Btrfs: Handle data checksumming on bios that span multiple ordered extents

Data checksumming is done right before the bio is sent down the IO stack,
which means a single bio might span more than one ordered extent.  In
this case, the checksumming data is split between two ordered extents.

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

fs/btrfs/ctree.h

@@ -1579,8 +1579,8 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, struct inode *inode,
 			   struct btrfs_ordered_sum *sums);
-int btrfs_csum_one_bio(struct btrfs_root *root,
+int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
-		       struct bio *bio, struct btrfs_ordered_sum **sums_ret);
+		       struct bio *bio);
 struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 					  struct btrfs_root *root,
 					  struct btrfs_path *path,

fs/btrfs/file-item.c

@@ -134,26 +134,53 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
 	return ret;
 }
 
-int btrfs_csum_one_bio(struct btrfs_root *root,
+int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
-		       struct bio *bio, struct btrfs_ordered_sum **sums_ret)
+		       struct bio *bio)
 {
 	struct btrfs_ordered_sum *sums;
 	struct btrfs_sector_sum *sector_sum;
+	struct btrfs_ordered_extent *ordered;
 	char *data;
 	struct bio_vec *bvec = bio->bi_io_vec;
 	int bio_index = 0;
+	unsigned long total_bytes = 0;
+	unsigned long this_sum_bytes = 0;
+	u64 offset;
 
 	WARN_ON(bio->bi_vcnt <= 0);
 	sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS);
 	if (!sums)
 		return -ENOMEM;
-	*sums_ret = sums;
+
 	sector_sum = &sums->sums;
-	sums->file_offset = page_offset(bvec->bv_page);
+	sums->file_offset = page_offset(bvec->bv_page) + bvec->bv_offset;
 	sums->len = bio->bi_size;
 	INIT_LIST_HEAD(&sums->list);
+	ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset);
+	BUG_ON(!ordered);
 
 	while(bio_index < bio->bi_vcnt) {
+		offset = page_offset(bvec->bv_page) + bvec->bv_offset;
+		if (offset >= ordered->file_offset + ordered->len) {
+			unsigned long bytes_left;
+			sums->len = this_sum_bytes;
+			this_sum_bytes = 0;
+			btrfs_add_ordered_sum(inode, ordered, sums);
+			btrfs_put_ordered_extent(ordered);
+
+			bytes_left = bio->bi_size - total_bytes;
+
+			sums = kzalloc(btrfs_ordered_sum_size(root, bytes_left),
+				       GFP_NOFS);
+			BUG_ON(!sums);
+			sector_sum = &sums->sums;
+			sums->len = bytes_left;
+			sums->file_offset = offset;
+			ordered = btrfs_lookup_ordered_extent(inode,
+						      sums->file_offset);
+			BUG_ON(!ordered);
+		}
+
 		data = kmap_atomic(bvec->bv_page, KM_USER0);
 		sector_sum->sum = ~(u32)0;
 		sector_sum->sum = btrfs_csum_data(root,
@@ -165,10 +192,18 @@ int btrfs_csum_one_bio(struct btrfs_root *root,
 				 (char *)&sector_sum->sum);
 		sector_sum->offset = page_offset(bvec->bv_page) +
 			bvec->bv_offset;
+
 		sector_sum++;
 		bio_index++;
+		total_bytes += bvec->bv_len;
+		this_sum_bytes += bvec->bv_len;
 		bvec++;
 	}
+	btrfs_add_ordered_sum(inode, ordered, sums);
+	btrfs_put_ordered_extent(ordered);
+	if (total_bytes != bio->bi_size) {
+printk("warning, total bytes %lu bio size %u\n", total_bytes, bio->bi_size);
+	}
 	return 0;
 }
 

fs/btrfs/inode.c

@@ -351,12 +351,8 @@ int __btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
 {
 	struct btrfs_root *root = BTRFS_I(inode)->root;
 	int ret = 0;
-	struct btrfs_ordered_sum *sums;
 
-	ret = btrfs_csum_one_bio(root, bio, &sums);
+	ret = btrfs_csum_one_bio(root, inode, bio);
-	BUG_ON(ret);
-
-	ret = btrfs_add_ordered_sum(inode, sums);
 	BUG_ON(ret);
 
 	return btrfs_map_bio(root, rw, bio, mirror_num, 1);

fs/btrfs/ordered-data.c

@@ -186,22 +186,17 @@ int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
 
 /*
  * Add a struct btrfs_ordered_sum into the list of checksums to be inserted
- * when an ordered extent is finished.
+ * when an ordered extent is finished.  If the list covers more than one
+ * ordered extent, it is split across multiples.
  */
-int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum)
+int btrfs_add_ordered_sum(struct inode *inode,
+			  struct btrfs_ordered_extent *entry,
+			  struct btrfs_ordered_sum *sum)
 {
 	struct btrfs_ordered_inode_tree *tree;
-	struct rb_node *node;
-	struct btrfs_ordered_extent *entry;
 
 	tree = &BTRFS_I(inode)->ordered_tree;
 	mutex_lock(&tree->mutex);
-	node = tree_search(tree, sum->file_offset);
-	BUG_ON(!node);
-
-	entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
-	BUG_ON(!offset_in_entry(entry, sum->file_offset));
-
 	list_add_tail(&sum->list, &entry->list);
 	mutex_unlock(&tree->mutex);
 	return 0;
@@ -524,8 +519,10 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum)
 	struct btrfs_ordered_extent *ordered;
 	struct btrfs_ordered_inode_tree *tree = &BTRFS_I(inode)->ordered_tree;
 	struct list_head *cur;
+	unsigned long num_sectors;
+	unsigned long i;
+	u32 sectorsize = BTRFS_I(inode)->root->sectorsize;
 	int ret = 1;
-	int index;
 
 	ordered = btrfs_lookup_ordered_extent(inode, offset);
 	if (!ordered)
@@ -534,14 +531,17 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum)
 	mutex_lock(&tree->mutex);
 	list_for_each_prev(cur, &ordered->list) {
 		ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list);
-		if (offset >= ordered_sum->file_offset &&
+		if (offset >= ordered_sum->file_offset) {
-		    offset < ordered_sum->file_offset + ordered_sum->len) {
+			num_sectors = ordered_sum->len / sectorsize;
-			index = (offset - ordered_sum->file_offset) /
-				BTRFS_I(inode)->root->sectorsize;;
 			sector_sums = &ordered_sum->sums;
-			*sum = sector_sums[index].sum;
+			for (i = 0; i < num_sectors; i++) {
-			ret = 0;
+				if (sector_sums[i].offset == offset) {
-			goto out;
+printk("find ordered sum inode %lu offset %Lu\n", inode->i_ino, offset);
+					*sum = sector_sums[i].sum;
+					ret = 0;
+					goto out;
+				}
+			}
 		}
 	}
 out:

fs/btrfs/ordered-data.h

@@ -39,7 +39,11 @@ struct btrfs_sector_sum {
 
 struct btrfs_ordered_sum {
 	u64 file_offset;
-	u64 len;
+	/*
+	 * this is the length in bytes covered by the sums array below.
+	 * But, the sums array may not be contiguous in the file.
+	 */
+	unsigned long len;
 	struct list_head list;
 	/* last field is a variable length array of btrfs_sector_sums */
 	struct btrfs_sector_sum sums;
@@ -95,6 +99,7 @@ static inline int btrfs_ordered_sum_size(struct btrfs_root *root, u64 bytes)
 {
 	unsigned long num_sectors = (bytes + root->sectorsize - 1) /
 		root->sectorsize;
+	num_sectors++;
 	return sizeof(struct btrfs_ordered_sum) +
 		num_sectors * sizeof(struct btrfs_sector_sum);
 }
@@ -114,7 +119,9 @@ int btrfs_dec_test_ordered_pending(struct inode *inode,
 				       u64 file_offset, u64 io_size);
 int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
 			     u64 start, u64 len);
-int btrfs_add_ordered_sum(struct inode *inode, struct btrfs_ordered_sum *sum);
+int btrfs_add_ordered_sum(struct inode *inode,
+			  struct btrfs_ordered_extent *entry,
+			  struct btrfs_ordered_sum *sum);
 struct btrfs_ordered_extent *btrfs_lookup_ordered_extent(struct inode *inode,
 							 u64 file_offset);
 void btrfs_start_ordered_extent(struct inode *inode,