- 15 Feb, 2023 40 commits
-
-
Christoph Hellwig authored
The only user in the zoned remap code is gone now, so remove the argument. Reviewed-by:
Josef Bacik <josef@toxicpanda.com> Signed-off-by:
Christoph Hellwig <hch@lst.de> Reviewed-by:
David Sterba <dsterba@suse.com> Signed-off-by:
-
Christoph Hellwig authored
btrfs_record_physical_zoned relies on a bio->bi_bdev samples in the bio_end_io handler to find the reverse map for remapping the zone append write, but stacked block device drivers can and usually do change bi_bdev when sending on the bio to a lower device. This can happen e.g. with the nvme-multipath driver when a NVMe SSD sets the shared namespace bit. But there is no real need for the bdev in btrfs_record_physical_zoned, as it is only passed to btrfs_rmap_block, which uses it to pick the mapping to report if there are multiple reverse mappings. As zone writes can only do simple non-mirror writes right now, and anything more complex will use the stripe tree there is no chance of the multiple mappings case actually happening. Instead open code the subset of btrfs_rmap_block in btrfs_record_physical_zoned, which also removes a memory allocation and remove the bdev field in the ordered extent. Fixes: d8e3fb10 ("btrfs: zoned: use ZONE_APPEND write for zoned mode") Reviewed-by:
-
Christoph Hellwig authored
Using Zone Append only makes sense for writes to the device, so check that in btrfs_use_zone_append. This avoids the possibility of artificially limited read size on zoned file systems. Reviewed-by:
Johannes Thumshirn <johannes.thumshirn@wdc.com> Signed-off-by:
-
Christoph Hellwig authored
struct btrfs_bio has all the information needed for btrfs_use_append, so pass that instead of a btrfs_inode and file_offset. Reviewed-by:
-
Christoph Hellwig authored
Instead of digging into the bio_vec in submit_one_bio, set file_offset at bio allocation time from the provided parameter. This also ensures that the file_offset is available all the time when building up the bio payload. Reviewed-by:
-
Christoph Hellwig authored
btrfs_ordered_extent->disk_bytenr can be rewritten by the zoned I/O completion handler, and thus in general is not a good idea to limit I/O size. But the maximum bio size calculation can easily be done using the file_offset fields in the btrfs_ordered_extent and btrfs_bio structures, so switch to that instead. Reviewed-by:
-
Filipe Manana authored
In the search loop of the binary search function, we are doing a division by 2 of the sum of the high and low slots. Because the slots are integers, the generated assembly code for it is the following on x86_64: 0x00000000000141f1 <+145>: mov %eax,%ebx 0x00000000000141f3 <+147>: shr $0x1f,%ebx 0x00000000000141f6 <+150>: add %eax,%ebx 0x00000000000141f8 <+152>: sar %ebx It's a few more instructions than a simple right shift, because signed integer division needs to round towards zero. However we know that slots can never be negative (btrfs_header_nritems() returns an u32), so we can instead use unsigned types for the low and high slots and therefore use unsigned integer division, which results in a single instruction on x86_64: 0x00000000000141f0 <+144>: shr %ebx So use unsigned types for the slots and therefore unsigned division. This is part of a small patchset comprised of the following two patches: btrfs: eliminate extra call when doing binary search on extent buffer btrfs: do unsigned integer division in the extent buffer binary search loop The following fs_mark test was run on a non-debug kernel (Debian's default kernel config) before and after applying the patchset: $ cat test.sh #!/bin/bash DEV=/dev/sdi MNT=/mnt/sdi MOUNT_OPTIONS="-o ssd" MKFS_OPTIONS="-O no-holes -R free-space-tree" FILES=100000 THREADS=$(nproc --all) FILE_SIZE=0 umount $DEV &> /dev/null mkfs.btrfs -f $MKFS_OPTIONS $DEV mount $MOUNT_OPTIONS $DEV $MNT OPTS="-S 0 -L 6 -n $FILES -s $FILE_SIZE -t $THREADS -k" for ((i = 1; i <= $THREADS; i++)); do OPTS="$OPTS -d $MNT/d$i" done fs_mark $OPTS umount $MNT Results before applying patchset: FSUse% Count Size Files/sec App Overhead 2 1200000 0 174472.0 11549868 4 2400000 0 253503.0 11694618 4 3600000 0 257833.1 11611508 6 4800000 0 247089.5 11665983 6 6000000 0 211296.1 12121244 10 7200000 0 187330.6 12548565 Results after applying patchset: FSUse% Count Size Files/sec App Overhead 2 1200000 0 207556.0 11393252 4 2400000 0 266751.1 11347909 4 3600000 0 274397.5 11270058 6 4800000 0 259608.4 11442250 6 6000000 0 238895.8 11635921 8 7200000 0 211942.2 11873825 Reviewed-by:Filipe Manana <fdmanana@suse.com> Reviewed-by:
-
Filipe Manana authored
The function btrfs_bin_search() is just a wrapper around the function generic_bin_search(), which passes the same arguments plus a default low slot with a value of 0. This adds an unnecessary extra function call, since btrfs_bin_search() is not static. So improve on this by making btrfs_bin_search() an inline function that calls generic_bin_search(), renaming the later to btrfs_generic_bin_search() and exporting it. Reviewed-by:
-
Christoph Hellwig authored
The only caller of scrub_rbio calls rbio_orig_end_io right after it, move it into scrub_rbio to match the other work item helpers. Reviewed-by:
Qu Wenruo <wqu@suse.com> Reviewed-by:
-
Christoph Hellwig authored
Both callers of recover_rbio call rbio_orig_end_io right after it, so move the call into the shared function. Reviewed-by:
-
Christoph Hellwig authored
Both callers of rmv_rbio call rbio_orig_end_io right after it, so move the call into the shared function. Reviewed-by:
-
Christoph Hellwig authored
Instead of filling in a bio_list and submitting the bios in the only caller, do that in scrub_assemble_read_bios. This removes the need to pass the bio_list, and also makes it clear that the extra bio_list cleanup in the caller is entirely pointless. Rename the function to scrub_read_bios to make it clear that the bios are not only assembled. Reviewed-by:
-
Christoph Hellwig authored
There is very little extra code in rmw_read_bios, and a large part of it is the superfluous extra cleanup of the bio list. Merge the two functions, and only clean up the bio list after it has been added to but before it has been emptied again by submit_read_wait_bio_list. Reviewed-by:
-
Christoph Hellwig authored
There is very little extra code in recover_rbio, and a large part of it is the superfluous extra cleanup of the bio list. Merge the two functions, and only clean up the bio list after it has been added to but before it has been emptied again by submit_read_wait_bio_list. Reviewed-by:
-
Christoph Hellwig authored
Add a helper to put all bios in a list. This does not need to be added to block layer as there are no other users of such code. Reviewed-by:
-
Christoph Hellwig authored
In addition to setting up the end_io handler and submitting the bios in submit_read_bios, also wait for them to be completed instead of waiting for the completion manually in all three callers. Rename submit_read_bios to submit_read_wait_bio_list to make it clear it waits for the bios as well. Reviewed-by:
-
Christoph Hellwig authored
Remove the write goto label by moving the data page allocation and data read into the branch. Reviewed-by:
-
Christoph Hellwig authored
Handle the error return on alloc_rbio failure directly instead of using a goto and remove the queue_rbio goto label by moving the plugged check into the if branch. Reviewed-by:
-
Josef Bacik authored
This is used in the tree-log code and is a holdover from previous iterations of extent buffer writeback. We can simply use wait_on_extent_buffer_writeback here, and remove btrfs_wait_tree_block_writeback completely as it's equivalent (waiting on page write writeback). Signed-off-by:
-
Josef Bacik authored
btrfs_clear_buffer_dirty just does the test_clear_bit() and then calls clear_extent_buffer_dirty and does the dirty metadata accounting. Combine this into clear_extent_buffer_dirty and make the result btrfs_clear_buffer_dirty. Signed-off-by:
-
Josef Bacik authored
btrfs_clean_tree_block is a misnomer, it's just clear_extent_buffer_dirty with some extra accounting around it. Rename this to btrfs_clear_buffer_dirty to make it more clear it belongs with it's setter, btrfs_mark_buffer_dirty. Signed-off-by:
-
Josef Bacik authored
We only add if we set the extent buffer dirty, and we subtract when we clear the extent buffer dirty. If we end up in set_btree_ioerr we have already cleared the buffer dirty, and we aren't resetting dirty on the extent buffer, so this is simply wrong. Signed-off-by:
-
Josef Bacik authored
Now that we're passing in the trans into btrfs_clean_tree_block, we can easily roll in the handling of the !trans case and replace all occurrences of if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) clear_extent_buffer_dirty(eb); with btrfs_tree_lock(eb); btrfs_clean_tree_block(eb); btrfs_tree_unlock(eb); We need the lock because if we are actually dirty we need to make sure we aren't racing with anything that's starting writeout currently. This also makes sure that we're accounting fs_info->dirty_metadata_bytes appropriately. Signed-off-by:
-
Josef Bacik authored
We check the header generation in the extent buffer against the current running transaction id to see if it's safe to clear DIRTY on this buffer. Generally speaking if we're clearing the buffer dirty we're holding the transaction open, but in the case of cleaning up an aborted transaction we don't, so we have extra checks in that path to check the transid. To allow for a future cleanup go ahead and pass in the trans handle so we don't have to rely on ->running_transaction being set. Signed-off-by:
-
Josef Bacik authored
We want to clean up the dirty handling for extent buffers so it's a little more consistent, so skip the check for generation == transid and simply always lock the extent buffer before calling btrfs_clean_tree_block. Signed-off-by:
-
Christoph Hellwig authored
No users left now that btrfs takes REQ_OP_WRITE bios from iomap and splits and converts them to REQ_OP_ZONE_APPEND internally. Reviewed-by:
Damien Le Moal <damien.lemoal@opensource.wdc.com> Reviewed-by:
-
Christoph Hellwig authored
The current btrfs zoned device support is a little cumbersome in the data I/O path as it requires the callers to not issue I/O larger than the supported ZONE_APPEND size of the underlying device. This leads to a lot of extra accounting. Instead change btrfs_submit_bio so that it can take write bios of arbitrary size and form from the upper layers, and just split them internally to the ZONE_APPEND queue limits. Then remove all the upper layer warts catering to limited write sized on zoned devices, including the extra refcount in the compressed_bio. Reviewed-by:
-
Christoph Hellwig authored
To be able to split a write into properly sized zone append commands, we need a queue_limits structure that contains the least common denominator suitable for all devices. Reviewed-by:
-
Christoph Hellwig authored
Call btrfs_submit_bio and btrfs_submit_compressed_read directly from submit_one_bio now that all additional functionality has moved into btrfs_submit_bio. Reviewed-by:
-
Christoph Hellwig authored
btrfs_submit_bio can derive it trivially from bbio->inode, so stop bothering in the callers. Reviewed-by:
-
Christoph Hellwig authored
Open code the functionality in the only caller and remove the now superfluous error handling there. Reviewed-by:
-
Christoph Hellwig authored
Now that btrfs_get_io_geometry has a single caller, we can massage it into a form that is more suitable for that caller and remove the marshalling into and out of struct btrfs_io_geometry. Reviewed-by:
-
Qu Wenruo authored
Stop looking at the stripe boundary in btrfs_encoded_read_regular_fill_pages() now that btrfs_submit_bio can split bios. Reviewed-by:
-
Qu Wenruo authored
Stop looking at the stripe boundary in alloc_compressed_bio() now that that btrfs_submit_bio can split bios, open code the now trivial code from alloc_compressed_bio() in btrfs_submit_compressed_read and stop maintaining the pending_ios count for reads as there is always just a single bio now. Reviewed-by:
-
Qu Wenruo authored
Remove btrfs_bio_ctrl::len_to_stripe_boundary, so that buffer I/O will no longer limit its bio size according to stripe length now that btrfs_submit_bio can split bios at stripe boundaries. Reviewed-by:
-
Christoph Hellwig authored
Now that btrfs_submit_bio splits the bio when crossing stripe boundaries, there is no need for the higher level code to do that manually. For direct I/O this is really helpful, as btrfs_submit_io can now simply take the bio allocated by iomap and send it on to btrfs_submit_bio instead of allocating clones. For that to work, the bio embedded into struct btrfs_dio_private needs to become a full btrfs_bio as expected by btrfs_submit_bio. With this change there is a single work item to offload the entire iomap bio so the heuristics to skip async processing for bios that were split isn't needed anymore either. Reviewed-by:
-
Christoph Hellwig authored
Currently the I/O submitters have to split bios according to the chunk stripe boundaries. This leads to extra lookups in the extent trees and a lot of boilerplate code. To drop this requirement, split the bio when __btrfs_map_block returns a mapping that is smaller than the requested size and keep a count of pending bios in the original btrfs_bio so that the upper level completion is only invoked when all clones have completed. Based on a patch from Qu Wenruo. Reviewed-by:
-
Christoph Hellwig authored
To allow splitting bios in btrfs_submit_bio, btree_csum_one_bio needs to be able to handle cloned bios. As btree_csum_one_bio is always called before handing the bio to the block layer that is trivially done by using bio_for_each_segment instead of bio_for_each_segment_all. Also switch the function to take a btrfs_bio and use that to derive the fs_info. Reviewed-by:
-
Christoph Hellwig authored
Move the code that splits the ordered extents and records the physical location for them to the storage layer so that the higher level consumers don't have to care about physical block numbers at all. This will also allow to eventually remove accounting for the zone append write sizes in the upper layer with a little bit more block layer work. Reviewed-by:
Naohiro Aota <naohiro.aota@wdc.com> Reviewed-by:
-
Christoph Hellwig authored
Instead of letting the callers of btrfs_submit_bio deal with checksumming the (meta)data in the bio and making decisions on when to offload the checksumming to the bio, leave that to btrfs_submit_bio. Do do so the existing btrfs_submit_bio function is split into an upper and a lower half, so that the lower half can be offloaded to a workqueue. Note that this changes the behavior for direct writes to raid56 volumes so that async checksum offloading is not skipped when more I/O is expected. This runs counter to the argument explaining why it was done, although I can't measure any affects of the change. Commits later in this series will make sure the entire direct writes is offloaded to the workqueue at once and thus make sure it is sent to the raid56 code from a single thread. Reviewed-by:
-
