RE: regression introduced by "block: Add support for DAX reads/writes to block devices"

["Wilcox, Matthew R" <matthew.r.wilcox@intel.com>]

Yes, that's the result I want.  Fundamentally, I think DAX should be able to support devices that are not multiples of PAGE_SIZE in size.

-----Original Message-----
From: Jeff Moyer [mailto:jmoyer@redhat.com] 
Sent: Thursday, August 06, 2015 8:34 AM
To: Wilcox, Matthew R
Cc: linda.knippers@hp.com; linux-kernel@vger.kernel.org; linux-fsdevel@vger.kernel.org
Subject: Re: regression introduced by "block: Add support for DAX reads/writes to block devices"

"Wilcox, Matthew R" <matthew.r.wilcox@intel.com> writes:

> I think I see the problem.  I'm kind of wrapped up in other things
> right now; can you try replacing the line in dax_io():
>
> -				bh->b_size = PAGE_ALIGN(end - pos);
> +				bh->b_size = ALIGN(end - pos, 1 << blkbits);

That's not gonna work either.  :)  You'll end up with -EINVAL since
bdev_direct_access wants the sector to be aligned to a page:

       if (sector % (PAGE_SIZE / 512))
                return -EINVAL;

I think you really want to call direct_access with the full page, and
then tease out the part you want up in dax_io, right?  I'll take a crack
at it if you're busy.

Cheers,
Jeff

> -----Original Message-----
> From: Jeff Moyer [mailto:jmoyer@redhat.com] 
> Sent: Wednesday, August 05, 2015 1:19 PM
> To: Wilcox, Matthew R; linda.knippers@hp.com
> Cc: linux-kernel@vger.kernel.org; linux-fsdevel@vger.kernel.org
> Subject: regression introduced by "block: Add support for DAX reads/writes to block devices"
>
> Hi, Matthew,
>
> Linda Knippers noticed that commit (bbab37ddc20b) breaks mkfs.xfs:
>
> # mkfs -t xfs -f /dev/pmem0
> meta-data=/dev/pmem0             isize=256    agcount=4, agsize=524288 blks
>          =                       sectsz=512   attr=2, projid32bit=1
>          =                       crc=0        finobt=0
> data     =                       bsize=4096   blocks=2097152, imaxpct=25
>          =                       sunit=0      swidth=0 blks
> naming   =version 2              bsize=4096   ascii-ci=0 ftype=0
> log      =internal log           bsize=4096   blocks=2560, version=2
>          =                       sectsz=512   sunit=0 blks, lazy-count=1
> realtime =none                   extsz=4096   blocks=0, rtextents=0
> mkfs.xfs: read failed: Numerical result out of range
>
> I sat down with Linda to look into it, and the problem is that mkfs.xfs
> sets the blocksize of the device to 512 (via BLKBSZSET), and then reads
> from the last sector of the device.  This results in dax_io trying to do
> a page-sized I/O at 512 bytes from the end of the device.
> bdev_direct_access, receiving this bogus pos/size combo, returns
> -ERANGE:
>
> 	if ((sector + DIV_ROUND_UP(size, 512)) >
> 					part_nr_sects_read(bdev->bd_part))
> 		return -ERANGE;
>
> Given that file systems supporting dax refuse to mount with a blocksize
> != page size, I'm guessing this is sort of expected behavior.  However,
> we really shouldn't be breaking direct I/O on pmem devices.
>
> So, what do you want to do?  We could make the pmem device's logical
> block size fixed at the sytem page size.  Or, we could modify the dax
> code to work with blocksize < pagesize.  Or, we could continue using the
> direct I/O codepath for direct block device access.  What do you think?
>
> Thaks,
> Jeff and Linda