Re: [PATCH 07/12] dm: convert to bioset_init()/mempool_init()

[Mike Snitzer]

On Sun, May 20 2018 at  6:25pm -0400,
Kent Overstreet <kent.overstreet@xxxxxxxxx> wrote:

> Signed-off-by: Kent Overstreet <kent.overstreet@xxxxxxxxx>
> ---
>  drivers/md/dm-bio-prison-v1.c      | 13 ++++---
>  drivers/md/dm-bio-prison-v2.c      | 13 ++++---
>  drivers/md/dm-cache-target.c       | 25 ++++++-------
>  drivers/md/dm-core.h               |  4 +-
>  drivers/md/dm-crypt.c              | 60 ++++++++++++++----------------
>  drivers/md/dm-integrity.c          | 15 ++++----
>  drivers/md/dm-io.c                 | 29 ++++++++-------
>  drivers/md/dm-kcopyd.c             | 22 ++++++-----
>  drivers/md/dm-log-userspace-base.c | 19 +++++-----
>  drivers/md/dm-region-hash.c        | 23 ++++++------
>  drivers/md/dm-rq.c                 |  2 +-
>  drivers/md/dm-snap.c               | 17 ++++-----
>  drivers/md/dm-thin.c               | 32 ++++++++--------
>  drivers/md/dm-verity-fec.c         | 55 +++++++++++++--------------
>  drivers/md/dm-verity-fec.h         |  8 ++--
>  drivers/md/dm-zoned-target.c       | 13 +++----
>  drivers/md/dm.c                    | 53 ++++++++++++--------------
>  17 files changed, 197 insertions(+), 206 deletions(-)

In general I think each of these driver patches in the series needs a
one-line summary in the description, e.g.:
Embed bioset and mempool_t to avoid pointer indirection in the fast IO path.

But that aside, on a code level, I had to double check bioset_exit() is
safe to call even if bioset_init() failed.  AFter confirming it is (and
double checking by chatting with Kent) I like how the code is simpler
(not having to check if the bioset is set before calling it).

Nice job Kent, like you said: mostly mechnical.  But even the changes
that weren't purely mechnical (e.g. passing r to ERR_PTR, rather than
using old hardcode) were fixed up nicely.

Acked-by: Mike Snitzer <snitzer@xxxxxxxxxx>