1 files changed, 79 insertions, 66 deletions

diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
index dd38dfe174df..c3c0c064c25d 100644
--- a/fs/btrfs/send.c
+++ b/fs/btrfs/send.c
@@ -686,7 +686,7 @@ static int send_cmd(struct send_ctx *sctx)
 	hdr->len = cpu_to_le32(sctx->send_size - sizeof(*hdr));
 	hdr->crc = 0;
 
-	crc = crc32c(0, (unsigned char *)sctx->send_buf, sctx->send_size);
+	crc = btrfs_crc32c(0, (unsigned char *)sctx->send_buf, sctx->send_size);
 	hdr->crc = cpu_to_le32(crc);
 
 	ret = write_buf(sctx->send_filp, sctx->send_buf, sctx->send_size,
@@ -4999,6 +4999,12 @@ static int send_hole(struct send_ctx *sctx, u64 end)
 	if (offset >= sctx->cur_inode_size)
 		return 0;
 
+	/*
+	 * Don't go beyond the inode's i_size due to prealloc extents that start
+	 * after the i_size.
+	 */
+	end = min_t(u64, end, sctx->cur_inode_size);
+
 	if (sctx->flags & BTRFS_SEND_FLAG_NO_FILE_DATA)
 		return send_update_extent(sctx, offset, end - offset);
 
@@ -5218,10 +5224,50 @@ static int clone_range(struct send_ctx *sctx,
 		clone_len = min_t(u64, ext_len, len);
 
 		if (btrfs_file_extent_disk_bytenr(leaf, ei) == disk_byte &&
-		    clone_data_offset == data_offset)
-			ret = send_clone(sctx, offset, clone_len, clone_root);
-		else
+		    clone_data_offset == data_offset) {
+			const u64 src_end = clone_root->offset + clone_len;
+			const u64 sectorsize = SZ_64K;
+
+			/*
+			 * We can't clone the last block, when its size is not
+			 * sector size aligned, into the middle of a file. If we
+			 * do so, the receiver will get a failure (-EINVAL) when
+			 * trying to clone or will silently corrupt the data in
+			 * the destination file if it's on a kernel without the
+			 * fix introduced by commit ac765f83f1397646
+			 * ("Btrfs: fix data corruption due to cloning of eof
+			 * block).
+			 *
+			 * So issue a clone of the aligned down range plus a
+			 * regular write for the eof block, if we hit that case.
+			 *
+			 * Also, we use the maximum possible sector size, 64K,
+			 * because we don't know what's the sector size of the
+			 * filesystem that receives the stream, so we have to
+			 * assume the largest possible sector size.
+			 */
+			if (src_end == clone_src_i_size &&
+			    !IS_ALIGNED(src_end, sectorsize) &&
+			    offset + clone_len < sctx->cur_inode_size) {
+				u64 slen;
+
+				slen = ALIGN_DOWN(src_end - clone_root->offset,
+						  sectorsize);
+				if (slen > 0) {
+					ret = send_clone(sctx, offset, slen,
+							 clone_root);
+					if (ret < 0)
+						goto out;
+				}
+				ret = send_extent_data(sctx, offset + slen,
+						       clone_len - slen);
+			} else {
+				ret = send_clone(sctx, offset, clone_len,
+						 clone_root);
+			}
+		} else {
 			ret = send_extent_data(sctx, offset, clone_len);
+		}
 
 		if (ret < 0)
 			goto out;
@@ -6276,68 +6322,21 @@ static int changed_extent(struct send_ctx *sctx,
 {
 	int ret = 0;
 
-	if (sctx->cur_ino != sctx->cmp_key->objectid) {
-
-		if (result == BTRFS_COMPARE_TREE_CHANGED) {
-			struct extent_buffer *leaf_l;
-			struct extent_buffer *leaf_r;
-			struct btrfs_file_extent_item *ei_l;
-			struct btrfs_file_extent_item *ei_r;
-
-			leaf_l = sctx->left_path->nodes[0];
-			leaf_r = sctx->right_path->nodes[0];
-			ei_l = btrfs_item_ptr(leaf_l,
-					      sctx->left_path->slots[0],
-					      struct btrfs_file_extent_item);
-			ei_r = btrfs_item_ptr(leaf_r,
-					      sctx->right_path->slots[0],
-					      struct btrfs_file_extent_item);
-
-			/*
-			 * We may have found an extent item that has changed
-			 * only its disk_bytenr field and the corresponding
-			 * inode item was not updated. This case happens due to
-			 * very specific timings during relocation when a leaf
-			 * that contains file extent items is COWed while
-			 * relocation is ongoing and its in the stage where it
-			 * updates data pointers. So when this happens we can
-			 * safely ignore it since we know it's the same extent,
-			 * but just at different logical and physical locations
-			 * (when an extent is fully replaced with a new one, we
-			 * know the generation number must have changed too,
-			 * since snapshot creation implies committing the current
-			 * transaction, and the inode item must have been updated
-			 * as well).
-			 * This replacement of the disk_bytenr happens at
-			 * relocation.c:replace_file_extents() through
-			 * relocation.c:btrfs_reloc_cow_block().
-			 */
-			if (btrfs_file_extent_generation(leaf_l, ei_l) ==
-			    btrfs_file_extent_generation(leaf_r, ei_r) &&
-			    btrfs_file_extent_ram_bytes(leaf_l, ei_l) ==
-			    btrfs_file_extent_ram_bytes(leaf_r, ei_r) &&
-			    btrfs_file_extent_compression(leaf_l, ei_l) ==
-			    btrfs_file_extent_compression(leaf_r, ei_r) &&
-			    btrfs_file_extent_encryption(leaf_l, ei_l) ==
-			    btrfs_file_extent_encryption(leaf_r, ei_r) &&
-			    btrfs_file_extent_other_encoding(leaf_l, ei_l) ==
-			    btrfs_file_extent_other_encoding(leaf_r, ei_r) &&
-			    btrfs_file_extent_type(leaf_l, ei_l) ==
-			    btrfs_file_extent_type(leaf_r, ei_r) &&
-			    btrfs_file_extent_disk_bytenr(leaf_l, ei_l) !=
-			    btrfs_file_extent_disk_bytenr(leaf_r, ei_r) &&
-			    btrfs_file_extent_disk_num_bytes(leaf_l, ei_l) ==
-			    btrfs_file_extent_disk_num_bytes(leaf_r, ei_r) &&
-			    btrfs_file_extent_offset(leaf_l, ei_l) ==
-			    btrfs_file_extent_offset(leaf_r, ei_r) &&
-			    btrfs_file_extent_num_bytes(leaf_l, ei_l) ==
-			    btrfs_file_extent_num_bytes(leaf_r, ei_r))
-				return 0;
-		}
-
-		inconsistent_snapshot_error(sctx, result, "extent");
-		return -EIO;
-	}
+	/*
+	 * We have found an extent item that changed without the inode item
+	 * having changed. This can happen either after relocation (where the
+	 * disk_bytenr of an extent item is replaced at
+	 * relocation.c:replace_file_extents()) or after deduplication into a
+	 * file in both the parent and send snapshots (where an extent item can
+	 * get modified or replaced with a new one). Note that deduplication
+	 * updates the inode item, but it only changes the iversion (sequence
+	 * field in the inode item) of the inode, so if a file is deduplicated
+	 * the same amount of times in both the parent and send snapshots, its
+	 * iversion becames the same in both snapshots, whence the inode item is
+	 * the same on both snapshots.
+	 */
+	if (sctx->cur_ino != sctx->cmp_key->objectid)
+		return 0;
 
 	if (!sctx->cur_inode_new_gen && !sctx->cur_inode_deleted) {
 		if (result != BTRFS_COMPARE_TREE_DELETED)
@@ -6883,9 +6882,23 @@ long btrfs_ioctl_send(struct file *mnt_file, struct btrfs_ioctl_send_args *arg)
 	if (ret)
 		goto out;
 
+	mutex_lock(&fs_info->balance_mutex);
+	if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
+		mutex_unlock(&fs_info->balance_mutex);
+		btrfs_warn_rl(fs_info,
+		"cannot run send because a balance operation is in progress");
+		ret = -EAGAIN;
+		goto out;
+	}
+	fs_info->send_in_progress++;
+	mutex_unlock(&fs_info->balance_mutex);
+
 	current->journal_info = BTRFS_SEND_TRANS_STUB;
 	ret = send_subvol(sctx);
 	current->journal_info = NULL;
+	mutex_lock(&fs_info->balance_mutex);
+	fs_info->send_in_progress--;
+	mutex_unlock(&fs_info->balance_mutex);
 	if (ret < 0)
 		goto out;