[Ocfs2-tools-devel] [PATCH] fsck.ocfs2(1.2): Modify e_clusters if it isn't equal to eb contains.

[Tao Ma]

In bug 7553939, we meet with a scenario that the clusters haven't
been allocated to the extent block(no update) while the inode already
have the i_clusters and e_clusters updated. The commit
09166ad03c83f39f2dcd77c74cd4672c3a371dfa has fixed the problem
of i_clusters corruption, so this one intends to fix e_clusters.
We calculate the total clusters within the extent block and then
modify the extent record if it doesn't show the right value.

Please note that to be safe, it only fix the corruption happens in the
last extent block. If there is any corruption happens in other extent block,
we don't touch it(actually if we modify it, we may leave a hole there and
make the system corrupted). The good thing is that in ocfs2-1.2, we don't
support sparse files, and we only extend files, so if there is one corruption
like 7553939, it must happen in the last extent block.

One more thing, in ocfs2-1.4, we support sparse files, so it can be handled
gracefully. So this is only a fix for ocfs2-tools-1.2.

Signed-off-by: Tao Ma <tao.ma at oracle.com>
---
 fsck.ocfs2/extent.c               |   36 ++++++++++++++++++++++++++++--------
 fsck.ocfs2/fsck.ocfs2.checks.8.in |    7 +++++++
 2 files changed, 35 insertions(+), 8 deletions(-)

diff --git a/fsck.ocfs2/extent.c b/fsck.ocfs2/extent.c
index 0690408..d4b5364 100644
--- a/fsck.ocfs2/extent.c
+++ b/fsck.ocfs2/extent.c
@@ -57,11 +57,12 @@ struct extent_info {
 static errcode_t check_el(o2fsck_state *ost, struct extent_info *ei,
 			  struct ocfs2_dinode *di,
 			  struct ocfs2_extent_list *el,
-			  uint16_t max_recs, int *changed);
+			  uint16_t max_recs, int *changed,
+			  uint32_t *total_clusters);
 
 static errcode_t check_eb(o2fsck_state *ost, struct extent_info *ei,
 			  struct ocfs2_dinode *di, uint64_t blkno,
-			  int *is_valid)
+			  int *is_valid, uint32_t *total_clusters)
 {
 	int changed = 0;
 	char *buf = NULL;
@@ -128,7 +129,7 @@ static errcode_t check_eb(o2fsck_state *ost, struct extent_info *ei,
 
 	check_el(ost, ei, di, &eb->h_list,
 		 ocfs2_extent_recs_per_eb(ost->ost_fs->fs_blocksize), 
-		 &changed);
+		 &changed, total_clusters);
 
 	if (changed) {
 		ret = ocfs2_write_extent_block(ost->ost_fs, blkno, buf);
@@ -155,7 +156,7 @@ static errcode_t check_er(o2fsck_state *ost, struct extent_info *ei,
 {
 	errcode_t ret = 0;
 	uint64_t first_block;
-	uint32_t last_cluster;
+	uint32_t last_cluster, total_clusters;
 
 	verbosef("cpos %u clusters %u blkno %"PRIu64"\n", er->e_cpos,
 		 er->e_clusters, er->e_blkno);
@@ -170,7 +171,9 @@ static errcode_t check_er(o2fsck_state *ost, struct extent_info *ei,
 		 * is checked */
 		ei->ei_expect_depth = 1;
 		ei->ei_expected_depth = el->l_tree_depth - 1;
-		check_eb(ost, ei, di, er->e_blkno, &is_valid);
+		ret = 0;
+		total_clusters = 0;
+		check_eb(ost, ei, di, er->e_blkno, &is_valid, &total_clusters);
 		if (!is_valid && 
 		    prompt(ost, PY, PR_EXTENT_EB_INVALID,
 			   "The extent record for cluster offset "
@@ -181,8 +184,22 @@ static errcode_t check_er(o2fsck_state *ost, struct extent_info *ei,
 
 			er->e_blkno = 0;
 			*changed = 1;
+			goto out;
 		}
-		ret = 0;
+
+		if (el->l_recs[el->l_next_free_rec-1].e_blkno == er->e_blkno &&
+		    total_clusters != er->e_clusters &&
+		    prompt(ost, PY, PR_EXTENT_EB_CLUSTER_RANGE,
+			   "The extent record for cluster offset %"PRIu32" "
+			   "in inode %"PRIu64" has a cluster count of %"PRIu32
+			   " while its extent block has %"PRIu32" clusters "
+			   "allocated. Change it to what extent block "
+			   "represents?", er->e_cpos, di->i_blkno,
+			   er->e_clusters, total_clusters)) {
+			er->e_clusters = total_clusters;
+			*changed = 1;
+		}
+
 		goto out;
 	}
 
@@ -230,7 +247,8 @@ out:
 static errcode_t check_el(o2fsck_state *ost, struct extent_info *ei,
 			  struct ocfs2_dinode *di,
 			  struct ocfs2_extent_list *el,
-			  uint16_t max_recs, int *changed)
+			  uint16_t max_recs, int *changed,
+			  uint32_t *total_clusters)
 {
 	int trust_next_free = 1;
 	struct ocfs2_extent_rec *er;
@@ -319,6 +337,8 @@ static errcode_t check_el(o2fsck_state *ost, struct extent_info *ei,
 			continue;
 		}
 
+		if (total_clusters)
+			*total_clusters += er->e_clusters;
 
 		/* we've already accounted for the extent block as part of
 		 * the extent block chain groups */
@@ -350,7 +370,7 @@ errcode_t o2fsck_check_extents(o2fsck_state *ost,
 	
 	ret = check_el(ost, &ei, di, &di->id2.i_list, 
 	         ocfs2_extent_recs_per_inode(ost->ost_fs->fs_blocksize),
-		 &changed);
+		 &changed, NULL);
 
 	if (changed)
 		o2fsck_write_inode(ost, di->i_blkno, di);
diff --git a/fsck.ocfs2/fsck.ocfs2.checks.8.in b/fsck.ocfs2/fsck.ocfs2.checks.8.in
index 2e30f9d..8551018 100644
--- a/fsck.ocfs2/fsck.ocfs2.checks.8.in
+++ b/fsck.ocfs2/fsck.ocfs2.checks.8.in
@@ -71,6 +71,13 @@ extent tree references an invalid extent block.
 Answering yes stops the tree from referencing the invalid extent block.  This
 may truncate data from the file which contains the tree.
 
+.SS "EXTENT_EB_CLUSTER_RANGE"
+Deep extent trees are built by forming a tree out of extent blocks.  An
+extent record represent a wrong clusters count than the extent block contains.
+
+Answering yes change clusters count in the extent record to the number the below
+extent block contains.
+
 .SS "EXTENT_LIST_DEPTH"
 Extent lists contain a record of their depth in the tree.  An extent list
 was found whose recorded depth doesn't match the position they have in the
-- 
1.5.5