]> git.proxmox.com Git - mirror_ubuntu-focal-kernel.git/commitdiff
blk-settings: align max_sectors on "logical_block_size" boundary
authorMikulas Patocka <mpatocka@redhat.com>
Wed, 24 Feb 2021 02:25:30 +0000 (19:25 -0700)
committerKleber Sacilotto de Souza <kleber.souza@canonical.com>
Wed, 14 Apr 2021 16:31:36 +0000 (18:31 +0200)
BugLink: https://bugs.launchpad.net/bugs/1918974
commit 97f433c3601a24d3513d06f575a389a2ca4e11e4 upstream.

We get I/O errors when we run md-raid1 on the top of dm-integrity on the
top of ramdisk.
device-mapper: integrity: Bio not aligned on 8 sectors: 0xff00, 0xff
device-mapper: integrity: Bio not aligned on 8 sectors: 0xff00, 0xff
device-mapper: integrity: Bio not aligned on 8 sectors: 0xffff, 0x1
device-mapper: integrity: Bio not aligned on 8 sectors: 0xffff, 0x1
device-mapper: integrity: Bio not aligned on 8 sectors: 0x8048, 0xff
device-mapper: integrity: Bio not aligned on 8 sectors: 0x8147, 0xff
device-mapper: integrity: Bio not aligned on 8 sectors: 0x8246, 0xff
device-mapper: integrity: Bio not aligned on 8 sectors: 0x8345, 0xbb

The ramdisk device has logical_block_size 512 and max_sectors 255. The
dm-integrity device uses logical_block_size 4096 and it doesn't affect the
"max_sectors" value - thus, it inherits 255 from the ramdisk. So, we have
a device with max_sectors not aligned on logical_block_size.

The md-raid device sees that the underlying leg has max_sectors 255 and it
will split the bios on 255-sector boundary, making the bios unaligned on
logical_block_size.

In order to fix the bug, we round down max_sectors to logical_block_size.

Cc: stable@vger.kernel.org
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Kelsey Skunberg <kelsey.skunberg@canonical.com>
block/blk-settings.c

index be1dca0103a455ab9ad67e4648f38ff653a7505c..13be635300a853eb22bb7ae338017a82df4aed64 100644 (file)
@@ -473,6 +473,14 @@ void blk_queue_stack_limits(struct request_queue *t, struct request_queue *b)
 }
 EXPORT_SYMBOL(blk_queue_stack_limits);
 
+static unsigned int blk_round_down_sectors(unsigned int sectors, unsigned int lbs)
+{
+       sectors = round_down(sectors, lbs >> SECTOR_SHIFT);
+       if (sectors < PAGE_SIZE >> SECTOR_SHIFT)
+               sectors = PAGE_SIZE >> SECTOR_SHIFT;
+       return sectors;
+}
+
 /**
  * blk_stack_limits - adjust queue_limits for stacked devices
  * @t: the stacking driver limits (top device)
@@ -586,6 +594,10 @@ int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
                ret = -1;
        }
 
+       t->max_sectors = blk_round_down_sectors(t->max_sectors, t->logical_block_size);
+       t->max_hw_sectors = blk_round_down_sectors(t->max_hw_sectors, t->logical_block_size);
+       t->max_dev_sectors = blk_round_down_sectors(t->max_dev_sectors, t->logical_block_size);
+
        /* Discard alignment and granularity */
        if (b->discard_granularity) {
                alignment = queue_limit_discard_alignment(b, start);