gfs2: Per-revoke accounting in transactions

In the log, revokes are stored as a revoke descriptor (struct
gfs2_log_descriptor), followed by zero or more additional revoke blocks
(struct gfs2_meta_header).  On filesystems with a blocksize of 4k, the
revoke descriptor contains up to 503 revokes, and the metadata blocks
contain up to 509 revokes each.  We've so far been reserving space for
revokes in transactions in block granularity, so a lot more space than
necessary was being allocated and then released again.

This patch switches to assigning revokes to transactions individually
instead.  Initially, space for the revoke descriptor is reserved and
handed out to transactions.  When more revokes than that are reserved,
additional revoke blocks are added.  When the log is flushed, the space
for the additional revoke blocks is released, but we keep the space for
the revoke descriptor block allocated.

Transactions may still reserve more revokes than they will actually need
in the end, but now we won't overshoot the target as much, and by only
returning the space for excess revokes at log flush time, we further
reduce the amount of contention between processes.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>

1 files changed, 26 insertions, 8 deletions

diff --git a/fs/gfs2/trans.c b/fs/gfs2/trans.c
index 231ca1a41a7362..ab96cf0bf26bea 100644
--- a/fs/gfs2/trans.c
+++ b/fs/gfs2/trans.c
@@ -31,16 +31,18 @@ static void gfs2_print_trans(struct gfs2_sbd *sdp, const struct gfs2_trans *tr)
 	fs_warn(sdp, "blocks=%u revokes=%u reserved=%u touched=%u\n",
 		tr->tr_blocks, tr->tr_revokes, tr->tr_reserved,
 		test_bit(TR_TOUCHED, &tr->tr_flags));
-	fs_warn(sdp, "Buf %u/%u Databuf %u/%u Revoke %u/%u\n",
+	fs_warn(sdp, "Buf %u/%u Databuf %u/%u Revoke %u\n",
 		tr->tr_num_buf_new, tr->tr_num_buf_rm,
 		tr->tr_num_databuf_new, tr->tr_num_databuf_rm,
-		tr->tr_num_revoke, tr->tr_num_revoke_rm);
+		tr->tr_num_revoke);
 }
 
 int __gfs2_trans_begin(struct gfs2_trans *tr, struct gfs2_sbd *sdp,
 		       unsigned int blocks, unsigned int revokes,
 		       unsigned long ip)
 {
+	unsigned int extra_revokes;
+
 	if (current->journal_info) {
 		gfs2_print_trans(sdp, current->journal_info);
 		BUG();
@@ -62,8 +64,6 @@ int __gfs2_trans_begin(struct gfs2_trans *tr, struct gfs2_sbd *sdp,
 		 */
 		tr->tr_reserved += blocks + 1 + DIV_ROUND_UP(blocks - 1, databuf_limit(sdp));
 	}
-	if (revokes)
-		tr->tr_reserved += gfs2_struct2blk(sdp, revokes) - 1;
 	INIT_LIST_HEAD(&tr->tr_databuf);
 	INIT_LIST_HEAD(&tr->tr_buf);
 	INIT_LIST_HEAD(&tr->tr_list);
@@ -75,10 +75,26 @@ int __gfs2_trans_begin(struct gfs2_trans *tr, struct gfs2_sbd *sdp,
 
 	sb_start_intwrite(sdp->sd_vfs);
 
-	gfs2_log_reserve(sdp, tr->tr_reserved);
+	/*
+	 * Try the reservations under sd_log_flush_lock to prevent log flushes
+	 * from creating inconsistencies between the number of allocated and
+	 * reserved revokes.  If that fails, do a full-block allocation outside
+	 * of the lock to avoid stalling log flushes.  Then, allot the
+	 * appropriate number of blocks to revokes, use as many revokes locally
+	 * as needed, and "release" the surplus into the revokes pool.
+	 */
 
 	down_read(&sdp->sd_log_flush_lock);
+	if (gfs2_log_try_reserve(sdp, tr, &extra_revokes))
+		goto reserved;
+	up_read(&sdp->sd_log_flush_lock);
+	gfs2_log_reserve(sdp, tr, &extra_revokes);
+	down_read(&sdp->sd_log_flush_lock);
+
+reserved:
+	gfs2_log_release_revokes(sdp, extra_revokes);
 	if (unlikely(!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))) {
+		gfs2_log_release_revokes(sdp, tr->tr_revokes);
 		up_read(&sdp->sd_log_flush_lock);
 		gfs2_log_release(sdp, tr->tr_reserved);
 		sb_end_intwrite(sdp->sd_vfs);
@@ -113,14 +129,17 @@ void gfs2_trans_end(struct gfs2_sbd *sdp)
 	current->journal_info = NULL;
 
 	if (!test_bit(TR_TOUCHED, &tr->tr_flags)) {
-		gfs2_log_release(sdp, tr->tr_reserved);
+		gfs2_log_release_revokes(sdp, tr->tr_revokes);
 		up_read(&sdp->sd_log_flush_lock);
+		gfs2_log_release(sdp, tr->tr_reserved);
 		if (!test_bit(TR_ONSTACK, &tr->tr_flags))
 			gfs2_trans_free(sdp, tr);
 		sb_end_intwrite(sdp->sd_vfs);
 		return;
 	}
 
+	gfs2_log_release_revokes(sdp, tr->tr_revokes - tr->tr_num_revoke);
+
 	nbuf = tr->tr_num_buf_new + tr->tr_num_databuf_new;
 	nbuf -= tr->tr_num_buf_rm;
 	nbuf -= tr->tr_num_databuf_rm;
@@ -278,7 +297,6 @@ void gfs2_trans_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd)
 void gfs2_trans_remove_revoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len)
 {
 	struct gfs2_bufdata *bd, *tmp;
-	struct gfs2_trans *tr = current->journal_info;
 	unsigned int n = len;
 
 	gfs2_log_lock(sdp);
@@ -290,7 +308,7 @@ void gfs2_trans_remove_revoke(struct gfs2_sbd *sdp, u64 blkno, unsigned int len)
 			if (bd->bd_gl)
 				gfs2_glock_remove_revoke(bd->bd_gl);
 			kmem_cache_free(gfs2_bufdata_cachep, bd);
-			tr->tr_num_revoke_rm++;
+			gfs2_log_release_revokes(sdp, 1);
 			if (--n == 0)
 				break;
 		}