[PATCH 1/2] btrfs-progs: find-root: Output matched root when searching all roots

[PATCH 1/2] btrfs-progs: find-root: Output matched root when searching all roots

[Qu Wenruo]

[Bug]
When given '-a' option, btrfs-find-root will output all possible tree
roots but the exact matched one.

[Reason]
Result printing skipes the exact match one, as it will normally be shown
before the alternative ones.
But when '-a' is given, that's not the case.

[Fix]
Just show the exact match one for search all case.

Reported-by: Marc Merlin <marc@merlins.org>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
 btrfs-find-root.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/btrfs-find-root.c b/btrfs-find-root.c
index 1cb3085..01b3603 100644
--- a/btrfs-find-root.c
+++ b/btrfs-find-root.c
@@ -109,6 +109,9 @@ static void print_one_result(struct cache_extent *tree_block,
 	       tree_block->start, generation, level);
 	if (unsure)
 		printf("but we are unsure about the correct generation/level\n");
+	else if (level == filter->match_level &&
+		 generation == filter->match_gen)
+		printf("and it matches superblock\n");
 	else
 		printf("but generation/level doesn't match, want gen: %llu level: %u\n",
 		       filter->match_gen, filter->match_level);
@@ -129,8 +132,10 @@ static void print_find_root_result(struct cache_tree *result,
 				struct btrfs_find_root_gen_cache, cache);
 		level = gen_cache->highest_level;
 		generation = cache->start;
+		/* For exact found one, skip it as it's output before */
 		if (level == filter->match_level &&
-		    generation == filter->match_gen)
+		    generation == filter->match_gen &&
+		    !filter->search_all)
 			continue;
 		for (tree_block = last_cache_extent(&gen_cache->eb_tree);
 		     tree_block; tree_block = prev_cache_extent(tree_block))
-- 
2.5.0

[PATCH 2/2] btrfs-progs: find-root: Fix a bug that will cause wrong result

[Qu Wenruo]

[BUG]
btrfs-find-root may not output desire result, as due to
search_extent_cache() may return result which doesn't cover the desired
range, generation cache can be screwed up if higher generation tree root
is found before lower generation tree root.

For example:
=======
./btrfs-find-root  /dev/sda6  -a
Superblock thinks the generation is 8
Superblock thinks the level is 0
adding bytenr: 4194304, gen: 8	<<< Debug output
adding bytenr: 24715264, gen: 7 <<< gen is 7 at read_tree_block time
Well block 4194304(gen: 8 level: 0) seems good, and it matches
superblock
Well block 24715264(gen: 8 level: 0) seems good, and it matches
superblock  <<< But its gen is wrong at result output time
=======

[Fix]
Add a new judgment to make sure the search_extent_cache() is returning
desired result.

Reported-by: Marc Merlin <marc@merlins.org>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
 find-root.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/find-root.c b/find-root.c
index 1af37b5..0eca0ab 100644
--- a/find-root.c
+++ b/find-root.c
@@ -46,9 +46,14 @@ static int add_eb_to_result(struct extent_buffer *eb,
 	    generation < filter->generation)
 		return ret;
 
-	/* Get the generation cache or create one */
+	/*
+	 * Get the generation cache or create one
+	 *
+	 * NOTE: search_cache_extent() may return cache doesn't cover
+	 * the range. So need extra judgement to make sure it's a match.
+	 */
 	cache = search_cache_extent(result, generation);
-	if (!cache) {
+	if (!cache || cache->start != generation) {
 		gen_cache = malloc(sizeof(*gen_cache));
 		BUG_ON(!gen_cache);
 		cache = &gen_cache->cache;
-- 
2.5.0

Re: [PATCH 1/2] btrfs-progs: find-root: Output matched root when searching all roots

[David Sterba]

On Tue, Aug 25, 2015 at 10:44:56AM +0800, Qu Wenruo wrote:
> [Bug]
> When given '-a' option, btrfs-find-root will output all possible tree
> roots but the exact matched one.
> 
> [Reason]
> Result printing skipes the exact match one, as it will normally be shown
> before the alternative ones.
> But when '-a' is given, that's not the case.
> 
> [Fix]
> Just show the exact match one for search all case.
> 
> Reported-by: Marc Merlin <marc@merlins.org>
> Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>

Applied, thanks.

Re: [PATCH 2/2] btrfs-progs: find-root: Fix a bug that will cause wrong result

[David Sterba]

On Tue, Aug 25, 2015 at 10:44:57AM +0800, Qu Wenruo wrote:
> [BUG]
> btrfs-find-root may not output desire result, as due to
> search_extent_cache() may return result which doesn't cover the desired
> range, generation cache can be screwed up if higher generation tree root
> is found before lower generation tree root.
> 
> For example:
> =======
> ./btrfs-find-root  /dev/sda6  -a
> Superblock thinks the generation is 8
> Superblock thinks the level is 0
> adding bytenr: 4194304, gen: 8	<<< Debug output
> adding bytenr: 24715264, gen: 7 <<< gen is 7 at read_tree_block time
> Well block 4194304(gen: 8 level: 0) seems good, and it matches
> superblock
> Well block 24715264(gen: 8 level: 0) seems good, and it matches
> superblock  <<< But its gen is wrong at result output time
> =======
> 
> [Fix]
> Add a new judgment to make sure the search_extent_cache() is returning
> desired result.
> 
> Reported-by: Marc Merlin <marc@merlins.org>
> Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>

Applied, thanks.