the problem is not dependent. This filesystem shutdown (?) is
trigerred by the suspend / resume event, but it also exists independently.

Oh, I nearly forgot: there is a strange issue involved with the filesystem:
during the normal shutdown process, the kernel spits out the traceback of
something connected with fs/buffer.c. Alas, it happens either when my
keyboard doesn't work (after a S3 resume... well, it isn't perfect), or I'm
not paying attention, so I'm not able to press scroll-lock in time. I'll
try to catch this message and see if it will be of any meaning.

Best regards,

-- 
Karol 'sziwan' Kozimor
sziwan@xxxxxxxxxxx


<Prev in Thread] Current Thread [Next in Thread>