[PATCH 26/60] btrfs: set NO_MP for request queues behind BTRFS

[PATCH 26/60] btrfs: set NO_MP for request queues behind BTRFS

[Ming Lei]

There are lots of direct access to .bi_vcnt & .bi_io_vec
of bio, and it isn't ready to support multipage bvecs
for BTRFS, so set NO_MP for these request queues.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
---
 fs/btrfs/volumes.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 71a60cc01451..2e7237a3b84d 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1011,6 +1011,9 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		if (blk_queue_discard(q))
 			device->can_discard = 1;
 
+		/* BTRFS isn't ready to support multipage bvecs */
+		set_bit(QUEUE_FLAG_NO_MP, &q->queue_flags);
+
 		device->bdev = bdev;
 		device->in_fs_metadata = 0;
 		device->mode = flags;
-- 
2.7.4

Re: [PATCH 26/60] btrfs: set NO_MP for request queues behind BTRFS

[Christoph Hellwig]

On Sat, Oct 29, 2016 at 04:08:25PM +0800, Ming Lei wrote:
> There are lots of direct access to .bi_vcnt & .bi_io_vec
> of bio, and it isn't ready to support multipage bvecs
> for BTRFS, so set NO_MP for these request queues.

For one bio is an I/O submitter, it has absolutely no business changing
queue flags - if we need to stick to this limitation it simply needs
a version of bio_add_page that doesn't create multi-page bvecs.

Second I don't think making it multipage bvec aware is all that hard,
and we should aim for doing the proper thing.

Re: [PATCH 26/60] btrfs: set NO_MP for request queues behind BTRFS

[Chris Mason]

On Mon, Oct 31, 2016 at 08:36:44AM -0700, Christoph Hellwig wrote:
>On Sat, Oct 29, 2016 at 04:08:25PM +0800, Ming Lei wrote:
>> There are lots of direct access to .bi_vcnt & .bi_io_vec
>> of bio, and it isn't ready to support multipage bvecs
>> for BTRFS, so set NO_MP for these request queues.
>
>For one bio is an I/O submitter, it has absolutely no business changing
>queue flags - if we need to stick to this limitation it simply needs
>a version of bio_add_page that doesn't create multi-page bvecs.
>
>Second I don't think making it multipage bvec aware is all that hard,
>and we should aim for doing the proper thing.

Yeah, I'd rather make us less special.  The direct access was a short 
term fix to adjust to the new bio interfaces, we should clean it up.

-chris

Re: [PATCH 26/60] btrfs: set NO_MP for request queues behind BTRFS

[Christoph Hellwig]

On Mon, Oct 31, 2016 at 11:58:29AM -0600, Chris Mason wrote:
> Yeah, I'd rather make us less special.  The direct access was a short term
> fix to adjust to the new bio interfaces, we should clean it up.

I've got patches for a few areas in progress, I'll send them your way
once I've finished testing.  There will be a few more areas left where
I'll need a little help, though.