All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dave Chinner <david@fromorbit.com>
To: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: linux-xfs@vger.kernel.org
Subject: Re: [PATCH 2/2] xfs: fix data corruption w/ unaligned reflink ranges
Date: Fri, 5 Oct 2018 15:21:21 +1000	[thread overview]
Message-ID: <20181005052121.GA12041@dastard> (raw)
In-Reply-To: <20181005014014.GO19324@magnolia>

On Thu, Oct 04, 2018 at 06:40:14PM -0700, Darrick J. Wong wrote:
> On Fri, Oct 05, 2018 at 11:23:36AM +1000, Dave Chinner wrote:
> > When reflinking sub-file ranges, a data corruption can occur when
> > the source file range includes a partial EOF block. This shares the
> > unknown data beyond EOF into the second file at a position inside
> > EOF, exposing stale data in the second file.
> > 
> > XFS only supports whole block sharing, but we still need to
> > support whole file reflink correctly.  Hence if the reflink
> > request includes the last block of the souce file, only proceed with
> > the reflink operation if it lands at or past the destination file's
> > current EOF. If it lands within the destination file EOF, reject the
> > entire request with -EINVAL and make the caller go the hard way.
> > 
> > This avoids the data corruption vector, but also avoids disruption
> > of returning EINVAL to userspace for the common case of whole file
> > cloning.
> > 
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > ---
> >  fs/xfs/xfs_reflink.c | 39 +++++++++++++++++++++++++++++++--------
> >  1 file changed, 31 insertions(+), 8 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_reflink.c b/fs/xfs/xfs_reflink.c
> > index 6b0da1b80103..2615271603ce 100644
> > --- a/fs/xfs/xfs_reflink.c
> > +++ b/fs/xfs/xfs_reflink.c
> > @@ -1229,12 +1229,24 @@ xfs_iolock_two_inodes_and_break_layout(
> >   * hence can introduce a corruption into the file that has it's
> >   * block replaced.
> >   *
> > - * Despite this issue, we still need to report that range as successfully
> > - * deduped to avoid confusing userspace with EINVAL errors on completely
> > - * matching file data. The only time that an unaligned length will be passed to
> > - * us is when it spans the EOF block of the source file, so if we simply mask it
> > - * down to be block aligned here the we will dedupe everything but that partial
> > - * EOF block.
> > + * In similar fashion, the VFS file cloning also allows partial EOF blocks to be
> > + * "block aligned" for the purposes of cloning entire files. 
> > + * However, if the source file range
> > + * includes the EOF block and it lands within the existing EOF of the
> > + * destination file, then we can expose stale data from beyond the source file
> > + * EOF in the destination file.
> > + *
> > + * XFs doesn't support partial block sharing, so in both cases we have check
> > + * these cases ourselves. For dedupe, we can simply round the length to dedupe
> > + * down to the previous whole block and ignore the partial EOF block. While this
> > + * means we can't dedupe the last block of a file, this is an acceptible
> > + * tradeoff for simplicity on implementation.
> > + *
> > + * For cloning, we want to share the partial EOF block if it is also the new EOF
> > + * block of the destination file. If the partial EOF blck lies inside the
> > + * existing destination EOF, then we have to abort the clone to avoid exposing
> > + * stale data int eh destination file. Hence we reject these clone attempts with
> > + * -EINVAL in this case.
> >   */
> >  int
> >  xfs_reflink_remap_range(
> > @@ -1255,6 +1267,7 @@ xfs_reflink_remap_range(
> >  	xfs_filblks_t		fsblen;
> >  	xfs_extlen_t		cowextsize;
> >  	ssize_t			ret;
> > +	u64			blkmask = i_blocksize(inode_in) - 1;
> >  
> >  	if (!xfs_sb_version_hasreflink(&mp->m_sb))
> >  		return -EOPNOTSUPP;
> > @@ -1292,8 +1305,18 @@ xfs_reflink_remap_range(
> >  	 * from the source file so we don't try to dedupe the partial
> >  	 * EOF block.
> >  	 */
> > -	if (is_dedupe)
> > -		len &= ~((u64)i_blocksize(inode_in) - 1);
> > +	if (is_dedupe) {
> > +		len &= ~blkmask;
> > +	} else if (len & blkmask) {
> > +		/*
> > +		 * The user is attempting to share a partial EOF block,
> > +		 * if it's inside the destination EOF then reject it
> > +		 */
> > +		if (pos_out + len < i_size_read(inode_out)) {
> > +			ret = -EINVAL;
> > +			goto out_unlock;
> 
> Hmm... to integrate this with the new series I just posted, I think we'd
> decrease len to be block aligned (perhaps in generic_clone_checks) so
> that copy_file_range would be able to pagecache copy the last bit
> instead of failing the whole operation.  IOWs,
> 
> if (is_dedupe) {
> 	len &= ~blkmask;
> } else if (len & blkmask) {
> 	if (pos_out + len < size_out) {
> 		len &= ~blkmask;
> 	}
> }

OK. But if I'm going to push it with just the EOF zeroing and
ctime/suid fixes, then this doesn't change until the handling of
partial completion is added to XFS later in the patchset, right?

Cheers,

Dave.

-- 
Dave Chinner
david@fromorbit.com

  reply	other threads:[~2018-10-05 12:18 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-10-05  1:23 [PATCH 0/2] xfs: fix another couple of reflink data corruptions Dave Chinner
2018-10-05  1:23 ` [PATCH 1/2] xfs: fix data corruption w/ unaligned dedupe ranges Dave Chinner
2018-10-05 23:58   ` Darrick J. Wong
2018-10-06 10:47   ` Christoph Hellwig
2018-10-05  1:23 ` [PATCH 2/2] xfs: fix data corruption w/ unaligned reflink ranges Dave Chinner
2018-10-05  1:40   ` Darrick J. Wong
2018-10-05  5:21     ` Dave Chinner [this message]
2018-10-05 17:11       ` Darrick J. Wong
2018-10-06  0:00   ` Darrick J. Wong
2018-10-06 10:54   ` Christoph Hellwig

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20181005052121.GA12041@dastard \
    --to=david@fromorbit.com \
    --cc=darrick.wong@oracle.com \
    --cc=linux-xfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.