Merge tag 'gfs2-for-5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2

Pull gfs2 updates from Andreas Gruenbacher:

- Log space and revoke accounting rework to fix some failed asserts.

- Local resource group glock sharing for better local performance.

- Add support for version 1802 filesystems: trusted xattr support and
'-o rgrplvb' mounts by default.

- Actually synchronize on the inode glock's FREEING bit during withdraw
("gfs2: fix glock confusion in function signal_our_withdraw").

- Fix parallel recovery of multiple journals ("gfs2: keep bios separate
for each journal").

- Various other bug fixes.

* tag 'gfs2-for-5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2: (49 commits)
gfs2: Don't get stuck with I/O plugged in gfs2_ail1_flush
gfs2: Per-revoke accounting in transactions
gfs2: Rework the log space allocation logic
gfs2: Minor calc_reserved cleanup
gfs2: Use resource group glock sharing
gfs2: Allow node-wide exclusive glock sharing
gfs2: Add local resource group locking
gfs2: Add per-reservation reserved block accounting
gfs2: Rename rs_{free -> requested} and rd_{reserved -> requested}
gfs2: Check for active reservation in gfs2_release
gfs2: Don't search for unreserved space twice
gfs2: Only pass reservation down to gfs2_rbm_find
gfs2: Also reflect single-block allocations in rgd->rd_extfail_pt
gfs2: Recursive gfs2_quota_hold in gfs2_iomap_end
gfs2: Add trusted xattr support
gfs2: Enable rgrplvb for sb_fs_format 1802
gfs2: Don't skip dlm unlock if glock has an lvb
gfs2: Lock imbalance on error path in gfs2_recover_one
gfs2: Move function gfs2_ail_empty_tr
gfs2: Get rid of current_tail()
...

fs/gfs2/bmap.c

···
 
 	gfs2_inplace_release(ip);
 
+	if (ip->i_qadata && ip->i_qadata->qa_qd_num)
+		gfs2_quota_unlock(ip);
+
 	if (length != written && (iomap->flags & IOMAP_F_NEW)) {
 		/* Deallocate blocks that were just allocated. */
 		loff_t blockmask = i_blocksize(inode) - 1;
···
 			punch_hole(ip, pos, end - pos);
 		}
 	}
-
-	if (ip->i_qadata && ip->i_qadata->qa_qd_num)
-		gfs2_quota_unlock(ip);
 
 	if (unlikely(!written))
 		goto out_unlock;
···
 				goto out;
 			}
 			ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE,
-						 0, rd_gh);
+						 LM_FLAG_NODE_SCOPE, rd_gh);
 			if (ret)
 				goto out;
 
 			/* Must be done with the rgrp glock held: */
 			if (gfs2_rs_active(&ip->i_res) &&
-			    rgd == ip->i_res.rs_rbm.rgd)
+			    rgd == ip->i_res.rs_rgd)
 				gfs2_rs_deltree(&ip->i_res);
 		}
 

fs/gfs2/file.c

···
 	kfree(file->private_data);
 	file->private_data = NULL;
 
-	if (file->f_mode & FMODE_WRITE) {
+	if (gfs2_rs_active(&ip->i_res))
 		gfs2_rs_delete(ip, &inode->i_writecount);
+	if (file->f_mode & FMODE_WRITE)
 		gfs2_qa_put(ip);
-	}
 	return 0;
 }
 
···
 			goto out_qunlock;
 
 		/* check if the selected rgrp limits our max_blks further */
-		if (ap.allowed && ap.allowed < max_blks)
-			max_blks = ap.allowed;
+		if (ip->i_res.rs_reserved < max_blks)
+			max_blks = ip->i_res.rs_reserved;
 
 		/* Almost done. Calculate bytes that can be written using
 		 * max_blks. We also recompute max_bytes, data_blocks and

fs/gfs2/glock.c

···
 static inline int may_grant(const struct gfs2_glock *gl, const struct gfs2_holder *gh)
 {
 	const struct gfs2_holder *gh_head = list_first_entry(&gl->gl_holders, const struct gfs2_holder, gh_list);
-	if ((gh->gh_state == LM_ST_EXCLUSIVE ||
-	     gh_head->gh_state == LM_ST_EXCLUSIVE) && gh != gh_head)
-		return 0;
+
+	if (gh != gh_head) {
+		/**
+		 * Here we make a special exception to grant holders who agree
+		 * to share the EX lock with other holders who also have the
+		 * bit set. If the original holder has the LM_FLAG_NODE_SCOPE bit
+		 * is set, we grant more holders with the bit set.
+		 */
+		if (gh_head->gh_state == LM_ST_EXCLUSIVE &&
+		    (gh_head->gh_flags & LM_FLAG_NODE_SCOPE) &&
+		    gh->gh_state == LM_ST_EXCLUSIVE &&
+		    (gh->gh_flags & LM_FLAG_NODE_SCOPE))
+			return 1;
+		if ((gh->gh_state == LM_ST_EXCLUSIVE ||
+		     gh_head->gh_state == LM_ST_EXCLUSIVE))
+			return 0;
+	}
 	if (gl->gl_state == gh->gh_state)
 		return 1;
 	if (gh->gh_flags & GL_EXACT)
···
 		*p++ = 'A';
 	if (flags & LM_FLAG_PRIORITY)
 		*p++ = 'p';
+	if (flags & LM_FLAG_NODE_SCOPE)
+		*p++ = 'n';
 	if (flags & GL_ASYNC)
 		*p++ = 'a';
 	if (flags & GL_EXACT)

fs/gfs2/glock.h

···
  * request and directly join the other shared lock.  A shared lock request
  * without the priority flag might be forced to wait until the deferred
  * requested had acquired and released the lock.
+ *
+ * LM_FLAG_NODE_SCOPE
+ * This holder agrees to share the lock within this node. In other words,
+ * the glock is held in EX mode according to DLM, but local holders on the
+ * same node can share it.
  */
 
 #define LM_FLAG_TRY		0x0001
···
 #define LM_FLAG_NOEXP		0x0004
 #define LM_FLAG_ANY		0x0008
 #define LM_FLAG_PRIORITY	0x0010
+#define LM_FLAG_NODE_SCOPE	0x0020
 #define GL_ASYNC		0x0040
 #define GL_EXACT		0x0080
 #define GL_SKIP			0x0100

fs/gfs2/glops.c

···
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 	struct gfs2_trans tr;
+	unsigned int revokes;
 	int ret;
 
-	memset(&tr, 0, sizeof(tr));
-	INIT_LIST_HEAD(&tr.tr_buf);
-	INIT_LIST_HEAD(&tr.tr_databuf);
-	INIT_LIST_HEAD(&tr.tr_ail1_list);
-	INIT_LIST_HEAD(&tr.tr_ail2_list);
-	tr.tr_revokes = atomic_read(&gl->gl_ail_count);
+	revokes = atomic_read(&gl->gl_ail_count);
 
-	if (!tr.tr_revokes) {
+	if (!revokes) {
 		bool have_revokes;
 		bool log_in_flight;
 
···
 		return 0;
 	}
 
-	/* A shortened, inline version of gfs2_trans_begin()
-         * tr->alloced is not set since the transaction structure is
-         * on the stack */
-	tr.tr_reserved = 1 + gfs2_struct2blk(sdp, tr.tr_revokes);
-	tr.tr_ip = _RET_IP_;
-	ret = gfs2_log_reserve(sdp, tr.tr_reserved);
-	if (ret < 0)
-		return ret;
-	WARN_ON_ONCE(current->journal_info);
-	current->journal_info = &tr;
-
-	__gfs2_ail_flush(gl, 0, tr.tr_revokes);
-
+	memset(&tr, 0, sizeof(tr));
+	set_bit(TR_ONSTACK, &tr.tr_flags);
+	ret = __gfs2_trans_begin(&tr, sdp, 0, revokes, _RET_IP_);
+	if (ret)
+		goto flush;
+	__gfs2_ail_flush(gl, 0, revokes);
 	gfs2_trans_end(sdp);
+
 flush:
 	gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL |
 		       GFS2_LFC_AIL_EMPTY_GL);
···
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 	unsigned int revokes = atomic_read(&gl->gl_ail_count);
-	unsigned int max_revokes = (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_log_descriptor)) / sizeof(u64);
 	int ret;
 
 	if (!revokes)
 		return;
 
-	while (revokes > max_revokes)
-		max_revokes += (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header)) / sizeof(u64);
-
-	ret = gfs2_trans_begin(sdp, 0, max_revokes);
+	ret = gfs2_trans_begin(sdp, 0, revokes);
 	if (ret)
 		return;
-	__gfs2_ail_flush(gl, fsync, max_revokes);
+	__gfs2_ail_flush(gl, fsync, revokes);
 	gfs2_trans_end(sdp);
 	gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL |
 		       GFS2_LFC_AIL_FLUSH);

fs/gfs2/incore.h

···
 #include <linux/percpu.h>
 #include <linux/lockref.h>
 #include <linux/rhashtable.h>
+#include <linux/mutex.h>
 
 #define DIO_WAIT	0x00000010
 #define DIO_METADATA	0x00000020
···
 	u32 rd_data;			/* num of data blocks in rgrp */
 	u32 rd_bitbytes;		/* number of bytes in data bitmaps */
 	u32 rd_free;
-	u32 rd_reserved;                /* number of blocks reserved */
+	u32 rd_requested;		/* number of blocks in rd_rstree */
+	u32 rd_reserved;		/* number of reserved blocks */
 	u32 rd_free_clone;
 	u32 rd_dinodes;
 	u64 rd_igeneration;
···
 #define GFS2_RDF_PREFERRED	0x80000000 /* This rgrp is preferred */
 #define GFS2_RDF_MASK		0xf0000000 /* mask for internal flags */
 	spinlock_t rd_rsspin;           /* protects reservation related vars */
+	struct mutex rd_mutex;
 	struct rb_root rd_rstree;       /* multi-block reservation tree */
 };
-
-struct gfs2_rbm {
-	struct gfs2_rgrpd *rgd;
-	u32 offset;		/* The offset is bitmap relative */
-	int bii;		/* Bitmap index */
-};
-
-static inline struct gfs2_bitmap *rbm_bi(const struct gfs2_rbm *rbm)
-{
-	return rbm->rgd->rd_bits + rbm->bii;
-}
-
-static inline u64 gfs2_rbm_to_block(const struct gfs2_rbm *rbm)
-{
-	BUG_ON(rbm->offset >= rbm->rgd->rd_data);
-	return rbm->rgd->rd_data0 + (rbm_bi(rbm)->bi_start * GFS2_NBBY) +
-		rbm->offset;
-}
-
-static inline bool gfs2_rbm_eq(const struct gfs2_rbm *rbm1,
-			       const struct gfs2_rbm *rbm2)
-{
-	return (rbm1->rgd == rbm2->rgd) && (rbm1->bii == rbm2->bii) &&
-	       (rbm1->offset == rbm2->offset);
-}
 
 enum gfs2_state_bits {
 	BH_Pinned = BH_PrivateStart,
···
 */
 
 struct gfs2_blkreserv {
-	struct rb_node rs_node;       /* link to other block reservations */
-	struct gfs2_rbm rs_rbm;       /* Start of reservation */
-	u32 rs_free;                  /* how many blocks are still free */
+	struct rb_node rs_node;       /* node within rd_rstree */
+	struct gfs2_rgrpd *rs_rgd;
+	u64 rs_start;
+	u32 rs_requested;
+	u32 rs_reserved;              /* number of reserved blocks */
 };
 
 /*
···
 enum {
 	TR_TOUCHED = 1,
 	TR_ATTACHED = 2,
-	TR_ALLOCED = 3,
+	TR_ONSTACK = 3,
 };
 
 struct gfs2_trans {
···
 	unsigned int tr_num_buf_rm;
 	unsigned int tr_num_databuf_rm;
 	unsigned int tr_num_revoke;
-	unsigned int tr_num_revoke_rm;
 
 	struct list_head tr_list;
 	struct list_head tr_databuf;
···
 	unsigned int nr_extents;
 	struct work_struct jd_work;
 	struct inode *jd_inode;
+	struct bio *jd_log_bio;
 	unsigned long jd_flags;
 #define JDF_RECOVERY 1
 	unsigned int jd_jid;
···
 	unsigned int ar_errors:2;               /* errors=withdraw | panic */
 	unsigned int ar_nobarrier:1;            /* do not send barriers */
 	unsigned int ar_rgrplvb:1;		/* use lvbs for rgrp info */
+	unsigned int ar_got_rgrplvb:1;		/* Was the rgrplvb opt given? */
 	unsigned int ar_loccookie:1;		/* use location based readdir
 						   cookies */
 	s32 ar_commit;				/* Commit interval */
···
 
 	struct gfs2_trans *sd_log_tr;
 	unsigned int sd_log_blks_reserved;
-	int sd_log_committed_revoke;
 
 	atomic_t sd_log_pinned;
 	unsigned int sd_log_num_revoke;
···
 	atomic_t sd_log_thresh2;
 	atomic_t sd_log_blks_free;
 	atomic_t sd_log_blks_needed;
+	atomic_t sd_log_revokes_available;
 	wait_queue_head_t sd_log_waitq;
 	wait_queue_head_t sd_logd_waitq;
 
 	u64 sd_log_sequence;
-	unsigned int sd_log_head;
-	unsigned int sd_log_tail;
 	int sd_log_idle;
 
 	struct rw_semaphore sd_log_flush_lock;
 	atomic_t sd_log_in_flight;
-	struct bio *sd_log_bio;
 	wait_queue_head_t sd_log_flush_wait;
 	int sd_log_error; /* First log error */
 	wait_queue_head_t sd_withdraw_wait;
 
-	atomic_t sd_reserving_log;
-	wait_queue_head_t sd_reserving_log_wait;
-
+	unsigned int sd_log_tail;
+	unsigned int sd_log_flush_tail;
+	unsigned int sd_log_head;
 	unsigned int sd_log_flush_head;
 
 	spinlock_t sd_ail_lock;

fs/gfs2/inode.c

···
 	if (!rgd)
 		goto out_inodes;
 
-	gfs2_holder_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, ghs + 2);
+	gfs2_holder_init(rgd->rd_gl, LM_ST_EXCLUSIVE, LM_FLAG_NODE_SCOPE, ghs + 2);
 
 
 	error = gfs2_glock_nq(ghs); /* parent */
···
 			error = -ENOENT;
 			goto out_gunlock;
 		}
-		error = gfs2_glock_nq_init(nrgd->rd_gl, LM_ST_EXCLUSIVE, 0,
-					   &rd_gh);
+		error = gfs2_glock_nq_init(nrgd->rd_gl, LM_ST_EXCLUSIVE,
+					   LM_FLAG_NODE_SCOPE, &rd_gh);
 		if (error)
 			goto out_gunlock;
 	}

fs/gfs2/lock_dlm.c

···
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_sbd;
 	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
-	int lvb_needs_unlock = 0;
 	int error;
 
 	if (gl->gl_lksb.sb_lkid == 0) {
···
 	gfs2_sbstats_inc(gl, GFS2_LKS_DCOUNT);
 	gfs2_update_request_times(gl);
 
-	/* don't want to skip dlm_unlock writing the lvb when lock is ex */
-
-	if (gl->gl_lksb.sb_lvbptr && (gl->gl_state == LM_ST_EXCLUSIVE))
-		lvb_needs_unlock = 1;
+	/* don't want to skip dlm_unlock writing the lvb when lock has one */
 
 	if (test_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags) &&
-	    !lvb_needs_unlock) {
+	    !gl->gl_lksb.sb_lvbptr) {
 		gfs2_glock_free(gl);
 		return;
 	}

fs/gfs2/log.c

···
 	unsigned int blks;
 	unsigned int first, second;
 
+	/* The initial struct gfs2_log_descriptor block */
 	blks = 1;
 	first = sdp->sd_ldptrs;
 
 	if (nstruct > first) {
+		/* Subsequent struct gfs2_meta_header blocks */
 		second = sdp->sd_inptrs;
 		blks += DIV_ROUND_UP(nstruct - first, second);
 	}
···
 
 static int gfs2_ail1_start_one(struct gfs2_sbd *sdp,
 			       struct writeback_control *wbc,
-			       struct gfs2_trans *tr)
+			       struct gfs2_trans *tr, struct blk_plug *plug)
 __releases(&sdp->sd_ail_lock)
 __acquires(&sdp->sd_ail_lock)
 {
···
 			continue;
 		spin_unlock(&sdp->sd_ail_lock);
 		ret = generic_writepages(mapping, wbc);
+		if (need_resched()) {
+			blk_finish_plug(plug);
+			cond_resched();
+			blk_start_plug(plug);
+		}
 		spin_lock(&sdp->sd_ail_lock);
 		if (ret == -ENODATA) /* if a jdata write into a new hole */
 			ret = 0; /* ignore it */
···
 	list_for_each_entry_reverse(tr, head, tr_list) {
 		if (wbc->nr_to_write <= 0)
 			break;
-		ret = gfs2_ail1_start_one(sdp, wbc, tr);
+		ret = gfs2_ail1_start_one(sdp, wbc, tr, &plug);
 		if (ret) {
 			if (ret == -EBUSY)
 				goto restart;
···
 	};
 
 	return gfs2_ail1_flush(sdp, &wbc);
+}
+
+static void gfs2_log_update_flush_tail(struct gfs2_sbd *sdp)
+{
+	unsigned int new_flush_tail = sdp->sd_log_head;
+	struct gfs2_trans *tr;
+
+	if (!list_empty(&sdp->sd_ail1_list)) {
+		tr = list_last_entry(&sdp->sd_ail1_list,
+				     struct gfs2_trans, tr_list);
+		new_flush_tail = tr->tr_first;
+	}
+	sdp->sd_log_flush_tail = new_flush_tail;
+}
+
+static void gfs2_log_update_head(struct gfs2_sbd *sdp)
+{
+	unsigned int new_head = sdp->sd_log_flush_head;
+
+	if (sdp->sd_log_flush_tail == sdp->sd_log_head)
+		sdp->sd_log_flush_tail = new_head;
+	sdp->sd_log_head = new_head;
+}
+
+/**
+ * gfs2_ail_empty_tr - empty one of the ail lists of a transaction
+ */
+
+static void gfs2_ail_empty_tr(struct gfs2_sbd *sdp, struct gfs2_trans *tr,
+			      struct list_head *head)
+{
+	struct gfs2_bufdata *bd;
+
+	while (!list_empty(head)) {
+		bd = list_first_entry(head, struct gfs2_bufdata,
+				      bd_ail_st_list);
+		gfs2_assert(sdp, bd->bd_tr == tr);
+		gfs2_remove_from_ail(bd);
+	}
 }
 
 /**
···
 		else
 			oldest_tr = 0;
 	}
+	gfs2_log_update_flush_tail(sdp);
 	ret = list_empty(&sdp->sd_ail1_list);
 	spin_unlock(&sdp->sd_ail_lock);
 
···
 	spin_unlock(&sdp->sd_ail_lock);
 }
 
-/**
- * gfs2_ail_empty_tr - empty one of the ail lists for a transaction
- */
-
-static void gfs2_ail_empty_tr(struct gfs2_sbd *sdp, struct gfs2_trans *tr,
-			      struct list_head *head)
+static void __ail2_empty(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
 {
-	struct gfs2_bufdata *bd;
-
-	while (!list_empty(head)) {
-		bd = list_first_entry(head, struct gfs2_bufdata,
-				      bd_ail_st_list);
-		gfs2_assert(sdp, bd->bd_tr == tr);
-		gfs2_remove_from_ail(bd);
-	}
+	gfs2_ail_empty_tr(sdp, tr, &tr->tr_ail2_list);
+	list_del(&tr->tr_list);
+	gfs2_assert_warn(sdp, list_empty(&tr->tr_ail1_list));
+	gfs2_assert_warn(sdp, list_empty(&tr->tr_ail2_list));
+	gfs2_trans_free(sdp, tr);
 }
 
 static void ail2_empty(struct gfs2_sbd *sdp, unsigned int new_tail)
 {
-	struct gfs2_trans *tr, *safe;
+	struct list_head *ail2_list = &sdp->sd_ail2_list;
 	unsigned int old_tail = sdp->sd_log_tail;
-	int wrap = (new_tail < old_tail);
-	int a, b, rm;
+	struct gfs2_trans *tr, *safe;
 
 	spin_lock(&sdp->sd_ail_lock);
-
-	list_for_each_entry_safe(tr, safe, &sdp->sd_ail2_list, tr_list) {
-		a = (old_tail <= tr->tr_first);
-		b = (tr->tr_first < new_tail);
-		rm = (wrap) ? (a || b) : (a && b);
-		if (!rm)
-			continue;
-
-		gfs2_ail_empty_tr(sdp, tr, &tr->tr_ail2_list);
-		list_del(&tr->tr_list);
-		gfs2_assert_warn(sdp, list_empty(&tr->tr_ail1_list));
-		gfs2_assert_warn(sdp, list_empty(&tr->tr_ail2_list));
-		gfs2_trans_free(sdp, tr);
+	if (old_tail <= new_tail) {
+		list_for_each_entry_safe(tr, safe, ail2_list, tr_list) {
+			if (old_tail <= tr->tr_first && tr->tr_first < new_tail)
+				__ail2_empty(sdp, tr);
+		}
+	} else {
+		list_for_each_entry_safe(tr, safe, ail2_list, tr_list) {
+			if (old_tail <= tr->tr_first || tr->tr_first < new_tail)
+				__ail2_empty(sdp, tr);
+		}
 	}
-
 	spin_unlock(&sdp->sd_ail_lock);
+}
+
+/**
+ * gfs2_log_is_empty - Check if the log is empty
+ * @sdp: The GFS2 superblock
+ */
+
+bool gfs2_log_is_empty(struct gfs2_sbd *sdp) {
+	return atomic_read(&sdp->sd_log_blks_free) == sdp->sd_jdesc->jd_blocks;
+}
+
+static bool __gfs2_log_try_reserve_revokes(struct gfs2_sbd *sdp, unsigned int revokes)
+{
+	unsigned int available;
+
+	available = atomic_read(&sdp->sd_log_revokes_available);
+	while (available >= revokes) {
+		if (atomic_try_cmpxchg(&sdp->sd_log_revokes_available,
+				       &available, available - revokes))
+			return true;
+	}
+	return false;
+}
+
+/**
+ * gfs2_log_release_revokes - Release a given number of revokes
+ * @sdp: The GFS2 superblock
+ * @revokes: The number of revokes to release
+ *
+ * sdp->sd_log_flush_lock must be held.
+ */
+void gfs2_log_release_revokes(struct gfs2_sbd *sdp, unsigned int revokes)
+{
+	if (revokes)
+		atomic_add(revokes, &sdp->sd_log_revokes_available);
 }
 
 /**
···
 
 void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks)
 {
-
 	atomic_add(blks, &sdp->sd_log_blks_free);
 	trace_gfs2_log_blocks(sdp, blks);
 	gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <=
 				  sdp->sd_jdesc->jd_blocks);
-	up_read(&sdp->sd_log_flush_lock);
+	if (atomic_read(&sdp->sd_log_blks_needed))
+		wake_up(&sdp->sd_log_waitq);
 }
 
 /**
- * gfs2_log_reserve - Make a log reservation
+ * __gfs2_log_try_reserve - Try to make a log reservation
  * @sdp: The GFS2 superblock
  * @blks: The number of blocks to reserve
+ * @taboo_blks: The number of blocks to leave free
  *
- * Note that we never give out the last few blocks of the journal. Thats
- * due to the fact that there is a small number of header blocks
- * associated with each log flush. The exact number can't be known until
- * flush time, so we ensure that we have just enough free blocks at all
- * times to avoid running out during a log flush.
+ * Try to do the same as __gfs2_log_reserve(), but fail if no more log
+ * space is immediately available.
+ */
+static bool __gfs2_log_try_reserve(struct gfs2_sbd *sdp, unsigned int blks,
+				   unsigned int taboo_blks)
+{
+	unsigned wanted = blks + taboo_blks;
+	unsigned int free_blocks;
+
+	free_blocks = atomic_read(&sdp->sd_log_blks_free);
+	while (free_blocks >= wanted) {
+		if (atomic_try_cmpxchg(&sdp->sd_log_blks_free, &free_blocks,
+				       free_blocks - blks)) {
+			trace_gfs2_log_blocks(sdp, -blks);
+			return true;
+		}
+	}
+	return false;
+}
+
+/**
+ * __gfs2_log_reserve - Make a log reservation
+ * @sdp: The GFS2 superblock
+ * @blks: The number of blocks to reserve
+ * @taboo_blks: The number of blocks to leave free
+ *
+ * @taboo_blks is set to 0 for logd, and to GFS2_LOG_FLUSH_MIN_BLOCKS
+ * for all other processes.  This ensures that when the log is almost full,
+ * logd will still be able to call gfs2_log_flush one more time  without
+ * blocking, which will advance the tail and make some more log space
+ * available.
  *
  * We no longer flush the log here, instead we wake up logd to do that
  * for us. To avoid the thundering herd and to ensure that we deal fairly
  * with queued waiters, we use an exclusive wait. This means that when we
  * get woken with enough journal space to get our reservation, we need to
  * wake the next waiter on the list.
- *
- * Returns: errno
  */
 
-int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks)
+static void __gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks,
+			       unsigned int taboo_blks)
 {
-	int ret = 0;
-	unsigned reserved_blks = 7 * (4096 / sdp->sd_vfs->s_blocksize);
-	unsigned wanted = blks + reserved_blks;
-	DEFINE_WAIT(wait);
-	int did_wait = 0;
+	unsigned wanted = blks + taboo_blks;
 	unsigned int free_blocks;
 
-	if (gfs2_assert_warn(sdp, blks) ||
-	    gfs2_assert_warn(sdp, blks <= sdp->sd_jdesc->jd_blocks))
-		return -EINVAL;
 	atomic_add(blks, &sdp->sd_log_blks_needed);
-retry:
-	free_blocks = atomic_read(&sdp->sd_log_blks_free);
-	if (unlikely(free_blocks <= wanted)) {
-		do {
-			prepare_to_wait_exclusive(&sdp->sd_log_waitq, &wait,
-					TASK_UNINTERRUPTIBLE);
+	for (;;) {
+		if (current != sdp->sd_logd_process)
 			wake_up(&sdp->sd_logd_waitq);
-			did_wait = 1;
-			if (atomic_read(&sdp->sd_log_blks_free) <= wanted)
-				io_schedule();
-			free_blocks = atomic_read(&sdp->sd_log_blks_free);
-		} while(free_blocks <= wanted);
-		finish_wait(&sdp->sd_log_waitq, &wait);
+		io_wait_event(sdp->sd_log_waitq,
+			(free_blocks = atomic_read(&sdp->sd_log_blks_free),
+			 free_blocks >= wanted));
+		do {
+			if (atomic_try_cmpxchg(&sdp->sd_log_blks_free,
+					       &free_blocks,
+					       free_blocks - blks))
+				goto reserved;
+		} while (free_blocks >= wanted);
 	}
-	atomic_inc(&sdp->sd_reserving_log);
-	if (atomic_cmpxchg(&sdp->sd_log_blks_free, free_blocks,
-				free_blocks - blks) != free_blocks) {
-		if (atomic_dec_and_test(&sdp->sd_reserving_log))
-			wake_up(&sdp->sd_reserving_log_wait);
-		goto retry;
-	}
-	atomic_sub(blks, &sdp->sd_log_blks_needed);
+
+reserved:
 	trace_gfs2_log_blocks(sdp, -blks);
-
-	/*
-	 * If we waited, then so might others, wake them up _after_ we get
-	 * our share of the log.
-	 */
-	if (unlikely(did_wait))
+	if (atomic_sub_return(blks, &sdp->sd_log_blks_needed))
 		wake_up(&sdp->sd_log_waitq);
+}
 
-	down_read(&sdp->sd_log_flush_lock);
-	if (unlikely(!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))) {
-		gfs2_log_release(sdp, blks);
-		ret = -EROFS;
+/**
+ * gfs2_log_try_reserve - Try to make a log reservation
+ * @sdp: The GFS2 superblock
+ * @tr: The transaction
+ * @extra_revokes: The number of additional revokes reserved (output)
+ *
+ * This is similar to gfs2_log_reserve, but sdp->sd_log_flush_lock must be
+ * held for correct revoke accounting.
+ */
+
+bool gfs2_log_try_reserve(struct gfs2_sbd *sdp, struct gfs2_trans *tr,
+			  unsigned int *extra_revokes)
+{
+	unsigned int blks = tr->tr_reserved;
+	unsigned int revokes = tr->tr_revokes;
+	unsigned int revoke_blks = 0;
+
+	*extra_revokes = 0;
+	if (revokes && !__gfs2_log_try_reserve_revokes(sdp, revokes)) {
+		revoke_blks = DIV_ROUND_UP(revokes, sdp->sd_inptrs);
+		*extra_revokes = revoke_blks * sdp->sd_inptrs - revokes;
+		blks += revoke_blks;
 	}
-	if (atomic_dec_and_test(&sdp->sd_reserving_log))
-		wake_up(&sdp->sd_reserving_log_wait);
-	return ret;
+	if (!blks)
+		return true;
+	if (__gfs2_log_try_reserve(sdp, blks, GFS2_LOG_FLUSH_MIN_BLOCKS))
+		return true;
+	if (!revoke_blks)
+		gfs2_log_release_revokes(sdp, revokes);
+	return false;
+}
+
+/**
+ * gfs2_log_reserve - Make a log reservation
+ * @sdp: The GFS2 superblock
+ * @tr: The transaction
+ * @extra_revokes: The number of additional revokes reserved (output)
+ *
+ * sdp->sd_log_flush_lock must not be held.
+ */
+
+void gfs2_log_reserve(struct gfs2_sbd *sdp, struct gfs2_trans *tr,
+		      unsigned int *extra_revokes)
+{
+	unsigned int blks = tr->tr_reserved;
+	unsigned int revokes = tr->tr_revokes;
+	unsigned int revoke_blks = 0;
+
+	*extra_revokes = 0;
+	if (revokes) {
+		revoke_blks = DIV_ROUND_UP(revokes, sdp->sd_inptrs);
+		*extra_revokes = revoke_blks * sdp->sd_inptrs - revokes;
+		blks += revoke_blks;
+	}
+	__gfs2_log_reserve(sdp, blks, GFS2_LOG_FLUSH_MIN_BLOCKS);
 }
 
 /**
···
 }
 
 /**
- * calc_reserved - Calculate the number of blocks to reserve when
- *                 refunding a transaction's unused buffers.
+ * calc_reserved - Calculate the number of blocks to keep reserved
  * @sdp: The GFS2 superblock
  *
  * This is complex.  We need to reserve room for all our currently used
- * metadata buffers (e.g. normal file I/O rewriting file time stamps) and 
- * all our journaled data buffers for journaled files (e.g. files in the 
+ * metadata blocks (e.g. normal file I/O rewriting file time stamps) and
+ * all our journaled data blocks for journaled files (e.g. files in the
  * meta_fs like rindex, or files for which chattr +j was done.)
- * If we don't reserve enough space, gfs2_log_refund and gfs2_log_flush
- * will count it as free space (sd_log_blks_free) and corruption will follow.
+ * If we don't reserve enough space, corruption will follow.
  *
- * We can have metadata bufs and jdata bufs in the same journal.  So each
- * type gets its own log header, for which we need to reserve a block.
- * In fact, each type has the potential for needing more than one header 
- * in cases where we have more buffers than will fit on a journal page.
+ * We can have metadata blocks and jdata blocks in the same journal.  Each
+ * type gets its own log descriptor, for which we need to reserve a block.
+ * In fact, each type has the potential for needing more than one log descriptor
+ * in cases where we have more blocks than will fit in a log descriptor.
  * Metadata journal entries take up half the space of journaled buffer entries.
- * Thus, metadata entries have buf_limit (502) and journaled buffers have
- * databuf_limit (251) before they cause a wrap around.
  *
  * Also, we need to reserve blocks for revoke journal entries and one for an
  * overall header for the lot.
···
  */
 static unsigned int calc_reserved(struct gfs2_sbd *sdp)
 {
-	unsigned int reserved = 0;
-	unsigned int mbuf;
-	unsigned int dbuf;
+	unsigned int reserved = GFS2_LOG_FLUSH_MIN_BLOCKS;
+	unsigned int blocks;
 	struct gfs2_trans *tr = sdp->sd_log_tr;
 
 	if (tr) {
-		mbuf = tr->tr_num_buf_new - tr->tr_num_buf_rm;
-		dbuf = tr->tr_num_databuf_new - tr->tr_num_databuf_rm;
-		reserved = mbuf + dbuf;
-		/* Account for header blocks */
-		reserved += DIV_ROUND_UP(mbuf, buf_limit(sdp));
-		reserved += DIV_ROUND_UP(dbuf, databuf_limit(sdp));
+		blocks = tr->tr_num_buf_new - tr->tr_num_buf_rm;
+		reserved += blocks + DIV_ROUND_UP(blocks, buf_limit(sdp));
+		blocks = tr->tr_num_databuf_new - tr->tr_num_databuf_rm;
+		reserved += blocks + DIV_ROUND_UP(blocks, databuf_limit(sdp));
 	}
-
-	if (sdp->sd_log_committed_revoke > 0)
-		reserved += gfs2_struct2blk(sdp, sdp->sd_log_committed_revoke);
-	/* One for the overall header */
-	if (reserved)
-		reserved++;
 	return reserved;
 }
 
-static unsigned int current_tail(struct gfs2_sbd *sdp)
+static void log_pull_tail(struct gfs2_sbd *sdp)
 {
-	struct gfs2_trans *tr;
-	unsigned int tail;
+	unsigned int new_tail = sdp->sd_log_flush_tail;
+	unsigned int dist;
 
-	spin_lock(&sdp->sd_ail_lock);
-
-	if (list_empty(&sdp->sd_ail1_list)) {
-		tail = sdp->sd_log_head;
-	} else {
-		tr = list_last_entry(&sdp->sd_ail1_list, struct gfs2_trans,
-				tr_list);
-		tail = tr->tr_first;
-	}
-
-	spin_unlock(&sdp->sd_ail_lock);
-
-	return tail;
-}
-
-static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail)
-{
-	unsigned int dist = log_distance(sdp, new_tail, sdp->sd_log_tail);
-
+	if (new_tail == sdp->sd_log_tail)
+		return;
+	dist = log_distance(sdp, new_tail, sdp->sd_log_tail);
 	ail2_empty(sdp, new_tail);
-
-	atomic_add(dist, &sdp->sd_log_blks_free);
-	trace_gfs2_log_blocks(sdp, dist);
-	gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <=
-			     sdp->sd_jdesc->jd_blocks);
-
+	gfs2_log_release(sdp, dist);
 	sdp->sd_log_tail = new_tail;
 }
 
···
 }
 
 /**
- * gfs2_write_revokes - Add as many revokes to the system transaction as we can
+ * gfs2_flush_revokes - Add as many revokes to the system transaction as we can
  * @sdp: The GFS2 superblock
  *
  * Our usual strategy is to defer writing revokes as much as we can in the hope
···
  * been written back.  This will basically come at no cost now, and will save
  * us from having to keep track of those blocks on the AIL2 list later.
  */
-void gfs2_write_revokes(struct gfs2_sbd *sdp)
+void gfs2_flush_revokes(struct gfs2_sbd *sdp)
 {
 	/* number of revokes we still have room for */
-	int max_revokes = (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_log_descriptor)) / sizeof(u64);
+	unsigned int max_revokes = atomic_read(&sdp->sd_log_revokes_available);
 
 	gfs2_log_lock(sdp);
-	while (sdp->sd_log_num_revoke > max_revokes)
-		max_revokes += (sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header)) / sizeof(u64);
-	max_revokes -= sdp->sd_log_num_revoke;
-	if (!sdp->sd_log_num_revoke) {
-		atomic_dec(&sdp->sd_log_blks_free);
-		/* If no blocks have been reserved, we need to also
-		 * reserve a block for the header */
-		if (!sdp->sd_log_blks_reserved) {
-			atomic_dec(&sdp->sd_log_blks_free);
-			trace_gfs2_log_blocks(sdp, -2);
-		} else {
-			trace_gfs2_log_blocks(sdp, -1);
-		}
-	}
 	gfs2_ail1_empty(sdp, max_revokes);
 	gfs2_log_unlock(sdp);
-
-	if (!sdp->sd_log_num_revoke) {
-		atomic_inc(&sdp->sd_log_blks_free);
-		if (!sdp->sd_log_blks_reserved) {
-			atomic_inc(&sdp->sd_log_blks_free);
-			trace_gfs2_log_blocks(sdp, 2);
-		} else {
-			trace_gfs2_log_blocks(sdp, 1);
-		}
-	}
 }
 
 /**
···
 	u64 dblock;
 
 	if (gfs2_withdrawn(sdp))
-		goto out;
+		return;
 
 	page = mempool_alloc(gfs2_page_pool, GFP_NOIO);
 	lh = page_address(page);
···
 		     sb->s_blocksize - LH_V1_SIZE - 4);
 	lh->lh_crc = cpu_to_be32(crc);
 
-	gfs2_log_write(sdp, page, sb->s_blocksize, 0, dblock);
-	gfs2_log_submit_bio(&sdp->sd_log_bio, REQ_OP_WRITE | op_flags);
-out:
-	log_flush_wait(sdp);
+	gfs2_log_write(sdp, jd, page, sb->s_blocksize, 0, dblock);
+	gfs2_log_submit_bio(&jd->jd_log_bio, REQ_OP_WRITE | op_flags);
 }
 
 /**
···
 
 static void log_write_header(struct gfs2_sbd *sdp, u32 flags)
 {
-	unsigned int tail;
 	int op_flags = REQ_PREFLUSH | REQ_FUA | REQ_META | REQ_SYNC;
 	enum gfs2_freeze_state state = atomic_read(&sdp->sd_freeze_state);
 
 	gfs2_assert_withdraw(sdp, (state != SFS_FROZEN));
-	tail = current_tail(sdp);
 
 	if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags)) {
 		gfs2_ordered_wait(sdp);
 		log_flush_wait(sdp);
 		op_flags = REQ_SYNC | REQ_META | REQ_PRIO;
 	}
-	sdp->sd_log_idle = (tail == sdp->sd_log_flush_head);
-	gfs2_write_log_header(sdp, sdp->sd_jdesc, sdp->sd_log_sequence++, tail,
-			      sdp->sd_log_flush_head, flags, op_flags);
+	sdp->sd_log_idle = (sdp->sd_log_flush_tail == sdp->sd_log_flush_head);
+	gfs2_write_log_header(sdp, sdp->sd_jdesc, sdp->sd_log_sequence++,
+			      sdp->sd_log_flush_tail, sdp->sd_log_flush_head,
+			      flags, op_flags);
 	gfs2_log_incr_head(sdp);
-
-	if (sdp->sd_log_tail != tail)
-		log_pull_tail(sdp, tail);
+	log_flush_wait(sdp);
+	log_pull_tail(sdp);
+	gfs2_log_update_head(sdp);
 }
 
 /**
···
 void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl, u32 flags)
 {
 	struct gfs2_trans *tr = NULL;
+	unsigned int reserved_blocks = 0, used_blocks = 0;
 	enum gfs2_freeze_state state = atomic_read(&sdp->sd_freeze_state);
+	unsigned int first_log_head;
+	unsigned int reserved_revokes = 0;
 
 	down_write(&sdp->sd_log_flush_lock);
+	trace_gfs2_log_flush(sdp, 1, flags);
 
+repeat:
 	/*
 	 * Do this check while holding the log_flush_lock to prevent new
 	 * buffers from being added to the ail via gfs2_pin()
···
 	/* Log might have been flushed while we waited for the flush lock */
 	if (gl && !test_bit(GLF_LFLUSH, &gl->gl_flags))
 		goto out;
-	trace_gfs2_log_flush(sdp, 1, flags);
+
+	first_log_head = sdp->sd_log_head;
+	sdp->sd_log_flush_head = first_log_head;
+
+	tr = sdp->sd_log_tr;
+	if (tr || sdp->sd_log_num_revoke) {
+		if (reserved_blocks)
+			gfs2_log_release(sdp, reserved_blocks);
+		reserved_blocks = sdp->sd_log_blks_reserved;
+		reserved_revokes = sdp->sd_log_num_revoke;
+		if (tr) {
+			sdp->sd_log_tr = NULL;
+			tr->tr_first = first_log_head;
+			if (unlikely (state == SFS_FROZEN)) {
+				if (gfs2_assert_withdraw_delayed(sdp,
+				       !tr->tr_num_buf_new && !tr->tr_num_databuf_new))
+					goto out_withdraw;
+			}
+		}
+	} else if (!reserved_blocks) {
+		unsigned int taboo_blocks = GFS2_LOG_FLUSH_MIN_BLOCKS;
+
+		reserved_blocks = GFS2_LOG_FLUSH_MIN_BLOCKS;
+		if (current == sdp->sd_logd_process)
+			taboo_blocks = 0;
+
+		if (!__gfs2_log_try_reserve(sdp, reserved_blocks, taboo_blocks)) {
+			up_write(&sdp->sd_log_flush_lock);
+			__gfs2_log_reserve(sdp, reserved_blocks, taboo_blocks);
+			down_write(&sdp->sd_log_flush_lock);
+			goto repeat;
+		}
+		BUG_ON(sdp->sd_log_num_revoke);
+	}
 
 	if (flags & GFS2_LOG_HEAD_FLUSH_SHUTDOWN)
 		clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
 
-	sdp->sd_log_flush_head = sdp->sd_log_head;
-	tr = sdp->sd_log_tr;
-	if (tr) {
-		sdp->sd_log_tr = NULL;
-		tr->tr_first = sdp->sd_log_flush_head;
-		if (unlikely (state == SFS_FROZEN))
-			if (gfs2_assert_withdraw_delayed(sdp,
-			       !tr->tr_num_buf_new && !tr->tr_num_databuf_new))
-				goto out_withdraw;
-	}
-
 	if (unlikely(state == SFS_FROZEN))
-		if (gfs2_assert_withdraw_delayed(sdp, !sdp->sd_log_num_revoke))
+		if (gfs2_assert_withdraw_delayed(sdp, !reserved_revokes))
 			goto out_withdraw;
-	if (gfs2_assert_withdraw_delayed(sdp,
-			sdp->sd_log_num_revoke == sdp->sd_log_committed_revoke))
-		goto out_withdraw;
 
 	gfs2_ordered_write(sdp);
 	if (gfs2_withdrawn(sdp))
···
 	lops_before_commit(sdp, tr);
 	if (gfs2_withdrawn(sdp))
 		goto out_withdraw;
-	gfs2_log_submit_bio(&sdp->sd_log_bio, REQ_OP_WRITE);
+	gfs2_log_submit_bio(&sdp->sd_jdesc->jd_log_bio, REQ_OP_WRITE);
 	if (gfs2_withdrawn(sdp))
 		goto out_withdraw;
 
 	if (sdp->sd_log_head != sdp->sd_log_flush_head) {
-		log_flush_wait(sdp);
 		log_write_header(sdp, flags);
-	} else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle){
-		atomic_dec(&sdp->sd_log_blks_free); /* Adjust for unreserved buffer */
-		trace_gfs2_log_blocks(sdp, -1);
+	} else if (sdp->sd_log_tail != sdp->sd_log_flush_tail && !sdp->sd_log_idle) {
 		log_write_header(sdp, flags);
 	}
 	if (gfs2_withdrawn(sdp))
···
 	lops_after_commit(sdp, tr);
 
 	gfs2_log_lock(sdp);
-	sdp->sd_log_head = sdp->sd_log_flush_head;
 	sdp->sd_log_blks_reserved = 0;
-	sdp->sd_log_committed_revoke = 0;
 
 	spin_lock(&sdp->sd_ail_lock);
 	if (tr && !list_empty(&tr->tr_ail1_list)) {
···
 			empty_ail1_list(sdp);
 			if (gfs2_withdrawn(sdp))
 				goto out_withdraw;
-			atomic_dec(&sdp->sd_log_blks_free); /* Adjust for unreserved buffer */
-			trace_gfs2_log_blocks(sdp, -1);
 			log_write_header(sdp, flags);
-			sdp->sd_log_head = sdp->sd_log_flush_head;
 		}
 		if (flags & (GFS2_LOG_HEAD_FLUSH_SHUTDOWN |
 			     GFS2_LOG_HEAD_FLUSH_FREEZE))
···
 	}
 
 out_end:
-	trace_gfs2_log_flush(sdp, 0, flags);
+	used_blocks = log_distance(sdp, sdp->sd_log_flush_head, first_log_head);
+	reserved_revokes += atomic_read(&sdp->sd_log_revokes_available);
+	atomic_set(&sdp->sd_log_revokes_available, sdp->sd_ldptrs);
+	gfs2_assert_withdraw(sdp, reserved_revokes % sdp->sd_inptrs == sdp->sd_ldptrs);
+	if (reserved_revokes > sdp->sd_ldptrs)
+		reserved_blocks += (reserved_revokes - sdp->sd_ldptrs) / sdp->sd_inptrs;
 out:
+	if (used_blocks != reserved_blocks) {
+		gfs2_assert_withdraw_delayed(sdp, used_blocks < reserved_blocks);
+		gfs2_log_release(sdp, reserved_blocks - used_blocks);
+	}
 	up_write(&sdp->sd_log_flush_lock);
 	gfs2_trans_free(sdp, tr);
 	if (gfs2_withdrawing(sdp))
 		gfs2_withdraw(sdp);
+	trace_gfs2_log_flush(sdp, 0, flags);
 	return;
 
 out_withdraw:
···
 	old->tr_num_databuf_new	+= new->tr_num_databuf_new;
 	old->tr_num_buf_rm	+= new->tr_num_buf_rm;
 	old->tr_num_databuf_rm	+= new->tr_num_databuf_rm;
+	old->tr_revokes		+= new->tr_revokes;
 	old->tr_num_revoke	+= new->tr_num_revoke;
-	old->tr_num_revoke_rm	+= new->tr_num_revoke_rm;
 
 	list_splice_tail_init(&new->tr_databuf, &old->tr_databuf);
 	list_splice_tail_init(&new->tr_buf, &old->tr_buf);
···
 	if (sdp->sd_log_tr) {
 		gfs2_merge_trans(sdp, tr);
 	} else if (tr->tr_num_buf_new || tr->tr_num_databuf_new) {
-		gfs2_assert_withdraw(sdp, test_bit(TR_ALLOCED, &tr->tr_flags));
+		gfs2_assert_withdraw(sdp, !test_bit(TR_ONSTACK, &tr->tr_flags));
 		sdp->sd_log_tr = tr;
 		set_bit(TR_ATTACHED, &tr->tr_flags);
 	}
 
-	sdp->sd_log_committed_revoke += tr->tr_num_revoke - tr->tr_num_revoke_rm;
 	reserved = calc_reserved(sdp);
 	maxres = sdp->sd_log_blks_reserved + tr->tr_reserved;
 	gfs2_assert_withdraw(sdp, maxres >= reserved);
 	unused = maxres - reserved;
-	atomic_add(unused, &sdp->sd_log_blks_free);
-	trace_gfs2_log_blocks(sdp, unused);
-	gfs2_assert_withdraw(sdp, atomic_read(&sdp->sd_log_blks_free) <=
-			     sdp->sd_jdesc->jd_blocks);
+	if (unused)
+		gfs2_log_release(sdp, unused);
 	sdp->sd_log_blks_reserved = reserved;
 
 	gfs2_log_unlock(sdp);
···
 	gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke);
 	gfs2_assert_withdraw(sdp, list_empty(&sdp->sd_ail1_list));
 
-	sdp->sd_log_flush_head = sdp->sd_log_head;
-
 	log_write_header(sdp, GFS2_LOG_HEAD_UNMOUNT | GFS2_LFC_SHUTDOWN);
+	log_pull_tail(sdp);
 
 	gfs2_assert_warn(sdp, sdp->sd_log_head == sdp->sd_log_tail);
 	gfs2_assert_warn(sdp, list_empty(&sdp->sd_ail2_list));
-
-	sdp->sd_log_head = sdp->sd_log_flush_head;
-	sdp->sd_log_tail = sdp->sd_log_head;
 }
 
 static inline int gfs2_jrnl_flush_reqd(struct gfs2_sbd *sdp)
···
 	struct gfs2_sbd *sdp = data;
 	unsigned long t = 1;
 	DEFINE_WAIT(wait);
-	bool did_flush;
 
 	while (!kthread_should_stop()) {
 
···
 			continue;
 		}
 
-		did_flush = false;
 		if (gfs2_jrnl_flush_reqd(sdp) || t == 0) {
 			gfs2_ail1_empty(sdp, 0);
 			gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL |
-				       GFS2_LFC_LOGD_JFLUSH_REQD);
-			did_flush = true;
+						  GFS2_LFC_LOGD_JFLUSH_REQD);
 		}
 
 		if (gfs2_ail_flush_reqd(sdp)) {
···
 			gfs2_ail1_wait(sdp);
 			gfs2_ail1_empty(sdp, 0);
 			gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL |
-				       GFS2_LFC_LOGD_AIL_FLUSH_REQD);
-			did_flush = true;
+						  GFS2_LFC_LOGD_AIL_FLUSH_REQD);
 		}
-
-		if (!gfs2_ail_flush_reqd(sdp) || did_flush)
-			wake_up(&sdp->sd_log_waitq);
 
 		t = gfs2_tune_get(sdp, gt_logd_secs) * HZ;
 

fs/gfs2/log.h

···
 #include "incore.h"
 #include "inode.h"
 
+/*
+ * The minimum amount of log space required for a log flush is one block for
+ * revokes and one block for the log header.  Log flushes other than
+ * GFS2_LOG_HEAD_FLUSH_NORMAL may write one or two more log headers.
+ */
+#define GFS2_LOG_FLUSH_MIN_BLOCKS 4
+
 /**
  * gfs2_log_lock - acquire the right to mess with the log manager
  * @sdp: the filesystem
···
 	if (++value == sdp->sd_jdesc->jd_blocks) {
 		value = 0;
 	}
-	sdp->sd_log_head = sdp->sd_log_tail = value;
+	sdp->sd_log_tail = value;
+	sdp->sd_log_flush_tail = value;
+	sdp->sd_log_head = value;
 }
 
 static inline void gfs2_ordered_add_inode(struct gfs2_inode *ip)
···
 extern void gfs2_ordered_del_inode(struct gfs2_inode *ip);
 extern unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct);
 extern void gfs2_remove_from_ail(struct gfs2_bufdata *bd);
+extern bool gfs2_log_is_empty(struct gfs2_sbd *sdp);
+extern void gfs2_log_release_revokes(struct gfs2_sbd *sdp, unsigned int revokes);
 extern void gfs2_log_release(struct gfs2_sbd *sdp, unsigned int blks);
-extern int gfs2_log_reserve(struct gfs2_sbd *sdp, unsigned int blks);
+extern bool gfs2_log_try_reserve(struct gfs2_sbd *sdp, struct gfs2_trans *tr,
+				 unsigned int *extra_revokes);
+extern void gfs2_log_reserve(struct gfs2_sbd *sdp, struct gfs2_trans *tr,
+			     unsigned int *extra_revokes);
 extern void gfs2_write_log_header(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd,
 				  u64 seq, u32 tail, u32 lblock, u32 flags,
 				  int op_flags);
···
 extern int gfs2_logd(void *data);
 extern void gfs2_add_revoke(struct gfs2_sbd *sdp, struct gfs2_bufdata *bd);
 extern void gfs2_glock_remove_revoke(struct gfs2_glock *gl);
-extern void gfs2_write_revokes(struct gfs2_sbd *sdp);
+extern void gfs2_flush_revokes(struct gfs2_sbd *sdp);
 
 #endif /* __LOG_DOT_H__ */

fs/gfs2/lops.c

···
 	unsigned int index = bd->bd_bh->b_blocknr - gl->gl_name.ln_number;
 	struct gfs2_bitmap *bi = rgd->rd_bits + index;
 
+	rgrp_lock_local(rgd);
 	if (bi->bi_clone == NULL)
-		return;
+		goto out;
 	if (sdp->sd_args.ar_discard)
 		gfs2_rgrp_send_discards(sdp, rgd->rd_data0, bd->bd_bh, bi, 1, NULL);
 	memcpy(bi->bi_clone + bi->bi_offset,
 	       bd->bd_bh->b_data + bi->bi_offset, bi->bi_bytes);
 	clear_bit(GBF_FULL, &bi->bi_flags);
 	rgd->rd_free_clone = rgd->rd_free;
+	BUG_ON(rgd->rd_free_clone < rgd->rd_reserved);
 	rgd->rd_extfail_pt = rgd->rd_free;
+
+out:
+	rgrp_unlock_local(rgd);
 }
 
 /**
···
  * then add the page segment to that.
  */
 
-void gfs2_log_write(struct gfs2_sbd *sdp, struct page *page,
-		    unsigned size, unsigned offset, u64 blkno)
+void gfs2_log_write(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd,
+		    struct page *page, unsigned size, unsigned offset,
+		    u64 blkno)
 {
 	struct bio *bio;
 	int ret;
 
-	bio = gfs2_log_get_bio(sdp, blkno, &sdp->sd_log_bio, REQ_OP_WRITE,
+	bio = gfs2_log_get_bio(sdp, blkno, &jd->jd_log_bio, REQ_OP_WRITE,
 			       gfs2_end_log_write, false);
 	ret = bio_add_page(bio, page, size, offset);
 	if (ret == 0) {
-		bio = gfs2_log_get_bio(sdp, blkno, &sdp->sd_log_bio,
+		bio = gfs2_log_get_bio(sdp, blkno, &jd->jd_log_bio,
 				       REQ_OP_WRITE, gfs2_end_log_write, true);
 		ret = bio_add_page(bio, page, size, offset);
 		WARN_ON(ret == 0);
···
 
 	dblock = gfs2_log_bmap(sdp->sd_jdesc, sdp->sd_log_flush_head);
 	gfs2_log_incr_head(sdp);
-	gfs2_log_write(sdp, bh->b_page, bh->b_size, bh_offset(bh), dblock);
+	gfs2_log_write(sdp, sdp->sd_jdesc, bh->b_page, bh->b_size,
+		       bh_offset(bh), dblock);
 }
 
 /**
···
  * the page may be freed at any time.
  */
 
-void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page)
+static void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page)
 {
 	struct super_block *sb = sdp->sd_vfs;
 	u64 dblock;
 
 	dblock = gfs2_log_bmap(sdp->sd_jdesc, sdp->sd_log_flush_head);
 	gfs2_log_incr_head(sdp);
-	gfs2_log_write(sdp, page, sb->s_blocksize, 0, dblock);
+	gfs2_log_write(sdp, sdp->sd_jdesc, page, sb->s_blocksize, 0, dblock);
 }
 
 /**
···
 	struct page *page;
 	unsigned int length;
 
-	gfs2_write_revokes(sdp);
+	gfs2_flush_revokes(sdp);
 	if (!sdp->sd_log_num_revoke)
 		return;
 
···
 		sdp->sd_log_num_revoke--;
 
 		if (offset + sizeof(u64) > sdp->sd_sb.sb_bsize) {
-
 			gfs2_log_write_page(sdp, page);
 			page = mempool_alloc(gfs2_page_pool, GFP_NOIO);
 			mh = page_address(page);

fs/gfs2/lops.h

···
 #include <linux/list.h>
 #include "incore.h"
 
-#define BUF_OFFSET \
-	((sizeof(struct gfs2_log_descriptor) + sizeof(__be64) - 1) & \
-	 ~(sizeof(__be64) - 1))
-#define DATABUF_OFFSET \
-	((sizeof(struct gfs2_log_descriptor) + (2 * sizeof(__be64) - 1)) & \
-	 ~(2 * sizeof(__be64) - 1))
-
 extern const struct gfs2_log_operations *gfs2_log_ops[];
 extern void gfs2_log_incr_head(struct gfs2_sbd *sdp);
 extern u64 gfs2_log_bmap(struct gfs2_jdesc *jd, unsigned int lbn);
-extern void gfs2_log_write(struct gfs2_sbd *sdp, struct page *page,
-			   unsigned size, unsigned offset, u64 blkno);
-extern void gfs2_log_write_page(struct gfs2_sbd *sdp, struct page *page);
+extern void gfs2_log_write(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd,
+			   struct page *page, unsigned size, unsigned offset,
+			   u64 blkno);
 extern void gfs2_log_submit_bio(struct bio **biop, int opf);
 extern void gfs2_pin(struct gfs2_sbd *sdp, struct buffer_head *bh);
 extern int gfs2_find_jhead(struct gfs2_jdesc *jd,
 			   struct gfs2_log_header_host *head, bool keep_cache);
 static inline unsigned int buf_limit(struct gfs2_sbd *sdp)
 {
-	unsigned int limit;
-
-	limit = (sdp->sd_sb.sb_bsize - BUF_OFFSET) / sizeof(__be64);
-	return limit;
+	return sdp->sd_ldptrs;
 }
 
 static inline unsigned int databuf_limit(struct gfs2_sbd *sdp)
 {
-	unsigned int limit;
-
-	limit = (sdp->sd_sb.sb_bsize - DATABUF_OFFSET) / (2 * sizeof(__be64));
-	return limit;
+	return sdp->sd_ldptrs / 2;
 }
 
 static inline void lops_before_commit(struct gfs2_sbd *sdp,

fs/gfs2/main.c

···
 	error = -ENOMEM;
 	gfs2_glock_cachep = kmem_cache_create("gfs2_glock",
 					      sizeof(struct gfs2_glock),
-					      0, 0,
+					      0, SLAB_RECLAIM_ACCOUNT,
 					      gfs2_init_glock_once);
 	if (!gfs2_glock_cachep)
 		goto fail_cachep1;
···
 
 	gfs2_quotad_cachep = kmem_cache_create("gfs2_quotad",
 					       sizeof(struct gfs2_quota_data),
-					       0, 0, NULL);
+					       0, SLAB_RECLAIM_ACCOUNT, NULL);
 	if (!gfs2_quotad_cachep)
 		goto fail_cachep6;
 

fs/gfs2/ops_fstype.c

···
 
 	init_rwsem(&sdp->sd_log_flush_lock);
 	atomic_set(&sdp->sd_log_in_flight, 0);
-	atomic_set(&sdp->sd_reserving_log, 0);
-	init_waitqueue_head(&sdp->sd_reserving_log_wait);
 	init_waitqueue_head(&sdp->sd_log_flush_wait);
 	atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN);
 	mutex_init(&sdp->sd_freeze_mutex);
···
 		return -EINVAL;
 	}
 
-	if (sb->sb_fs_format != GFS2_FORMAT_FS ||
+	if (sb->sb_fs_format < GFS2_FS_FORMAT_MIN ||
+	    sb->sb_fs_format > GFS2_FS_FORMAT_MAX ||
 	    sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
 		fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
 		return -EINVAL;
···
 
 	if (sb->sb_bsize < 512 || sb->sb_bsize > PAGE_SIZE ||
 	    (sb->sb_bsize & (sb->sb_bsize - 1))) {
-		pr_warn("Invalid superblock size\n");
+		pr_warn("Invalid block size\n");
 		return -EINVAL;
 	}
 
···
 	sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
 				     sizeof(struct gfs2_meta_header))
 		* GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
+
+	/*
+	 * We always keep at least one block reserved for revokes in
+	 * transactions.  This greatly simplifies allocating additional
+	 * revoke blocks.
+	 */
+	atomic_set(&sdp->sd_log_revokes_available, sdp->sd_ldptrs);
 
 	/* Compute maximum reservation required to add a entry to a directory */
 
···
 	if (ret) {
 		fs_err(sdp, "can't read superblock: %d\n", ret);
 		goto out;
+	}
+
+	switch(sdp->sd_sb.sb_fs_format) {
+	case GFS2_FS_FORMAT_MAX:
+		sb->s_xattr = gfs2_xattr_handlers_max;
+		break;
+
+	case GFS2_FS_FORMAT_MIN:
+		sb->s_xattr = gfs2_xattr_handlers_min;
+		break;
+
+	default:
+		BUG();
 	}
 
 	/* Set up the buffer cache and SB for real */
···
 	}
 
 	if (lm->lm_mount == NULL) {
-		fs_info(sdp, "Now mounting FS...\n");
+		fs_info(sdp, "Now mounting FS (format %u)...\n", sdp->sd_sb.sb_fs_format);
 		complete_all(&sdp->sd_locking_init);
 		return 0;
 	}
 	ret = lm->lm_mount(sdp, table);
 	if (ret == 0)
-		fs_info(sdp, "Joined cluster. Now mounting FS...\n");
+		fs_info(sdp, "Joined cluster. Now mounting FS (format %u)...\n",
+		        sdp->sd_sb.sb_fs_format);
 	complete_all(&sdp->sd_locking_init);
 	return ret;
 }
···
 	int silent = fc->sb_flags & SB_SILENT;
 	struct gfs2_sbd *sdp;
 	struct gfs2_holder mount_gh;
+	struct gfs2_holder freeze_gh;
 	int error;
 
 	sdp = init_sbd(sb);
···
 	sb->s_op = &gfs2_super_ops;
 	sb->s_d_op = &gfs2_dops;
 	sb->s_export_op = &gfs2_export_ops;
-	sb->s_xattr = gfs2_xattr_handlers;
 	sb->s_qcop = &gfs2_quotactl_ops;
 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
 	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
···
 	if (error)
 		goto fail_locking;
 
+	/* Turn rgrplvb on by default if fs format is recent enough */
+	if (!sdp->sd_args.ar_got_rgrplvb && sdp->sd_sb.sb_fs_format > 1801)
+		sdp->sd_args.ar_rgrplvb = 1;
+
 	error = wait_on_journal(sdp);
 	if (error)
 		goto fail_sb;
···
 		goto fail_per_node;
 	}
 
-	if (sb_rdonly(sb)) {
-		struct gfs2_holder freeze_gh;
+	error = gfs2_freeze_lock(sdp, &freeze_gh, 0);
+	if (error)
+		goto fail_per_node;
 
-		error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
-					   LM_FLAG_NOEXP | GL_EXACT,
-					   &freeze_gh);
-		if (error) {
-			fs_err(sdp, "can't make FS RO: %d\n", error);
-			goto fail_per_node;
-		}
-		gfs2_glock_dq_uninit(&freeze_gh);
-	} else {
+	if (!sb_rdonly(sb))
 		error = gfs2_make_fs_rw(sdp);
-		if (error) {
-			fs_err(sdp, "can't make FS RW: %d\n", error);
-			goto fail_per_node;
-		}
-	}
 
+	gfs2_freeze_unlock(&freeze_gh);
+	if (error) {
+		fs_err(sdp, "can't make FS RW: %d\n", error);
+		goto fail_per_node;
+	}
 	gfs2_glock_dq_uninit(&mount_gh);
 	gfs2_online_uevent(sdp);
 	return 0;
···
 		break;
 	case Opt_rgrplvb:
 		args->ar_rgrplvb = result.boolean;
+		args->ar_got_rgrplvb = 1;
 		break;
 	case Opt_loccookie:
 		args->ar_loccookie = result.boolean;
···
 		fc->sb_flags |= SB_RDONLY;
 
 	if ((sb->s_flags ^ fc->sb_flags) & SB_RDONLY) {
+		struct gfs2_holder freeze_gh;
+
+		error = gfs2_freeze_lock(sdp, &freeze_gh, 0);
+		if (error)
+			return -EINVAL;
+
 		if (fc->sb_flags & SB_RDONLY) {
 			error = gfs2_make_fs_ro(sdp);
 			if (error)
···
 			if (error)
 				errorfc(fc, "unable to remount read-write");
 		}
+		gfs2_freeze_unlock(&freeze_gh);
 	}
 	sdp->sd_args = *newargs;
 

fs/gfs2/recovery.c

···
 
 		/* Acquire a shared hold on the freeze lock */
 
-		error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
-					   LM_FLAG_NOEXP | LM_FLAG_PRIORITY |
-					   GL_EXACT, &thaw_gh);
+		error = gfs2_freeze_lock(sdp, &thaw_gh, LM_FLAG_PRIORITY);
 		if (error)
 			goto fail_gunlock_ji;
 
···
 
 		/* We take the sd_log_flush_lock here primarily to prevent log
 		 * flushes and simultaneous journal replays from stomping on
-		 * each other wrt sd_log_bio. */
+		 * each other wrt jd_log_bio. */
 		down_read(&sdp->sd_log_flush_lock);
 		for (pass = 0; pass < 2; pass++) {
 			lops_before_scan(jd, &head, pass);
 			error = foreach_descriptor(jd, head.lh_tail,
 						   head.lh_blkno, pass);
 			lops_after_scan(jd, error, pass);
-			if (error)
+			if (error) {
+				up_read(&sdp->sd_log_flush_lock);
 				goto fail_gunlock_thaw;
+			}
 		}
 
 		recover_local_statfs(jd, &head);
 		clean_journal(jd, &head);
 		up_read(&sdp->sd_log_flush_lock);
 
-		gfs2_glock_dq_uninit(&thaw_gh);
+		gfs2_freeze_unlock(&thaw_gh);
 		t_rep = ktime_get();
 		fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
 			"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
···
 	goto done;
 
 fail_gunlock_thaw:
-	gfs2_glock_dq_uninit(&thaw_gh);
+	gfs2_freeze_unlock(&thaw_gh);
 fail_gunlock_ji:
 	if (jlocked) {
 		gfs2_glock_dq_uninit(&ji_gh);

fs/gfs2/rgrp.c

···
 #define BFITNOENT ((u32)~0)
 #define NO_BLOCK ((u64)~0)
 
+struct gfs2_rbm {
+	struct gfs2_rgrpd *rgd;
+	u32 offset;		/* The offset is bitmap relative */
+	int bii;		/* Bitmap index */
+};
+
+static inline struct gfs2_bitmap *rbm_bi(const struct gfs2_rbm *rbm)
+{
+	return rbm->rgd->rd_bits + rbm->bii;
+}
+
+static inline u64 gfs2_rbm_to_block(const struct gfs2_rbm *rbm)
+{
+	BUG_ON(rbm->offset >= rbm->rgd->rd_data);
+	return rbm->rgd->rd_data0 + (rbm_bi(rbm)->bi_start * GFS2_NBBY) +
+		rbm->offset;
+}
+
 /*
  * These routines are used by the resource group routines (rgrp.c)
  * to keep track of block allocation.  Each block is represented by two
···
 };
 
 static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 *minext,
-			 const struct gfs2_inode *ip, bool nowrap);
+			 struct gfs2_blkreserv *rs, bool nowrap);
 
 
 /**
···
 
 /**
  * rs_cmp - multi-block reservation range compare
- * @blk: absolute file system block number of the new reservation
+ * @start: start of the new reservation
  * @len: number of blocks in the new reservation
  * @rs: existing reservation to compare against
  *
···
  *         -1 if the block range is before the start of the reservation
  *          0 if the block range overlaps with the reservation
  */
-static inline int rs_cmp(u64 blk, u32 len, struct gfs2_blkreserv *rs)
+static inline int rs_cmp(u64 start, u32 len, struct gfs2_blkreserv *rs)
 {
-	u64 startblk = gfs2_rbm_to_block(&rs->rs_rbm);
-
-	if (blk >= startblk + rs->rs_free)
+	if (start >= rs->rs_start + rs->rs_requested)
 		return 1;
-	if (blk + len - 1 < startblk)
+	if (rs->rs_start >= start + len)
 		return -1;
 	return 0;
 }
···
 }
 
 /**
- * gfs2_rbm_incr - increment an rbm structure
+ * gfs2_rbm_add - add a number of blocks to an rbm
  * @rbm: The rbm with rgd already set correctly
+ * @blocks: The number of blocks to add to rpm
  *
- * This function takes an existing rbm structure and increments it to the next
- * viable block offset.
+ * This function takes an existing rbm structure and adds a number of blocks to
+ * it.
  *
- * Returns: If incrementing the offset would cause the rbm to go past the
- *          end of the rgrp, true is returned, otherwise false.
- *
+ * Returns: True if the new rbm would point past the end of the rgrp.
  */
 
-static bool gfs2_rbm_incr(struct gfs2_rbm *rbm)
+static bool gfs2_rbm_add(struct gfs2_rbm *rbm, u32 blocks)
 {
-	if (rbm->offset + 1 < rbm_bi(rbm)->bi_blocks) { /* in the same bitmap */
-		rbm->offset++;
+	struct gfs2_rgrpd *rgd = rbm->rgd;
+	struct gfs2_bitmap *bi = rgd->rd_bits + rbm->bii;
+
+	if (rbm->offset + blocks < bi->bi_blocks) {
+		rbm->offset += blocks;
 		return false;
 	}
-	if (rbm->bii == rbm->rgd->rd_length - 1) /* at the last bitmap */
-		return true;
+	blocks -= bi->bi_blocks - rbm->offset;
 
-	rbm->offset = 0;
-	rbm->bii++;
-	return false;
+	for(;;) {
+		bi++;
+		if (bi == rgd->rd_bits + rgd->rd_length)
+			return true;
+		if (blocks < bi->bi_blocks) {
+			rbm->offset = blocks;
+			rbm->bii = bi - rgd->rd_bits;
+			return false;
+		}
+		blocks -= bi->bi_blocks;
+	}
 }
 
 /**
···
  * @n_unaligned: Number of unaligned blocks to check
  * @len: Decremented for each block found (terminate on zero)
  *
- * Returns: true if a non-free block is encountered
+ * Returns: true if a non-free block is encountered or the end of the resource
+ *	    group is reached.
  */
 
 static bool gfs2_unaligned_extlen(struct gfs2_rbm *rbm, u32 n_unaligned, u32 *len)
···
 		(*len)--;
 		if (*len == 0)
 			return true;
-		if (gfs2_rbm_incr(rbm))
+		if (gfs2_rbm_add(rbm, 1))
 			return true;
 	}
 
···
 {
 	struct gfs2_inode *ip = container_of(rs, struct gfs2_inode, i_res);
 
-	gfs2_print_dbg(seq, "%s  B: n:%llu s:%llu b:%u f:%u\n", fs_id_buf,
+	gfs2_print_dbg(seq, "%s  B: n:%llu s:%llu f:%u\n",
+		       fs_id_buf,
 		       (unsigned long long)ip->i_no_addr,
-		       (unsigned long long)gfs2_rbm_to_block(&rs->rs_rbm),
-		       rs->rs_rbm.offset, rs->rs_free);
+		       (unsigned long long)rs->rs_start,
+		       rs->rs_requested);
 }
 
 /**
···
 	if (!gfs2_rs_active(rs))
 		return;
 
-	rgd = rs->rs_rbm.rgd;
+	rgd = rs->rs_rgd;
 	trace_gfs2_rs(rs, TRACE_RS_TREEDEL);
 	rb_erase(&rs->rs_node, &rgd->rd_rstree);
 	RB_CLEAR_NODE(&rs->rs_node);
 
-	if (rs->rs_free) {
-		u64 last_block = gfs2_rbm_to_block(&rs->rs_rbm) +
-				 rs->rs_free - 1;
-		struct gfs2_rbm last_rbm = { .rgd = rs->rs_rbm.rgd, };
-		struct gfs2_bitmap *start, *last;
+	if (rs->rs_requested) {
+		/* return requested blocks to the rgrp */
+		BUG_ON(rs->rs_rgd->rd_requested < rs->rs_requested);
+		rs->rs_rgd->rd_requested -= rs->rs_requested;
 
-		/* return reserved blocks to the rgrp */
-		BUG_ON(rs->rs_rbm.rgd->rd_reserved < rs->rs_free);
-		rs->rs_rbm.rgd->rd_reserved -= rs->rs_free;
 		/* The rgrp extent failure point is likely not to increase;
 		   it will only do so if the freed blocks are somehow
 		   contiguous with a span of free blocks that follows. Still,
 		   it will force the number to be recalculated later. */
-		rgd->rd_extfail_pt += rs->rs_free;
-		rs->rs_free = 0;
-		if (gfs2_rbm_from_block(&last_rbm, last_block))
-			return;
-		start = rbm_bi(&rs->rs_rbm);
-		last = rbm_bi(&last_rbm);
-		do
-			clear_bit(GBF_FULL, &start->bi_flags);
-		while (start++ != last);
+		rgd->rd_extfail_pt += rs->rs_requested;
+		rs->rs_requested = 0;
 	}
 }
 
···
 {
 	struct gfs2_rgrpd *rgd;
 
-	rgd = rs->rs_rbm.rgd;
+	rgd = rs->rs_rgd;
 	if (rgd) {
 		spin_lock(&rgd->rd_rsspin);
 		__rs_deltree(rs);
-		BUG_ON(rs->rs_free);
+		BUG_ON(rs->rs_requested);
 		spin_unlock(&rgd->rd_rsspin);
 	}
 }
···
 	rgd->rd_data = be32_to_cpu(buf.ri_data);
 	rgd->rd_bitbytes = be32_to_cpu(buf.ri_bitbytes);
 	spin_lock_init(&rgd->rd_rsspin);
+	mutex_init(&rgd->rd_mutex);
 
 	error = compute_bitstructs(rgd);
 	if (error)
···
 	return count;
 }
 
+static void rgrp_set_bitmap_flags(struct gfs2_rgrpd *rgd)
+{
+	struct gfs2_bitmap *bi;
+	int x;
+
+	if (rgd->rd_free) {
+		for (x = 0; x < rgd->rd_length; x++) {
+			bi = rgd->rd_bits + x;
+			clear_bit(GBF_FULL, &bi->bi_flags);
+		}
+	} else {
+		for (x = 0; x < rgd->rd_length; x++) {
+			bi = rgd->rd_bits + x;
+			set_bit(GBF_FULL, &bi->bi_flags);
+		}
+	}
+}
 
 /**
  * gfs2_rgrp_bh_get - Read in a RG's header and bitmaps
···
 	}
 
 	if (!(rgd->rd_flags & GFS2_RDF_UPTODATE)) {
-		for (x = 0; x < length; x++)
-			clear_bit(GBF_FULL, &rgd->rd_bits[x].bi_flags);
 		gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data);
+		rgrp_set_bitmap_flags(rgd);
 		rgd->rd_flags |= (GFS2_RDF_UPTODATE | GFS2_RDF_CHECK);
 		rgd->rd_free_clone = rgd->rd_free;
+		BUG_ON(rgd->rd_reserved);
 		/* max out the rgrp allocation failure point */
 		rgd->rd_extfail_pt = rgd->rd_free;
 	}
···
 	if (rgd->rd_rgl->rl_unlinked == 0)
 		rgd->rd_flags &= ~GFS2_RDF_CHECK;
 	rgd->rd_free = be32_to_cpu(rgd->rd_rgl->rl_free);
+	rgrp_set_bitmap_flags(rgd);
 	rgd->rd_free_clone = rgd->rd_free;
+	BUG_ON(rgd->rd_reserved);
+	/* max out the rgrp allocation failure point */
+	rgd->rd_extfail_pt = rgd->rd_free;
 	rgd->rd_dinodes = be32_to_cpu(rgd->rd_rgl->rl_dinodes);
 	rgd->rd_igeneration = be64_to_cpu(rgd->rd_rgl->rl_igeneration);
 	return 0;
···
 
 	while (1) {
 
-		ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh);
+		ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE,
+					 LM_FLAG_NODE_SCOPE, &gh);
 		if (ret)
 			goto out;
 
···
 			/* Trim each bitmap in the rgrp */
 			for (x = 0; x < rgd->rd_length; x++) {
 				struct gfs2_bitmap *bi = rgd->rd_bits + x;
+				rgrp_lock_local(rgd);
 				ret = gfs2_rgrp_send_discards(sdp,
 						rgd->rd_data0, NULL, bi, minlen,
 						&amt);
+				rgrp_unlock_local(rgd);
 				if (ret) {
 					gfs2_glock_dq_uninit(&gh);
 					goto out;
···
 			ret = gfs2_trans_begin(sdp, RES_RG_HDR, 0);
 			if (ret == 0) {
 				bh = rgd->rd_bits[0].bi_bh;
+				rgrp_lock_local(rgd);
 				rgd->rd_flags |= GFS2_RGF_TRIMMED;
 				gfs2_trans_add_meta(rgd->rd_gl, bh);
 				gfs2_rgrp_out(rgd, bh->b_data);
+				rgrp_unlock_local(rgd);
 				gfs2_trans_end(sdp);
 			}
 		}
···
 	struct rb_node **newn, *parent = NULL;
 	int rc;
 	struct gfs2_blkreserv *rs = &ip->i_res;
-	struct gfs2_rgrpd *rgd = rs->rs_rbm.rgd;
-	u64 fsblock = gfs2_rbm_to_block(&rs->rs_rbm);
+	struct gfs2_rgrpd *rgd = rs->rs_rgd;
 
 	BUG_ON(gfs2_rs_active(rs));
 
···
 			rb_entry(*newn, struct gfs2_blkreserv, rs_node);
 
 		parent = *newn;
-		rc = rs_cmp(fsblock, rs->rs_free, cur);
+		rc = rs_cmp(rs->rs_start, rs->rs_requested, cur);
 		if (rc > 0)
 			newn = &((*newn)->rb_right);
 		else if (rc < 0)
···
 	rb_insert_color(&rs->rs_node, &rgd->rd_rstree);
 
 	/* Do our rgrp accounting for the reservation */
-	rgd->rd_reserved += rs->rs_free; /* blocks reserved */
+	rgd->rd_requested += rs->rs_requested; /* blocks requested */
 	spin_unlock(&rgd->rd_rsspin);
 	trace_gfs2_rs(rs, TRACE_RS_INSERT);
 }
···
 {
 	u32 tot_reserved, tot_free;
 
-	if (WARN_ON_ONCE(rgd->rd_reserved < rs->rs_free))
+	if (WARN_ON_ONCE(rgd->rd_requested < rs->rs_requested))
 		return 0;
-	tot_reserved = rgd->rd_reserved - rs->rs_free;
+	tot_reserved = rgd->rd_requested - rs->rs_requested;
 
 	if (rgd->rd_free_clone < tot_reserved)
 		tot_reserved = 0;
···
 	u64 goal;
 	struct gfs2_blkreserv *rs = &ip->i_res;
 	u32 extlen;
-	u32 free_blocks = rgd_free(rgd, rs);
+	u32 free_blocks, blocks_available;
 	int ret;
 	struct inode *inode = &ip->i_inode;
+
+	spin_lock(&rgd->rd_rsspin);
+	free_blocks = rgd_free(rgd, rs);
+	if (rgd->rd_free_clone < rgd->rd_requested)
+		free_blocks = 0;
+	blocks_available = rgd->rd_free_clone - rgd->rd_reserved;
+	if (rgd == rs->rs_rgd)
+		blocks_available += rs->rs_reserved;
+	spin_unlock(&rgd->rd_rsspin);
 
 	if (S_ISDIR(inode->i_mode))
 		extlen = 1;
···
 		extlen = max_t(u32, atomic_read(&ip->i_sizehint), ap->target);
 		extlen = clamp(extlen, (u32)RGRP_RSRV_MINBLKS, free_blocks);
 	}
-	if ((rgd->rd_free_clone < rgd->rd_reserved) || (free_blocks < extlen))
+	if (free_blocks < extlen || blocks_available < extlen)
 		return;
 
 	/* Find bitmap block that contains bits for goal block */
···
 	if (WARN_ON(gfs2_rbm_from_block(&rbm, goal)))
 		return;
 
-	ret = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, &extlen, ip, true);
+	ret = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, &extlen, &ip->i_res, true);
 	if (ret == 0) {
-		rs->rs_rbm = rbm;
-		rs->rs_free = extlen;
+		rs->rs_start = gfs2_rbm_to_block(&rbm);
+		rs->rs_requested = extlen;
 		rs_insert(ip);
 	} else {
 		if (goal == rgd->rd_last_alloc + rgd->rd_data0)
···
  * @rgd: The resource group
  * @block: The starting block
  * @length: The required length
- * @ip: Ignore any reservations for this inode
+ * @ignore_rs: Reservation to ignore
  *
  * If the block does not appear in any reservation, then return the
  * block number unchanged. If it does appear in the reservation, then
···
 
 static u64 gfs2_next_unreserved_block(struct gfs2_rgrpd *rgd, u64 block,
 				      u32 length,
-				      const struct gfs2_inode *ip)
+				      struct gfs2_blkreserv *ignore_rs)
 {
 	struct gfs2_blkreserv *rs;
 	struct rb_node *n;
···
 	}
 
 	if (n) {
-		while ((rs_cmp(block, length, rs) == 0) && (&ip->i_res != rs)) {
-			block = gfs2_rbm_to_block(&rs->rs_rbm) + rs->rs_free;
+		while (rs_cmp(block, length, rs) == 0 && rs != ignore_rs) {
+			block = rs->rs_start + rs->rs_requested;
 			n = n->rb_right;
 			if (n == NULL)
 				break;
···
 /**
  * gfs2_reservation_check_and_update - Check for reservations during block alloc
  * @rbm: The current position in the resource group
- * @ip: The inode for which we are searching for blocks
+ * @rs: Our own reservation
  * @minext: The minimum extent length
  * @maxext: A pointer to the maximum extent structure
  *
···
  */
 
 static int gfs2_reservation_check_and_update(struct gfs2_rbm *rbm,
-					     const struct gfs2_inode *ip,
+					     struct gfs2_blkreserv *rs,
 					     u32 minext,
 					     struct gfs2_extent *maxext)
 {
 	u64 block = gfs2_rbm_to_block(rbm);
 	u32 extlen = 1;
 	u64 nblock;
-	int ret;
 
 	/*
 	 * If we have a minimum extent length, then skip over any extent
 	 * which is less than the min extent length in size.
 	 */
-	if (minext) {
+	if (minext > 1) {
 		extlen = gfs2_free_extlen(rbm, minext);
 		if (extlen <= maxext->len)
 			goto fail;
···
 	 * Check the extent which has been found against the reservations
 	 * and skip if parts of it are already reserved
 	 */
-	nblock = gfs2_next_unreserved_block(rbm->rgd, block, extlen, ip);
+	nblock = gfs2_next_unreserved_block(rbm->rgd, block, extlen, rs);
 	if (nblock == block) {
 		if (!minext || extlen >= minext)
 			return 0;
···
 			maxext->len = extlen;
 			maxext->rbm = *rbm;
 		}
-fail:
-		nblock = block + extlen;
+	} else {
+		u64 len = nblock - block;
+		if (len >= (u64)1 << 32)
+			return -E2BIG;
+		extlen = len;
 	}
-	ret = gfs2_rbm_from_block(rbm, nblock);
-	if (ret < 0)
-		return ret;
+fail:
+	if (gfs2_rbm_add(rbm, extlen))
+		return -E2BIG;
 	return 1;
 }
 
···
  * gfs2_rbm_find - Look for blocks of a particular state
  * @rbm: Value/result starting position and final position
  * @state: The state which we want to find
- * @minext: Pointer to the requested extent length (NULL for a single block)
+ * @minext: Pointer to the requested extent length
  *          This is updated to be the actual reservation size.
- * @ip: If set, check for reservations
+ * @rs: Our own reservation (NULL to skip checking for reservations)
  * @nowrap: Stop looking at the end of the rgrp, rather than wrapping
  *          around until we've reached the starting point.
  *
···
  */
 
 static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 *minext,
-			 const struct gfs2_inode *ip, bool nowrap)
+			 struct gfs2_blkreserv *rs, bool nowrap)
 {
 	bool scan_from_start = rbm->bii == 0 && rbm->offset == 0;
 	struct buffer_head *bh;
···
 
 	while(1) {
 		bi = rbm_bi(rbm);
-		if ((ip == NULL || !gfs2_rs_active(&ip->i_res)) &&
-		    test_bit(GBF_FULL, &bi->bi_flags) &&
+		if (test_bit(GBF_FULL, &bi->bi_flags) &&
 		    (state == GFS2_BLKST_FREE))
 			goto next_bitmap;
 
···
 			goto next_bitmap;
 		}
 		rbm->offset = offset;
-		if (ip == NULL)
+		if (!rs)
 			return 0;
 
-		ret = gfs2_reservation_check_and_update(rbm, ip,
-							minext ? *minext : 0,
+		ret = gfs2_reservation_check_and_update(rbm, rs, *minext,
 							&maxext);
 		if (ret == 0)
 			return 0;
···
 			break;
 	}
 
-	if (minext == NULL || state != GFS2_BLKST_FREE)
+	if (state != GFS2_BLKST_FREE)
 		return -ENOSPC;
 
 	/* If the extent was too small, and it's smaller than the smallest
···
 	   useless to search this rgrp again for this amount or more. */
 	if (wrapped && (scan_from_start || rbm->bii > last_bii) &&
 	    *minext < rbm->rgd->rd_extfail_pt)
-		rbm->rgd->rd_extfail_pt = *minext;
+		rbm->rgd->rd_extfail_pt = *minext - 1;
 
 	/* If the maximum extent we found is big enough to fulfill the
 	   minimum requirements, use it anyway. */
···
 	u64 tdiff;
 
 	tdiff = ktime_to_ns(ktime_sub(ktime_get_real(),
-                            rs->rs_rbm.rgd->rd_gl->gl_dstamp));
+                            rs->rs_rgd->rd_gl->gl_dstamp));
 
 	return tdiff > (msecs * 1000 * 1000);
 }
···
  * We try our best to find an rgrp that has at least ap->target blocks
  * available. After a couple of passes (loops == 2), the prospects of finding
  * such an rgrp diminish. At this stage, we return the first rgrp that has
- * at least ap->min_target blocks available. Either way, we set ap->allowed to
- * the number of blocks available in the chosen rgrp.
+ * at least ap->min_target blocks available.
  *
  * Returns: 0 on success,
  *          -ENOMEM if a suitable rgrp can't be found
···
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_rgrpd *begin = NULL;
 	struct gfs2_blkreserv *rs = &ip->i_res;
-	int error = 0, rg_locked, flags = 0;
+	int error = 0, flags = LM_FLAG_NODE_SCOPE;
+	bool rg_locked;
 	u64 last_unlinked = NO_BLOCK;
+	u32 target = ap->target;
 	int loops = 0;
-	u32 free_blocks, skip = 0;
+	u32 free_blocks, blocks_available, skip = 0;
+
+	BUG_ON(rs->rs_reserved);
 
 	if (sdp->sd_args.ar_rgrplvb)
 		flags |= GL_SKIP;
-	if (gfs2_assert_warn(sdp, ap->target))
+	if (gfs2_assert_warn(sdp, target))
 		return -EINVAL;
 	if (gfs2_rs_active(rs)) {
-		begin = rs->rs_rbm.rgd;
-	} else if (rs->rs_rbm.rgd &&
-		   rgrp_contains_block(rs->rs_rbm.rgd, ip->i_goal)) {
-		begin = rs->rs_rbm.rgd;
+		begin = rs->rs_rgd;
+	} else if (rs->rs_rgd &&
+		   rgrp_contains_block(rs->rs_rgd, ip->i_goal)) {
+		begin = rs->rs_rgd;
 	} else {
 		check_and_update_goal(ip);
-		rs->rs_rbm.rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1);
+		rs->rs_rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1);
 	}
 	if (S_ISDIR(ip->i_inode.i_mode) && (ap->aflags & GFS2_AF_ORLOV))
 		skip = gfs2_orlov_skip(ip);
-	if (rs->rs_rbm.rgd == NULL)
+	if (rs->rs_rgd == NULL)
 		return -EBADSLT;
 
 	while (loops < 3) {
-		rg_locked = 1;
+		struct gfs2_rgrpd *rgd;
 
-		if (!gfs2_glock_is_locked_by_me(rs->rs_rbm.rgd->rd_gl)) {
-			rg_locked = 0;
+		rg_locked = gfs2_glock_is_locked_by_me(rs->rs_rgd->rd_gl);
+		if (rg_locked) {
+			rgrp_lock_local(rs->rs_rgd);
+		} else {
 			if (skip && skip--)
 				goto next_rgrp;
 			if (!gfs2_rs_active(rs)) {
 				if (loops == 0 &&
-				    !fast_to_acquire(rs->rs_rbm.rgd))
+				    !fast_to_acquire(rs->rs_rgd))
 					goto next_rgrp;
 				if ((loops < 2) &&
 				    gfs2_rgrp_used_recently(rs, 1000) &&
-				    gfs2_rgrp_congested(rs->rs_rbm.rgd, loops))
+				    gfs2_rgrp_congested(rs->rs_rgd, loops))
 					goto next_rgrp;
 			}
-			error = gfs2_glock_nq_init(rs->rs_rbm.rgd->rd_gl,
+			error = gfs2_glock_nq_init(rs->rs_rgd->rd_gl,
 						   LM_ST_EXCLUSIVE, flags,
 						   &ip->i_rgd_gh);
 			if (unlikely(error))
 				return error;
+			rgrp_lock_local(rs->rs_rgd);
 			if (!gfs2_rs_active(rs) && (loops < 2) &&
-			    gfs2_rgrp_congested(rs->rs_rbm.rgd, loops))
+			    gfs2_rgrp_congested(rs->rs_rgd, loops))
 				goto skip_rgrp;
 			if (sdp->sd_args.ar_rgrplvb) {
-				error = update_rgrp_lvb(rs->rs_rbm.rgd);
+				error = update_rgrp_lvb(rs->rs_rgd);
 				if (unlikely(error)) {
+					rgrp_unlock_local(rs->rs_rgd);
 					gfs2_glock_dq_uninit(&ip->i_rgd_gh);
 					return error;
 				}
···
 		}
 
 		/* Skip unusable resource groups */
-		if ((rs->rs_rbm.rgd->rd_flags & (GFS2_RGF_NOALLOC |
+		if ((rs->rs_rgd->rd_flags & (GFS2_RGF_NOALLOC |
 						 GFS2_RDF_ERROR)) ||
-		    (loops == 0 && ap->target > rs->rs_rbm.rgd->rd_extfail_pt))
+		    (loops == 0 && target > rs->rs_rgd->rd_extfail_pt))
 			goto skip_rgrp;
 
 		if (sdp->sd_args.ar_rgrplvb)
-			gfs2_rgrp_bh_get(rs->rs_rbm.rgd);
+			gfs2_rgrp_bh_get(rs->rs_rgd);
 
 		/* Get a reservation if we don't already have one */
 		if (!gfs2_rs_active(rs))
-			rg_mblk_search(rs->rs_rbm.rgd, ip, ap);
+			rg_mblk_search(rs->rs_rgd, ip, ap);
 
 		/* Skip rgrps when we can't get a reservation on first pass */
 		if (!gfs2_rs_active(rs) && (loops < 1))
 			goto check_rgrp;
 
 		/* If rgrp has enough free space, use it */
-		free_blocks = rgd_free(rs->rs_rbm.rgd, rs);
-		if (free_blocks >= ap->target ||
-		    (loops == 2 && ap->min_target &&
-		     free_blocks >= ap->min_target)) {
-			ap->allowed = free_blocks;
-			return 0;
+		rgd = rs->rs_rgd;
+		spin_lock(&rgd->rd_rsspin);
+		free_blocks = rgd_free(rgd, rs);
+		blocks_available = rgd->rd_free_clone - rgd->rd_reserved;
+		if (free_blocks < target || blocks_available < target) {
+			spin_unlock(&rgd->rd_rsspin);
+			goto check_rgrp;
 		}
+		rs->rs_reserved = ap->target;
+		if (rs->rs_reserved > blocks_available)
+			rs->rs_reserved = blocks_available;
+		rgd->rd_reserved += rs->rs_reserved;
+		spin_unlock(&rgd->rd_rsspin);
+		rgrp_unlock_local(rs->rs_rgd);
+		return 0;
 check_rgrp:
 		/* Check for unlinked inodes which can be reclaimed */
-		if (rs->rs_rbm.rgd->rd_flags & GFS2_RDF_CHECK)
-			try_rgrp_unlink(rs->rs_rbm.rgd, &last_unlinked,
+		if (rs->rs_rgd->rd_flags & GFS2_RDF_CHECK)
+			try_rgrp_unlink(rs->rs_rgd, &last_unlinked,
 					ip->i_no_addr);
 skip_rgrp:
+		rgrp_unlock_local(rs->rs_rgd);
+
 		/* Drop reservation, if we couldn't use reserved rgrp */
 		if (gfs2_rs_active(rs))
 			gfs2_rs_deltree(rs);
···
 			gfs2_glock_dq_uninit(&ip->i_rgd_gh);
 next_rgrp:
 		/* Find the next rgrp, and continue looking */
-		if (gfs2_select_rgrp(&rs->rs_rbm.rgd, begin))
+		if (gfs2_select_rgrp(&rs->rs_rgd, begin))
 			continue;
 		if (skip)
 			continue;
···
 				return error;
 		}
 		/* Flushing the log may release space */
-		if (loops == 2)
+		if (loops == 2) {
+			if (ap->min_target)
+				target = ap->min_target;
 			gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL |
 				       GFS2_LFC_INPLACE_RESERVE);
+		}
 	}
 
 	return -ENOSPC;
···
 
 void gfs2_inplace_release(struct gfs2_inode *ip)
 {
+	struct gfs2_blkreserv *rs = &ip->i_res;
+
+	if (rs->rs_reserved) {
+		struct gfs2_rgrpd *rgd = rs->rs_rgd;
+
+		spin_lock(&rgd->rd_rsspin);
+		BUG_ON(rgd->rd_reserved < rs->rs_reserved);
+		rgd->rd_reserved -= rs->rs_reserved;
+		spin_unlock(&rgd->rd_rsspin);
+		rs->rs_reserved = 0;
+	}
 	if (gfs2_holder_initialized(&ip->i_rgd_gh))
 		gfs2_glock_dq_uninit(&ip->i_rgd_gh);
 }
···
 			bi_prev = bi;
 		}
 		gfs2_setbit(&rbm, false, new_state);
-		gfs2_rbm_incr(&rbm);
+		gfs2_rbm_add(&rbm, 1);
 	}
 }
 
···
 	struct gfs2_blkreserv *trs;
 	const struct rb_node *n;
 
-	gfs2_print_dbg(seq, "%s R: n:%llu f:%02x b:%u/%u i:%u r:%u e:%u\n",
+	spin_lock(&rgd->rd_rsspin);
+	gfs2_print_dbg(seq, "%s R: n:%llu f:%02x b:%u/%u i:%u q:%u r:%u e:%u\n",
 		       fs_id_buf,
 		       (unsigned long long)rgd->rd_addr, rgd->rd_flags,
 		       rgd->rd_free, rgd->rd_free_clone, rgd->rd_dinodes,
-		       rgd->rd_reserved, rgd->rd_extfail_pt);
+		       rgd->rd_requested, rgd->rd_reserved, rgd->rd_extfail_pt);
 	if (rgd->rd_sbd->sd_args.ar_rgrplvb) {
 		struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl;
 
···
 			       be32_to_cpu(rgl->rl_free),
 			       be32_to_cpu(rgl->rl_dinodes));
 	}
-	spin_lock(&rgd->rd_rsspin);
 	for (n = rb_first(&rgd->rd_rstree); n; n = rb_next(&trs->rs_node)) {
 		trs = rb_entry(n, struct gfs2_blkreserv, rs_node);
 		dump_rs(seq, trs, fs_id_buf);
···
 {
 	struct gfs2_blkreserv *rs = &ip->i_res;
 	struct gfs2_rgrpd *rgd = rbm->rgd;
-	unsigned rlen;
-	u64 block;
-	int ret;
 
-	spin_lock(&rgd->rd_rsspin);
+	BUG_ON(rs->rs_reserved < len);
+	rs->rs_reserved -= len;
 	if (gfs2_rs_active(rs)) {
-		if (gfs2_rbm_eq(&rs->rs_rbm, rbm)) {
-			block = gfs2_rbm_to_block(rbm);
-			ret = gfs2_rbm_from_block(&rs->rs_rbm, block + len);
-			rlen = min(rs->rs_free, len);
-			rs->rs_free -= rlen;
-			rgd->rd_reserved -= rlen;
+		u64 start = gfs2_rbm_to_block(rbm);
+
+		if (rs->rs_start == start) {
+			unsigned int rlen;
+
+			rs->rs_start += len;
+			rlen = min(rs->rs_requested, len);
+			rs->rs_requested -= rlen;
+			rgd->rd_requested -= rlen;
 			trace_gfs2_rs(rs, TRACE_RS_CLAIM);
-			if (rs->rs_free && !ret)
-				goto out;
+			if (rs->rs_start < rgd->rd_data0 + rgd->rd_data &&
+			    rs->rs_requested)
+				return;
 			/* We used up our block reservation, so we should
 			   reserve more blocks next time. */
 			atomic_add(RGRP_RSRV_ADDBLKS, &ip->i_sizehint);
 		}
 		__rs_deltree(rs);
 	}
-out:
-	spin_unlock(&rgd->rd_rsspin);
 }
 
 /**
···
 	u64 goal;
 
 	if (gfs2_rs_active(&ip->i_res)) {
-		*rbm = ip->i_res.rs_rbm;
-		return;
+		goal = ip->i_res.rs_start;
+	} else {
+		if (!dinode && rgrp_contains_block(rbm->rgd, ip->i_goal))
+			goal = ip->i_goal;
+		else
+			goal = rbm->rgd->rd_last_alloc + rbm->rgd->rd_data0;
 	}
-
-	if (!dinode && rgrp_contains_block(rbm->rgd, ip->i_goal))
-		goal = ip->i_goal;
-	else
-		goal = rbm->rgd->rd_last_alloc + rbm->rgd->rd_data0;
-
 	if (WARN_ON_ONCE(gfs2_rbm_from_block(rbm, goal))) {
 		rbm->bii = 0;
 		rbm->offset = 0;
···
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct buffer_head *dibh;
-	struct gfs2_rbm rbm = { .rgd = ip->i_res.rs_rbm.rgd, };
-	unsigned int ndata;
+	struct gfs2_rbm rbm = { .rgd = ip->i_res.rs_rgd, };
 	u64 block; /* block, within the file system scope */
-	int error;
+	u32 minext = 1;
+	int error = -ENOSPC;
 
-	gfs2_set_alloc_start(&rbm, ip, dinode);
-	error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, NULL, ip, false);
+	BUG_ON(ip->i_res.rs_reserved < *nblocks);
 
+	rgrp_lock_local(rbm.rgd);
+	if (gfs2_rs_active(&ip->i_res)) {
+		gfs2_set_alloc_start(&rbm, ip, dinode);
+		error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, &minext, &ip->i_res, false);
+	}
 	if (error == -ENOSPC) {
 		gfs2_set_alloc_start(&rbm, ip, dinode);
-		error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, NULL, NULL, false);
+		error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, &minext, NULL, false);
 	}
 
 	/* Since all blocks are reserved in advance, this shouldn't happen */
···
 	gfs2_alloc_extent(&rbm, dinode, nblocks);
 	block = gfs2_rbm_to_block(&rbm);
 	rbm.rgd->rd_last_alloc = block - rbm.rgd->rd_data0;
-	if (gfs2_rs_active(&ip->i_res))
-		gfs2_adjust_reservation(ip, &rbm, *nblocks);
-	ndata = *nblocks;
-	if (dinode)
-		ndata--;
-
 	if (!dinode) {
-		ip->i_goal = block + ndata - 1;
+		ip->i_goal = block + *nblocks - 1;
 		error = gfs2_meta_inode_buffer(ip, &dibh);
 		if (error == 0) {
 			struct gfs2_dinode *di =
···
 			brelse(dibh);
 		}
 	}
-	if (rbm.rgd->rd_free < *nblocks) {
+	spin_lock(&rbm.rgd->rd_rsspin);
+	gfs2_adjust_reservation(ip, &rbm, *nblocks);
+	if (rbm.rgd->rd_free < *nblocks || rbm.rgd->rd_reserved < *nblocks) {
 		fs_warn(sdp, "nblocks=%u\n", *nblocks);
+		spin_unlock(&rbm.rgd->rd_rsspin);
 		goto rgrp_error;
 	}
-
+	BUG_ON(rbm.rgd->rd_reserved < *nblocks);
+	BUG_ON(rbm.rgd->rd_free_clone < *nblocks);
+	BUG_ON(rbm.rgd->rd_free < *nblocks);
+	rbm.rgd->rd_reserved -= *nblocks;
+	rbm.rgd->rd_free_clone -= *nblocks;
 	rbm.rgd->rd_free -= *nblocks;
+	spin_unlock(&rbm.rgd->rd_rsspin);
 	if (dinode) {
 		rbm.rgd->rd_dinodes++;
 		*generation = rbm.rgd->rd_igeneration++;
···
 
 	gfs2_trans_add_meta(rbm.rgd->rd_gl, rbm.rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rbm.rgd, rbm.rgd->rd_bits[0].bi_bh->b_data);
+	rgrp_unlock_local(rbm.rgd);
 
 	gfs2_statfs_change(sdp, 0, -(s64)*nblocks, dinode ? 1 : 0);
 	if (dinode)
···
 
 	gfs2_quota_change(ip, *nblocks, ip->i_inode.i_uid, ip->i_inode.i_gid);
 
-	rbm.rgd->rd_free_clone -= *nblocks;
 	trace_gfs2_block_alloc(ip, rbm.rgd, block, *nblocks,
 			       dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED);
 	*bn = block;
 	return 0;
 
 rgrp_error:
+	rgrp_unlock_local(rbm.rgd);
 	gfs2_rgrp_error(rbm.rgd);
 	return -EIO;
 }
···
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 
+	rgrp_lock_local(rgd);
 	rgblk_free(sdp, rgd, bstart, blen, GFS2_BLKST_FREE);
 	trace_gfs2_block_alloc(ip, rgd, bstart, blen, GFS2_BLKST_FREE);
 	rgd->rd_free += blen;
 	rgd->rd_flags &= ~GFS2_RGF_TRIMMED;
 	gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
+	rgrp_unlock_local(rgd);
 
 	/* Directories keep their data in the metadata address space */
 	if (meta || ip->i_depth || gfs2_is_jdata(ip))
···
 	rgd = gfs2_blk2rgrpd(sdp, blkno, true);
 	if (!rgd)
 		return;
+	rgrp_lock_local(rgd);
 	rgblk_free(sdp, rgd, blkno, 1, GFS2_BLKST_UNLINKED);
 	trace_gfs2_block_alloc(ip, rgd, blkno, 1, GFS2_BLKST_UNLINKED);
 	gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
 	be32_add_cpu(&rgd->rd_rgl->rl_unlinked, 1);
+	rgrp_unlock_local(rgd);
 }
 
 void gfs2_free_di(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip)
 {
 	struct gfs2_sbd *sdp = rgd->rd_sbd;
 
+	rgrp_lock_local(rgd);
 	rgblk_free(sdp, rgd, ip->i_no_addr, 1, GFS2_BLKST_FREE);
 	if (!rgd->rd_dinodes)
 		gfs2_consist_rgrpd(rgd);
···
 
 	gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh);
 	gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data);
+	rgrp_unlock_local(rgd);
 	be32_add_cpu(&rgd->rd_rgl->rl_unlinked, -1);
 
 	gfs2_statfs_change(sdp, 0, +1, -1);
···
  * @sdp: The superblock
  * @no_addr: The block number to check
  * @type: The block type we are looking for
+ *
+ * The inode glock of @no_addr must be held.  The @type to check for is either
+ * GFS2_BLKST_DINODE or GFS2_BLKST_UNLINKED; checking for type GFS2_BLKST_FREE
+ * or GFS2_BLKST_USED would make no sense.
  *
  * Returns: 0 if the block type matches the expected type
  *          -ESTALE if it doesn't match
···
 	rbm.rgd = rgd;
 	error = gfs2_rbm_from_block(&rbm, no_addr);
 	if (!WARN_ON_ONCE(error)) {
+		/*
+		 * No need to take the local resource group lock here; the
+		 * inode glock of @no_addr provides the necessary
+		 * synchronization in case the block is an inode.  (In case
+		 * the block is not an inode, the block type will not match
+		 * the @type we are looking for.)
+		 */
 		if (gfs2_testbit(&rbm, false) != type)
 			error = -ESTALE;
 	}
···
 			return;
 		rgd = gfs2_blk2rgrpd(sdp, block, 1);
 	} else {
-		rgd = ip->i_res.rs_rbm.rgd;
+		rgd = ip->i_res.rs_rgd;
 		if (!rgd || !rgrp_contains_block(rgd, block))
 			rgd = gfs2_blk2rgrpd(sdp, block, 1);
 	}
···
 				      sizeof(struct gfs2_holder),
 				      GFP_NOFS | __GFP_NOFAIL);
 	for (x = 0; x < rlist->rl_rgrps; x++)
-		gfs2_holder_init(rlist->rl_rgd[x]->rd_gl,
-				LM_ST_EXCLUSIVE, 0,
-				&rlist->rl_ghs[x]);
+		gfs2_holder_init(rlist->rl_rgd[x]->rd_gl, LM_ST_EXCLUSIVE,
+				 LM_FLAG_NODE_SCOPE, &rlist->rl_ghs[x]);
 }
 
 /**
···
 	}
 }
 
+void rgrp_lock_local(struct gfs2_rgrpd *rgd)
+{
+	BUG_ON(!gfs2_glock_is_held_excl(rgd->rd_gl) &&
+	       !test_bit(SDF_NORECOVERY, &rgd->rd_sbd->sd_flags));
+	mutex_lock(&rgd->rd_mutex);
+}
+
+void rgrp_unlock_local(struct gfs2_rgrpd *rgd)
+{
+	mutex_unlock(&rgd->rd_mutex);
+}

fs/gfs2/rgrp.h

···
 /* This is how to tell if a reservation is in the rgrp tree: */
 static inline bool gfs2_rs_active(const struct gfs2_blkreserv *rs)
 {
-	return rs && !RB_EMPTY_NODE(&rs->rs_node);
+	return !RB_EMPTY_NODE(&rs->rs_node);
 }
 
 static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block)
···
 }
 
 extern void check_and_update_goal(struct gfs2_inode *ip);
+
+extern void rgrp_lock_local(struct gfs2_rgrpd *rgd);
+extern void rgrp_unlock_local(struct gfs2_rgrpd *rgd);
+
 #endif /* __RGRP_DOT_H__ */

fs/gfs2/super.c

···
 static struct gfs2_jdesc *jdesc_find_i(struct list_head *head, unsigned int jid)
 {
 	struct gfs2_jdesc *jd;
-	int found = 0;
 
 	list_for_each_entry(jd, head, jd_list) {
-		if (jd->jd_jid == jid) {
-			found = 1;
-			break;
-		}
+		if (jd->jd_jid == jid)
+			return jd;
 	}
-
-	if (!found)
-		jd = NULL;
-
-	return jd;
+	return NULL;
 }
 
 struct gfs2_jdesc *gfs2_jdesc_find(struct gfs2_sbd *sdp, unsigned int jid)
···
 {
 	struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode);
 	struct gfs2_glock *j_gl = ip->i_gl;
-	struct gfs2_holder freeze_gh;
 	struct gfs2_log_header_host head;
 	int error;
 
 	error = init_threads(sdp);
 	if (error)
 		return error;
-
-	error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
-				   LM_FLAG_NOEXP | GL_EXACT,
-				   &freeze_gh);
-	if (error)
-		goto fail_threads;
 
 	j_gl->gl_ops->go_inval(j_gl, DIO_METADATA);
 	if (gfs2_withdrawn(sdp)) {
···
 
 	set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
 
-	gfs2_glock_dq_uninit(&freeze_gh);
-
 	return 0;
 
 fail:
-	gfs2_glock_dq_uninit(&freeze_gh);
-fail_threads:
 	if (sdp->sd_quotad_process)
 		kthread_stop(sdp->sd_quotad_process);
 	sdp->sd_quotad_process = NULL;
···
 	}
 
 	if (error)
-		gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
+		gfs2_freeze_unlock(&sdp->sd_freeze_gh);
 
 out:
 	while (!list_empty(&list)) {
···
 
 int gfs2_make_fs_ro(struct gfs2_sbd *sdp)
 {
-	struct gfs2_holder freeze_gh;
 	int error = 0;
 	int log_write_allowed = test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
-
-	gfs2_holder_mark_uninitialized(&freeze_gh);
-	if (sdp->sd_freeze_gl &&
-	    !gfs2_glock_is_locked_by_me(sdp->sd_freeze_gl)) {
-		if (!log_write_allowed) {
-			error = gfs2_glock_nq_init(sdp->sd_freeze_gl,
-						   LM_ST_SHARED, LM_FLAG_TRY |
-						   LM_FLAG_NOEXP | GL_EXACT,
-						   &freeze_gh);
-			if (error == GLR_TRYFAILED)
-				error = 0;
-		} else {
-			error = gfs2_glock_nq_init(sdp->sd_freeze_gl,
-						   LM_ST_SHARED,
-						   LM_FLAG_NOEXP | GL_EXACT,
-						   &freeze_gh);
-			if (error && !gfs2_withdrawn(sdp))
-				return error;
-		}
-	}
 
 	gfs2_flush_delete_work(sdp);
 	if (!log_write_allowed && current == sdp->sd_quotad_process)
···
 
 		gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_SHUTDOWN |
 			       GFS2_LFC_MAKE_FS_RO);
-		wait_event(sdp->sd_reserving_log_wait,
-			   atomic_read(&sdp->sd_reserving_log) == 0);
-		gfs2_assert_warn(sdp, atomic_read(&sdp->sd_log_blks_free) ==
-				 sdp->sd_jdesc->jd_blocks);
+		wait_event_timeout(sdp->sd_log_waitq,
+				   gfs2_log_is_empty(sdp),
+				   HZ * 5);
+		gfs2_assert_warn(sdp, gfs2_log_is_empty(sdp));
 	} else {
-		wait_event_timeout(sdp->sd_reserving_log_wait,
-				   atomic_read(&sdp->sd_reserving_log) == 0,
+		wait_event_timeout(sdp->sd_log_waitq,
+				   gfs2_log_is_empty(sdp),
 				   HZ * 5);
 	}
-	if (gfs2_holder_initialized(&freeze_gh))
-		gfs2_glock_dq_uninit(&freeze_gh);
-
 	gfs2_quota_cleanup(sdp);
 
 	if (!log_write_allowed)
···
 	struct super_block *sb = sdp->sd_vfs;
 
 	atomic_inc(&sb->s_active);
-	error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
-				   LM_FLAG_NOEXP | GL_EXACT, &freeze_gh);
+	error = gfs2_freeze_lock(sdp, &freeze_gh, 0);
 	if (error) {
-		fs_info(sdp, "GFS2: couldn't get freeze lock : %d\n", error);
 		gfs2_assert_withdraw(sdp, 0);
 	} else {
 		atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN);
···
 				error);
 			gfs2_assert_withdraw(sdp, 0);
 		}
-		gfs2_glock_dq_uninit(&freeze_gh);
+		gfs2_freeze_unlock(&freeze_gh);
 	}
 	deactivate_super(sb);
 	clear_bit_unlock(SDF_FS_FROZEN, &sdp->sd_flags);
···
                 return 0;
 	}
 
-	gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
+	gfs2_freeze_unlock(&sdp->sd_freeze_gh);
 	mutex_unlock(&sdp->sd_freeze_mutex);
 	return wait_on_bit(&sdp->sd_flags, SDF_FS_FROZEN, TASK_INTERRUPTIBLE);
 }
···
 		goto out_qs;
 	}
 
-	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh);
+	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE,
+				   LM_FLAG_NODE_SCOPE, &gh);
 	if (error)
 		goto out_qs;
 

fs/gfs2/super.h

···
 #include <linux/dcache.h>
 #include "incore.h"
 
+/* Supported fs format version range */
+#define GFS2_FS_FORMAT_MIN (1801)
+#define GFS2_FS_FORMAT_MAX (1802)
+
 extern void gfs2_lm_unmount(struct gfs2_sbd *sdp);
 
 static inline unsigned int gfs2_jindex_size(struct gfs2_sbd *sdp)
···
 extern const struct export_operations gfs2_export_ops;
 extern const struct super_operations gfs2_super_ops;
 extern const struct dentry_operations gfs2_dops;
-extern const struct xattr_handler *gfs2_xattr_handlers[];
+
+extern const struct xattr_handler *gfs2_xattr_handlers_max[];
+extern const struct xattr_handler **gfs2_xattr_handlers_min;
 
 #endif /* __SUPER_DOT_H__ */
 

fs/gfs2/trace_gfs2.h

···
 		__field(	u8,	block_state		)
 		__field(        u64,	rd_addr			)
 		__field(        u32,	rd_free_clone		)
+		__field(	u32,	rd_requested		)
 		__field(	u32,	rd_reserved		)
 	),
 
···
 		__entry->block_state	= block_state;
 		__entry->rd_addr	= rgd->rd_addr;
 		__entry->rd_free_clone	= rgd->rd_free_clone;
+		__entry->rd_requested	= rgd->rd_requested;
 		__entry->rd_reserved	= rgd->rd_reserved;
 	),
 
-	TP_printk("%u,%u bmap %llu alloc %llu/%lu %s rg:%llu rf:%u rr:%lu",
+	TP_printk("%u,%u bmap %llu alloc %llu/%lu %s rg:%llu rf:%u rq:%u rr:%u",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  (unsigned long long)__entry->inum,
 		  (unsigned long long)__entry->start,
 		  (unsigned long)__entry->len,
 		  block_state_name(__entry->block_state),
 		  (unsigned long long)__entry->rd_addr,
-		  __entry->rd_free_clone, (unsigned long)__entry->rd_reserved)
+		  __entry->rd_free_clone,
+		  __entry->rd_requested,
+		  __entry->rd_reserved)
 );
 
 /* Keep track of multi-block reservations as they are allocated/freed */
···
 		__field(        dev_t,  dev                     )
 		__field(	u64,	rd_addr			)
 		__field(	u32,	rd_free_clone		)
+		__field(	u32,	rd_requested		)
 		__field(	u32,	rd_reserved		)
 		__field(	u64,	inum			)
 		__field(	u64,	start			)
-		__field(	u32,	free			)
+		__field(	u32,	requested		)
+		__field(	u32,	reserved		)
 		__field(	u8,	func			)
 	),
 
 	TP_fast_assign(
-		__entry->dev		= rs->rs_rbm.rgd->rd_sbd->sd_vfs->s_dev;
-		__entry->rd_addr	= rs->rs_rbm.rgd->rd_addr;
-		__entry->rd_free_clone	= rs->rs_rbm.rgd->rd_free_clone;
-		__entry->rd_reserved	= rs->rs_rbm.rgd->rd_reserved;
+		__entry->dev		= rs->rs_rgd->rd_sbd->sd_vfs->s_dev;
+		__entry->rd_addr	= rs->rs_rgd->rd_addr;
+		__entry->rd_free_clone	= rs->rs_rgd->rd_free_clone;
+		__entry->rd_requested	= rs->rs_rgd->rd_requested;
+		__entry->rd_reserved	= rs->rs_rgd->rd_reserved;
 		__entry->inum		= container_of(rs, struct gfs2_inode,
 						       i_res)->i_no_addr;
-		__entry->start		= gfs2_rbm_to_block(&rs->rs_rbm);
-		__entry->free		= rs->rs_free;
+		__entry->start		= rs->rs_start;
+		__entry->requested	= rs->rs_requested;
+		__entry->reserved	= rs->rs_reserved;
 		__entry->func		= func;
 	),
 
-	TP_printk("%u,%u bmap %llu resrv %llu rg:%llu rf:%lu rr:%lu %s f:%lu",
+	TP_printk("%u,%u bmap %llu resrv %llu rg:%llu rf:%u rq:%u rr:%u %s q:%u r:%u",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  (unsigned long long)__entry->inum,
 		  (unsigned long long)__entry->start,
 		  (unsigned long long)__entry->rd_addr,
-		  (unsigned long)__entry->rd_free_clone,
-		  (unsigned long)__entry->rd_reserved,
-		  rs_func_name(__entry->func), (unsigned long)__entry->free)
+		  __entry->rd_free_clone,
+		  __entry->rd_requested,
+		  __entry->rd_reserved,
+		  rs_func_name(__entry->func),
+		  __entry->requested,
+		  __entry->reserved)
 );
 
 #endif /* _TRACE_GFS2_H */

fs/gfs2/trans.c

···
 	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_sbd *sdp, unsigned int blocks,
-		     unsigned int revokes)
+int __gfs2_trans_begin(struct gfs2_trans *tr, struct gfs2_sbd *sdp,
+		       unsigned int blocks, unsigned int revokes,
+		       unsigned long ip)
 {
-	struct gfs2_trans *tr;
-	int error;
+	unsigned int extra_revokes;
 
 	if (current->journal_info) {
 		gfs2_print_trans(sdp, current->journal_info);
···
 	if (!test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags))
 		return -EROFS;
 
-	tr = kmem_cache_zalloc(gfs2_trans_cachep, GFP_NOFS);
-	if (!tr)
-		return -ENOMEM;
-
-	tr->tr_ip = _RET_IP_;
+	tr->tr_ip = ip;
 	tr->tr_blocks = blocks;
 	tr->tr_revokes = revokes;
-	tr->tr_reserved = 1;
-	set_bit(TR_ALLOCED, &tr->tr_flags);
-	if (blocks)
-		tr->tr_reserved += 6 + blocks;
-	if (revokes)
-		tr->tr_reserved += gfs2_struct2blk(sdp, revokes);
+	tr->tr_reserved = GFS2_LOG_FLUSH_MIN_BLOCKS;
+	if (blocks) {
+		/*
+		 * The reserved blocks are either used for data or metadata.
+		 * We can have mixed data and metadata, each with its own log
+		 * descriptor block; see calc_reserved().
+		 */
+		tr->tr_reserved += blocks + 1 + DIV_ROUND_UP(blocks - 1, databuf_limit(sdp));
+	}
 	INIT_LIST_HEAD(&tr->tr_databuf);
 	INIT_LIST_HEAD(&tr->tr_buf);
 	INIT_LIST_HEAD(&tr->tr_list);
 	INIT_LIST_HEAD(&tr->tr_ail1_list);
 	INIT_LIST_HEAD(&tr->tr_ail2_list);
 
+	if (gfs2_assert_warn(sdp, tr->tr_reserved <= sdp->sd_jdesc->jd_blocks))
+		return -EINVAL;
+
 	sb_start_intwrite(sdp->sd_vfs);
 
-	error = gfs2_log_reserve(sdp, tr->tr_reserved);
-	if (error)
-		goto fail;
+	/*
+	 * 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);
+		return -EROFS;
+	}
 
 	current->journal_info = tr;
 
 	return 0;
+}
 
-fail:
-	sb_end_intwrite(sdp->sd_vfs);
-	kmem_cache_free(gfs2_trans_cachep, tr);
+int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks,
+		     unsigned int revokes)
+{
+	struct gfs2_trans *tr;
+	int error;
 
+	tr = kmem_cache_zalloc(gfs2_trans_cachep, GFP_NOFS);
+	if (!tr)
+		return -ENOMEM;
+	error = __gfs2_trans_begin(tr, sdp, blocks, revokes, _RET_IP_);
+	if (error)
+		kmem_cache_free(gfs2_trans_cachep, tr);
 	return error;
 }
 
···
 {
 	struct gfs2_trans *tr = current->journal_info;
 	s64 nbuf;
-	int alloced = test_bit(TR_ALLOCED, &tr->tr_flags);
 
 	current->journal_info = NULL;
 
 	if (!test_bit(TR_TOUCHED, &tr->tr_flags)) {
+		gfs2_log_release_revokes(sdp, tr->tr_revokes);
+		up_read(&sdp->sd_log_flush_lock);
 		gfs2_log_release(sdp, tr->tr_reserved);
-		if (alloced) {
+		if (!test_bit(TR_ONSTACK, &tr->tr_flags))
 			gfs2_trans_free(sdp, tr);
-			sb_end_intwrite(sdp->sd_vfs);
-		}
+		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;
 
-	if (gfs2_assert_withdraw(sdp, (nbuf <= tr->tr_blocks) &&
-				       (tr->tr_num_revoke <= tr->tr_revokes)))
+	if (gfs2_assert_withdraw(sdp, nbuf <= tr->tr_blocks) ||
+	    gfs2_assert_withdraw(sdp, tr->tr_num_revoke <= tr->tr_revokes))
 		gfs2_print_trans(sdp, tr);
 
 	gfs2_log_commit(sdp, tr);
-	if (alloced && !test_bit(TR_ATTACHED, &tr->tr_flags))
+	if (!test_bit(TR_ONSTACK, &tr->tr_flags) &&
+	    !test_bit(TR_ATTACHED, &tr->tr_flags))
 		gfs2_trans_free(sdp, tr);
 	up_read(&sdp->sd_log_flush_lock);
 
 	if (sdp->sd_vfs->s_flags & SB_SYNCHRONOUS)
 		gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL |
 			       GFS2_LFC_TRANS_END);
-	if (alloced)
-		sb_end_intwrite(sdp->sd_vfs);
+	sb_end_intwrite(sdp->sd_vfs);
 }
 
 static struct gfs2_bufdata *gfs2_alloc_bufdata(struct gfs2_glock *gl,
···
 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);
···
 			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;
 		}

fs/gfs2/trans.h

···
  * block, or all of the blocks in the rg, whichever is smaller */
 static inline unsigned int gfs2_rg_blocks(const struct gfs2_inode *ip, unsigned requested)
 {
-	struct gfs2_rgrpd *rgd = ip->i_res.rs_rbm.rgd;
+	struct gfs2_rgrpd *rgd = ip->i_res.rs_rgd;
 
 	if (requested < rgd->rd_length)
 		return requested + 1;
 	return rgd->rd_length;
 }
 
+extern int __gfs2_trans_begin(struct gfs2_trans *tr, struct gfs2_sbd *sdp,
+			      unsigned int blocks, unsigned int revokes,
+			      unsigned long ip);
 extern int gfs2_trans_begin(struct gfs2_sbd *sdp, unsigned int blocks,
 			    unsigned int revokes);
 

fs/gfs2/util.c

···
 	return error;
 }
 
+/**
+ * gfs2_freeze_lock - hold the freeze glock
+ * @sdp: the superblock
+ * @freeze_gh: pointer to the requested holder
+ * @caller_flags: any additional flags needed by the caller
+ */
+int gfs2_freeze_lock(struct gfs2_sbd *sdp, struct gfs2_holder *freeze_gh,
+		     int caller_flags)
+{
+	int flags = LM_FLAG_NOEXP | GL_EXACT | caller_flags;
+	int error;
+
+	error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED, flags,
+				   freeze_gh);
+	if (error && error != GLR_TRYFAILED)
+		fs_err(sdp, "can't lock the freeze lock: %d\n", error);
+	return error;
+}
+
+void gfs2_freeze_unlock(struct gfs2_holder *freeze_gh)
+{
+	if (gfs2_holder_initialized(freeze_gh))
+		gfs2_glock_dq_uninit(freeze_gh);
+}
+
 static void signal_our_withdraw(struct gfs2_sbd *sdp)
 {
-	struct gfs2_glock *gl = sdp->sd_live_gh.gh_gl;
+	struct gfs2_glock *live_gl = sdp->sd_live_gh.gh_gl;
 	struct inode *inode = sdp->sd_jdesc->jd_inode;
 	struct gfs2_inode *ip = GFS2_I(inode);
+	struct gfs2_glock *i_gl = ip->i_gl;
 	u64 no_formal_ino = ip->i_no_formal_ino;
+	int log_write_allowed = test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
 	int ret = 0;
 	int tries;
 
···
 	 * therefore we need to clear SDF_JOURNAL_LIVE manually.
 	 */
 	clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
-	if (!sb_rdonly(sdp->sd_vfs))
-		ret = gfs2_make_fs_ro(sdp);
+	if (!sb_rdonly(sdp->sd_vfs)) {
+		struct gfs2_holder freeze_gh;
+
+		gfs2_holder_mark_uninitialized(&freeze_gh);
+		if (sdp->sd_freeze_gl &&
+		    !gfs2_glock_is_locked_by_me(sdp->sd_freeze_gl)) {
+			ret = gfs2_freeze_lock(sdp, &freeze_gh,
+				       log_write_allowed ? 0 : LM_FLAG_TRY);
+			if (ret == GLR_TRYFAILED)
+				ret = 0;
+		}
+		if (!ret)
+			ret = gfs2_make_fs_ro(sdp);
+		gfs2_freeze_unlock(&freeze_gh);
+	}
 
 	if (sdp->sd_lockstruct.ls_ops->lm_lock == NULL) { /* lock_nolock */
 		if (!ret)
···
 		atomic_set(&sdp->sd_freeze_state, SFS_FROZEN);
 		thaw_super(sdp->sd_vfs);
 	} else {
-		wait_on_bit(&gl->gl_flags, GLF_DEMOTE, TASK_UNINTERRUPTIBLE);
+		wait_on_bit(&i_gl->gl_flags, GLF_DEMOTE,
+			    TASK_UNINTERRUPTIBLE);
 	}
 
 	/*
···
 	 * on other nodes to be successful, otherwise we remain the owner of
 	 * the glock as far as dlm is concerned.
 	 */
-	if (gl->gl_ops->go_free) {
-		set_bit(GLF_FREEING, &gl->gl_flags);
-		wait_on_bit(&gl->gl_flags, GLF_FREEING, TASK_UNINTERRUPTIBLE);
+	if (i_gl->gl_ops->go_free) {
+		set_bit(GLF_FREEING, &i_gl->gl_flags);
+		wait_on_bit(&i_gl->gl_flags, GLF_FREEING, TASK_UNINTERRUPTIBLE);
 	}
 
 	/*
 	 * Dequeue the "live" glock, but keep a reference so it's never freed.
 	 */
-	gfs2_glock_hold(gl);
+	gfs2_glock_hold(live_gl);
 	gfs2_glock_dq_wait(&sdp->sd_live_gh);
 	/*
 	 * We enqueue the "live" glock in EX so that all other nodes
···
 		gfs2_glock_nq(&sdp->sd_live_gh);
 	}
 
-	gfs2_glock_queue_put(gl); /* drop the extra reference we acquired */
+	gfs2_glock_queue_put(live_gl); /* drop extra reference we acquired */
 	clear_bit(SDF_WITHDRAW_RECOVERY, &sdp->sd_flags);
 
 	/*

fs/gfs2/util.h

···
 
 extern int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd,
 			       bool verbose);
+extern int gfs2_freeze_lock(struct gfs2_sbd *sdp,
+			    struct gfs2_holder *freeze_gh, int caller_flags);
+extern void gfs2_freeze_unlock(struct gfs2_holder *freeze_gh);
 
 #define gfs2_io_error(sdp) \
 gfs2_io_error_i((sdp), __func__, __FILE__, __LINE__)

fs/gfs2/xattr.c

···
 	return 0;
 }
 
+static bool gfs2_eatype_valid(struct gfs2_sbd *sdp, u8 type)
+{
+	switch(sdp->sd_sb.sb_fs_format) {
+	case GFS2_FS_FORMAT_MAX:
+		return true;
+
+	case GFS2_FS_FORMAT_MIN:
+		return type <= GFS2_EATYPE_SECURITY;
+
+	default:
+		return false;
+	}
+}
+
 typedef int (*ea_call_t) (struct gfs2_inode *ip, struct buffer_head *bh,
 			  struct gfs2_ea_header *ea,
 			  struct gfs2_ea_header *prev, void *private);
···
 static int ea_foreach_i(struct gfs2_inode *ip, struct buffer_head *bh,
 			ea_call_t ea_call, void *data)
 {
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct gfs2_ea_header *ea, *prev = NULL;
 	int error = 0;
 
···
 		if (!(bh->b_data <= (char *)ea && (char *)GFS2_EA2NEXT(ea) <=
 						  bh->b_data + bh->b_size))
 			goto fail;
-		if (!GFS2_EATYPE_VALID(ea->ea_type))
+		if (!gfs2_eatype_valid(sdp, ea->ea_type))
 			goto fail;
-
 		error = ea_call(ip, bh, ea, prev, data);
 		if (error)
 			return error;
···
 		return -EIO;
 	}
 
-	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &rg_gh);
+	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE,
+				   LM_FLAG_NODE_SCOPE, &rg_gh);
 	if (error)
 		return error;
 
···
 		     struct gfs2_ea_header *ea, struct gfs2_ea_header *prev,
 		     void *private)
 {
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	struct ea_list *ei = private;
 	struct gfs2_ea_request *er = ei->ei_er;
 	unsigned int ea_size;
···
 	if (ea->ea_type == GFS2_EATYPE_UNUSED)
 		return 0;
 
+	BUG_ON(ea->ea_type > GFS2_EATYPE_SECURITY &&
+	       sdp->sd_sb.sb_fs_format == GFS2_FS_FORMAT_MIN);
 	switch (ea->ea_type) {
 	case GFS2_EATYPE_USR:
 		prefix = "user.";
···
 		prefix = "security.";
 		l = 9;
 		break;
+	case GFS2_EATYPE_TRUSTED:
+		prefix = "trusted.";
+		l = 8;
+		break;
 	default:
-		BUG();
+		return 0;
 	}
 
 	ea_size = l + ea->ea_name_len + 1;
···
 		return -EIO;
 	}
 
-	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh);
+	error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE,
+				   LM_FLAG_NODE_SCOPE, &gh);
 	if (error)
 		return error;
 
···
 	.set    = gfs2_xattr_set,
 };
 
-const struct xattr_handler *gfs2_xattr_handlers[] = {
+static bool
+gfs2_xattr_trusted_list(struct dentry *dentry)
+{
+	return capable(CAP_SYS_ADMIN);
+}
+
+static const struct xattr_handler gfs2_xattr_trusted_handler = {
+	.prefix = XATTR_TRUSTED_PREFIX,
+	.flags  = GFS2_EATYPE_TRUSTED,
+	.list	= gfs2_xattr_trusted_list,
+	.get    = gfs2_xattr_get,
+	.set    = gfs2_xattr_set,
+};
+
+const struct xattr_handler *gfs2_xattr_handlers_max[] = {
+	/* GFS2_FS_FORMAT_MAX */
+	&gfs2_xattr_trusted_handler,
+
+	/* GFS2_FS_FORMAT_MIN */
 	&gfs2_xattr_user_handler,
 	&gfs2_xattr_security_handler,
 	&posix_acl_access_xattr_handler,
···
 	NULL,
 };
 
+const struct xattr_handler **gfs2_xattr_handlers_min = gfs2_xattr_handlers_max + 1;

include/uapi/linux/gfs2_ondisk.h

···
 #define GFS2_FORMAT_DE		1200
 #define GFS2_FORMAT_QU		1500
 /* These are part of the superblock */
-#define GFS2_FORMAT_FS		1801
+#define GFS2_FORMAT_FS		1802
 #define GFS2_FORMAT_MULTI	1900
 
 /*
···
 #define GFS2_EATYPE_USR		1
 #define GFS2_EATYPE_SYS		2
 #define GFS2_EATYPE_SECURITY	3
+#define GFS2_EATYPE_TRUSTED	4
 
-#define GFS2_EATYPE_LAST	3
+#define GFS2_EATYPE_LAST	4
 #define GFS2_EATYPE_VALID(x)	((x) <= GFS2_EATYPE_LAST)
 
 #define GFS2_EAFLAG_LAST	0x01	/* last ea in block */