[XFS] Lazy Superblock Counters

When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.

When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.

The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.

The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.

This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.

It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.

One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.

As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....

SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a

Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>

1 files changed, 148 insertions, 6 deletions

diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 5de1f392e632..f6fe47d8c4dc 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -643,6 +643,64 @@ xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
 					sbp->sb_inopblock);
 	mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
 }
+
+/*
+ * xfs_initialize_perag_data
+ *
+ * Read in each per-ag structure so we can count up the number of
+ * allocated inodes, free inodes and used filesystem blocks as this
+ * information is no longer persistent in the superblock. Once we have
+ * this information, write it into the in-core superblock structure.
+ */
+STATIC int
+xfs_initialize_perag_data(xfs_mount_t *mp, xfs_agnumber_t agcount)
+{
+	xfs_agnumber_t	index;
+	xfs_perag_t	*pag;
+	xfs_sb_t	*sbp = &mp->m_sb;
+	uint64_t	ifree = 0;
+	uint64_t	ialloc = 0;
+	uint64_t	bfree = 0;
+	uint64_t	bfreelst = 0;
+	uint64_t	btree = 0;
+	int		error;
+	int		s;
+
+	for (index = 0; index < agcount; index++) {
+		/*
+		 * read the agf, then the agi. This gets us
+		 * all the inforamtion we need and populates the
+		 * per-ag structures for us.
+		 */
+		error = xfs_alloc_pagf_init(mp, NULL, index, 0);
+		if (error)
+			return error;
+
+		error = xfs_ialloc_pagi_init(mp, NULL, index);
+		if (error)
+			return error;
+		pag = &mp->m_perag[index];
+		ifree += pag->pagi_freecount;
+		ialloc += pag->pagi_count;
+		bfree += pag->pagf_freeblks;
+		bfreelst += pag->pagf_flcount;
+		btree += pag->pagf_btreeblks;
+	}
+	/*
+	 * Overwrite incore superblock counters with just-read data
+	 */
+	s = XFS_SB_LOCK(mp);
+	sbp->sb_ifree = ifree;
+	sbp->sb_icount = ialloc;
+	sbp->sb_fdblocks = bfree + bfreelst + btree;
+	XFS_SB_UNLOCK(mp, s);
+
+	/* Fixup the per-cpu counters as well. */
+	xfs_icsb_reinit_counters(mp);
+
+	return 0;
+}
+
 /*
  * xfs_mountfs
  *
@@ -987,6 +1045,34 @@ xfs_mountfs(
 	}
 
 	/*
+	 * Now the log is mounted, we know if it was an unclean shutdown or
+	 * not. If it was, with the first phase of recovery has completed, we
+	 * have consistent AG blocks on disk. We have not recovered EFIs yet,
+	 * but they are recovered transactionally in the second recovery phase
+	 * later.
+	 *
+	 * Hence we can safely re-initialise incore superblock counters from
+	 * the per-ag data. These may not be correct if the filesystem was not
+	 * cleanly unmounted, so we need to wait for recovery to finish before
+	 * doing this.
+	 *
+	 * If the filesystem was cleanly unmounted, then we can trust the
+	 * values in the superblock to be correct and we don't need to do
+	 * anything here.
+	 *
+	 * If we are currently making the filesystem, the initialisation will
+	 * fail as the perag data is in an undefined state.
+	 */
+
+	if (xfs_sb_version_haslazysbcount(&mp->m_sb) &&
+	    !XFS_LAST_UNMOUNT_WAS_CLEAN(mp) &&
+	     !mp->m_sb.sb_inprogress) {
+		error = xfs_initialize_perag_data(mp, sbp->sb_agcount);
+		if (error) {
+			goto error2;
+		}
+	}
+	/*
 	 * Get and sanity-check the root inode.
 	 * Save the pointer to it in the mount structure.
 	 */
@@ -1049,6 +1135,7 @@ xfs_mountfs(
 		goto error4;
 	}
 
+
 	/*
 	 * Complete the quota initialisation, post-log-replay component.
 	 */
@@ -1111,10 +1198,9 @@ xfs_unmountfs(xfs_mount_t *mp, struct cred *cr)
 		xfs_binval(mp->m_rtdev_targp);
 	}
 
+	xfs_log_sbcount(mp, 1);
 	xfs_unmountfs_writesb(mp);
-
 	xfs_unmountfs_wait(mp); 		/* wait for async bufs */
-
 	xfs_log_unmount(mp);			/* Done! No more fs ops. */
 
 	xfs_freesb(mp);
@@ -1161,6 +1247,62 @@ xfs_unmountfs_wait(xfs_mount_t *mp)
 }
 
 int
+xfs_fs_writable(xfs_mount_t *mp)
+{
+	bhv_vfs_t	*vfsp = XFS_MTOVFS(mp);
+
+	return !(vfs_test_for_freeze(vfsp) || XFS_FORCED_SHUTDOWN(mp) ||
+		(vfsp->vfs_flag & VFS_RDONLY));
+}
+
+/*
+ * xfs_log_sbcount
+ *
+ * Called either periodically to keep the on disk superblock values
+ * roughly up to date or from unmount to make sure the values are
+ * correct on a clean unmount.
+ *
+ * Note this code can be called during the process of freezing, so
+ * we may need to use the transaction allocator which does not not
+ * block when the transaction subsystem is in its frozen state.
+ */
+int
+xfs_log_sbcount(
+	xfs_mount_t	*mp,
+	uint		sync)
+{
+	xfs_trans_t	*tp;
+	int		error;
+
+	if (!xfs_fs_writable(mp))
+		return 0;
+
+	xfs_icsb_sync_counters(mp);
+
+	/*
+	 * we don't need to do this if we are updating the superblock
+	 * counters on every modification.
+	 */
+	if (!xfs_sb_version_haslazysbcount(&mp->m_sb))
+		return 0;
+
+	tp = _xfs_trans_alloc(mp, XFS_TRANS_SB_COUNT);
+	error = xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
+					XFS_DEFAULT_LOG_COUNT);
+	if (error) {
+		xfs_trans_cancel(tp, 0);
+		return error;
+	}
+
+	xfs_mod_sb(tp, XFS_SB_IFREE | XFS_SB_ICOUNT | XFS_SB_FDBLOCKS);
+	if (sync)
+		xfs_trans_set_sync(tp);
+	xfs_trans_commit(tp, 0);
+
+	return 0;
+}
+
+int
 xfs_unmountfs_writesb(xfs_mount_t *mp)
 {
 	xfs_buf_t	*sbp;
@@ -1171,16 +1313,15 @@ xfs_unmountfs_writesb(xfs_mount_t *mp)
 	 * skip superblock write if fs is read-only, or
 	 * if we are doing a forced umount.
 	 */
-	sbp = xfs_getsb(mp, 0);
 	if (!(XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY ||
 		XFS_FORCED_SHUTDOWN(mp))) {
 
-		xfs_icsb_sync_counters(mp);
+		sbp = xfs_getsb(mp, 0);
+ 		sb = XFS_BUF_TO_SBP(sbp);
 
 		/*
 		 * mark shared-readonly if desired
 		 */
-		sb = XFS_BUF_TO_SBP(sbp);
 		if (mp->m_mk_sharedro) {
 			if (!(sb->sb_flags & XFS_SBF_READONLY))
 				sb->sb_flags |= XFS_SBF_READONLY;
@@ -1189,6 +1330,7 @@ xfs_unmountfs_writesb(xfs_mount_t *mp)
 			xfs_fs_cmn_err(CE_NOTE, mp,
 				"Unmounting, marking shared read-only");
 		}
+
 		XFS_BUF_UNDONE(sbp);
 		XFS_BUF_UNREAD(sbp);
 		XFS_BUF_UNDELAYWRITE(sbp);
@@ -1203,8 +1345,8 @@ xfs_unmountfs_writesb(xfs_mount_t *mp)
 					  mp, sbp, XFS_BUF_ADDR(sbp));
 		if (error && mp->m_mk_sharedro)
 			xfs_fs_cmn_err(CE_ALERT, mp, "Superblock write error detected while unmounting.  Filesystem may not be marked shared readonly");
+		xfs_buf_relse(sbp);
 	}
-	xfs_buf_relse(sbp);
 	return error;
 }