Re: stop using buffer heads in xfs and iomap

On Fri, May 11, 2018 at 08:22:08AM +0200, Christoph Hellwig wrote:

On Thu, May 10, 2018 at 08:13:03AM -0700, Darrick J. Wong wrote:

quoted

I ran xfstests on this for fun last night but hung in g/095:

FSTYP         -- xfs (debug)
PLATFORM      -- Linux/x86_64 submarine-djwong-mtr01 4.17.0-rc4-djw
MKFS_OPTIONS  -- -f -m reflink=1,rmapbt=1, -i sparse=1, -b size=1024, /dev/sdf
MOUNT_OPTIONS -- /dev/sdf /opt

FWIW the stock v4 and the 'v5 with everything and 4k blocks' vms
passed, so I guess there's a bug somewhere in the sub-page block size
code paths...

I haven't seen that in my -b size 1024 -m relink test  I'll try your above
exact setup, too.  Is this disk or SSD?  How much memory and how many
CPUs?

4 CPUs in a VM on a Nehalem-era machine, 2GB RAM, two 2.3GB virtio-scsi
disks...

...the VM host itself is a quad-core Nehalem, 24G RAM, atop an ext4 fs
on spinning rust in a raid1.

Btw, I think the series might be worthwhile even if we have to delay
the sub-page blocksize support - the blocksize == pagesize code is
basically entirely separate and already very useful.  Only the last
three patches contain the small blocksize support, without that we'll
just continue using buffer heads for that case for now.

<shrug> I'll keep reading, it seemed generally ok until I hit the
sub-page part and my eyes glazed over. :)

--D

--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html