| Message ID | 1c7f75c138cba65d0af23ad4eda7c1bab1cc21fe.1649666810.git.naohiro.aota@wdc.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | btrfs: do not sleep unnecessary on successful batch page allocation | expand |
On 11.04.22 г. 12:02 ч., Naohiro Aota wrote: > After commit 727fd577af04 ("btrfs: wait between incomplete batch memory > allocations"), fstests btrfs/187 becomes incredibly slow. Before the commit > it takes 335 seconds to run the test on 8GB ZRAM device running on QEMU. > But, after that patch, it never finish after 4 hours. > > This is because of unnecessary memalloc_retry_wait() call. If > alloc_pages_bulk_array() successfully allocated all the requested pages, it > is no use to call memalloc_retry_wait() to wait for retrying. > > I confirmed the test run time back to 353 seconds with this patch applied. > > Link: https://lore.kernel.org/linux-btrfs/20220411071124.zwtcarqngqqkdd6q@naota-xeon/ > Cc: Sweet Tea Dorminy <sweettea-kernel@dorminy.me> > Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> Reviewed-by: Nikolay Borisov <nborisov@suse.com> > --- > fs/btrfs/extent_io.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c > index 1a5a7ded3175..2681a998ee1c 100644 > --- a/fs/btrfs/extent_io.c > +++ b/fs/btrfs/extent_io.c > @@ -3150,6 +3150,9 @@ int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array) > > allocated = alloc_pages_bulk_array(GFP_NOFS, nr_pages, page_array); > > + if (allocated == nr_pages) > + return 0; > + > /* > * During this iteration, no page could be allocated, even > * though alloc_pages_bulk_array() falls back to alloc_page()
On Mon, Apr 11, 2022 at 06:02:52PM +0900, Naohiro Aota wrote: > After commit 727fd577af04 ("btrfs: wait between incomplete batch memory > allocations"), fstests btrfs/187 becomes incredibly slow. Before the commit > it takes 335 seconds to run the test on 8GB ZRAM device running on QEMU. > But, after that patch, it never finish after 4 hours. > > This is because of unnecessary memalloc_retry_wait() call. If > alloc_pages_bulk_array() successfully allocated all the requested pages, it > is no use to call memalloc_retry_wait() to wait for retrying. > > I confirmed the test run time back to 353 seconds with this patch applied. Thanks, I've folded the fixup to the patch.
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c index 1a5a7ded3175..2681a998ee1c 100644 --- a/fs/btrfs/extent_io.c +++ b/fs/btrfs/extent_io.c @@ -3150,6 +3150,9 @@ int btrfs_alloc_page_array(unsigned int nr_pages, struct page **page_array) allocated = alloc_pages_bulk_array(GFP_NOFS, nr_pages, page_array); + if (allocated == nr_pages) + return 0; + /* * During this iteration, no page could be allocated, even * though alloc_pages_bulk_array() falls back to alloc_page()
After commit 727fd577af04 ("btrfs: wait between incomplete batch memory allocations"), fstests btrfs/187 becomes incredibly slow. Before the commit it takes 335 seconds to run the test on 8GB ZRAM device running on QEMU. But, after that patch, it never finish after 4 hours. This is because of unnecessary memalloc_retry_wait() call. If alloc_pages_bulk_array() successfully allocated all the requested pages, it is no use to call memalloc_retry_wait() to wait for retrying. I confirmed the test run time back to 353 seconds with this patch applied. Link: https://lore.kernel.org/linux-btrfs/20220411071124.zwtcarqngqqkdd6q@naota-xeon/ Cc: Sweet Tea Dorminy <sweettea-kernel@dorminy.me> Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com> --- fs/btrfs/extent_io.c | 3 +++ 1 file changed, 3 insertions(+)