[PATCH] ext4: Fix file fragmentation during large file write.

[PATCH] ext4: Fix file fragmentation during large file write.

[Aneesh Kumar K.V]

The range_cyclic writeback mode use the address_space
writeback_index as the start index for writeback. With
delayed allocation we were updating writeback_index
wrongly resulting in highly fragmented file. Number of
extents reduced from 4000 to 27 for a 3GB file with
the below patch.

The patch also cleanup the ext4 delayed allocation writepages
by implementing write_cache_pages locally with needed changes
for cleanup. Also it drops the range_cont writeback mode added
for ext4 delayed allocation writeback

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/ext4/inode.c           |  192 +++++++++++++++++++++++++++++++++------------
 include/linux/writeback.h |    1 -
 mm/page-writeback.c       |    2 -
 3 files changed, 143 insertions(+), 52 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 21f1d3a..b6b0985 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1648,6 +1648,7 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
 	int ret = 0, err, nr_pages, i;
 	unsigned long index, end;
 	struct pagevec pvec;
+	long pages_skipped;
 
 	BUG_ON(mpd->next_page <= mpd->first_page);
 	pagevec_init(&pvec, 0);
@@ -1655,20 +1656,30 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
 	end = mpd->next_page - 1;
 
 	while (index <= end) {
-		/* XXX: optimize tail */
-		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
+		/*
+		 * We can use PAGECACHE_TAG_DIRTY lookup here because
+		 * even though we have cleared the dirty flag on the page
+		 * We still keep the page in the radix tree with tag
+		 * PAGECACHE_TAG_DIRTY. See clear_page_dirty_for_io.
+		 * The PAGECACHE_TAG_DIRTY is cleared in set_page_writeback
+		 * which is called via the below writepage callback.
+		 */
+		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
+					PAGECACHE_TAG_DIRTY,
+					min(end - index,
+					(pgoff_t)PAGEVEC_SIZE-1) + 1);
 		if (nr_pages == 0)
 			break;
 		for (i = 0; i < nr_pages; i++) {
 			struct page *page = pvec.pages[i];
 
-			index = page->index;
-			if (index > end)
-				break;
-			index++;
-
+			pages_skipped = mpd->wbc->pages_skipped;
 			err = mapping->a_ops->writepage(page, mpd->wbc);
-			if (!err)
+			if (!err && (pages_skipped == mpd->wbc->pages_skipped))
+				/*
+				 * have successfully written the page
+				 * without skipping the same
+				 */
 				mpd->pages_written++;
 			/*
 			 * In error case, we have to continue because
@@ -2088,6 +2099,100 @@ static int __mpage_da_writepage(struct page *page,
 	return 0;
 }
 
+static int ext4_write_cache_pages(struct address_space *mapping,
+		      struct writeback_control *wbc, writepage_t writepage,
+		      void *data)
+{
+	struct pagevec pvec;
+	pgoff_t index, end;
+	long to_write = wbc->nr_to_write;
+	int ret = 0, done = 0, scanned = 0, nr_pages;
+	struct backing_dev_info *bdi = mapping->backing_dev_info;
+
+	if (wbc->nonblocking && bdi_write_congested(bdi)) {
+		wbc->encountered_congestion = 1;
+		return 0;
+	}
+
+	pagevec_init(&pvec, 0);
+	if (wbc->range_cyclic) {
+		index = mapping->writeback_index; /* Start from prev offset */
+		end = -1;
+	} else {
+		index = wbc->range_start >> PAGE_CACHE_SHIFT;
+		end = wbc->range_end >> PAGE_CACHE_SHIFT;
+		scanned = 1;
+	}
+retry:
+	while (!done && (index <= end) &&
+	       (nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
+					      PAGECACHE_TAG_DIRTY,
+					      min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1))) {
+		unsigned i;
+
+		scanned = 1;
+		for (i = 0; i < nr_pages; i++) {
+			struct page *page = pvec.pages[i];
+
+			/*
+			 * At this point we hold neither mapping->tree_lock nor
+			 * lock on the page itself: the page may be truncated or
+			 * invalidated (changing page->mapping to NULL), or even
+			 * swizzled back from swapper_space to tmpfs file
+			 * mapping
+			 */
+			lock_page(page);
+
+			if (unlikely(page->mapping != mapping)) {
+				unlock_page(page);
+				continue;
+			}
+			if (!wbc->range_cyclic && page->index > end) {
+				done = 1;
+				unlock_page(page);
+				continue;
+			}
+			if (wbc->sync_mode != WB_SYNC_NONE)
+				wait_on_page_writeback(page);
+
+			if (PageWriteback(page) ||
+			    !clear_page_dirty_for_io(page)) {
+				unlock_page(page);
+				continue;
+			}
+			ret = (*writepage)(page, wbc, data);
+			if (unlikely(ret == AOP_WRITEPAGE_ACTIVATE)) {
+				unlock_page(page);
+				ret = 0;
+			}
+			if (ret || (--(to_write) <= 0))
+				/*
+				 * writepage either failed.
+				 * or did an extent write.
+				 * We wrote what we are asked to
+				 * write
+				 */
+				done = 1;
+			if (wbc->nonblocking && bdi_write_congested(bdi)) {
+				wbc->encountered_congestion = 1;
+				done = 1;
+			}
+		}
+		pagevec_release(&pvec);
+		cond_resched();
+	}
+	if (!scanned && !done) {
+		/*
+		 * We hit the last page and there is more work to be done: wrap
+		 * back to the start of the file
+		 */
+		scanned = 1;
+		index = 0;
+		goto retry;
+	}
+	return ret;
+}
+
 /*
  * mpage_da_writepages - walk the list of dirty pages of the given
  * address space, allocates non-allocated blocks, maps newly-allocated
@@ -2104,7 +2209,6 @@ static int mpage_da_writepages(struct address_space *mapping,
 			       struct writeback_control *wbc,
 			       struct mpage_da_data *mpd)
 {
-	long to_write;
 	int ret;
 
 	if (!mpd->get_block)
@@ -2119,10 +2223,7 @@ static int mpage_da_writepages(struct address_space *mapping,
 	mpd->pages_written = 0;
 	mpd->retval = 0;
 
-	to_write = wbc->nr_to_write;
-
-	ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, mpd);
-
+	ret = ext4_write_cache_pages(mapping, wbc, __mpage_da_writepage, mpd);
 	/*
 	 * Handle last extent of pages
 	 */
@@ -2131,7 +2232,7 @@ static int mpage_da_writepages(struct address_space *mapping,
 			mpage_da_submit_io(mpd);
 	}
 
-	wbc->nr_to_write = to_write - mpd->pages_written;
+	wbc->nr_to_write -= mpd->pages_written;
 	return ret;
 }
 
@@ -2360,12 +2461,13 @@ static int ext4_da_writepages_trans_blocks(struct inode *inode)
 static int ext4_da_writepages(struct address_space *mapping,
 			      struct writeback_control *wbc)
 {
+	pgoff_t	index;
+	int range_whole = 0;
 	handle_t *handle = NULL;
-	loff_t range_start = 0;
+	long pages_written = 0;
 	struct mpage_da_data mpd;
 	struct inode *inode = mapping->host;
 	int needed_blocks, ret = 0, nr_to_writebump = 0;
-	long to_write, pages_skipped = 0;
 	struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
 
 	/*
@@ -2385,23 +2487,18 @@ static int ext4_da_writepages(struct address_space *mapping,
 		nr_to_writebump = sbi->s_mb_stream_request - wbc->nr_to_write;
 		wbc->nr_to_write = sbi->s_mb_stream_request;
 	}
+	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
+		range_whole = 1;
 
-	if (!wbc->range_cyclic)
-		/*
-		 * If range_cyclic is not set force range_cont
-		 * and save the old writeback_index
-		 */
-		wbc->range_cont = 1;
-
-	range_start =  wbc->range_start;
-	pages_skipped = wbc->pages_skipped;
+	if (wbc->range_cyclic)
+		index = mapping->writeback_index;
+	else
+		index = wbc->range_start >> PAGE_CACHE_SHIFT;
 
 	mpd.wbc = wbc;
 	mpd.inode = mapping->host;
 
-restart_loop:
-	to_write = wbc->nr_to_write;
-	while (!ret && to_write > 0) {
+	while (!ret && wbc->nr_to_write > 0) {
 
 		/*
 		 * we  insert one extent at a time. So we need
@@ -2422,48 +2519,45 @@ static int ext4_da_writepages(struct address_space *mapping,
 			dump_stack();
 			goto out_writepages;
 		}
-		to_write -= wbc->nr_to_write;
-
 		mpd.get_block = ext4_da_get_block_write;
 		ret = mpage_da_writepages(mapping, wbc, &mpd);
 
 		ext4_journal_stop(handle);
 
-		if (mpd.retval == -ENOSPC)
+		if (mpd.retval == -ENOSPC) {
+			/* commit the transaction which would
+			 * free blocks released in the transaction
+			 * and try again
+			 */
 			jbd2_journal_force_commit_nested(sbi->s_journal);
-
-		/* reset the retry count */
-		if (ret == MPAGE_DA_EXTENT_TAIL) {
+			ret = 0;
+		} else if (ret == MPAGE_DA_EXTENT_TAIL) {
 			/*
 			 * got one extent now try with
 			 * rest of the pages
 			 */
-			to_write += wbc->nr_to_write;
+			pages_written += mpd.pages_written;
 			ret = 0;
-		} else if (wbc->nr_to_write) {
+		} else if (wbc->nr_to_write)
 			/*
 			 * There is no more writeout needed
 			 * or we requested for a noblocking writeout
 			 * and we found the device congested
 			 */
-			to_write += wbc->nr_to_write;
 			break;
-		}
-		wbc->nr_to_write = to_write;
 	}
 
-	if (wbc->range_cont && (pages_skipped != wbc->pages_skipped)) {
-		/* We skipped pages in this loop */
-		wbc->range_start = range_start;
-		wbc->nr_to_write = to_write +
-				wbc->pages_skipped - pages_skipped;
-		wbc->pages_skipped = pages_skipped;
-		goto restart_loop;
-	}
+	/* Update index */
+	index += pages_written;
+	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
+		/*
+		 * set the writeback_index so that range_cyclic
+		 * mode will write it back later
+		 */
+		mapping->writeback_index = index;
 
 out_writepages:
-	wbc->nr_to_write = to_write - nr_to_writebump;
-	wbc->range_start = range_start;
+	wbc->nr_to_write -= nr_to_writebump;
 	return ret;
 }
 
diff --git a/include/linux/writeback.h b/include/linux/writeback.h
index 12b15c5..bd91987 100644
--- a/include/linux/writeback.h
+++ b/include/linux/writeback.h
@@ -63,7 +63,6 @@ struct writeback_control {
 	unsigned for_writepages:1;	/* This is a writepages() call */
 	unsigned range_cyclic:1;	/* range_start is cyclic */
 	unsigned more_io:1;		/* more io to be dispatched */
-	unsigned range_cont:1;
 };
 
 /*
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 24de8b6..718efa6 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -961,8 +961,6 @@ int write_cache_pages(struct address_space *mapping,
 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
 		mapping->writeback_index = index;
 
-	if (wbc->range_cont)
-		wbc->range_start = index << PAGE_CACHE_SHIFT;
 	return ret;
 }
 EXPORT_SYMBOL(write_cache_pages);
-- 
1.6.0.1.285.g1070

Patches for patchqueue

[Aneesh Kumar K.V]

Hi Ted,

Below patches are updates for patches in the patch queue.
0001-ext4-Use-an-rbtree-for-tracking-blocks-freed-during.patch
Updated with the fix from you. I also removed the definition of
EXT4_BB_MAX_BLOCKS

0004-ext4-Use-tag-dirty-lookup-during-mpage_da_submit_io.patch
0005-vfs-Remove-the-range_cont-writeback-mode.patch
0006-vfs-Add-no_nrwrite_update-and-no_index_update-write.patch
0007-ext4-Fix-file-fragmentation-during-large-file-write.patch

Split the single patch fix_file_fragmentation_during_large_file_write
into 4 patches. The VFS changes are cc to linux-fsdevel@vger.kernel.org

0010-ext4-Free-ext4_prealloc_space-using-kmem_cache_free.patch

-aneesh

[PATCH -V5] ext4: Use an rbtree for tracking blocks freed during transaction.

[Aneesh Kumar K.V]

With this patch we track the block freed during a transaction using
rb tree. We also make sure contiguous blocks freed are collected
in one rb node.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
---
 fs/ext4/mballoc.c |  184 +++++++++++++++++++++++++++++++++-------------------
 fs/ext4/mballoc.h |   26 +++++---
 2 files changed, 133 insertions(+), 77 deletions(-)

diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index b580714..7023228 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -2300,6 +2300,7 @@ int ext4_mb_add_groupinfo(struct super_block *sb, ext4_group_t group,
 	}
 
 	INIT_LIST_HEAD(&meta_group_info[i]->bb_prealloc_list);
+	meta_group_info[i]->bb_free_root.rb_node = NULL;;
 
 #ifdef DOUBLE_CHECK
 	{
@@ -2647,13 +2648,11 @@ int ext4_mb_release(struct super_block *sb)
 static noinline_for_stack void
 ext4_mb_free_committed_blocks(struct super_block *sb)
 {
-	struct ext4_sb_info *sbi = EXT4_SB(sb);
-	int err;
-	int i;
-	int count = 0;
-	int count2 = 0;
-	struct ext4_free_metadata *md;
 	struct ext4_buddy e4b;
+	struct ext4_group_info *db;
+	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	int err, count = 0, count2 = 0;
+	struct ext4_free_data *entry;
 
 	if (list_empty(&sbi->s_committed_transaction))
 		return;
@@ -2661,44 +2660,46 @@ ext4_mb_free_committed_blocks(struct super_block *sb)
 	/* there is committed blocks to be freed yet */
 	do {
 		/* get next array of blocks */
-		md = NULL;
+		entry = NULL;
 		spin_lock(&sbi->s_md_lock);
 		if (!list_empty(&sbi->s_committed_transaction)) {
-			md = list_entry(sbi->s_committed_transaction.next,
-					struct ext4_free_metadata, list);
-			list_del(&md->list);
+			entry = list_entry(sbi->s_committed_transaction.next,
+					struct ext4_free_data, list);
+			list_del(&entry->list);
 		}
 		spin_unlock(&sbi->s_md_lock);
 
-		if (md == NULL)
+		if (entry == NULL)
 			break;
 
 		mb_debug("gonna free %u blocks in group %lu (0x%p):",
-				md->num, md->group, md);
+				entry->count, entry->group, entry);
 
-		err = ext4_mb_load_buddy(sb, md->group, &e4b);
+		err = ext4_mb_load_buddy(sb, entry->group, &e4b);
 		/* we expect to find existing buddy because it's pinned */
 		BUG_ON(err != 0);
 
+		db = e4b.bd_info;
 		/* there are blocks to put in buddy to make them really free */
-		count += md->num;
+		count += entry->count;
 		count2++;
-		ext4_lock_group(sb, md->group);
-		for (i = 0; i < md->num; i++) {
-			mb_debug(" %u", md->blocks[i]);
-			mb_free_blocks(NULL, &e4b, md->blocks[i], 1);
+		ext4_lock_group(sb, entry->group);
+		/* Take it out of per group rb tree */
+		rb_erase(&entry->node, &(db->bb_free_root));
+		mb_free_blocks(NULL, &e4b, entry->start_blk, entry->count);
+
+		if (!db->bb_free_root.rb_node) {
+			/* No more items in the per group rb tree
+			 * balance refcounts from ext4_mb_free_metadata()
+			 */
+			page_cache_release(e4b.bd_buddy_page);
+			page_cache_release(e4b.bd_bitmap_page);
 		}
-		mb_debug("\n");
-		ext4_unlock_group(sb, md->group);
-
-		/* balance refcounts from ext4_mb_free_metadata() */
-		page_cache_release(e4b.bd_buddy_page);
-		page_cache_release(e4b.bd_bitmap_page);
+		ext4_unlock_group(sb, entry->group);
 
-		kfree(md);
+		kmem_cache_free(ext4_free_ext_cachep, entry);
 		ext4_mb_release_desc(&e4b);
-
-	} while (md);
+	} while (1);
 
 	mb_debug("freed %u blocks in %u structures\n", count, count2);
 }
@@ -2771,6 +2772,16 @@ int __init init_ext4_mballoc(void)
 		kmem_cache_destroy(ext4_pspace_cachep);
 		return -ENOMEM;
 	}
+
+	ext4_free_ext_cachep =
+		kmem_cache_create("ext4_free_block_extents",
+				     sizeof(struct ext4_free_data),
+				     0, SLAB_RECLAIM_ACCOUNT, NULL);
+	if (ext4_free_ext_cachep == NULL) {
+		kmem_cache_destroy(ext4_pspace_cachep);
+		kmem_cache_destroy(ext4_ac_cachep);
+		return -ENOMEM;
+	}
 	return 0;
 }
 
@@ -2779,6 +2790,7 @@ void exit_ext4_mballoc(void)
 	/* XXX: synchronize_rcu(); */
 	kmem_cache_destroy(ext4_pspace_cachep);
 	kmem_cache_destroy(ext4_ac_cachep);
+	kmem_cache_destroy(ext4_free_ext_cachep);
 }
 
 
@@ -4415,6 +4427,21 @@ static void ext4_mb_poll_new_transaction(struct super_block *sb,
 	ext4_mb_free_committed_blocks(sb);
 }
 
+/*
+ * We can merge two free data extents only if the physical blocks
+ * are contiguous, AND the extents were freed by the same transaction,
+ * AND the blocks are associated with the same group.
+ */
+static int can_merge(struct ext4_free_data *entry1,
+			struct ext4_free_data *entry2)
+{
+	if (entry1->t_tid == entry2->t_tid &&
+		(entry1->group == entry2->group) &&
+		(entry1->start_blk + entry1->count) == entry2->start_blk)
+		return 1;
+	return 0;
+}
+
 static noinline_for_stack int
 ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b,
 			  ext4_group_t group, ext4_grpblk_t block, int count)
@@ -4422,57 +4449,80 @@ ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b,
 	struct ext4_group_info *db = e4b->bd_info;
 	struct super_block *sb = e4b->bd_sb;
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
-	struct ext4_free_metadata *md;
-	int i;
+	struct ext4_free_data *entry, *new_entry;
+	struct rb_node **n = &db->bb_free_root.rb_node, *node;
+	struct rb_node *parent = NULL, *new_node;
+
 
 	BUG_ON(e4b->bd_bitmap_page == NULL);
 	BUG_ON(e4b->bd_buddy_page == NULL);
 
+	new_entry  = kmem_cache_alloc(ext4_free_ext_cachep, GFP_NOFS);
+	new_entry->start_blk = block;
+	new_entry->group  = group;
+	new_entry->count = count;
+	new_entry->t_tid = handle->h_transaction->t_tid;
+	new_node = &new_entry->node;
+
 	ext4_lock_group(sb, group);
-	for (i = 0; i < count; i++) {
-		md = db->bb_md_cur;
-		if (md && db->bb_tid != handle->h_transaction->t_tid) {
-			db->bb_md_cur = NULL;
-			md = NULL;
+	if (!*n) {
+		/* first free block exent. We need to
+		   protect buddy cache from being freed,
+		 * otherwise we'll refresh it from
+		 * on-disk bitmap and lose not-yet-available
+		 * blocks */
+		page_cache_get(e4b->bd_buddy_page);
+		page_cache_get(e4b->bd_bitmap_page);
+	}
+	while (*n) {
+		parent = *n;
+		entry = rb_entry(parent, struct ext4_free_data, node);
+		if (block < entry->start_blk)
+			n = &(*n)->rb_left;
+		else if (block >= (entry->start_blk + entry->count))
+			n = &(*n)->rb_right;
+		else {
+			ext4_error(sb, __func__,
+			    "Double free of blocks %d (%d %d)\n",
+			    block, entry->start_blk, entry->count);
+			return 0;
 		}
+	}
 
-		if (md == NULL) {
-			ext4_unlock_group(sb, group);
-			md = kmalloc(sizeof(*md), GFP_NOFS);
-			if (md == NULL)
-				return -ENOMEM;
-			md->num = 0;
-			md->group = group;
-
-			ext4_lock_group(sb, group);
-			if (db->bb_md_cur == NULL) {
-				spin_lock(&sbi->s_md_lock);
-				list_add(&md->list, &sbi->s_active_transaction);
-				spin_unlock(&sbi->s_md_lock);
-				/* protect buddy cache from being freed,
-				 * otherwise we'll refresh it from
-				 * on-disk bitmap and lose not-yet-available
-				 * blocks */
-				page_cache_get(e4b->bd_buddy_page);
-				page_cache_get(e4b->bd_bitmap_page);
-				db->bb_md_cur = md;
-				db->bb_tid = handle->h_transaction->t_tid;
-				mb_debug("new md 0x%p for group %lu\n",
-						md, md->group);
-			} else {
-				kfree(md);
-				md = db->bb_md_cur;
-			}
+	rb_link_node(new_node, parent, n);
+	rb_insert_color(new_node, &db->bb_free_root);
+
+	/* Now try to see the extent can be merged to left and right */
+	node = rb_prev(new_node);
+	if (node) {
+		entry = rb_entry(node, struct ext4_free_data, node);
+		if (can_merge(entry, new_entry)) {
+			new_entry->start_blk = entry->start_blk;
+			new_entry->count += entry->count;
+			rb_erase(node, &(db->bb_free_root));
+			spin_lock(&sbi->s_md_lock);
+			list_del(&entry->list);
+			spin_unlock(&sbi->s_md_lock);
+			kmem_cache_free(ext4_free_ext_cachep, entry);
 		}
+	}
 
-		BUG_ON(md->num >= EXT4_BB_MAX_BLOCKS);
-		md->blocks[md->num] = block + i;
-		md->num++;
-		if (md->num == EXT4_BB_MAX_BLOCKS) {
-			/* no more space, put full container on a sb's list */
-			db->bb_md_cur = NULL;
+	node = rb_next(new_node);
+	if (node) {
+		entry = rb_entry(node, struct ext4_free_data, node);
+		if (can_merge(new_entry, entry)) {
+			new_entry->count += entry->count;
+			rb_erase(node, &(db->bb_free_root));
+			spin_lock(&sbi->s_md_lock);
+			list_del(&entry->list);
+			spin_unlock(&sbi->s_md_lock);
+			kmem_cache_free(ext4_free_ext_cachep, entry);
 		}
 	}
+	/* Add the extent to active_transaction list */
+	spin_lock(&sbi->s_md_lock);
+	list_add(&new_entry->list, &sbi->s_active_transaction);
+	spin_unlock(&sbi->s_md_lock);
 	ext4_unlock_group(sb, group);
 	return 0;
 }
diff --git a/fs/ext4/mballoc.h b/fs/ext4/mballoc.h
index b3b4828..9e815c4 100644
--- a/fs/ext4/mballoc.h
+++ b/fs/ext4/mballoc.h
@@ -98,23 +98,29 @@
 
 static struct kmem_cache *ext4_pspace_cachep;
 static struct kmem_cache *ext4_ac_cachep;
+static struct kmem_cache *ext4_free_ext_cachep;
 
-#ifdef EXT4_BB_MAX_BLOCKS
-#undef EXT4_BB_MAX_BLOCKS
-#endif
-#define EXT4_BB_MAX_BLOCKS	30
+struct ext4_free_data {
+	/* this links the free block information from group_info */
+	struct rb_node node;
 
-struct ext4_free_metadata {
-	ext4_group_t group;
-	unsigned short num;
-	ext4_grpblk_t  blocks[EXT4_BB_MAX_BLOCKS];
+	/* this links the free block information from ext4_sb_info */
 	struct list_head list;
+
+	/* group which free block extent belongs */
+	ext4_group_t group;
+
+	/* free block extent */
+	ext4_grpblk_t start_blk;
+	ext4_grpblk_t count;
+
+	/* transaction which freed this extent */
+	tid_t	t_tid;
 };
 
 struct ext4_group_info {
 	unsigned long	bb_state;
-	unsigned long	bb_tid;
-	struct ext4_free_metadata *bb_md_cur;
+	struct rb_root  bb_free_root;
 	unsigned short	bb_first_free;
 	unsigned short	bb_free;
 	unsigned short	bb_fragments;
-- 
1.6.0.2.526.g5c283

[PATCH] ext4: Use tag dirty lookup during mpage_da_submit_io

[Aneesh Kumar K.V]

This enables us to drop the range_cont writeback mode
use from ext4_da_writepages.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 fs/ext4/inode.c |   30 +++++++++++++-----------------
 1 files changed, 13 insertions(+), 17 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 7c2820e..cba7960 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1656,17 +1656,23 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
 
 	while (index <= end) {
 		/* XXX: optimize tail */
-		nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
+		/*
+		 * We can use PAGECACHE_TAG_DIRTY lookup here because
+		 * even though we have cleared the dirty flag on the page
+		 * We still keep the page in the radix tree with tag
+		 * PAGECACHE_TAG_DIRTY. See clear_page_dirty_for_io.
+		 * The PAGECACHE_TAG_DIRTY is cleared in set_page_writeback
+		 * which is called via the below writepage callback.
+		 */
+		nr_pages = pagevec_lookup_tag(&pvec, mapping, &index,
+					PAGECACHE_TAG_DIRTY,
+					min(end - index,
+					(pgoff_t)PAGEVEC_SIZE-1) + 1);
 		if (nr_pages == 0)
 			break;
 		for (i = 0; i < nr_pages; i++) {
 			struct page *page = pvec.pages[i];
 
-			index = page->index;
-			if (index > end)
-				break;
-			index++;
-
 			err = mapping->a_ops->writepage(page, mpd->wbc);
 			if (!err)
 				mpd->pages_written++;
@@ -2361,7 +2367,6 @@ static int ext4_da_writepages(struct address_space *mapping,
 			      struct writeback_control *wbc)
 {
 	handle_t *handle = NULL;
-	loff_t range_start = 0;
 	struct mpage_da_data mpd;
 	struct inode *inode = mapping->host;
 	int needed_blocks, ret = 0, nr_to_writebump = 0;
@@ -2386,14 +2391,7 @@ static int ext4_da_writepages(struct address_space *mapping,
 		wbc->nr_to_write = sbi->s_mb_stream_request;
 	}
 
-	if (!wbc->range_cyclic)
-		/*
-		 * If range_cyclic is not set force range_cont
-		 * and save the old writeback_index
-		 */
-		wbc->range_cont = 1;
 
-	range_start =  wbc->range_start;
 	pages_skipped = wbc->pages_skipped;
 
 	mpd.wbc = wbc;
@@ -2452,9 +2450,8 @@ static int ext4_da_writepages(struct address_space *mapping,
 		wbc->nr_to_write = to_write;
 	}
 
-	if (wbc->range_cont && (pages_skipped != wbc->pages_skipped)) {
+	if (!wbc->range_cyclic && (pages_skipped != wbc->pages_skipped)) {
 		/* We skipped pages in this loop */
-		wbc->range_start = range_start;
 		wbc->nr_to_write = to_write +
 				wbc->pages_skipped - pages_skipped;
 		wbc->pages_skipped = pages_skipped;
@@ -2463,7 +2460,6 @@ static int ext4_da_writepages(struct address_space *mapping,
 
 out_writepages:
 	wbc->nr_to_write = to_write - nr_to_writebump;
-	wbc->range_start = range_start;
 	return ret;
 }
 
-- 
1.6.0.2.526.g5c283

[PATCH] vfs: Remove the range_cont writeback mode.

[Aneesh Kumar K.V]

Ext4 was the only user of range_cont writeback mode
and ext4 switched to a different method. So remove
the range_cont mode which is not used in the kernel.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
CC: linux-fsdevel@vger.kernel.org
---
 include/linux/writeback.h |    1 -
 mm/page-writeback.c       |    2 --
 2 files changed, 0 insertions(+), 3 deletions(-)

diff --git a/include/linux/writeback.h b/include/linux/writeback.h
index 12b15c5..bd91987 100644
--- a/include/linux/writeback.h
+++ b/include/linux/writeback.h
@@ -63,7 +63,6 @@ struct writeback_control {
 	unsigned for_writepages:1;	/* This is a writepages() call */
 	unsigned range_cyclic:1;	/* range_start is cyclic */
 	unsigned more_io:1;		/* more io to be dispatched */
-	unsigned range_cont:1;
 };
 
 /*
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 24de8b6..718efa6 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -961,8 +961,6 @@ int write_cache_pages(struct address_space *mapping,
 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
 		mapping->writeback_index = index;
 
-	if (wbc->range_cont)
-		wbc->range_start = index << PAGE_CACHE_SHIFT;
 	return ret;
 }
 EXPORT_SYMBOL(write_cache_pages);
-- 
1.6.0.2.526.g5c283

[PATCH] vfs: Add no_nrwrite_update and no_index_update writeback control flags

[Aneesh Kumar K.V]

If no_nrwrite_update is set we don't update nr_to_write in
write_cache_pages. Similarly if no_index_update is we don't
update address space writeback_index. These changes enable a
file system to skip these updates in write_cache_pages and do
them in the writepages() callback. This patch will be followed
by an ext4 patch that make use of these new flags.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
CC: linux-fsdevel@vger.kernel.org
---
 include/linux/writeback.h |    4 ++++
 mm/page-writeback.c       |    9 +++++++--
 2 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/include/linux/writeback.h b/include/linux/writeback.h
index bd91987..b04287e 100644
--- a/include/linux/writeback.h
+++ b/include/linux/writeback.h
@@ -63,6 +63,10 @@ struct writeback_control {
 	unsigned for_writepages:1;	/* This is a writepages() call */
 	unsigned range_cyclic:1;	/* range_start is cyclic */
 	unsigned more_io:1;		/* more io to be dispatched */
+
+	/* write_cache_pages() control */
+	unsigned no_nrwrite_update:1;	/* don't update nr_to_write */
+	unsigned no_index_update:1;	/* don't update writeback_index */
 };
 
 /*
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 718efa6..4f359f4 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -876,6 +876,7 @@ int write_cache_pages(struct address_space *mapping,
 	pgoff_t end;		/* Inclusive */
 	int scanned = 0;
 	int range_whole = 0;
+	long nr_to_write = wbc->nr_to_write;
 
 	if (wbc->nonblocking && bdi_write_congested(bdi)) {
 		wbc->encountered_congestion = 1;
@@ -939,7 +940,7 @@ int write_cache_pages(struct address_space *mapping,
 				unlock_page(page);
 				ret = 0;
 			}
-			if (ret || (--(wbc->nr_to_write) <= 0))
+			if (ret || (--nr_to_write <= 0))
 				done = 1;
 			if (wbc->nonblocking && bdi_write_congested(bdi)) {
 				wbc->encountered_congestion = 1;
@@ -958,8 +959,12 @@ int write_cache_pages(struct address_space *mapping,
 		index = 0;
 		goto retry;
 	}
-	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
+	if (!wbc->no_index_update &&
+		(wbc->range_cyclic || (range_whole && nr_to_write > 0))) {
 		mapping->writeback_index = index;
+	}
+	if (!wbc->no_nrwrite_update)
+		wbc->nr_to_write = nr_to_write;
 
 	return ret;
 }
-- 
1.6.0.2.526.g5c283

[PATCH] ext4: Fix file fragmentation during large file write.

[Aneesh Kumar K.V]

The range_cyclic writeback mode uses the address_space writeback_index
as the start index for writeback.  With delayed allocation we were
updating writeback_index wrongly resulting in highly fragmented file.
Number of extents reduced from 4000 to 27 for a 3GB file with the below
patch.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
---
 fs/ext4/inode.c |   83 +++++++++++++++++++++++++++++++++----------------------
 1 files changed, 50 insertions(+), 33 deletions(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index cba7960..844c136 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1648,6 +1648,7 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
 	int ret = 0, err, nr_pages, i;
 	unsigned long index, end;
 	struct pagevec pvec;
+	long pages_skipped;
 
 	BUG_ON(mpd->next_page <= mpd->first_page);
 	pagevec_init(&pvec, 0);
@@ -1655,7 +1656,6 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
 	end = mpd->next_page - 1;
 
 	while (index <= end) {
-		/* XXX: optimize tail */
 		/*
 		 * We can use PAGECACHE_TAG_DIRTY lookup here because
 		 * even though we have cleared the dirty flag on the page
@@ -1673,8 +1673,13 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
 		for (i = 0; i < nr_pages; i++) {
 			struct page *page = pvec.pages[i];
 
+			pages_skipped = mpd->wbc->pages_skipped;
 			err = mapping->a_ops->writepage(page, mpd->wbc);
-			if (!err)
+			if (!err && (pages_skipped == mpd->wbc->pages_skipped))
+				/*
+				 * have successfully written the page
+				 * without skipping the same
+				 */
 				mpd->pages_written++;
 			/*
 			 * In error case, we have to continue because
@@ -2110,7 +2115,6 @@ static int mpage_da_writepages(struct address_space *mapping,
 			       struct writeback_control *wbc,
 			       struct mpage_da_data *mpd)
 {
-	long to_write;
 	int ret;
 
 	if (!mpd->get_block)
@@ -2125,10 +2129,7 @@ static int mpage_da_writepages(struct address_space *mapping,
 	mpd->pages_written = 0;
 	mpd->retval = 0;
 
-	to_write = wbc->nr_to_write;
-
 	ret = write_cache_pages(mapping, wbc, __mpage_da_writepage, mpd);
-
 	/*
 	 * Handle last extent of pages
 	 */
@@ -2137,7 +2138,7 @@ static int mpage_da_writepages(struct address_space *mapping,
 			mpage_da_submit_io(mpd);
 	}
 
-	wbc->nr_to_write = to_write - mpd->pages_written;
+	wbc->nr_to_write -= mpd->pages_written;
 	return ret;
 }
 
@@ -2366,11 +2367,14 @@ static int ext4_da_writepages_trans_blocks(struct inode *inode)
 static int ext4_da_writepages(struct address_space *mapping,
 			      struct writeback_control *wbc)
 {
+	pgoff_t	index;
+	int range_whole = 0;
 	handle_t *handle = NULL;
+	long pages_written = 0;
 	struct mpage_da_data mpd;
 	struct inode *inode = mapping->host;
+	int no_nrwrite_update, no_index_update;
 	int needed_blocks, ret = 0, nr_to_writebump = 0;
-	long to_write, pages_skipped = 0;
 	struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
 
 	/*
@@ -2390,16 +2394,27 @@ static int ext4_da_writepages(struct address_space *mapping,
 		nr_to_writebump = sbi->s_mb_stream_request - wbc->nr_to_write;
 		wbc->nr_to_write = sbi->s_mb_stream_request;
 	}
+	if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
+		range_whole = 1;
 
-
-	pages_skipped = wbc->pages_skipped;
+	if (wbc->range_cyclic)
+		index = mapping->writeback_index;
+	else
+		index = wbc->range_start >> PAGE_CACHE_SHIFT;
 
 	mpd.wbc = wbc;
 	mpd.inode = mapping->host;
 
-restart_loop:
-	to_write = wbc->nr_to_write;
-	while (!ret && to_write > 0) {
+	/*
+	 * we don't want write_cache_pages to update
+	 * nr_to_write and writeback_index
+	 */
+	no_nrwrite_update = wbc->no_nrwrite_update;
+	wbc->no_nrwrite_update = 1;
+	no_index_update = wbc->no_index_update;
+	wbc->no_index_update   = 1;
+
+	while (!ret && wbc->nr_to_write > 0) {
 
 		/*
 		 * we  insert one extent at a time. So we need
@@ -2420,46 +2435,48 @@ static int ext4_da_writepages(struct address_space *mapping,
 			dump_stack();
 			goto out_writepages;
 		}
-		to_write -= wbc->nr_to_write;
-
 		mpd.get_block = ext4_da_get_block_write;
 		ret = mpage_da_writepages(mapping, wbc, &mpd);
 
 		ext4_journal_stop(handle);
 
-		if (mpd.retval == -ENOSPC)
+		if (mpd.retval == -ENOSPC) {
+			/* commit the transaction which would
+			 * free blocks released in the transaction
+			 * and try again
+			 */
 			jbd2_journal_force_commit_nested(sbi->s_journal);
-
-		/* reset the retry count */
-		if (ret == MPAGE_DA_EXTENT_TAIL) {
+			ret = 0;
+		} else if (ret == MPAGE_DA_EXTENT_TAIL) {
 			/*
 			 * got one extent now try with
 			 * rest of the pages
 			 */
-			to_write += wbc->nr_to_write;
+			pages_written += mpd.pages_written;
 			ret = 0;
-		} else if (wbc->nr_to_write) {
+		} else if (wbc->nr_to_write)
 			/*
 			 * There is no more writeout needed
 			 * or we requested for a noblocking writeout
 			 * and we found the device congested
 			 */
-			to_write += wbc->nr_to_write;
 			break;
-		}
-		wbc->nr_to_write = to_write;
-	}
-
-	if (!wbc->range_cyclic && (pages_skipped != wbc->pages_skipped)) {
-		/* We skipped pages in this loop */
-		wbc->nr_to_write = to_write +
-				wbc->pages_skipped - pages_skipped;
-		wbc->pages_skipped = pages_skipped;
-		goto restart_loop;
 	}
+	/* Update index */
+	index += pages_written;
+	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
+		/*
+		 * set the writeback_index so that range_cyclic
+		 * mode will write it back later
+		 */
+		mapping->writeback_index = index;
 
 out_writepages:
-	wbc->nr_to_write = to_write - nr_to_writebump;
+	if (!no_nrwrite_update)
+		wbc->no_nrwrite_update = 0;
+	if (!no_index_update)
+		wbc->no_index_update   = 0;
+	wbc->nr_to_write -= nr_to_writebump;
 	return ret;
 }
 
-- 
1.6.0.2.526.g5c283