Re: [PATCH 037/119] xfs: remove an extent from the rmap btree

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Thu, Jun 16, 2016 at 06:21:49PM -0700, Darrick J. Wong wrote:
> From: Dave Chinner <dchinner@xxxxxxxxxx>
> 
> Now that we have records in the rmap btree, we need to remove them
> when extents are freed. This needs to find the relevant record in
> the btree and remove/trim/split it accordingly.
> 
> v2: Update the free function to deal with non-shared file data, and
> isolate the part that does the rmap update from the part that deals
> with cursors.  This will be useful for deferred ops.
> 
> [darrick.wong@xxxxxxxxxx: make rmap routines handle the enlarged keyspace]
> [dchinner: remove remaining unused debug printks]
> [darrick: fix a bug when growfs in an AG with an rmap ending at EOFS]
> 
> Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> Reviewed-by: Dave Chinner <dchinner@xxxxxxxxxx>
> Signed-off-by: Dave Chinner <david@xxxxxxxxxxxxx>
> ---

Reviewed-by: Brian Foster <bfoster@xxxxxxxxxx>

>  fs/xfs/libxfs/xfs_rmap.c |  220 +++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 215 insertions(+), 5 deletions(-)
> 
> 
> diff --git a/fs/xfs/libxfs/xfs_rmap.c b/fs/xfs/libxfs/xfs_rmap.c
> index 196e952..1043c63 100644
> --- a/fs/xfs/libxfs/xfs_rmap.c
> +++ b/fs/xfs/libxfs/xfs_rmap.c
> @@ -133,6 +133,212 @@ xfs_rmap_get_rec(
>  	return xfs_rmapbt_btrec_to_irec(rec, irec);
>  }
>  
> +/*
> + * Find the extent in the rmap btree and remove it.
> + *
> + * The record we find should always be an exact match for the extent that we're
> + * looking for, since we insert them into the btree without modification.
> + *
> + * Special Case #1: when growing the filesystem, we "free" an extent when
> + * growing the last AG. This extent is new space and so it is not tracked as
> + * used space in the btree. The growfs code will pass in an owner of
> + * XFS_RMAP_OWN_NULL to indicate that it expected that there is no owner of this
> + * extent. We verify that - the extent lookup result in a record that does not
> + * overlap.
> + *
> + * Special Case #2: EFIs do not record the owner of the extent, so when
> + * recovering EFIs from the log we pass in XFS_RMAP_OWN_UNKNOWN to tell the rmap
> + * btree to ignore the owner (i.e. wildcard match) so we don't trigger
> + * corruption checks during log recovery.
> + */
> +STATIC int
> +__xfs_rmap_free(
> +	struct xfs_btree_cur	*cur,
> +	xfs_agblock_t		bno,
> +	xfs_extlen_t		len,
> +	bool			unwritten,
> +	struct xfs_owner_info	*oinfo)
> +{
> +	struct xfs_mount	*mp = cur->bc_mp;
> +	struct xfs_rmap_irec	ltrec;
> +	uint64_t		ltoff;
> +	int			error = 0;
> +	int			i;
> +	uint64_t		owner;
> +	uint64_t		offset;
> +	unsigned int		flags;
> +	bool			ignore_off;
> +
> +	xfs_owner_info_unpack(oinfo, &owner, &offset, &flags);
> +	ignore_off = XFS_RMAP_NON_INODE_OWNER(owner) ||
> +			(flags & XFS_RMAP_BMBT_BLOCK);
> +	if (unwritten)
> +		flags |= XFS_RMAP_UNWRITTEN;
> +	trace_xfs_rmap_free_extent(mp, cur->bc_private.a.agno, bno, len,
> +			unwritten, oinfo);
> +
> +	/*
> +	 * We should always have a left record because there's a static record
> +	 * for the AG headers at rm_startblock == 0 created by mkfs/growfs that
> +	 * will not ever be removed from the tree.
> +	 */
> +	error = xfs_rmap_lookup_le(cur, bno, len, owner, offset, flags, &i);
> +	if (error)
> +		goto out_error;
> +	XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
> +
> +	error = xfs_rmap_get_rec(cur, &ltrec, &i);
> +	if (error)
> +		goto out_error;
> +	XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
> +	trace_xfs_rmap_lookup_le_range_result(cur->bc_mp,
> +			cur->bc_private.a.agno, ltrec.rm_startblock,
> +			ltrec.rm_blockcount, ltrec.rm_owner,
> +			ltrec.rm_offset, ltrec.rm_flags);
> +	ltoff = ltrec.rm_offset;
> +
> +	/*
> +	 * For growfs, the incoming extent must be beyond the left record we
> +	 * just found as it is new space and won't be used by anyone. This is
> +	 * just a corruption check as we don't actually do anything with this
> +	 * extent.  Note that we need to use >= instead of > because it might
> +	 * be the case that the "left" extent goes all the way to EOFS.
> +	 */
> +	if (owner == XFS_RMAP_OWN_NULL) {
> +		XFS_WANT_CORRUPTED_GOTO(mp, bno >= ltrec.rm_startblock +
> +						ltrec.rm_blockcount, out_error);
> +		goto out_done;
> +	}
> +
> +	/* Make sure the unwritten flag matches. */
> +	XFS_WANT_CORRUPTED_GOTO(mp, (flags & XFS_RMAP_UNWRITTEN) ==
> +			(ltrec.rm_flags & XFS_RMAP_UNWRITTEN), out_error);
> +
> +	/* Make sure the extent we found covers the entire freeing range. */
> +	XFS_WANT_CORRUPTED_GOTO(mp, ltrec.rm_startblock <= bno &&
> +		ltrec.rm_startblock + ltrec.rm_blockcount >=
> +		bno + len, out_error);
> +
> +	/* Make sure the owner matches what we expect to find in the tree. */
> +	XFS_WANT_CORRUPTED_GOTO(mp, owner == ltrec.rm_owner ||
> +				    XFS_RMAP_NON_INODE_OWNER(owner), out_error);
> +
> +	/* Check the offset, if necessary. */
> +	if (!XFS_RMAP_NON_INODE_OWNER(owner)) {
> +		if (flags & XFS_RMAP_BMBT_BLOCK) {
> +			XFS_WANT_CORRUPTED_GOTO(mp,
> +					ltrec.rm_flags & XFS_RMAP_BMBT_BLOCK,
> +					out_error);
> +		} else {
> +			XFS_WANT_CORRUPTED_GOTO(mp,
> +					ltrec.rm_offset <= offset, out_error);
> +			XFS_WANT_CORRUPTED_GOTO(mp,
> +					ltoff + ltrec.rm_blockcount >= offset + len,
> +					out_error);
> +		}
> +	}
> +
> +	if (ltrec.rm_startblock == bno && ltrec.rm_blockcount == len) {
> +		/* exact match, simply remove the record from rmap tree */
> +		trace_xfs_rmapbt_delete(mp, cur->bc_private.a.agno,
> +				ltrec.rm_startblock, ltrec.rm_blockcount,
> +				ltrec.rm_owner, ltrec.rm_offset,
> +				ltrec.rm_flags);
> +		error = xfs_btree_delete(cur, &i);
> +		if (error)
> +			goto out_error;
> +		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_error);
> +	} else if (ltrec.rm_startblock == bno) {
> +		/*
> +		 * overlap left hand side of extent: move the start, trim the
> +		 * length and update the current record.
> +		 *
> +		 *       ltbno                ltlen
> +		 * Orig:    |oooooooooooooooooooo|
> +		 * Freeing: |fffffffff|
> +		 * Result:            |rrrrrrrrrr|
> +		 *         bno       len
> +		 */
> +		ltrec.rm_startblock += len;
> +		ltrec.rm_blockcount -= len;
> +		if (!ignore_off)
> +			ltrec.rm_offset += len;
> +		error = xfs_rmap_update(cur, &ltrec);
> +		if (error)
> +			goto out_error;
> +	} else if (ltrec.rm_startblock + ltrec.rm_blockcount == bno + len) {
> +		/*
> +		 * overlap right hand side of extent: trim the length and update
> +		 * the current record.
> +		 *
> +		 *       ltbno                ltlen
> +		 * Orig:    |oooooooooooooooooooo|
> +		 * Freeing:            |fffffffff|
> +		 * Result:  |rrrrrrrrrr|
> +		 *                    bno       len
> +		 */
> +		ltrec.rm_blockcount -= len;
> +		error = xfs_rmap_update(cur, &ltrec);
> +		if (error)
> +			goto out_error;
> +	} else {
> +
> +		/*
> +		 * overlap middle of extent: trim the length of the existing
> +		 * record to the length of the new left-extent size, increment
> +		 * the insertion position so we can insert a new record
> +		 * containing the remaining right-extent space.
> +		 *
> +		 *       ltbno                ltlen
> +		 * Orig:    |oooooooooooooooooooo|
> +		 * Freeing:       |fffffffff|
> +		 * Result:  |rrrrr|         |rrrr|
> +		 *               bno       len
> +		 */
> +		xfs_extlen_t	orig_len = ltrec.rm_blockcount;
> +
> +		ltrec.rm_blockcount = bno - ltrec.rm_startblock;
> +		error = xfs_rmap_update(cur, &ltrec);
> +		if (error)
> +			goto out_error;
> +
> +		error = xfs_btree_increment(cur, 0, &i);
> +		if (error)
> +			goto out_error;
> +
> +		cur->bc_rec.r.rm_startblock = bno + len;
> +		cur->bc_rec.r.rm_blockcount = orig_len - len -
> +						     ltrec.rm_blockcount;
> +		cur->bc_rec.r.rm_owner = ltrec.rm_owner;
> +		if (ignore_off)
> +			cur->bc_rec.r.rm_offset = 0;
> +		else
> +			cur->bc_rec.r.rm_offset = offset + len;
> +		cur->bc_rec.r.rm_flags = flags;
> +		trace_xfs_rmapbt_insert(mp, cur->bc_private.a.agno,
> +				cur->bc_rec.r.rm_startblock,
> +				cur->bc_rec.r.rm_blockcount,
> +				cur->bc_rec.r.rm_owner,
> +				cur->bc_rec.r.rm_offset,
> +				cur->bc_rec.r.rm_flags);
> +		error = xfs_btree_insert(cur, &i);
> +		if (error)
> +			goto out_error;
> +	}
> +
> +out_done:
> +	trace_xfs_rmap_free_extent_done(mp, cur->bc_private.a.agno, bno, len,
> +			unwritten, oinfo);
> +out_error:
> +	if (error)
> +		trace_xfs_rmap_free_extent_error(mp, cur->bc_private.a.agno,
> +				bno, len, unwritten, oinfo);
> +	return error;
> +}
> +
> +/*
> + * Remove a reference to an extent in the rmap btree.
> + */
>  int
>  xfs_rmap_free(
>  	struct xfs_trans	*tp,
> @@ -143,19 +349,23 @@ xfs_rmap_free(
>  	struct xfs_owner_info	*oinfo)
>  {
>  	struct xfs_mount	*mp = tp->t_mountp;
> -	int			error = 0;
> +	struct xfs_btree_cur	*cur;
> +	int			error;
>  
>  	if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
>  		return 0;
>  
> -	trace_xfs_rmap_free_extent(mp, agno, bno, len, false, oinfo);
> -	if (1)
> +	cur = xfs_rmapbt_init_cursor(mp, tp, agbp, agno);
> +
> +	error = __xfs_rmap_free(cur, bno, len, false, oinfo);
> +	if (error)
>  		goto out_error;
> -	trace_xfs_rmap_free_extent_done(mp, agno, bno, len, false, oinfo);
> +
> +	xfs_btree_del_cursor(cur, XFS_BTREE_NOERROR);
>  	return 0;
>  
>  out_error:
> -	trace_xfs_rmap_free_extent_error(mp, agno, bno, len, false, oinfo);
> +	xfs_btree_del_cursor(cur, XFS_BTREE_ERROR);
>  	return error;
>  }
>  
> 
> _______________________________________________
> xfs mailing list
> xfs@xxxxxxxxxxx
> http://oss.sgi.com/mailman/listinfo/xfs
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [Samba]     [Device Mapper]     [CEPH Development]
  Powered by Linux