[Cluster-devel] [PATCH 0/9] GFS2: Pre-pull patch posting (merge window)

[Cluster-devel] [PATCH 0/9] GFS2: Pre-pull patch posting (merge window)

[Bob Peterson]

Hi,

We've got nine GFS2 patches for this merge window.

1. Andreas Gruenbacher wrote a patch to replace the deprecated
   call to rhashtable_walk_init with rhashtable_walk_enter.
2. Andreas also wrote a patch to eliminate redundant code in
   two of our debugfs sequence files.
3. Andreas also cleaned up the rhashtable key ugliness Linus
   pointed out during this cycle, following Linus's suggestions.
4. Andreas also wrote a patch to take advantage of his new
   function rhashtable_lookup_get_insert_fast. This makes glock
   lookup faster and more bullet-proof.
5. Andreas also wrote a patch to revert a patch in the evict
   path that caused occasional deadlocks, and is no longer
   needed.

6. Andrew Price wrote a patch to re-enable fallocate for the
   rindex system file to enable gfs2_grow to grow properly on
   secondary file system grow operations.

7. I wrote a patch to initialize an inode number field to make
   certain kernel trace points more understandable.
8. I also wrote a patch that makes GFS2 file system "withdraw"
   work more like it should by ignoring operations after a
   withdraw that would formerly cause a BUG() and kernel panic.
9. I also reworked the entire truncate/delete algorithm,
   scrapping the old recursive algorithm in favor of a new
   non-recursive algorithm. This was done for performance:
   This way, GFS2 no longer needs to lock multiple resource
   groups while doing truncates and deletes of files that cross
   multiple resource group boundaries, allowing for better
   parallelism. It also solves a problem whereby deleting large
   files would request a large chunk of kernel memory, which
   resulted in a get_page_from_freelist warning.

Regards,

Bob Peterson
---
Andreas Gruenbacher (5):
  gfs2: Replace rhashtable_walk_init with rhashtable_walk_enter
  gfs2: Deduplicate gfs2_{glocks,glstats}_open
  gfs2: Don't pack struct lm_lockname
  gfs2: Switch to rhashtable_lookup_get_insert_fast
  Revert "GFS2: Wait for iopen glock dequeues"

Andrew Price (1):
  gfs2: Re-enable fallocate for the rindex

Bob Peterson (3):
  GFS2: Prevent BUG from occurring when normal Withdraws occur
  GFS2: Temporarily zero i_no_addr when creating a dinode
  GFS2: Non-recursive delete

[Cluster-devel] [PATCH 1/9] GFS2: Prevent BUG from occurring when normal Withdraws occur

[Bob Peterson]

When the GFS2 file system withdraws due to metadata corruption, it
often has outstanding transactions in the journal and delayed work
queued for its glocks. This patch adds some new checks for a
withdrawn file system before proceeding with operations that would
obviously cause a BUG() to be triggered. That allows GFS2 to be
safely unmounted rather than cause the system to go down.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/glock.c | 2 ++
 fs/gfs2/super.c | 3 ++-
 2 files changed, 4 insertions(+), 1 deletion(-)

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index ec0848f..694167b 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -449,6 +449,8 @@ __acquires(&gl->gl_lockref.lock)
 	unsigned int lck_flags = (unsigned int)(gh ? gh->gh_flags : 0);
 	int ret;
 
+	if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		return;
 	lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP |
 		      LM_FLAG_PRIORITY);
 	GLOCK_BUG_ON(gl, gl->gl_state == target);
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 361796a..2a9a830 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -793,7 +793,8 @@ static void gfs2_dirty_inode(struct inode *inode, int flags)
 
 	if (!(flags & (I_DIRTY_DATASYNC|I_DIRTY_SYNC)))
 		return;
-
+	if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
+		return;
 	if (!gfs2_glock_is_locked_by_me(ip->i_gl)) {
 		ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
 		if (ret) {
-- 
2.9.3

[Cluster-devel] [PATCH 2/9] gfs2: Replace rhashtable_walk_init with rhashtable_walk_enter

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Function rhashtable_walk_init is deprecated.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/glock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 694167b..b2674dd 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1934,7 +1934,7 @@ static int gfs2_glocks_open(struct inode *inode, struct file *file)
 		if (seq->buf)
 			seq->size = GFS2_SEQ_GOODSIZE;
 		gi->gl = NULL;
-		ret = rhashtable_walk_init(&gl_hash_table, &gi->hti, GFP_KERNEL);
+		rhashtable_walk_enter(&gl_hash_table, &gi->hti);
 	}
 	return ret;
 }
-- 
2.9.3

[Cluster-devel] [PATCH 3/9] gfs2: Deduplicate gfs2_{glocks, glstats}_open

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Both functions are identical except for the seq_operations used.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/glock.c | 26 +++++++++-----------------
 1 file changed, 9 insertions(+), 17 deletions(-)

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index b2674dd..b2638d8 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -1920,10 +1920,10 @@ static const struct seq_operations gfs2_sbstats_seq_ops = {
 
 #define GFS2_SEQ_GOODSIZE min(PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER, 65536UL)
 
-static int gfs2_glocks_open(struct inode *inode, struct file *file)
+static int __gfs2_glocks_open(struct inode *inode, struct file *file,
+			      const struct seq_operations *ops)
 {
-	int ret = seq_open_private(file, &gfs2_glock_seq_ops,
-				   sizeof(struct gfs2_glock_iter));
+	int ret = seq_open_private(file, ops, sizeof(struct gfs2_glock_iter));
 	if (ret == 0) {
 		struct seq_file *seq = file->private_data;
 		struct gfs2_glock_iter *gi = seq->private;
@@ -1939,6 +1939,11 @@ static int gfs2_glocks_open(struct inode *inode, struct file *file)
 	return ret;
 }
 
+static int gfs2_glocks_open(struct inode *inode, struct file *file)
+{
+	return __gfs2_glocks_open(inode, file, &gfs2_glock_seq_ops);
+}
+
 static int gfs2_glocks_release(struct inode *inode, struct file *file)
 {
 	struct seq_file *seq = file->private_data;
@@ -1951,20 +1956,7 @@ static int gfs2_glocks_release(struct inode *inode, struct file *file)
 
 static int gfs2_glstats_open(struct inode *inode, struct file *file)
 {
-	int ret = seq_open_private(file, &gfs2_glstats_seq_ops,
-				   sizeof(struct gfs2_glock_iter));
-	if (ret == 0) {
-		struct seq_file *seq = file->private_data;
-		struct gfs2_glock_iter *gi = seq->private;
-		gi->sdp = inode->i_private;
-		gi->last_pos = 0;
-		seq->buf = kmalloc(GFS2_SEQ_GOODSIZE, GFP_KERNEL | __GFP_NOWARN);
-		if (seq->buf)
-			seq->size = GFS2_SEQ_GOODSIZE;
-		gi->gl = NULL;
-		ret = rhashtable_walk_init(&gl_hash_table, &gi->hti, GFP_KERNEL);
-	}
-	return ret;
+	return __gfs2_glocks_open(inode, file, &gfs2_glstats_seq_ops);
 }
 
 static int gfs2_sbstats_open(struct inode *inode, struct file *file)
-- 
2.9.3

[Cluster-devel] [PATCH 4/9] gfs2: Don't pack struct lm_lockname

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

As per a suggestion by Linus, don't pack struct lm_lockname: we did that
because the struct is used as a rhashtable key, but packing tells the
compiler that the 64-bit fields in the struct may be unaligned, causing
it to generate worse code on some architectures.  Instead, rearrange the
fields in the struct so that there is no padding between fields, and
exclude any tail padding from the hash key size.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/glock.c  | 2 +-
 fs/gfs2/incore.h | 8 ++++++--
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index b2638d8..35f3b0a 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -73,7 +73,7 @@ static DEFINE_SPINLOCK(lru_lock);
 
 static struct rhashtable_params ht_parms = {
 	.nelem_hint = GFS2_GL_HASH_SIZE * 3 / 4,
-	.key_len = sizeof(struct lm_lockname),
+	.key_len = offsetofend(struct lm_lockname, ln_type),
 	.key_offset = offsetof(struct gfs2_glock, gl_name),
 	.head_offset = offsetof(struct gfs2_glock, gl_node),
 };
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index 511e1ed..b7cf65d 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -203,11 +203,15 @@ enum {
 	DFL_DLM_RECOVERY	= 6,
 };
 
+/*
+ * We are using struct lm_lockname as an rhashtable key.  Avoid holes within
+ * the struct; padding at the end is fine.
+ */
 struct lm_lockname {
-	struct gfs2_sbd *ln_sbd;
 	u64 ln_number;
+	struct gfs2_sbd *ln_sbd;
 	unsigned int ln_type;
-} __packed __aligned(sizeof(int));
+};
 
 #define lm_name_equal(name1, name2) \
         (((name1)->ln_number == (name2)->ln_number) &&	\
-- 
2.9.3

[Cluster-devel] [PATCH 5/9] GFS2: Temporarily zero i_no_addr when creating a dinode

[Bob Peterson]

Before this patch i_no_addr was not initialized until after the
return from allocating its block. That meant the i_no_addr was
temporarily uninitialized storage. Ordinarily that's not a concern,
but if inplace_reserve can't find space, it can call try_rgrp_unlink
which references i_no_addr as a block to avoid. That can result in
unpredictable behavior. More importantly, the trace point in
gfs2_alloc_blocks references ip->i_no_addr before it is set, which
is misleading when reading the kernel traces. This patch makes it
look like the new dinode block was assigned in the name of inode 0
rather than a random inode that's completely unrelated.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/inode.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index e279c3c..4f405d4 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -667,6 +667,7 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	ip->i_height = 0;
 	ip->i_depth = 0;
 	ip->i_entries = 0;
+	ip->i_no_addr = 0; /* Temporarily zero until real addr is assigned */
 
 	switch(mode & S_IFMT) {
 	case S_IFREG:
-- 
2.9.3

[Cluster-devel] [PATCH 6/9] gfs2: Switch to rhashtable_lookup_get_insert_fast

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Switch from rhashtable_lookup_insert_fast + rhashtable_lookup_fast to
rhashtable_lookup_get_insert_fast, which is cleaner and avoids an extra
rhashtable lookup.

At the same time, turn the retry loop in gfs2_glock_get into an infinite
loop.  The lookup or insert will eventually succeed, usually very fast,
but there is no reason to give up trying at a fixed number of
iterations.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/glock.c | 45 +++++++++++++++++++++------------------------
 1 file changed, 21 insertions(+), 24 deletions(-)

diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c
index 35f3b0a..9e81219 100644
--- a/fs/gfs2/glock.c
+++ b/fs/gfs2/glock.c
@@ -655,10 +655,10 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	struct lm_lockname name = { .ln_number = number,
 				    .ln_type = glops->go_type,
 				    .ln_sbd = sdp };
-	struct gfs2_glock *gl, *tmp = NULL;
+	struct gfs2_glock *gl, *tmp;
 	struct address_space *mapping;
 	struct kmem_cache *cachep;
-	int ret, tries = 0;
+	int ret = 0;
 
 	rcu_read_lock();
 	gl = rhashtable_lookup_fast(&gl_hash_table, &name, ht_parms);
@@ -723,35 +723,32 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	}
 
 again:
-	ret = rhashtable_lookup_insert_fast(&gl_hash_table, &gl->gl_node,
-					    ht_parms);
-	if (ret == 0) {
+	rcu_read_lock();
+	tmp = rhashtable_lookup_get_insert_fast(&gl_hash_table, &gl->gl_node,
+						ht_parms);
+	if (!tmp) {
 		*glp = gl;
-		return 0;
+		goto out;
 	}
-
-	if (ret == -EEXIST) {
-		ret = 0;
-		rcu_read_lock();
-		tmp = rhashtable_lookup_fast(&gl_hash_table, &name, ht_parms);
-		if (tmp == NULL || !lockref_get_not_dead(&tmp->gl_lockref)) {
-			if (++tries < 100) {
-				rcu_read_unlock();
-				cond_resched();
-				goto again;
-			}
-			tmp = NULL;
-			ret = -ENOMEM;
-		}
-		rcu_read_unlock();
-	} else {
-		WARN_ON_ONCE(ret);
+	if (IS_ERR(tmp)) {
+		ret = PTR_ERR(tmp);
+		goto out_free;
+	}
+	if (lockref_get_not_dead(&tmp->gl_lockref)) {
+		*glp = tmp;
+		goto out_free;
 	}
+	rcu_read_unlock();
+	cond_resched();
+	goto again;
+
+out_free:
 	kfree(gl->gl_lksb.sb_lvbptr);
 	kmem_cache_free(cachep, gl);
 	atomic_dec(&sdp->sd_glock_disposal);
-	*glp = tmp;
 
+out:
+	rcu_read_unlock();
 	return ret;
 }
 
-- 
2.9.3

[Cluster-devel] [PATCH 7/9] Revert "GFS2: Wait for iopen glock dequeues"

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Revert commit 86d067a797d4e8546a7c92b985f31e8cd3ec39ad: it turns out
that waiting for iopen glock dequeues here isn't needed anymore because
the bugs that commit was meant to fix have been fixed otherwise.

In addition, we want to avoid waiting on glocks in gfs2_evict_inode in
shrinker context because the shrinker may be invoked on behalf of DLM,
in which case calling into DLM again would deadlock.  This commit makes
the described scenario less likely without completely avoiding it; it's
still a step in the right direction, though.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/inode.c | 3 +--
 fs/gfs2/super.c | 8 +++-----
 2 files changed, 4 insertions(+), 7 deletions(-)

diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 4f405d4..9f605ea 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -202,8 +202,7 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 fail_refresh:
 	ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
 	ip->i_iopen_gh.gh_gl->gl_object = NULL;
-	gfs2_glock_dq_wait(&ip->i_iopen_gh);
-	gfs2_holder_uninit(&ip->i_iopen_gh);
+	gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 fail_put:
 	if (io_gl)
 		gfs2_glock_put(io_gl);
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index 2a9a830..29b0473 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -1539,8 +1539,7 @@ static void gfs2_evict_inode(struct inode *inode)
 	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_SKIP, &gh);
 	if (unlikely(error)) {
 		ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
-		gfs2_glock_dq_wait(&ip->i_iopen_gh);
-		gfs2_holder_uninit(&ip->i_iopen_gh);
+		gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 		goto out;
 	}
 
@@ -1618,7 +1617,7 @@ static void gfs2_evict_inode(struct inode *inode)
 	if (gfs2_holder_initialized(&ip->i_iopen_gh)) {
 		if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags)) {
 			ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
-			gfs2_glock_dq_wait(&ip->i_iopen_gh);
+			gfs2_glock_dq(&ip->i_iopen_gh);
 		}
 		gfs2_holder_uninit(&ip->i_iopen_gh);
 	}
@@ -1640,8 +1639,7 @@ static void gfs2_evict_inode(struct inode *inode)
 	if (gfs2_holder_initialized(&ip->i_iopen_gh)) {
 		ip->i_iopen_gh.gh_gl->gl_object = NULL;
 		ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
-		gfs2_glock_dq_wait(&ip->i_iopen_gh);
-		gfs2_holder_uninit(&ip->i_iopen_gh);
+		gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 	}
 }
 
-- 
2.9.3

[Cluster-devel] [PATCH 8/9] gfs2: Re-enable fallocate for the rindex

[Bob Peterson]

From: Andrew Price <anprice@redhat.com>

Commit 86066914edff2316cbed63aac8a87d5001441a16 "gfs2: Don't support
fallocate on jdata files" removed the ability of gfs2_grow to reserve
space at the end of the rindex, which could prevent a second gfs2_grow
from succeeding if the fs is full. Allow fallocate to work on the rindex
once again.

Signed-off-by: Andrew Price <anprice@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/file.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 6fe2a59..c2062a1 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -911,11 +911,15 @@ static long __gfs2_fallocate(struct file *file, int mode, loff_t offset, loff_t
 static long gfs2_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 {
 	struct inode *inode = file_inode(file);
+	struct gfs2_sbd *sdp = GFS2_SB(inode);
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_holder gh;
 	int ret;
 
-	if ((mode & ~FALLOC_FL_KEEP_SIZE) || gfs2_is_jdata(ip))
+	if (mode & ~FALLOC_FL_KEEP_SIZE)
+		return -EOPNOTSUPP;
+	/* fallocate is needed by gfs2_grow to reserve space in the rindex */
+	if (gfs2_is_jdata(ip) && inode != sdp->sd_rindex)
 		return -EOPNOTSUPP;
 
 	inode_lock(inode);
-- 
2.9.3

[Cluster-devel] [PATCH 9/9] GFS2: Non-recursive delete

[Bob Peterson]

Implement truncate/delete as a non-recursive algorithm. The older
algorithm was implemented with recursion to strip off each layer
at a time (going by height, starting with the maximum height.
This version tries to do the same thing but without recursion,
and without needing to allocate new structures or lists in memory.

For example, say you want to truncate a very large file to 1 byte,
and its end-of-file metapath is: 0.505.463.428. The starting
metapath would be 0.0.0.0. Since it's a truncate to non-zero, it
needs to preserve that byte, and all metadata pointing to it.
So it would start at 0.0.0.0, look up all its metadata buffers,
then free all data blocks pointed to at the highest level.
After that buffer is "swept", it moves on to 0.0.0.1, then
0.0.0.2, etc., reading in buffers and sweeping them clean.
When it gets to the end of the 0.0.0 metadata buffer (for 4K
blocks the last valid one is 0.0.0.508), it backs up to the
previous height and starts working on 0.0.1.0, then 0.0.1.1,
and so forth. After it reaches the end and sweeps 0.0.1.508,
it continues with 0.0.2.0, and so on. When that height is
exhausted, and it reaches 0.0.508.508 it backs up another level,
to 0.1.0.0, then 0.1.0.1, through 0.1.0.508. So it has to keep
marching backwards and forwards through the metadata until it's
all swept clean. Once it has all the data blocks freed, it
lowers the strip height, and begins the process all over again,
but with one less height. This time it sweeps 0.0.0 through
0.505.463. When that's clean, it lowers the strip height again
and works to free 0.505. Eventually it strips the lowest height, 0.
For a delete or truncate to 0, all metadata for all heights of
0.0.0.0 would be freed. For a truncate to 1 byte, 0.0.0.0 would
be preserved.

This isn't much different from normal integer incrementing,
where an integer gets incremented from 0000 (0.0.0.0) to 3021
(3.0.2.1). So 0000 gets increments to 0001, 0002, up to 0009,
then on to 0010, 0011 up to 0099, then 0100 and so forth. It's
just that each "digit" goes from 0 to 508 (for a total of 509
pointers) rather than from 0 to 9.

Note that the dinode will only have 483 pointers due to the
dinode structure itself.

Also note: this is just an example. These numbers (509 and 483)
are based on a standard 4K block size. Smaller block sizes will
yield smaller numbers of indirect pointers accordingly.

The truncation process is accomplished with the help of two
major functions and a few helper functions.

Functions do_strip and recursive_scan are obsolete, so removed.

New function sweep_bh_for_rgrps cleans a buffer_head pointed to
by the given metapath and height. By cleaning, I mean it frees
all blocks starting at the offset passed in metapath. It starts
at the first block in the buffer pointed to by the metapath and
identifies its resource group (rgrp). From there it frees all
subsequent block pointers that lie within that rgrp. If it's
already inside a transaction, it stays within it as long as it
can. In other words, it doesn't close a transaction until it knows
it's freed what it can from the resource group. In this way,
multiple buffers may be cleaned in a single transaction, as long
as those blocks in the buffer all lie within the same rgrp.

If it's not in a transaction, it starts one. If the buffer_head
has references to blocks within multiple rgrps, it frees all the
blocks inside the first rgrp it finds, then closes the
transaction. Then it repeats the cycle: identifies the next
unfreed block, uses it to find its rgrp, then starts a new
transaction for that set. It repeats this process repeatedly
until the buffer_head contains no more references to any blocks
past the given metapath.

Function trunc_dealloc has been reworked into a finite state
automaton. It has basically 3 active states:
DEALLOC_MP_FULL, DEALLOC_MP_LOWER, and DEALLOC_FILL_MP:

The DEALLOC_MP_FULL state implies the metapath has a full set
of buffers out to the "shrink height", and therefore, it can
call function sweep_bh_for_rgrps to free the blocks within the
highest height of the metapath. If it's just swept the lowest
level (or an error has occurred) the state machine is ended.
Otherwise it proceeds to the DEALLOC_MP_LOWER state.

The DEALLOC_MP_LOWER state implies we are finished with a given
buffer_head, which may now be released, and therefore we are
then missing some buffer information from the metapath. So we
need to find more buffers to read in. In most cases, this is
just a matter of releasing the buffer_head and moving to the
next pointer from the previous height, so it may be read in and
swept as well. If it can't find another non-null pointer to
process, it checks whether it's reached the end of a height
and needs to lower the strip height, or whether it still needs
move forward through the previous height's metadata. In this
state, all zero-pointers are skipped. From this state, it can
only loop around (once more backing up another height) or,
once a valid metapath is found (one that has non-zero
pointers), proceed to state DEALLOC_FILL_MP.

The DEALLOC_FILL_MP state implies that we have a metapath
but not all its buffers are read in. So we must proceed to read
in buffer_heads until the metapath has a valid buffer for every
height. If the previous state backed us up 3 heights, we may
need to read in a buffer, increment the height, then repeat the
process until buffers have been read in for all required heights.
If it's successful reading a buffer, and it's at the highest
height we need, it proceeds back to the DEALLOC_MP_FULL state.
If it's unable to fill in a buffer, (encounters a hole, etc.)
it tries to find another non-zero block pointer. If they're all
zero, it lowers the height and returns to the DEALLOC_MP_LOWER
state. If it finds a good non-null pointer, it loops around and
reads it in, while keeping the metapath in lock-step with the
pointers it examines.

The state machine runs until the truncation request is
satisfied. Then any transactions are ended, the quota and
statfs data are updated, and the function is complete.

Helper function metaptr1 was introduced to be an easy way to
determine the start of a buffer_head's indirect pointers.

Helper function lookup_mp_height was introduced to find a
metapath index and read in the buffer that corresponds to it.
In this way, function lookup_metapath becomes a simple loop to
call it for every height.

Helper function fillup_metapath is similar to lookup_metapath
except it can do partial lookups. If the state machine
backed up multiple levels (like 2999 wrapping to 3000) it
needs to find out the next starting point and start issuing
metadata reads at that point.

Helper function hptrs is a shortcut to determine how many
pointers should be expected in a buffer. Height 0 is the dinode
which has fewer pointers than the others.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/bmap.c | 741 +++++++++++++++++++++++++++++++++++----------------------
 fs/gfs2/rgrp.c |   7 -
 fs/gfs2/rgrp.h |   7 +
 3 files changed, 463 insertions(+), 292 deletions(-)

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 01b97c0..3814a60 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -38,11 +38,6 @@ struct metapath {
 	__u16 mp_list[GFS2_MAX_META_HEIGHT];
 };
 
-struct strip_mine {
-	int sm_first;
-	unsigned int sm_height;
-};
-
 /**
  * gfs2_unstuffer_page - unstuff a stuffed inode into a block cached by a page
  * @ip: the inode
@@ -253,6 +248,19 @@ static inline unsigned int metapath_branch_start(const struct metapath *mp)
 }
 
 /**
+ * metaptr1 - Return the first possible metadata pointer in a metaath buffer
+ * @height: The metadata height (0 = dinode)
+ * @mp: The metapath
+ */
+static inline __be64 *metaptr1(unsigned int height, const struct metapath *mp)
+{
+	struct buffer_head *bh = mp->mp_bh[height];
+	if (height == 0)
+		return ((__be64 *)(bh->b_data + sizeof(struct gfs2_dinode)));
+	return ((__be64 *)(bh->b_data + sizeof(struct gfs2_meta_header)));
+}
+
+/**
  * metapointer - Return pointer to start of metadata in a buffer
  * @height: The metadata height (0 = dinode)
  * @mp: The metapath
@@ -264,10 +272,8 @@ static inline unsigned int metapath_branch_start(const struct metapath *mp)
 
 static inline __be64 *metapointer(unsigned int height, const struct metapath *mp)
 {
-	struct buffer_head *bh = mp->mp_bh[height];
-	unsigned int head_size = (height > 0) ?
-		sizeof(struct gfs2_meta_header) : sizeof(struct gfs2_dinode);
-	return ((__be64 *)(bh->b_data + head_size)) + mp->mp_list[height];
+	__be64 *p = metaptr1(height, mp);
+	return p + mp->mp_list[height];
 }
 
 static void gfs2_metapath_ra(struct gfs2_glock *gl,
@@ -296,6 +302,23 @@ static void gfs2_metapath_ra(struct gfs2_glock *gl,
 }
 
 /**
+ * lookup_mp_height - helper function for lookup_metapath
+ * @ip: the inode
+ * @mp: the metapath
+ * @h: the height which needs looking up
+ */
+static int lookup_mp_height(struct gfs2_inode *ip, struct metapath *mp, int h)
+{
+	__be64 *ptr = metapointer(h, mp);
+	u64 dblock = be64_to_cpu(*ptr);
+
+	if (!dblock)
+		return h + 1;
+
+	return gfs2_meta_indirect_buffer(ip, h + 1, dblock, &mp->mp_bh[h + 1]);
+}
+
+/**
  * lookup_metapath - Walk the metadata tree to a specific point
  * @ip: The inode
  * @mp: The metapath
@@ -316,17 +339,10 @@ static int lookup_metapath(struct gfs2_inode *ip, struct metapath *mp)
 {
 	unsigned int end_of_metadata = ip->i_height - 1;
 	unsigned int x;
-	__be64 *ptr;
-	u64 dblock;
 	int ret;
 
 	for (x = 0; x < end_of_metadata; x++) {
-		ptr = metapointer(x, mp);
-		dblock = be64_to_cpu(*ptr);
-		if (!dblock)
-			return x + 1;
-
-		ret = gfs2_meta_indirect_buffer(ip, x+1, dblock, &mp->mp_bh[x+1]);
+		ret = lookup_mp_height(ip, mp, x);
 		if (ret)
 			return ret;
 	}
@@ -334,6 +350,35 @@ static int lookup_metapath(struct gfs2_inode *ip, struct metapath *mp)
 	return ip->i_height;
 }
 
+/**
+ * fillup_metapath - fill up buffers for the metadata path to a specific height
+ * @ip: The inode
+ * @mp: The metapath
+ * @h: The height to which it should be mapped
+ *
+ * Similar to lookup_metapath, but does lookups for a range of heights
+ *
+ * Returns: error or height of metadata tree
+ */
+
+static int fillup_metapath(struct gfs2_inode *ip, struct metapath *mp, int h)
+{
+	unsigned int start_h = h - 1;
+	int ret;
+
+	if (h) {
+		/* find the first buffer we need to look up. */
+		while (start_h > 0 && mp->mp_bh[start_h] == NULL)
+			start_h--;
+		for (; start_h < h; start_h++) {
+			ret = lookup_mp_height(ip, mp, start_h);
+			if (ret)
+				return ret;
+		}
+	}
+	return ip->i_height;
+}
+
 static inline void release_metapath(struct metapath *mp)
 {
 	int i;
@@ -422,6 +467,13 @@ enum alloc_state {
 	/* ALLOC_UNSTUFF = 3,   TBD and rather complicated */
 };
 
+static inline unsigned int hptrs(struct gfs2_sbd *sdp, const unsigned int hgt)
+{
+	if (hgt)
+		return sdp->sd_inptrs;
+	return sdp->sd_diptrs;
+}
+
 /**
  * gfs2_bmap_alloc - Build a metadata tree of the requested height
  * @inode: The GFS2 inode
@@ -620,7 +672,7 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 
 	BUG_ON(maxlen == 0);
 
-	memset(mp.mp_bh, 0, sizeof(mp.mp_bh));
+	memset(&mp, 0, sizeof(mp));
 	bmap_lock(ip, create);
 	clear_buffer_mapped(bh_map);
 	clear_buffer_new(bh_map);
@@ -702,252 +754,6 @@ int gfs2_extent_map(struct inode *inode, u64 lblock, int *new, u64 *dblock, unsi
 }
 
 /**
- * do_strip - Look for a layer a particular layer of the file and strip it off
- * @ip: the inode
- * @dibh: the dinode buffer
- * @bh: A buffer of pointers
- * @top: The first pointer in the buffer
- * @bottom: One more than the last pointer
- * @height: the height this buffer is at
- * @sm: a pointer to a struct strip_mine
- *
- * Returns: errno
- */
-
-static int do_strip(struct gfs2_inode *ip, struct buffer_head *dibh,
-		    struct buffer_head *bh, __be64 *top, __be64 *bottom,
-		    unsigned int height, struct strip_mine *sm)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	struct gfs2_rgrp_list rlist;
-	struct gfs2_trans *tr;
-	u64 bn, bstart;
-	u32 blen, btotal;
-	__be64 *p;
-	unsigned int rg_blocks = 0;
-	int metadata;
-	unsigned int revokes = 0;
-	int x;
-	int error;
-	int jblocks_rqsted;
-
-	error = gfs2_rindex_update(sdp);
-	if (error)
-		return error;
-
-	if (!*top)
-		sm->sm_first = 0;
-
-	if (height != sm->sm_height)
-		return 0;
-
-	if (sm->sm_first) {
-		top++;
-		sm->sm_first = 0;
-	}
-
-	metadata = (height != ip->i_height - 1);
-	if (metadata)
-		revokes = (height) ? sdp->sd_inptrs : sdp->sd_diptrs;
-	else if (ip->i_depth)
-		revokes = sdp->sd_inptrs;
-
-	memset(&rlist, 0, sizeof(struct gfs2_rgrp_list));
-	bstart = 0;
-	blen = 0;
-
-	for (p = top; p < bottom; p++) {
-		if (!*p)
-			continue;
-
-		bn = be64_to_cpu(*p);
-
-		if (bstart + blen == bn)
-			blen++;
-		else {
-			if (bstart)
-				gfs2_rlist_add(ip, &rlist, bstart);
-
-			bstart = bn;
-			blen = 1;
-		}
-	}
-
-	if (bstart)
-		gfs2_rlist_add(ip, &rlist, bstart);
-	else
-		goto out; /* Nothing to do */
-
-	gfs2_rlist_alloc(&rlist, LM_ST_EXCLUSIVE);
-
-	for (x = 0; x < rlist.rl_rgrps; x++) {
-		struct gfs2_rgrpd *rgd;
-		rgd = rlist.rl_ghs[x].gh_gl->gl_object;
-		rg_blocks += rgd->rd_length;
-	}
-
-	error = gfs2_glock_nq_m(rlist.rl_rgrps, rlist.rl_ghs);
-	if (error)
-		goto out_rlist;
-
-	if (gfs2_rs_active(&ip->i_res)) /* needs to be done with the rgrp glock held */
-		gfs2_rs_deltree(&ip->i_res);
-
-restart:
-	jblocks_rqsted = rg_blocks + RES_DINODE +
-		RES_INDIRECT + RES_STATFS + RES_QUOTA +
-		gfs2_struct2blk(sdp, revokes, sizeof(u64));
-	if (jblocks_rqsted > atomic_read(&sdp->sd_log_thresh2))
-		jblocks_rqsted = atomic_read(&sdp->sd_log_thresh2);
-	error = gfs2_trans_begin(sdp, jblocks_rqsted, revokes);
-	if (error)
-		goto out_rg_gunlock;
-
-	tr = current->journal_info;
-	down_write(&ip->i_rw_mutex);
-
-	gfs2_trans_add_meta(ip->i_gl, dibh);
-	gfs2_trans_add_meta(ip->i_gl, bh);
-
-	bstart = 0;
-	blen = 0;
-	btotal = 0;
-
-	for (p = top; p < bottom; p++) {
-		if (!*p)
-			continue;
-
-		/* check for max reasonable journal transaction blocks */
-		if (tr->tr_num_buf_new + RES_STATFS +
-		    RES_QUOTA >= atomic_read(&sdp->sd_log_thresh2)) {
-			if (rg_blocks >= tr->tr_num_buf_new)
-				rg_blocks -= tr->tr_num_buf_new;
-			else
-				rg_blocks = 0;
-			break;
-		}
-
-		bn = be64_to_cpu(*p);
-
-		if (bstart + blen == bn)
-			blen++;
-		else {
-			if (bstart) {
-				__gfs2_free_blocks(ip, bstart, blen, metadata);
-				btotal += blen;
-			}
-
-			bstart = bn;
-			blen = 1;
-		}
-
-		*p = 0;
-		gfs2_add_inode_blocks(&ip->i_inode, -1);
-	}
-	if (p == bottom)
-		rg_blocks = 0;
-
-	if (bstart) {
-		__gfs2_free_blocks(ip, bstart, blen, metadata);
-		btotal += blen;
-	}
-
-	gfs2_statfs_change(sdp, 0, +btotal, 0);
-	gfs2_quota_change(ip, -(s64)btotal, ip->i_inode.i_uid,
-			  ip->i_inode.i_gid);
-
-	ip->i_inode.i_mtime = ip->i_inode.i_ctime = current_time(&ip->i_inode);
-
-	gfs2_dinode_out(ip, dibh->b_data);
-
-	up_write(&ip->i_rw_mutex);
-
-	gfs2_trans_end(sdp);
-
-	if (rg_blocks)
-		goto restart;
-
-out_rg_gunlock:
-	gfs2_glock_dq_m(rlist.rl_rgrps, rlist.rl_ghs);
-out_rlist:
-	gfs2_rlist_free(&rlist);
-out:
-	return error;
-}
-
-/**
- * recursive_scan - recursively scan through the end of a file
- * @ip: the inode
- * @dibh: the dinode buffer
- * @mp: the path through the metadata to the point to start
- * @height: the height the recursion is at
- * @block: the indirect block to look at
- * @first: 1 if this is the first block
- * @sm: data opaque to this function to pass to @bc
- *
- * When this is first called @height and @block should be zero and
- * @first should be 1.
- *
- * Returns: errno
- */
-
-static int recursive_scan(struct gfs2_inode *ip, struct buffer_head *dibh,
-			  struct metapath *mp, unsigned int height,
-			  u64 block, int first, struct strip_mine *sm)
-{
-	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	struct buffer_head *bh = NULL;
-	__be64 *top, *bottom;
-	u64 bn;
-	int error;
-	int mh_size = sizeof(struct gfs2_meta_header);
-
-	if (!height) {
-		error = gfs2_meta_inode_buffer(ip, &bh);
-		if (error)
-			return error;
-		dibh = bh;
-
-		top = (__be64 *)(bh->b_data + sizeof(struct gfs2_dinode)) + mp->mp_list[0];
-		bottom = (__be64 *)(bh->b_data + sizeof(struct gfs2_dinode)) + sdp->sd_diptrs;
-	} else {
-		error = gfs2_meta_indirect_buffer(ip, height, block, &bh);
-		if (error)
-			return error;
-
-		top = (__be64 *)(bh->b_data + mh_size) +
-				  (first ? mp->mp_list[height] : 0);
-
-		bottom = (__be64 *)(bh->b_data + mh_size) + sdp->sd_inptrs;
-	}
-
-	error = do_strip(ip, dibh, bh, top, bottom, height, sm);
-	if (error)
-		goto out;
-
-	if (height < ip->i_height - 1) {
-
-		gfs2_metapath_ra(ip->i_gl, bh, top);
-
-		for (; top < bottom; top++, first = 0) {
-			if (!*top)
-				continue;
-
-			bn = be64_to_cpu(*top);
-
-			error = recursive_scan(ip, dibh, mp, height + 1, bn,
-					       first, sm);
-			if (error)
-				break;
-		}
-	}
-out:
-	brelse(bh);
-	return error;
-}
-
-
-/**
  * gfs2_block_truncate_page - Deal with zeroing out data for truncate
  *
  * This is partly borrowed from ext3.
@@ -1106,41 +912,406 @@ static int trunc_start(struct inode *inode, u64 oldsize, u64 newsize)
 	return error;
 }
 
-static int trunc_dealloc(struct gfs2_inode *ip, u64 size)
+/**
+ * sweep_bh_for_rgrps - find an rgrp in a meta buffer and free blocks therein
+ * @ip: inode
+ * @rg_gh: holder of resource group glock
+ * @mp: current metapath fully populated with buffers
+ * @btotal: place to keep count of total blocks freed
+ * @hgt: height we're processing
+ * @first: true if this is the first call to this function for this height
+ *
+ * We sweep a metadata buffer (provided by the metapath) for blocks we need to
+ * free, and free them all. However, we do it one rgrp@a time. If this
+ * block has references to multiple rgrps, we break it into individual
+ * transactions. This allows other processes to use the rgrps while we're
+ * focused on a single one, for better concurrency / performance.
+ * At every transaction boundary, we rewrite the inode into the journal.
+ * That way the bitmaps are kept consistent with the inode and we can recover
+ * if we're interrupted by power-outages.
+ *
+ * Returns: 0, or return code if an error occurred.
+ *          *btotal has the total number of blocks freed
+ */
+static int sweep_bh_for_rgrps(struct gfs2_inode *ip, struct gfs2_holder *rd_gh,
+			      const struct metapath *mp, u32 *btotal, int hgt,
+			      bool preserve1)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
-	unsigned int height = ip->i_height;
-	u64 lblock;
-	struct metapath mp;
-	int error;
+	struct gfs2_rgrpd *rgd;
+	struct gfs2_trans *tr;
+	struct buffer_head *bh = mp->mp_bh[hgt];
+	__be64 *top, *bottom, *p;
+	int blks_outside_rgrp;
+	u64 bn, bstart, isize_blks;
+	s64 blen; /* needs to be s64 or gfs2_add_inode_blocks breaks */
+	int meta = ((hgt != ip->i_height - 1) ? 1 : 0);
+	int ret = 0;
+	bool buf_in_tr = false; /* buffer was added to transaction */
+
+	if (gfs2_metatype_check(sdp, bh,
+				(hgt ? GFS2_METATYPE_IN : GFS2_METATYPE_DI)))
+		return -EIO;
+
+more_rgrps:
+	blks_outside_rgrp = 0;
+	bstart = 0;
+	blen = 0;
+	top = metapointer(hgt, mp); /* first ptr from metapath */
+	/* If we're keeping some data at the truncation point, we've got to
+	   preserve the metadata tree by adding 1 to the starting metapath. */
+	if (preserve1)
+		top++;
+
+	bottom = (__be64 *)(bh->b_data + bh->b_size);
+
+	for (p = top; p < bottom; p++) {
+		if (!*p)
+			continue;
+		bn = be64_to_cpu(*p);
+		if (gfs2_holder_initialized(rd_gh)) {
+			rgd = (struct gfs2_rgrpd *)rd_gh->gh_gl->gl_object;
+			gfs2_assert_withdraw(sdp,
+				     gfs2_glock_is_locked_by_me(rd_gh->gh_gl));
+		} else {
+			rgd = gfs2_blk2rgrpd(sdp, bn, false);
+			ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE,
+						 0, 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)
+				gfs2_rs_deltree(&ip->i_res);
+		}
+
+		if (!rgrp_contains_block(rgd, bn)) {
+			blks_outside_rgrp++;
+			continue;
+		}
+
+		/* The size of our transactions will be unknown until we
+		   actually process all the metadata blocks that relate to
+		   the rgrp. So we estimate. We know it can't be more than
+		   the dinode's i_blocks and we don't want to exceed the
+		   journal flush threshold, sd_log_thresh2. */
+		if (current->journal_info == NULL) {
+			unsigned int jblocks_rqsted, revokes;
+
+			jblocks_rqsted = rgd->rd_length + RES_DINODE +
+				RES_INDIRECT;
+			isize_blks = gfs2_get_inode_blocks(&ip->i_inode);
+			if (isize_blks > atomic_read(&sdp->sd_log_thresh2))
+				jblocks_rqsted +=
+					atomic_read(&sdp->sd_log_thresh2);
+			else
+				jblocks_rqsted += isize_blks;
+			revokes = jblocks_rqsted;
+			if (meta)
+				revokes += hptrs(sdp, hgt);
+			else if (ip->i_depth)
+				revokes += sdp->sd_inptrs;
+			ret = gfs2_trans_begin(sdp, jblocks_rqsted, revokes);
+			if (ret)
+				goto out_unlock;
+			down_write(&ip->i_rw_mutex);
+		}
+		/* check if we will exceed the transaction blocks requested */
+		tr = current->journal_info;
+		if (tr->tr_num_buf_new + RES_STATFS +
+		    RES_QUOTA >= atomic_read(&sdp->sd_log_thresh2)) {
+			/* We set blks_outside_rgrp to ensure the loop will
+			   be repeated for the same rgrp, but with a new
+			   transaction. */
+			blks_outside_rgrp++;
+			/* This next part is tricky. If the buffer was added
+			   to the transaction, we've already set some block
+			   pointers to 0, so we better follow through and free
+			   them, or we will introduce corruption (so break).
+			   This may be impossible, or at least rare, but I
+			   decided to cover the case regardless.
+
+			   If the buffer was not added to the transaction
+			   (this call), doing so would exceed our transaction
+			   size, so we need to end the transaction and start a
+			   new one (so goto). */
+
+			if (buf_in_tr)
+				break;
+			goto out_unlock;
+		}
+
+		gfs2_trans_add_meta(ip->i_gl, bh);
+		buf_in_tr = true;
+		*p = 0;
+		if (bstart + blen == bn) {
+			blen++;
+			continue;
+		}
+		if (bstart) {
+			__gfs2_free_blocks(ip, bstart, (u32)blen, meta);
+			(*btotal) += blen;
+			gfs2_add_inode_blocks(&ip->i_inode, -blen);
+		}
+		bstart = bn;
+		blen = 1;
+	}
+	if (bstart) {
+		__gfs2_free_blocks(ip, bstart, (u32)blen, meta);
+		(*btotal) += blen;
+		gfs2_add_inode_blocks(&ip->i_inode, -blen);
+	}
+out_unlock:
+	if (!ret && blks_outside_rgrp) { /* If buffer still has non-zero blocks
+					    outside the rgrp we just processed,
+					    do it all over again. */
+		if (current->journal_info) {
+			struct buffer_head *dibh = mp->mp_bh[0];
+
+			/* Every transaction boundary, we rewrite the dinode
+			   to keep its di_blocks current in case of failure. */
+			ip->i_inode.i_mtime = ip->i_inode.i_ctime =
+				CURRENT_TIME;
+			gfs2_trans_add_meta(ip->i_gl, dibh);
+			gfs2_dinode_out(ip, dibh->b_data);
+			up_write(&ip->i_rw_mutex);
+			gfs2_trans_end(sdp);
+		}
+		gfs2_glock_dq_uninit(rd_gh);
+		cond_resched();
+		goto more_rgrps;
+	}
+out:
+	return ret;
+}
+
+/**
+ * find_nonnull_ptr - find a non-null pointer given a metapath and height
+ * assumes the metapath is valid (with buffers) out to height h
+ * @mp: starting metapath
+ * @h: desired height to search
+ *
+ * Returns: true if a non-null pointer was found in the metapath buffer
+ *          false if all remaining pointers are NULL in the buffer
+ */
+static bool find_nonnull_ptr(struct gfs2_sbd *sdp, struct metapath *mp,
+			     unsigned int h)
+{
+	__be64 *ptr;
+	unsigned int ptrs = hptrs(sdp, h) - 1;
+
+	while (true) {
+		ptr = metapointer(h, mp);
+		if (*ptr) /* if we have a non-null pointer */
+			return true;
+
+		if (mp->mp_list[h] < ptrs)
+			mp->mp_list[h]++;
+		else
+			return false; /* no more pointers in this buffer */
+	}
+}
+
+enum dealloc_states {
+	DEALLOC_MP_FULL = 0,    /* Strip a metapath with all buffers read in */
+	DEALLOC_MP_LOWER = 1,   /* lower the metapath strip height */
+	DEALLOC_FILL_MP = 2,  /* Fill in the metapath to the given height. */
+	DEALLOC_DONE = 3,       /* process complete */
+};
 
-	if (!size)
+/**
+ * trunc_dealloc - truncate a file down to a desired size
+ * @ip: inode to truncate
+ * @newsize: The desired size of the file
+ *
+ * This function truncates a file to newsize. It works from the
+ * bottom up, and from the right to the left. In other words, it strips off
+ * the highest layer (data) before stripping any of the metadata. Doing it
+ * this way is best in case the operation is interrupted by power failure, etc.
+ * The dinode is rewritten in every transaction to guarantee integrity.
+ */
+static int trunc_dealloc(struct gfs2_inode *ip, u64 newsize)
+{
+	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
+	struct metapath mp;
+	struct buffer_head *dibh, *bh;
+	struct gfs2_holder rd_gh;
+	u64 lblock;
+	__u16 nbof[GFS2_MAX_META_HEIGHT]; /* new beginning of truncation */
+	unsigned int strip_h = ip->i_height - 1;
+	u32 btotal = 0;
+	int ret, state;
+	int mp_h; /* metapath buffers are read in to this height */
+	sector_t last_ra = 0;
+	u64 prev_bnr = 0;
+	bool preserve1; /* need to preserve the first meta pointer? */
+
+	if (!newsize)
 		lblock = 0;
 	else
-		lblock = (size - 1) >> sdp->sd_sb.sb_bsize_shift;
+		lblock = (newsize - 1) >> sdp->sd_sb.sb_bsize_shift;
 
+	memset(&mp, 0, sizeof(mp));
 	find_metapath(sdp, lblock, &mp, ip->i_height);
-	error = gfs2_rindex_update(sdp);
-	if (error)
-		return error;
 
-	error = gfs2_quota_hold(ip, NO_UID_QUOTA_CHANGE, NO_GID_QUOTA_CHANGE);
-	if (error)
-		return error;
+	memcpy(&nbof, &mp.mp_list, sizeof(nbof));
+
+	ret = gfs2_meta_inode_buffer(ip, &dibh);
+	if (ret)
+		return ret;
 
-	while (height--) {
-		struct strip_mine sm;
-		sm.sm_first = !!size;
-		sm.sm_height = height;
+	mp.mp_bh[0] = dibh;
+	ret = lookup_metapath(ip, &mp);
+	if (ret == ip->i_height)
+		state = DEALLOC_MP_FULL; /* We have a complete metapath */
+	else
+		state = DEALLOC_FILL_MP; /* deal with partial metapath */
 
-		error = recursive_scan(ip, NULL, &mp, 0, 0, 1, &sm);
-		if (error)
+	ret = gfs2_rindex_update(sdp);
+	if (ret)
+		goto out_metapath;
+
+	ret = gfs2_quota_hold(ip, NO_UID_QUOTA_CHANGE, NO_GID_QUOTA_CHANGE);
+	if (ret)
+		goto out_metapath;
+	gfs2_holder_mark_uninitialized(&rd_gh);
+
+	mp_h = strip_h;
+
+	while (state != DEALLOC_DONE) {
+		switch (state) {
+		/* Truncate a full metapath at the given strip height.
+		 * Note that strip_h == mp_h in order to be in this state. */
+		case DEALLOC_MP_FULL:
+			if (mp_h > 0) { /* issue read-ahead on metadata */
+				__be64 *top;
+
+				bh = mp.mp_bh[mp_h - 1];
+				if (bh->b_blocknr != last_ra) {
+					last_ra = bh->b_blocknr;
+					top = metaptr1(mp_h - 1, &mp);
+					gfs2_metapath_ra(ip->i_gl, bh, top);
+				}
+			}
+			/* If we're truncating to a non-zero size and the mp is
+			   at the beginning of file for the strip height, we
+			   need to preserve the first metadata pointer. */
+			preserve1 = (newsize &&
+				     (mp.mp_list[mp_h] == nbof[mp_h]));
+			bh = mp.mp_bh[mp_h];
+			gfs2_assert_withdraw(sdp, bh);
+			if (gfs2_assert_withdraw(sdp,
+						 prev_bnr != bh->b_blocknr)) {
+				printk(KERN_EMERG "GFS2: fsid=%s:inode %llu, "
+				       "block:%llu, i_h:%u, s_h:%u, mp_h:%u\n",
+				       sdp->sd_fsname,
+				       (unsigned long long)ip->i_no_addr,
+				       prev_bnr, ip->i_height, strip_h, mp_h);
+			}
+			prev_bnr = bh->b_blocknr;
+			ret = sweep_bh_for_rgrps(ip, &rd_gh, &mp, &btotal,
+						 mp_h, preserve1);
+			/* If we hit an error or just swept dinode buffer,
+			   just exit. */
+			if (ret || !mp_h) {
+				state = DEALLOC_DONE;
+				break;
+			}
+			state = DEALLOC_MP_LOWER;
+			break;
+
+		/* lower the metapath strip height */
+		case DEALLOC_MP_LOWER:
+			/* We're done with the current buffer, so release it,
+			   unless it's the dinode buffer. Then back up to the
+			   previous pointer. */
+			if (mp_h) {
+				brelse(mp.mp_bh[mp_h]);
+				mp.mp_bh[mp_h] = NULL;
+			}
+			/* If we can't get any lower in height, we've stripped
+			   off all we can. Next step is to back up and start
+			   stripping the previous level of metadata. */
+			if (mp_h == 0) {
+				strip_h--;
+				memcpy(&mp.mp_list, &nbof, sizeof(nbof));
+				mp_h = strip_h;
+				state = DEALLOC_FILL_MP;
+				break;
+			}
+			mp.mp_list[mp_h] = 0;
+			mp_h--; /* search one metadata height down */
+			if (mp.mp_list[mp_h] >= hptrs(sdp, mp_h) - 1)
+				break; /* loop around in the same state */
+			mp.mp_list[mp_h]++;
+			/* Here we've found a part of the metapath that is not
+			 * allocated. We need to search at that height for the
+			 * next non-null pointer. */
+			if (find_nonnull_ptr(sdp, &mp, mp_h)) {
+				state = DEALLOC_FILL_MP;
+				mp_h++;
+			}
+			/* No more non-null pointers at this height. Back up
+			   to the previous height and try again. */
+			break; /* loop around in the same state */
+
+		/* Fill the metapath with buffers to the given height. */
+		case DEALLOC_FILL_MP:
+			/* Fill the buffers out to the current height. */
+			ret = fillup_metapath(ip, &mp, mp_h);
+			if (ret < 0)
+				goto out;
+
+			/* If buffers found for the entire strip height */
+			if ((ret == ip->i_height) && (mp_h == strip_h)) {
+				state = DEALLOC_MP_FULL;
+				break;
+			}
+			if (ret < ip->i_height) /* We have a partial height */
+				mp_h = ret - 1;
+
+			/* If we find a non-null block pointer, crawl a bit
+			   higher up in the metapath and try again, otherwise
+			   we need to look lower for a new starting point. */
+			if (find_nonnull_ptr(sdp, &mp, mp_h))
+				mp_h++;
+			else
+				state = DEALLOC_MP_LOWER;
 			break;
+		}
 	}
 
-	gfs2_quota_unhold(ip);
+	if (btotal) {
+		if (current->journal_info == NULL) {
+			ret = gfs2_trans_begin(sdp, RES_DINODE + RES_STATFS +
+					       RES_QUOTA, 0);
+			if (ret)
+				goto out;
+			down_write(&ip->i_rw_mutex);
+		}
+		gfs2_statfs_change(sdp, 0, +btotal, 0);
+		gfs2_quota_change(ip, -(s64)btotal, ip->i_inode.i_uid,
+				  ip->i_inode.i_gid);
+		ip->i_inode.i_mtime = ip->i_inode.i_ctime = CURRENT_TIME;
+		gfs2_trans_add_meta(ip->i_gl, dibh);
+		gfs2_dinode_out(ip, dibh->b_data);
+		up_write(&ip->i_rw_mutex);
+		gfs2_trans_end(sdp);
+	}
 
-	return error;
+out:
+	if (gfs2_holder_initialized(&rd_gh))
+		gfs2_glock_dq_uninit(&rd_gh);
+	if (current->journal_info) {
+		up_write(&ip->i_rw_mutex);
+		gfs2_trans_end(sdp);
+		cond_resched();
+	}
+	gfs2_quota_unhold(ip);
+out_metapath:
+	release_metapath(&mp);
+	return ret;
 }
 
 static int trunc_end(struct gfs2_inode *ip)
diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 86ccc015..83c9909 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -483,13 +483,6 @@ void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd)
 	}
 }
 
-static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block)
-{
-	u64 first = rgd->rd_data0;
-	u64 last = first + rgd->rd_data;
-	return first <= block && block < last;
-}
-
 /**
  * gfs2_blk2rgrpd - Find resource group for a given data/meta block number
  * @sdp: The GFS2 superblock
diff --git a/fs/gfs2/rgrp.h b/fs/gfs2/rgrp.h
index 66b51cf..e90478e 100644
--- a/fs/gfs2/rgrp.h
+++ b/fs/gfs2/rgrp.h
@@ -83,5 +83,12 @@ static inline bool gfs2_rs_active(const struct gfs2_blkreserv *rs)
 	return rs && !RB_EMPTY_NODE(&rs->rs_node);
 }
 
+static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block)
+{
+	u64 first = rgd->rd_data0;
+	u64 last = first + rgd->rd_data;
+	return first <= block && block < last;
+}
+
 extern void check_and_update_goal(struct gfs2_inode *ip);
 #endif /* __RGRP_DOT_H__ */
-- 
2.9.3