- 27 Jul, 2020 40 commits
-
-
David Sterba authored
On a filesystem with exhausted metadata, but still enough to start balance, it's possible to hit this error: [324402.053842] BTRFS info (device loop0): 1 enospc errors during balance [324402.060769] BTRFS info (device loop0): balance: ended with status: -28 [324402.172295] BTRFS: error (device loop0) in reset_balance_state:3321: errno=-28 No space left It fails inside reset_balance_state and turns the filesystem to read-only, which is unnecessary and should be fixed too, but the problem is caused by lack for space when the balance item is deleted. This is a one-time operation and from the same rank as unlink that is allowed to use the global block reserve. So do the same for the balance item. Status of the filesystem (100GiB) just after the balance fails: $ btrfs fi df mnt Data, single: total=80.01GiB, used=38.58GiB System, single: total=4.00MiB, used=16.00KiB Metadata, single: total=19.99GiB, used=19.48GiB GlobalReserve, single: total=512.00MiB, used=50.11MiB CC: stable@vger.kernel.org # 4.4+ Reviewed-by:
Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by:
David Sterba <dsterba@suse.com>
-
Qu Wenruo authored
The function btrfs_check_can_nocow() now has two completely different call patterns. For nowait variant, callers don't need to do any cleanup. While for wait variant, callers need to release the lock if they can do nocow write. This is somehow confusing, and is already a problem for the exported btrfs_check_can_nocow(). So this patch will separate the different patterns into different functions. For nowait variant, the function will be called check_nocow_nolock(). For wait variant, the function pair will be btrfs_check_nocow_lock() btrfs_check_nocow_unlock(). Reviewed-by:
Anand Jain <anand.jain@oracle.com> Signed-off-by:
Qu Wenruo <wqu@suse.com> Reviewed-by:
-
Qu Wenruo authored
These two functions have extra conditions that their callers need to meet, and some not-that-common parameters used for return value. So adding some comments may save reviewers some time. Signed-off-by:
-
Qu Wenruo authored
[BUG] When the data space is exhausted, even if the inode has NOCOW attribute, we will still refuse to truncate unaligned range due to ENOSPC. The following script can reproduce it pretty easily: #!/bin/bash dev=/dev/test/test mnt=/mnt/btrfs umount $dev &> /dev/null umount $mnt &> /dev/null mkfs.btrfs -f $dev -b 1G mount -o nospace_cache $dev $mnt touch $mnt/foobar chattr +C $mnt/foobar xfs_io -f -c "pwrite -b 4k 0 4k" $mnt/foobar > /dev/null xfs_io -f -c "pwrite -b 4k 0 1G" $mnt/padding &> /dev/null sync xfs_io -c "fpunch 0 2k" $mnt/foobar umount $mnt Currently this will fail at the fpunch part. [CAUSE] Because btrfs_truncate_block() always reserves space without checking the NOCOW attribute. Since the writeback path follows NOCOW bit, we only need to bother the space reservation code in btrfs_truncate_block(). [FIX] Make btrfs_truncate_block() follow btrfs_buffered_write() to try to reserve data space first, and fall back to NOCOW check only when we don't have enough space. Such always-try-reserve is an optimization introduced in btrfs_buffered_write(), to avoid expensive btrfs_check_can_nocow() call. This patch will export check_can_nocow() as btrfs_check_can_nocow(), and use it in btrfs_truncate_block() to fix the problem. Reported-by:
Martin Doucha <martin.doucha@suse.com> Reviewed-by:
Filipe Manana <fdmanana@suse.com> Reviewed-by:
-
David Sterba authored
Estimated time of removal of the functionality is 5.11, the option will be still parsed but will have no effect. Reasons for deprecation and removal: - very poor naming choice of the mount option, it's supposed to cache and reuse the inode _numbers_, but it sounds a some generic cache for inodes - the only known usecase where this option would make sense is on a 32bit architecture where inode numbers in one subvolume would be exhausted due to 32bit inode::i_ino - the cache is stored on disk, consumes space, needs to be loaded and written back - new inode number allocation is slower due to lookups into the cache (compared to a simple increment which is the default) - uses the free-space-cache code that is going to be deprecated as well in the future Known problems: - since 2011, returning EEXIST when there's not enough space in a page to store all checksums, see commit 4b9465cb ("Btrfs: add mount -o inode_cache") Remaining issues: - if the option was enabled, new inodes created, the option disabled again, the cache is still stored on the devices and there's currently no way to remove it Signed-off-by:
-
David Sterba authored
Last touched in 2013 by commit de78b51a ("btrfs: remove cache only arguments from defrag path") that was the only code that used the value. Now it's only set but never used for anything, so we can remove it. Reviewed-by:
Nikolay Borisov <nborisov@suse.com> Reviewed-by:
-
David Sterba authored
The fiemap callback is not part of UAPI interface and the prototypes don't have the __u64 types either. Reviewed-by:
-
Denis Efremov authored
num_extents is already checked in the next if condition and can be safely removed. Signed-off-by:
Denis Efremov <efremov@linux.com> Reviewed-by:
-
Johannes Thumshirn authored
In btrfs_put_root() we're freeing a btrfs_root's 'node' and 'commit_root' extent buffers manually via kfree(), while we're using free_root_extent_buffers() in the free_root_pointers() function above. free_root_extent_buffers() also NULLs the pointers after freeing, which mitigates potential double frees. Reviewed-by:
-
Nikolay Borisov authored
This function iterates all extents in the extent cluster, make this intention obvious by using a for loop. No functional chanes. Reviewed-by:
-
Nikolay Borisov authored
btrfs_alloc_data_chunk_ondemand and btrfs_free_reserved_data_space_noquota don't really use the guts of the inodes being passed to them. This implies it's not required to call them under extent lock. Move code around in prealloc_file_extent_cluster to do the heavy, data alloc/free operations outside of the lock. This also makes the 'out' label unnecessary, so remove it. Reviewed-by:
-
Nikolay Borisov authored
Extents in the extent cluster are guaranteed to be contiguous as such the hole check inside the loop can never trigger. In fact this check was never functional since it was added in 18513091 ("btrfs: update btrfs_space_info's bytes_may_use timely") which came after the commit introducing clustered/contiguous extents 0257bb82 ("Btrfs: relocate file extents in clusters"). Let's just remove it as it adds noise to the source. Reviewed-by:
-
Nikolay Borisov authored
It has only 4 uses of a vfs_inode for inode_sub_bytes but unifies the interface with the non __ prefixed version. Will also makes converting its callers to btrfs_inode easier. Signed-off-by:
-
Nikolay Borisov authored
Will enable converting btrfs_submit_compressed_write to btrfs_inode more easily. Signed-off-by:
-
Nikolay Borisov authored
It has one VFS and 1 btrfs inode usages but converting it to btrfs_inode interface will allow seamless conversion of its callers. Signed-off-by:
-
Nikolay Borisov authored
It really wants a btrfs_inode and will allow submit_compressed_extents to be completely converted to btrfs_inode in follow up patches. Signed-off-by:
-
Nikolay Borisov authored
It really wants btrfs_inode and not a vfs inode. Signed-off-by:
-
Nikolay Borisov authored
It doesn't use the generic vfs inode for anything use btrfs_inode directly. Signed-off-by:
-
Nikolay Borisov authored
It doesn't use the vfs inode for anything, can just as easily take btrfs_inode. Follow up patches will convert callers as well. Signed-off-by:
-
Nikolay Borisov authored
This is internal btrfs function what really needs the vfs_inode only for igrab and a tracepoint. Signed-off-by:
-
Filipe Manana authored
The 'trans_list' member of an ordered extent was used to keep track of the ordered extents for which a transaction commit had to wait. These were ordered extents that were started and logged by an fsync. However we don't do that anymore and before we stopped doing it we changed the approach to wait for the ordered extents in commit 161c3549 ("Btrfs: change how we wait for pending ordered extents"), which stopped using that list and therefore the 'trans_list' member is not used anymore since that commit. So just remove it since it's doing nothing and making each ordered extent structure waste memory (2 pointers). Reviewed-by:
-
Filipe Manana authored
The 'log_list' member of an ordered extent was used keep track of which ordered extents we needed to wait after logging metadata, but is not used anymore since commit 5636cf7d ("btrfs: remove the logged extents infrastructure"), as we now always wait on ordered extent completion before logging metadata. So just remove it since it's doing nothing and making each ordered extent structure waste more memory (2 pointers). Reviewed-by:
-
David Sterba authored
The CPU and on-disk keys are mapped to two different structures because of the endianness. There's an intermediate buffer used to do the conversion, but this is not necessary when CPU and on-disk endianness match. Add optimized versions of helpers that take disk_key and use the buffer directly for CPU keys or drop the intermediate buffer and conversion. This saves a lot of stack space accross many functions and removes about 6K of generated binary code: text data bss dec hex filename 1090439 17468 14912 1122819 112203 pre/btrfs.ko 1084613 17456 14912 1116981 110b35 post/btrfs.ko Delta: -5826 Reviewed-by:
-
Qu Wenruo authored
Before this patch, qgroup completely relies on per-inode extent io tree to detect reserved data space leak. However previous bug has already shown how release page before btrfs_finish_ordered_io() could lead to leak, and since it's QGROUP_RESERVED bit cleared without triggering qgroup rsv, it can't be detected by per-inode extent io tree. So this patch adds another (and hopefully the final) safety net to catch qgroup data reserved space leak. At least the new safety net catches all the leaks during development, so it should be pretty useful in the real world. Reviewed-by:
Josef Bacik <josef@toxicpanda.com> Signed-off-by:
-
Qu Wenruo authored
[BUG] The following simple workload from fsstress can lead to qgroup reserved data space leak: 0/0: creat f0 x:0 0 0 0/0: creat add id=0,parent=-1 0/1: write f0[259 1 0 0 0 0] [600030,27288] 0 0/4: dwrite - xfsctl(XFS_IOC_DIOINFO) f0[259 1 0 0 64 627318] return 25, fallback to stat() 0/4: dwrite f0[259 1 0 0 64 627318] [610304,106496] 0 This would cause btrfs qgroup to leak 20480 bytes for data reserved space. If btrfs qgroup limit is enabled, such leak can lead to unexpected early EDQUOT and unusable space. [CAUSE] When doing direct IO, kernel will try to writeback existing buffered page cache, then invalidate them: generic_file_direct_write() |- filemap_write_and_wait_range(); |- invalidate_inode_pages2_range(); However for btrfs, the bi_end_io hook doesn't finish all its heavy work right after bio ends. In fact, it delays its work further: submit_extent_page(end_io_func=end_bio_extent_writepage); end_bio_extent_writepage() |- btrfs_writepage_endio_finish_ordered() |- btrfs_init_work(finish_ordered_fn); <<< Work queue execution >>> finish_ordered_fn() |- btrfs_finish_ordered_io(); |- Clear qgroup bits This means, when filemap_write_and_wait_range() returns, btrfs_finish_ordered_io() is not guaranteed to be executed, thus the qgroup bits for related range are not cleared. Now into how the leak happens, this will only focus on the overlapping part of buffered and direct IO part. 1. After buffered write The inode had the following range with QGROUP_RESERVED bit: 596 616K |///////////////| Qgroup reserved data space: 20K 2. Writeback part for range [596K, 616K) Write back finished, but btrfs_finish_ordered_io() not get called yet. So we still have: 596K 616K |///////////////| Qgroup reserved data space: 20K 3. Pages for range [596K, 616K) get released This will clear all qgroup bits, but don't update the reserved data space. So we have: 596K 616K | | Qgroup reserved data space: 20K That number doesn't match the qgroup bit range anymore. 4. Dio prepare space for range [596K, 700K) Qgroup reserved data space for that range, we got: 596K 616K 700K |///////////////|///////////////////////| Qgroup reserved data space: 20K + 104K = 124K 5. btrfs_finish_ordered_range() gets executed for range [596K, 616K) Qgroup free reserved space for that range, we got: 596K 616K 700K | |///////////////////////| We need to free that range of reserved space. Qgroup reserved data space: 124K - 20K = 104K 6. btrfs_finish_ordered_range() gets executed for range [596K, 700K) However qgroup bit for range [596K, 616K) is already cleared in previous step, so we only free 84K for qgroup reserved space. 596K 616K 700K | | | We need to free that range of reserved space. Qgroup reserved data space: 104K - 84K = 20K Now there is no way to release that 20K unless disabling qgroup or unmounting the fs. [FIX] This patch will change the timing of btrfs_qgroup_release/free_data() call. Here it uses buffered COW write as an example. The new timing | The old timing ----------------------------------------+--------------------------------------- btrfs_buffered_write() | btrfs_buffered_write() |- btrfs_qgroup_reserve_data() | |- btrfs_qgroup_reserve_data() | btrfs_run_delalloc_range() | btrfs_run_delalloc_range() |- btrfs_add_ordered_extent() | |- btrfs_qgroup_release_data() | The reserved is passed into | btrfs_ordered_extent structure | | btrfs_finish_ordered_io() | btrfs_finish_ordered_io() |- The reserved space is passed to | |- btrfs_qgroup_release_data() btrfs_qgroup_record | The resereved space is passed | to btrfs_qgroup_recrod | btrfs_qgroup_account_extents() | btrfs_qgroup_account_extents() |- btrfs_qgroup_free_refroot() | |- btrfs_qgroup_free_refroot() The point of such change is to ensure, when ordered extents are submitted, the qgroup reserved space is already released, to keep the timing aligned with file_write_and_wait_range(). So that qgroup data reserved space is all bound to btrfs_ordered_extent and solve the timing mismatch. Fixes: f695fdce ("btrfs: qgroup: Introduce functions to release/free qgroup reserve data space") Suggested-by: -
Qu Wenruo authored
The incoming qgroup reserved space timing will move the data reservation to ordered extent completely. However in btrfs_punch_hole_lock_range() will call btrfs_invalidate_page(), which will clear QGROUP_RESERVED bit for the range. In current stage it's OK, but if we're making ordered extents handle the reserved space, then btrfs_punch_hole_lock_range() can clear the QGROUP_RESERVED bit before we submit ordered extent, leading to qgroup reserved space leakage. So here change the timing to make reserve data space after btrfs_punch_hole_lock_range(). The new timing is fine for either current code or the new code. Reviewed-by:
-
Qu Wenruo authored
This is to prepare for the incoming timing change of qgroup reserved data space and ordered extent. Reviewed-by:
-
Qu Wenruo authored
Function insert_reserved_file_extent() takes a long list of parameters, which are all for btrfs_file_extent_item, even including two reserved members, encryption and other_encoding. This makes the parameter list unnecessary long for a function which only gets called twice. This patch will refactor the parameter list, by using btrfs_file_extent_item as parameter directly to hugely reduce the number of parameters. Also, since there are only two callers, one in btrfs_finish_ordered_io() which inserts file extent for ordered extent, and one __btrfs_prealloc_file_range(). These two call sites have completely different context, where ordered extent can be compressed, but will always be regular extent, while the preallocated one is never going to be compressed and always has PREALLOC type. So use two small wrapper for these two different call sites to improve readability. Reviewed-by:
-
David Sterba authored
Add proper variable for the scrub page and use it instead of repeatedly dereferencing the other structures. Signed-off-by: -
David Sterba authored
Use a simpler iteration over tree block pages, same what csum_tree_block does: first page always exists, loop over the rest. Signed-off-by: -
David Sterba authored
Add proper variable for the scrub page and use it instead of repeatedly dereferencing the other structures. Signed-off-by: -
David Sterba authored
We have sectorsize same as PAGE_SIZE, the checksum can be calculated in one go. Signed-off-by: -
David Sterba authored
Add proper variable for the scrub page and use it instead of repeatedly dereferencing the other structures. Signed-off-by: -
David Sterba authored
The page contents with the checksum is available during the entire function so we don't need to make a copy. Signed-off-by: -
David Sterba authored
BTRFS_SUPER_INFO_SIZE is 4096, and fits to a page on all supported architectures, so we can calculate the checksum in one go. Signed-off-by: -
David Sterba authored
As the page mapping has been removed, rename the variables to 'kaddr' that we use everywhere else. The type is changed to 'char *' so pointer arithmetic works without casts. Signed-off-by: -
David Sterba authored
All pages that scrub uses in the scrub_block::pagev array are allocated with GFP_KERNEL and never part of any mapping, so kmap is not necessary, we only need to know the page address. In scrub_write_page_to_dev_replace we don't even need to call flush_dcache_page because of the same reason as above. Signed-off-by: -
Qu Wenruo authored
This patch introduces a new "rescue=" mount option group for all mount options for data recovery. Different rescue sub options are seperated by ':'. E.g "ro,rescue=nologreplay:usebackuproot". The original plan was to use ';', but ';' needs to be escaped/quoted, or it will be interpreted by bash, similar to '|'. And obviously, user can specify rescue options one by one like: "ro,rescue=nologreplay,rescue=usebackuproot". The following mount options are converted to "rescue=", old mount options are deprecated but still available for compatibility purpose: - usebackuproot Now it's "rescue=usebackuproot" - nologreplay Now it's "rescue=nologreplay" Signed-off-by:
-
Filipe Manana authored
We currently use btrfs_check_data_free_space() when allocating space for relocating data extents, but that is not necessary because that function combines btrfs_alloc_data_chunk_ondemand(), which does the actual space reservation, and btrfs_qgroup_reserve_data(). We can use btrfs_alloc_data_chunk_ondemand() directly because we know we do not need to reserve qgroup space since we are dealing with a relocation tree, which can never have qgroups (btrfs_qgroup_reserve_data() does nothing as is_fstree() returns false for a relocation tree). Conversely we can use btrfs_free_reserved_data_space_noquota() directly instead of btrfs_free_reserved_data_space(), since we had no qgroup reservation when allocating space. This change is preparatory work for another patch in this series that makes relocation reserve the exact amount of space it needs to relocate a data block group. The function btrfs_check_data_free_space() has the incovenient of requiring a start offset argument and we will want to be able to allocate space for multiple ranges, which are not consecutive, at once. Reviewed-by:
-
Filipe Manana authored
The start argument for btrfs_free_reserved_data_space_noquota() is only used to make sure the amount of bytes we decrement from the bytes_may_use counter of the data space_info object is aligned to the filesystem's sector size. It serves no other purpose. All its current callers always pass a length argument that is already aligned to the sector size, so we can make the start argument go away. In fact its presence makes it impossible to use it in a context where we just want to free a number of bytes for a range for which either we do not know its start offset or for freeing multiple ranges at once (which are not contiguous). This change is preparatory work for a patch (third patch in this series) that makes relocation of data block groups that are not full reserve less data space. Reviewed-by:
-
