Incorporated various review feedback, and I reordered/reworked the patches to
make them more coherent.
This also has the latest version of the closure code, incorporating review
feedback from Joe and Tejun. I may rewrite the bio splitting stuff to not
depend on it so it can go in sooner (Tejun is not a fan of closures), but IMO
the patch is better for using them.
First 5 patches: Kill bi_destructor
Next 3 patches: Better bio splitting
Next 3 patches: Rework bio cloning to not clone entire bio vec
Tejun also pointed out that with the bio splitting improvements we ought to be
able to get rid of merge_bvec_fn, so I started on that. Those patches are
definitely a WIP, though.
For most of the md code, it looks like things will almost just work - they
already had bio_pair_split() calls to handle single page bios that crossed bvec
boundries, but bio_pair_split() now splits arbitrary bios so things should just work.
(I just thought of one problem - previously, in practice you'd never need to
split a bio more than once, but that's no longer true - so the
bio_pair_split() calls do need to be stuck in a loop, and the as is the
patches aren't correct.)
For raid5 though I couldn't find any splitting in the existing code - it looks
like it's depending on merge_bvec_fn to make sure bios never need to be split.
Adding bio splitting looks to be not quite trivial - Neil, any thoughts?
For dm, it was already splitting multi page bios when necessary, it looked like
its merge_bvec_fn was mostly for propagating device limits - which my
generic_make_request() patch handles. Any dm folks want to take a look?
There are only 4 merge_bvec_fns left in my kernel:
I haven't looked at the last three, but if the maintainers want to help out a
bit it should be pretty easy to get rid of merge_bvec_fn entirely.
Kent Overstreet (16):
block: Generalized bio pool freeing
dm: Use bioset's front_pad for dm_rq_clone_bio_info
block: Add bio_reset()
pktcdvd: Switch to bio_kmalloc()
block: Kill bi_destructor
block: Add an explicit bio flag for bios that own their bvec
block: Rename bio_split() -> bio_pair_split()
block: Rework bio splitting
block: Add bio_clone_kmalloc()
block: Add bio_clone_bioset()
block: Only clone bio vecs that are in use
Closures
Make generic_make_request handle arbitrarily large bios
Gut bio_add_page()
md: Kill merge_bvec_fn()s
dm: Kill merge_bvec_fn()
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
07-24-2012, 08:11 PM
Kent Overstreet
Block cleanups
CHANGES SINCE LAST VERSION:
Split off the "make generic_make_request() handle arbitrary size bios" - this
patch series now just kills bi_destructor and reworks bio splitting.
Split out the "rework bio splitting" patch based on Boaz's feedback - now I've
got one patch that introduces the new bio_split() and another that rewrites
bio_pair_split(), which makes the patches much more readable.
Going to try and finish off the rest of the patches that get rid of
merge_bvec_fn, but that'll be a separate patch series.
Kent Overstreet (12):
block: Generalized bio pool freeing
dm: Use bioset's front_pad for dm_rq_clone_bio_info
block: Add bio_reset()
pktcdvd: Switch to bio_kmalloc()
block: Kill bi_destructor
block: Add an explicit bio flag for bios that own their bvec
block: Rename bio_split() -> bio_pair_split()
block: Introduce new bio_split()
block: Rework bio_pair_split()
block: Add bio_clone_kmalloc()
block: Add bio_clone_bioset()
block: Only clone bio vecs that are in use
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
08-06-2012, 10:08 PM
Kent Overstreet
Block cleanups
Various cleanups for the generic block layer - the patches should be pretty
self explanatory.
CHANGES SINCE LAST VERSION:
Review feedback - should all be noted in the patch descriptions. Fixed
retarded rebase conflicts. Added some acked-bys.
Kent Overstreet (12):
block: Generalized bio pool freeing
dm: Use bioset's front_pad for dm_rq_clone_bio_info
block: Add bio_reset()
pktcdvd: Switch to bio_kmalloc()
block: Kill bi_destructor
block: Add an explicit bio flag for bios that own their bvec
block: Rename bio_split() -> bio_pair_split()
block: Introduce new bio_split()
block: Rework bio_pair_split()
block: Add bio_clone_kmalloc()
block: Add bio_clone_bioset()
block: Only clone bio vecs that are in use
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
08-22-2012, 05:03 PM
Kent Overstreet
Block cleanups
Buncha changes in this version.
I came up with a new, much better solution to the "bio allocation from
bio sets under generic_make_request()" problem. Previously, I was
working around this in all of my code that allocated bios; it'd mask out
GFP_WAIT if current->bio_list != NULL, then punt to workqueue if the
allocation failed - so as to avoid the deadlock from allocating multiple
bios from the same bio set.
This approach worked, but it required it to be explicitly handled in any
stacking driver that might split a bio an arbitrary number of times -
less than ideal.
My new patch inverts this - when we go to allocate a bio and we're
blocking another bio from being submitted, we punt the bio(s) that are
being blocked off to a workqueue to be submitted. This means we only
have to handle it in one place, in bio_alloc_bioset() - stack block
drivers no longer have to do anything special (I got to delete some code
from bcache, and dm wasn't handling this at all before).
Other big change - I consolidated bio allocation. Now bio_alloc(),
bio_kmalloc(), bio_clone(), and bio_clone_kmalloc() are all one line
wrappers.
Incorporated a bunch of review feedback, most of which I hopefully
remembered to note in the patch descriptions this time.
Kent Overstreet (13):
block: Generalized bio pool freeing
dm: Use bioset's front_pad for dm_rq_clone_bio_info
block: Add bio_reset()
pktcdvd: Switch to bio_kmalloc()
block: Kill bi_destructor
block: Consolidate bio_alloc_bioset(), bio_kmalloc()
block: Avoid deadlocks with bio allocation by stacking drivers
block: Add an explicit bio flag for bios that own their bvec
block: Rename bio_split() -> bio_pair_split()
block: Introduce new bio_split()
block: Rework bio_pair_split()
block: Add bio_clone_bioset(), bio_clone_kmalloc()
block: Only clone bio vecs that are in use
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
08-23-2012, 06:00 PM
Vivek Goyal
Block cleanups
On Wed, Aug 22, 2012 at 10:03:57AM -0700, Kent Overstreet wrote:
[..]
> Kent Overstreet (13):
> block: Generalized bio pool freeing
> dm: Use bioset's front_pad for dm_rq_clone_bio_info
> block: Add bio_reset()
> pktcdvd: Switch to bio_kmalloc()
> block: Kill bi_destructor
> block: Consolidate bio_alloc_bioset(), bio_kmalloc()
This patch series says "Block Cleanups" in the subject. I think that
it can be divided in two parts. First part seems to be cleanup where
we get rid of ->bi_destructor and merge kmalloc() and allocation
from bio_set.
Second part seems to be more of enhancing bio splitting functionality
and not necessarily a cleanup. So may be breaking this series in two
parts makes sense. First part is cleanup and second part is enhanced
bio splitting capabilities of block layer.
Thanks
Vivek
> block: Avoid deadlocks with bio allocation by stacking drivers
> block: Add an explicit bio flag for bios that own their bvec
> block: Rename bio_split() -> bio_pair_split()
> block: Introduce new bio_split()
> block: Rework bio_pair_split()
> block: Add bio_clone_bioset(), bio_clone_kmalloc()
> block: Only clone bio vecs that are in use
>
> Documentation/block/biodoc.txt | 5 -
> block/blk-core.c | 10 +-
> drivers/block/drbd/drbd_main.c | 13 +-
> drivers/block/drbd/drbd_req.c | 18 +-
> drivers/block/osdblk.c | 3 +-
> drivers/block/pktcdvd.c | 73 ++-----
> drivers/block/rbd.c | 8 +-
> drivers/md/dm-crypt.c | 9 -
> drivers/md/dm-io.c | 11 -
> drivers/md/dm.c | 68 ++----
> drivers/md/linear.c | 6 +-
> drivers/md/md.c | 44 +---
> drivers/md/raid0.c | 8 +-
> drivers/md/raid10.c | 23 +-
> drivers/target/target_core_iblock.c | 9 -
> fs/bio-integrity.c | 45 ----
> fs/bio.c | 419 +++++++++++++++++++++++-------------
> fs/exofs/ore.c | 5 +-
> include/linux/bio.h | 155 +++++++------
> include/linux/blk_types.h | 16 +-
> 20 files changed, 434 insertions(+), 514 deletions(-)
>
> --
> 1.7.12
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
08-24-2012, 12:46 PM
Kent Overstreet
Block cleanups
On Thu, Aug 23, 2012 at 02:00:41PM -0400, Vivek Goyal wrote:
> On Wed, Aug 22, 2012 at 10:03:57AM -0700, Kent Overstreet wrote:
>
> [..]
> > Kent Overstreet (13):
> > block: Generalized bio pool freeing
> > dm: Use bioset's front_pad for dm_rq_clone_bio_info
> > block: Add bio_reset()
> > pktcdvd: Switch to bio_kmalloc()
> > block: Kill bi_destructor
> > block: Consolidate bio_alloc_bioset(), bio_kmalloc()
>
> This patch series says "Block Cleanups" in the subject. I think that
> it can be divided in two parts. First part seems to be cleanup where
> we get rid of ->bi_destructor and merge kmalloc() and allocation
> from bio_set.
>
> Second part seems to be more of enhancing bio splitting functionality
> and not necessarily a cleanup. So may be breaking this series in two
> parts makes sense. First part is cleanup and second part is enhanced
> bio splitting capabilities of block layer.
Yeah, that's a good idea - will do.
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
09-05-2012, 08:27 PM
Kent Overstreet
Block cleanups
Since v7:
Split off the deadlock fix, this is now just cleanups
Noticed a minor issue with bio integrity freeing, and decided the best way to
fix it was to just make it use bi_pool directly which simplifies the code
anyways.
Tested the integrity stuff with the scsi debug module.
Cleaned up bio_reset()/bio_free() a bit, introduce a __bio_free() helper used
by both and made bio_free() static.
Kent Overstreet (8):
block: Generalized bio pool freeing
block: Ues bi_pool for bio_integrity_alloc()
dm: Use bioset's front_pad for dm_rq_clone_bio_info
block: Add bio_reset()
pktcdvd: Switch to bio_kmalloc()
block: Kill bi_destructor
block: Consolidate bio_alloc_bioset(), bio_kmalloc()
block: Add bio_clone_bioset(), bio_clone_kmalloc()
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
09-06-2012, 11:48 PM
Tejun Heo
Block cleanups
Hello, guys.
(cc'ing Jens, Alasdair and Neil)
On Thu, Sep 06, 2012 at 03:34:54PM -0700, Kent Overstreet wrote:
> Screwed up the bio_reset() patch in the last patch series when I went to edit
> the description, fixed that here.
>
> Only other change is the dm patch - made the front_pad conditional on
> DM_TYPE_BIO_BASED.
>
> Kent Overstreet (8):
> block: Generalized bio pool freeing
> block: Ues bi_pool for bio_integrity_alloc()
> dm: Use bioset's front_pad for dm_rq_clone_bio_info
> block: Add bio_reset()
> pktcdvd: Switch to bio_kmalloc()
> block: Kill bi_destructor
> block: Consolidate bio_alloc_bioset(), bio_kmalloc()
> block: Add bio_clone_bioset(), bio_clone_kmalloc()
This series looks good to me now. If someone can ack the dm patch, I
think it's good to go. Jens, what do you think?
Thanks.
--
tejun
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
Block cleanups
