- 16 May, 2022 40 commits
-
-
Qu Wenruo authored
[BUG] There is a report that a btrfs has a bad super block num devices. This makes btrfs to reject the fs completely. BTRFS error (device sdd3): super_num_devices 3 mismatch with num_devices 2 found here BTRFS error (device sdd3): failed to read chunk tree: -22 BTRFS error (device sdd3): open_ctree failed [CAUSE] During btrfs device removal, chunk tree and super block num devs are updated in two different transactions: btrfs_rm_device() |- btrfs_rm_dev_item(device) | |- trans = btrfs_start_transaction() | | Now we got transaction X | | | |- btrfs_del_item() | | Now device item is removed from chunk tree | | | |- btrfs_commit_transaction() | Transaction X got committed, super num devs untouched, | but device item removed from chunk tree. | (AKA, super num devs is already incorrect) | |- cur_devices->num_devices--; |- cur_devices->total_devices--; |- btrfs_set_super_num_devices() All those operations are not in transaction X, thus it will only be written back to disk in next transaction. So after the transaction X in btrfs_rm_dev_item() committed, but before transaction X+1 (which can be minutes away), a power loss happen, then we got the super num mismatch. This has been fixed by commit bbac5869 ("btrfs: remove device item and update super block in the same transaction"). [FIX] Make the super_num_devices check less strict, converting it from a hard error to a warning, and reset the value to a correct one for the current or next transaction commit. As the number of device items is the critical information where the super block num_devices is only a cached value (and also useful for cross checking), it's safe to automatically update it. Other device related problems like missing device are handled after that and may require other means to resolve, like degraded mount. With this fix, potentially affected filesystems won't fail mount and require the manual repair by btrfs check. Reported-by:Luca Béla Palkovics <luca.bela.palkovics@gmail.com> Link: https://lore.kernel.org/linux-btrfs/CA+8xDSpvdm_U0QLBAnrH=zqDq_cWCOH5TiV46CKmp3igr44okQ@mail.gmail.com/ CC: stable@vger.kernel.org # 4.14+ Signed-off-by:
Qu Wenruo <wqu@suse.com> Reviewed-by:
David Sterba <dsterba@suse.com> Signed-off-by:
-
Goldwyn Rodrigues authored
Parameter struct compressed_bio is not used by the function submit_compressed_bio(). Remove it. Signed-off-by:
Goldwyn Rodrigues <rgoldwyn@suse.com> Reviewed-by:
-
Filipe Manana authored
When doing a NOCOW write, either through direct IO or buffered IO, we do two lookups for the block group that contains the target extent: once when we call btrfs_inc_nocow_writers() and then later again when we call btrfs_dec_nocow_writers() after creating the ordered extent. The lookups require taking a lock and navigating the red black tree used to track all block groups, which can take a non-negligible amount of time for a large filesystem with thousands of block groups, as well as lock contention and cache line bouncing. Improve on this by having a single block group search: making btrfs_inc_nocow_writers() return the block group to its caller and then have the caller pass that block group to btrfs_dec_nocow_writers(). This is part of a patchset comprised of the following patches: btrfs: remove search start argument from first_logical_byte() btrfs: use rbtree with leftmost node cached for tracking lowest block group btrfs: use a read/write lock for protecting the block groups tree btrfs: return block group directly at btrfs_next_block_group() btrfs: avoid double search for block group during NOCOW writes The following test was used to test these changes from a performance perspective: $ cat test.sh #!/bin/bash modprobe null_blk nr_devices=0 NULL_DEV_PATH=/sys/kernel/config/nullb/nullb0 mkdir $NULL_DEV_PATH if [ $? -ne 0 ]; then echo "Failed to create nullb0 directory." exit 1 fi echo 2 > $NULL_DEV_PATH/submit_queues echo 16384 > $NULL_DEV_PATH/size # 16G echo 1 > $NULL_DEV_PATH/memory_backed echo 1 > $NULL_DEV_PATH/power DEV=/dev/nullb0 MNT=/mnt/nullb0 LOOP_MNT="$MNT/loop" MOUNT_OPTIONS="-o ssd -o nodatacow" MKFS_OPTIONS="-R free-space-tree -O no-holes" cat <<EOF > /tmp/fio-job.ini [io_uring_writes] rw=randwrite fsync=0 fallocate=posix group_reporting=1 direct=1 ioengine=io_uring iodepth=64 bs=64k filesize=1g runtime=300 time_based directory=$LOOP_MNT numjobs=8 thread EOF echo performance | \ tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor echo echo "Using config:" echo cat /tmp/fio-job.ini echo umount $MNT &> /dev/null mkfs.btrfs -f $MKFS_OPTIONS $DEV &> /dev/null mount $MOUNT_OPTIONS $DEV $MNT mkdir $LOOP_MNT truncate -s 4T $MNT/loopfile mkfs.btrfs -f $MKFS_OPTIONS $MNT/loopfile &> /dev/null mount $MOUNT_OPTIONS $MNT/loopfile $LOOP_MNT # Trigger the allocation of about 3500 data block groups, without # actually consuming space on underlying filesystem, just to make # the tree of block group large. fallocate -l 3500G $LOOP_MNT/filler fio /tmp/fio-job.ini umount $LOOP_MNT umount $MNT echo 0 > $NULL_DEV_PATH/power rmdir $NULL_DEV_PATH The test was run on a non-debug kernel (Debian's default kernel config), the result were the following. Before patchset: WRITE: bw=1455MiB/s (1526MB/s), 1455MiB/s-1455MiB/s (1526MB/s-1526MB/s), io=426GiB (458GB), run=300006-300006msec After patchset: WRITE: bw=1503MiB/s (1577MB/s), 1503MiB/s-1503MiB/s (1577MB/s-1577MB/s), io=440GiB (473GB), run=300006-300006msec +3.3% write throughput and +3.3% IO done in the same time period. The test has somewhat limited coverage scope, as with only NOCOW writes we get less contention on the red black tree of block groups, since we don't have the extra contention caused by COW writes, namely when allocating data extents, pinning and unpinning data extents, but on the hand there's access to tree in the NOCOW path, when incrementing a block group's number of NOCOW writers. Reviewed-by:Nikolay Borisov <nborisov@suse.com> Signed-off-by:
Filipe Manana <fdmanana@suse.com> Signed-off-by:
-
Filipe Manana authored
At btrfs_next_block_group(), we have this long line with two statements: cache = btrfs_lookup_first_block_group(...); return cache; This makes it a bit harder to read due to two statements on the same line, so change that to directly return the result of the call to btrfs_lookup_first_block_group(). Reviewed-by:
-
Filipe Manana authored
Currently we use a spin lock to protect the red black tree that we use to track block groups. Most accesses to that tree are actually read only and for large filesystems, with thousands of block groups, it actually has a bad impact on performance, as concurrent read only searches on the tree are serialized. Read only searches on the tree are very frequent and done when: 1) Pinning and unpinning extents, as we need to lookup the respective block group from the tree; 2) Freeing the last reference of a tree block, regardless if we pin the underlying extent or add it back to free space cache/tree; 3) During NOCOW writes, both buffered IO and direct IO, we need to check if the block group that contains an extent is read only or not and to increment the number of NOCOW writers in the block group. For those operations we need to search for the block group in the tree. Similarly, after creating the ordered extent for the NOCOW write, we need to decrement the number of NOCOW writers from the same block group, which requires searching for it in the tree; 4) Decreasing the number of extent reservations in a block group; 5) When allocating extents and freeing reserved extents; 6) Adding and removing free space to the free space tree; 7) When releasing delalloc bytes during ordered extent completion; 8) When relocating a block group; 9) During fitrim, to iterate over the block groups; 10) etc; Write accesses to the tree, to add or remove block groups, are much less frequent as they happen only when allocating a new block group or when deleting a block group. We also use the same spin lock to protect the list of currently caching block groups. Additions to this list are made when we need to cache a block group, because we don't have a free space cache for it (or we have but it's invalid), and removals from this list are done when caching of the block group's free space finishes. These cases are also not very common, but when they happen, they happen only once when the filesystem is mounted. So switch the lock that protects the tree of block groups from a spinning lock to a read/write lock. Reviewed-by:
-
Filipe Manana authored
We keep track of the start offset of the block group with the lowest start offset at fs_info->first_logical_byte. This requires explicitly updating that field every time we add, delete or lookup a block group to/from the red black tree at fs_info->block_group_cache_tree. Since the block group with the lowest start address happens to always be the one that is the leftmost node of the tree, we can use a red black tree that caches the left most node. Then when we need the start address of that block group, we can just quickly get the leftmost node in the tree and extract the start offset of that node's block group. This avoids the need to explicitly keep track of that address in the dedicated member fs_info->first_logical_byte, and it also allows the next patch in the series to switch the lock that protects the red black tree from a spin lock to a read/write lock - without this change it would be tricky because block group searches also update fs_info->first_logical_byte. Reviewed-by:
-
Filipe Manana authored
The search start argument passed to first_logical_byte() is always 0, as we always want to get the logical start address of the block group with the lowest logical start address. So remove it, as not only it is not necessary, it also makes the following patches that change the lock that protects the red black tree of block groups from a spin lock to a read/write lock. Reviewed-by:
-
Qu Wenruo authored
[BUG] If we hit an error from submit_extent_page() inside __extent_writepage_io(), we could still return 0 to the caller, and even trigger the warning in btrfs_page_assert_not_dirty(). [CAUSE] In __extent_writepage_io(), if we hit an error from submit_extent_page(), we will just clean up the range and continue. This is completely fine for regular PAGE_SIZE == sectorsize, as we can only hit one sector in one page, thus after the error we're ensured to exit and @ret will be saved. But for subpage case, we may have other dirty subpage range in the page, and in the next loop, we may succeeded submitting the next range. In that case, @ret will be overwritten, and we return 0 to the caller, while we have hit some error. [FIX] Introduce @has_error and @saved_ret to record the first error we hit, so we will never forget what error we hit. CC: stable@vger.kernel.org # 5.15+ Signed-off-by:
-
Qu Wenruo authored
[BUG] Test case generic/475 have a very high chance (almost 100%) to hit a fs hang, where a data page will never be unlocked and hang all later operations. [CAUSE] In btrfs_do_readpage(), if we hit an error from submit_extent_page() we will try to do the cleanup for our current io range, and exit. This works fine for PAGE_SIZE == sectorsize cases, but not for subpage. For subpage btrfs_do_readpage() will lock the full page first, which can contain several different sectors and extents: btrfs_do_readpage() |- begin_page_read() | |- btrfs_subpage_start_reader(); | Now the page will have PAGE_SIZE / sectorsize reader pending, | and the page is locked. | |- end_page_read() for different branches | This function will reduce subpage readers, and when readers | reach 0, it will unlock the page. But when submit_extent_page() failed, we only cleanup the current io range, while the remaining io range will never be cleaned up, and the page remains locked forever. [FIX] Update the error handling of submit_extent_page() to cleanup all the remaining subpage range before exiting the loop. Please note that, now submit_extent_page() can only fail due to sanity check in alloc_new_bio(). Thus regular IO errors are impossible to trigger the error path. CC: stable@vger.kernel.org # 5.15+ Signed-off-by:
-
Qu Wenruo authored
[BUG] When running generic/475 with 64K page size and 4K sector size, it has a very high chance (almost 100%) to hang, with mostly data page locked but no one is going to unlock it. [CAUSE] With commit 1784b7d5 ("btrfs: handle csum lookup errors properly on reads"), if we failed to lookup checksum due to metadata IO error, we will return error for btrfs_submit_data_bio(). This will cause the page to be unlocked twice in btrfs_do_readpage(): btrfs_do_readpage() |- submit_extent_page() | |- submit_one_bio() | |- btrfs_submit_data_bio() | |- if (ret) { | |- bio->bi_status = ret; | |- bio_endio(bio); } | In the endio function, we will call end_page_read() | and unlock_extent() to cleanup the subpage range. | |- if (ret) { |- unlock_extent(); end_page_read() } Here we unlock the extent and cleanup the subpage range again. For unlock_extent(), it's mostly double unlock safe. But for end_page_read(), it's not, especially for subpage case, as for subpage case we will call btrfs_subpage_end_reader() to reduce the reader number, and use that to number to determine if we need to unlock the full page. If double accounted, it can underflow the number and leave the page locked without anyone to unlock it. [FIX] The commit 1784b7d5 ("btrfs: handle csum lookup errors properly on reads") itself is completely fine, it's our existing code not properly handling the error from bio submission hook properly. This patch will make submit_one_bio() to return void so that the callers will never be able to do cleanup when bio submission hook fails. Signed-off-by:
-
Schspa Shi authored
This is an optimization for fix fee13fe9 ("btrfs: correct zstd workspace manager lock to use spin_lock_bh()") The critical region for wsm.lock is only accessed by the process context and the softirq context. Because in the soft interrupt, the critical section will not be preempted by the soft interrupt again, there is no need to call spin_lock_bh(&wsm.lock) to turn off the soft interrupt, spin_lock(&wsm.lock) is enough for this situation. Signed-off-by:
Schspa Shi <schspa@gmail.com> [ minor comment update ] Signed-off-by:
-
Filipe Manana authored
We are still using the magic value of 2 at btrfs_create_new_inode(), but there's now a constant for that, named BTRFS_DIR_START_INDEX, which was introduced in commit 528ee697 ("btrfs: put initial index value of a directory in a constant"). So change that to use the constant. Reviewed-by:
-
Qu Wenruo authored
Cleanup the function submit_read_repair() by: - Remove the fixed argument submit_bio_hook() The function is only called on buffered data read path, so the @submit_bio_hook argument is always btrfs_submit_data_bio(). Since it's fixed, then there is no need to pass that argument at all. - Rename the function to submit_data_read_repair() Just to be more explicit on all the 3 things, data, read and repair. Reviewed-by:
-
Christoph Hellwig authored
Reading a value from a different member of a union is not just a great way to obfuscate code, but also creates an aliasing violation. Switch btrfs_is_zoned to look at ->zone_size and remove the union. Note: union was to simplify the detection of zoned filesystem but now this is wrapped behind btrfs_is_zoned so we can drop the union. Reviewed-by:
Johannes Thumshirn <johannes.thumshirn@wdc.com> Reviewed-by:
Naohiro Aota <naohiro.aota@wdc.com> Signed-off-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
-
Lv Ruyi authored
iput() already handles NULL and non-NULL parameter, so it is not needed to check that. This unifies all iput calls. Reported-by:
Zeal Robot <zealci@zte.com.cn> Signed-off-by:
Lv Ruyi <lv.ruyi@zte.com.cn> Reviewed-by:
-
Christoph Hellwig authored
The bios added to ->bio_list are the original bios fed into btrfs_map_bio, which are never advanced. Just use the iter in the bio itself. Reviewed-by:
-
Christoph Hellwig authored
All the scrub bios go straight to the block device or the raid56 code, none of which looks at the btrfs_bio. Reviewed-by:
-
Christoph Hellwig authored
Except for the spurious initialization of ->device just after allocation nothing uses the btrfs_bio, so just allocate a normal bio without extra data. Reviewed-by:
-
Christoph Hellwig authored
Prepare for further refactoring by moving this initialization to a single place instead of setting it in the callers. Signed-off-by:
-
Christoph Hellwig authored
Pass the block_device to bio_alloc_clone instead of setting it later. Signed-off-by:
-
Christoph Hellwig authored
Prepare for additional refactoring, btrfs_map_bio is direct caller of submit_stripe_bio. Signed-off-by:
-
Christoph Hellwig authored
The I/O in repair_io_failue is synchronous and doesn't need a btrfs_bio, so just use an on-stack bio. Reviewed-by:
-
Christoph Hellwig authored
The I/O in repair_io_failue is synchronous and doesn't need a btrfs_bio, so just use an on-stack bio. Reviewed-by:
-
Christoph Hellwig authored
The I/O in repair_io_failue is synchronous and doesn't need a btrfs_bio, so just use an on-stack bio. Also cleanup the error handling to use goto labels and not discard the actual return values. Reviewed-by:
-
Christoph Hellwig authored
btrfsic_read_block does not need the btrfs_bio structure, so switch to plain bio_alloc (that also does not fail as it's backed by a bioset). Reviewed-by:
-
Christoph Hellwig authored
Require a separate call to the integrity checking helpers from the actual bio submission. Reviewed-by:
-
Christoph Hellwig authored
Split out two helpers to make __btrfsic_submit_bio more readable. Reviewed-by:
-
Johannes Thumshirn authored
The current auto-reclaim algorithm starts reclaiming all block groups with a zone_unusable value above a configured threshold. This is causing a lot of reclaim IO even if there would be enough free zones on the device. Instead of only accounting a block groups zone_unusable value, also take the ratio of free and not usable (written as well as zone_unusable) bytes a device has into account. Tested-by:
Pankaj Raghav <p.raghav@samsung.com> Signed-off-by:
-
Josef Bacik authored
For the non-zoned case we may want to set the threshold for reclaim to something below 50%. Change the acceptable threshold from 50-100 to 0-100. Tested-by:
Josef Bacik <josef@toxicpanda.com> Signed-off-by:
-
Josef Bacik authored
This will allow us to set a threshold for block groups to be automatically relocated even if we don't have zoned devices. We have found this feature invaluable at Facebook due to how our workload interacts with the allocator. We have been using this in production for months with only a single problem that has already been fixed. Tested-by:
-
Josef Bacik authored
For non-zoned file systems it's useful to have the auto reclaim feature, however there are different use cases for non-zoned, for example we may not want to reclaim metadata chunks ever, only data chunks. Move this sysfs flag to per-space_info. This won't affect current users because this tunable only ever did anything for zoned, and that is currently hidden behind BTRFS_CONFIG_DEBUG. Tested-by:
-
Filipe Manana authored
When checking if we can do a NOCOW write against a range covered by a file extent item, we do a quick a check to determine if the inode's root was snapshotted in a generation older than the generation of the file extent item or not. This is to quickly determine if the extent is likely shared and avoid the expensive check for cross references (this was added in commit 78d4295b ("btrfs: lift some btrfs_cross_ref_exist checks in nocow path"). We restrict that check to the case where the inode is not a free space inode (since commit 27a7ff55 ("btrfs: skip file_extent generation check for free_space_inode in run_delalloc_nocow")). That is because when we had the inode cache feature, inode caches were backed by a free space inode that belonged to the inode's root. However we don't have support for the inode cache feature since kernel 5.11, so we don't need this check anymore since free space inodes are now always related to free space caches, which are always associated to the root tree (which can't be snapshotted, and its last_snapshot field is always 0). So remove that condition. Signed-off-by:
-
Filipe Manana authored
Verifying if we can do a NOCOW write against a range fully or partially covered by a file extent item requires verifying several constraints, and these are currently duplicated at two different places: can_nocow_extent() and run_delalloc_nocow(). This change moves those checks into a common helper function to avoid duplication. It adds some comments and also preserves all existing behaviour like for example can_nocow_extent() treating errors from the calls to btrfs_cross_ref_exist() and csum_exist_in_range() as meaning we can not NOCOW, instead of propagating the error back to the caller. That specific behaviour is questionable but also reasonable to some degree. Signed-off-by:
-
Sweet Tea Dorminy authored
When allocating memory in a loop, each iteration should call memalloc_retry_wait() in order to prevent starving memory-freeing processes (and to mark where allocation loops are). Other filesystems do that as well. The bulk page allocation is the only place in btrfs with an allocation retry loop, so add an appropriate call to it. Signed-off-by:
Sweet Tea Dorminy <sweettea-kernel@dorminy.me> Reviewed-by:
-
Sweet Tea Dorminy authored
While calling alloc_page() in a loop is an effective way to populate an array of pages, the MM subsystem provides a method to allocate pages in bulk. alloc_pages_bulk_array() populates the NULL slots in a page array, trying to grab more than one page at a time. Unfortunately, it doesn't guarantee allocating all slots in the array, but it's easy to call it in a loop and return an error if no progress occurs. Reviewed-by:
-
Sweet Tea Dorminy authored
Several functions currently populate an array of page pointers one allocated page at a time. Factor out the common code so as to allow improvements to all of the sites at once. Reviewed-by:
-
Yu Zhe authored
Explicit type casts are not necessary when it's void* to another pointer type. Signed-off-by:
Yu Zhe <yuzhe@nfschina.com> Reviewed-by:
-
Qu Wenruo authored
With the recent change in metadata handling, we can handle metadata in the following cases: - nodesize < PAGE_SIZE and sectorsize < PAGE_SIZE Go subpage routine for both metadata and data. - nodesize < PAGE_SIZE and sectorsize >= PAGE_SIZE Invalid case for now. As we require nodesize >= sectorsize. - nodesize >= PAGE_SIZE and sectorsize < PAGE_SIZE Go subpage routine for data, but regular page routine for metadata. - nodesize >= PAGE_SIZE and sectorsize >= PAGE_SIZE Go regular page routine for both metadata and data. Now we can handle any sectorsize < PAGE_SIZE, plus the existing sectorsize == PAGE_SIZE support. But here we introduce an artificial limit, any PAGE_SIZE > 4K case, we will only support 4K and PAGE_SIZE as sector size. The idea here is to reduce the test combinations, and push 4K as the default standard in the future. Signed-off-by:
-
Qu Wenruo authored
The reason why we only support 64K page size for subpage is, for 64K page size we can ensure no matter what the nodesize is, we can fit it into one page. When other page size come, especially like 16K, the limitation is a bit limiting. To remove such limitation, we allow nodesize >= PAGE_SIZE case to go the non-subpage routine. By this, we can allow 4K sectorsize on 16K page size. Although this introduces another smaller limitation, the metadata can not cross page boundary, which is already met by most recent mkfs. Another small improvement is, we can avoid the overhead for metadata if nodesize >= PAGE_SIZE. For 4K sector size and 64K page size/node size, or 4K sector size and 16K page size/node size, we don't need to allocate extra memory for the metadata pages. Please note that, this patch will not yet enable other page size support yet. Signed-off-by:
-
Qu Wenruo authored
In function btrfs_read_sys_array(), we allocate a real extent buffer using btrfs_find_create_tree_block(). Such extent buffer will be even cached into buffer_radix tree, and using btree inode address space. However we only use such extent buffer to enable the accessors, thus we don't even need to bother using real extent buffer, a dummy one is what we really need. And for dummy extent buffer, we no longer need to do any special handling for the first page, as subpage helper is already doing it properly. Signed-off-by:
-
