On Tue, 2017-06-06 at 17:19 +0800, Eryu Guan wrote:

On Wed, May 31, 2017 at 09:08:20AM -0400, Jeff Layton wrote:

quoted

With btrfs, we can't really put the log on a separate device. What we
can do however is mirror the metadata across two devices and make the
data striped across all devices. When we turn on dmerror then the
metadata can fall back to using the other mirror while the data errors
out.

Note that the current incarnation of btrfs has a fixed 64k stripe
width. If that ever changes or becomes settable, we may need to adjust
the amount of data that the test program writes.

Signed-off-by: Jeff Layton <redacted>
---
 common/rc | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/common/rc b/common/rc
index 83765aacfb06..078270451b53 100644
--- a/common/rc
+++ b/common/rc

@@ -830,6 +830,8 @@ _scratch_mkfs()
 		;;
 	btrfs)
 		mkfs_cmd="$MKFS_BTRFS_PROG"
+		[ "$USE_EXTERNAL" = yes -a ! -z "$SCRATCH_LOGDEV" ] && \
+			mkfs_cmd="$mkfs_cmd -d raid0 -m raid1 $SCRATCH_LOGDEV"

I don't think this is the correct way to do it. If btrfs doesn't support
external log device, then this test doesn't fit btrfs, or we need other
ways to test btrfs.

One of the problems of this hack is that raid1 requires all devices are
in the same size, we have a _require_scratch_dev_pool_equal_size() rule
to check on it, but this hack doesn't do the proper check and test fails
if SCRATCH_LOGDEV is smaller or bigger in size.

If btrfs "-d raid0 -m raid1" is capable to do this writeback error test,
perhaps you can write a new btrfs test and mkfs with "-d raid0 -m raid1"
explicitly. e.g.

...
_require_scratch_dev_pool 2
_require_scratch_dev_pool_equal_size
...
_scratch_mkfs "-d raid0 -m raid1"
...

Thanks,
Eryu

Yeah, that's probably the right way to do this. It looks like btrfs also
has $SCRATCH_DEV_POOL, and we can probably base it on that. I'll look at
reworking it.

-- 
Jeff Layton [off-list ref]