Skip to contentSkip to conversation

btrfs-progs: map-logical: remove the extent check

PATCH In review Qu Wenruo 2 participants 7 messages Aug 12, 2021 Btrfs file system

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

cover btrfs-progs: map-logical: remove the extent check

Commit eb81f8d263ce ("btrfs-progs: map-logical: Rework map-logical logics") introduced a strict extent item check for btrfs-map-logical.

It not only makes it harder to use btrfs-map-logical, but also prevents btrfs-map-logical to handle fs with corrupted extent tree.

This patchset will address the behavior by completely removing the extent item check, and allow btrfs-map-logical to work on fs with corrupted extent tree.

Since we're here, also change the default mapping length to sectorsize, and reject any unaligned logical/bytes pair.

Qu Wenruo (4):
btrfs-progs: map-logical: use sectorsize as default size
btrfs-progs: map-logical: reject unaligned logical/bytes pair
btrfs-progs: map-logical: loosen the required trees to open the
filesystem
btrfs-progs: map-logical: remove the extent item check

 btrfs-map-logical.c | 158 +++++---------------------------------------
 1 file changed, 17 insertions(+), 141 deletions(-)
Signature · 1 line
2.32.0
author
#1 lore raw

PATCH 1/4 btrfs-progs: map-logical: use sectorsize as default size

Originally btrfs-map-logical is utilizing function debug_read_block(), which pretents to read an extent buffer from disk, thus it uses nodesize as default mapping size.

But after commit eb81f8d263ce ("btrfs-progs: map-logical: Rework map-logical logics") it no longer requires an extent buffer to do the mapping.

Thus the default mapping size is no longer needed to be nodesize.

Thus change the default size to sectorsize.

---
 btrfs-map-logical.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
btrfs-map-logical.c +1 -1
index b35677730374..eff0b89dbec6 100644
@@ -288,7 +288,7 @@ int main(int argc, char **argv)
 	}
 
 	if (bytes == 0)
-		bytes = root->fs_info->nodesize;
+		bytes = root->fs_info->sectorsize;
 	cur_logical = logical;
 	cur_len = bytes;
 
Signature · 1 line
2.32.0
author
#2 lore raw

PATCH 2/4 btrfs-progs: map-logical: reject unaligned logical/bytes pair

In reply to Qu Wenruo

Btrfs filesystem is a block filesystem, there is no sense to support unaligned logical/bytes pair.

---
 btrfs-map-logical.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
btrfs-map-logical.c +10 -0
index eff0b89dbec6..21f00fa20ce8 100644
@@ -289,6 +289,16 @@ int main(int argc, char **argv)
 
 	if (bytes == 0)
 		bytes = root->fs_info->sectorsize;
+
+	if (!IS_ALIGNED(logical, root->fs_info->sectorsize) ||
+	    !IS_ALIGNED(bytes, root->fs_info->sectorsize)) {
+		ret = -EINVAL;
+		error(
+	"invalid logical/bytes, both need to aligned to %u, have %llu and %llu",
+			root->fs_info->sectorsize, logical, bytes);
+		goto close;
+	}
+
 	cur_logical = logical;
 	cur_len = bytes;
 
Signature · 1 line
2.32.0
#3 lore raw

Re: [PATCH 2/4] btrfs-progs: map-logical: reject unaligned logical/bytes pair

Quoted 13 lines · Qu Wenruo

Btrfs filesystem is a block filesystem, there is no sense to support unaligned logical/bytes pair.

---
 btrfs-map-logical.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
btrfs-map-logical.c +0 -0
index eff0b89dbec6..21f00fa20ce8 100644

btrfs-map-logical.c int main(int argc, char **argv) · line 298

Quoted 10 lines
btrfs-map-logical.c +7 -0
@@ -289,6 +289,16 @@ int main(int argc, char **argv)
 
 	if (bytes == 0)
 		bytes = root->fs_info->sectorsize;
+
+	if (!IS_ALIGNED(logical, root->fs_info->sectorsize) ||
+	    !IS_ALIGNED(bytes, root->fs_info->sectorsize)) {
+		ret = -EINVAL;
+		error(
+	"invalid logical/bytes, both need to aligned to %u, have %llu and %llu",
+			root->fs_info->sectorsize, logical, bytes);

In order to be more graceful I'd say print this as a warning and do the alignment automatically, informing the user this has been performed.

Quoted 7 lines
btrfs-map-logical.c +3 -0
+		goto close;
+	}
+
 	cur_logical = logical;
 	cur_len = bytes;
 
author
#4 lore raw

PATCH 3/4 btrfs-progs: map-logical: loosen the required trees to open the filesystem

In reply to Qu Wenruo

For btrfs-map-logical, it only requires chunk tree to do the logicla->physical mapping.

All othe trees are not really needed.

Loosen the required trees to make btrfs-map-logical to work better on corrupted fs.

---
 btrfs-map-logical.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
btrfs-map-logical.c +2 -1
index 21f00fa20ce8..9f119d08bad8 100644
@@ -261,7 +261,8 @@ int main(int argc, char **argv)
 	radix_tree_init();
 	cache_tree_init(&root_cache);
 
-	root = open_ctree(dev, 0, 0);
+	root = open_ctree(dev, 0, OPEN_CTREE_PARTIAL |
+				  OPEN_CTREE_NO_BLOCK_GROUPS);
 	if (!root) {
 		fprintf(stderr, "Open ctree failed\n");
 		free(output_file);
Signature · 1 line
2.32.0
#5 lore raw

Re: [PATCH 3/4] btrfs-progs: map-logical: loosen the required trees to open the filesystem

Quoted 18 lines · Qu Wenruo

For btrfs-map-logical, it only requires chunk tree to do the logicla->physical mapping.

All othe trees are not really needed.

Loosen the required trees to make btrfs-map-logical to work better on corrupted fs.

---
 btrfs-map-logical.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
btrfs-map-logical.c +0 -0
index 21f00fa20ce8..9f119d08bad8 100644

btrfs-map-logical.c int main(int argc, char **argv) · line 265

Quoted 6 lines
btrfs-map-logical.c +2 -1
@@ -261,7 +261,8 @@ int main(int argc, char **argv)
 	radix_tree_init();
 	cache_tree_init(&root_cache);
 
-	root = open_ctree(dev, 0, 0);
+	root = open_ctree(dev, 0, OPEN_CTREE_PARTIAL |
+				  OPEN_CTREE_NO_BLOCK_GROUPS);

nit: OPEN_CTREE_PARTIAL implies ignoring error when opening the csum and extent tress, since those trees are opened via setup_root_or_create_block. This tells me that OPEN_CTREE_NO_BLOCK_GROUPS is redundant and can be removed and replaced with simply OPEN_CTREE_PARTIAL.

Quoted 4 lines
btrfs-map-logical.c +0 -0
 	if (!root) {
 		fprintf(stderr, "Open ctree failed\n");
 		free(output_file);
author
#6 lore raw

PATCH 4/4 btrfs-progs: map-logical: remove the extent item check

In reply to Qu Wenruo

Tool btrfs-map-logical is really only doing logical -> physical mapping, mostly utilized by developers or experienced users.

There is really no need to check whether the specified range has an extent or not.

In fact the extent check behavior is a big blockage for corrupted fs as such fs can have corrupted extent tree, doing an extent item search can lead to -EIO error.

This patch will just remove the extent item check, allowing users to do any logical -> physical mapping lookup as long as there is a chunk.

---
 btrfs-map-logical.c | 145 ++------------------------------------------
 1 file changed, 5 insertions(+), 140 deletions(-)
btrfs-map-logical.c +5 -140
index 9f119d08bad8..96263013a5c6 100644
@@ -38,65 +38,6 @@
  * */
 static FILE *info_file;
 
-static int map_one_extent(struct btrfs_fs_info *fs_info,
-			  u64 *logical_ret, u64 *len_ret, int search_forward)
-{
-	struct btrfs_path *path;
-	struct btrfs_key key;
-	u64 logical;
-	u64 len = 0;
-	int ret = 0;
-
-	BUG_ON(!logical_ret);
-	logical = *logical_ret;
-
-	path = btrfs_alloc_path();
-	if (!path)
-		return -ENOMEM;
-
-	key.objectid = logical;
-	key.type = 0;
-	key.offset = 0;
-
-	ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path,
-				0, 0);
-	if (ret < 0)
-		goto out;
-	BUG_ON(ret == 0);
-	ret = 0;
-
-again:
-	btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
-	if ((search_forward && key.objectid < logical) ||
-	    (!search_forward && key.objectid > logical) ||
-	    (key.type != BTRFS_EXTENT_ITEM_KEY &&
-	     key.type != BTRFS_METADATA_ITEM_KEY)) {
-		if (!search_forward)
-			ret = btrfs_previous_extent_item(fs_info->extent_root,
-							 path, 0);
-		else
-			ret = btrfs_next_extent_item(fs_info->extent_root,
-						     path, 0);
-		if (ret)
-			goto out;
-		goto again;
-	}
-	logical = key.objectid;
-	if (key.type == BTRFS_METADATA_ITEM_KEY)
-		len = fs_info->nodesize;
-	else
-		len = key.offset;
-
-out:
-	btrfs_free_path(path);
-	if (!ret) {
-		*logical_ret = logical;
-		if (len_ret)
-			*len_ret = len;
-	}
-	return ret;
-}
-
 static int __print_mapping_info(struct btrfs_fs_info *fs_info, u64 logical,
 				u64 len, int mirror_num)
 {
@@ -134,16 +75,6 @@ static int __print_mapping_info(struct btrfs_fs_info *fs_info, u64 logical,
 	return ret;
 }
 
-/*
- * Logical and len is the exact value of a extent.
- * And offset is the offset inside the extent. It's only used for case
- * where user only want to print part of the extent.
- *
- * Caller *MUST* ensure the range [logical,logical+len) are in one extent.
- * Or we can encounter the following case, causing a -ENOENT error:
- * |<-----given parameter------>|
- *		|<------ Extent A ----->|
- */
 static int print_mapping_info(struct btrfs_fs_info *fs_info, u64 logical,
 			      u64 len)
 {
@@ -213,10 +144,7 @@ int main(int argc, char **argv)
 	u64 copy = 0;
 	u64 logical = 0;
 	u64 bytes = 0;
-	u64 cur_logical = 0;
-	u64 cur_len = 0;
 	int out_fd = -1;
-	int found = 0;
 	int ret = 0;
 
 	while(1) {
@@ -300,76 +228,13 @@ int main(int argc, char **argv)
 		goto close;
 	}
 
-	cur_logical = logical;
-	cur_len = bytes;
-
-	/* First find the nearest extent */
-	ret = map_one_extent(root->fs_info, &cur_logical, &cur_len, 0);
-	if (ret < 0) {
-		errno = -ret;
-		fprintf(stderr, "Failed to find extent at [%llu,%llu): %m\n",
-			cur_logical, cur_logical + cur_len);
+	ret = print_mapping_info(root->fs_info, logical, bytes);
+	if (ret < 0)
 		goto out_close_fd;
-	}
-	/*
-	 * Normally, search backward should be OK, but for special case like
-	 * given logical is quite small where no extents are before it,
-	 * we need to search forward.
-	 */
-	if (ret > 0) {
-		ret = map_one_extent(root->fs_info, &cur_logical, &cur_len, 1);
-		if (ret < 0) {
-			errno = -ret;
-			fprintf(stderr,
-				"Failed to find extent at [%llu,%llu): %m\n",
-				cur_logical, cur_logical + cur_len);
-			goto out_close_fd;
-		}
-		if (ret > 0) {
-			fprintf(stderr,
-				"Failed to find any extent at [%llu,%llu)\n",
-				cur_logical, cur_logical + cur_len);
-			goto out_close_fd;
-		}
-	}
-
-	while (cur_logical + cur_len >= logical && cur_logical < logical +
-	       bytes) {
-		u64 real_logical;
-		u64 real_len;
+	if (output_file && out_fd != -1)
+		ret = write_extent_content(root->fs_info, out_fd, logical,
+					   bytes, copy);
 
-		found = 1;
-		ret = map_one_extent(root->fs_info, &cur_logical, &cur_len, 1);
-		if (ret < 0)
-			goto out_close_fd;
-		if (ret > 0)
-			break;
-		/* check again if there is overlap. */
-		if (cur_logical + cur_len < logical ||
-		    cur_logical >= logical + bytes)
-			break;
-
-		real_logical = max(logical, cur_logical);
-		real_len = min(logical + bytes, cur_logical + cur_len) -
-			   real_logical;
-
-		ret = print_mapping_info(root->fs_info, real_logical, real_len);
-		if (ret < 0)
-			goto out_close_fd;
-		if (output_file && out_fd != -1) {
-			ret = write_extent_content(root->fs_info, out_fd,
-					real_logical, real_len, copy);
-			if (ret < 0)
-				goto out_close_fd;
-		}
-
-		cur_logical += cur_len;
-	}
-
-	if (!found) {
-		fprintf(stderr, "No extent found at range [%llu,%llu)\n",
-			logical, logical + bytes);
-	}
 out_close_fd:
 	if (output_file && out_fd != 1)
 		close(out_fd);
Signature · 1 line
2.32.0