[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 only got 9 GFS2 patches for this merge window:

1. Abhi Das contributed a patch to report journal recovery times more
   accurately during journal replay.
2. Andreas Gruenbacher contributed a patch to fix fallocate chunk size.
3. Andreas added a patch to correctly dirty inodes during rename.
4. Andreas added a patch to improve the comment for function gfs2_block_map.
5. Andreas added a patch to improve kernel trace point iomap end:
   The physical block address was added.
6. Andreas added a patch to fix a nasty file system corruption bug that
   surfaced in xfstests 476 in punch-hole/truncate.
7. Andreas fixed a problem Christoph Helwig pointed out, namely, that GFS2
   was misusing the IOMAP_ZERO flag. The zeroing of new blocks was moved
   to the proper fallocate code.
8. I contributed a patch to declare function gfs2_remove_from_ail as static.
9. I added a patch to only set PageChecked for jdata page writes.

Regards,

Bob Peterson
---
Abhi Das (1):
  gfs2: time journal recovery steps accurately

Andreas Gruenbacher (6):
  gfs2: Fix fallocate chunk size
  gfs2: Dirty source inode during rename
  gfs2: Improve gfs2_block_map comment
  gfs2: gfs2_iomap_end tracepoint: log block address
  gfs2: Check for the end of metadata in punch_hole
  gfs2: Zero out fallocated blocks in fallocate_chunk

Bob Peterson (2):
  GFS2: Make function gfs2_remove_from_ail static
  GFS2: Only set PageChecked for jdata pages

 fs/gfs2/aops.c       |  8 ++++----
 fs/gfs2/bmap.c       | 38 ++++++++++++++++++++------------------
 fs/gfs2/dir.c        | 13 ++-----------
 fs/gfs2/file.c       | 34 +++++++++++++++++-----------------
 fs/gfs2/incore.h     |  3 ---
 fs/gfs2/inode.c      | 10 +---------
 fs/gfs2/log.c        |  2 +-
 fs/gfs2/log.h        |  1 -
 fs/gfs2/quota.h      |  2 ++
 fs/gfs2/recovery.c   | 20 ++++++++++++++------
 fs/gfs2/trace_gfs2.h |  9 +++++++--
 11 files changed, 68 insertions(+), 72 deletions(-)

-- 
2.14.3

[Cluster-devel] [PATCH 1/9] gfs2: Fix fallocate chunk size

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

The chunk size of allocations in __gfs2_fallocate is calculated
incorrectly.  The size can collapse, causing __gfs2_fallocate to
allocate one block at a time, which is very inefficient.  This needs
fixing in two places:

In gfs2_quota_lock_check, always set ap->allowed to UINT_MAX to indicate
that there is no quota limit.  This fixes callers that rely on
ap->allowed to be set even when quotas are off.

In __gfs2_fallocate, reset max_blks to UINT_MAX in each iteration of the
loop to make sure that allocation limits from one resource group won't
spill over into another resource group.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/file.c  | 5 +++--
 fs/gfs2/quota.h | 2 ++
 2 files changed, 5 insertions(+), 2 deletions(-)

diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 4f88e201b3f0..2edd3a9a7b79 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -809,7 +809,7 @@ static long __gfs2_fallocate(struct file *file, int mode, loff_t offset, loff_t
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_alloc_parms ap = { .aflags = 0, };
 	unsigned int data_blocks = 0, ind_blocks = 0, rblocks;
-	loff_t bytes, max_bytes, max_blks = UINT_MAX;
+	loff_t bytes, max_bytes, max_blks;
 	int error;
 	const loff_t pos = offset;
 	const loff_t count = len;
@@ -861,7 +861,8 @@ static long __gfs2_fallocate(struct file *file, int mode, loff_t offset, loff_t
 			return error;
 		/* ap.allowed tells us how many blocks quota will allow
 		 * us to write. Check if this reduces max_blks */
-		if (ap.allowed && ap.allowed < max_blks)
+		max_blks = UINT_MAX;
+		if (ap.allowed)
 			max_blks = ap.allowed;
 
 		error = gfs2_inplace_reserve(ip, &ap);
diff --git a/fs/gfs2/quota.h b/fs/gfs2/quota.h
index 5e47c935a515..836f29480be6 100644
--- a/fs/gfs2/quota.h
+++ b/fs/gfs2/quota.h
@@ -45,6 +45,8 @@ static inline int gfs2_quota_lock_check(struct gfs2_inode *ip,
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
 	int ret;
+
+	ap->allowed = UINT_MAX; /* Assume we are permitted a whole lot */
 	if (sdp->sd_args.ar_quota == GFS2_QUOTA_OFF)
 		return 0;
 	ret = gfs2_quota_lock(ip, NO_UID_QUOTA_CHANGE, NO_GID_QUOTA_CHANGE);
-- 
2.14.3

[Cluster-devel] [PATCH 2/9] gfs2: Dirty source inode during rename

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Mark the source inode dirty during a rename instead of just updating the
underlying buffer head.  Otherwise, fsync may find the inode clean and
will then skip flushing the journal.  A subsequent power failure will
cause the rename to be lost.  This happens in command sequences like:

  xfs_io -f -c 'pwrite 0 4096' -c 'fsync' foo
  mv foo bar
  xfs_io -c 'fsync' bar
  # power failure

Fixes xfstests generic/322, generic/376.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/dir.c   | 13 ++-----------
 fs/gfs2/inode.c | 10 +---------
 2 files changed, 3 insertions(+), 20 deletions(-)

diff --git a/fs/gfs2/dir.c b/fs/gfs2/dir.c
index 7c21aea0266b..d9fb0ad6cc30 100644
--- a/fs/gfs2/dir.c
+++ b/fs/gfs2/dir.c
@@ -1940,7 +1940,6 @@ int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
 {
 	struct buffer_head *bh;
 	struct gfs2_dirent *dent;
-	int error;
 
 	dent = gfs2_dirent_search(&dip->i_inode, filename, gfs2_dirent_find, &bh);
 	if (!dent) {
@@ -1953,18 +1952,10 @@ int gfs2_dir_mvino(struct gfs2_inode *dip, const struct qstr *filename,
 	gfs2_trans_add_meta(dip->i_gl, bh);
 	gfs2_inum_out(nip, dent);
 	dent->de_type = cpu_to_be16(new_type);
-
-	if (dip->i_diskflags & GFS2_DIF_EXHASH) {
-		brelse(bh);
-		error = gfs2_meta_inode_buffer(dip, &bh);
-		if (error)
-			return error;
-		gfs2_trans_add_meta(dip->i_gl, bh);
-	}
+	brelse(bh);
 
 	dip->i_inode.i_mtime = dip->i_inode.i_ctime = current_time(&dip->i_inode);
-	gfs2_dinode_out(dip, bh->b_data);
-	brelse(bh);
+	mark_inode_dirty_sync(&dip->i_inode);
 	return 0;
 }
 
diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c
index 59e0560180ec..8700eb815638 100644
--- a/fs/gfs2/inode.c
+++ b/fs/gfs2/inode.c
@@ -1326,19 +1326,11 @@ static int gfs2_ok_to_move(struct gfs2_inode *this, struct gfs2_inode *to)
 static int update_moved_ino(struct gfs2_inode *ip, struct gfs2_inode *ndip,
 			    int dir_rename)
 {
-	int error;
-	struct buffer_head *dibh;
-
 	if (dir_rename)
 		return gfs2_dir_mvino(ip, &gfs2_qdotdot, ndip, DT_DIR);
 
-	error = gfs2_meta_inode_buffer(ip, &dibh);
-	if (error)
-		return error;
 	ip->i_inode.i_ctime = current_time(&ip->i_inode);
-	gfs2_trans_add_meta(ip->i_gl, dibh);
-	gfs2_dinode_out(ip, dibh->b_data);
-	brelse(dibh);
+	mark_inode_dirty_sync(&ip->i_inode);
 	return 0;
 }
 
-- 
2.14.3

[Cluster-devel] [PATCH 3/9] GFS2: Make function gfs2_remove_from_ail static

[Bob Peterson]

Function gfs2_remove_from_ail is only ever used from log.c, so there
is no reason to declare it extern. This patch removes the extern and
declares it static.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/log.c | 2 +-
 fs/gfs2/log.h | 1 -
 2 files changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index cf6b46247df4..0248835625f1 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -73,7 +73,7 @@ unsigned int gfs2_struct2blk(struct gfs2_sbd *sdp, unsigned int nstruct,
  *
  */
 
-void gfs2_remove_from_ail(struct gfs2_bufdata *bd)
+static void gfs2_remove_from_ail(struct gfs2_bufdata *bd)
 {
 	bd->bd_tr = NULL;
 	list_del_init(&bd->bd_ail_st_list);
diff --git a/fs/gfs2/log.h b/fs/gfs2/log.h
index 93b52ac1ca1f..1862e310a067 100644
--- a/fs/gfs2/log.h
+++ b/fs/gfs2/log.h
@@ -70,7 +70,6 @@ extern void gfs2_write_log_header(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd,
 extern void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl,
 			   u32 type);
 extern void gfs2_log_commit(struct gfs2_sbd *sdp, struct gfs2_trans *trans);
-extern void gfs2_remove_from_ail(struct gfs2_bufdata *bd);
 extern void gfs2_ail1_flush(struct gfs2_sbd *sdp, struct writeback_control *wbc);
 
 extern void gfs2_log_shutdown(struct gfs2_sbd *sdp);
-- 
2.14.3

[Cluster-devel] [PATCH 4/9] GFS2: Only set PageChecked for jdata pages

[Bob Peterson]

Before this patch, GFS2 was setting the PageChecked flag for ordered
write pages. This is unnecessary. The ext3 file system only does it
for jdata, and it's only used in jdata circumstances. It only muddies
the already murky waters of writing pages in the aops.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/aops.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
index 2f725b4a386b..f58716567972 100644
--- a/fs/gfs2/aops.c
+++ b/fs/gfs2/aops.c
@@ -940,13 +940,13 @@ static int gfs2_write_end(struct file *file, struct address_space *mapping,
 }
 
 /**
- * gfs2_set_page_dirty - Page dirtying function
+ * jdata_set_page_dirty - Page dirtying function
  * @page: The page to dirty
  *
  * Returns: 1 if it dirtyed the page, or 0 otherwise
  */
  
-static int gfs2_set_page_dirty(struct page *page)
+static int jdata_set_page_dirty(struct page *page)
 {
 	SetPageChecked(page);
 	return __set_page_dirty_buffers(page);
@@ -1214,7 +1214,7 @@ static const struct address_space_operations gfs2_ordered_aops = {
 	.readpages = gfs2_readpages,
 	.write_begin = gfs2_write_begin,
 	.write_end = gfs2_write_end,
-	.set_page_dirty = gfs2_set_page_dirty,
+	.set_page_dirty = __set_page_dirty_buffers,
 	.bmap = gfs2_bmap,
 	.invalidatepage = gfs2_invalidatepage,
 	.releasepage = gfs2_releasepage,
@@ -1231,7 +1231,7 @@ static const struct address_space_operations gfs2_jdata_aops = {
 	.readpages = gfs2_readpages,
 	.write_begin = gfs2_write_begin,
 	.write_end = gfs2_write_end,
-	.set_page_dirty = gfs2_set_page_dirty,
+	.set_page_dirty = jdata_set_page_dirty,
 	.bmap = gfs2_bmap,
 	.invalidatepage = gfs2_invalidatepage,
 	.releasepage = gfs2_releasepage,
-- 
2.14.3

[Cluster-devel] [PATCH 5/9] gfs2: Improve gfs2_block_map comment

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/bmap.c | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 51f940e76c5e..8c25a64d8ae0 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -812,15 +812,22 @@ int gfs2_iomap_begin(struct inode *inode, loff_t pos, loff_t length,
 }
 
 /**
- * gfs2_block_map - Map a block from an inode to a disk block
+ * gfs2_block_map - Map one or more blocks of an inode to a disk block
  * @inode: The inode
  * @lblock: The logical block number
  * @bh_map: The bh to be mapped
  * @create: True if its ok to alloc blocks to satify the request
  *
- * Sets buffer_mapped() if successful, sets buffer_boundary() if a
- * read of metadata will be required before the next block can be
- * mapped. Sets buffer_new() if new blocks were allocated.
+ * The size of the requested mapping is defined in bh_map->b_size.
+ *
+ * Clears buffer_mapped(bh_map) and leaves bh_map->b_size unchanged
+ * when @lblock is not mapped.  Sets buffer_mapped(bh_map) and
+ * bh_map->b_size to indicate the size of the mapping when @lblock and
+ * successive blocks are mapped, up to the requested size.
+ *
+ * Sets buffer_boundary() if a read of metadata will be required
+ * before the next block can be mapped. Sets buffer_new() if new
+ * blocks were allocated.
  *
  * Returns: errno
  */
-- 
2.14.3

[Cluster-devel] [PATCH 6/9] gfs2: gfs2_iomap_end tracepoint: log block address

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

In the gfs2_iomap_end tracepoint, log the physical block address, just
as in the gfs2_bmap tracepoint.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/trace_gfs2.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/fs/gfs2/trace_gfs2.h b/fs/gfs2/trace_gfs2.h
index b9318b49ff8f..cb10b95efe0f 100644
--- a/fs/gfs2/trace_gfs2.h
+++ b/fs/gfs2/trace_gfs2.h
@@ -515,6 +515,7 @@ TRACE_EVENT(gfs2_iomap_end,
 		__field(	u64,	inum			)
 		__field(	loff_t, offset			)
 		__field(	ssize_t, length			)
+		__field(	sector_t, pblock		)
 		__field(	u16,	flags			)
 		__field(	u16,	type			)
 		__field(	int,	ret			)
@@ -525,16 +526,20 @@ TRACE_EVENT(gfs2_iomap_end,
 		__entry->inum		= ip->i_no_addr;
 		__entry->offset		= iomap->offset;
 		__entry->length		= iomap->length;
+		__entry->pblock		= iomap->addr == IOMAP_NULL_ADDR ? 0 :
+					 (iomap->addr >> ip->i_inode.i_blkbits);
 		__entry->flags		= iomap->flags;
 		__entry->type		= iomap->type;
 		__entry->ret		= ret;
 	),
 
-	TP_printk("%u,%u bmap %llu iomap end %llu/%lu ty:%d flags:%08x rc:%d",
+	TP_printk("%u,%u bmap %llu iomap end %llu/%lu to %llu ty:%d flags:%08x rc:%d",
 		  MAJOR(__entry->dev), MINOR(__entry->dev),
 		  (unsigned long long)__entry->inum,
 		  (unsigned long long)__entry->offset,
-		  (unsigned long)__entry->length, (u16)__entry->type,
+		  (unsigned long)__entry->length,
+		  (long long)__entry->pblock,
+		  (u16)__entry->type,
 		  (u16)__entry->flags, __entry->ret)
 );
 
-- 
2.14.3

[Cluster-devel] [PATCH 7/9] gfs2: Check for the end of metadata in punch_hole

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

When punching a hole or truncating an inode down to a given size, also
check if the truncate point / start of the hole is within the range we
have metadata for.  Otherwise, we can end up freeing blocks that
shouldn't be freed, corrupting the inode, or crashing the machine when
trying to punch a hole into the void.

When growing an inode via truncate, we set the new size but we don't
allocate additional levels of indirect blocks and grow the inode height.
When shrinking that inode again, the new size may still point beyond the
end of the inode's metadata.

Fixes xfstest generic/476.

Debugged-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/bmap.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index 8c25a64d8ae0..ce4a978e0c18 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -1351,6 +1351,7 @@ static inline bool walk_done(struct gfs2_sbd *sdp,
 static int punch_hole(struct gfs2_inode *ip, u64 offset, u64 length)
 {
 	struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
+	u64 maxsize = sdp->sd_heightsize[ip->i_height];
 	struct metapath mp = {};
 	struct buffer_head *dibh, *bh;
 	struct gfs2_holder rd_gh;
@@ -1366,6 +1367,14 @@ static int punch_hole(struct gfs2_inode *ip, u64 offset, u64 length)
 	u64 prev_bnr = 0;
 	__be64 *start, *end;
 
+	if (offset >= maxsize) {
+		/*
+		 * The starting point lies beyond the allocated meta-data;
+		 * there are no blocks do deallocate.
+		 */
+		return 0;
+	}
+
 	/*
 	 * The start position of the hole is defined by lblock, start_list, and
 	 * start_aligned.  The end position of the hole is defined by lend,
@@ -1379,7 +1388,6 @@ static int punch_hole(struct gfs2_inode *ip, u64 offset, u64 length)
 	 */
 
 	if (length) {
-		u64 maxsize = sdp->sd_heightsize[ip->i_height];
 		u64 end_offset = offset + length;
 		u64 lend;
 
-- 
2.14.3

[Cluster-devel] [PATCH 8/9] gfs2: Zero out fallocated blocks in fallocate_chunk

[Bob Peterson]

From: Andreas Gruenbacher <agruenba@redhat.com>

Instead of zeroing out fallocated blocks in gfs2_iomap_alloc, zero them
out in fallocate_chunk, much higher up the call stack.  This gets rid of
gfs2's abuse of the IOMAP_ZERO flag as well as the gfs2 specific zeronew
buffer flag.  I can't think of a reason why zeroing out the blocks in
gfs2_iomap_alloc would have any benefits: there is no additional locking
at that level that would add protection to the newly allocated blocks.

While at it, change fallocate over from gs2_block_map to gfs2_iomap_begin.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Acked-by: Christoph Hellwig <hch@lst.de>
---
 fs/gfs2/bmap.c   | 13 -------------
 fs/gfs2/file.c   | 29 ++++++++++++++---------------
 fs/gfs2/incore.h |  3 ---
 3 files changed, 14 insertions(+), 31 deletions(-)

diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c
index ce4a978e0c18..685c305cbeb6 100644
--- a/fs/gfs2/bmap.c
+++ b/fs/gfs2/bmap.c
@@ -491,14 +491,12 @@ static int gfs2_iomap_alloc(struct inode *inode, struct iomap *iomap,
 {
 	struct gfs2_inode *ip = GFS2_I(inode);
 	struct gfs2_sbd *sdp = GFS2_SB(inode);
-	struct super_block *sb = sdp->sd_vfs;
 	struct buffer_head *dibh = mp->mp_bh[0];
 	u64 bn;
 	unsigned n, i, blks, alloced = 0, iblks = 0, branch_start = 0;
 	unsigned dblks = 0;
 	unsigned ptrs_per_blk;
 	const unsigned end_of_metadata = mp->mp_fheight - 1;
-	int ret;
 	enum alloc_state state;
 	__be64 *ptr;
 	__be64 zero_bn = 0;
@@ -607,15 +605,6 @@ static int gfs2_iomap_alloc(struct inode *inode, struct iomap *iomap,
 			iomap->flags |= IOMAP_F_NEW;
 			while (n-- > 0)
 				*ptr++ = cpu_to_be64(bn++);
-			if (flags & IOMAP_ZERO) {
-				ret = sb_issue_zeroout(sb, iomap->addr >> inode->i_blkbits,
-						       dblks, GFP_NOFS);
-				if (ret) {
-					fs_err(sdp,
-					       "Failed to zero data buffers\n");
-					flags &= ~IOMAP_ZERO;
-				}
-			}
 			break;
 		}
 	} while (iomap->addr == IOMAP_NULL_ADDR);
@@ -846,8 +835,6 @@ int gfs2_block_map(struct inode *inode, sector_t lblock,
 
 	if (create)
 		flags |= IOMAP_WRITE;
-	if (buffer_zeronew(bh_map))
-		flags |= IOMAP_ZERO;
 	ret = gfs2_iomap_begin(inode, (loff_t)lblock << inode->i_blkbits,
 			       bh_map->b_size, flags, &iomap);
 	if (ret) {
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 2edd3a9a7b79..4b71f021a9e2 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -729,11 +729,12 @@ static ssize_t gfs2_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
 static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len,
 			   int mode)
 {
+	struct super_block *sb = inode->i_sb;
 	struct gfs2_inode *ip = GFS2_I(inode);
+	loff_t end = offset + len;
 	struct buffer_head *dibh;
+	struct iomap iomap;
 	int error;
-	unsigned int nr_blks;
-	sector_t lblock = offset >> inode->i_blkbits;
 
 	error = gfs2_meta_inode_buffer(ip, &dibh);
 	if (unlikely(error))
@@ -747,21 +748,19 @@ static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len,
 			goto out;
 	}
 
-	while (len) {
-		struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };
-		bh_map.b_size = len;
-		set_buffer_zeronew(&bh_map);
-
-		error = gfs2_block_map(inode, lblock, &bh_map, 1);
-		if (unlikely(error))
+	while (offset < end) {
+		error = gfs2_iomap_begin(inode, offset, end - offset,
+					 IOMAP_WRITE, &iomap);
+		if (error)
 			goto out;
-		len -= bh_map.b_size;
-		nr_blks = bh_map.b_size >> inode->i_blkbits;
-		lblock += nr_blks;
-		if (!buffer_new(&bh_map))
+		offset = iomap.offset + iomap.length;
+		if (iomap.type != IOMAP_HOLE)
 			continue;
-		if (unlikely(!buffer_zeronew(&bh_map))) {
-			error = -EIO;
+		error = sb_issue_zeroout(sb, iomap.addr >> inode->i_blkbits,
+					 iomap.length >> inode->i_blkbits,
+					 GFP_NOFS);
+		if (error) {
+			fs_err(GFS2_SB(inode), "Failed to zero data buffers\n");
 			goto out;
 		}
 	}
diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h
index e0557b8a590a..1b6b1e3f5caf 100644
--- a/fs/gfs2/incore.h
+++ b/fs/gfs2/incore.h
@@ -130,15 +130,12 @@ static inline bool gfs2_rbm_eq(const struct gfs2_rbm *rbm1,
 enum gfs2_state_bits {
 	BH_Pinned = BH_PrivateStart,
 	BH_Escaped = BH_PrivateStart + 1,
-	BH_Zeronew = BH_PrivateStart + 2,
 };
 
 BUFFER_FNS(Pinned, pinned)
 TAS_BUFFER_FNS(Pinned, pinned)
 BUFFER_FNS(Escaped, escaped)
 TAS_BUFFER_FNS(Escaped, escaped)
-BUFFER_FNS(Zeronew, zeronew)
-TAS_BUFFER_FNS(Zeronew, zeronew)
 
 struct gfs2_bufdata {
 	struct buffer_head *bd_bh;
-- 
2.14.3

[Cluster-devel] [PATCH 9/9] gfs2: time journal recovery steps accurately

[Bob Peterson]

From: Abhi Das <adas@redhat.com>

This patch spits out the time taken by the various steps in the
journal recover process. Previously, the journal recovery time
didn't account for finding the journal head in the log which takes
up a significant portion of time.

Signed-off-by: Abhi Das <adas@redhat.com>
Signed-off-by: Bob Peterson <rpeterso@redhat.com>
---
 fs/gfs2/recovery.c | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

diff --git a/fs/gfs2/recovery.c b/fs/gfs2/recovery.c
index b6b258998bcd..d8b622c375ab 100644
--- a/fs/gfs2/recovery.c
+++ b/fs/gfs2/recovery.c
@@ -15,6 +15,7 @@
 #include <linux/gfs2_ondisk.h>
 #include <linux/crc32.h>
 #include <linux/crc32c.h>
+#include <linux/ktime.h>
 
 #include "gfs2.h"
 #include "incore.h"
@@ -409,12 +410,13 @@ void gfs2_recover_func(struct work_struct *work)
 	struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
 	struct gfs2_log_header_host head;
 	struct gfs2_holder j_gh, ji_gh, thaw_gh;
-	unsigned long t;
+	ktime_t t_start, t_jlck, t_jhd, t_tlck, t_rep;
 	int ro = 0;
 	unsigned int pass;
 	int error;
 	int jlocked = 0;
 
+	t_start = ktime_get();
 	if (sdp->sd_args.ar_spectator ||
 	    (jd->jd_jid != sdp->sd_lockstruct.ls_jid)) {
 		fs_info(sdp, "jid=%u: Trying to acquire journal lock...\n",
@@ -446,6 +448,7 @@ void gfs2_recover_func(struct work_struct *work)
 		fs_info(sdp, "jid=%u, already locked for use\n", jd->jd_jid);
 	}
 
+	t_jlck = ktime_get();
 	fs_info(sdp, "jid=%u: Looking at journal...\n", jd->jd_jid);
 
 	error = gfs2_jdesc_check(jd);
@@ -455,13 +458,12 @@ void gfs2_recover_func(struct work_struct *work)
 	error = gfs2_find_jhead(jd, &head);
 	if (error)
 		goto fail_gunlock_ji;
+	t_jhd = ktime_get();
 
 	if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
 		fs_info(sdp, "jid=%u: Acquiring the transaction lock...\n",
 			jd->jd_jid);
 
-		t = jiffies;
-
 		/* Acquire a shared hold on the freeze lock */
 
 		error = gfs2_glock_nq_init(sdp->sd_freeze_gl, LM_ST_SHARED,
@@ -495,6 +497,7 @@ void gfs2_recover_func(struct work_struct *work)
 			goto fail_gunlock_thaw;
 		}
 
+		t_tlck = ktime_get();
 		fs_info(sdp, "jid=%u: Replaying journal...\n", jd->jd_jid);
 
 		for (pass = 0; pass < 2; pass++) {
@@ -509,9 +512,14 @@ void gfs2_recover_func(struct work_struct *work)
 		clean_journal(jd, &head);
 
 		gfs2_glock_dq_uninit(&thaw_gh);
-		t = DIV_ROUND_UP(jiffies - t, HZ);
-		fs_info(sdp, "jid=%u: Journal replayed in %lus\n",
-			jd->jd_jid, t);
+		t_rep = ktime_get();
+		fs_info(sdp, "jid=%u: Journal replayed in %lldms [jlck:%lldms, "
+			"jhead:%lldms, tlck:%lldms, replay:%lldms]\n",
+			jd->jd_jid, ktime_ms_delta(t_rep, t_start),
+			ktime_ms_delta(t_jlck, t_start),
+			ktime_ms_delta(t_jhd, t_jlck),
+			ktime_ms_delta(t_tlck, t_jhd),
+			ktime_ms_delta(t_rep, t_tlck));
 	}
 
 	gfs2_recovery_done(sdp, jd->jd_jid, LM_RD_SUCCESS);
-- 
2.14.3