On 2012-08-06 00:16:14, Ben Hutchings wrote:
> On Wed, 2012-08-01 at 12:11 -0300, Herton Ronaldo Krzesinski wrote:
> > Hi,
> > 
> > please consider including the following commit into 2.6.32.y, 3.0.y,
> > 3.2.y:
> > 
> > commit 4a26620df451ad46151ad21d711ed43e963c004e
> > Author: Tyler Hicks <[email protected]>
> > Date:   Sat Nov 5 13:45:08 2011 -0400
> > 
> >     eCryptfs: Improve statfs reporting
> > 
> > It fixes bug https://bugs.launchpad.net/ecryptfs/+bug/885744, but didn't
> > had an Cc stable on it.
> > 
> > I noted while running ecryptfs tests on stables that the test for the bug
> > was failing.
> > 
> > The fix cherry-picks cleanly on 3.0 and 3.2. For 2.6.32, an backport is
> > needed, the attached backport from Ubuntu 2.6.32 should work. If 2.6.34
> > or 2.6.35 are still maintained, feel free to include it too, I can
> > provide backported versions if needed.
> 
> This isn't a bug fix; it's a regression.  pathconf(_PC_NAME_MAX) needs
> to report an upper bound on the maximum name length, not a lower bound,
> so that readdir_r() can be used safely.

You're right. I was obviously only thinking about using
pathconf(_PC_NAME_MAX) to make sure a given filename would be usable in
an eCryptfs mount, not to determine buffer sizes when reading existing
filenames.

I can't remember the details now, but I was having trouble getting a
good calculation on the upper bound. I'll revisit it, get a fix in, mark
it for the stable tree and mention that it requires
4a26620df451ad46151ad21d711ed43e963c004e.

Thanks for the review, Ben.

Tyler

> 
> Ben.
> 
> -- 
> Ben Hutchings
> Theory and practice are closer in theory than in practice.
>                                 - John Levine, moderator of comp.compilers


Attachment: signature.asc
Description: Digital signature

Reply via email to