dax: fix O_DIRECT I/O to the last block of a blockdev
authorJeff Moyer <jmoyer@redhat.com>
Fri, 14 Aug 2015 20:15:31 +0000 (16:15 -0400)
committerDan Williams <dan.j.williams@intel.com>
Wed, 16 Sep 2015 00:07:35 +0000 (20:07 -0400)
commite94f5a2285fc94202a9efb2c687481f29b64132c
tree2cb07ecaaaf15389bad53415ecf7c6143f07a9fa
parent6ff33f3902c3b1c5d0db6b1e2c70b6d76fba357f
dax: fix O_DIRECT I/O to the last block of a blockdev

commit bbab37ddc20b (block: Add support for DAX reads/writes to
block devices) caused a regression in mkfs.xfs.  That utility
sets the block size of the device to the logical block size
using the BLKBSZSET ioctl, and then issues a single sector read
from the last sector of the device.  This results in the dax_io
code trying to do a page-sized read from 512 bytes from the end
of the device.  The result is -ERANGE being returned to userspace.

The fix is to align the block to the page size before calling
get_block.

Thanks to willy for simplifying my original patch.

Cc: <stable@vger.kernel.org>
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Tested-by: Linda Knippers <linda.knippers@hp.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
fs/dax.c