zram: set BDI_CAP_STABLE_WRITES once
authorMinchan Kim <minchan@kernel.org>
Thu, 16 Nov 2017 01:32:56 +0000 (17:32 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 16 Nov 2017 02:21:02 +0000 (18:21 -0800)
With fast swap storage, the platform wants to use swap more aggressively
and swap-in is crucial to application latency.

The rw_page() based synchronous devices like zram, pmem and btt are such
fast storage.  When I profile swapin performance with zram lz4
decompress test, S/W overhead is more than 70%.  Maybe, it would be
bigger in nvdimm.

This patchset reduces swap-in latency by skipping swapcache if the swap
device is a synchronous device like a rw_page() based device.

It enhances by 45% my swapin test (5G sequential swapin, no readahead)
from 2.41sec to 1.64sec.

This patch (of 4):

Commit 19b7ccf8651d ("block: get rid of blk_integrity_revalidate()")
fixed a weird thing (i.e., reset BDI_CAP_STABLE_WRITES flag
unconditionally whenever revalidat_disk is called) so zram doesn't need
to reset the flag any more when revalidating the bdev.  Instead, set the
flag just once when the zram device is created.

It shouldn't change any behavior.

Link: http://lkml.kernel.org/r/1505886205-9671-2-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Hugh Dickins <hughd@google.com>
Cc: Huang Ying <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

index f149d3e612341d1f5e70797a8f62d0330fa3973d..d95bb8ce50928122cc2c8c00444fdaace10beab0 100644 (file)
@@ -122,14 +122,6 @@ static inline bool is_partial_io(struct bio_vec *bvec)
-static void zram_revalidate_disk(struct zram *zram)
-       revalidate_disk(zram->disk);
-       /* revalidate_disk reset the BDI_CAP_STABLE_WRITES so set again */
-       zram->disk->queue->backing_dev_info->capabilities |=
-               BDI_CAP_STABLE_WRITES;
  * Check if request is within bounds and aligned on zram logical blocks.
@@ -1373,7 +1365,8 @@ static ssize_t disksize_store(struct device *dev,
        zram->comp = comp;
        zram->disksize = disksize;
        set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT);
-       zram_revalidate_disk(zram);
+       revalidate_disk(zram->disk);
        return len;
@@ -1420,7 +1413,7 @@ static ssize_t reset_store(struct device *dev,
        /* Make sure all the pending I/O are finished */
-       zram_revalidate_disk(zram);
+       revalidate_disk(zram->disk);
@@ -1539,6 +1532,7 @@ static int zram_add(void)
        /* zram devices sort of resembles non-rotational disks */
        queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zram->disk->queue);
        queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zram->disk->queue);
         * To ensure that we always get PAGE_SIZE aligned
         * and n*PAGE_SIZED sized I/O requests.
@@ -1563,6 +1557,8 @@ static int zram_add(void)
                blk_queue_max_write_zeroes_sectors(zram->disk->queue, UINT_MAX);
+       zram->disk->queue->backing_dev_info->capabilities |=
+                                       BDI_CAP_STABLE_WRITES;
        ret = sysfs_create_group(&disk_to_dev(zram->disk)->kobj,