linux-kernel - Re: [PATCH v2 1/3] ext4: Add alignment check for DAX mount

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Tue, 03 May 2016 09:41:36 -0600
From:	Toshi Kani <toshi.kani@....com>
To:	Ross Zwisler <ross.zwisler@...ux.intel.com>,
	Jan Kara <jack@...e.cz>
Cc:	Christoph Hellwig <hch@...radead.org>, dan.j.williams@...el.com,
	david@...morbit.com, boaz@...xistor.com, tytso@....edu,
	adilger.kernel@...ger.ca, micah.parrish@....com,
	linux-nvdimm@...1.01.org, linux-fsdevel@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 1/3] ext4: Add alignment check for DAX mount

On Tue, 2016-05-03 at 08:43 -0600, Ross Zwisler wrote:
> On Tue, May 03, 2016 at 11:00:21AM +0200, Jan Kara wrote:
> > 
> > On Tue 03-05-16 01:44:10, Christoph Hellwig wrote:
> > > 
> > > Please come up with a version that doesn't require tons of
> > > boilerplate code in every file system.
> >
> > Well, I was thinking about some helper as well but we could save ~4
> > lines with that and that didn't seem significant to me. Most of the
> > lines is actually reporting appropriate mount error in dmesg and that
> > is fs-dependent so it needs to stay in the filesystem... So what do you
> > have in mind?
>
> I guess if you wanted to reduce the code needed in each filesystem, you
> could avoid having different error messages for each of the failure
> conditions, and just print the error value.  All the error cases caught
> by the current code are unique, so we aren't losing any information.  The
> resulting patch for ext4 would look like this:

I'd prefer to keep the "error: device does not support dax" and "error:
unaligned partition for dax" messages since they clarify the problem as a
user error. The former case is especially common with BTT. The "error: dax
access failed (%d)" message is helpful when we need to look into the case.

If XFS is OK to use the messages similar to ext2/4, then we can do: 

 * Add a new helper function, say bdev_check_dax_mount(), which logs common
error messages with pr_err(), such as:
       "VFS (pmem0): error: device does not support dax"
 * When bdev_check_dax_mount() returns with a negative value:
   - XFS logs an additional message below via xfs_alert() and proceeds
without dax option.
       "XFS (pmem0): Turning DAX off."
   - ext2/4 fails the mount.

Thanks,
-Toshi