From Artur:
The wait_for_overlap wait queue is currently used in two cases, which
are not really related:
- waiting for actual overlapping bios, which uses R5_Overlap bit,
- waiting for events related to reshape.
Handling every write request in raid5_make_request() involves adding to
and removing from this wait queue, which uses a spinlock. With fast
storage and multiple submitting threads the contention on this lock is
noticeable.
This patch series aims to resolve this by separating the two cases
mentioned above and using this wait queue only when reshape is in
progress.
The results when testing 4k random writes on raid5 with null_blk
(8 jobs, qd=64, group_thread_cnt=8):
before: 463k IOPS
after: 523k IOPS
The improvement is not huge with this series alone but it is just one of
the bottlenecks. When applied onto some other changes I'm working on, it
allowed to go from 845k IOPS to 975k IOPS on the same test.
* md-6.12-raid5-opt:
md/raid5: rename wait_for_overlap to wait_for_reshape
md/raid5: only add to wq if reshape is in progress
md/raid5: use wait_on_bit() for R5_Overlap
Signed-off-by: Song Liu <song@kernel.org>
Now that actual overlaps are not handled on the wait_for_overlap wq
anymore, the remaining cases when we wait on this wq are limited to
reshape. If reshape is not in progress, don't add to the wq in
raid5_make_request() because add_wait_queue() / remove_wait_queue()
operations take a spinlock and cause noticeable contention when multiple
threads are submitting requests to the mddev.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Link: https://lore.kernel.org/r/20240827153536.6743-3-artur.paszkiewicz@intel.com
Signed-off-by: Song Liu <song@kernel.org>
From Yu Kuai (with minor changes by Song Liu):
The background is that currently bitmap is using a global spin_lock,
causing lock contention and huge IO performance degradation for all raid
levels.
However, it's impossible to implement a new lock free bitmap with
current situation that md-bitmap exposes the internal implementation
with lots of exported apis. Hence bitmap_operations is invented, to
describe bitmap core implementation, and a new bitmap can be introduced
with a new bitmap_operations, we only need to switch to the new one
during initialization.
And with this we can build bitmap as kernel module, but that's not
our concern for now.
This version was tested with mdadm tests and lvm2 tests. This set does
not introduce new errors in these tests.
* md-6.12-bitmap: (42 commits)
md/md-bitmap: make in memory structure internal
md/md-bitmap: merge md_bitmap_enabled() into bitmap_operations
md/md-bitmap: merge md_bitmap_wait_behind_writes() into bitmap_operations
md/md-bitmap: merge md_bitmap_free() into bitmap_operations
md/md-bitmap: merge md_bitmap_set_pages() into struct bitmap_operations
md/md-bitmap: merge md_bitmap_copy_from_slot() into struct bitmap_operation.
md/md-bitmap: merge get_bitmap_from_slot() into bitmap_operations
md/md-bitmap: merge md_bitmap_resize() into bitmap_operations
md/md-bitmap: pass in mddev directly for md_bitmap_resize()
md/md-bitmap: merge md_bitmap_daemon_work() into bitmap_operations
md/md-bitmap: merge bitmap_unplug() into bitmap_operations
md/md-bitmap: merge md_bitmap_unplug_async() into md_bitmap_unplug()
md/md-bitmap: merge md_bitmap_sync_with_cluster() into bitmap_operations
md/md-bitmap: merge md_bitmap_cond_end_sync() into bitmap_operations
md/md-bitmap: merge md_bitmap_close_sync() into bitmap_operations
md/md-bitmap: merge md_bitmap_end_sync() into bitmap_operations
md/md-bitmap: remove the parameter 'aborted' for md_bitmap_end_sync()
md/md-bitmap: merge md_bitmap_start_sync() into bitmap_operations
md/md-bitmap: merge md_bitmap_endwrite() into bitmap_operations
md/md-bitmap: merge md_bitmap_startwrite() into bitmap_operations
...
Signed-off-by: Song Liu <song@kernel.org>
For flush request, md has a special flush handling to merge concurrent
flush request into single one, however, the whole mechanism is based on
a disk level spin_lock 'mddev->lock'. And fsync can be called quite
often in some user cases, for consequence, spin lock from IO fast path can
cause performance degradation.
Fortunately, the block layer already has flush handling to merge
concurrent flush request, and it only acquires hctx level spin lock. (see
details in blk-flush.c)
This patch removes the flush handling in md, and converts to use general
block layer flush handling in underlying disks.
Flush test for 4 nvme raid10:
start 128 threads to do fsync 100000 times, on arm64, see how long it
takes.
Test script:
void* thread_func(void* arg) {
int fd = *(int*)arg;
for (int i = 0; i < FSYNC_COUNT; i++) {
fsync(fd);
}
return NULL;
}
int main() {
int fd = open("/dev/md0", O_RDWR);
if (fd < 0) {
perror("open");
exit(1);
}
pthread_t threads[THREADS];
struct timeval start, end;
gettimeofday(&start, NULL);
for (int i = 0; i < THREADS; i++) {
pthread_create(&threads[i], NULL, thread_func, &fd);
}
for (int i = 0; i < THREADS; i++) {
pthread_join(threads[i], NULL);
}
gettimeofday(&end, NULL);
close(fd);
long long elapsed = (end.tv_sec - start.tv_sec) * 1000000LL + (end.tv_usec - start.tv_usec);
printf("Elapsed time: %lld microseconds\n", elapsed);
return 0;
}
Test result: about 10 times faster:
Before this patch: 50943374 microseconds
After this patch: 5096347 microseconds
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240827110616.3860190-1-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
So that the implementation won't be exposed, and it'll be possible
to invent a new bitmap by replacing bitmap_operations.
Also change the parameter from bitmap to mddev, to avoid access
bitmap outside md-bitmap.c as much as possible. And change the type
of 'success' and 'behind' from int to bool.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240826074452.1490072-25-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
So that the implementation won't be exposed, and it'll be possible
to invent a new bitmap by replacing bitmap_operations.
Also change the parameter from bitmap to mddev, to avoid access
bitmap outside md-bitmap.c as much as possible. And change the type
of 'behind' from int to bool.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240826074452.1490072-24-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
So that the implementation won't be exposed, and it'll be possible
to invent a new bitmap by replacing bitmap_operations.
Also change the parameter from bitmap to mddev, to avoid access
bitmap outside md-bitmap.c as much as possible.
And while we're here, also fix coding style for bitmap_store().
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240826074452.1490072-23-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
To avoid dereferencing bitmap directly in md-cluster to prepare
inventing a new bitmap.
BTW, also fix following checkpatch warnings:
WARNING: Deprecated use of 'kmap_atomic', prefer 'kmap_local_page' instead
WARNING: Deprecated use of 'kunmap_atomic', prefer 'kunmap_local' instead
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240826074452.1490072-7-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
drivers/md/md-cluster.c:1220:22: warning: incorrect type in assignment (different base types)
drivers/md/md-cluster.c:1220:22: expected unsigned long my_sync_size
drivers/md/md-cluster.c:1220:22: got restricted __le64 [usertype] sync_size
drivers/md/md-cluster.c:1252:35: warning: incorrect type in assignment (different base types)
drivers/md/md-cluster.c:1252:35: expected unsigned long sync_size
drivers/md/md-cluster.c:1252:35: got restricted __le64 [usertype] sync_size
drivers/md/md-cluster.c:1253:41: warning: restricted __le64 degrades to integer
Fix the warnings by using le64_to_cpu() to convet __le64 to integer.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240826074452.1490072-6-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
Use the existed helper instead of open coding it to make the code cleaner.
There are no functional changes, and also avoid dereferencing bitmap
directly to prepare inventing a new bitmap.
Noted that this patch also export md_bitmap_wait_behind_writes(), which
is necessary for now, and the exported api will be removed in following
patches to convert bitmap apis into ops.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240826074452.1490072-2-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
The local variable will only be used onced, in the error path that
read_balance() failed to find a valid rdev to read. Since now the rdev
is ensured can't be removed from conf while IO is still pending,
remove the local variable and dereference rdev directly.
Since we're here, also remove an extra empty line, and unnecessary
type conversion from sector_t(u64) to unsigned long long.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240801133008.459998-1-yukuai1@huaweicloud.com
Signed-off-by: Song Liu <song@kernel.org>
