On Sun, Feb 03, 2008 at 08:09:39PM -0800, Petr Vandrovec wrote:
> I suspect that problem is that volume in question was repartitioned &
> reformatted recently - apparently it took udev some time to create nodes in
> /dev, so blkid somehow found devices in /dev/.static/dev/sdX1, and that
> triggered ENOMEDIUM problems on next boot - as for udev /dev/sdX1 would not
> exist, and so blkid would notice.
>
> (*) I've looked at sources, and blkid has special case for
> ENXIO/ENOENT/ENODEV. For ENOMEDIUM it reports back "returning unverified
> data", and returns whatever it found in cache. I have no idea what correct
> behavior should be - I'm sure in some cases users are trying to mount
> drives without medium inside.
That's very helpful. Thank you for doing the detailed debugging!
I should probably just change it so that any error (including
ENOMEDIUM) that's not EPERM causes the cache entry to be flushed.
That should address your problem, and that should cause blkid to do
the right thing.
- Ted
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
