Revert "block: remove artifical max_hw_sectors cap"

This reverts commit 34b48db66e08ca1c1bc07cf305d672ac940268dc.
That commit caused performance regressions for streaming I/O
workloads on a number of different storage devices, from
SATA disks to external RAID arrays.  It also managed to
trip up some buggy firmware in at least one drive, causing
data corruption.

The next patch will bump the default max_sectors_kb value to
1280, which will accommodate a 10-data-disk stripe write
with chunk size 128k.  In the testing I've done using iozone,
fio, and aio-stress, a value of 1280 does not show a big
performance difference from 512.  This will hopefully still
help the software RAID setup that Christoph saw the original
performance gains with while still not regressing other
storage configurations.

Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

diff --git a/block/blk-settings.c b/block/blk-settings.c
index 9df73991b23175d5390470b1501c0e3f862cfff6..d27b4e272356212530d6d5917fbb59ea2c914a25 100644 (file)
--- a/block/blk-settings.c
+++ b/block/blk-settings.c
@@ -236,7 +236,9 @@ void blk_limits_max_hw_sectors(struct queue_limits *limits, unsigned int max_hw_
                       __func__, max_hw_sectors);
        }
 
-       limits->max_sectors = limits->max_hw_sectors = max_hw_sectors;
+       limits->max_hw_sectors = max_hw_sectors;
+       limits->max_sectors = min_t(unsigned int, max_hw_sectors,
+                                   BLK_DEF_MAX_SECTORS);
 }
 EXPORT_SYMBOL(blk_limits_max_hw_sectors);
 

diff --git a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
index 46c282fff104cdbd013780460751c84b79d558bc..dd73e1ff1759c902db1734ac94975abea11e1b02 100644 (file)
--- a/drivers/block/aoe/aoeblk.c
+++ b/drivers/block/aoe/aoeblk.c
@@ -395,7 +395,7 @@ aoeblk_gdalloc(void *vp)
        WARN_ON(d->flags & DEVFL_TKILL);
        WARN_ON(d->gd);
        WARN_ON(d->flags & DEVFL_UP);
-       blk_queue_max_hw_sectors(q, 1024);
+       blk_queue_max_hw_sectors(q, BLK_DEF_MAX_SECTORS);
        q->backing_dev_info.name = "aoe";
        q->backing_dev_info.ra_pages = READ_AHEAD / PAGE_CACHE_SIZE;
        d->bufpool = mp;

diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index a1feff54aeab0b7fe9155fcc0b1a1fee3d9ac5e0..9f1e3f8eeed7d08a41f0ddb6d199366a4cc77926 100644 (file)
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1132,6 +1132,7 @@ extern int blk_verify_command(unsigned char *cmd, fmode_t has_write_perm);
 enum blk_default_limits {
        BLK_MAX_SEGMENTS        = 128,
        BLK_SAFE_MAX_SECTORS    = 255,
+       BLK_DEF_MAX_SECTORS     = 1024,
        BLK_MAX_SEGMENT_SIZE    = 65536,
        BLK_SEG_BOUNDARY_MASK   = 0xFFFFFFFFUL,
 };