Invalidate: busy buffer + MD RAID 1

From: Igmar Palsenberg (i.palsenberg@jdimedia.nl)
Date: Wed Dec 18 2002 - 13:45:00 EST


Hi,

I get a 'invalidate: busy buffer' about 20 times at reboot. Only at
reboot however.

Setup :

linux-2.4.19 + grsecurity-1.9.7d + acl+xattr 0.8.53 + freeswan (inc. aes
and that kind of stuff)

The machine (Compaq ML350) has 2 scsi devices (sda, sdb) and a RAID 1
setup :

md0 : sda1 + sdb1
md1 : sda3 + sdb3
swap is done on sda2 + sdb2, using default prio's.

Triggering the 'invalidate: busy buffer' is easiest done by letting squid
create it's cache dirs and then rebooting.

No data corruption is occuring (at last not any that a force fsck can
detect), but I removed the RAID1 setup to make sure I sleep well tonight
:)

Looking at the md.c code, line 1708 :

    ITERATE_RDEV(mddev,rdev,tmp) {
        if (rdev->faulty)
            continue;
        invalidate_device(rdev->dev, 1);
        if (get_hardsect_size(rdev->dev)
> md_hardsect_sizes[mdidx(mddev)])
            md_hardsect_sizes[mdidx(mddev)] =
                get_hardsect_size(rdev->dev);
    }

Looks like it is invalidating the underlying devices (sda[13], sdb[13] in
my case.

Since my RAID array doesn't get screwed I suspect that the md code does
the above again on a do_md_stop(), but I can't find it.

Anyone got any comments on this ??

        Regards,

                Igmar

Please CC all responses.

-- 

Igmar Palsenberg JDI Media Solutions

Helhoek 30 6923PE Groessen Tel: +31 (0)316 - 596695 Fax: +31 (0)316 - 596699 The Netherlands

mailto: i.palsenberg@jdimedia.nl PGP/GPG key : http://www.jdimedia.nl/formulier/pgp/igmar

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Mon Dec 23 2002 - 22:00:20 EST