Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (61 commits)
  jbd2: Add MAINTAINERS entry
  jbd2: fix a potential leak of a journal_head on an error path
  ext4: teach ext4_ext_split to calculate extents efficiently
  ext4: Convert ext4 to new truncate calling convention
  ext4: do not normalize block requests from fallocate()
  ext4: enable "punch hole" functionality
  ext4: add "punch hole" flag to ext4_map_blocks()
  ext4: punch out extents
  ext4: add new function ext4_block_zero_page_range()
  ext4: add flag to ext4_has_free_blocks
  ext4: reserve inodes and feature code for 'quota' feature
  ext4: add support for multiple mount protection
  ext4: ensure f_bfree returned by ext4_statfs() is non-negative
  ext4: protect bb_first_free in ext4_trim_all_free() with group lock
  ext4: only load buddy bitmap in ext4_trim_fs() when it is needed
  jbd2: Fix comment to match the code in jbd2__journal_start()
  ext4: fix waiting and sending of a barrier in ext4_sync_file()
  jbd2: Add function jbd2_trans_will_send_data_barrier()
  jbd2: fix sending of data flush on journal commit
  ext4: fix ext4_ext_fiemap_cb() to handle blocks before request range correctly
  ...

Documentation/filesystems/ext4.txt

@@ -226,10 +226,6 @@ acl			Enables POSIX Access Control Lists support.
 noacl			This option disables POSIX Access Control List
 			support.
 
-reservation
-
-noreservation
-
 bsddf		(*)	Make 'df' act like BSD.
 minixdf			Make 'df' act like Minix.
 

MAINTAINERS

@@ -3572,9 +3572,16 @@ M:	Andrew Morton <akpm@linux-foundation.org>
 M:	Jan Kara <jack@suse.cz>
 L:	linux-ext4@vger.kernel.org
 S:	Maintained
-F:	fs/jbd*/
-F:	include/linux/ext*jbd*.h
-F:	include/linux/jbd*.h
+F:	fs/jbd/
+F:	include/linux/ext3_jbd.h
+F:	include/linux/jbd.h
+
+JOURNALLING LAYER FOR BLOCK DEVICES (JBD2)
+M:	"Theodore Ts'o" <tytso@mit.edu>
+L:	linux-ext4@vger.kernel.org
+S:	Maintained
+F:	fs/jbd2/
+F:	include/linux/jbd2.h
 
 JSM Neo PCI based serial card
 M:	Breno Leitao <leitao@linux.vnet.ibm.com>

fs/ext4/Makefile

@@ -6,7 +6,8 @@ obj-$(CONFIG_EXT4_FS) += ext4.o
 
 ext4-y	:= balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o page-io.o \
 		ioctl.o namei.o super.o symlink.o hash.o resize.o extents.o \
-		ext4_jbd2.o migrate.o mballoc.o block_validity.o move_extent.o
+		ext4_jbd2.o migrate.o mballoc.o block_validity.o move_extent.o \
+		mmp.o
 
 ext4-$(CONFIG_EXT4_FS_XATTR)		+= xattr.o xattr_user.o xattr_trusted.o
 ext4-$(CONFIG_EXT4_FS_POSIX_ACL)	+= acl.o

fs/ext4/balloc.c

@@ -361,130 +361,6 @@ ext4_read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
 	return bh;
 }
 
-/**
- * ext4_add_groupblocks() -- Add given blocks to an existing group
- * @handle:			handle to this transaction
- * @sb:				super block
- * @block:			start physcial block to add to the block group
- * @count:			number of blocks to free
- *
- * This marks the blocks as free in the bitmap. We ask the
- * mballoc to reload the buddy after this by setting group
- * EXT4_GROUP_INFO_NEED_INIT_BIT flag
- */
-void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
-			 ext4_fsblk_t block, unsigned long count)
-{
-	struct buffer_head *bitmap_bh = NULL;
-	struct buffer_head *gd_bh;
-	ext4_group_t block_group;
-	ext4_grpblk_t bit;
-	unsigned int i;
-	struct ext4_group_desc *desc;
-	struct ext4_sb_info *sbi = EXT4_SB(sb);
-	int err = 0, ret, blk_free_count;
-	ext4_grpblk_t blocks_freed;
-	struct ext4_group_info *grp;
-
-	ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
-
-	ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
-	grp = ext4_get_group_info(sb, block_group);
-	/*
-	 * Check to see if we are freeing blocks across a group
-	 * boundary.
-	 */
-	if (bit + count > EXT4_BLOCKS_PER_GROUP(sb)) {
-		goto error_return;
-	}
-	bitmap_bh = ext4_read_block_bitmap(sb, block_group);
-	if (!bitmap_bh)
-		goto error_return;
-	desc = ext4_get_group_desc(sb, block_group, &gd_bh);
-	if (!desc)
-		goto error_return;
-
-	if (in_range(ext4_block_bitmap(sb, desc), block, count) ||
-	    in_range(ext4_inode_bitmap(sb, desc), block, count) ||
-	    in_range(block, ext4_inode_table(sb, desc), sbi->s_itb_per_group) ||
-	    in_range(block + count - 1, ext4_inode_table(sb, desc),
-		     sbi->s_itb_per_group)) {
-		ext4_error(sb, "Adding blocks in system zones - "
-			   "Block = %llu, count = %lu",
-			   block, count);
-		goto error_return;
-	}
-
-	/*
-	 * We are about to add blocks to the bitmap,
-	 * so we need undo access.
-	 */
-	BUFFER_TRACE(bitmap_bh, "getting undo access");
-	err = ext4_journal_get_undo_access(handle, bitmap_bh);
-	if (err)
-		goto error_return;
-
-	/*
-	 * We are about to modify some metadata.  Call the journal APIs
-	 * to unshare ->b_data if a currently-committing transaction is
-	 * using it
-	 */
-	BUFFER_TRACE(gd_bh, "get_write_access");
-	err = ext4_journal_get_write_access(handle, gd_bh);
-	if (err)
-		goto error_return;
-	/*
-	 * make sure we don't allow a parallel init on other groups in the
-	 * same buddy cache
-	 */
-	down_write(&grp->alloc_sem);
-	for (i = 0, blocks_freed = 0; i < count; i++) {
-		BUFFER_TRACE(bitmap_bh, "clear bit");
-		if (!ext4_clear_bit_atomic(ext4_group_lock_ptr(sb, block_group),
-						bit + i, bitmap_bh->b_data)) {
-			ext4_error(sb, "bit already cleared for block %llu",
-				   (ext4_fsblk_t)(block + i));
-			BUFFER_TRACE(bitmap_bh, "bit already cleared");
-		} else {
-			blocks_freed++;
-		}
-	}
-	ext4_lock_group(sb, block_group);
-	blk_free_count = blocks_freed + ext4_free_blks_count(sb, desc);
-	ext4_free_blks_set(sb, desc, blk_free_count);
-	desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
-	ext4_unlock_group(sb, block_group);
-	percpu_counter_add(&sbi->s_freeblocks_counter, blocks_freed);
-
-	if (sbi->s_log_groups_per_flex) {
-		ext4_group_t flex_group = ext4_flex_group(sbi, block_group);
-		atomic_add(blocks_freed,
-			   &sbi->s_flex_groups[flex_group].free_blocks);
-	}
-	/*
-	 * request to reload the buddy with the
-	 * new bitmap information
-	 */
-	set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &(grp->bb_state));
-	grp->bb_free += blocks_freed;
-	up_write(&grp->alloc_sem);
-
-	/* We dirtied the bitmap block */
-	BUFFER_TRACE(bitmap_bh, "dirtied bitmap block");
-	err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
-
-	/* And the group descriptor block */
-	BUFFER_TRACE(gd_bh, "dirtied group descriptor block");
-	ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
-	if (!err)
-		err = ret;
-
-error_return:
-	brelse(bitmap_bh);
-	ext4_std_error(sb, err);
-	return;
-}
-
 /**
  * ext4_has_free_blocks()
  * @sbi:	in-core super block structure.
@@ -493,7 +369,8 @@ void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
  * Check if filesystem has nblocks free & available for allocation.
  * On success return 1, return 0 on failure.
  */
-static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
+static int ext4_has_free_blocks(struct ext4_sb_info *sbi,
+				s64 nblocks, unsigned int flags)
 {
 	s64 free_blocks, dirty_blocks, root_blocks;
 	struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
@@ -507,11 +384,6 @@ static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
 						EXT4_FREEBLOCKS_WATERMARK) {
 		free_blocks  = percpu_counter_sum_positive(fbc);
 		dirty_blocks = percpu_counter_sum_positive(dbc);
-		if (dirty_blocks < 0) {
-			printk(KERN_CRIT "Dirty block accounting "
-					"went wrong %lld\n",
-					(long long)dirty_blocks);
-		}
 	}
 	/* Check whether we have space after
 	 * accounting for current dirty blocks & root reserved blocks.
@@ -522,7 +394,9 @@ static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
 	/* Hm, nope.  Are (enough) root reserved blocks available? */
 	if (sbi->s_resuid == current_fsuid() ||
 	    ((sbi->s_resgid != 0) && in_group_p(sbi->s_resgid)) ||
-	    capable(CAP_SYS_RESOURCE)) {
+	    capable(CAP_SYS_RESOURCE) ||
+		(flags & EXT4_MB_USE_ROOT_BLOCKS)) {
+
 		if (free_blocks >= (nblocks + dirty_blocks))
 			return 1;
 	}
@@ -531,9 +405,9 @@ static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
 }
 
 int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
-						s64 nblocks)
+			   s64 nblocks, unsigned int flags)
 {
-	if (ext4_has_free_blocks(sbi, nblocks)) {
+	if (ext4_has_free_blocks(sbi, nblocks, flags)) {
 		percpu_counter_add(&sbi->s_dirtyblocks_counter, nblocks);
 		return 0;
 	} else
@@ -554,7 +428,7 @@ int ext4_claim_free_blocks(struct ext4_sb_info *sbi,
  */
 int ext4_should_retry_alloc(struct super_block *sb, int *retries)
 {
-	if (!ext4_has_free_blocks(EXT4_SB(sb), 1) ||
+	if (!ext4_has_free_blocks(EXT4_SB(sb), 1, 0) ||
 	    (*retries)++ > 3 ||
 	    !EXT4_SB(sb)->s_journal)
 		return 0;
@@ -577,7 +451,8 @@ int ext4_should_retry_alloc(struct super_block *sb, int *retries)
  * error stores in errp pointer
  */
 ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
-		ext4_fsblk_t goal, unsigned long *count, int *errp)
+				  ext4_fsblk_t goal, unsigned int flags,
+				  unsigned long *count, int *errp)
 {
 	struct ext4_allocation_request ar;
 	ext4_fsblk_t ret;
@@ -587,6 +462,7 @@ ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
 	ar.inode = inode;
 	ar.goal = goal;
 	ar.len = count ? *count : 1;
+	ar.flags = flags;
 
 	ret = ext4_mb_new_blocks(handle, &ar, errp);
 	if (count)

fs/ext4/ext4_jbd2.c

@@ -6,20 +6,6 @@
 
 #include <trace/events/ext4.h>
 
-int __ext4_journal_get_undo_access(const char *where, unsigned int line,
-				   handle_t *handle, struct buffer_head *bh)
-{
-	int err = 0;
-
-	if (ext4_handle_valid(handle)) {
-		err = jbd2_journal_get_undo_access(handle, bh);
-		if (err)
-			ext4_journal_abort_handle(where, line, __func__, bh,
-						  handle, err);
-	}
-	return err;
-}
-
 int __ext4_journal_get_write_access(const char *where, unsigned int line,
 				    handle_t *handle, struct buffer_head *bh)
 {

fs/ext4/ext4_jbd2.h

@@ -126,9 +126,6 @@ void ext4_journal_abort_handle(const char *caller, unsigned int line,
 			       const char *err_fn,
 		struct buffer_head *bh, handle_t *handle, int err);
 
-int __ext4_journal_get_undo_access(const char *where, unsigned int line,
-				   handle_t *handle, struct buffer_head *bh);
-
 int __ext4_journal_get_write_access(const char *where, unsigned int line,
 				    handle_t *handle, struct buffer_head *bh);
 
@@ -146,8 +143,6 @@ int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
 int __ext4_handle_dirty_super(const char *where, unsigned int line,
 			      handle_t *handle, struct super_block *sb);
 
-#define ext4_journal_get_undo_access(handle, bh) \
-	__ext4_journal_get_undo_access(__func__, __LINE__, (handle), (bh))
 #define ext4_journal_get_write_access(handle, bh) \
 	__ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \

fs/ext4/file.c

@@ -272,7 +272,6 @@ const struct file_operations ext4_file_operations = {
 };
 
 const struct inode_operations ext4_file_inode_operations = {
-	.truncate	= ext4_truncate,
 	.setattr	= ext4_setattr,
 	.getattr	= ext4_getattr,
 #ifdef CONFIG_EXT4_FS_XATTR

fs/ext4/fsync.c

@@ -36,7 +36,7 @@
 
 static void dump_completed_IO(struct inode * inode)
 {
-#ifdef	EXT4_DEBUG
+#ifdef	EXT4FS_DEBUG
 	struct list_head *cur, *before, *after;
 	ext4_io_end_t *io, *io0, *io1;
 	unsigned long flags;
@@ -172,6 +172,7 @@ int ext4_sync_file(struct file *file, int datasync)
 	journal_t *journal = EXT4_SB(inode->i_sb)->s_journal;
 	int ret;
 	tid_t commit_tid;
+	bool needs_barrier = false;
 
 	J_ASSERT(ext4_journal_current_handle() == NULL);
 
@@ -211,22 +212,12 @@ int ext4_sync_file(struct file *file, int datasync)
 	}
 
 	commit_tid = datasync ? ei->i_datasync_tid : ei->i_sync_tid;
-	if (jbd2_log_start_commit(journal, commit_tid)) {
-		/*
-		 * When the journal is on a different device than the
-		 * fs data disk, we need to issue the barrier in
-		 * writeback mode.  (In ordered mode, the jbd2 layer
-		 * will take care of issuing the barrier.  In
-		 * data=journal, all of the data blocks are written to
-		 * the journal device.)
-		 */
-		if (ext4_should_writeback_data(inode) &&
-		    (journal->j_fs_dev != journal->j_dev) &&
-		    (journal->j_flags & JBD2_BARRIER))
-			blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL,
-					NULL);
-		ret = jbd2_log_wait_commit(journal, commit_tid);
-	} else if (journal->j_flags & JBD2_BARRIER)
+	if (journal->j_flags & JBD2_BARRIER &&
+	    !jbd2_trans_will_send_data_barrier(journal, commit_tid))
+		needs_barrier = true;
+	jbd2_log_start_commit(journal, commit_tid);
+	ret = jbd2_log_wait_commit(journal, commit_tid);
+	if (needs_barrier)
 		blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
  out:
 	trace_ext4_sync_file_exit(inode, ret);

fs/ext4/inode.c

@@ -639,8 +639,8 @@ static int ext4_alloc_blocks(handle_t *handle, struct inode *inode,
 	while (target > 0) {
 		count = target;
 		/* allocating blocks for indirect blocks and direct blocks */
-		current_block = ext4_new_meta_blocks(handle, inode,
-							goal, &count, err);
+		current_block = ext4_new_meta_blocks(handle, inode, goal,
+						     0, &count, err);
 		if (*err)
 			goto failed_out;
 
@@ -1930,7 +1930,7 @@ static int ext4_da_reserve_space(struct inode *inode, ext4_lblk_t lblock)
 	 * We do still charge estimated metadata to the sb though;
 	 * we cannot afford to run out of free blocks.
 	 */
-	if (ext4_claim_free_blocks(sbi, md_needed + 1)) {
+	if (ext4_claim_free_blocks(sbi, md_needed + 1, 0)) {
 		dquot_release_reservation_block(inode, 1);
 		if (ext4_should_retry_alloc(inode->i_sb, &retries)) {
 			yield();
@@ -2796,9 +2796,7 @@ static int write_cache_pages_da(struct address_space *mapping,
 				continue;
 			}
 
-			if (PageWriteback(page))
-				wait_on_page_writeback(page);
-
+			wait_on_page_writeback(page);
 			BUG_ON(PageWriteback(page));
 
 			if (mpd->next_page != page->index)
@@ -3513,7 +3511,7 @@ static ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
 			loff_t end = offset + iov_length(iov, nr_segs);
 
 			if (end > isize)
-				vmtruncate(inode, isize);
+				ext4_truncate_failed_write(inode);
 		}
 	}
 	if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
@@ -3915,10 +3913,31 @@ void ext4_set_aops(struct inode *inode)
  */
 int ext4_block_truncate_page(handle_t *handle,
 		struct address_space *mapping, loff_t from)
+{
+	unsigned offset = from & (PAGE_CACHE_SIZE-1);
+	unsigned length;
+	unsigned blocksize;
+	struct inode *inode = mapping->host;
+
+	blocksize = inode->i_sb->s_blocksize;
+	length = blocksize - (offset & (blocksize - 1));
+
+	return ext4_block_zero_page_range(handle, mapping, from, length);
+}
+
+/*
+ * ext4_block_zero_page_range() zeros out a mapping of length 'length'
+ * starting from file offset 'from'.  The range to be zero'd must
+ * be contained with in one block.  If the specified range exceeds
+ * the end of the block it will be shortened to end of the block
+ * that cooresponds to 'from'
+ */
+int ext4_block_zero_page_range(handle_t *handle,
+		struct address_space *mapping, loff_t from, loff_t length)
 {
 	ext4_fsblk_t index = from >> PAGE_CACHE_SHIFT;
 	unsigned offset = from & (PAGE_CACHE_SIZE-1);
-	unsigned blocksize, length, pos;
+	unsigned blocksize, max, pos;
 	ext4_lblk_t iblock;
 	struct inode *inode = mapping->host;
 	struct buffer_head *bh;
@@ -3931,7 +3950,15 @@ int ext4_block_truncate_page(handle_t *handle,
 		return -EINVAL;
 
 	blocksize = inode->i_sb->s_blocksize;
-	length = blocksize - (offset & (blocksize - 1));
+	max = blocksize - (offset & (blocksize - 1));
+
+	/*
+	 * correct length if it does not fall between
+	 * 'from' and the end of the block
+	 */
+	if (length > max || length < 0)
+		length = max;
+
 	iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);
 
 	if (!page_has_buffers(page))
@@ -4380,8 +4407,6 @@ static void ext4_free_branches(handle_t *handle, struct inode *inode,
 
 int ext4_can_truncate(struct inode *inode)
 {
-	if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
-		return 0;
 	if (S_ISREG(inode->i_mode))
 		return 1;
 	if (S_ISDIR(inode->i_mode))
@@ -4391,6 +4416,31 @@ int ext4_can_truncate(struct inode *inode)
 	return 0;
 }
 
+/*
+ * ext4_punch_hole: punches a hole in a file by releaseing the blocks
+ * associated with the given offset and length
+ *
+ * @inode:  File inode
+ * @offset: The offset where the hole will begin
+ * @len:    The length of the hole
+ *
+ * Returns: 0 on sucess or negative on failure
+ */
+
+int ext4_punch_hole(struct file *file, loff_t offset, loff_t length)
+{
+	struct inode *inode = file->f_path.dentry->d_inode;
+	if (!S_ISREG(inode->i_mode))
+		return -ENOTSUPP;
+
+	if (!ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
+		/* TODO: Add support for non extent hole punching */
+		return -ENOTSUPP;
+	}
+
+	return ext4_ext_punch_hole(file, offset, length);
+}
+
 /*
  * ext4_truncate()
  *
@@ -4617,7 +4667,7 @@ static int __ext4_get_inode_loc(struct inode *inode,
 	/*
 	 * Figure out the offset within the block group inode table
 	 */
-	inodes_per_block = (EXT4_BLOCK_SIZE(sb) / EXT4_INODE_SIZE(sb));
+	inodes_per_block = EXT4_SB(sb)->s_inodes_per_block;
 	inode_offset = ((inode->i_ino - 1) %
 			EXT4_INODES_PER_GROUP(sb));
 	block = ext4_inode_table(sb, gdp) + (inode_offset / inodes_per_block);
@@ -5311,8 +5361,7 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
 
 	if (S_ISREG(inode->i_mode) &&
 	    attr->ia_valid & ATTR_SIZE &&
-	    (attr->ia_size < inode->i_size ||
-	     (ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)))) {
+	    (attr->ia_size < inode->i_size)) {
 		handle_t *handle;
 
 		handle = ext4_journal_start(inode, 3);
@@ -5346,14 +5395,15 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
 				goto err_out;
 			}
 		}
-		/* ext4_truncate will clear the flag */
-		if ((ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS)))
-			ext4_truncate(inode);
 	}
 
-	if ((attr->ia_valid & ATTR_SIZE) &&
-	    attr->ia_size != i_size_read(inode))
-		rc = vmtruncate(inode, attr->ia_size);
+	if (attr->ia_valid & ATTR_SIZE) {
+		if (attr->ia_size != i_size_read(inode)) {
+			truncate_setsize(inode, attr->ia_size);
+			ext4_truncate(inode);
+		} else if (ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))
+			ext4_truncate(inode);
+	}
 
 	if (!rc) {
 		setattr_copy(inode, attr);
@@ -5811,15 +5861,19 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 		goto out_unlock;
 	}
 	ret = 0;
-	if (PageMappedToDisk(page))
-		goto out_unlock;
+
+	lock_page(page);
+	wait_on_page_writeback(page);
+	if (PageMappedToDisk(page)) {
+		up_read(&inode->i_alloc_sem);
+		return VM_FAULT_LOCKED;
+	}
 
 	if (page->index == size >> PAGE_CACHE_SHIFT)
 		len = size & ~PAGE_CACHE_MASK;
 	else
 		len = PAGE_CACHE_SIZE;
 
-	lock_page(page);
 	/*
 	 * return if we have all the buffers mapped. This avoid
 	 * the need to call write_begin/write_end which does a
@@ -5829,8 +5883,8 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	if (page_has_buffers(page)) {
 		if (!walk_page_buffers(NULL, page_buffers(page), 0, len, NULL,
 					ext4_bh_unmapped)) {
-			unlock_page(page);
-			goto out_unlock;
+			up_read(&inode->i_alloc_sem);
+			return VM_FAULT_LOCKED;
 		}
 	}
 	unlock_page(page);
@@ -5850,6 +5904,16 @@ int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
 	if (ret < 0)
 		goto out_unlock;
 	ret = 0;
+
+	/*
+	 * write_begin/end might have created a dirty page and someone
+	 * could wander in and start the IO.  Make sure that hasn't
+	 * happened.
+	 */
+	lock_page(page);
+	wait_on_page_writeback(page);
+	up_read(&inode->i_alloc_sem);
+	return VM_FAULT_LOCKED;
 out_unlock:
 	if (ret)
 		ret = VM_FAULT_SIGBUS;

fs/ext4/mballoc.h

@@ -193,11 +193,6 @@ struct ext4_allocation_context {
 	__u8 ac_op;		/* operation, for history only */
 	struct page *ac_bitmap_page;
 	struct page *ac_buddy_page;
-	/*
-	 * pointer to the held semaphore upon successful
-	 * block allocation
-	 */
-	struct rw_semaphore *alloc_semp;
 	struct ext4_prealloc_space *ac_pa;
 	struct ext4_locality_group *ac_lg;
 };
@@ -215,7 +210,6 @@ struct ext4_buddy {
 	struct super_block *bd_sb;
 	__u16 bd_blkbits;
 	ext4_group_t bd_group;
-	struct rw_semaphore *alloc_semp;
 };
 #define EXT4_MB_BITMAP(e4b)	((e4b)->bd_bitmap)
 #define EXT4_MB_BUDDY(e4b)	((e4b)->bd_buddy)

fs/ext4/migrate.c

@@ -376,7 +376,7 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
 	 * We have the extent map build with the tmp inode.
 	 * Now copy the i_data across
 	 */
-	ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS);
+	ext4_set_inode_flag(inode, EXT4_INODE_EXTENTS);
 	memcpy(ei->i_data, tmp_ei->i_data, sizeof(ei->i_data));
 
 	/*

fs/ext4/mmp.c

+#include <linux/fs.h>
+#include <linux/random.h>
+#include <linux/buffer_head.h>
+#include <linux/utsname.h>
+#include <linux/kthread.h>
+
+#include "ext4.h"
+
+/*
+ * Write the MMP block using WRITE_SYNC to try to get the block on-disk
+ * faster.
+ */
+static int write_mmp_block(struct buffer_head *bh)
+{
+	mark_buffer_dirty(bh);
+	lock_buffer(bh);
+	bh->b_end_io = end_buffer_write_sync;
+	get_bh(bh);
+	submit_bh(WRITE_SYNC, bh);
+	wait_on_buffer(bh);
+	if (unlikely(!buffer_uptodate(bh)))
+		return 1;
+
+	return 0;
+}
+
+/*
+ * Read the MMP block. It _must_ be read from disk and hence we clear the
+ * uptodate flag on the buffer.
+ */
+static int read_mmp_block(struct super_block *sb, struct buffer_head **bh,
+			  ext4_fsblk_t mmp_block)
+{
+	struct mmp_struct *mmp;
+
+	if (*bh)
+		clear_buffer_uptodate(*bh);
+
+	/* This would be sb_bread(sb, mmp_block), except we need to be sure
+	 * that the MD RAID device cache has been bypassed, and that the read
+	 * is not blocked in the elevator. */
+	if (!*bh)
+		*bh = sb_getblk(sb, mmp_block);
+	if (*bh) {
+		get_bh(*bh);
+		lock_buffer(*bh);
+		(*bh)->b_end_io = end_buffer_read_sync;
+		submit_bh(READ_SYNC, *bh);
+		wait_on_buffer(*bh);
+		if (!buffer_uptodate(*bh)) {
+			brelse(*bh);
+			*bh = NULL;
+		}
+	}
+	if (!*bh) {
+		ext4_warning(sb, "Error while reading MMP block %llu",
+			     mmp_block);
+		return -EIO;
+	}
+
+	mmp = (struct mmp_struct *)((*bh)->b_data);
+	if (le32_to_cpu(mmp->mmp_magic) != EXT4_MMP_MAGIC)
+		return -EINVAL;
+
+	return 0;
+}
+
+/*
+ * Dump as much information as possible to help the admin.
+ */
+void __dump_mmp_msg(struct super_block *sb, struct mmp_struct *mmp,
+		    const char *function, unsigned int line, const char *msg)
+{
+	__ext4_warning(sb, function, line, msg);
+	__ext4_warning(sb, function, line,
+		       "MMP failure info: last update time: %llu, last update "
+		       "node: %s, last update device: %s\n",
+		       (long long unsigned int) le64_to_cpu(mmp->mmp_time),
+		       mmp->mmp_nodename, mmp->mmp_bdevname);
+}
+
+/*
+ * kmmpd will update the MMP sequence every s_mmp_update_interval seconds
+ */
+static int kmmpd(void *data)
+{
+	struct super_block *sb = ((struct mmpd_data *) data)->sb;
+	struct buffer_head *bh = ((struct mmpd_data *) data)->bh;
+	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
+	struct mmp_struct *mmp;
+	ext4_fsblk_t mmp_block;
+	u32 seq = 0;
+	unsigned long failed_writes = 0;
+	int mmp_update_interval = le16_to_cpu(es->s_mmp_update_interval);
+	unsigned mmp_check_interval;
+	unsigned long last_update_time;
+	unsigned long diff;
+	int retval;
+
+	mmp_block = le64_to_cpu(es->s_mmp_block);
+	mmp = (struct mmp_struct *)(bh->b_data);
+	mmp->mmp_time = cpu_to_le64(get_seconds());
+	/*
+	 * Start with the higher mmp_check_interval and reduce it if
+	 * the MMP block is being updated on time.
+	 */
+	mmp_check_interval = max(EXT4_MMP_CHECK_MULT * mmp_update_interval,
+				 EXT4_MMP_MIN_CHECK_INTERVAL);
+	mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
+	bdevname(bh->b_bdev, mmp->mmp_bdevname);
+
+	memcpy(mmp->mmp_nodename, init_utsname()->sysname,
+	       sizeof(mmp->mmp_nodename));
+
+	while (!kthread_should_stop()) {
+		if (++seq > EXT4_MMP_SEQ_MAX)
+			seq = 1;
+
+		mmp->mmp_seq = cpu_to_le32(seq);
+		mmp->mmp_time = cpu_to_le64(get_seconds());
+		last_update_time = jiffies;
+
+		retval = write_mmp_block(bh);
+		/*
+		 * Don't spew too many error messages. Print one every
+		 * (s_mmp_update_interval * 60) seconds.
+		 */
+		if (retval && (failed_writes % 60) == 0) {
+			ext4_error(sb, "Error writing to MMP block");
+			failed_writes++;
+		}
+
+		if (!(le32_to_cpu(es->s_feature_incompat) &
+		    EXT4_FEATURE_INCOMPAT_MMP)) {
+			ext4_warning(sb, "kmmpd being stopped since MMP feature"
+				     " has been disabled.");
+			EXT4_SB(sb)->s_mmp_tsk = NULL;
+			goto failed;
+		}
+
+		if (sb->s_flags & MS_RDONLY) {
+			ext4_warning(sb, "kmmpd being stopped since filesystem "
+				     "has been remounted as readonly.");
+			EXT4_SB(sb)->s_mmp_tsk = NULL;
+			goto failed;
+		}
+
+		diff = jiffies - last_update_time;
+		if (diff < mmp_update_interval * HZ)
+			schedule_timeout_interruptible(mmp_update_interval *
+						       HZ - diff);
+
+		/*
+		 * We need to make sure that more than mmp_check_interval
+		 * seconds have not passed since writing. If that has happened
+		 * we need to check if the MMP block is as we left it.
+		 */
+		diff = jiffies - last_update_time;
+		if (diff > mmp_check_interval * HZ) {
+			struct buffer_head *bh_check = NULL;
+			struct mmp_struct *mmp_check;
+
+			retval = read_mmp_block(sb, &bh_check, mmp_block);
+			if (retval) {
+				ext4_error(sb, "error reading MMP data: %d",
+					   retval);
+
+				EXT4_SB(sb)->s_mmp_tsk = NULL;
+				goto failed;
+			}
+
+			mmp_check = (struct mmp_struct *)(bh_check->b_data);
+			if (mmp->mmp_seq != mmp_check->mmp_seq ||
+			    memcmp(mmp->mmp_nodename, mmp_check->mmp_nodename,
+				   sizeof(mmp->mmp_nodename))) {
+				dump_mmp_msg(sb, mmp_check,
+					     "Error while updating MMP info. "
+					     "The filesystem seems to have been"
+					     " multiply mounted.");
+				ext4_error(sb, "abort");
+				goto failed;
+			}
+			put_bh(bh_check);
+		}
+
+		 /*
+		 * Adjust the mmp_check_interval depending on how much time
+		 * it took for the MMP block to be written.
+		 */
+		mmp_check_interval = max(min(EXT4_MMP_CHECK_MULT * diff / HZ,
+					     EXT4_MMP_MAX_CHECK_INTERVAL),
+					 EXT4_MMP_MIN_CHECK_INTERVAL);
+		mmp->mmp_check_interval = cpu_to_le16(mmp_check_interval);
+	}
+
+	/*
+	 * Unmount seems to be clean.
+	 */
+	mmp->mmp_seq = cpu_to_le32(EXT4_MMP_SEQ_CLEAN);
+	mmp->mmp_time = cpu_to_le64(get_seconds());
+
+	retval = write_mmp_block(bh);
+
+failed:
+	kfree(data);
+	brelse(bh);
+	return retval;
+}
+
+/*
+ * Get a random new sequence number but make sure it is not greater than
+ * EXT4_MMP_SEQ_MAX.
+ */
+static unsigned int mmp_new_seq(void)
+{
+	u32 new_seq;
+
+	do {
+		get_random_bytes(&new_seq, sizeof(u32));
+	} while (new_seq > EXT4_MMP_SEQ_MAX);
+
+	return new_seq;
+}
+
+/*
+ * Protect the filesystem from being mounted more than once.
+ */
+int ext4_multi_mount_protect(struct super_block *sb,
+				    ext4_fsblk_t mmp_block)
+{
+	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
+	struct buffer_head *bh = NULL;
+	struct mmp_struct *mmp = NULL;
+	struct mmpd_data *mmpd_data;
+	u32 seq;
+	unsigned int mmp_check_interval = le16_to_cpu(es->s_mmp_update_interval);
+	unsigned int wait_time = 0;
+	int retval;
+
+	if (mmp_block < le32_to_cpu(es->s_first_data_block) ||
+	    mmp_block >= ext4_blocks_count(es)) {
+		ext4_warning(sb, "Invalid MMP block in superblock");
+		goto failed;
+	}
+
+	retval = read_mmp_block(sb, &bh, mmp_block);
+	if (retval)
+		goto failed;
+
+	mmp = (struct mmp_struct *)(bh->b_data);
+
+	if (mmp_check_interval < EXT4_MMP_MIN_CHECK_INTERVAL)
+		mmp_check_interval = EXT4_MMP_MIN_CHECK_INTERVAL;
+
+	/*
+	 * If check_interval in MMP block is larger, use that instead of
+	 * update_interval from the superblock.
+	 */
+	if (mmp->mmp_check_interval > mmp_check_interval)
+		mmp_check_interval = mmp->mmp_check_interval;
+
+	seq = le32_to_cpu(mmp->mmp_seq);
+	if (seq == EXT4_MMP_SEQ_CLEAN)
+		goto skip;
+
+	if (seq == EXT4_MMP_SEQ_FSCK) {
+		dump_mmp_msg(sb, mmp, "fsck is running on the filesystem");
+		goto failed;
+	}
+
+	wait_time = min(mmp_check_interval * 2 + 1,
+			mmp_check_interval + 60);
+
+	/* Print MMP interval if more than 20 secs. */
+	if (wait_time > EXT4_MMP_MIN_CHECK_INTERVAL * 4)
+		ext4_warning(sb, "MMP interval %u higher than expected, please"
+			     " wait.\n", wait_time * 2);
+
+	if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
+		ext4_warning(sb, "MMP startup interrupted, failing mount\n");
+		goto failed;
+	}
+
+	retval = read_mmp_block(sb, &bh, mmp_block);
+	if (retval)
+		goto failed;
+	mmp = (struct mmp_struct *)(bh->b_data);
+	if (seq != le32_to_cpu(mmp->mmp_seq)) {
+		dump_mmp_msg(sb, mmp,
+			     "Device is already active on another node.");
+		goto failed;
+	}
+
+skip:
+	/*
+	 * write a new random sequence number.
+	 */
+	mmp->mmp_seq = seq = cpu_to_le32(mmp_new_seq());
+
+	retval = write_mmp_block(bh);
+	if (retval)
+		goto failed;
+
+	/*
+	 * wait for MMP interval and check mmp_seq.
+	 */
+	if (schedule_timeout_interruptible(HZ * wait_time) != 0) {
+		ext4_warning(sb, "MMP startup interrupted, failing mount\n");
+		goto failed;
+	}
+
+	retval = read_mmp_block(sb, &bh, mmp_block);
+	if (retval)
+		goto failed;
+	mmp = (struct mmp_struct *)(bh->b_data);
+	if (seq != le32_to_cpu(mmp->mmp_seq)) {
+		dump_mmp_msg(sb, mmp,
+			     "Device is already active on another node.");
+		goto failed;
+	}
+
+	mmpd_data = kmalloc(sizeof(struct mmpd_data), GFP_KERNEL);
+	if (!mmpd_data) {
+		ext4_warning(sb, "not enough memory for mmpd_data");
+		goto failed;
+	}
+	mmpd_data->sb = sb;
+	mmpd_data->bh = bh;
+
+	/*
+	 * Start a kernel thread to update the MMP block periodically.
+	 */
+	EXT4_SB(sb)->s_mmp_tsk = kthread_run(kmmpd, mmpd_data, "kmmpd-%s",
+					     bdevname(bh->b_bdev,
+						      mmp->mmp_bdevname));
+	if (IS_ERR(EXT4_SB(sb)->s_mmp_tsk)) {
+		EXT4_SB(sb)->s_mmp_tsk = NULL;
+		kfree(mmpd_data);
+		ext4_warning(sb, "Unable to create kmmpd thread for %s.",
+			     sb->s_id);
+		goto failed;
+	}
+
+	return 0;
+
+failed:
+	brelse(bh);
+	return 1;
+}
+
+

fs/ext4/move_extent.c

@@ -876,8 +876,7 @@ move_extent_per_page(struct file *o_filp, struct inode *donor_inode,
 	 * It needs to call wait_on_page_writeback() to wait for the
 	 * writeback of the page.
 	 */
-	if (PageWriteback(page))
-		wait_on_page_writeback(page);
+	wait_on_page_writeback(page);
 
 	/* Release old bh and drop refs */
 	try_to_release_page(page, 0);

fs/ext4/namei.c

@@ -1413,10 +1413,22 @@ static int make_indexed_dir(handle_t *handle, struct dentry *dentry,
 	frame->at = entries;
 	frame->bh = bh;
 	bh = bh2;
+
+	ext4_handle_dirty_metadata(handle, dir, frame->bh);
+	ext4_handle_dirty_metadata(handle, dir, bh);
+
 	de = do_split(handle,dir, &bh, frame, &hinfo, &retval);
-	dx_release (frames);
-	if (!(de))
+	if (!de) {
+		/*
+		 * Even if the block split failed, we have to properly write
+		 * out all the changes we did so far. Otherwise we can end up
+		 * with corrupted filesystem.
+		 */
+		ext4_mark_inode_dirty(handle, dir);
+		dx_release(frames);
 		return retval;
+	}
+	dx_release(frames);
 
 	retval = add_dirent_to_buf(handle, dentry, inode, de, bh);
 	brelse(bh);
@@ -2240,6 +2252,7 @@ static int ext4_symlink(struct inode *dir,
 	handle_t *handle;
 	struct inode *inode;
 	int l, err, retries = 0;
+	int credits;
 
 	l = strlen(symname)+1;
 	if (l > dir->i_sb->s_blocksize)
@@ -2247,10 +2260,26 @@ static int ext4_symlink(struct inode *dir,
 
 	dquot_initialize(dir);
 
+	if (l > EXT4_N_BLOCKS * 4) {
+		/*
+		 * For non-fast symlinks, we just allocate inode and put it on
+		 * orphan list in the first transaction => we need bitmap,
+		 * group descriptor, sb, inode block, quota blocks.
+		 */
+		credits = 4 + EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb);
+	} else {
+		/*
+		 * Fast symlink. We have to add entry to directory
+		 * (EXT4_DATA_TRANS_BLOCKS + EXT4_INDEX_EXTRA_TRANS_BLOCKS),
+		 * allocate new inode (bitmap, group descriptor, inode block,
+		 * quota blocks, sb is already counted in previous macros).
+		 */
+		credits = EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
+			  EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 +
+			  EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb);
+	}
 retry:
-	handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
-					EXT4_INDEX_EXTRA_TRANS_BLOCKS + 5 +
-					EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb));
+	handle = ext4_journal_start(dir, credits);
 	if (IS_ERR(handle))
 		return PTR_ERR(handle);
 
@@ -2263,21 +2292,44 @@ static int ext4_symlink(struct inode *dir,
 	if (IS_ERR(inode))
 		goto out_stop;
 
-	if (l > sizeof(EXT4_I(inode)->i_data)) {
+	if (l > EXT4_N_BLOCKS * 4) {
 		inode->i_op = &ext4_symlink_inode_operations;
 		ext4_set_aops(inode);
 		/*
-		 * page_symlink() calls into ext4_prepare/commit_write.
-		 * We have a transaction open.  All is sweetness.  It also sets
-		 * i_size in generic_commit_write().
+		 * We cannot call page_symlink() with transaction started
+		 * because it calls into ext4_write_begin() which can wait
+		 * for transaction commit if we are running out of space
+		 * and thus we deadlock. So we have to stop transaction now
+		 * and restart it when symlink contents is written.
+		 * 
+		 * To keep fs consistent in case of crash, we have to put inode
+		 * to orphan list in the mean time.
 		 */
+		drop_nlink(inode);
+		err = ext4_orphan_add(handle, inode);
+		ext4_journal_stop(handle);
+		if (err)
+			goto err_drop_inode;
 		err = __page_symlink(inode, symname, l, 1);
+		if (err)
+			goto err_drop_inode;
+		/*
+		 * Now inode is being linked into dir (EXT4_DATA_TRANS_BLOCKS
+		 * + EXT4_INDEX_EXTRA_TRANS_BLOCKS), inode is also modified
+		 */
+		handle = ext4_journal_start(dir,
+				EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
+				EXT4_INDEX_EXTRA_TRANS_BLOCKS + 1);
+		if (IS_ERR(handle)) {
+			err = PTR_ERR(handle);
+			goto err_drop_inode;
+		}
+		inc_nlink(inode);
+		err = ext4_orphan_del(handle, inode);
 		if (err) {
+			ext4_journal_stop(handle);
 			clear_nlink(inode);
-			unlock_new_inode(inode);
-			ext4_mark_inode_dirty(handle, inode);
-			iput(inode);
-			goto out_stop;
+			goto err_drop_inode;
 		}
 	} else {
 		/* clear the extent format for fast symlink */
@@ -2293,6 +2345,10 @@ static int ext4_symlink(struct inode *dir,
 	if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
 		goto retry;
 	return err;
+err_drop_inode:
+	unlock_new_inode(inode);
+	iput(inode);
+	return err;
 }
 
 static int ext4_link(struct dentry *old_dentry,

fs/ext4/page-io.c

@@ -203,46 +203,29 @@ static void ext4_end_bio(struct bio *bio, int error)
 	for (i = 0; i < io_end->num_io_pages; i++) {
 		struct page *page = io_end->pages[i]->p_page;
 		struct buffer_head *bh, *head;
-		int partial_write = 0;
+		loff_t offset;
+		loff_t io_end_offset;
 
-		head = page_buffers(page);
-		if (error)
+		if (error) {
 			SetPageError(page);
-		BUG_ON(!head);
-		if (head->b_size != PAGE_CACHE_SIZE) {
-			loff_t offset;
-			loff_t io_end_offset = io_end->offset + io_end->size;
+			set_bit(AS_EIO, &page->mapping->flags);
+			head = page_buffers(page);
+			BUG_ON(!head);
+
+			io_end_offset = io_end->offset + io_end->size;
 
 			offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
 			bh = head;
 			do {
 				if ((offset >= io_end->offset) &&
-				    (offset+bh->b_size <= io_end_offset)) {
-					if (error)
-						buffer_io_error(bh);
-
-				}
-				if (buffer_delay(bh))
-					partial_write = 1;
-				else if (!buffer_mapped(bh))
-					clear_buffer_dirty(bh);
-				else if (buffer_dirty(bh))
-					partial_write = 1;
+				    (offset+bh->b_size <= io_end_offset))
+					buffer_io_error(bh);
+
 				offset += bh->b_size;
 				bh = bh->b_this_page;
 			} while (bh != head);
 		}
 
-		/*
-		 * If this is a partial write which happened to make
-		 * all buffers uptodate then we can optimize away a
-		 * bogus readpage() for the next read(). Here we
-		 * 'discover' whether the page went uptodate as a
-		 * result of this (potentially partial) write.
-		 */
-		if (!partial_write)
-			SetPageUptodate(page);
-
 		put_io_page(io_end->pages[i]);
 	}
 	io_end->num_io_pages = 0;

fs/ext4/xattr.c

@@ -820,8 +820,8 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
 			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
 				goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
 
-			block = ext4_new_meta_blocks(handle, inode,
-						  goal, NULL, &error);
+			block = ext4_new_meta_blocks(handle, inode, goal, 0,
+						     NULL, &error);
 			if (error)
 				goto cleanup;
 

fs/jbd2/commit.c

@@ -219,7 +219,6 @@ static int journal_submit_data_buffers(journal_t *journal,
 			ret = err;
 		spin_lock(&journal->j_list_lock);
 		J_ASSERT(jinode->i_transaction == commit_transaction);
-		commit_transaction->t_flushed_data_blocks = 1;
 		clear_bit(__JI_COMMIT_RUNNING, &jinode->i_flags);
 		smp_mb__after_clear_bit();
 		wake_up_bit(&jinode->i_flags, __JI_COMMIT_RUNNING);
@@ -672,12 +671,16 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 		err = 0;
 	}
 
+	write_lock(&journal->j_state_lock);
+	J_ASSERT(commit_transaction->t_state == T_COMMIT);
+	commit_transaction->t_state = T_COMMIT_DFLUSH;
+	write_unlock(&journal->j_state_lock);
 	/* 
 	 * If the journal is not located on the file system device,
 	 * then we must flush the file system device before we issue
 	 * the commit record
 	 */
-	if (commit_transaction->t_flushed_data_blocks &&
+	if (commit_transaction->t_need_data_flush &&
 	    (journal->j_fs_dev != journal->j_dev) &&
 	    (journal->j_flags & JBD2_BARRIER))
 		blkdev_issue_flush(journal->j_fs_dev, GFP_KERNEL, NULL);
@@ -754,8 +757,13 @@ void jbd2_journal_commit_transaction(journal_t *journal)
                    required. */
 		JBUFFER_TRACE(jh, "file as BJ_Forget");
 		jbd2_journal_file_buffer(jh, commit_transaction, BJ_Forget);
-		/* Wake up any transactions which were waiting for this
-		   IO to complete */
+		/*
+		 * Wake up any transactions which were waiting for this IO to
+		 * complete. The barrier must be here so that changes by
+		 * jbd2_journal_file_buffer() take effect before wake_up_bit()
+		 * does the waitqueue check.
+		 */
+		smp_mb();
 		wake_up_bit(&bh->b_state, BH_Unshadow);
 		JBUFFER_TRACE(jh, "brelse shadowed buffer");
 		__brelse(bh);
@@ -794,6 +802,10 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 		jbd2_journal_abort(journal, err);
 
 	jbd_debug(3, "JBD: commit phase 5\n");
+	write_lock(&journal->j_state_lock);
+	J_ASSERT(commit_transaction->t_state == T_COMMIT_DFLUSH);
+	commit_transaction->t_state = T_COMMIT_JFLUSH;
+	write_unlock(&journal->j_state_lock);
 
 	if (!JBD2_HAS_INCOMPAT_FEATURE(journal,
 				       JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)) {
@@ -949,7 +961,7 @@ void jbd2_journal_commit_transaction(journal_t *journal)
 
 	jbd_debug(3, "JBD: commit phase 7\n");
 
-	J_ASSERT(commit_transaction->t_state == T_COMMIT);
+	J_ASSERT(commit_transaction->t_state == T_COMMIT_JFLUSH);
 
 	commit_transaction->t_start = jiffies;
 	stats.run.rs_logging = jbd2_time_diff(stats.run.rs_logging,

fs/jbd2/journal.c

@@ -479,9 +479,12 @@ int __jbd2_log_space_left(journal_t *journal)
 int __jbd2_log_start_commit(journal_t *journal, tid_t target)
 {
 	/*
-	 * Are we already doing a recent enough commit?
+	 * The only transaction we can possibly wait upon is the
+	 * currently running transaction (if it exists).  Otherwise,
+	 * the target tid must be an old one.
 	 */
-	if (!tid_geq(journal->j_commit_request, target)) {
+	if (journal->j_running_transaction &&
+	    journal->j_running_transaction->t_tid == target) {
 		/*
 		 * We want a new commit: OK, mark the request and wakeup the
 		 * commit thread.  We do _not_ do the commit ourselves.
@@ -493,7 +496,15 @@ int __jbd2_log_start_commit(journal_t *journal, tid_t target)
 			  journal->j_commit_sequence);
 		wake_up(&journal->j_wait_commit);
 		return 1;
-	}
+	} else if (!tid_geq(journal->j_commit_request, target))
+		/* This should never happen, but if it does, preserve
+		   the evidence before kjournald goes into a loop and
+		   increments j_commit_sequence beyond all recognition. */
+		WARN_ONCE(1, "jbd: bad log_start_commit: %u %u %u %u\n",
+			  journal->j_commit_request,
+			  journal->j_commit_sequence,
+			  target, journal->j_running_transaction ? 
+			  journal->j_running_transaction->t_tid : 0);
 	return 0;
 }
 
@@ -576,6 +587,47 @@ int jbd2_journal_start_commit(journal_t *journal, tid_t *ptid)
 	return ret;
 }
 
+/*
+ * Return 1 if a given transaction has not yet sent barrier request
+ * connected with a transaction commit. If 0 is returned, transaction
+ * may or may not have sent the barrier. Used to avoid sending barrier
+ * twice in common cases.
+ */
+int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid)
+{
+	int ret = 0;
+	transaction_t *commit_trans;
+
+	if (!(journal->j_flags & JBD2_BARRIER))
+		return 0;
+	read_lock(&journal->j_state_lock);
+	/* Transaction already committed? */
+	if (tid_geq(journal->j_commit_sequence, tid))
+		goto out;
+	commit_trans = journal->j_committing_transaction;
+	if (!commit_trans || commit_trans->t_tid != tid) {
+		ret = 1;
+		goto out;
+	}
+	/*
+	 * Transaction is being committed and we already proceeded to
+	 * submitting a flush to fs partition?
+	 */
+	if (journal->j_fs_dev != journal->j_dev) {
+		if (!commit_trans->t_need_data_flush ||
+		    commit_trans->t_state >= T_COMMIT_DFLUSH)
+			goto out;
+	} else {
+		if (commit_trans->t_state >= T_COMMIT_JFLUSH)
+			goto out;
+	}
+	ret = 1;
+out:
+	read_unlock(&journal->j_state_lock);
+	return ret;
+}
+EXPORT_SYMBOL(jbd2_trans_will_send_data_barrier);
+
 /*
  * Wait for a specified commit to complete.
  * The caller may not hold the journal lock.

fs/jbd2/transaction.c

@@ -82,7 +82,7 @@ jbd2_get_transaction(journal_t *journal, transaction_t *transaction)
  */
 
 /*
- * Update transiaction's maximum wait time, if debugging is enabled.
+ * Update transaction's maximum wait time, if debugging is enabled.
  *
  * In order for t_max_wait to be reliable, it must be protected by a
  * lock.  But doing so will mean that start_this_handle() can not be
@@ -91,11 +91,10 @@ jbd2_get_transaction(journal_t *journal, transaction_t *transaction)
  * means that maximum wait time reported by the jbd2_run_stats
  * tracepoint will always be zero.
  */
-static inline void update_t_max_wait(transaction_t *transaction)
+static inline void update_t_max_wait(transaction_t *transaction,
+				     unsigned long ts)
 {
 #ifdef CONFIG_JBD2_DEBUG
-	unsigned long ts = jiffies;
-
 	if (jbd2_journal_enable_debug &&
 	    time_after(transaction->t_start, ts)) {
 		ts = jbd2_time_diff(ts, transaction->t_start);
@@ -121,6 +120,7 @@ static int start_this_handle(journal_t *journal, handle_t *handle,
 	tid_t		tid;
 	int		needed, need_to_start;
 	int		nblocks = handle->h_buffer_credits;
+	unsigned long ts = jiffies;
 
 	if (nblocks > journal->j_max_transaction_buffers) {
 		printk(KERN_ERR "JBD: %s wants too many credits (%d > %d)\n",
@@ -271,7 +271,7 @@ static int start_this_handle(journal_t *journal, handle_t *handle,
 	/* OK, account for the buffers that this operation expects to
 	 * use and add the handle to the running transaction. 
 	 */
-	update_t_max_wait(transaction);
+	update_t_max_wait(transaction, ts);
 	handle->h_transaction = transaction;
 	atomic_inc(&transaction->t_updates);
 	atomic_inc(&transaction->t_handle_count);
@@ -316,7 +316,8 @@ static handle_t *new_handle(int nblocks)
  * This function is visible to journal users (like ext3fs), so is not
  * called with the journal already locked.
  *
- * Return a pointer to a newly allocated handle, or NULL on failure
+ * Return a pointer to a newly allocated handle, or an ERR_PTR() value
+ * on failure.
  */
 handle_t *jbd2__journal_start(journal_t *journal, int nblocks, int gfp_mask)
 {
@@ -921,8 +922,8 @@ int jbd2_journal_get_create_access(handle_t *handle, struct buffer_head *bh)
 	 */
 	JBUFFER_TRACE(jh, "cancelling revoke");
 	jbd2_journal_cancel_revoke(handle, jh);
-	jbd2_journal_put_journal_head(jh);
 out:
+	jbd2_journal_put_journal_head(jh);
 	return err;
 }
 
@@ -2147,6 +2148,13 @@ int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *jinode)
 	    jinode->i_next_transaction == transaction)
 		goto done;
 
+	/*
+	 * We only ever set this variable to 1 so the test is safe. Since
+	 * t_need_data_flush is likely to be set, we do the test to save some
+	 * cacheline bouncing
+	 */
+	if (!transaction->t_need_data_flush)
+		transaction->t_need_data_flush = 1;
 	/* On some different transaction's list - should be
 	 * the committing one */
 	if (jinode->i_transaction) {

include/linux/jbd2.h

@@ -529,9 +529,10 @@ struct transaction_s
 	enum {
 		T_RUNNING,
 		T_LOCKED,
-		T_RUNDOWN,
 		T_FLUSH,
 		T_COMMIT,
+		T_COMMIT_DFLUSH,
+		T_COMMIT_JFLUSH,
 		T_FINISHED
 	}			t_state;
 
@@ -658,7 +659,9 @@ struct transaction_s
 	 * waiting for it to finish.
 	 */
 	unsigned int t_synchronous_commit:1;
-	unsigned int t_flushed_data_blocks:1;
+
+	/* Disk flush needs to be sent to fs partition [no locking] */
+	int			t_need_data_flush;
 
 	/*
 	 * For use by the filesystem to store fs-specific data
@@ -1228,6 +1231,7 @@ int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
 int jbd2_journal_force_commit_nested(journal_t *journal);
 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
 int jbd2_log_do_checkpoint(journal_t *journal);
+int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
 
 void __jbd2_log_wait_for_space(journal_t *journal);
 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);