- 08 Feb, 2021 40 commits
-
-
Qu Wenruo authored
Introduce the following functions to handle subpage error status: - btrfs_subpage_set_error() - btrfs_subpage_clear_error() - btrfs_subpage_test_error() These helpers can only be called when the page has subpage attached and the range is ensured to be inside the page. - btrfs_page_set_error() - btrfs_page_clear_error() - btrfs_page_test_error() These helpers can handle both regular sector size and subpage without problem. Reviewed-by:
Josef Bacik <josef@toxicpanda.com> Signed-off-by:
Qu Wenruo <wqu@suse.com> Reviewed-by:
David Sterba <dsterba@suse.com> Signed-off-by:
-
Qu Wenruo authored
Introduce the following functions to handle subpage uptodate status: - btrfs_subpage_set_uptodate() - btrfs_subpage_clear_uptodate() - btrfs_subpage_test_uptodate() These helpers can only be called when the page has subpage attached and the range is ensured to be inside the page. - btrfs_page_set_uptodate() - btrfs_page_clear_uptodate() - btrfs_page_test_uptodate() These helpers can handle both regular sector size and subpage. Although caller should still ensure that the range is inside the page. Reviewed-by:
-
Qu Wenruo authored
There are locations where we allocate dummy extent buffers for temporary usage, like in tree_mod_log_rewind() or get_old_root(). These dummy extent buffers will be handled by the same eb accessors, and if they don't have page::private subpage eb accessors could fail. To address such problems, make __alloc_dummy_extent_buffer() attach page private for dummy extent buffers too. Reviewed-by:
-
Qu Wenruo authored
In btrfs_release_extent_buffer_pages(), we need to add extra handling for subpage. Introduce a helper, detach_extent_buffer_page(), to do different handling for regular and subpage cases. For subpage case, handle detaching page private. For unmapped (dummy or cloned) ebs, we can detach the page private immediately as the page can only be attached to one unmapped eb. For mapped ebs, we have to ensure there are no eb in the page range before we delete it, as page->private is shared between all ebs in the same page. But there is a subpage specific race, where we can race with extent buffer allocation, and clear the page private while new eb is still being utilized, like this: Extent buffer A is the new extent buffer which will be allocated, while extent buffer B is the last existing extent buffer of the page. T1 (eb A) | T2 (eb B) -------------------------------+------------------------------ alloc_extent_buffer() | btrfs_release_extent_buffer_pages() |- p = find_or_create_page() | | |- attach_extent_buffer_page() | | | | |- detach_extent_buffer_page() | | |- if (!page_range_has_eb()) | | | No new eb in the page range yet | | | As new eb A hasn't yet been | | | inserted into radix tree. | | |- btrfs_detach_subpage() | | |- detach_page_private(); |- radix_tree_insert() | Then we have a metadata eb whose page has no private bit. To avoid such race, we introduce a subpage metadata-specific member, btrfs_subpage::eb_refs. In alloc_extent_buffer() we increase eb_refs in the critical section of private_lock. Then page_range_has_eb() will return true for detach_extent_buffer_page(), and will not detach page private. The section is marked by: - btrfs_page_inc_eb_refs() - btrfs_page_dec_eb_refs() Signed-off-by:
-
Qu Wenruo authored
For subpage case, grab_extent_buffer() can't really get an extent buffer just from btrfs_subpage. We have radix tree lock protecting us from inserting the same eb into the tree. Thus we don't really need to do the extra hassle, just let alloc_extent_buffer() handle the existing eb in radix tree. Now if two ebs are being allocated as the same time, one will fail with -EEIXST when inserting into the radix tree. So for grab_extent_buffer(), just always return NULL for subpage case. Reviewed-by:
-
Qu Wenruo authored
For subpage case, we need to allocate additional memory for each metadata page. So we need to: - Allow attach_extent_buffer_page() to return int to indicate allocation failure - Allow manually pre-allocate subpage memory for alloc_extent_buffer() As we don't want to use GFP_ATOMIC under spinlock, we introduce btrfs_alloc_subpage() and btrfs_free_subpage() functions for this purpose. (The simple wrap for btrfs_free_subpage() is for later convert to kmem_cache. Already internally tested without problem) - Preallocate btrfs_subpage structure for alloc_extent_buffer() We don't want to call memory allocation with spinlock held, so do preallocation before we acquire mapping->private_lock. - Handle subpage and regular case differently in attach_extent_buffer_page() For regular case, no change, just do the usual thing. For subpage case, allocate new memory or use the preallocated memory. For future subpage metadata, we will make use of radix tree to grab extent buffer. Reviewed-by:
-
Qu Wenruo authored
For sectorsize < page size support, we need a structure to record extra status info for each sector of a page. Introduce the skeleton structure, all subpage related code would go to subpage.[ch]. Reviewed-by:
-
Qu Wenruo authored
For the incoming subpage support, UNMAPPED extent buffer will have different behavior in btrfs_release_extent_buffer(). This means we need to set UNMAPPED bit early before calling btrfs_release_extent_buffer(). Currently there is only one caller which relies on btrfs_release_extent_buffer() in its error path while set UNMAPPED bit late: - btrfs_clone_extent_buffer() Make it subpage compatible by setting the UNMAPPED bit early, since we're here, also move the UPTODATE bit early. There is another caller, __alloc_dummy_extent_buffer(), setting UNMAPPED bit late, but that function clean up the allocated page manually, thus no need for any modification. Reviewed-by:
-
Qu Wenruo authored
PAGE_CLEAR_DIRTY and PAGE_SET_WRITEBACK are two defines used in __process_pages_contig(), to let the function know to clear page dirty bit and then set page writeback. However page writeback and dirty bits are conflicting (at least for sector size == PAGE_SIZE case), this means these two have to be always updated together. This means we can merge PAGE_CLEAR_DIRTY and PAGE_SET_WRITEBACK to PAGE_START_WRITEBACK. Reviewed-by:
-
Filipe Manana authored
Often an fsync needs to fallback to a transaction commit for several reasons (to ensure consistency after a power failure, a new block group was allocated or a temporary error such as ENOMEM or ENOSPC happened). In that case the log is marked as needing a full commit and any concurrent tasks attempting to log inodes or commit the log will also fallback to the transaction commit. When this happens they all wait for the task that first started the transaction commit to finish the transaction commit - however they wait until the full transaction commit happens, which is not needed, as they only need to wait for the superblocks to be persisted and not for unpinning all the extents pinned during the transaction's lifetime, which even for short lived transactions can be a few thousand and take some significant amount of time to complete - for dbench workloads I have observed up to 4~5 milliseconds of time spent unpinning extents in the worst cases, and the number of pinned extents was between 2 to 3 thousand. So allow fsync tasks to skip waiting for the unpinning of extents when they call btrfs_commit_transaction() and they were not the task that started the transaction commit (that one has to do it, the alternative would be to offload the transaction commit to another task so that it could avoid waiting for the extent unpinning or offload the extent unpinning to another task). This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit After applying the entire patchset, dbench shows improvements in respect to throughput and latency. The script used to measure it is the following: $ cat dbench-test.sh #!/bin/bash DEV=/dev/sdk MNT=/mnt/sdk MOUNT_OPTIONS="-o ssd" MKFS_OPTIONS="-m single -d single" echo "performance" | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor umount $DEV &> /dev/null mkfs.btrfs -f $MKFS_OPTIONS $DEV mount $MOUNT_OPTIONS $DEV $MNT dbench -D $MNT -t 300 64 umount $MNT The test was run on a physical machine with 12 cores (Intel corei7), 64G of ram, using a NVMe device and a non-debug kernel configuration (Debian's default configuration). Before applying patchset, 32 clients: Operation Count AvgLat MaxLat ---------------------------------------- NTCreateX 9627107 0.153 61.938 Close 7072076 0.001 3.175 Rename 407633 1.222 44.439 Unlink 1943895 0.658 44.440 Deltree 256 17.339 110.891 Mkdir 128 0.003 0.009 Qpathinfo 8725406 0.064 17.850 Qfileinfo 1529516 0.001 2.188 Qfsinfo 1599884 0.002 1.457 Sfileinfo 784200 0.005 3.562 Find 3373513 0.411 30.312 WriteX 4802132 0.053 29.054 ReadX 15089959 0.002 5.801 LockX 31344 0.002 0.425 UnlockX 31344 0.001 0.173 Flush 674724 5.952 341.830 Throughput 1008.02 MB/sec 32 clients 32 procs max_latency=341.833 ms After applying patchset, 32 clients: After patchset, with 32 clients: Operation Count AvgLat MaxLat ---------------------------------------- NTCreateX 9931568 0.111 25.597 Close 7295730 0.001 2.171 Rename 420549 0.982 49.714 Unlink 2005366 0.497 39.015 Deltree 256 11.149 89.242 Mkdir 128 0.002 0.014 Qpathinfo 9001863 0.049 20.761 Qfileinfo 1577730 0.001 2.546 Qfsinfo 1650508 0.002 3.531 Sfileinfo 809031 0.005 5.846 Find 3480259 0.309 23.977 WriteX 4952505 0.043 41.283 ReadX 15568127 0.002 5.476 LockX 32338 0.002 0.978 UnlockX 32338 0.001 2.032 Flush 696017 7.485 228.835 Throughput 1049.91 MB/sec 32 clients 32 procs max_latency=228.847 ms --> +4.1% throughput, -39.6% max latency Before applying patchset, 64 clients: Operation Count AvgLat MaxLat ---------------------------------------- NTCreateX 8956748 0.342 108.312 Close 6579660 0.001 3.823 Rename 379209 2.396 81.897 Unlink 1808625 1.108 131.148 Deltree 256 25.632 172.176 Mkdir 128 0.003 0.018 Qpathinfo 8117615 0.131 55.916 Qfileinfo 1423495 0.001 2.635 Qfsinfo 1488496 0.002 5.412 Sfileinfo 729472 0.007 8.643 Find 3138598 0.855 78.321 WriteX 4470783 0.102 79.442 ReadX 14038139 0.002 7.578 LockX 29158 0.002 0.844 UnlockX 29158 0.001 0.567 Flush 627746 14.168 506.151 Throughput 924.738 MB/sec 64 clients 64 procs max_latency=506.154 ms After applying patchset, 64 clients: Operation Count AvgLat MaxLat ---------------------------------------- NTCreateX 9069003 0.303 43.193 Close 6662328 0.001 3.888 Rename 383976 2.194 46.418 Unlink 1831080 1.022 43.873 Deltree 256 24.037 155.763 Mkdir 128 0.002 0.005 Qpathinfo 8219173 0.137 30.233 Qfileinfo 1441203 0.001 3.204 Qfsinfo 1507092 0.002 4.055 Sfileinfo 738775 0.006 5.431 Find 3177874 0.936 38.170 WriteX 4526152 0.084 39.518 ReadX 14213562 0.002 24.760 LockX 29522 0.002 1.221 UnlockX 29522 0.001 0.694 Flush 635652 14.358 422.039 Throughput 990.13 MB/sec 64 clients 64 procs max_latency=422.043 ms --> +6.8% throughput, -18.1% max latency Reviewed-by:
Filipe Manana <fdmanana@suse.com> Signed-off-by:
-
Filipe Manana authored
Whenever we fsync an inode, if it is a directory, a regular file that was created in the current transaction or has last_unlink_trans set to the generation of the current transaction, we check if any of its ancestor inodes (and the inode itself if it is a directory) can not be logged and need a fallback to a full transaction commit - if so, we return with a value of 1 in order to fallback to a transaction commit. However we often do not need to fallback to a transaction commit because: 1) The ancestor inode is not an immediate parent, and therefore there is not an explicit request to log it and it is not needed neither to guarantee the consistency of the inode originally asked to be logged (fsynced) nor its immediate parent; 2) The ancestor inode was already logged before, in which case any link, unlink or rename operation updates the log as needed. So for these two cases we can avoid an unnecessary transaction commit. Therefore remove check_parent_dirs_for_sync() and add a check at the top of btrfs_log_inode() to make us fallback immediately to a transaction commit when we are logging a directory inode that can not be logged and needs a full transaction commit. All we need to protect is the case where after renaming a file someone fsyncs only the old directory, which would result is losing the renamed file after a log replay. This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit Performance results, after applying all patches, are mentioned in the change log of the last patch. Reviewed-by:
-
Filipe Manana authored
When logging new directory entries of a directory, we log the inodes of new dentries and the inodes of dentries pointing to directories that may have been created in past transactions. For the case of directories we log in full mode, which can be particularly expensive for large directories. We do use btrfs_inode_in_log() to skip already logged inodes, however for that helper to return true, it requires that the log transaction used to log the inode to be already committed. This means that when we have more than one task using the same log transaction we can end up logging an inode multiple times, which is a waste of time and not necessary since the log will be committed by one of the tasks and the others will wait for the log transaction to be committed before returning to user space. So simply replace the use of btrfs_inode_in_log() with the new helper function need_log_inode(), introduced in a previous commit. This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit Performance results, after applying all patches, are mentioned in the change log of the last patch. Reviewed-by:
-
Filipe Manana authored
Some times when we fsync an inode we need to do a full log of all its ancestors (due to unlink, link or rename operations), which can be an expensive operation, specially if the directories are large. However if we find an ancestor directory inode that is already logged in the current transaction, and has no inserted/updated/deleted xattrs since it was last logged, we can skip logging the directory again. We are safe to skip that since we know that for logged directories, any link, unlink or rename operations that implicate the directory will update the log as necessary. So use the helper need_log_dir(), introduced in a previous commit, to detect already logged directories that can be skipped. This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit Performance results, after applying all patches, are mentioned in the change log of the last patch. Reviewed-by:
-
Filipe Manana authored
When we fsync a new file, created in the current transaction, we check all its ancestor inodes and always log them if they were created in the current transaction - even if we have already logged them before, which is a waste of time. So avoid logging new ancestor inodes if they were already logged before and have no xattrs added/updated/removed since they were last logged. This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit Performance results, after applying all patches, are mentioned in the change log of the last patch. Reviewed-by:
-
Filipe Manana authored
When we fill an inode item for logging we are setting its nbytes field with the value returned by inode_get_bytes() (a VFS API), however we do not need it because it is not used during log replay. In fact, for fast fsyncs, when we call inode_get_bytes() we may even get an outdated value for nbytes because the nbytes field of the inode is only updated when ordered extents complete, and a fast fsync only waits for writeback to complete, it does not wait for ordered extent completion. So just remove the setup of nbytes and add an explicit comment mentioning why we do not set it. This also avoids adding contention on the inode's i_lock (VFS) with concurrent stat() calls, since that spinlock is used by inode_get_bytes() which is also called by our stat callback (btrfs_getattr()). This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit Performance results, after applying all patches, are mentioned in the change log of the last patch. Reviewed-by:
-
Filipe Manana authored
When we remove a directory entry, as part of an unlink operation, if the directory was logged before we must remove the directory index items from the log. We are also updating the inode item of the directory to update its i_size, but that is not necessary because during log replay we do not need it and we correctly adjust the i_size in the inode item of the subvolume as we process directory index items and replay deletes. This is not needed since commit d555438b ("Btrfs: drop dir i_size when adding new names on replay"), where we explicitly ignore the i_size of directory inode items on log replay. Before that we used it but it was buggy as mentioned in that commit's change log (i_size got a larger value then it should have). So stop updating the i_size of the directory inode item in the log, as that is a waste of time, adds more log contention to the log tree and often results in COWing more extent buffers for the log tree. This code path is triggered often during dbench workloads for example. This patch is part of a patchset comprised of the following patches: btrfs: remove unnecessary directory inode item update when deleting dir entry btrfs: stop setting nbytes when filling inode item for logging btrfs: avoid logging new ancestor inodes when logging new inode btrfs: skip logging directories already logged when logging all parents btrfs: skip logging inodes already logged when logging new entries btrfs: remove unnecessary check_parent_dirs_for_sync() btrfs: make concurrent fsyncs wait less when waiting for a transaction commit Performance results, after applying all patches, are mentioned in the change log of the last patch. Reviewed-by:
-
Michal Rostecki authored
Before this change, the btrfs_get_io_geometry() function was calling btrfs_get_chunk_map() to get the extent mapping, necessary for calculating the I/O geometry. It was using that extent mapping only internally and freeing the pointer after its execution. That resulted in calling btrfs_get_chunk_map() de facto twice by the __btrfs_map_block() function. It was calling btrfs_get_io_geometry() first and then calling btrfs_get_chunk_map() directly to get the extent mapping, used by the rest of the function. Change that to passing the extent mapping to the btrfs_get_io_geometry() function as an argument. This could improve performance in some cases. For very large filesystems, i.e. several thousands of allocated chunks, not only this avoids searching two times the rbtree, saving time, it may also help reducing contention on the lock that protects the tree - thinking of writeback starting for multiple inodes, other tasks allocating or removing chunks, and anything else that requires access to the rbtree. Reviewed-by:
Michal Rostecki <mrostecki@suse.com> Reviewed-by:
-
Qu Wenruo authored
Commit dbfdb6d1 ("Btrfs: Search for all ordered extents that could span across a page") make btrfs_invalidapage() to search all ordered extents. The offending code looks like this: again: start = page_start; ordered = btrfs_lookup_ordered_range(inode, start, page_end - start + 1); if (ordred) { end = min(page_end, ordered->file_offset + ordered->num_bytes - 1); /* Do the cleanup */ start = end + 1; if (start < page_end) goto again; } The behavior is indeed necessary for the incoming subpage support, but when it iterates through all the ordered extents, it also resets the search range @start. This means, for the following cases, we can double account the ordered extents, causing its bytes_left underflow: Page offset 0 16K 32K |<--- OE 1 --->|<--- OE 2 ---->| As the first iteration will find ordered extent (OE) 1, which doesn't cover the full page, thus after cleanup code, we need to retry again. But again label will reset start to page_start, and we got OE 1 again, which causes double accounting on OE 1, and cause OE 1's byte_left to underflow. This problem can only happen for subpage case, as for regular sectorsize == PAGE_SIZE case, we will always find a OE ends at or after page end, thus no way to trigger the problem. Move the again label after start = page_start. There will be more comprehensive rework to convert the open coded loop to a proper while loop for subpage support. Fixes: dbfdb6d1 ("Btrfs: Search for all ordered extents that could span across a page") Reviewed-by:
-
Abaci Team authored
Fix the following coccicheck warnings: ./fs/btrfs/delayed-inode.c:1157:39-41: WARNING !A || A && B is equivalent to !A || B. Reported-by:
Abaci Robot <abaci@linux.alibaba.com> Suggested-by:
Jiapeng Zhong <oswb@linux.alibaba.com> Reviewed-by:
Abaci Team <abaci-bugfix@linux.alibaba.com> Reviewed-by:
-
Filipe Manana authored
The comment for can_nocow_extent() says that the function will flush ordered extents, however that never happens and was never true before the comment was added in commit e4ecaf90 ("btrfs: add comments for btrfs_check_can_nocow() and can_nocow_extent()"). This is true only for the function btrfs_check_can_nocow(), which after that commit was renamed to check_can_nocow(). So just remove that part of the comment. Reviewed-by:
-
Josef Bacik authored
Often when I'm debugging ENOSPC related issues I have to resort to printing the entire ENOSPC state with trace_printk() in different spots. This gets pretty annoying, so add a trace state that does this for us. Then add a trace point at the end of preemptive flushing so you can see the state of the space_info when we decide to exit preemptive flushing. This helped me figure out we weren't kicking in the preemptive flushing soon enough. Reviewed-by:
Nikolay Borisov <nborisov@suse.com> Signed-off-by:
-
Josef Bacik authored
Since we have normal ticketed flushing and preemptive flushing, adjust the tracepoint so that we know the source of the flushing action to make it easier to debug problems. Reviewed-by:
-
Josef Bacik authored
Starting preemptive flushing at 50% of available free space is a good start, but some workloads are particularly abusive and can quickly overwhelm the preemptive flushing code and drive us into using tickets. Handle this by clamping down on our threshold for starting and continuing to run preemptive flushing. This is particularly important for our overcommit case, as we can really drive the file system into overages and then it's more difficult to pull it back as we start to actually fill up the file system. The clamping is essentially 2^CLAMP, but we start at 1 so whatever we calculate for overcommit is the baseline. Reviewed-by:
-
Josef Bacik authored
A lot of this was added all in one go with no explanation, and is a bit unwieldy and confusing. Simplify the logic to start preemptive flushing if we've reserved more than half of our available free space. Reviewed-by:
-
Josef Bacik authored
Currently btrfs_calc_reclaim_metadata_size does two things, it returns the space currently required for flushing by the tickets, and if there are no tickets it calculates a value for the preemptive flushing. However for the normal ticketed flushing we really only care about the space required for tickets. We will accidentally come in and flush one time, but as soon as we see there are no tickets we bail out of our flushing. Fix this by making btrfs_calc_reclaim_metadata_size really only tell us what is required for flushing if we have people waiting on space. Then move the preemptive flushing logic into need_preemptive_reclaim(). We ignore btrfs_calc_reclaim_metadata_size() in need_preemptive_reclaim() because if we are in this path then we made our reservation and there are not pending tickets currently, so we do not need to check it, simply do the fuzzy logic to check if we're getting low on space. Reviewed-by:
-
Josef Bacik authored
If we're flushing space for tickets then we have space_info->reclaim_size set and we do not need to do background reclaim. Reviewed-by:
-
Josef Bacik authored
All of our normal flushing is asynchronous reclaim, so this helper is poorly named. This is more checking if we need to preemptively flush space, so rename it to need_preemptive_reclaim. Also switch it to bool and make it plain static as followup patches will move more code here. Reviewed-by:
-
Josef Bacik authored
Currently if we ever have to flush space because we do not have enough we allocate a ticket and attach it to the space_info, and then systematically flush things in the filesystem that hold space reservations until our space is reclaimed. However this has a latency cost, we must go to sleep and wait for the flushing to make progress before we are woken up and allowed to continue doing our work. In order to address that we used to kick off the async worker to flush space preemptively, so that we could be reclaiming space hopefully before any tasks needed to stop and wait for space to reclaim. When I introduced the ticketed ENOSPC stuff this broke slightly in the fact that we were using tickets to indicate if we were done flushing. No tickets, no more flushing. However this meant that we essentially never preemptively flushed. This caused a write performance regression that Nikolay noticed in an unrelated patch that removed the committing of the transaction during btrfs_end_transaction. The behavior that happened pre that patch was btrfs_end_transaction() would see that we were low on space, and it would commit the transaction. This was bad because in this particular case you could end up with thousands and thousands of transactions being committed during the 5 minute reproducer. With the patch to remove this behavior we got much more sane transaction commits, but we ended up slower because we would write for a while, flush, write for a while, flush again. To address this we need to reinstate a preemptive flushing mechanism. However it is distinctly different from our ticketing flushing in that it doesn't have tickets to base it's decisions on. Instead of bolting this logic into our existing flushing work, add another worker to handle this preemptive flushing. Here we will attempt to be slightly intelligent about the things that we flushing, attempting to balance between whichever pool is taking up the most space. Reviewed-by:
-
Josef Bacik authored
Solely for preemptive flushing, we want to be able to force the transaction commit without any of the ambiguity of may_commit_transaction(). This is because may_commit_transaction() checks tickets and such, and in preemptive flushing we already know it'll be helpful, so use this to keep the code nice and clean and straightforward. Reviewed-by:
-
Josef Bacik authored
We track dio_bytes because the shrink delalloc code needs to know if we have more DIO in flight than we have normal buffered IO. The reason for this is because we can't "flush" DIO, we have to just wait on the ordered extents to finish. However this is true of all ordered extents. If we have more ordered space outstanding than dirty pages we should be waiting on ordered extents. We already are ok on this front technically, because we always do a FLUSH_DELALLOC_WAIT loop, but I want to use the ordered counter in the preemptive flushing code as well, so change this to count all ordered bytes instead of just DIO ordered bytes. Reviewed-by:
-
Josef Bacik authored
While debugging a ENOSPC related performance problem I needed to see the time difference between start and end of a reserve ticket, so add a trace point to report when we handle a reserve ticket. I opted to spit out start_ns itself without calculating the difference because there could be a gap between enabling the tracepoint and setting start_ns. Doing it this way allows us to filter on 0 start_ns so we don't get bogus entries, and we can easily calculate the time difference with bpftrace or something else. Reviewed-by:
-
Josef Bacik authored
I got a automated message from somebody who runs clang against our kernels and it's because I used the wrong enum type for what I passed into flush_space, caught by -Wenum-conversion. Change the argument to be explicitly the enum we're expecting to make everything consistent. Maybe eventually gcc will catch errors like this. Reviewed-by:
-
Roman Anasal authored
btrfs_compare_trees and changed_cb use a void *ctx parameter instead of struct send_ctx *sctx but when used in changed_cb it is immediately cast to `struct send_ctx *sctx = ctx;`. changed_cb is only ever called from btrfs_compare_trees and full_send_tree: - full_send_tree already passes a struct send_ctx *sctx - btrfs_compare_trees is only called by send_subvol with a struct send_ctx *sctx - void *ctx in btrfs_compare_trees is only used to be passed to changed_cb So casting to/from void *ctx seems unnecessary and directly using struct send_ctx *sctx instead provides better type-safety. The original reason for using void *ctx in the first place seems to have been dropped with 1b51d6fc ("btrfs: send: remove indirect callback parameter for changed_cb"). Signed-off-by:
Roman Anasal <roman.anasal@bdsu.de> Reviewed-by:
-
Josef Bacik authored
We love running delayed refs in commit_cowonly_roots, but it is a bit excessive. I was seeing cases of running 3 or 4 refs a few times in a row during this time. Instead simply: - update all of the roots first - then run delayed refs - then handle the empty block groups case - and then if we have any more dirty roots do the whole thing again This allows us to be much more efficient with our delayed ref running, as we can batch a few more operations at once. Reviewed-by:
-
Josef Bacik authored
This was added in commit 361048f5 ("Btrfs: fix full backref problem when inserting shared block reference") to address a problem where we hit the following BUG_ON() in alloc_reserved_tree_block if (node->type == BTRFS_SHARED_BLOCK_REF_KEY) { BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF)); However this BUG_ON() is bogus, and was removed by previous commit: btrfs: remove bogus BUG_ON in alloc_reserved_tree_block We no longer need to run delayed refs because of this, and can remove this flushing here. Signed-off-by:
-
Josef Bacik authored
The fix 361048f5 ("Btrfs: fix full backref problem when inserting shared block reference") added a delayed ref flushing at subvolume creation time in order to avoid hitting this particular BUG_ON(). Before this fix, we were tripping the BUG_ON() by 1. Modify snapshot A, which creates blocks with a normal reference for snapshot A, as A is the owner of these blocks. We now have delayed refs for these blocks. 2. Create a snapshot of A named B, which pushes references for the children blocks of the root node for the new root B, thus creating more delayed refs for newly allocated blocks. 3. A is modified, and because the metadata blocks can now be shared, it must push FULL_BACKREF references to the children of any block that A COWs down it's path to its target key. 4. Delayed refs are run. Because these are newly allocated blocks, we have ->must_insert_reserved reserved set on the delayed ref head, we call into alloc_reserved_tree_block() to add the extent item, and then add our ref. At the time of this fix, we were ordering FULL_BACKREF delayed ref operations first, so we'd go to add this reference and then BUG_ON() because we didn't have the FULL_BACKREF flag set. The patch fixed this problem by making sure we ran the delayed refs before we had the chance to modify A. This meant that any *new* blocks would have had their extent items created _before_ we would ever actually COW down and generate FULL_BACKREF entries. Thus the problem went away. However this BUG_ON() is actually completely bogus. The existence of a full backref doesn't necessarily mean that FULL_BACKREF must be set on that block, it must only be set on the actual parent itself. Consider the example provided above. If we COW down one path from A, any nodes are going to have a FULL_BACKREF ref pushed down to _all_ of their children, but not all of the children are going to have FULL_BACKREF set. It is completely valid to have an extent item with normal and full backrefs without FULL_BACKREF actually set on the block itself. As a final note, I have been testing with the patch (applied after this one) btrfs: stop running all delayed refs during snapshot which removed this flushing. My test was a torture test which did a lot of operations while snapshotting and deleting snapshots as well as relocation, and I never tripped this BUG_ON(). This is actually because at the time of 361048f5, we ordered SHARED keys _before_ normal references, and thus they would get run first. However currently they are ordered _after_ normal references, so we'd do the initial creation without having a shared reference, and thus not hit this BUG_ON(), which explains why I didn't start hitting this problem during my testing with my other patch applied. Signed-off-by:
-
Josef Bacik authored
The commit d6726335 ("btrfs: qgroup: Make snapshot accounting work with new extent-oriented qgroup.") added a flush of the delayed refs during snapshot creation in order to get the qgroup accounting properly. However this code has changed and been moved to it's own helper that is skipped if qgroups are turned off. Move the flushing to the helper, as we do not need it when qgroups are turned off. Also add a comment explaining why it exists, and why it doesn't actually save us. This will be helpful later when we try to fix qgroup accounting properly. Signed-off-by:
-
Josef Bacik authored
We try to pre-flush the delayed refs when committing, because we want to do as little work as possible in the critical section of the transaction commit. However doing this twice can lead to very long transaction commit delays as other threads are allowed to continue to generate more delayed refs, which potentially delays the commit by multiple minutes in very extreme cases. So simply stick to one pre-flush, and then continue the rest of the transaction commit. Reviewed-by:
-
Josef Bacik authored
Previously our delayed ref running used the total number of items as the items to run. However we changed that to number of heads to run with the delayed_refs_rsv, as generally we want to run all of the operations for one bytenr. But with btrfs_run_delayed_refs(trans, 0) we set our count to 2x the number of items that we have. This is generally fine, but if we have some operation generation loads of delayed refs while we're doing this pre-flushing in the transaction commit, we'll just spin forever doing delayed refs. Fix this to simply pick the number of delayed refs we currently have, that way we do not end up doing a lot of extra work that's being generated in other threads. Reviewed-by:
-
Josef Bacik authored
I've been running a stress test that runs 20 workers in their own subvolume, which are running an fsstress instance with 4 threads per worker, which is 80 total fsstress threads. In addition to this I'm running balance in the background as well as creating and deleting snapshots. This test takes around 12 hours to run normally, going slower and slower as the test goes on. The reason for this is because fsstress is running fsync sometimes, and because we're messing with block groups we often fall through to btrfs_commit_transaction, so will often have 20-30 threads all calling btrfs_commit_transaction at the same time. These all get stuck contending on the extent tree while they try to run delayed refs during the initial part of the commit. This is suboptimal, really because the extent tree is a single point of failure we only want one thread acting on that tree at once to reduce lock contention. Fix this by making the flushing mechanism a bit operation, to make it easy to use test_and_set_bit() in order to make sure only one task does this initial flush. Once we're into the transaction commit we only have one thread doing delayed ref running, it's just this initial pre-flush that is problematic. With this patch my stress test takes around 90 minutes to run, instead of 12 hours. The memory barrier is not necessary for the flushing bit as it's ordered, unlike plain int. The transaction state accessed in btrfs_should_end_transaction could be affected by that too as it's not always used under transaction lock. Upon Nikolay's analysis in [1] it's not necessary: In should_end_transaction it's read without holding any locks. (U) It's modified in btrfs_cleanup_transaction without holding the fs_info->trans_lock (U), but the STATE_ERROR flag is going to be set. set in cleanup_transaction under fs_info->trans_lock (L) set in btrfs_commit_trans to COMMIT_START under fs_info->trans_lock.(L) set in btrfs_commit_trans to COMMIT_DOING under fs_info->trans_lock.(L) set in btrfs_commit_trans to COMMIT_UNBLOCK under fs_info->trans_lock.(L) set in btrfs_commit_trans to COMMIT_COMPLETED without locks but at this point the transaction is finished and fs_info->running_trans is NULL (U but irrelevant). So by the looks of it we can have a concurrent READ race with a WRITE, due to reads not taking a lock. In this case what we want to ensure is we either see new or old state. I consulted with Will Deacon and he said that in such a case we'd want to annotate the accesses to ->state with (READ|WRITE)_ONCE so as to avoid a theoretical tear, in this case I don't think this could happen but I imagine at some point KCSAN would flag such an access as racy (which it is). [1] https://lore.kernel.org/linux-btrfs/e1fd5cc1-0f28-f670-69f4-e9958b4964e6@suse.comReviewed-by:
-
