[BUG]
The following script will cause false alert on devid check.
#!/bin/bash
dev1=/dev/test/test
dev2=/dev/test/scratch1
mnt=/mnt/btrfs
umount $dev1 &> /dev/null
umount $dev2 &> /dev/null
umount $mnt &> /dev/null
mkfs.btrfs -f $dev1
mount $dev1 $mnt
_fail()
{
echo "!!! FAILED !!!"
exit 1
}
for ((i = 0; i < 4096; i++)); do
btrfs dev add -f $dev2 $mnt || _fail
btrfs dev del $dev1 $mnt || _fail
dev_tmp=$dev1
dev1=$dev2
dev2=$dev_tmp
done
[CAUSE]
Tree-checker uses BTRFS_MAX_DEVS() and BTRFS_MAX_DEVS_SYS_CHUNK() as
upper limit for devid. But we can have devid holes just like above
script.
So the check for devid is incorrect and could cause false alert.
[FIX]
Just remove the whole devid check. We don't have any hard requirement
for devid assignment.
Furthermore, even devid could get corrupted by a bitflip, we still have
dev extents verification at mount time, so corrupted data won't sneak
in.
This fixes fstests btrfs/194.
Reported-by: Anand Jain <anand.jain@oracle.com>
Fixes: ab4ba2e133 ("btrfs: tree-checker: Verify dev item")
CC: stable@vger.kernel.org # 5.2+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For SYSTEM chunks, despite the regular chunk item size limit, there is
another limit due to system chunk array size.
The extra limit was removed in a refactoring, so add it back.
Fixes: e3ecdb3fde ("btrfs: factor out devs_max setting in __btrfs_alloc_chunk")
CC: stable@vger.kernel.org # 5.3+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
For btrfs:qgroup_meta_reserve event, the trace event can output garbage:
qgroup_meta_reserve: 9c7f6acc-b342-4037-bc47-7f6e4d2232d7: refroot=5(FS_TREE) type=DATA diff=2
qgroup_meta_reserve: 9c7f6acc-b342-4037-bc47-7f6e4d2232d7: refroot=5(FS_TREE) type=0x258792 diff=2
The @type can be completely garbage, as DATA type is not possible for
trace_qgroup_meta_reserve() trace event.
[CAUSE]
Ther are several problems related to qgroup trace events:
- Unassigned entry member
Member entry::type of trace_qgroup_update_reserve() and
trace_qgourp_meta_reserve() is not assigned
- Redundant entry member
Member entry::type is completely useless in
trace_qgroup_meta_convert()
Fixes: 4ee0d8832c ("btrfs: qgroup: Update trace events for metadata reservation")
CC: stable@vger.kernel.org # 4.10+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
For btrfs:qgroup_meta_reserve event, the trace event can output garbage:
qgroup_meta_reserve: 9c7f6acc-b342-4037-bc47-7f6e4d2232d7: refroot=5(FS_TREE) type=DATA diff=2
The diff should always be alinged to sector size (4k), so there is
definitely something wrong.
[CAUSE]
For the wrong @diff, it's caused by wrong parameter order.
The correct parameters are:
struct btrfs_root, s64 diff, int type.
However the parameters used are:
struct btrfs_root, int type, s64 diff.
Fixes: 4ee0d8832c ("btrfs: qgroup: Update trace events for metadata reservation")
CC: stable@vger.kernel.org # 4.19+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[Background]
Btrfs qgroup uses two types of reserved space for METADATA space,
PERTRANS and PREALLOC.
PERTRANS is metadata space reserved for each transaction started by
btrfs_start_transaction().
While PREALLOC is for delalloc, where we reserve space before joining a
transaction, and finally it will be converted to PERTRANS after the
writeback is done.
[Inconsistency]
However there is inconsistency in how we handle PREALLOC metadata space.
The most obvious one is:
In btrfs_buffered_write():
btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes, true);
We always free qgroup PREALLOC meta space.
While in btrfs_truncate_block():
btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize, (ret != 0));
We only free qgroup PREALLOC meta space when something went wrong.
[The Correct Behavior]
The correct behavior should be the one in btrfs_buffered_write(), we
should always free PREALLOC metadata space.
The reason is, the btrfs_delalloc_* mechanism works by:
- Reserve metadata first, even it's not necessary
In btrfs_delalloc_reserve_metadata()
- Free the unused metadata space
Normally in:
btrfs_delalloc_release_extents()
|- btrfs_inode_rsv_release()
Here we do calculation on whether we should release or not.
E.g. for 64K buffered write, the metadata rsv works like:
/* The first page */
reserve_meta: num_bytes=calc_inode_reservations()
free_meta: num_bytes=0
total: num_bytes=calc_inode_reservations()
/* The first page caused one outstanding extent, thus needs metadata
rsv */
/* The 2nd page */
reserve_meta: num_bytes=calc_inode_reservations()
free_meta: num_bytes=calc_inode_reservations()
total: not changed
/* The 2nd page doesn't cause new outstanding extent, needs no new meta
rsv, so we free what we have reserved */
/* The 3rd~16th pages */
reserve_meta: num_bytes=calc_inode_reservations()
free_meta: num_bytes=calc_inode_reservations()
total: not changed (still space for one outstanding extent)
This means, if btrfs_delalloc_release_extents() determines to free some
space, then those space should be freed NOW.
So for qgroup, we should call btrfs_qgroup_free_meta_prealloc() other
than btrfs_qgroup_convert_reserved_meta().
The good news is:
- The callers are not that hot
The hottest caller is in btrfs_buffered_write(), which is already
fixed by commit 336a8bb8e3 ("btrfs: Fix wrong
btrfs_delalloc_release_extents parameter"). Thus it's not that
easy to cause false EDQUOT.
- The trans commit in advance for qgroup would hide the bug
Since commit f5fef45936 ("btrfs: qgroup: Make qgroup async transaction
commit more aggressive"), when btrfs qgroup metadata free space is slow,
it will try to commit transaction and free the wrongly converted
PERTRANS space, so it's not that easy to hit such bug.
[FIX]
So to fix the problem, remove the @qgroup_free parameter for
btrfs_delalloc_release_extents(), and always pass true to
btrfs_inode_rsv_release().
Reported-by: Filipe Manana <fdmanana@suse.com>
Fixes: 43b18595d6 ("btrfs: qgroup: Use separate meta reservation type for delalloc")
CC: stable@vger.kernel.org # 4.19+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In btrfs_read_block_groups(), if we have an invalid block group which
has mixed type (DATA|METADATA) while the fs doesn't have MIXED_GROUPS
feature, we error out without freeing the block group cache.
This patch will add the missing btrfs_put_block_group() to prevent
memory leak.
Note for stable backports: the file to patch in versions <= 5.3 is
fs/btrfs/extent-tree.c
Fixes: 49303381f1 ("Btrfs: bail out if block group has different mixed flag")
CC: stable@vger.kernel.org # 4.9+
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
The following script can cause btrfs qgroup data space leak:
mkfs.btrfs -f $dev
mount $dev -o nospace_cache $mnt
btrfs subv create $mnt/subv
btrfs quota en $mnt
btrfs quota rescan -w $mnt
btrfs qgroup limit 128m $mnt/subv
for (( i = 0; i < 3; i++)); do
# Create 3 64M holes for latter fallocate to fail
truncate -s 192m $mnt/subv/file
xfs_io -c "pwrite 64m 4k" $mnt/subv/file > /dev/null
xfs_io -c "pwrite 128m 4k" $mnt/subv/file > /dev/null
sync
# it's supposed to fail, and each failure will leak at least 64M
# data space
xfs_io -f -c "falloc 0 192m" $mnt/subv/file &> /dev/null
rm $mnt/subv/file
sync
done
# Shouldn't fail after we removed the file
xfs_io -f -c "falloc 0 64m" $mnt/subv/file
[CAUSE]
Btrfs qgroup data reserve code allow multiple reservations to happen on
a single extent_changeset:
E.g:
btrfs_qgroup_reserve_data(inode, &data_reserved, 0, SZ_1M);
btrfs_qgroup_reserve_data(inode, &data_reserved, SZ_1M, SZ_2M);
btrfs_qgroup_reserve_data(inode, &data_reserved, 0, SZ_4M);
Btrfs qgroup code has its internal tracking to make sure we don't
double-reserve in above example.
The only pattern utilizing this feature is in the main while loop of
btrfs_fallocate() function.
However btrfs_qgroup_reserve_data()'s error handling has a bug in that
on error it clears all ranges in the io_tree with EXTENT_QGROUP_RESERVED
flag but doesn't free previously reserved bytes.
This bug has a two fold effect:
- Clearing EXTENT_QGROUP_RESERVED ranges
This is the correct behavior, but it prevents
btrfs_qgroup_check_reserved_leak() to catch the leakage as the
detector is purely EXTENT_QGROUP_RESERVED flag based.
- Leak the previously reserved data bytes.
The bug manifests when N calls to btrfs_qgroup_reserve_data are made and
the last one fails, leaking space reserved in the previous ones.
[FIX]
Also free previously reserved data bytes when btrfs_qgroup_reserve_data
fails.
Fixes: 5247255370 ("btrfs: qgroup: Introduce btrfs_qgroup_reserve_data function")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
Under the following case with qgroup enabled, if some error happened
after we have reserved delalloc space, then in error handling path, we
could cause qgroup data space leakage:
From btrfs_truncate_block() in inode.c:
ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
block_start, blocksize);
if (ret)
goto out;
again:
page = find_or_create_page(mapping, index, mask);
if (!page) {
btrfs_delalloc_release_space(inode, data_reserved,
block_start, blocksize, true);
btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize, true);
ret = -ENOMEM;
goto out;
}
[CAUSE]
In the above case, btrfs_delalloc_reserve_space() will call
btrfs_qgroup_reserve_data() and mark the io_tree range with
EXTENT_QGROUP_RESERVED flag.
In the error handling path, we have the following call stack:
btrfs_delalloc_release_space()
|- btrfs_free_reserved_data_space()
|- btrsf_qgroup_free_data()
|- __btrfs_qgroup_release_data(reserved=@reserved, free=1)
|- qgroup_free_reserved_data(reserved=@reserved)
|- clear_record_extent_bits();
|- freed += changeset.bytes_changed;
However due to a completion bug, qgroup_free_reserved_data() will clear
EXTENT_QGROUP_RESERVED flag in BTRFS_I(inode)->io_failure_tree, other
than the correct BTRFS_I(inode)->io_tree.
Since io_failure_tree is never marked with that flag,
btrfs_qgroup_free_data() will not free any data reserved space at all,
causing a leakage.
This type of error handling can only be triggered by errors outside of
qgroup code. So EDQUOT error from qgroup can't trigger it.
[FIX]
Fix the wrong target io_tree.
Reported-by: Josef Bacik <josef@toxicpanda.com>
Fixes: bc42bda223 ("btrfs: qgroup: Fix qgroup reserved space underflow by only freeing reserved ranges")
CC: stable@vger.kernel.org # 4.14+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
With v5.3 kernel, we can't convert to SINGLE profile:
# btrfs balance start -f -dconvert=single $mnt
ERROR: error during balancing '/mnt/btrfs': Invalid argument
# dmesg -t | tail
validate_convert_profile: data profile=0x1000000000000 allowed=0x20 is_valid=1 final=0x1000000000000 ret=1
BTRFS error (device dm-3): balance: invalid convert data profile single
[CAUSE]
With the extra debug output added, it shows that the @allowed bit is
lacking the special in-memory only SINGLE profile bit.
Thus we fail at that (profile & ~allowed) check.
This regression is caused by commit 081db89b13 ("btrfs: use raid_attr
to get allowed profiles for balance conversion") and the fact that we
don't use any bit to indicate SINGLE profile on-disk, but uses special
in-memory only bit to help distinguish different profiles.
[FIX]
Add that BTRFS_AVAIL_ALLOC_BIT_SINGLE to @allowed, so the code should be
the same as it was and fix the regression.
Reported-by: Chris Murphy <lists@colorremedies.com>
Fixes: 081db89b13 ("btrfs: use raid_attr to get allowed profiles for balance conversion")
CC: stable@vger.kernel.org # 5.3+
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
One user reported a reproducible KASAN report about use-after-free:
BTRFS info (device sdi1): balance: start -dvrange=1256811659264..1256811659265
BTRFS info (device sdi1): relocating block group 1256811659264 flags data|raid0
==================================================================
BUG: KASAN: use-after-free in btrfs_init_reloc_root+0x2cd/0x340 [btrfs]
Write of size 8 at addr ffff88856f671710 by task kworker/u24:10/261579
CPU: 2 PID: 261579 Comm: kworker/u24:10 Tainted: P OE 5.2.11-arch1-1-kasan #4
Hardware name: To Be Filled By O.E.M. To Be Filled By O.E.M./X99 Extreme4, BIOS P3.80 04/06/2018
Workqueue: btrfs-endio-write btrfs_endio_write_helper [btrfs]
Call Trace:
dump_stack+0x7b/0xba
print_address_description+0x6c/0x22e
? btrfs_init_reloc_root+0x2cd/0x340 [btrfs]
__kasan_report.cold+0x1b/0x3b
? btrfs_init_reloc_root+0x2cd/0x340 [btrfs]
kasan_report+0x12/0x17
__asan_report_store8_noabort+0x17/0x20
btrfs_init_reloc_root+0x2cd/0x340 [btrfs]
record_root_in_trans+0x2a0/0x370 [btrfs]
btrfs_record_root_in_trans+0xf4/0x140 [btrfs]
start_transaction+0x1ab/0xe90 [btrfs]
btrfs_join_transaction+0x1d/0x20 [btrfs]
btrfs_finish_ordered_io+0x7bf/0x18a0 [btrfs]
? lock_repin_lock+0x400/0x400
? __kmem_cache_shutdown.cold+0x140/0x1ad
? btrfs_unlink_subvol+0x9b0/0x9b0 [btrfs]
finish_ordered_fn+0x15/0x20 [btrfs]
normal_work_helper+0x1bd/0xca0 [btrfs]
? process_one_work+0x819/0x1720
? kasan_check_read+0x11/0x20
btrfs_endio_write_helper+0x12/0x20 [btrfs]
process_one_work+0x8c9/0x1720
? pwq_dec_nr_in_flight+0x2f0/0x2f0
? worker_thread+0x1d9/0x1030
worker_thread+0x98/0x1030
kthread+0x2bb/0x3b0
? process_one_work+0x1720/0x1720
? kthread_park+0x120/0x120
ret_from_fork+0x35/0x40
Allocated by task 369692:
__kasan_kmalloc.part.0+0x44/0xc0
__kasan_kmalloc.constprop.0+0xba/0xc0
kasan_kmalloc+0x9/0x10
kmem_cache_alloc_trace+0x138/0x260
btrfs_read_tree_root+0x92/0x360 [btrfs]
btrfs_read_fs_root+0x10/0xb0 [btrfs]
create_reloc_root+0x47d/0xa10 [btrfs]
btrfs_init_reloc_root+0x1e2/0x340 [btrfs]
record_root_in_trans+0x2a0/0x370 [btrfs]
btrfs_record_root_in_trans+0xf4/0x140 [btrfs]
start_transaction+0x1ab/0xe90 [btrfs]
btrfs_start_transaction+0x1e/0x20 [btrfs]
__btrfs_prealloc_file_range+0x1c2/0xa00 [btrfs]
btrfs_prealloc_file_range+0x13/0x20 [btrfs]
prealloc_file_extent_cluster+0x29f/0x570 [btrfs]
relocate_file_extent_cluster+0x193/0xc30 [btrfs]
relocate_data_extent+0x1f8/0x490 [btrfs]
relocate_block_group+0x600/0x1060 [btrfs]
btrfs_relocate_block_group+0x3a0/0xa00 [btrfs]
btrfs_relocate_chunk+0x9e/0x180 [btrfs]
btrfs_balance+0x14e4/0x2fc0 [btrfs]
btrfs_ioctl_balance+0x47f/0x640 [btrfs]
btrfs_ioctl+0x119d/0x8380 [btrfs]
do_vfs_ioctl+0x9f5/0x1060
ksys_ioctl+0x67/0x90
__x64_sys_ioctl+0x73/0xb0
do_syscall_64+0xa5/0x370
entry_SYSCALL_64_after_hwframe+0x44/0xa9
Freed by task 369692:
__kasan_slab_free+0x14f/0x210
kasan_slab_free+0xe/0x10
kfree+0xd8/0x270
btrfs_drop_snapshot+0x154c/0x1eb0 [btrfs]
clean_dirty_subvols+0x227/0x340 [btrfs]
relocate_block_group+0x972/0x1060 [btrfs]
btrfs_relocate_block_group+0x3a0/0xa00 [btrfs]
btrfs_relocate_chunk+0x9e/0x180 [btrfs]
btrfs_balance+0x14e4/0x2fc0 [btrfs]
btrfs_ioctl_balance+0x47f/0x640 [btrfs]
btrfs_ioctl+0x119d/0x8380 [btrfs]
do_vfs_ioctl+0x9f5/0x1060
ksys_ioctl+0x67/0x90
__x64_sys_ioctl+0x73/0xb0
do_syscall_64+0xa5/0x370
entry_SYSCALL_64_after_hwframe+0x44/0xa9
The buggy address belongs to the object at ffff88856f671100
which belongs to the cache kmalloc-4k of size 4096
The buggy address is located 1552 bytes inside of
4096-byte region [ffff88856f671100, ffff88856f672100)
The buggy address belongs to the page:
page:ffffea0015bd9c00 refcount:1 mapcount:0 mapping:ffff88864400e600 index:0x0 compound_mapcount: 0
flags: 0x2ffff0000010200(slab|head)
raw: 02ffff0000010200 dead000000000100 dead000000000200 ffff88864400e600
raw: 0000000000000000 0000000000070007 00000001ffffffff 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffff88856f671600: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff88856f671680: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
>ffff88856f671700: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
^
ffff88856f671780: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
ffff88856f671800: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================
BTRFS info (device sdi1): 1 enospc errors during balance
BTRFS info (device sdi1): balance: ended with status: -28
[CAUSE]
The problem happens when finish_ordered_io() get called with balance
still running, while the reloc root of that subvolume is already dead.
(Tree is swap already done, but tree not yet deleted for possible qgroup
usage.)
That means root->reloc_root still exists, but that reloc_root can be
under btrfs_drop_snapshot(), thus we shouldn't access it.
The following race could cause the use-after-free problem:
CPU1 | CPU2
--------------------------------------------------------------------------
| relocate_block_group()
| |- unset_reloc_control(rc)
| |- btrfs_commit_transaction()
btrfs_finish_ordered_io() | |- clean_dirty_subvols()
|- btrfs_join_transaction() | |
|- record_root_in_trans() | |
|- btrfs_init_reloc_root() | |
|- if (root->reloc_root) | |
| | |- root->reloc_root = NULL
| | |- btrfs_drop_snapshot(reloc_root);
|- reloc_root->last_trans|
= trans->transid |
^^^^^^^^^^^^^^^^^^^^^^
Use after free
[FIX]
Fix it by the following modifications:
- Test if the root has dead reloc tree before accessing root->reloc_root
If the root has BTRFS_ROOT_DEAD_RELOC_TREE, then we don't need to
create or update root->reloc_tree
- Clear the BTRFS_ROOT_DEAD_RELOC_TREE flag until we have fully dropped
reloc tree
To co-operate with above modification, so as long as
BTRFS_ROOT_DEAD_RELOC_TREE is still set, we won't try to re-create
reloc tree at record_root_in_trans().
Reported-by: Cebtenzzre <cebtenzzre@gmail.com>
Fixes: d2311e6985 ("btrfs: relocation: Delay reloc tree deletion after merge_reloc_roots")
CC: stable@vger.kernel.org # 5.1+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
EXTENT_DATA_REF is a little like DIR_ITEM which contains hash in its
key->offset.
This patch will check the following contents:
- Key->objectid
Basic alignment check.
- Hash
Hash of each extent_data_ref item must match key->offset.
- Offset
Basic alignment check.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For TREE_BLOCK_REF, SHARED_DATA_REF and SHARED_BLOCK_REF we need to
check:
| TREE_BLOCK_REF | SHARED_BLOCK_REF | SHARED_BLOCK_REF
--------------+----------------+-----------------+------------------
key->objectid | Alignment | Alignment | Alignment
key->offset | Any value | Alignment | Alignment
item_size | 0 | 0 | sizeof(le32) (*)
*: sizeof(struct btrfs_shared_data_ref)
So introduce a check to check all these 3 key types together.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This patch introduces the ability to check extent items.
This check involves:
- key->objectid check
Basic alignment check.
- key->type check
Against btrfs_extent_item::type and SKINNY_METADATA feature.
- key->offset alignment check for EXTENT_ITEM
- key->offset check for METADATA_ITEM
- item size check
Both against minimal size and stepping check.
- btrfs_extent_item check
Checks its flags and generation.
- btrfs_extent_inline_ref checks
Against 4 types inline ref.
Checks bytenr alignment and tree level.
- btrfs_extent_item::refs check
Check against total refs found in inline refs.
This check would be the most complex single item check due to its nature
of inlined items.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When we try to delete qgroups, we're pretty cautious, we make sure both
qgroups exist and there is a relationship between them, then try to
delete the relation.
This behavior is OK, but the problem is we need to two relation items,
and if we failed the first item deletion, we error out, leaving the
other relation item in qgroup tree.
Sometimes the error from del_qgroup_relation_item() could just be
-ENOENT, thus we can ignore that error and continue without any problem.
Further more, such cautious behavior makes qgroup relation deletion
impossible for orphan relation items.
This patch will enhance __del_qgroup_relation():
- If both qgroups and their relation items exist
Go the regular deletion routine and update their accounting if needed.
- If any qgroup or relation item doesn't exist
Then we still try to delete the orphan items anyway, but don't trigger
the accounting update.
By this, we try our best to remove relation items, and can handle orphan
relation items properly, while still keep the existing behavior for good
qgroup tree.
Reported-by: Andrei Borzenkov <arvidjaar@gmail.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This patch will introduce ROOT_ITEM check, which includes:
- Key->objectid and key->offset check
Currently only some easy check, e.g. 0 as rootid is invalid.
- Item size check
Root item size is fixed.
- Generation checks
Generation, generation_v2 and last_snapshot should not be greater than
super generation + 1
- Level and alignment check
Level should be in [0, 7], and bytenr must be aligned to sector size.
- Flags check
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=203261
Reported-by: Jungyeon Yoon <jungyeon.yoon@gmail.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
With fuzzed image and MIXED_GROUPS super flag, we can hit the following
BUG_ON():
kernel BUG at fs/btrfs/delayed-ref.c:491!
invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
CPU: 0 PID: 1849 Comm: sync Tainted: G O 5.2.0-custom #27
RIP: 0010:update_existing_head_ref.cold+0x44/0x46 [btrfs]
Call Trace:
add_delayed_ref_head+0x20c/0x2d0 [btrfs]
btrfs_add_delayed_tree_ref+0x1fc/0x490 [btrfs]
btrfs_free_tree_block+0x123/0x380 [btrfs]
__btrfs_cow_block+0x435/0x500 [btrfs]
btrfs_cow_block+0x110/0x240 [btrfs]
btrfs_search_slot+0x230/0xa00 [btrfs]
? __lock_acquire+0x105e/0x1e20
btrfs_insert_empty_items+0x67/0xc0 [btrfs]
alloc_reserved_file_extent+0x9e/0x340 [btrfs]
__btrfs_run_delayed_refs+0x78e/0x1240 [btrfs]
? kvm_clock_read+0x18/0x30
? __sched_clock_gtod_offset+0x21/0x50
btrfs_run_delayed_refs.part.0+0x4e/0x180 [btrfs]
btrfs_run_delayed_refs+0x23/0x30 [btrfs]
btrfs_commit_transaction+0x53/0x9f0 [btrfs]
btrfs_sync_fs+0x7c/0x1c0 [btrfs]
? __ia32_sys_fdatasync+0x20/0x20
sync_fs_one_sb+0x23/0x30
iterate_supers+0x95/0x100
ksys_sync+0x62/0xb0
__ia32_sys_sync+0xe/0x20
do_syscall_64+0x65/0x240
entry_SYSCALL_64_after_hwframe+0x49/0xbe
[CAUSE]
This situation is caused by several factors:
- Fuzzed image
The extent tree of this fs missed one backref for extent tree root.
So we can allocated space from that slot.
- MIXED_BG feature
Super block has MIXED_BG flag.
- No mixed block groups exists
All block groups are just regular ones.
This makes data space_info->block_groups[] contains metadata block
groups. And when we reserve space for data, we can use space in
metadata block group.
Then we hit the following file operations:
- fallocate
We need to allocate data extents.
find_free_extent() choose to use the metadata block to allocate space
from, and choose the space of extent tree root, since its backref is
missing.
This generate one delayed ref head with is_data = 1.
- extent tree update
We need to update extent tree at run_delayed_ref time.
This generate one delayed ref head with is_data = 0, for the same
bytenr of old extent tree root.
Then we trigger the BUG_ON().
[FIX]
The quick fix here is to check block_group->flags before using it.
The problem can only happen for MIXED_GROUPS fs. Regular filesystems
won't have space_info with DATA|METADATA flag, and no way to hit the
bug.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=203255
Reported-by: Jungyeon Yoon <jungyeon.yoon@gmail.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is one report of fuzzed image which leads to BUG_ON() in
btrfs_delete_delayed_dir_index().
Although that fuzzed image can already be addressed by enhanced
extent-tree error handler, it's still better to hunt down more BUG_ON().
This patch will hunt down two BUG_ON()s in
btrfs_delete_delayed_dir_index():
- One for error from btrfs_delayed_item_reserve_metadata()
Instead of BUG_ON(), we output an error message and free the item.
And return the error.
All callers of this function handles the error by aborting current
trasaction.
- One for possible EEXIST from __btrfs_add_delayed_deletion_item()
That function can return -EEXIST.
We already have a good enough error message for that, only need to
clean up the reserved metadata space and allocated item.
To help above cleanup, also modifiy __btrfs_remove_delayed_item() called
in btrfs_release_delayed_item(), to skip unassociated item.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=203253
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
Test case btrfs/156 fails since commit 302167c50b ("btrfs: don't end
the transaction for delayed refs in throttle") with ENOSPC.
[CAUSE]
The ENOSPC is reported from btrfs_can_relocate().
This function will check:
- If this block group is empty, we can relocate
- If we can enough free space, we can relocate
Above checks are valid but the following check is vague due to its
implementation:
- If and only if we can allocated a new block group to contain all the
used space, we can relocate
This design itself is OK, but the way to determine if we can allocate a
new block group is problematic.
btrfs_can_relocate() uses find_free_dev_extent() to find free space on a
device.
However find_free_dev_extent() only searches commit root and excludes
dev extents allocated in current trans, this makes it unable to use dev
extent just freed in current transaction.
So for the following example, btrfs_can_relocate() will report ENOSPC:
The example block group layout:
1M 129M 257M 385M 513M 550M
|///////|///////////|//////////| | |
// = Used bg, consider all bg is 100% used for easy calculation.
And all block groups are SINGLE, on-disk bytenr is the same as the
logical bytenr.
1) Bg in [129M, 257M) get relocated to [385M, 513M), transid=100
1M 129M 257M 385M 513M 550M
|///////| |//////////|/////////|
In transid 100, bg in [129M, 257M) get relocated to [385M, 513M)
However transid 100 is not committed yet, so in dev commit tree, we
still have the old dev extents layout:
1M 129M 257M 385M 513M 550M
|///////|///////////|//////////| | |
2) Try to relocate bg [257M, 385M)
We goes into btrfs_can_relocate(), no free space in current bgs, so we
check if we can find large enough free dev extents.
The first slot is [385M, 513M), but that is already used by new bg at
[385M, 513M), so we continue search.
The remaining slot is [512M, 550M), smaller than the bg's length 128M.
So btrfs_can_relocate report ENOSPC.
However this is over killed, in fact if we just skip btrfs_can_relocate()
check, and go into regular relocation routine, at extent reservation time,
if we can't find free extent, then we fallback to commit transaction,
which will free up the dev extents and allow new block group to be created.
[FIX]
The fix here is to remove btrfs_can_relocate() completely.
If we hit the false ENOSPC case just like btrfs/156, extent allocator
will push harder by committing transaction and we will have space for
new block group, avoiding the false ENOSPC.
If we really ran out of space, we will hit ENOSPC at
relocate_block_group(), and btrfs will just reports the ENOSPC error as
usual.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
inc_block_group_ro() is only designed to mark one block group read-only,
it doesn't really care if other block groups have enough free space to
contain the used space in the block group.
However due to the close connection between this function and
relocation, sometimes we can be confused and think this function is
responsible for balance space reservation, which is not true.
Add some comment to make the functionality clear.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since commit 6df9a95e63 ("Btrfs: make the chunk allocator completely
tree lockless") we search commit root of device tree to avoid deadlock.
This introduced a safety feature, find_free_dev_extent_start() won't
use dev extents which just get freed in current transaction.
This safety feature makes sure we won't allocate new block group using
just freed dev extents to break CoW.
However, this feature also makes find_free_dev_extent_start() not
reliable reporting free device space. Just add such comment to make
later viewer careful about this behavior.
This behavior makes one caller, btrfs_can_relocate() unreliable
determining the device free space.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This function is only used locally in find_free_dev_extent(), no
external callers.
So unexport it.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Normally the range->len is set to default value (U64_MAX), but when it's
not default value, we should check if the range overflows.
And if it overflows, return -EINVAL before doing anything.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
As btrfs(5) specified:
Note
If nodatacow or nodatasum are enabled, compression is disabled.
If NODATASUM or NODATACOW set, we should not compress the extent.
Normally NODATACOW is detected properly in run_delalloc_range() so
compression won't happen for NODATACOW.
However for NODATASUM we don't have any check, and it can cause
compressed extent without csum pretty easily, just by:
mkfs.btrfs -f $dev
mount $dev $mnt -o nodatasum
touch $mnt/foobar
mount -o remount,datasum,compress $mnt
xfs_io -f -c "pwrite 0 128K" $mnt/foobar
And in fact, we have a bug report about corrupted compressed extent
without proper data checksum so even RAID1 can't recover the corruption.
(https://bugzilla.kernel.org/show_bug.cgi?id=199707)
Running compression without proper checksum could cause more damage when
corruption happens, as compressed data could make the whole extent
unreadable, so there is no need to allow compression for
NODATACSUM.
The fix will refactor the inode compression check into two parts:
- inode_can_compress()
As the hard requirement, checked at btrfs_run_delalloc_range(), so no
compression will happen for NODATASUM inode at all.
- inode_need_compress()
As the soft requirement, checked at btrfs_run_delalloc_range() and
compress_file_range().
Reported-by: James Harvey <jamespharvey20@gmail.com>
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
dm-log-writes records the sequence of completion, not submission, thus
for the following sequence (W=write, C=complete):
Wa,Wb,Wc,Cc,Ca,FLUSH,FUAd,Cb,CFLUSH,CFUAd
The logged results in log device should be:
c,a,b,flush,fua
Fix the comment to give a better example.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
[BUG]
Lockdep will report the following circular locking dependency:
WARNING: possible circular locking dependency detected
5.2.0-rc2-custom #24 Tainted: G O
------------------------------------------------------
btrfs/8631 is trying to acquire lock:
000000002536438c (&fs_info->qgroup_ioctl_lock#2){+.+.}, at: btrfs_qgroup_inherit+0x40/0x620 [btrfs]
but task is already holding lock:
000000003d52cc23 (&fs_info->tree_log_mutex){+.+.}, at: create_pending_snapshot+0x8b6/0xe60 [btrfs]
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #2 (&fs_info->tree_log_mutex){+.+.}:
__mutex_lock+0x76/0x940
mutex_lock_nested+0x1b/0x20
btrfs_commit_transaction+0x475/0xa00 [btrfs]
btrfs_commit_super+0x71/0x80 [btrfs]
close_ctree+0x2bd/0x320 [btrfs]
btrfs_put_super+0x15/0x20 [btrfs]
generic_shutdown_super+0x72/0x110
kill_anon_super+0x18/0x30
btrfs_kill_super+0x16/0xa0 [btrfs]
deactivate_locked_super+0x3a/0x80
deactivate_super+0x51/0x60
cleanup_mnt+0x3f/0x80
__cleanup_mnt+0x12/0x20
task_work_run+0x94/0xb0
exit_to_usermode_loop+0xd8/0xe0
do_syscall_64+0x210/0x240
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #1 (&fs_info->reloc_mutex){+.+.}:
__mutex_lock+0x76/0x940
mutex_lock_nested+0x1b/0x20
btrfs_commit_transaction+0x40d/0xa00 [btrfs]
btrfs_quota_enable+0x2da/0x730 [btrfs]
btrfs_ioctl+0x2691/0x2b40 [btrfs]
do_vfs_ioctl+0xa9/0x6d0
ksys_ioctl+0x67/0x90
__x64_sys_ioctl+0x1a/0x20
do_syscall_64+0x65/0x240
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #0 (&fs_info->qgroup_ioctl_lock#2){+.+.}:
lock_acquire+0xa7/0x190
__mutex_lock+0x76/0x940
mutex_lock_nested+0x1b/0x20
btrfs_qgroup_inherit+0x40/0x620 [btrfs]
create_pending_snapshot+0x9d7/0xe60 [btrfs]
create_pending_snapshots+0x94/0xb0 [btrfs]
btrfs_commit_transaction+0x415/0xa00 [btrfs]
btrfs_mksubvol+0x496/0x4e0 [btrfs]
btrfs_ioctl_snap_create_transid+0x174/0x180 [btrfs]
btrfs_ioctl_snap_create_v2+0x11c/0x180 [btrfs]
btrfs_ioctl+0xa90/0x2b40 [btrfs]
do_vfs_ioctl+0xa9/0x6d0
ksys_ioctl+0x67/0x90
__x64_sys_ioctl+0x1a/0x20
do_syscall_64+0x65/0x240
entry_SYSCALL_64_after_hwframe+0x49/0xbe
other info that might help us debug this:
Chain exists of:
&fs_info->qgroup_ioctl_lock#2 --> &fs_info->reloc_mutex --> &fs_info->tree_log_mutex
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&fs_info->tree_log_mutex);
lock(&fs_info->reloc_mutex);
lock(&fs_info->tree_log_mutex);
lock(&fs_info->qgroup_ioctl_lock#2);
*** DEADLOCK ***
6 locks held by btrfs/8631:
#0: 00000000ed8f23f6 (sb_writers#12){.+.+}, at: mnt_want_write_file+0x28/0x60
#1: 000000009fb1597a (&type->i_mutex_dir_key#10/1){+.+.}, at: btrfs_mksubvol+0x70/0x4e0 [btrfs]
#2: 0000000088c5ad88 (&fs_info->subvol_sem){++++}, at: btrfs_mksubvol+0x128/0x4e0 [btrfs]
#3: 000000009606fc3e (sb_internal#2){.+.+}, at: start_transaction+0x37a/0x520 [btrfs]
#4: 00000000f82bbdf5 (&fs_info->reloc_mutex){+.+.}, at: btrfs_commit_transaction+0x40d/0xa00 [btrfs]
#5: 000000003d52cc23 (&fs_info->tree_log_mutex){+.+.}, at: create_pending_snapshot+0x8b6/0xe60 [btrfs]
[CAUSE]
Due to the delayed subvolume creation, we need to call
btrfs_qgroup_inherit() inside commit transaction code, with a lot of
other mutex hold.
This hell of lock chain can lead to above problem.
[FIX]
On the other hand, we don't really need to hold qgroup_ioctl_lock if
we're in the context of create_pending_snapshot().
As in that context, we're the only one being able to modify qgroup.
All other qgroup functions which needs qgroup_ioctl_lock are either
holding a transaction handle, or will start a new transaction:
Functions will start a new transaction():
* btrfs_quota_enable()
* btrfs_quota_disable()
Functions hold a transaction handler:
* btrfs_add_qgroup_relation()
* btrfs_del_qgroup_relation()
* btrfs_create_qgroup()
* btrfs_remove_qgroup()
* btrfs_limit_qgroup()
* btrfs_qgroup_inherit() call inside create_subvol()
So we have a higher level protection provided by transaction, thus we
don't need to always hold qgroup_ioctl_lock in btrfs_qgroup_inherit().
Only the btrfs_qgroup_inherit() call in create_subvol() needs to hold
qgroup_ioctl_lock, while the btrfs_qgroup_inherit() call in
create_pending_snapshot() is already protected by transaction.
So the fix is to detect the context by checking
trans->transaction->state.
If we're at TRANS_STATE_COMMIT_DOING, then we're in commit transaction
context and no need to get the mutex.
Reported-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
The following script can cause unexpected fsync failure:
#!/bin/bash
dev=/dev/test/test
mnt=/mnt/btrfs
mkfs.btrfs -f $dev -b 512M > /dev/null
mount $dev $mnt -o nospace_cache
# Prealloc one extent
xfs_io -f -c "falloc 8k 64m" $mnt/file1
# Fill the remaining data space
xfs_io -f -c "pwrite 0 -b 4k 512M" $mnt/padding
sync
# Write into the prealloc extent
xfs_io -c "pwrite 1m 16m" $mnt/file1
# Reflink then fsync, fsync would fail due to ENOSPC
xfs_io -c "reflink $mnt/file1 8k 0 4k" -c "fsync" $mnt/file1
umount $dev
The fsync fails with ENOSPC, and the last page of the buffered write is
lost.
[CAUSE]
This is caused by:
- Btrfs' back reference only has extent level granularity
So write into shared extent must be COWed even only part of the extent
is shared.
So for above script we have:
- fallocate
Create a preallocated extent where we can do NOCOW write.
- fill all the remaining data and unallocated space
- buffered write into preallocated space
As we have not enough space available for data and the extent is not
shared (yet) we fall into NOCOW mode.
- reflink
Now part of the large preallocated extent is shared, later write
into that extent must be COWed.
- fsync triggers writeback
But now the extent is shared and therefore we must fallback into COW
mode, which fails with ENOSPC since there's not enough space to
allocate data extents.
[WORKAROUND]
The workaround is to ensure any buffered write in the related extents
(not just the reflink source range) get flushed before reflink/dedupe,
so that NOCOW writes succeed that happened before reflinking succeed.
The workaround is expensive, we could do it better by only flushing
NOCOW range, but that needs extra accounting for NOCOW range.
For now, fix the possible data loss first.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
At the context of btrfs_run_delalloc_range(), we haven't started/joined
a transaction, thus even something went wrong, we can't and won't abort
transaction, thus no way to make the fs RO.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Just add a safe net for btrfs_space_info member updating.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Under certain conditions, we could have strange file extent item in log
tree like:
item 18 key (69599 108 397312) itemoff 15208 itemsize 53
extent data disk bytenr 0 nr 0
extent data offset 0 nr 18446744073709547520 ram 18446744073709547520
The num_bytes + ram_bytes overflow 64 bit type.
For num_bytes part, we can detect such overflow along with file offset
(key->offset), as file_offset + num_bytes should never go beyond u64.
For ram_bytes part, it's about the decompressed size of the extent, not
directly related to the size.
In theory it is OK to have a large value, and put extra limitation
on RAM bytes may cause unexpected false alerts.
So in tree-checker, we only check if the file offset and num bytes
overflow.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Instead of using @sign to determine whether we're adding or subtracting.
Even it only has 3 callers, it's still (and in fact already caused
problem in the past) confusing to use.
Refactor add_pinned_bytes() to add_pinned_bytes() and sub_pinned_bytes()
to explicitly show what we're doing.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
When mounting a fs with reloc tree and has qgroup enabled, it can cause
NULL pointer dereference at mount time:
BUG: kernel NULL pointer dereference, address: 00000000000000a8
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT SMP NOPTI
RIP: 0010:btrfs_qgroup_add_swapped_blocks+0x186/0x300 [btrfs]
Call Trace:
replace_path.isra.23+0x685/0x900 [btrfs]
merge_reloc_root+0x26e/0x5f0 [btrfs]
merge_reloc_roots+0x10a/0x1a0 [btrfs]
btrfs_recover_relocation+0x3cd/0x420 [btrfs]
open_ctree+0x1bc8/0x1ed0 [btrfs]
btrfs_mount_root+0x544/0x680 [btrfs]
legacy_get_tree+0x34/0x60
vfs_get_tree+0x2d/0xf0
fc_mount+0x12/0x40
vfs_kern_mount.part.12+0x61/0xa0
vfs_kern_mount+0x13/0x20
btrfs_mount+0x16f/0x860 [btrfs]
legacy_get_tree+0x34/0x60
vfs_get_tree+0x2d/0xf0
do_mount+0x81f/0xac0
ksys_mount+0xbf/0xe0
__x64_sys_mount+0x25/0x30
do_syscall_64+0x65/0x240
entry_SYSCALL_64_after_hwframe+0x49/0xbe
[CAUSE]
In btrfs_recover_relocation(), we don't have enough info to determine
which block group we're relocating, but only to merge existing reloc
trees.
Thus in btrfs_recover_relocation(), rc->block_group is NULL.
btrfs_qgroup_add_swapped_blocks() hasn't taken this into consideration,
and causes a NULL pointer dereference.
The bug is introduced by commit 3d0174f78e ("btrfs: qgroup: Only trace
data extents in leaves if we're relocating data block group"), and
later qgroup refactoring still keeps this optimization.
[FIX]
Thankfully in the context of btrfs_recover_relocation(), there is no
other progress can modify tree blocks, thus those swapped tree blocks
pair will never affect qgroup numbers, no matter whatever we set for
block->trace_leaf.
So we only need to check if @bg is NULL before accessing @bg->flags.
Reported-by: Juan Erbes <jerbes@gmail.com>
Link: https://bugzilla.opensuse.org/show_bug.cgi?id=1134806
Fixes: 3d0174f78e ("btrfs: qgroup: Only trace data extents in leaves if we're relocating data block group")
CC: stable@vger.kernel.org # 4.20+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
When a fs has orphan reloc tree along with unfinished balance:
...
item 16 key (TREE_RELOC ROOT_ITEM FS_TREE) itemoff 12090 itemsize 439
generation 12 root_dirid 256 bytenr 300400640 level 1 refs 0 <<<
lastsnap 8 byte_limit 0 bytes_used 1359872 flags 0x0(none)
uuid 7c48d938-33a3-4aae-ab19-6e5c9d406e46
item 17 key (BALANCE TEMPORARY_ITEM 0) itemoff 11642 itemsize 448
temporary item objectid BALANCE offset 0
balance status flags 14
Then at mount time, we can hit the following kernel BUG_ON():
BTRFS info (device dm-3): relocating block group 298844160 flags metadata|dup
------------[ cut here ]------------
kernel BUG at fs/btrfs/relocation.c:1413!
invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
CPU: 1 PID: 897 Comm: btrfs-balance Tainted: G O 5.2.0-rc1-custom #15
RIP: 0010:create_reloc_root+0x1eb/0x200 [btrfs]
Call Trace:
btrfs_init_reloc_root+0x96/0xb0 [btrfs]
record_root_in_trans+0xb2/0xe0 [btrfs]
btrfs_record_root_in_trans+0x55/0x70 [btrfs]
select_reloc_root+0x7e/0x230 [btrfs]
do_relocation+0xc4/0x620 [btrfs]
relocate_tree_blocks+0x592/0x6a0 [btrfs]
relocate_block_group+0x47b/0x5d0 [btrfs]
btrfs_relocate_block_group+0x183/0x2f0 [btrfs]
btrfs_relocate_chunk+0x4e/0xe0 [btrfs]
btrfs_balance+0x864/0xfa0 [btrfs]
balance_kthread+0x3b/0x50 [btrfs]
kthread+0x123/0x140
ret_from_fork+0x27/0x50
[CAUSE]
In btrfs, reloc trees are used to record swapped tree blocks during
balance.
Reloc tree either get merged (replace old tree blocks of its parent
subvolume) in next transaction if its ref is 1 (fresh).
Or is already merged and will be cleaned up if its ref is 0 (orphan).
After commit d2311e6985 ("btrfs: relocation: Delay reloc tree deletion
after merge_reloc_roots"), reloc tree cleanup is delayed until one block
group is balanced.
Since fresh reloc roots are recorded during merge, as long as there
is no power loss, those orphan reloc roots converted from fresh ones are
handled without problem.
However when power loss happens, orphan reloc roots can be recorded
on-disk, thus at next mount time, we will have orphan reloc roots from
on-disk data directly, and ignored by clean_dirty_subvols() routine.
Then when background balance starts to balance another block group, and
needs to create new reloc root for the same root, btrfs_insert_item()
returns -EEXIST, and trigger that BUG_ON().
[FIX]
For orphan reloc roots, also queue them to rc->dirty_subvol_roots, so
all reloc roots no matter orphan or not, can be cleaned up properly and
avoid above BUG_ON().
And to cooperate with above change, clean_dirty_subvols() will check if
the queued root is a reloc root or a subvol root.
For a subvol root, do the old work, and for a orphan reloc root, clean it
up.
Fixes: d2311e6985 ("btrfs: relocation: Delay reloc tree deletion after merge_reloc_roots")
CC: stable@vger.kernel.org # 5.1
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Commit ddf30cf03f ("btrfs: extent-tree: Use btrfs_ref to refactor
add_pinned_bytes()") refactored add_pinned_bytes(), but during that
refactor, there are two callers which add the pinned bytes instead
of subtracting.
That refactor misses those two caller, causing incorrect pinned bytes
calculation and resulting unexpected ENOSPC error.
Fix it by adding a new parameter @sign to restore the original behavior.
Reported-by: kernel test robot <rong.a.chen@intel.com>
Fixes: ddf30cf03f ("btrfs: extent-tree: Use btrfs_ref to refactor add_pinned_bytes()")
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Allowing error injection for btrfs_check_leaf_full() and
btrfs_check_node() is useful to test the failure path of btrfs write
time tree check.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since reloc tree doesn't contribute to qgroup numbers, just skip them.
This should catch the final cause of unnecessary data ref processing
when running balance of metadata with qgroups on.
The 4G data 16 snapshots test (*) should explain it pretty well:
| delayed subtree | refactor delayed ref | this patch
---------------------------------------------------------------------
relocated | 22653 | 22673 | 22744
qgroup dirty | 122792 | 48360 | 70
time | 24.494 | 11.606 | 3.944
Finally, we're at the stage where qgroup + metadata balance cost no
obvious overhead.
Test environment:
Test VM:
- vRAM 8G
- vCPU 8
- block dev vitrio-blk, 'unsafe' cache mode
- host block 850evo
Test workload:
- Copy 4G data from /usr/ to one subvolume
- Create 16 snapshots of that subvolume, and modify 3 files in each
snapshot
- Enable quota, rescan
- Time "btrfs balance start -m"
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Similar to btrfs_inc_extent_ref(), use btrfs_ref to replace the long
parameter list and the confusing @owner parameter.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Use the new btrfs_ref structure and replace parameter list to clean up
the usage of owner and level to distinguish the extent types.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since add_pinned_bytes() only needs to know if the extent is metadata
and if it's a chunk tree extent, btrfs_ref is a perfect match for it, as
we don't need various owner/level trick to determine extent type.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
It's a perfect match for btrfs_ref_tree_mod() to use btrfs_ref, as
btrfs_ref describes a metadata/data reference update comprehensively.
Now we have one less function use confusing owner/level trick.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Just like btrfs_add_delayed_tree_ref(), use btrfs_ref to refactor
btrfs_add_delayed_data_ref().
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_add_delayed_tree_ref() has a longer and longer parameter list, and
some callers like btrfs_inc_extent_ref() are using @owner as level for
delayed tree ref.
Instead of making the parameter list longer, use btrfs_ref to refactor
it, so each parameter assignment should be self-explaining without dirty
level/owner trick, and provides the basis for later refactoring.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The process_func function pointer is local to __btrfs_mod_ref() and
points to either btrfs_inc_extent_ref() or btrfs_free_extent().
Open code it to make later delayed ref refactor easier, so we can
refactor btrfs_inc_extent_ref() and btrfs_free_extent() in different
patches.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Current delayed ref interface has several problems:
- Longer and longer parameter lists
bytenr
num_bytes
parent
---------- so far so good
ref_root
owner
offset
---------- I don't feel good now
- Different interpretation of the same parameter
Above @owner for data ref is inode number (u64),
while for tree ref, it's level (int).
They are even in different size range.
For level we only need 0 ~ 8, while for ino it's
BTRFS_FIRST_FREE_OBJECTID ~ BTRFS_LAST_FREE_OBJECTID.
And @offset doesn't even make sense for tree ref.
Such parameter reuse may look clever as an hidden union, but it
destroys code readability.
To solve both problems, we introduce a new structure, btrfs_ref to solve
them:
- Structure instead of long parameter list
This makes later expansion easier, and is better documented.
- Use btrfs_ref::type to distinguish data and tree ref
- Use proper union to store data/tree ref specific structures.
- Use separate functions to fill data/tree ref data, with a common generic
function to fill common bytenr/num_bytes members.
All parameters will find its place in btrfs_ref, and an extra member,
@real_root, inspired by ref-verify code, is newly introduced for later
qgroup code, to record which tree is triggered by this extent modification.
This patch doesn't touch any code, but provides the basis for further
refactoring.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Unlike btrfs_tree_lock() and btrfs_tree_read_lock(), the remaining
functions in locking.c will not sleep, thus doesn't make much sense to
record their execution time.
Those events are introduced mainly for user space tool to audit and
detect lock leakage or dead lock.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are two tree lock events which can sleep:
- btrfs_tree_read_lock()
- btrfs_tree_lock()
Sometimes we may need to look into the concurrency picture of the fs.
For that case, we need the execution time of above two functions and the
owner of @eb.
Here we introduce a trace events for user space tools like bcc, to get
the execution time of above two functions, and get detailed owner info
where eBPF code can't.
All the overhead is hidden behind the trace events, so if events are not
enabled, there is no overhead.
These trace events also output bytenr and generation, allow them to be
pared with unlock events to pin down deadlock.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are at least 2 reports about a memory bit flip sneaking into
on-disk data.
Currently we only have a relaxed check triggered at
btrfs_mark_buffer_dirty() time, as it's not mandatory and only for
CONFIG_BTRFS_FS_CHECK_INTEGRITY enabled build, it doesn't help users to
detect such problem.
This patch will address the hole by triggering comprehensive check on
tree blocks before writing it back to disk.
The design points are:
- Timing of the check: Tree block write hook
This timing is chosen to reduce the overhead.
The comprehensive check should be as expensive as a checksum
calculation.
Doing full check at btrfs_mark_buffer_dirty() is too expensive for end
user.
- Loose empty leaf check
Originally for an empty leaf, tree-checker will report error if it's
not a tree root.
The problem for such check at write time is:
* False alert for tree root created in current transaction
In that case, the commit root still needs to be written to disk.
And since current root can differ from commit root, then it will
cause false alert.
This happens for log tree.
* False alert for relocated tree block
Relocated tree block can be written to disk due to memory pressure,
in that case an empty csum tree root can be written to disk and
cause false alert, since csum root node hasn't been updated.
Previous patch of removing comprehensive empty leaf owner check has
paved the way for this patch.
The example error output will be something like:
BTRFS critical (device dm-3): corrupt leaf: root=2 block=1350630375424 slot=68, bad key order, prev (10510212874240 169 0) current (1714119868416 169 0)
BTRFS error (device dm-3): block=1350630375424 write time tree block corruption detected
BTRFS: error (device dm-3) in btrfs_commit_transaction:2220: errno=-5 IO failure (Error while writing out transaction)
BTRFS info (device dm-3): forced readonly
BTRFS warning (device dm-3): Skipping commit of aborted transaction.
BTRFS: error (device dm-3) in cleanup_transaction:1839: errno=-5 IO failure
BTRFS info (device dm-3): delayed_refs has NO entry
Reported-by: Leonard Lausen <leonard@lausen.nl>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Commit 1ba98d086f ("Btrfs: detect corruption when non-root leaf has
zero item") introduced comprehensive root owner checker.
However it's pretty expensive tree search to locate the owner root,
especially when it get reused by mandatory read and write time
tree-checker.
This patch will remove that check, and completely rely on owner based
empty leaf check, which is much faster and still works fine for most
case.
And since we skip the old root owner check, now write time tree check
can be merged with btrfs_check_leaf_full().
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
When accessing a file on a crafted image, btrfs can crash in block layer:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
PGD 136501067 P4D 136501067 PUD 124519067 PMD 0
CPU: 3 PID: 0 Comm: swapper/3 Not tainted 5.0.0-rc8-default #252
RIP: 0010:end_bio_extent_readpage+0x144/0x700
Call Trace:
<IRQ>
blk_update_request+0x8f/0x350
blk_mq_end_request+0x1a/0x120
blk_done_softirq+0x99/0xc0
__do_softirq+0xc7/0x467
irq_exit+0xd1/0xe0
call_function_single_interrupt+0xf/0x20
</IRQ>
RIP: 0010:default_idle+0x1e/0x170
[CAUSE]
The crafted image has a tricky corruption, the INODE_ITEM has a
different type against its parent dir:
item 20 key (268 INODE_ITEM 0) itemoff 2808 itemsize 160
generation 13 transid 13 size 1048576 nbytes 1048576
block group 0 mode 121644 links 1 uid 0 gid 0 rdev 0
sequence 9 flags 0x0(none)
This mode number 0120000 means it's a symlink.
But the dir item think it's still a regular file:
item 8 key (264 DIR_INDEX 5) itemoff 3707 itemsize 32
location key (268 INODE_ITEM 0) type FILE
transid 13 data_len 0 name_len 2
name: f4
item 40 key (264 DIR_ITEM 51821248) itemoff 1573 itemsize 32
location key (268 INODE_ITEM 0) type FILE
transid 13 data_len 0 name_len 2
name: f4
For symlink, we don't set BTRFS_I(inode)->io_tree.ops and leave it
empty, as symlink is only designed to have inlined extent, all handled
by tree block read. Thus no need to trigger btrfs_submit_bio_hook() for
inline file extent.
However end_bio_extent_readpage() expects tree->ops populated, as it's
reading regular data extent. This causes NULL pointer dereference.
[FIX]
This patch fixes the problem in two ways:
- Verify inode mode against its dir item when looking up inode
So in btrfs_lookup_dentry() if we find inode mode mismatch with dir
item, we error out so that corrupted inode will not be accessed.
- Verify inode mode when getting extent mapping
Only regular file should have regular or preallocated extent.
If we found regular/preallocated file extent for symlink or
the rest, we error out before submitting the read bio.
With this fix that crafted image can be rejected gracefully:
BTRFS critical (device loop0): inode mode mismatch with dir: inode mode=0121644 btrfs type=7 dir type=1
Reported-by: Yoon Jungyeon <jungyeon@gatech.edu>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202763
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is a report in kernel bugzilla about mismatch file type in dir
item and inode item.
This inspires us to check inode mode in inode item.
This patch will check the following members:
- inode key objectid
Should be ROOT_DIR_DIR or [256, (u64)-256] or FREE_INO.
- inode key offset
Should be 0
- inode item generation
- inode item transid
No newer than sb generation + 1.
The +1 is for log tree.
- inode item mode
No unknown bits.
No invalid S_IF* bit.
NOTE: S_IFMT check is not enough, need to check every know type.
- inode item nlink
Dir should have no more link than 1.
- inode item flags
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
For fuzzed image whose DEV_ITEM has invalid total_bytes as 0, then
kernel will just panic:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000098
#PF error: [normal kernel read fault]
PGD 800000022b2bd067 P4D 800000022b2bd067 PUD 22b2bc067 PMD 0
Oops: 0000 [#1] SMP PTI
CPU: 0 PID: 1106 Comm: mount Not tainted 5.0.0-rc8+ #9
RIP: 0010:btrfs_verify_dev_extents+0x2a5/0x5a0
Call Trace:
open_ctree+0x160d/0x2149
btrfs_mount_root+0x5b2/0x680
[CAUSE]
If device extent verification finds a deivce with 0 total_bytes, then it
assumes it's a seed dummy, then search for seed devices.
But in this case, there is no seed device at all, causing NULL pointer.
[FIX]
Since this is caused by fuzzed image, let's go the tree-check way, just
add a new verification for device item.
Reported-by: Yoon Jungyeon <jungyeon@gatech.edu>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202691
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
Since we have btrfs_check_chunk_valid() in tree-checker, let's do
chunk item verification in tree-checker too.
Since the tree-checker is run at endio time, if one chunk leaf fails
chunk verification, we can still retry the other copy, making btrfs more
robust to fuzzed image as we may still get a good chunk item.
Also since we have done chunk verification in tree block read time, skip
the btrfs_check_chunk_valid() call in read_one_chunk() if we're reading
chunk items from leaf.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Old error message would be something like:
BTRFS error (device dm-3): invalid chunk num_stipres: 0
New error message would be:
Btrfs critical (device dm-3): corrupt superblock syschunk array: chunk_start=2097152, invalid chunk num_stripes: 0
Or
Btrfs critical (device dm-3): corrupt leaf: root=3 block=8388608 slot=3 chunk_start=2097152, invalid chunk num_stripes: 0
And for certain error message, also output expected value.
The error message levels are changed from error to critical.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
By function, chunk item verification is more suitable to be done inside
tree-checker.
So move btrfs_check_chunk_valid() to tree-checker.c and export it.
And since it's now moved to tree-checker, also add a better comment for
what this function is doing.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Just add one extra line to show when the corruption is detected.
Currently only read time detection is possible.
The planned distinguish line would be:
read time:
<detailed report>
block=XXXXX read time tree block corruption detected
write time:
<detailed report>
block=XXXXX write time tree block corruption detected
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We can only get <=0 from extent_write_cache_pages, add an ASSERT() for
it just in case.
Then instead of submitting the write bio even if we got some error,
check the return value first.
If we have already hit some error, just clean up the corrupted or
half-baked bio, and return error.
If there is no error so far, then call flush_write_bio() and return the
result.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This function needs some extra checks on locked pages and eb. For error
handling we need to unlock locked pages and the eb.
There is a rare >0 return value branch, where all pages get locked
while write bio is not flushed.
Thankfully it's handled by the only caller, btree_write_cache_pages(),
as later write_one_eb() call will trigger submit_one_bio(). So there
shouldn't be any problem.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We can only get @ret <= 0. Add an ASSERT() for it just in case.
Then, instead of submitting the write bio even we got some error, check
the return value first.
If we have already hit some error, just clean up the corrupted or
half-baked bio, and return error.
If there is no error so far, then call flush_write_bio() and return the
result.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since __extent_writepage() will no longer return >0 value,
(ret == AOP_WRITEPAGE_ACTIVATE) will never be true.
Kill that dead branch.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In btree_write_cache_pages(), we can only get @ret <= 0.
Add an ASSERT() for it just in case.
Then instead of submitting the write bio even we got some error, check
the return value first.
If we have already hit some error, just clean up the corrupted or
half-baked bio, and return error.
If there is no error so far, then call flush_write_bio() and return the
result.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since now flush_write_bio() could return error, kill the BUG_ON() first.
Then don't call flush_write_bio() unconditionally, instead we check the
return value from __extent_writepage() first.
If __extent_writepage() fails, we do cleanup, and return error without
submitting the possible corrupted or half-baked bio.
If __extent_writepage() successes, then we call flush_write_bio() and
return the result.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have a BUG_ON() in flush_write_bio() to handle the return value of
submit_one_bio().
Move the BUG_ON() one level up to all its callers.
This patch will introduce temporary variable, @flush_ret to keep code
change minimal in this patch. That variable will be cleaned up when
enhancing the error handling later.
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have internal report of strange transaction abort due to EUCLEAN
without any error message.
Since error message inside verify_level_key() is only enabled for
CONFIG_BTRFS_DEBUG, the error message won't be printed on most builds.
This patch will make the error message mandatory, so when problem
happens we know what's causing the problem.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
When reading a file from a fuzzed image, kernel can panic like:
BTRFS warning (device loop0): csum failed root 5 ino 270 off 0 csum 0x98f94189 expected csum 0x00000000 mirror 1
assertion failed: !memcmp_extent_buffer(b, &disk_key, offsetof(struct btrfs_leaf, items[0].key), sizeof(disk_key)), file: fs/btrfs/ctree.c, line: 2544
------------[ cut here ]------------
kernel BUG at fs/btrfs/ctree.h:3500!
invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
RIP: 0010:btrfs_search_slot.cold.24+0x61/0x63 [btrfs]
Call Trace:
btrfs_lookup_csum+0x52/0x150 [btrfs]
__btrfs_lookup_bio_sums+0x209/0x640 [btrfs]
btrfs_submit_bio_hook+0x103/0x170 [btrfs]
submit_one_bio+0x59/0x80 [btrfs]
extent_read_full_page+0x58/0x80 [btrfs]
generic_file_read_iter+0x2f6/0x9d0
__vfs_read+0x14d/0x1a0
vfs_read+0x8d/0x140
ksys_read+0x52/0xc0
do_syscall_64+0x60/0x210
entry_SYSCALL_64_after_hwframe+0x49/0xbe
[CAUSE]
The fuzzed image has a corrupted leaf whose first key doesn't match its
parent:
checksum tree key (CSUM_TREE ROOT_ITEM 0)
node 29741056 level 1 items 14 free 107 generation 19 owner CSUM_TREE
fs uuid 3381d111-94a3-4ac7-8f39-611bbbdab7e6
chunk uuid 9af1c3c7-2af5-488b-8553-530bd515f14c
...
key (EXTENT_CSUM EXTENT_CSUM 79691776) block 29761536 gen 19
leaf 29761536 items 1 free space 1726 generation 19 owner CSUM_TREE
leaf 29761536 flags 0x1(WRITTEN) backref revision 1
fs uuid 3381d111-94a3-4ac7-8f39-611bbbdab7e6
chunk uuid 9af1c3c7-2af5-488b-8553-530bd515f14c
item 0 key (EXTENT_CSUM EXTENT_CSUM 8798638964736) itemoff 1751 itemsize 2244
range start 8798638964736 end 8798641262592 length 2297856
When reading the above tree block, we have extent_buffer->refs = 2 in
the context:
- initial one from __alloc_extent_buffer()
alloc_extent_buffer()
|- __alloc_extent_buffer()
|- atomic_set(&eb->refs, 1)
- one being added to fs_info->buffer_radix
alloc_extent_buffer()
|- check_buffer_tree_ref()
|- atomic_inc(&eb->refs)
So if even we call free_extent_buffer() in read_tree_block or other
similar situation, we only decrease the refs by 1, it doesn't reach 0
and won't be freed right now.
The staled eb and its corrupted content will still be kept cached.
Furthermore, we have several extra cases where we either don't do first
key check or the check is not proper for all callers:
- scrub
We just don't have first key in this context.
- shared tree block
One tree block can be shared by several snapshot/subvolume trees.
In that case, the first key check for one subvolume doesn't apply to
another.
So for the above reasons, a corrupted extent buffer can sneak into the
buffer cache.
[FIX]
Call verify_level_key in read_block_for_search to do another
verification. For that purpose the function is exported.
Due to above reasons, although we can free corrupted extent buffer from
cache, we still need the check in read_block_for_search(), for scrub and
shared tree blocks.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202755
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202757
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202759
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202761
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202767
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202769
Reported-by: Yoon Jungyeon <jungyeon@gatech.edu>
CC: stable@vger.kernel.org # 4.19+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
From the introduction of btrfs_(set|clear)_header_flag, there is no
usage of its return value. So just make it return void.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Commit d2311e6985 ("btrfs: relocation: Delay reloc tree deletion after
merge_reloc_roots()") expands the life span of root->reloc_root.
This breaks certain checs of fs_info->reloc_ctl. Before that commit, if
we have a root with valid reloc_root, then it's ensured to have
fs_info->reloc_ctl.
But now since reloc_root doesn't always mean a valid fs_info->reloc_ctl,
such check is unreliable and can cause the following NULL pointer
dereference:
BUG: unable to handle kernel NULL pointer dereference at 00000000000005c1
IP: btrfs_reloc_pre_snapshot+0x20/0x50 [btrfs]
PGD 0 P4D 0
Oops: 0000 [#1] SMP PTI
CPU: 0 PID: 10379 Comm: snapperd Not tainted
Call Trace:
create_pending_snapshot+0xd7/0xfc0 [btrfs]
create_pending_snapshots+0x8e/0xb0 [btrfs]
btrfs_commit_transaction+0x2ac/0x8f0 [btrfs]
btrfs_mksubvol+0x561/0x570 [btrfs]
btrfs_ioctl_snap_create_transid+0x189/0x190 [btrfs]
btrfs_ioctl_snap_create_v2+0x102/0x150 [btrfs]
btrfs_ioctl+0x5c9/0x1e60 [btrfs]
do_vfs_ioctl+0x90/0x5f0
SyS_ioctl+0x74/0x80
do_syscall_64+0x7b/0x150
entry_SYSCALL_64_after_hwframe+0x3d/0xa2
RIP: 0033:0x7fd7cdab8467
Fix it by explicitly checking fs_info->reloc_ctl other than using the
implied root->reloc_root.
Fixes: d2311e6985 ("btrfs: relocation: Delay reloc tree deletion after merge_reloc_roots")
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Although btrfs heavily relies on extent_io_tree, we don't really have
any good trace events for them.
This patch will add the folowing trace events:
- trace_btrfs_set_extent_bit()
- trace_btrfs_clear_extent_bit()
- trace_btrfs_convert_extent_bit()
Since selftests could create temporary extent_io_tree without fs_info,
modify TP_fast_assign_fsid() to accept NULL as fs_info. NULL fs_info
will lead to all zero fsid.
The output would be:
btrfs_set_extent_bit: <FDID>: io_tree=INODE_IO ino=1 root=1 start=22036480 len=4096 set_bits=LOCKED
btrfs_set_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22040576 len=4096 set_bits=LOCKED
btrfs_set_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22044672 len=4096 set_bits=LOCKED
btrfs_set_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22048768 len=4096 set_bits=LOCKED
btrfs_clear_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22036480 len=16384 clear_bits=LOCKED
^^^ Extent buffer 22036480 read from disk, the locking progress
btrfs_set_extent_bit: <FSID>: io_tree=TRANS_DIRTY_PAGES ino=1 root=1 start=30425088 len=16384 set_bits=DIRTY
btrfs_set_extent_bit: <FSID>: io_tree=TRANS_DIRTY_PAGES ino=1 root=1 start=30441472 len=16384 set_bits=DIRTY
^^^ 2 new tree blocks allocated in one transaction
btrfs_set_extent_bit: <FSID>: io_tree=FREED_EXTENTS0 ino=0 root=0 start=30523392 len=16384 set_bits=DIRTY
btrfs_set_extent_bit: <FSID>: io_tree=FREED_EXTENTS0 ino=0 root=0 start=30556160 len=16384 set_bits=DIRTY
^^^ 2 old tree blocks get pinned down
There is one point which need attention:
1) Those trace events can be pretty heavy:
The following workload would generate over 400 trace events.
mkfs.btrfs -f $dev
start_trace
mount $dev $mnt -o enospc_debug
sync
touch $mnt/file1
touch $mnt/file2
touch $mnt/file3
xfs_io -f -c "pwrite 0 16k" $mnt/file4
umount $mnt
end_trace
It's not recommended to use them in real world environment.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ rename enums ]
Signed-off-by: David Sterba <dsterba@suse.com>
Btrfs has the following different extent_io_trees used:
- fs_info::free_extents[2]
- btrfs_inode::io_tree - for both normal inodes and the btree inode
- btrfs_inode::io_failure_tree
- btrfs_transaction::dirty_pages
- btrfs_root::dirty_log_pages
If we want to trace changes in those trees, it will be pretty hard to
distinguish them.
Instead of using hard-to-read pointer address, this patch will introduce
a new member extent_io_tree::owner to track the owner.
This modification needs all the callers of extent_io_tree_init() to
accept a new parameter @owner.
This patch provides the basis for later trace events.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This patch will add a new member fs_info to extent_io_tree.
This provides the basis for later trace events to distinguish the output
between different btrfs filesystems. While this increases the size of
the structure, we want to know the source of the trace events and
passing the fs_info as an argument to all contexts is not possible.
The selftests are now allowed to set it to NULL as they don't use the
tracepoints.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When we failed to find a root key in btrfs_update_root(), we just panic.
That's definitely not cool, fix it by outputting an unique error
message, aborting current transaction and return -EUCLEAN. This should
not normally happen as the root has been used by the callers in some
way.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
Btrfs qgroup will still hit EDQUOT under the following case:
$ dev=/dev/test/test
$ mnt=/mnt/btrfs
$ umount $mnt &> /dev/null
$ umount $dev &> /dev/null
$ mkfs.btrfs -f $dev
$ mount $dev $mnt -o nospace_cache
$ btrfs subv create $mnt/subv
$ btrfs quota enable $mnt
$ btrfs quota rescan -w $mnt
$ btrfs qgroup limit -e 1G $mnt/subv
$ fallocate -l 900M $mnt/subv/padding
$ sync
$ rm $mnt/subv/padding
# Hit EDQUOT
$ xfs_io -f -c "pwrite 0 512M" $mnt/subv/real_file
[CAUSE]
Since commit a514d63882 ("btrfs: qgroup: Commit transaction in advance
to reduce early EDQUOT"), btrfs is not forced to commit transaction to
reclaim more quota space.
Instead, we just check pertrans metadata reservation against some
threshold and try to do asynchronously transaction commit.
However in above case, the pertrans metadata reservation is pretty small
thus it will never trigger asynchronous transaction commit.
[FIX]
Instead of only accounting pertrans metadata reservation, we calculate
how much free space we have, and if there isn't much free space left,
commit transaction asynchronously to try to free some space.
This may slow down the fs when we have less than 32M free qgroup space,
but should reduce a lot of false EDQUOT, so the cost should be
acceptable.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
Btrfs/139 will fail with a high probability if the testing machine (VM)
has only 2G RAM.
Resulting the final write success while it should fail due to EDQUOT,
and the fs will have quota exceeding the limit by 16K.
The simplified reproducer will be: (needs a 2G ram VM)
$ mkfs.btrfs -f $dev
$ mount $dev $mnt
$ btrfs subv create $mnt/subv
$ btrfs quota enable $mnt
$ btrfs quota rescan -w $mnt
$ btrfs qgroup limit -e 1G $mnt/subv
$ for i in $(seq -w 1 8); do
xfs_io -f -c "pwrite 0 128M" $mnt/subv/file_$i > /dev/null
echo "file $i written" > /dev/kmsg
done
$ sync
$ btrfs qgroup show -pcre --raw $mnt
The last pwrite will not trigger EDQUOT and final 'qgroup show' will
show something like:
qgroupid rfer excl max_rfer max_excl parent child
-------- ---- ---- -------- -------- ------ -----
0/5 16384 16384 none none --- ---
0/256 1073758208 1073758208 none 1073741824 --- ---
And 1073758208 is larger than
> 1073741824.
[CAUSE]
It's a bug in btrfs qgroup data reserved space management.
For quota limit, we must ensure that:
reserved (data + metadata) + rfer/excl <= limit
Since rfer/excl is only updated at transaction commmit time, reserved
space needs to be taken special care.
One important part of reserved space is data, and for a new data extent
written to disk, we still need to take the reserved space until
rfer/excl numbers get updated.
Originally when an ordered extent finishes, we migrate the reserved
qgroup data space from extent_io tree to delayed ref head of the data
extent, expecting delayed ref will only be cleaned up at commit
transaction time.
However for small RAM machine, due to memory pressure dirty pages can be
flushed back to disk without committing a transaction.
The related events will be something like:
file 1 written
btrfs_finish_ordered_io: ino=258 ordered offset=0 len=54947840
btrfs_finish_ordered_io: ino=258 ordered offset=54947840 len=5636096
btrfs_finish_ordered_io: ino=258 ordered offset=61153280 len=57344
btrfs_finish_ordered_io: ino=258 ordered offset=61210624 len=8192
btrfs_finish_ordered_io: ino=258 ordered offset=60583936 len=569344
cleanup_ref_head: num_bytes=54947840
cleanup_ref_head: num_bytes=5636096
cleanup_ref_head: num_bytes=569344
cleanup_ref_head: num_bytes=57344
cleanup_ref_head: num_bytes=8192
^^^^^^^^^^^^^^^^ This will free qgroup data reserved space
file 2 written
...
file 8 written
cleanup_ref_head: num_bytes=8192
...
btrfs_commit_transaction <<< the only transaction committed during
the test
When file 2 is written, we have already freed 128M reserved qgroup data
space for ino 258. Thus later write won't trigger EDQUOT.
This allows us to write more data beyond qgroup limit.
In my 2G ram VM, it could reach about 1.2G before hitting EDQUOT.
[FIX]
By moving reserved qgroup data space from btrfs_delayed_ref_head to
btrfs_qgroup_extent_record, we can ensure that reserved qgroup data
space won't be freed half way before commit transaction, thus fix the
problem.
Fixes: f64d5ca868 ("btrfs: delayed_ref: Add new function to record reserved space into delayed ref")
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is no need to forward declare flush_write_bio(), as it only
depends on submit_one_bio(). Both of them are pretty small, just move
them to kill the forward declaration.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since inc_block_group_ro() would return -ENOSPC, outputting debug info
for enospc_debug mount option would be helpful to debug some balance
false ENOSPC report.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Inside qgroup_rsv_add/release(), we have trace events
trace_qgroup_update_reserve() to catch reserved space update.
However we still have two manual trace_qgroup_update_reserve() calls
just outside these functions. Remove these duplicated calls.
Fixes: 64ee4e751a ("btrfs: qgroup: Update trace events to use new separate rsv types")
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Since it's replaced by new delayed subtree swap code, remove the
original code.
The cleanup is small since most of its core function is still used by
delayed subtree swap trace.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Before this patch, qgroup code traces the whole subtree of subvolume and
reloc trees unconditionally.
This makes qgroup numbers consistent, but it could cause tons of
unnecessary extent tracing, which causes a lot of overhead.
However for subtree swap of balance, just swap both subtrees because
they contain the same contents and tree structure, so qgroup numbers
won't change.
It's the race window between subtree swap and transaction commit could
cause qgroup number change.
This patch will delay the qgroup subtree scan until COW happens for the
subtree root.
So if there is no other operations for the fs, balance won't cause extra
qgroup overhead. (best case scenario)
Depending on the workload, most of the subtree scan can still be
avoided.
Only for worst case scenario, it will fall back to old subtree swap
overhead. (scan all swapped subtrees)
[[Benchmark]]
Hardware:
VM 4G vRAM, 8 vCPUs,
disk is using 'unsafe' cache mode,
backing device is SAMSUNG 850 evo SSD.
Host has 16G ram.
Mkfs parameter:
--nodesize 4K (To bump up tree size)
Initial subvolume contents:
4G data copied from /usr and /lib.
(With enough regular small files)
Snapshots:
16 snapshots of the original subvolume.
each snapshot has 3 random files modified.
balance parameter:
-m
So the content should be pretty similar to a real world root fs layout.
And after file system population, there is no other activity, so it
should be the best case scenario.
| v4.20-rc1 | w/ patchset | diff
-----------------------------------------------------------------------
relocated extents | 22615 | 22457 | -0.1%
qgroup dirty extents | 163457 | 121606 | -25.6%
time (sys) | 22.884s | 18.842s | -17.6%
time (real) | 27.724s | 22.884s | -17.5%
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
To allow delayed subtree swap rescan, btrfs needs to record per-root
information about which tree blocks get swapped. This patch introduces
the required infrastructure.
The designed workflow will be:
1) Record the subtree root block that gets swapped.
During subtree swap:
O = Old tree blocks
N = New tree blocks
reloc tree subvolume tree X
Root Root
/ \ / \
NA OB OA OB
/ | | \ / | | \
NC ND OE OF OC OD OE OF
In this case, NA and OA are going to be swapped, record (NA, OA) into
subvolume tree X.
2) After subtree swap.
reloc tree subvolume tree X
Root Root
/ \ / \
OA OB NA OB
/ | | \ / | | \
OC OD OE OF NC ND OE OF
3a) COW happens for OB
If we are going to COW tree block OB, we check OB's bytenr against
tree X's swapped_blocks structure.
If it doesn't fit any, nothing will happen.
3b) COW happens for NA
Check NA's bytenr against tree X's swapped_blocks, and get a hit.
Then we do subtree scan on both subtrees OA and NA.
Resulting 6 tree blocks to be scanned (OA, OC, OD, NA, NC, ND).
Then no matter what we do to subvolume tree X, qgroup numbers will
still be correct.
Then NA's record gets removed from X's swapped_blocks.
4) Transaction commit
Any record in X's swapped_blocks gets removed, since there is no
modification to swapped subtrees, no need to trigger heavy qgroup
subtree rescan for them.
This will introduce 128 bytes overhead for each btrfs_root even qgroup
is not enabled. This is to reduce memory allocations and potential
failures.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Refactor btrfs_qgroup_trace_subtree_swap() into
qgroup_trace_subtree_swap(), which only needs two extent buffer and some
other bool to control the behavior.
This provides the basis for later delayed subtree scan work.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Relocation code will drop btrfs_root::reloc_root as soon as
merge_reloc_root() finishes.
However later qgroup code will need to access btrfs_root::reloc_root
after merge_reloc_root() for delayed subtree rescan.
So alter the timming of resetting btrfs_root:::reloc_root, make it
happens after transaction commit.
With this patch, we will introduce a new btrfs_root::state,
BTRFS_ROOT_DEAD_RELOC_TREE, to info part of btrfs_root::reloc_tree user
that although btrfs_root::reloc_tree is still non-NULL, but still it's
not used any more.
The lifespan of btrfs_root::reloc tree will become:
Old behavior | New
------------------------------------------------------------------------
btrfs_init_reloc_root() --- | btrfs_init_reloc_root() ---
set reloc_root | | set reloc_root |
| | |
| | |
merge_reloc_root() | | merge_reloc_root() |
|- btrfs_update_reloc_root() --- | |- btrfs_update_reloc_root() -+-
clear btrfs_root::reloc_root | set ROOT_DEAD_RELOC_TREE |
| record root into dirty |
| roots rbtree |
| |
| reloc_block_group() Or |
| btrfs_recover_relocation() |
| | After transaction commit |
| |- clean_dirty_subvols() ---
| clear btrfs_root::reloc_root
During ROOT_DEAD_RELOC_TREE set lifespan, the only user of
btrfs_root::reloc_tree should be qgroup.
Since reloc root needs a longer life-span, this patch will also delay
btrfs_drop_snapshot() call.
Now btrfs_drop_snapshot() is called in clean_dirty_subvols().
This patch will increase the size of btrfs_root by 16 bytes.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
Linux v5.0-rc1 will fail fstests/btrfs/163 with the following kernel
message:
BTRFS error (device dm-6): dev extent devid 1 physical offset 13631488 len 8388608 is beyond device boundary 0
BTRFS error (device dm-6): failed to verify dev extents against chunks: -117
BTRFS error (device dm-6): open_ctree failed
[CAUSE]
Commit cf90d884b3 ("btrfs: Introduce mount time chunk <-> dev extent
mapping check") introduced strict check on dev extents.
We use btrfs_find_device() with dev uuid and fs uuid set to NULL, and
only dependent on @devid to find the real device.
For seed devices, we call clone_fs_devices() in open_seed_devices() to
allow us search seed devices directly.
However clone_fs_devices() just populates devices with devid and dev
uuid, without populating other essential members, like disk_total_bytes.
This makes any device returned by btrfs_find_device(fs_info, devid,
NULL, NULL) is just a dummy, with 0 disk_total_bytes, and any dev
extents on the seed device will not pass the device boundary check.
[FIX]
This patch will try to verify the device returned by btrfs_find_device()
and if it's a dummy then re-search in seed devices.
Fixes: cf90d884b3 ("btrfs: Introduce mount time chunk <-> dev extent mapping check")
CC: stable@vger.kernel.org # 4.19+
Reported-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
@blocksize variable in do_walk_down() is only used once, really no need
to declare it.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In extent-io self test, we need 2 ordered extents at its maximum size to
do the test.
Instead of using the intermediate numbers, use BTRFS_MAX_EXTENT_SIZE for
@max_bytes, and twice @max_bytes for @total_dirty. This should explain
why we need all these magic numbers and prevent people to modify them by
accident.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add extra dev extent end check against device boundary.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Enhance btrfs_verify_dev_extents() to remember previous checked dev
extents, so it can verify no dev extents can overlap.
Analysis from Hans:
"Imagine allocating a DATA|DUP chunk.
In the chunk allocator, we first set...
max_stripe_size = SZ_1G;
max_chunk_size = BTRFS_MAX_DATA_CHUNK_SIZE
... which is 10GiB.
Then...
/* we don't want a chunk larger than 10% of writeable space */
max_chunk_size = min(div_factor(fs_devices->total_rw_bytes, 1),
max_chunk_size);
Imagine we only have one 7880MiB block device in this filesystem. Now
max_chunk_size is down to 788MiB.
The next step in the code is to search for max_stripe_size * dev_stripes
amount of free space on the device, which is in our example 1GiB * 2 =
2GiB. Imagine the device has exactly 1578MiB free in one contiguous
piece. This amount of bytes will be put in devices_info[ndevs - 1].max_avail
Next we recalculate the stripe_size (which is actually the device extent
length), based on the actual maximum amount of available raw disk space:
stripe_size = div_u64(devices_info[ndevs - 1].max_avail, dev_stripes);
stripe_size is now 789MiB
Next we do...
data_stripes = num_stripes / ncopies
...where data_stripes ends up as 1, because num_stripes is 2 (the amount
of device extents we're going to have), and DUP has ncopies 2.
Next there's a check...
if (stripe_size * data_stripes > max_chunk_size)
...which matches because 789MiB * 1 > 788MiB.
We go into the if code, and next is...
stripe_size = div_u64(max_chunk_size, data_stripes);
...which resets stripe_size to max_chunk_size: 788MiB
Next is a fun one...
/* bump the answer up to a 16MB boundary */
stripe_size = round_up(stripe_size, SZ_16M);
...which changes stripe_size from 788MiB to 800MiB.
We're not done changing stripe_size yet...
/* But don't go higher than the limits we found while searching
* for free extents
*/
stripe_size = min(devices_info[ndevs - 1].max_avail,
stripe_size);
This is bad. max_avail is twice the stripe_size (we need to fit 2 device
extents on the same device for DUP).
The result here is that 800MiB < 1578MiB, so it's unchanged. However,
the resulting DUP chunk will need 1600MiB disk space, which isn't there,
and the second dev_extent might extend into the next thing (next
dev_extent? end of device?) for 22MiB.
The last shown line of code relies on a situation where there's twice
the value of stripe_size present as value for the variable stripe_size
when it's DUP. This was actually the case before commit 92e222df7b
"btrfs: alloc_chunk: fix DUP stripe size handling", from which I quote:
"[...] in the meantime there's a check to see if the stripe_size does
not exceed max_chunk_size. Since during this check stripe_size is twice
the amount as intended, the check will reduce the stripe_size to
max_chunk_size if the actual correct to be used stripe_size is more than
half the amount of max_chunk_size."
In the previous version of the code, the 16MiB alignment (why is this
done, by the way?) would result in a 50% chance that it would actually
do an 8MiB alignment for the individual dev_extents, since it was
operating on double the size. Does this matter?
Does it matter that stripe_size can be set to anything which is not
16MiB aligned because of the amount of remaining available disk space
which is just taken?
What is the main purpose of this round_up?
The most straightforward thing to do seems something like...
stripe_size = min(
div_u64(devices_info[ndevs - 1].max_avail, dev_stripes),
stripe_size
)
..just putting half of the max_avail into stripe_size."
Link: https://lore.kernel.org/linux-btrfs/b3461a38-e5f8-f41d-c67c-2efac8129054@mendix.com/
Reported-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
[ add analysis from report ]
Signed-off-by: David Sterba <dsterba@suse.com>
We have a complex loop design for find_free_extent(), that has different
behavior for each loop, some even includes new chunk allocation.
Instead of putting such a long code into find_free_extent() and makes it
harder to read, just extract them into find_free_extent_update_loop().
With all the cleanups, the main find_free_extent() should be pretty
barebone:
find_free_extent()
|- Iterate through all block groups
| |- Get a valid block group
| |- Try to do clustered allocation in that block group
| |- Try to do unclustered allocation in that block group
| |- Check if the result is valid
| | |- If valid, then exit
| |- Jump to next block group
|
|- Push harder to find free extents
|- If not found, re-iterate all block groups
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
[ copy callchain from changelog to function comment ]
Signed-off-by: David Sterba <dsterba@suse.com>
This patch will extract unclsutered extent allocation code into
find_free_extent_unclustered().
And this helper function will use return value to indicate what to do
next.
This should make find_free_extent() a little easier to read.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
[Update merge conflict with fb5c39d7a8 ("btrfs: don't use ctl->free_space for max_extent_size")]
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have two main methods to find free extents inside a block group:
1) clustered allocation
2) unclustered allocation
This patch will extract the clustered allocation into
find_free_extent_clustered() to make it a little easier to read.
Instead of jumping between different labels in find_free_extent(), the
helper function will use return value to indicate different behavior.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Instead of tons of different local variables in find_free_extent(),
extract them into find_free_extent_ctl structure, and add better
explanation for them.
Some modification may looks redundant, but will later greatly simplify
function parameter list during find_free_extent() refactor.
Also add two comments to co-operate with fb5c39d7a8 ("btrfs: don't use
ctl->free_space for max_extent_size"), to make ffe_ctl->max_extent_size
update more reader-friendly.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Although we have space_info::bytes_may_use underflow detection in
btrfs_free_reserved_data_space_noquota(), we have more callers who are
subtracting number from space_info::bytes_may_use.
So instead of doing underflow detection for every caller, introduce a
new wrapper update_bytes_may_use() to replace open coded bytes_may_use
modifiers.
This also introduce a macro to declare more wrappers, but currently
space_info::bytes_may_use is the mostly interesting one.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
A completely valid btrfs will refuse to mount, with error message like:
BTRFS critical (device sdb2): corrupt leaf: root=2 block=239681536 slot=172 \
bg_start=12018974720 bg_len=10888413184, invalid block group size, \
have 10888413184 expect (0, 10737418240]
This has been reported several times as the 4.19 kernel is now being
used. The filesystem refuses to mount, but is otherwise ok and booting
4.18 is a workaround.
Btrfs check returns no error, and all kernels used on this fs is later
than 2011, which should all have the 10G size limit commit.
[CAUSE]
For a 12 devices btrfs, we could allocate a chunk larger than 10G due to
stripe stripe bump up.
__btrfs_alloc_chunk()
|- max_stripe_size = 1G
|- max_chunk_size = 10G
|- data_stripe = 11
|- if (1G * 11 > 10G) {
stripe_size = 976128930;
stripe_size = round_up(976128930, SZ_16M) = 989855744
However the final stripe_size (989855744) * 11 = 10888413184, which is
still larger than 10G.
[FIX]
For the comprehensive check, we need to do the full check at chunk read
time, and rely on bg <-> chunk mapping to do the check.
We could just skip the length check for now.
Fixes: fce466eab7 ("btrfs: tree-checker: Verify block_group_item")
Cc: stable@vger.kernel.org # v4.19+
Reported-by: Wang Yugui <wangyugui@e16-tech.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Commit 581c176041 ("btrfs: Validate child tree block's level and first
key") has made tree block level check mandatory.
So if tree block level doesn't match, we won't get a valid extent
buffer. The extra WARN_ON() check can be removed completely.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
And add one line comment explaining what we're doing for each loop.
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Some qgroup trace events like btrfs_qgroup_release_data() and
btrfs_qgroup_free_delayed_ref() can still be triggered even if qgroup is
not enabled.
This is caused by the lack of qgroup status check before calling some
qgroup functions. Thankfully the functions can handle quota disabled
case well and just do nothing for qgroup disabled case.
This patch will do earlier check before triggering related trace events.
And for enabled <-> disabled race case:
1) For enabled->disabled case
Disable will wipe out all qgroups data including reservation and
excl/rfer. Even if we leak some reservation or numbers, it will
still be cleared, so nothing will go wrong.
2) For disabled -> enabled case
Current btrfs_qgroup_release_data() will use extent_io tree to ensure
we won't underflow reservation. And for delayed_ref we use
head->qgroup_reserved to record the reserved space, so in that case
head->qgroup_reserved should be 0 and we won't underflow.
CC: stable@vger.kernel.org # 4.14+
Reported-by: Chris Murphy <lists@colorremedies.com>
Link: https://lore.kernel.org/linux-btrfs/CAJCQCtQau7DtuUUeycCkZ36qjbKuxNzsgqJ7+sJ6W0dK_NLE3w@mail.gmail.com/
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Although we have tree level check at tree read runtime, it's completely
based on its parent level.
We still need to do accurate level check to avoid invalid tree blocks
sneak into kernel space.
The check itself is simple, for leaf its level should always be 0.
For nodes its level should be in range [1, BTRFS_MAX_LEVEL - 1].
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Su Yue <suy.fnst@cn.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
For qgroup_trace_extent_swap(), if we find one leaf that needs to be
traced, we will also iterate all file extents and trace them.
This is OK if we're relocating data block groups, but if we're
relocating metadata block groups, balance code itself has ensured that
both subtree of file tree and reloc tree contain the same contents.
That's to say, if we're relocating metadata block groups, all file
extents in reloc and file tree should match, thus no need to trace them.
This should reduce the total number of dirty extents processed in metadata
block group balance.
[[Benchmark]] (with all previous enhancement)
Hardware:
VM 4G vRAM, 8 vCPUs,
disk is using 'unsafe' cache mode,
backing device is SAMSUNG 850 evo SSD.
Host has 16G ram.
Mkfs parameter:
--nodesize 4K (To bump up tree size)
Initial subvolume contents:
4G data copied from /usr and /lib.
(With enough regular small files)
Snapshots:
16 snapshots of the original subvolume.
each snapshot has 3 random files modified.
balance parameter:
-m
So the content should be pretty similar to a real world root fs layout.
| v4.19-rc1 | w/ patchset | diff (*)
---------------------------------------------------------------
relocated extents | 22929 | 22851 | -0.3%
qgroup dirty extents | 227757 | 140886 | -38.1%
time (sys) | 65.253s | 37.464s | -42.6%
time (real) | 74.032s | 44.722s | -39.6%
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Qu Wenruo