Re: [PATCH V5 00/30] loop: Issue O_DIRECT aio using bio_vec

[Jeff Moyer <jmoyer@redhat.com>]

Dave Kleikamp <dave.kleikamp@oracle.com> writes:

> This patchset was begun by Zach Brown and was originally submitted for
> review in October, 2009. Feedback was positive, and I have picked up
> where he left off, porting his patches to the latest mainline kernel
> and adding support more file systems.

[snip]

> My hopes are that this patchset is finally ready for linux-next.

Hi, Shaggy,

I'm finally getting around to testing this using xfstests.  This is my
setup:

dd if=/dev/zero of=testdev.img bs=1M count=1024
dd if=/dev/zero of=scratchdev.img bs=1M count=1024
losetup -f ./testdev.img
losetup -f ./scratchdev.img
mkfs -t xfs /dev/loop0
mount /dev/loop0 /mnt/test
export TEST_DIR=/mnt/test
export TEST_DEV=/dev/loop0
export SCRATCH_MNT=/mnt/scratch
export SCRATCH_DEV=/dev/loop1

I then ran:

./check -g aio

and here is the summary:

Ran: 112 113 198 207 208 209 210 211 212 239 240
Failures: 112 198 207 239 240
Failed 5 of 11 tests

To be fair, I have not yet run this on a kernel without your changes.
I'm going to do that next, but wanted to give you a heads-up, in case
you had time to dig into the failures.  Let me know if you want the full
output from the run, but I assume you'll want to reproduce this anyway.

Oh, I pulled your git branch into a 3.8.0-rc4 kernel tree, FYI.

Cheers,
Jeff