Similar to the changes where strerror(errno) was converted, continue
with the remaining cases where the argument was stored in another
variable.
The savings in object size are about 4500 bytes:
$ size btrfs.old btrfs.new
text data bss dec hex filename
805055 24248 19748 849051 cf49b btrfs.old
804527 24248 19748 848523 cf28b btrfs.new
Signed-off-by: David Sterba <dsterba@suse.com>
Now that all the prerequisite code for proper support of free space
tree repair is in, it's time to wire it in. This is achieved by first
hooking the freespace tree to the __free_extent/alloc_reserved_tree_block
functions. And then introducing a wrapper function to contains the
existing check_space_cache and the newly introduced repair code.
Finally, it's important to note that FST repair code first clears the
existing FST in case of any problem found and rebuilds it from scratch.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This commit enables the delayed refs infrastructures. This entails doing
the following:
1. Replacing existing calls of btrfs_extent_post_op (which is the
equivalent of delayed refs) with the proper btrfs_run_delayed_refs.
As well as eliminating open-coded calls to finish_current_insert and
del_pending_extents which execute the delayed ops.
2. Wiring up the addition of delayed refs when freeing extents
(btrfs_free_extent) and when adding new extents (alloc_tree_block).
3. Adding calls to btrfs_run_delayed refs in the transaction commit
path alongside comments why every call is needed, since it's not
always obvious (those call sites were derived empirically by running
and debugging existing tests)
4. Correctly flagging the transaction in which we are reinitialising
the extent tree.
5. Moving btrfs_write_dirty_block_groups to
btrfs_write_dirty_block_groups since blockgroups should be written to
disk after the last delayed refs have been run.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
An infinite loop can be triggered during fuzz/003:
====== RUN MAYFAIL btrfs check --repair tests/fuzz-tests/images/bko-199833-reloc-recovery-crash.raw.restored
[1/7] checking root items
Fixed 0 roots.
[2/7] checking extents
ctree.c:1650: leaf_space_used: Warning: assertion `data_len < 0` failed, value 1
bad key ordering 18 19
ctree.c:1650: leaf_space_used: Warning: assertion `data_len < 0` failed, value 1
bad key ordering 18 19
ctree.c:1650: leaf_space_used: Warning: assertion `data_len < 0` failed, value 1
bad key ordering 18 19
[CAUSE]
In try_to_fix_bad_block() it's possible that btrfs_find_all_roots()
finds no root referring to that tree block, thus we can't do any repair.
However in that case, we still return 0 since the last caller assigning
@ret is btrfs_find_all_roots(), and the ulist while loop doesn't get run
at all.
And since try_to_fix_bad_block() returns 0, check_block() in
check/main.c will return -EAGAIN to re-check the tree block.
This leads to the infinite loop.
[FIX]
Change the default return value from 0 to -EIO in
try_to_fix_bad_block(), so if there is no tree referring to the bad tree
block, it won't cause infinite loop anymore.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In check_inode_recs(), for repair mode we always reset @ret to 0. It
makes no sense and later we check @ret to determine if the repair is
successful.
Fix it by removing the offending overwrite.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Gu Jinxiang <gujx@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Exposed by fuzz-tests/003-multi-check-unmounted/ on fuzzed image
bko-161811.raw.xz.
It's caused by the fact when check_fs_roots() finds tree root is
modified, it re-search tree root by goto again: label.
However again: label. will also reset root objectid to 0.
If we failed to repair one fs root but still modified tree root, we will
go into such infinite loop.
Fix it by recording which root we should skip for repair mode.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Gu Jinxiang <gujx@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have function btrfs_fsck_reinit_root() to reinit csum or extent tree.
However this function allows us to let it overwrite existing tree blocks
using @overwrite parameter.
Such behavior is pretty dangerous while no caller is using this feature
explicitly.
So just remove @overwrite parameter and allow btrfs_fsck_reinit_root()
to error out when it fails to allocate tree block.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Gu Jinxiang <gujx@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We reuse the task_position enum and task_ctx struct of the original progress
indicator, adding more values and fields for our needs.
Then add hooks in all steps of the check to properly record progress.
Here's how the output looks like on a 22 Tb 5-disk RAID1 FS:
Opening filesystem to check...
Checking filesystem on /dev/mapper/luks-ST10000VN0004-XXXXXXXX
UUID: xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
[1/7] checking extents (0:20:21 elapsed, 950958 items checked)
[2/7] checking root items (0:01:29 elapsed, 15121 items checked)
[3/7] checking free space cache (0:00:11 elapsed, 4928 items checked)
[4/7] checking fs roots (0:51:31 elapsed, 600892 items checked)
[5/7] checking csums (0:14:35 elapsed, 754522 items checked)
[6/7] checking root refs (0:00:00 elapsed, 232 items checked)
[7/7] checking quota groups skipped (not enabled on this FS)
found 5286458060800 bytes used, no error found
Signed-off-by: Stéphane Lesimple <stephane_btrfs@lesimple.fr>
Signed-off-by: David Sterba <dsterba@suse.com>
With crafted image, expected root item can refer to certain extent, and
original mode uses BUG_ON() to handle such case.
Fix it by gracefully return error.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=200403
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
They are not really needed, what free_extent_hook wants is really a
pointer to fs_info so give it to it directly. This is in preparation
of delayed refs code.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The experimental flag is already carried in the manpage, but was removed
from the btrfs check usage message as part of refactoring via
87c1bd13c1. Add it back.
Signed-off-by: David Disseldorp <ddiss@suse.de>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It looks like that around 2014, btrfs kernel has a regression that would
cause offset-by-one ram_bytes for inline extent.
Add the ability to repair it in original mode.
Reported-by: Steve Leung <sjleung@shaw.ca>
Tested-by: Steve Leung <sjleung@shaw.ca>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
If one uncompressed inline extent has incorrect ram_bytes, neither btrfs
check nor dump-tree could detect such corruption.
[CAUSE]
Every caller tries to read inline extent ram_bytes is using
btrfs_file_extent_inline_len(), other than directly calling
btrfs_file_extent_ram_bytes().
For compressed extent, it's just calling btrfs_file_extent_ram_bytes().
However for uncompressed extent, it falls back to
btrfs_file_extent_inline_item_len(), makes us unable to detect anything
wrong in ram_bytes.
[FIX]
Just get rid of such confusing btrfs_file_extent_inline_len() function.
Reported-by: Steve Leung <sjleung@shaw.ca>
Tested-by: Steve Leung <sjleung@shaw.ca>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
check_chunks_and_extents does quite a number of distinct things. The
first of those is going through all root items in the root tree and
classify every root depending on whether they have a dropping operation
in progress or not. Lets factor out this code and move the variables
specific to this in a separate function so clean up check_chunks_and_extents
a bit. Accidentally, this patch fixes some reference leaks since
in error conditions in the loop the code does "goto out" but at that
label we don't really release the path. Having this code extracted in a
separate function which always releases the path avoids this problem
entirely.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The first thing that check_chunks_and_extents does is to iterate all
the root items in the root tree and link them to either the "normal_list"
or "dropping_trees" list. If a leaf has to be crossed during this
operation btrfs_next_leaf is called to do that. However, currently it's
called with a wrong argument for its 'root' parameter. Since we are
iterating the root tree the passed root should be fs_info->tree_rot,
whereas right now we are passing the local variable 'root' which is
assigned to the fs_tree. As it stands, this bug is actually benign since
the passed root is only passed to reada_for_search, where it's used to
reference the fs_info. Nevertheless the code is wrong and at the very least
misleading, so fix it by passing the correct root.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Define new macro I_ERR_ODD_INODE_FLAGS to represents odd inode flags.
Symlinks should never have append/immutable flags.
While processing inodes, if found a symlink with append/immutable
flags, mark the inode record with I_ERR_ODD_INODE_FLAGS.
This is for original mode.
Issue: #133
Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
That function really wants an fs_info and not a root. Accidentally,
this also makes the kernel/user space signatures to be coherent.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is no longer used by the callees of that function so remove it.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It's always set to extent_root and the function already takes a
transaction handle where fs_info could be referenced and in turn
the extent_tree.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This function is always passed the extent_root as "root" parameter. In
turn it uses the root parameter to mostly access fs_info and performs
only a single call to btrfs_update_block_group where it passses the
passed root. This is all redundant since fs_info can be referenced
from the transaction handle and in turn extent_root can be referenced
from the fs_info. So do that to simplify the function's signature.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For btrfs_leaf_free_space(), to get leaf data size, we have two way to
get it:
1) leaf->fs_info->nodesize
2) leaf->len
Anyway, we could get rid of @fs_info parameter for
btrfs_leaf_free_space().
And here we choose method 2), as it provides extra benefit to get leaf
free space without initializing a real fs_info.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Current btrfs-check will check qgroup consistency, but even when it
finds something wrong, the return value is still 0.
Fix it by allowing report_qgroups() to return int to indicate qgroup
mismatch, and also add extra logic to return no error if qgroup repair
is successful.
Without this patch, fstests can't detect qgroup corruption by its fsck
alone.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This parameter was introduced with the original implementation of the
function but has never really been used, so just remove it.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
If options '--init-extent-tree' and '--mode=lowmem' are both
input, all metadata blocks will be traversed twice.
First one is done by pin_metadata_blocks() in reinit_extent_tree().
Second one is in check_chunks_and_extents_v2().
Excluding instead of pinning metadata blocks before reinitializing th
extent tree in lowmem can save some time.
Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Move pin_down_tree_blocks from main.c to mode-common.c for
further patches.
And export pin_metadata_blocks to mode-common.h.
Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
No matter --check-data-csum is passed or not, btrfs check will always
output message like "checking csum".
This message could be a little confusing, change it according to
--check-data-csum option.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since data csum mismatch is not a fatal error compared to fs/extent
trees, continue check.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When --check-data-csum option found csum mismatch, btrfs check still
return 0.
Fix it so log-replay could automatically pause when it finds csum error.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Original --check-data-csum option will skip the extra copy if the first
copy matches csum.
Since offline scrub (with recoverability report) is still out-of-tree, at
least enhance --check-data-csum option to handle multiple copies.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Instead of struct btrfs_root, use struct btrfs_fs_info, since nodesize
is now a per-fs setting, and with the need to pass a @root, caller don't
need to wonder which root should be passed.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For metadata dump (fs restored by btrfs-image), no data is restored
and check sum verification will definitely report error.
Add such check in check_csums() and prompt for user input.
Issue: #103
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This @first_key variable is introduced in f5c4c4f3b7
("btrfsck: add code to rebuild extent records"), however it's not only
unused, but also used incorrectly.
It's calling btrfs_item_key_to_cpu() on an node extent buffer.
Anyway, just remove it.
Fixes: f5c4c4f3b7 ("btrfsck: add code to rebuild extent records")
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For inline compressed file extent, kernel doesn't allow inline extent
ram size larger than sector size and on-disk inline extent size should
not exceed BTRFS_MAX_INLINE_DATA_SIZE().
For inline uncompressed file extent, kernel doesn't allow inline extent
ram and on-disk size larger than either BTRFS_MAX_INLINE_DATA_SIZE() or
sector size.
Check it in original mode.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Under some cases the filesystem checker reports an error when it finds
checksum items for an extent that is referenced by an inode as a prealloc
extent. Such cases are not an error when the extent is actually shared
(was cloned/reflinked) with other inodes and was written through one of
those other inodes.
Example:
$ mkfs.btrfs -f /dev/sdb
$ mount /dev/sdb /mnt
$ touch /mnt/foo
$ xfs_io -c "falloc 0 256K" /mnt/foo
$ sync
$ xfs_io -c "pwrite -S 0xab 0 256K" /mnt/foo
$ touch /mnt/bar
$ xfs_io -c "reflink /mnt/foo 0 0 256K" /mnt/bar
$ xfs_io -c "fsync" /mnt/bar
<power fail>
$ mount /dev/sdb /mnt
$ umount /mnt
$ btrfs check /dev/sdc
Checking filesystem on /dev/sdb
UUID: 52d3006e-ee3b-40eb-aa21-e56253a03d39
checking extents
checking free space cache
checking fs roots
root 5 inode 257 errors 800, odd csum item
ERROR: errors found in fs roots
found 688128 bytes used, error(s) found
total csum bytes: 256
total tree bytes: 163840
total fs tree bytes: 65536
total extent tree bytes: 16384
btree space waste bytes: 138819
file data blocks allocated: 10747904
referenced 10747904
$ echo $?
1
So teach check to not report such cases as errors by checking if the
extent is shared with other inodes and if so, consider it an error the
existence of checksum items only if all those other inodes are referencing
the extent as a prealloc extent.
This case can be hit often when running the generic/475 testcase from
fstests.
A test case will follow in a separate patch.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When we found free space difference between free space cache and block
group item, we just discard this free space cache.
Normally such difference is caused by btrfs_reserve_extent() called by
delalloc which is out of a transaction.
And since all btrfs_release_extent() is called with a transaction, under
heavy race free space cache can have less free space than block group
item.
Normally kernel will detect such difference and just discard that cache.
However we must be more careful if free space cache has more free space
cache, and if that happens, paried with above race one invalid free
space cache can be loaded into kernel.
So if we find any free space cache who has more free space then block
group item, we report it as an error other than ignoring it.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This bug is exposed by fsck-test with D=asan, hit by test case 020, with
the following error report:
=================================================================
==10740==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x621000061580 at pc 0x56051f0db6cd bp 0x7ffe170f3e20 sp 0x7ffe170f3e10
READ of size 1 at 0x621000061580 thread T0
#0 0x56051f0db6cc in btrfs_extent_inline_ref_type /home/adam/btrfs/btrfs-progs/ctree.h:1727
#1 0x56051f13b669 in build_roots_info_cache /home/adam/btrfs/btrfs-progs/cmds-check.c:14306
#2 0x56051f13c86a in repair_root_items /home/adam/btrfs/btrfs-progs/cmds-check.c:14450
#3 0x56051f13ea89 in cmd_check /home/adam/btrfs/btrfs-progs/cmds-check.c:14965
#4 0x56051efe75bb in main /home/adam/btrfs/btrfs-progs/btrfs.c:302
#5 0x7f04ddbb0f49 in __libc_start_main (/usr/lib/libc.so.6+0x20f49)
#6 0x56051efe68c9 in _start (/home/adam/btrfs/btrfs-progs/btrfs+0x5b8c9)
0x621000061580 is located 0 bytes to the right of 4224-byte region [0x621000060500,0x621000061580)
allocated by thread T0 here:
#0 0x7f04ded50ce1 in __interceptor_calloc /build/gcc/src/gcc/libsanitizer/asan/asan_malloc_linux.cc:70
#1 0x56051f04685e in __alloc_extent_buffer /home/adam/btrfs/btrfs-progs/extent_io.c:553
#2 0x56051f047563 in alloc_extent_buffer /home/adam/btrfs/btrfs-progs/extent_io.c:687
#3 0x56051efff1d1 in btrfs_find_create_tree_block /home/adam/btrfs/btrfs-progs/disk-io.c:187
#4 0x56051f000133 in read_tree_block /home/adam/btrfs/btrfs-progs/disk-io.c:327
#5 0x56051efeddb8 in read_node_slot /home/adam/btrfs/btrfs-progs/ctree.c:652
#6 0x56051effb0d9 in btrfs_next_leaf /home/adam/btrfs/btrfs-progs/ctree.c:2853
#7 0x56051f13b343 in build_roots_info_cache /home/adam/btrfs/btrfs-progs/cmds-check.c:14267
#8 0x56051f13c86a in repair_root_items /home/adam/btrfs/btrfs-progs/cmds-check.c:14450
#9 0x56051f13ea89 in cmd_check /home/adam/btrfs/btrfs-progs/cmds-check.c:14965
#10 0x56051efe75bb in main /home/adam/btrfs/btrfs-progs/btrfs.c:302
#11 0x7f04ddbb0f49 in __libc_start_main (/usr/lib/libc.so.6+0x20f49)
It's completely possible that one extent/metadata item has no inline
reference, while build_roots_info_cache() doesn't have such check.
Fix it by checking @iref against item end to avoid such problem.
Issue: #92
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since we're moving tons of code, it's a good idea to fix all errors and
warnings from checkpatch.
Note: we don't rely on checkpatch, take its warnings and errors as a
hint and do not blindly fix anything that gets reported. The code
follows some style that should be kept and is not in 100% with
what checkpatch finds. Always apply your common sense and decide
if the change makes sense or if it is really making the code
better.
Additional fixes from dsterba:
- %Lu -> %llu
- lowercase first word or messages
Signed-off-by: Qu Wenruo <wqu@suse.com>
[ added note ]
Signed-off-by: David Sterba <dsterba@suse.com>
There used to be some functions with _v2 suffix to distinguish them from
original mode that had similar functions.
However now when the lowmem mode is moved to own check/lowmem.[ch],
cleanup the _v2 suffixes, and for functions that really need to be
distinguished from original mode (exported functions), change the _v2
suffix to _lowmem.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since lowmem mode code is highly internally connected, it's pretty hard to
move them piece by piece.
In theory it's possible to move part of the functions and temporarily
export them, but it will just cause extra temporarily modifications.
So this patch moves the whole lowmem check part into its own
check/lowmem.[ch].
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Both original and lowmem modes share this function to do readahead.
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Josh Soref