The nid and ino of an inode are the same. So during building
nat_area_bitmap, we can know which nat entry represents an
inode, thus, we can count the number of inodes.
Then in fsck_chk_meta, the amount of inodes can be compared
with that recorded in CP.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch introduces a new option '-p' to do more checks on NAT/SIT areas.
'-p' has 2 levels: level 1 has the same sematics as '-a'; level 2 checks
NAT/SIT counters to see if they matches the status in SB and CP.
A new function, fsck_chk_meta, is called by '-p 1' to implement these
comparsion. If errors are detected, fix_on is set, which means fsck will
do a 'fsck -f' immediately.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Commit 4ea4f1db ("fsck.f2fs: large volume support") added the support
for large volume over about 3TB. The "cp_payload" is added to start_blk,
but not removed from orphan_blkaddr.
Fixes: 4ea4f1db ("fsck.f2fs: large volume support")
Signed-off-by: Junling Zheng <zhengjunling@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
There are some issues in original approach:
- memory allocated in i_extent.map will leak
- do not convert fields in i_ext from on-disk format to cpu format
- do not support checking file offset with extent info
This patch refactors the flow for fixing above issues and supporting
file offset check.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch enables fsck.f2fs to detect incorrect position where dirent
locates in an hierarchical hash structure directory.
Signed-off-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Commit 843f5b9388 ("fsck.f2fs: check ino of an inode") checks
nat_entry->ino for inode. However, the checking is irrelevant to node
type, and if ino of other kinds of nodes gets corrupted, fsck.f2fs
cannot detect it. This patch fixes the commit.
Fixes: 843f5b9388 ("fsck.f2fs: check ino of an inode")
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Check valid values for '.' and '..', such as ino, hash_code, and the
number of dot or dotdot in the directory.
Signed-off-by: Xue Liu <liuxueliu.liu@huawei.com>
Signed-off-by: Yong Sheng <shengyong1@huawei.com>
Reviewed-by: Chao Yu <chao2.yu@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
For indirect/double-indirect node, if the blk_addr in its nat entry is
corrupted, fsck could figure this out and nullify the nat entry. However,
the indirect/double-indirect node still keeps the corrupted nid. As a
result, fsck reports valid blkaddr but fixes nothing each time during
scanning all nodes, like:
=============================
[ASSERT] (sanity_check_nid: 356) --> blkaddres is not valid. [0x0]
delete in.nid[i] = 0;
[ASSERT] (sanity_check_nid: 356) --> blkaddres is not valid. [0x0]
delete in.nid[i] = 0;
[FSCK] Unreachable nat entries [Ok..] [0x0]
[FSCK] SIT valid block bitmap checking [Ok..]
[FSCK] Hard link checking for regular file [Ok..] [0x0]
[FSCK] valid_block_count matching with CP [Ok..] [0xa51b]
[FSCK] valid_node_count matcing with CP (de lookup) [Ok..] [0x98]
[FSCK] valid_node_count matcing with CP (nat lookup) [Ok..] [0x98]
[FSCK] valid_inode_count matched with CP [Ok..] [0x7b]
[FSCK] free segment_count matched with CP [Ok..] [0x1a]
[FSCK] next block offset is free [Ok..]
[FSCK] fixing SIT types
[FSCK] other corrupted bugs [Fail]
=============================
So let's clean it.
Reported-by: Chen Yinchao <chengyinchao@huawei.com>
Signed-off-by: Liu Xue <liuxueliu.liu@huawei.com>
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If the node is an inode, its ino in the nat entry should be equal to the
ino in its node footer. Otherwise, we drop the file.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Signed-off-by: Xue Liu <liuxueliu.liu@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If a nid is valid, the ino in its nat entry should record its parent nid.
However if the ino is zero, we should drop the node.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Signed-off-by: Xue Liu <liuxueliu.liu@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In some case, we need confirm node or data blkaddr is ok at first,
and then fix the conflict in ssa block. If the node blkaddr happen to
locate in data type segment caused by something error, may be we 'd
better ignore it, and it's same with data blkaddr.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Deal with realloc failure to avoid memory leak and memory free,
and assert realloc size to avoid double free for tree_mark.
Signed-off-by: Xue Liu <liuxueliu.liu@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This patch checks alloc_type of current segment type.
If it is LFS, the last of segment should have no valid block.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Deal with realloc failure to avoid memory leak and memory free.
Signed-off-by: Xue Liu <liuxueliu.liu@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, check_sit_types didn't handle different numbers of active logs and
SSR cases. But, it didn't cause any problem since fixing sit types is harmless.
Nevertheless, we still don't need to change them all.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If a directory has no dot and dotdot dentries, fsck.f2fs sets inline_dots for
the inode so that f2fs module can handle that properly.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If a child is a directory, we should increase parent's i_links.
Previously, it counts direct dentry blocks excluding indirect blocks.
This patch fixes to count child_cnt correctly in order to get precise i_links.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If some of children including directories are corrupted, we should not include
them into i_links of their parent.
This patch fixes that.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, if data and node summary was corrupted, the block was deallocated.
But, this patch fixes their summary first, so that we can keep their blocks.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If fsck.f2fs found some orphan blocks, previously fsck.f2fs drops them and
fixes checkpoint, resulting in false alarm on corruption.
This patch preserves orphan blocks to avoid such the alarms.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If i_links is wrong, we should not check block count, since it doesn't count
correctly which results in changing the block count to 1.
This patch fixes that.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Each inode has its filename inside inode block.
For directory, the name should be matched all the time.
In other cases, it shouldn't, since there is able to be linked to other file
names.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Previously, only data blocks locating in its inode block could be fixed.
This patch investigates direct node blocks to fix indices too.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
One of the cases would update the dentries but not set
fixed=1 causing the update to not be written out.
Change-Id: I111d599f93f7d5306dacd7d62da946fdb5c262ee
Signed-off-by: JP Abgrall <jpa@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
This is a followup from 056e4b04fc
Now the blk is marked as fixed so that the callers can write it back.
Change-Id: I99c94bc39e772e0d0baab81bd741ce55416ce0e6
Signed-off-by: JP Abgrall <jpa@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
