Date: Sun, 30 Mar 2008 00:50:14 -0400
From: "Josef 'Jeff' Sipek" <jeffpc@josefsipek.net>
To: Eric Sandeen <sandeen@sandeen.net>
Cc: David Chinner <dgc@sgi.com>, xfs-dev <xfs-dev@sgi.com>,
        xfs-oss <xfs@oss.sgi.com>
Subject: Re: [patch] detect and correct bad features2 superblock field
Message-ID: <20080330045014.GA26934@josefsipek.net>
References: <20080220054041.GM155407@sgi.com> <47EEED18.9090206@sandeen.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <47EEED18.9090206@sandeen.net>
User-Agent: Mutt/1.5.16 (2007-06-11)
Sender: xfs-bounce@oss.sgi.com
Errors-to: xfs-bounce@oss.sgi.com
Precedence: bulk

On Sat, Mar 29, 2008 at 08:30:00PM -0500, Eric Sandeen wrote:
> David Chinner wrote:
> > There is a bug in mkfs.xfs that can result in writing the features2
> > field in the superblock to the wrong location. This only occurs
> > on some architectures, typically those with 32 bit userspace and
> > 64 bit kernels.
> > 
> > This patch detects the defect at mount time, logs a warning
> > such as:
> > 
> > XFS: correcting sb_features alignment problem
> > 
> > in dmesg and corrects the problem so that everything is OK.
> > it also blacklists the bad field in the superblock so it does
> > not get used for something else later on.
> 
> ...
> 
> >  	/*
> > +	 * Check for a bad features2 field alignment. This happened on
> > +	 * some platforms due to xfs_sb_t not being 64bit size aligned
> > +	 * when sb_features was added and hence the compiler put it in
> > +	 * the wrong place.
> > +	 *
> > +	 * If we detect a bad field, we or the set bits into the existing
> > +	 * features2 field in case it has already been modified and we
> > +	 * don't want to lose any features. Zero the bad one and mark
> > +	 * the two fields as needing updates once the transaction subsystem
> > +	 * is online.
> > +	 */
> > +	if (xfs_sb_has_bad_features2(sbp)) {
> > +		cmn_err(CE_WARN,
> > +			"XFS: correcting sb_features alignment problem");
> > +		sbp->sb_features2 |= sbp->sb_bad_features2;
> > +		sbp->sb_bad_features2 = 0;
> > +		update_flags |= XFS_SB_FEATURES2 | XFS_SB_BAD_FEATURES2;
> > +	}
> 
> Hm, the other problem here may be that if we zero bad_features2, then
> any older kernel will mount up as attr2... and run into the corruption
> problem I found on F8...
> 
> Should we make features2 and bad_features2 match rather than zeroing
> bad_features2?

I thought that was discussed here (or was it on IRC?), and the conclusion
was the best way is to always have features2 == bad_features2.  It is the
safest way to handle things - the filesystem is guaranteed to work
everywhere properly (old & new kernels).  Both the userspace (xfs_repair)
and kernel have to of course do the same thing (or bad_features2 with
features2, and save the result in both locations).

At least that's what I seem to remember.

Josef 'Jeff' Sipek.

-- 
Once you have their hardware. Never give it back.
(The First Rule of Hardware Acquisition)