Searched +hist:0 +hist:ed586d0 (Results 1 - 1 of 1) sorted by relevance
/linux-master/fs/btrfs/ | ||
H A D | tree-log.c | diff e383e158 Thu Jan 25 17:59:23 MST 2024 Filipe Manana <fdmanana@suse.com> btrfs: preallocate temporary extent buffer for inode logging when needed When logging an inode and we require to copy items from subvolume leaves to the log tree, we clone each subvolume leaf and than use that clone to copy items to the log tree. This is required to avoid possible deadlocks as stated in commit 796787c978ef ("btrfs: do not modify log tree while holding a leaf from fs tree locked"). The cloning requires allocating an extent buffer (struct extent_buffer) and then allocating pages (folios) to attach to the extent buffer. This may be slow in case we are under memory pressure, and since we are doing the cloning while holding a read lock on a subvolume leaf, it means we can be blocking other operations on that leaf for significant periods of time, which can increase latency on operations like creating other files, renaming files, etc. Similarly because we're under a log transaction, we may also cause extra delay on other tasks doing an fsync, because syncing the log requires waiting for tasks that joined a log transaction to exit the transaction. So to improve this, for any inode logging operation that needs to copy items from a subvolume leaf ("full sync" or "copy everything" bit set in the inode), preallocate a dummy extent buffer before locking any extent buffer from the subvolume tree, and even before joining a log transaction, add it to the log context and then use it when we need to copy items from a subvolume leaf to the log tree. This avoids making other operations get extra latency when waiting to lock a subvolume leaf that is used during inode logging and we are under heavy memory pressure. The following test script with bonnie++ was used to test this: $ cat test.sh #!/bin/bash DEV=/dev/sdh MNT=/mnt/sdh MOUNT_OPTIONS="-o ssd" MEMTOTAL_BYTES=`free -b | grep Mem: | awk '{ print $2 }'` NR_DIRECTORIES=20 NR_FILES=20480 DATASET_SIZE=$((MEMTOTAL_BYTES * 2 / 1048576)) DIRECTORY_SIZE=$((MEMTOTAL_BYTES * 2 / NR_FILES)) NR_FILES=$((NR_FILES / 1024)) 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 bonnie++ -u root -d $MNT \ -n $NR_FILES:$DIRECTORY_SIZE:$DIRECTORY_SIZE:$NR_DIRECTORIES \ -r 0 -s $DATASET_SIZE -b umount $MNT The results of this test on a 8G VM running a non-debug kernel (Debian's default kernel config), were the following. Before this change: Version 2.00a ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Name:Size etc /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP debian0 7501M 376k 99 1.4g 96 117m 14 1510k 99 2.5g 95 +++++ +++ Latency 35068us 24976us 2944ms 30725us 71770us 26152us Version 2.00a ------Sequential Create------ --------Random Create-------- debian0 -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files:max:min /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 20:384100:384100/20 20480 32 20480 58 20480 48 20480 39 20480 56 20480 61 Latency 411ms 11914us 119ms 617ms 10296us 110ms After this change: Version 2.00a ------Sequential Output------ --Sequential Input- --Random- -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Name:Size etc /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP debian0 7501M 375k 99 1.4g 97 117m 14 1546k 99 2.3g 98 +++++ +++ Latency 35975us 20945us 2144ms 10297us 2217us 6004us Version 2.00a ------Sequential Create------ --------Random Create-------- debian0 -Create-- --Read--- -Delete-- -Create-- --Read--- -Delete-- files:max:min /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP /sec %CP 20:384100:384100/20 20480 35 20480 58 20480 48 20480 40 20480 57 20480 59 Latency 320ms 11237us 77779us 518ms 6470us 86389us Reviewed-by: Josef Bacik <josef@toxicpanda.com> Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff 0f8ac74d Fri Sep 22 04:39:08 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use extent_io_tree_release() to empty dirty log pages When freeing a log tree, during a transaction commit, we clear its dirty log pages io tree by calling clear_extent_bits() using a range from 0 to (u64)-1. This will iterate the io tree's rbtree and call rb_erase() on each node before freeing it, which will often trigger rebalance operations on the rbtree. A better alternative it to use extent_io_tree_release(), which will not do deletions and trigger rebalances. So use extent_io_tree_release() instead of clear_extent_bits(). Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff 0f8ac74d Fri Sep 22 04:39:08 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use extent_io_tree_release() to empty dirty log pages When freeing a log tree, during a transaction commit, we clear its dirty log pages io tree by calling clear_extent_bits() using a range from 0 to (u64)-1. This will iterate the io tree's rbtree and call rb_erase() on each node before freeing it, which will often trigger rebalance operations on the rbtree. A better alternative it to use extent_io_tree_release(), which will not do deletions and trigger rebalances. So use extent_io_tree_release() instead of clear_extent_bits(). Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff 50564b65 Tue Sep 12 06:04:29 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: abort transaction on generation mismatch when marking eb as dirty When marking an extent buffer as dirty, at btrfs_mark_buffer_dirty(), we check if its generation matches the running transaction and if not we just print a warning. Such mismatch is an indicator that something really went wrong and only printing a warning message (and stack trace) is not enough to prevent a corruption. Allowing a transaction to commit with such an extent buffer will trigger an error if we ever try to read it from disk due to a generation mismatch with its parent generation. So abort the current transaction with -EUCLEAN if we notice a generation mismatch. For this we need to pass a transaction handle to btrfs_mark_buffer_dirty() which is always available except in test code, in which case we can pass NULL since it operates on dummy extent buffers and all test roots have a single node/leaf (root node at level 0). Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff b4c639f6 Mon Sep 04 22:15:24 MDT 2023 Josef Bacik <josef@toxicpanda.com> btrfs: initialize start_slot in btrfs_log_prealloc_extents Jens reported a compiler warning when using CONFIG_CC_OPTIMIZE_FOR_SIZE=y that looks like this fs/btrfs/tree-log.c: In function ‘btrfs_log_prealloc_extents’: fs/btrfs/tree-log.c:4828:23: warning: ‘start_slot’ may be used uninitialized [-Wmaybe-uninitialized] 4828 | ret = copy_items(trans, inode, dst_path, path, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4829 | start_slot, ins_nr, 1, 0); | ~~~~~~~~~~~~~~~~~~~~~~~~~ fs/btrfs/tree-log.c:4725:13: note: ‘start_slot’ was declared here 4725 | int start_slot; | ^~~~~~~~~~ The compiler is incorrect, as we only use this code when ins_len > 0, and when ins_len > 0 we have start_slot properly initialized. However we generally find the -Wmaybe-uninitialized warnings valuable, so initialize start_slot to get rid of the warning. Reported-by: Jens Axboe <axboe@kernel.dk> Tested-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Josef Bacik <josef@toxicpanda.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff b4c639f6 Mon Sep 04 22:15:24 MDT 2023 Josef Bacik <josef@toxicpanda.com> btrfs: initialize start_slot in btrfs_log_prealloc_extents Jens reported a compiler warning when using CONFIG_CC_OPTIMIZE_FOR_SIZE=y that looks like this fs/btrfs/tree-log.c: In function ‘btrfs_log_prealloc_extents’: fs/btrfs/tree-log.c:4828:23: warning: ‘start_slot’ may be used uninitialized [-Wmaybe-uninitialized] 4828 | ret = copy_items(trans, inode, dst_path, path, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4829 | start_slot, ins_nr, 1, 0); | ~~~~~~~~~~~~~~~~~~~~~~~~~ fs/btrfs/tree-log.c:4725:13: note: ‘start_slot’ was declared here 4725 | int start_slot; | ^~~~~~~~~~ The compiler is incorrect, as we only use this code when ins_len > 0, and when ins_len > 0 we have start_slot properly initialized. However we generally find the -Wmaybe-uninitialized warnings valuable, so initialize start_slot to get rid of the warning. Reported-by: Jens Axboe <axboe@kernel.dk> Tested-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Josef Bacik <josef@toxicpanda.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff b4c639f6 Mon Sep 04 22:15:24 MDT 2023 Josef Bacik <josef@toxicpanda.com> btrfs: initialize start_slot in btrfs_log_prealloc_extents Jens reported a compiler warning when using CONFIG_CC_OPTIMIZE_FOR_SIZE=y that looks like this fs/btrfs/tree-log.c: In function ‘btrfs_log_prealloc_extents’: fs/btrfs/tree-log.c:4828:23: warning: ‘start_slot’ may be used uninitialized [-Wmaybe-uninitialized] 4828 | ret = copy_items(trans, inode, dst_path, path, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4829 | start_slot, ins_nr, 1, 0); | ~~~~~~~~~~~~~~~~~~~~~~~~~ fs/btrfs/tree-log.c:4725:13: note: ‘start_slot’ was declared here 4725 | int start_slot; | ^~~~~~~~~~ The compiler is incorrect, as we only use this code when ins_len > 0, and when ins_len > 0 we have start_slot properly initialized. However we generally find the -Wmaybe-uninitialized warnings valuable, so initialize start_slot to get rid of the warning. Reported-by: Jens Axboe <axboe@kernel.dk> Tested-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Josef Bacik <josef@toxicpanda.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff d67ba263 Tue May 16 17:02:13 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use inode_logged() at btrfs_record_unlink_dir() At btrfs_record_unlink_dir() we directly check the logged_trans field of the given inodes to check if they were previously logged in the current transaction, and if any of them were, then we can avoid setting the field last_unlink_trans of the directory to the id of the current transaction if we are in a rename path. Avoiding that can later prevent falling back to a transaction commit if anyone attempts to log the directory. However the logged_trans field, store in struct btrfs_inode, is transient, not persisted in the inode item on its subvolume b+tree, so that means that if an inode is evicted and then loaded again, its original value is lost and it's reset to 0. So directly checking the logged_trans field can lead to some false negative, and that only results in a performance impact as mentioned before. Instead of directly checking the logged_trans field of the inodes, use the inode_logged() helper, which will check in the log tree if an inode was logged before in case its logged_trans field has a value of 0. This way we can avoid setting the directory inode's last_unlink_trans and cause future logging attempts of it to fallback to transaction commits. The following test script shows one example where this happens without this patch: $ cat test.sh #!/bin/bash DEV=/dev/nullb0 MNT=/mnt/nullb0 num_init_files=10000 num_new_files=10000 mkfs.btrfs -f $DEV mount -o ssd $DEV $MNT mkdir $MNT/testdir for ((i = 1; i <= $num_init_files; i++)); do echo -n > $MNT/testdir/file_$i done echo -n > $MNT/testdir/foo sync # Add some files so that there's more work in the transaction other # than just renaming file foo. for ((i = 1; i <= $num_new_files; i++)); do echo -n > $MNT/testdir/new_file_$i done # Change the file, fsync it. setfattr -n user.x1 -v 123 $MNT/testdir/foo xfs_io -c "fsync" $MNT/testdir/foo # Now triggger eviction of file foo but no eviction for our test # directory, since it is being used by the process below. This will # set logged_trans of the file's inode to 0 once it is loaded again. ( cd $MNT/testdir while true; do : done ) & pid=$! echo 2 > /proc/sys/vm/drop_caches kill $pid wait $pid # Move foo out of our testdir. This will set last_unlink_trans # of the directory inode to the current transaction, because # logged_trans of both the directory and the file are set to 0. mv $MNT/testdir/foo $MNT/foo # Change file foo again and fsync it. # This fsync will result in a transaction commit because the rename # above has set last_unlink_trans of the parent directory to the id # of the current transaction and because our inode for file foo has # last_unlink_trans set to the current transaction, since it was # evicted and reloaded and it was previously modified in the current # transaction (the xattr addition). xfs_io -c "pwrite 0 64K" $MNT/foo start=$(date +%s%N) xfs_io -c "fsync" $MNT/foo end=$(date +%s%N) dur=$(( (end - start) / 1000000 )) echo "file fsync took: $dur milliseconds" umount $MNT Before this patch: fsync took 19 milliseconds After this patch: fsync took 5 milliseconds Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff d67ba263 Tue May 16 17:02:13 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use inode_logged() at btrfs_record_unlink_dir() At btrfs_record_unlink_dir() we directly check the logged_trans field of the given inodes to check if they were previously logged in the current transaction, and if any of them were, then we can avoid setting the field last_unlink_trans of the directory to the id of the current transaction if we are in a rename path. Avoiding that can later prevent falling back to a transaction commit if anyone attempts to log the directory. However the logged_trans field, store in struct btrfs_inode, is transient, not persisted in the inode item on its subvolume b+tree, so that means that if an inode is evicted and then loaded again, its original value is lost and it's reset to 0. So directly checking the logged_trans field can lead to some false negative, and that only results in a performance impact as mentioned before. Instead of directly checking the logged_trans field of the inodes, use the inode_logged() helper, which will check in the log tree if an inode was logged before in case its logged_trans field has a value of 0. This way we can avoid setting the directory inode's last_unlink_trans and cause future logging attempts of it to fallback to transaction commits. The following test script shows one example where this happens without this patch: $ cat test.sh #!/bin/bash DEV=/dev/nullb0 MNT=/mnt/nullb0 num_init_files=10000 num_new_files=10000 mkfs.btrfs -f $DEV mount -o ssd $DEV $MNT mkdir $MNT/testdir for ((i = 1; i <= $num_init_files; i++)); do echo -n > $MNT/testdir/file_$i done echo -n > $MNT/testdir/foo sync # Add some files so that there's more work in the transaction other # than just renaming file foo. for ((i = 1; i <= $num_new_files; i++)); do echo -n > $MNT/testdir/new_file_$i done # Change the file, fsync it. setfattr -n user.x1 -v 123 $MNT/testdir/foo xfs_io -c "fsync" $MNT/testdir/foo # Now triggger eviction of file foo but no eviction for our test # directory, since it is being used by the process below. This will # set logged_trans of the file's inode to 0 once it is loaded again. ( cd $MNT/testdir while true; do : done ) & pid=$! echo 2 > /proc/sys/vm/drop_caches kill $pid wait $pid # Move foo out of our testdir. This will set last_unlink_trans # of the directory inode to the current transaction, because # logged_trans of both the directory and the file are set to 0. mv $MNT/testdir/foo $MNT/foo # Change file foo again and fsync it. # This fsync will result in a transaction commit because the rename # above has set last_unlink_trans of the parent directory to the id # of the current transaction and because our inode for file foo has # last_unlink_trans set to the current transaction, since it was # evicted and reloaded and it was previously modified in the current # transaction (the xattr addition). xfs_io -c "pwrite 0 64K" $MNT/foo start=$(date +%s%N) xfs_io -c "fsync" $MNT/foo end=$(date +%s%N) dur=$(( (end - start) / 1000000 )) echo "file fsync took: $dur milliseconds" umount $MNT Before this patch: fsync took 19 milliseconds After this patch: fsync took 5 milliseconds Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff d67ba263 Tue May 16 17:02:13 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use inode_logged() at btrfs_record_unlink_dir() At btrfs_record_unlink_dir() we directly check the logged_trans field of the given inodes to check if they were previously logged in the current transaction, and if any of them were, then we can avoid setting the field last_unlink_trans of the directory to the id of the current transaction if we are in a rename path. Avoiding that can later prevent falling back to a transaction commit if anyone attempts to log the directory. However the logged_trans field, store in struct btrfs_inode, is transient, not persisted in the inode item on its subvolume b+tree, so that means that if an inode is evicted and then loaded again, its original value is lost and it's reset to 0. So directly checking the logged_trans field can lead to some false negative, and that only results in a performance impact as mentioned before. Instead of directly checking the logged_trans field of the inodes, use the inode_logged() helper, which will check in the log tree if an inode was logged before in case its logged_trans field has a value of 0. This way we can avoid setting the directory inode's last_unlink_trans and cause future logging attempts of it to fallback to transaction commits. The following test script shows one example where this happens without this patch: $ cat test.sh #!/bin/bash DEV=/dev/nullb0 MNT=/mnt/nullb0 num_init_files=10000 num_new_files=10000 mkfs.btrfs -f $DEV mount -o ssd $DEV $MNT mkdir $MNT/testdir for ((i = 1; i <= $num_init_files; i++)); do echo -n > $MNT/testdir/file_$i done echo -n > $MNT/testdir/foo sync # Add some files so that there's more work in the transaction other # than just renaming file foo. for ((i = 1; i <= $num_new_files; i++)); do echo -n > $MNT/testdir/new_file_$i done # Change the file, fsync it. setfattr -n user.x1 -v 123 $MNT/testdir/foo xfs_io -c "fsync" $MNT/testdir/foo # Now triggger eviction of file foo but no eviction for our test # directory, since it is being used by the process below. This will # set logged_trans of the file's inode to 0 once it is loaded again. ( cd $MNT/testdir while true; do : done ) & pid=$! echo 2 > /proc/sys/vm/drop_caches kill $pid wait $pid # Move foo out of our testdir. This will set last_unlink_trans # of the directory inode to the current transaction, because # logged_trans of both the directory and the file are set to 0. mv $MNT/testdir/foo $MNT/foo # Change file foo again and fsync it. # This fsync will result in a transaction commit because the rename # above has set last_unlink_trans of the parent directory to the id # of the current transaction and because our inode for file foo has # last_unlink_trans set to the current transaction, since it was # evicted and reloaded and it was previously modified in the current # transaction (the xattr addition). xfs_io -c "pwrite 0 64K" $MNT/foo start=$(date +%s%N) xfs_io -c "fsync" $MNT/foo end=$(date +%s%N) dur=$(( (end - start) / 1000000 )) echo "file fsync took: $dur milliseconds" umount $MNT Before this patch: fsync took 19 milliseconds After this patch: fsync took 5 milliseconds Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff d67ba263 Tue May 16 17:02:13 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use inode_logged() at btrfs_record_unlink_dir() At btrfs_record_unlink_dir() we directly check the logged_trans field of the given inodes to check if they were previously logged in the current transaction, and if any of them were, then we can avoid setting the field last_unlink_trans of the directory to the id of the current transaction if we are in a rename path. Avoiding that can later prevent falling back to a transaction commit if anyone attempts to log the directory. However the logged_trans field, store in struct btrfs_inode, is transient, not persisted in the inode item on its subvolume b+tree, so that means that if an inode is evicted and then loaded again, its original value is lost and it's reset to 0. So directly checking the logged_trans field can lead to some false negative, and that only results in a performance impact as mentioned before. Instead of directly checking the logged_trans field of the inodes, use the inode_logged() helper, which will check in the log tree if an inode was logged before in case its logged_trans field has a value of 0. This way we can avoid setting the directory inode's last_unlink_trans and cause future logging attempts of it to fallback to transaction commits. The following test script shows one example where this happens without this patch: $ cat test.sh #!/bin/bash DEV=/dev/nullb0 MNT=/mnt/nullb0 num_init_files=10000 num_new_files=10000 mkfs.btrfs -f $DEV mount -o ssd $DEV $MNT mkdir $MNT/testdir for ((i = 1; i <= $num_init_files; i++)); do echo -n > $MNT/testdir/file_$i done echo -n > $MNT/testdir/foo sync # Add some files so that there's more work in the transaction other # than just renaming file foo. for ((i = 1; i <= $num_new_files; i++)); do echo -n > $MNT/testdir/new_file_$i done # Change the file, fsync it. setfattr -n user.x1 -v 123 $MNT/testdir/foo xfs_io -c "fsync" $MNT/testdir/foo # Now triggger eviction of file foo but no eviction for our test # directory, since it is being used by the process below. This will # set logged_trans of the file's inode to 0 once it is loaded again. ( cd $MNT/testdir while true; do : done ) & pid=$! echo 2 > /proc/sys/vm/drop_caches kill $pid wait $pid # Move foo out of our testdir. This will set last_unlink_trans # of the directory inode to the current transaction, because # logged_trans of both the directory and the file are set to 0. mv $MNT/testdir/foo $MNT/foo # Change file foo again and fsync it. # This fsync will result in a transaction commit because the rename # above has set last_unlink_trans of the parent directory to the id # of the current transaction and because our inode for file foo has # last_unlink_trans set to the current transaction, since it was # evicted and reloaded and it was previously modified in the current # transaction (the xattr addition). xfs_io -c "pwrite 0 64K" $MNT/foo start=$(date +%s%N) xfs_io -c "fsync" $MNT/foo end=$(date +%s%N) dur=$(( (end - start) / 1000000 )) echo "file fsync took: $dur milliseconds" umount $MNT Before this patch: fsync took 19 milliseconds After this patch: fsync took 5 milliseconds Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> diff d67ba263 Tue May 16 17:02:13 MDT 2023 Filipe Manana <fdmanana@suse.com> btrfs: use inode_logged() at btrfs_record_unlink_dir() At btrfs_record_unlink_dir() we directly check the logged_trans field of the given inodes to check if they were previously logged in the current transaction, and if any of them were, then we can avoid setting the field last_unlink_trans of the directory to the id of the current transaction if we are in a rename path. Avoiding that can later prevent falling back to a transaction commit if anyone attempts to log the directory. However the logged_trans field, store in struct btrfs_inode, is transient, not persisted in the inode item on its subvolume b+tree, so that means that if an inode is evicted and then loaded again, its original value is lost and it's reset to 0. So directly checking the logged_trans field can lead to some false negative, and that only results in a performance impact as mentioned before. Instead of directly checking the logged_trans field of the inodes, use the inode_logged() helper, which will check in the log tree if an inode was logged before in case its logged_trans field has a value of 0. This way we can avoid setting the directory inode's last_unlink_trans and cause future logging attempts of it to fallback to transaction commits. The following test script shows one example where this happens without this patch: $ cat test.sh #!/bin/bash DEV=/dev/nullb0 MNT=/mnt/nullb0 num_init_files=10000 num_new_files=10000 mkfs.btrfs -f $DEV mount -o ssd $DEV $MNT mkdir $MNT/testdir for ((i = 1; i <= $num_init_files; i++)); do echo -n > $MNT/testdir/file_$i done echo -n > $MNT/testdir/foo sync # Add some files so that there's more work in the transaction other # than just renaming file foo. for ((i = 1; i <= $num_new_files; i++)); do echo -n > $MNT/testdir/new_file_$i done # Change the file, fsync it. setfattr -n user.x1 -v 123 $MNT/testdir/foo xfs_io -c "fsync" $MNT/testdir/foo # Now triggger eviction of file foo but no eviction for our test # directory, since it is being used by the process below. This will # set logged_trans of the file's inode to 0 once it is loaded again. ( cd $MNT/testdir while true; do : done ) & pid=$! echo 2 > /proc/sys/vm/drop_caches kill $pid wait $pid # Move foo out of our testdir. This will set last_unlink_trans # of the directory inode to the current transaction, because # logged_trans of both the directory and the file are set to 0. mv $MNT/testdir/foo $MNT/foo # Change file foo again and fsync it. # This fsync will result in a transaction commit because the rename # above has set last_unlink_trans of the parent directory to the id # of the current transaction and because our inode for file foo has # last_unlink_trans set to the current transaction, since it was # evicted and reloaded and it was previously modified in the current # transaction (the xattr addition). xfs_io -c "pwrite 0 64K" $MNT/foo start=$(date +%s%N) xfs_io -c "fsync" $MNT/foo end=$(date +%s%N) dur=$(( (end - start) / 1000000 )) echo "file fsync took: $dur milliseconds" umount $MNT Before this patch: fsync took 19 milliseconds After this patch: fsync took 5 milliseconds Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> |
Completed in 457 milliseconds