Btrfs: Do metadata checksums for reads via a workqueue

Before, metadata checksumming was done by the callers of read_tree_block,
which would set EXTENT_CSUM bits in the extent tree to show that a given
range of pages was already checksummed and didn't need to be verified
again.

But, those bits could go away via try_to_releasepage, and the end
result was bogus checksum failures on pages that never left the cache.

The new code validates checksums when the page is read.  It is a little
tricky because metadata blocks can span pages and a single read may
end up going via multiple bios.

Signed-off-by: Chris Mason <chris.mason@oracle.com>

fs/btrfs/ctree.c

@@ -158,9 +158,6 @@ int __btrfs_cow_block(struct btrfs_trans_handle *trans,
 	} else {
 		root_gen = 0;
 	}
-	if (!(buf->flags & EXTENT_CSUM))
-		WARN_ON(1);
-
 	WARN_ON(root->ref_cows && trans->transid !=
 		root->fs_info->running_transaction->transid);
 	WARN_ON(root->ref_cows && trans->transid != root->last_trans);
@@ -247,8 +244,6 @@ int btrfs_cow_block(struct btrfs_trans_handle *trans,
 		       root->fs_info->generation);
 		WARN_ON(1);
 	}
-	if (!(buf->flags & EXTENT_CSUM))
-		WARN_ON(1);
 
 	header_trans = btrfs_header_generation(buf);
 	spin_lock(&root->fs_info->hash_lock);

fs/btrfs/ctree.h

@@ -485,6 +485,10 @@ struct btrfs_fs_info {
 	struct list_head trans_list;
 	struct list_head hashers;
 	struct list_head dead_roots;
+	struct list_head end_io_work_list;
+	struct work_struct end_io_work;
+	spinlock_t end_io_work_lock;
+
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 	struct work_struct trans_work;
 #else

fs/btrfs/disk-io.c

@@ -24,6 +24,7 @@
 #include <linux/radix-tree.h>
 #include <linux/writeback.h>
 #include <linux/buffer_head.h> // for block_sync_page
+#include <linux/workqueue.h>
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
@@ -45,6 +46,16 @@ static int check_tree_block(struct btrfs_root *root, struct extent_buffer *buf)
 #endif
 
 static struct extent_io_ops btree_extent_io_ops;
+static struct workqueue_struct *end_io_workqueue;
+
+struct end_io_wq {
+	struct bio *bio;
+	bio_end_io_t *end_io;
+	void *private;
+	struct btrfs_fs_info *info;
+	int error;
+	struct list_head list;
+};
 
 struct extent_map *btree_get_extent(struct inode *inode, struct page *page,
 				    size_t page_offset, u64 start, u64 len,
@@ -219,11 +230,108 @@ static int btree_writepage_io_hook(struct page *page, u64 start, u64 end)
 	return 0;
 }
 
+int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
+			       struct extent_state *state)
+{
+	struct extent_io_tree *tree;
+	u64 found_start;
+	int found_level;
+	unsigned long len;
+	struct extent_buffer *eb;
+	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+	int ret;
+
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
+	if (page->private == EXTENT_PAGE_PRIVATE)
+		goto out;
+	if (!page->private)
+		goto out;
+	len = page->private >> 2;
+	if (len == 0) {
+		WARN_ON(1);
+	}
+	eb = alloc_extent_buffer(tree, start, len, page, GFP_NOFS);
+	read_extent_buffer_pages(tree, eb, start + PAGE_CACHE_SIZE, 1,
+				 btree_get_extent);
+	btrfs_clear_buffer_defrag(eb);
+	found_start = btrfs_header_bytenr(eb);
+	if (found_start != start) {
+		printk("warning: eb start incorrect %Lu buffer %Lu len %lu\n",
+		       start, found_start, len);
+		WARN_ON(1);
+		goto err;
+	}
+	if (eb->first_page != page) {
+		printk("bad first page %lu %lu\n", eb->first_page->index,
+		       page->index);
+		WARN_ON(1);
+		goto err;
+	}
+	found_level = btrfs_header_level(eb);
+
+	ret = csum_tree_block(root, eb, 1);
+
+	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
+	end = eb->start + end - 1;
+	release_extent_buffer_tail_pages(eb);
+err:
+	free_extent_buffer(eb);
+out:
+	return 0;
+}
+
+#if LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)
+static void end_workqueue_bio(struct bio *bio, int err)
+#else
+static int end_workqueue_bio(struct bio *bio,
+				   unsigned int bytes_done, int err)
+#endif
+{
+	struct end_io_wq *end_io_wq = bio->bi_private;
+	struct btrfs_fs_info *fs_info;
+	unsigned long flags;
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+	if (bio->bi_size)
+		return 1;
+#endif
+
+	fs_info = end_io_wq->info;
+	spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
+	end_io_wq->error = err;
+	list_add_tail(&end_io_wq->list, &fs_info->end_io_work_list);
+	spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
+	queue_work(end_io_workqueue, &fs_info->end_io_work);
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,23)
+	return 0;
+#endif
+}
+
 static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio)
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct end_io_wq *end_io_wq;
 	u64 offset;
 	offset = bio->bi_sector << 9;
+
+	if (rw & (1 << BIO_RW)) {
+		return btrfs_map_bio(BTRFS_I(inode)->root, rw, bio);
+	}
+
+	end_io_wq = kmalloc(sizeof(*end_io_wq), GFP_NOFS);
+	if (!end_io_wq)
+		return -ENOMEM;
+
+	end_io_wq->private = bio->bi_private;
+	end_io_wq->end_io = bio->bi_end_io;
+	end_io_wq->info = root->fs_info;
+	end_io_wq->error = 0;
+	end_io_wq->bio = bio;
+
+	bio->bi_private = end_io_wq;
+	bio->bi_end_io = end_workqueue_bio;
+
 	if (offset == BTRFS_SUPER_INFO_OFFSET) {
 		bio->bi_bdev = root->fs_info->sb->s_bdev;
 		submit_bio(rw, bio);
@@ -363,36 +471,7 @@ static int close_all_devices(struct btrfs_fs_info *fs_info)
 int btrfs_verify_block_csum(struct btrfs_root *root,
 			    struct extent_buffer *buf)
 {
-	struct extent_io_tree *io_tree;
-	u64 end;
-	int ret;
-
-	io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
-	if (buf->flags & EXTENT_CSUM)
-		return 0;
-
-	end = min_t(u64, buf->len, PAGE_CACHE_SIZE);
-	end = buf->start + end - 1;
-	if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
-		buf->flags |= EXTENT_CSUM;
-		return 0;
-	}
-	lock_extent(io_tree, buf->start, end, GFP_NOFS);
-
-	if (test_range_bit(io_tree, buf->start, end, EXTENT_CSUM, 1)) {
-		buf->flags |= EXTENT_CSUM;
-		ret = 0;
-		goto out_unlock;
-	}
-WARN_ON(buf->flags & EXTENT_CSUM);
-
-	ret = csum_tree_block(root, buf, 1);
-	set_extent_bits(io_tree, buf->start, end, EXTENT_CSUM, GFP_NOFS);
-	buf->flags |= EXTENT_CSUM;
-
-out_unlock:
-	unlock_extent(io_tree, buf->start, end, GFP_NOFS);
-	return ret;
+	return btrfs_buffer_uptodate(buf);
 }
 
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
@@ -430,11 +509,15 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
 	if (!buf)
 		return NULL;
-	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0, 1,
-				 btree_get_extent);
 
-	ret = btrfs_verify_block_csum(root, buf);
+	ret = read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree, buf, 0,
+				       1, btree_get_extent);
+
+	if (ret == 0) {
+		buf->flags |= EXTENT_UPTODATE;
+	}
 	return buf;
+
 }
 
 int clean_tree_block(struct btrfs_trans_handle *trans, struct btrfs_root *root,
@@ -724,6 +807,99 @@ static int setup_bdi(struct btrfs_fs_info *info, struct backing_dev_info *bdi)
 	return 0;
 }
 
+static int bio_ready_for_csum(struct bio *bio)
+{
+	u64 length = 0;
+	u64 buf_len = 0;
+	u64 start = 0;
+	struct page *page;
+	struct extent_io_tree *io_tree = NULL;
+	struct btrfs_fs_info *info = NULL;
+	struct bio_vec *bvec;
+	int i;
+	int ret;
+
+	bio_for_each_segment(bvec, bio, i) {
+		page = bvec->bv_page;
+		if (page->private == EXTENT_PAGE_PRIVATE) {
+			length += bvec->bv_len;
+			continue;
+		}
+		if (!page->private) {
+			length += bvec->bv_len;
+			continue;
+		}
+		length = bvec->bv_len;
+		buf_len = page->private >> 2;
+		start = page_offset(page) + bvec->bv_offset;
+		io_tree = &BTRFS_I(page->mapping->host)->io_tree;
+		info = BTRFS_I(page->mapping->host)->root->fs_info;
+	}
+	/* are we fully contained in this bio? */
+	if (buf_len <= length)
+		return 1;
+
+	ret = extent_range_uptodate(io_tree, start + length,
+				    start + buf_len - 1);
+	if (ret == 1)
+		return ret;
+	return ret;
+}
+
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+void btrfs_end_io_csum(void *p)
+#else
+void btrfs_end_io_csum(struct work_struct *work)
+#endif
+{
+#if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
+	struct btrfs_fs_info *fs_info = p;
+#else
+	struct btrfs_fs_info *fs_info = container_of(work,
+						     struct btrfs_fs_info,
+						     end_io_work);
+#endif
+	unsigned long flags;
+	struct end_io_wq *end_io_wq;
+	struct bio *bio;
+	struct list_head *next;
+	int error;
+	int was_empty;
+
+	while(1) {
+		spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
+		if (list_empty(&fs_info->end_io_work_list)) {
+			spin_unlock_irqrestore(&fs_info->end_io_work_lock,
+					       flags);
+			return;
+		}
+		next = fs_info->end_io_work_list.next;
+		list_del(next);
+		spin_unlock_irqrestore(&fs_info->end_io_work_lock, flags);
+
+		end_io_wq = list_entry(next, struct end_io_wq, list);
+
+		bio = end_io_wq->bio;
+		if (!bio_ready_for_csum(bio)) {
+			spin_lock_irqsave(&fs_info->end_io_work_lock, flags);
+			was_empty = list_empty(&fs_info->end_io_work_list);
+			list_add_tail(&end_io_wq->list,
+				      &fs_info->end_io_work_list);
+			spin_unlock_irqrestore(&fs_info->end_io_work_lock,
+					       flags);
+			if (was_empty)
+				return;
+			continue;
+		}
+		error = end_io_wq->error;
+		bio->bi_private = end_io_wq->private;
+		bio->bi_end_io = end_io_wq->end_io;
+		kfree(end_io_wq);
+		bio_endio(bio, error);
+	}
+}
+
+
 struct btrfs_root *open_ctree(struct super_block *sb,
 			      struct btrfs_fs_devices *fs_devices)
 {
@@ -750,11 +926,16 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		err = -ENOMEM;
 		goto fail;
 	}
+	end_io_workqueue = create_workqueue("btrfs-end-io");
+	BUG_ON(!end_io_workqueue);
+
 	INIT_RADIX_TREE(&fs_info->fs_roots_radix, GFP_NOFS);
 	INIT_LIST_HEAD(&fs_info->trans_list);
 	INIT_LIST_HEAD(&fs_info->dead_roots);
 	INIT_LIST_HEAD(&fs_info->hashers);
+	INIT_LIST_HEAD(&fs_info->end_io_work_list);
 	spin_lock_init(&fs_info->hash_lock);
+	spin_lock_init(&fs_info->end_io_work_lock);
 	spin_lock_init(&fs_info->delalloc_lock);
 	spin_lock_init(&fs_info->new_trans_lock);
 
@@ -799,6 +980,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 			     fs_info->btree_inode->i_mapping, GFP_NOFS);
 	fs_info->do_barriers = 1;
 
+	INIT_WORK(&fs_info->end_io_work, btrfs_end_io_csum);
 #if LINUX_VERSION_CODE <= KERNEL_VERSION(2,6,18)
 	INIT_WORK(&fs_info->trans_work, btrfs_transaction_cleaner, fs_info);
 #else
@@ -1044,6 +1226,8 @@ int close_ctree(struct btrfs_root *root)
 	extent_io_tree_empty_lru(&BTRFS_I(fs_info->btree_inode)->io_tree);
 
 	truncate_inode_pages(fs_info->btree_inode->i_mapping, 0);
+	flush_workqueue(end_io_workqueue);
+	destroy_workqueue(end_io_workqueue);
 
 	iput(fs_info->btree_inode);
 #if 0
@@ -1171,12 +1355,18 @@ int btrfs_read_buffer(struct extent_buffer *buf)
 {
 	struct btrfs_root *root = BTRFS_I(buf->first_page->mapping->host)->root;
 	struct inode *btree_inode = root->fs_info->btree_inode;
-	return read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
+	int ret;
+	ret = read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
 					buf, 0, 1, btree_get_extent);
+	if (ret == 0) {
+		buf->flags |= EXTENT_UPTODATE;
+	}
+	return ret;
 }
 
 static struct extent_io_ops btree_extent_io_ops = {
 	.writepage_io_hook = btree_writepage_io_hook,
+	.readpage_end_io_hook = btree_readpage_end_io_hook,
 	.submit_bio_hook = btree_submit_bio_hook,
 	/* note we're sharing with inode.c for the merge bio hook */
 	.merge_bio_hook = btrfs_merge_bio_hook,

fs/btrfs/extent-tree.c

@@ -1898,10 +1898,6 @@ struct extent_buffer *__btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 
 	set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
 			 buf->start + buf->len - 1, GFP_NOFS);
-	set_extent_bits(&BTRFS_I(root->fs_info->btree_inode)->io_tree,
-			buf->start, buf->start + buf->len - 1,
-			EXTENT_CSUM, GFP_NOFS);
-	buf->flags |= EXTENT_CSUM;
 	if (!btrfs_test_opt(root, SSD))
 		btrfs_set_buffer_defrag(buf);
 	trans->blocks_used++;

fs/btrfs/extent_io.c

@@ -2592,6 +2592,22 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb,
 	return p;
 }
 
+int release_extent_buffer_tail_pages(struct extent_buffer *eb)
+{
+	unsigned long num_pages = num_extent_pages(eb->start, eb->len);
+	struct page *page;
+	unsigned long i;
+
+	if (num_pages == 1)
+		return 0;
+	for (i = 1; i < num_pages; i++) {
+		page = extent_buffer_page(eb, i);
+		page_cache_release(page);
+	}
+	return 0;
+}
+
+
 int invalidate_extent_lru(struct extent_io_tree *tree, u64 start,
 			  unsigned long len)
 {
@@ -2609,9 +2625,6 @@ int invalidate_extent_lru(struct extent_io_tree *tree, u64 start,
 		if (eb->start <= start && eb->start + eb->len > start) {
 			eb->flags &= ~EXTENT_UPTODATE;
 		}
-		if (eb->start == start) {
-			eb->flags &= ~EXTENT_CSUM;
-		}
 		cur = cur->next;
 	} while (cur != lru);
 out:
@@ -2682,7 +2695,6 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
 		page_cache_get(page0);
 		mark_page_accessed(page0);
 		set_page_extent_mapped(page0);
-		WARN_ON(!PageUptodate(page0));
 		set_page_extent_head(page0, len);
 	} else {
 		i = 0;
@@ -2933,13 +2945,39 @@ int set_extent_buffer_uptodate(struct extent_io_tree *tree,
 }
 EXPORT_SYMBOL(set_extent_buffer_uptodate);
 
+int extent_range_uptodate(struct extent_io_tree *tree,
+			  u64 start, u64 end)
+{
+	struct page *page;
+	int ret;
+	int pg_uptodate = 1;
+	int uptodate;
+	unsigned long index;
+
+	ret = test_range_bit(tree, start, end, EXTENT_UPTODATE, 1);
+	if (ret)
+		return 1;
+	while(start <= end) {
+		index = start >> PAGE_CACHE_SHIFT;
+		page = find_get_page(tree->mapping, index);
+		uptodate = PageUptodate(page);
+		page_cache_release(page);
+		if (!uptodate) {
+			pg_uptodate = 0;
+			break;
+		}
+		start += PAGE_CACHE_SIZE;
+	}
+	return pg_uptodate;
+}
+
 int extent_buffer_uptodate(struct extent_io_tree *tree,
-			     struct extent_buffer *eb)
+			   struct extent_buffer *eb)
 {
 	int ret = 0;
 	int ret2;
-	int num_pages;
-	int i;
+	unsigned long num_pages;
+	unsigned long i;
 	struct page *page;
 	int pg_uptodate = 1;
 
@@ -2975,13 +3013,16 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	struct page *page;
 	int err;
 	int ret = 0;
+	int locked_pages = 0;
+	int all_uptodate = 1;
+	int inc_all_pages = 0;
 	unsigned long num_pages;
 	struct bio *bio = NULL;
 
 	if (eb->flags & EXTENT_UPTODATE)
 		return 0;
 
-	if (0 && test_range_bit(tree, eb->start, eb->start + eb->len - 1,
+	if (test_range_bit(tree, eb->start, eb->start + eb->len - 1,
 			   EXTENT_UPTODATE, 1)) {
 		return 0;
 	}
@@ -2997,17 +3038,30 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	num_pages = num_extent_pages(eb->start, eb->len);
 	for (i = start_i; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
-		if (PageUptodate(page)) {
-			continue;
-		}
 		if (!wait) {
-			if (TestSetPageLocked(page)) {
-				continue;
-			}
+			if (TestSetPageLocked(page))
+				goto unlock_exit;
 		} else {
 			lock_page(page);
 		}
+		locked_pages++;
+		if (!PageUptodate(page)) {
+			all_uptodate = 0;
+		}
+	}
+	if (all_uptodate) {
+		if (start_i == 0)
+			eb->flags |= EXTENT_UPTODATE;
+		goto unlock_exit;
+	}
+
+	for (i = start_i; i < num_pages; i++) {
+		page = extent_buffer_page(eb, i);
+		if (inc_all_pages)
+			page_cache_get(page);
 		if (!PageUptodate(page)) {
+			if (start_i == 0)
+				inc_all_pages = 1;
 			err = __extent_read_full_page(tree, page,
 						      get_extent, &bio);
 			if (err) {
@@ -3034,6 +3088,16 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	if (!ret)
 		eb->flags |= EXTENT_UPTODATE;
 	return ret;
+
+unlock_exit:
+	i = start_i;
+	while(locked_pages > 0) {
+		page = extent_buffer_page(eb, i);
+		i++;
+		unlock_page(page);
+		locked_pages--;
+	}
+	return ret;
 }
 EXPORT_SYMBOL(read_extent_buffer_pages);
 
@@ -3048,7 +3112,6 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv,
 	char *dst = (char *)dstv;
 	size_t start_offset = eb->start & ((u64)PAGE_CACHE_SIZE - 1);
 	unsigned long i = (start_offset + start) >> PAGE_CACHE_SHIFT;
-	unsigned long num_pages = num_extent_pages(eb->start, eb->len);
 
 	WARN_ON(start > eb->len);
 	WARN_ON(start + len > eb->start + eb->len);
@@ -3057,11 +3120,6 @@ void read_extent_buffer(struct extent_buffer *eb, void *dstv,
 
 	while(len > 0) {
 		page = extent_buffer_page(eb, i);
-		if (!PageUptodate(page)) {
-			printk("page %lu not up to date i %lu, total %lu, len %lu\n", page->index, i, num_pages, eb->len);
-			WARN_ON(1);
-		}
-		WARN_ON(!PageUptodate(page));
 
 		cur = min(len, (PAGE_CACHE_SIZE - offset));
 		kaddr = kmap_atomic(page, KM_USER1);
@@ -3105,7 +3163,6 @@ printk("bad mapping eb start %Lu len %lu, wanted %lu %lu\n", eb->start, eb->len,
 	}
 
 	p = extent_buffer_page(eb, i);
-	WARN_ON(!PageUptodate(p));
 	kaddr = kmap_atomic(p, km);
 	*token = kaddr;
 	*map = kaddr + offset;
@@ -3165,7 +3222,6 @@ int memcmp_extent_buffer(struct extent_buffer *eb, const void *ptrv,
 
 	while(len > 0) {
 		page = extent_buffer_page(eb, i);
-		WARN_ON(!PageUptodate(page));
 
 		cur = min(len, (PAGE_CACHE_SIZE - offset));
 

fs/btrfs/extent_io.h

@@ -13,7 +13,6 @@
 #define EXTENT_DEFRAG (1 << 6)
 #define EXTENT_DEFRAG_DONE (1 << 7)
 #define EXTENT_BUFFER_FILLED (1 << 8)
-#define EXTENT_CSUM (1 << 9)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
 
 /*
@@ -218,4 +217,7 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long offset,
 void unmap_extent_buffer(struct extent_buffer *eb, char *token, int km);
 int invalidate_extent_lru(struct extent_io_tree *tree, u64 start,
 			  unsigned long len);
+int release_extent_buffer_tail_pages(struct extent_buffer *eb);
+int extent_range_uptodate(struct extent_io_tree *tree,
+			  u64 start, u64 end);
 #endif

fs/btrfs/transaction.c

@@ -843,7 +843,7 @@ void btrfs_transaction_flush_work(struct btrfs_root *root)
 
 void __init btrfs_init_transaction_sys(void)
 {
-	trans_wq = create_workqueue("btrfs");
+	trans_wq = create_workqueue("btrfs-transaction");
 }
 
 void btrfs_exit_transaction_sys(void)