Date: Thu, 12 May 2011 14:24:49 +0200
From: Jan Kara <jack@suse.cz>
To: Robert Morell <rmorell@nvidia.com>
Cc: Jan Kara <jack@suse.cz>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "linux-fsdevel@vger.kernel.org" <linux-fsdevel@vger.kernel.org>
Subject: Re: mmap vs. ctime bug?
Message-ID: <20110512122449.GB4690@quack.suse.cz>
References: <20110510012348.GJ3848@morell.nvidia.com>
 <20110511104358.GD5057@quack.suse.cz>
 <20110511162441.GC22847@morell.nvidia.com>
 <20110511170130.GL5057@quack.suse.cz>
 <20110511173726.GA7030@morell.nvidia.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20110511173726.GA7030@morell.nvidia.com>
User-Agent: Mutt/1.5.20 (2009-06-14)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 4843
Lines: 87

On Wed 11-05-11 10:37:26, Robert Morell wrote:
> On Wed, May 11, 2011 at 10:01:30AM -0700, Jan Kara wrote:
> >   I agree with the transparency as far as data is concerned. But it simply
> > cannot work for metadata - we don't know some things (like the number of
> > used blocks) in advance until the file is written. 
> > 
> > > If we want to quote specifications, see:
> > > http://pubs.opengroup.org/onlinepubs/9699919799/
> > > 
> > > "Section 4.8 "File Times Update"
> > > [...]
> > > An implementation may update timestamps that are marked for update
> > > immediately, or it may update such timestamps periodically. At the point
> > > in time when an update occurs, any marked timestamps shall be set to the
> > > current time and the update marks shall be cleared. All timestamps that
> > > are marked for update shall be updated when the file ceases to be open
> > > by any process or before a fstat(), fstatat(), fsync(), futimens(),
> > > lstat(), stat(), utime(), utimensat(), or utimes() is successfully
> > > performed on the file."
> >   But the allocation of a disk block (and thus a change to the inode) happens
> > after the file is closed. So the timestamp is marked for update after the
> > file is closed and we are consistent with the above paragraph. In fact we
> > should not avoid updating the time stamp because then applications would
> > miss that the metadata information in inode has actually changed.
> 
> Practically speaking, does anything that monitors ctimes actually care
> about st_blocks changes?  Certainly tar and other similar backup or
> archive-type programs shouldn't care; they only care about data that can
> be restored on a new filesystem.  Maybe an acceptable change would be to
> simply not trigger ctime updates based solely on disk block allocations?
  Yes, tar shouldn't care but if you had e.g. disk space monitor, it could
care. Actually if tar cared about data only, it should check mtime, but I
guess it also cares about owners, permissions, etc. which is why it's
checking ctime.

> I realize that this is not spec-compliant since "file status" has
> changed, but this behavior could be tweaked with filesystem mount
> options to turn on struct ctime behavior, similar to strictatime.
  This is possible but you'll run into problems when you want to run both
tar and imaginatory disk monitor...

> > > > So although I can see why the combination of this behavior and your
> > > > libelf+tar usecase causes problems the kernel behaves according to the spec
> > > > and I don't think changing the kernel is the right solution. I'd rather
> > > > think that you should be able to disable the ctime check in tar.
> > > 
> > > This really breaks basic assumptions about process lifetime and I/O.  In
> > > the basic shell flow:
> > > $ ./a && ./b
> > > When b is invoked, it is assumed that a has been terminated and any
> > > I/O it has performed will be reflected if b tries to read it.  (I assume
> > > the shell achieves this with wait(pid)?()).  Again, it is not guaranteed
> > > that the output be flushed to disk, but the cache should be transparent
> > > to software.
> >   Again, cache is transparent for data, not for metadata. So if b is
> > dependent on metadata changed by a, things get complicated. There are some
> > basic things defined by POSIX but apart from that all bets are off. Basically
> > the only way to get some guarantees is to use fsync/sync which is dumb but
> > that's how it is. Sorry. If you wanted that perfectly metadata consistent
> > behavior, kernel would have to basically fsync the file behind the scenes
> > and people certainly would not like that.
> 
> fsync/sync are much heavier-weight than should be necessary, though.
> None of the data has to actually hit the disk; filesystem blocks are at
> the end of the day just software state; requiring disk latency here is
> rather unfortunate.
  It may seem so but then you have to guarantee that you don't expose stale
data after a system crash so you have to be sure that data hits the disk
before metadata changes. And you also have to guarantee that after a crash
you can replay filesystem to a consistent state. On the other hand
filesystem allocation structures are shared among processes which modify
them. So all these requirements create a rather complex system which isn't
that easy to handle and theoretically possible "it's just software state"
ends up being tough practical problem.

>  An alternative fstatsync() or so that tar could
> call on its files would be sufficient as well.

									Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR
--
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/