From 320ae51feed5c2f13664aa05a76bec198967e04d Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Thu, 24 Oct 2013 09:20:05 +0100 Subject: blk-mq: new multi-queue block IO queueing mechanism Linux currently has two models for block devices: - The classic request_fn based approach, where drivers use struct request units for IO. The block layer provides various helper functionalities to let drivers share code, things like tag management, timeout handling, queueing, etc. - The "stacked" approach, where a driver squeezes in between the block layer and IO submitter. Since this bypasses the IO stack, driver generally have to manage everything themselves. With drivers being written for new high IOPS devices, the classic request_fn based driver doesn't work well enough. The design dates back to when both SMP and high IOPS was rare. It has problems with scaling to bigger machines, and runs into scaling issues even on smaller machines when you have IOPS in the hundreds of thousands per device. The stacked approach is then most often selected as the model for the driver. But this means that everybody has to re-invent everything, and along with that we get all the problems again that the shared approach solved. This commit introduces blk-mq, block multi queue support. The design is centered around per-cpu queues for queueing IO, which then funnel down into x number of hardware submission queues. We might have a 1:1 mapping between the two, or it might be an N:M mapping. That all depends on what the hardware supports. blk-mq provides various helper functions, which include: - Scalable support for request tagging. Most devices need to be able to uniquely identify a request both in the driver and to the hardware. The tagging uses per-cpu caches for freed tags, to enable cache hot reuse. - Timeout handling without tracking request on a per-device basis. Basically the driver should be able to get a notification, if a request happens to fail. - Optional support for non 1:1 mappings between issue and submission queues. blk-mq can redirect IO completions to the desired location. - Support for per-request payloads. Drivers almost always need to associate a request structure with some driver private command structure. Drivers can tell blk-mq this at init time, and then any request handed to the driver will have the required size of memory associated with it. - Support for merging of IO, and plugging. The stacked model gets neither of these. Even for high IOPS devices, merging sequential IO reduces per-command overhead and thus increases bandwidth. For now, this is provided as a potential 3rd queueing model, with the hope being that, as it matures, it can replace both the classic and stacked model. That would get us back to having just 1 real model for block devices, leaving the stacked approach to dm/md devices (as it was originally intended). Contributions in this patch from the following people: Shaohua Li Alexander Gordeev Christoph Hellwig Mike Christie Matias Bjorling Jeff Moyer Acked-by: Christoph Hellwig Signed-off-by: Jens Axboe --- include/linux/bio.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include/linux/bio.h') diff --git a/include/linux/bio.h b/include/linux/bio.h index ec48bac5b039..4c2775443dcf 100644 --- a/include/linux/bio.h +++ b/include/linux/bio.h @@ -419,6 +419,8 @@ static inline void bio_list_init(struct bio_list *bl) bl->head = bl->tail = NULL; } +#define BIO_EMPTY_LIST { NULL, NULL } + #define bio_list_for_each(bio, bl) \ for (bio = (bl)->head; bio; bio = bio->bi_next) -- cgit v1.2.3 From 6678d83f18386eb103f8345024e52c5abe61725c Mon Sep 17 00:00:00 2001 From: Kent Overstreet Date: Wed, 7 Aug 2013 11:14:32 -0700 Subject: block: Consolidate duplicated bio_trim() implementations Someone cut and pasted md's md_trim_bio() into xen-blkfront.c. Come on, we should know better than this. Signed-off-by: Kent Overstreet Cc: Jens Axboe Cc: Neil Brown Cc: Konrad Rzeszutek Wilk Cc: Jeremy Fitzhardinge Signed-off-by: Jens Axboe --- drivers/block/xen-blkfront.c | 53 +------------------------------------------- drivers/md/md.c | 40 --------------------------------- drivers/md/md.h | 1 - drivers/md/raid1.c | 10 ++++----- drivers/md/raid10.c | 18 +++++++-------- fs/bio.c | 46 ++++++++++++++++++++++++++++++++++++++ include/linux/bio.h | 1 + 7 files changed, 61 insertions(+), 108 deletions(-) (limited to 'include/linux/bio.h') diff --git a/drivers/block/xen-blkfront.c b/drivers/block/xen-blkfront.c index a4660bbee8a6..8d53ed293606 100644 --- a/drivers/block/xen-blkfront.c +++ b/drivers/block/xen-blkfront.c @@ -1336,57 +1336,6 @@ static int blkfront_probe(struct xenbus_device *dev, return 0; } -/* - * This is a clone of md_trim_bio, used to split a bio into smaller ones - */ -static void trim_bio(struct bio *bio, int offset, int size) -{ - /* 'bio' is a cloned bio which we need to trim to match - * the given offset and size. - * This requires adjusting bi_sector, bi_size, and bi_io_vec - */ - int i; - struct bio_vec *bvec; - int sofar = 0; - - size <<= 9; - if (offset == 0 && size == bio->bi_size) - return; - - bio->bi_sector += offset; - bio->bi_size = size; - offset <<= 9; - clear_bit(BIO_SEG_VALID, &bio->bi_flags); - - while (bio->bi_idx < bio->bi_vcnt && - bio->bi_io_vec[bio->bi_idx].bv_len <= offset) { - /* remove this whole bio_vec */ - offset -= bio->bi_io_vec[bio->bi_idx].bv_len; - bio->bi_idx++; - } - if (bio->bi_idx < bio->bi_vcnt) { - bio->bi_io_vec[bio->bi_idx].bv_offset += offset; - bio->bi_io_vec[bio->bi_idx].bv_len -= offset; - } - /* avoid any complications with bi_idx being non-zero*/ - if (bio->bi_idx) { - memmove(bio->bi_io_vec, bio->bi_io_vec+bio->bi_idx, - (bio->bi_vcnt - bio->bi_idx) * sizeof(struct bio_vec)); - bio->bi_vcnt -= bio->bi_idx; - bio->bi_idx = 0; - } - /* Make sure vcnt and last bv are not too big */ - bio_for_each_segment(bvec, bio, i) { - if (sofar + bvec->bv_len > size) - bvec->bv_len = size - sofar; - if (bvec->bv_len == 0) { - bio->bi_vcnt = i; - break; - } - sofar += bvec->bv_len; - } -} - static void split_bio_end(struct bio *bio, int error) { struct split_bio *split_bio = bio->bi_private; @@ -1522,7 +1471,7 @@ static int blkif_recover(struct blkfront_info *info) (unsigned int)(bio->bi_size >> 9) - offset); cloned_bio = bio_clone(bio, GFP_NOIO); BUG_ON(cloned_bio == NULL); - trim_bio(cloned_bio, offset, size); + bio_trim(cloned_bio, offset, size); cloned_bio->bi_private = split_bio; cloned_bio->bi_end_io = split_bio_end; submit_bio(cloned_bio->bi_rw, cloned_bio); diff --git a/drivers/md/md.c b/drivers/md/md.c index 561a65f82e26..752119068d66 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -183,46 +183,6 @@ struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, } EXPORT_SYMBOL_GPL(bio_clone_mddev); -void md_trim_bio(struct bio *bio, int offset, int size) -{ - /* 'bio' is a cloned bio which we need to trim to match - * the given offset and size. - * This requires adjusting bi_sector, bi_size, and bi_io_vec - */ - int i; - struct bio_vec *bvec; - int sofar = 0; - - size <<= 9; - if (offset == 0 && size == bio->bi_size) - return; - - clear_bit(BIO_SEG_VALID, &bio->bi_flags); - - bio_advance(bio, offset << 9); - - bio->bi_size = size; - - /* avoid any complications with bi_idx being non-zero*/ - if (bio->bi_idx) { - memmove(bio->bi_io_vec, bio->bi_io_vec+bio->bi_idx, - (bio->bi_vcnt - bio->bi_idx) * sizeof(struct bio_vec)); - bio->bi_vcnt -= bio->bi_idx; - bio->bi_idx = 0; - } - /* Make sure vcnt and last bv are not too big */ - bio_for_each_segment(bvec, bio, i) { - if (sofar + bvec->bv_len > size) - bvec->bv_len = size - sofar; - if (bvec->bv_len == 0) { - bio->bi_vcnt = i; - break; - } - sofar += bvec->bv_len; - } -} -EXPORT_SYMBOL_GPL(md_trim_bio); - /* * We have a system wide 'event count' that is incremented * on any 'interesting' event, and readers of /proc/mdstat diff --git a/drivers/md/md.h b/drivers/md/md.h index 608050c43f17..c96456c16700 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -617,7 +617,6 @@ extern struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask, struct mddev *mddev); extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, struct mddev *mddev); -extern void md_trim_bio(struct bio *bio, int offset, int size); extern void md_unplug(struct blk_plug_cb *cb, bool from_schedule); static inline int mddev_check_plugged(struct mddev *mddev) diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index aacf6bf352d8..af6681b19776 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -1097,8 +1097,8 @@ read_again: r1_bio->read_disk = rdisk; read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev); - md_trim_bio(read_bio, r1_bio->sector - bio->bi_sector, - max_sectors); + bio_trim(read_bio, r1_bio->sector - bio->bi_sector, + max_sectors); r1_bio->bios[rdisk] = read_bio; @@ -1266,7 +1266,7 @@ read_again: continue; mbio = bio_clone_mddev(bio, GFP_NOIO, mddev); - md_trim_bio(mbio, r1_bio->sector - bio->bi_sector, max_sectors); + bio_trim(mbio, r1_bio->sector - bio->bi_sector, max_sectors); if (first_clone) { /* do behind I/O ? @@ -2126,7 +2126,7 @@ static int narrow_write_error(struct r1bio *r1_bio, int i) wbio->bi_sector = r1_bio->sector; wbio->bi_size = r1_bio->sectors << 9; - md_trim_bio(wbio, sector - r1_bio->sector, sectors); + bio_trim(wbio, sector - r1_bio->sector, sectors); wbio->bi_sector += rdev->data_offset; wbio->bi_bdev = rdev->bdev; if (submit_bio_wait(WRITE, wbio) == 0) @@ -2241,7 +2241,7 @@ read_more: } r1_bio->read_disk = disk; bio = bio_clone_mddev(r1_bio->master_bio, GFP_NOIO, mddev); - md_trim_bio(bio, r1_bio->sector - bio->bi_sector, max_sectors); + bio_trim(bio, r1_bio->sector - bio->bi_sector, max_sectors); r1_bio->bios[r1_bio->read_disk] = bio; rdev = conf->mirrors[disk].rdev; printk_ratelimited(KERN_ERR diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c index 73dc8a377522..7c3508abb5e1 100644 --- a/drivers/md/raid10.c +++ b/drivers/md/raid10.c @@ -1302,8 +1302,8 @@ read_again: slot = r10_bio->read_slot; read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev); - md_trim_bio(read_bio, r10_bio->sector - bio->bi_sector, - max_sectors); + bio_trim(read_bio, r10_bio->sector - bio->bi_sector, + max_sectors); r10_bio->devs[slot].bio = read_bio; r10_bio->devs[slot].rdev = rdev; @@ -1510,8 +1510,8 @@ retry_write: if (r10_bio->devs[i].bio) { struct md_rdev *rdev = conf->mirrors[d].rdev; mbio = bio_clone_mddev(bio, GFP_NOIO, mddev); - md_trim_bio(mbio, r10_bio->sector - bio->bi_sector, - max_sectors); + bio_trim(mbio, r10_bio->sector - bio->bi_sector, + max_sectors); r10_bio->devs[i].bio = mbio; mbio->bi_sector = (r10_bio->devs[i].addr+ @@ -1553,8 +1553,8 @@ retry_write: rdev = conf->mirrors[d].rdev; } mbio = bio_clone_mddev(bio, GFP_NOIO, mddev); - md_trim_bio(mbio, r10_bio->sector - bio->bi_sector, - max_sectors); + bio_trim(mbio, r10_bio->sector - bio->bi_sector, + max_sectors); r10_bio->devs[i].repl_bio = mbio; mbio->bi_sector = (r10_bio->devs[i].addr + @@ -2614,7 +2614,7 @@ static int narrow_write_error(struct r10bio *r10_bio, int i) sectors = sect_to_write; /* Write at 'sector' for 'sectors' */ wbio = bio_clone_mddev(bio, GFP_NOIO, mddev); - md_trim_bio(wbio, sector - bio->bi_sector, sectors); + bio_trim(wbio, sector - bio->bi_sector, sectors); wbio->bi_sector = (r10_bio->devs[i].addr+ choose_data_offset(r10_bio, rdev) + (sector - r10_bio->sector)); @@ -2687,9 +2687,7 @@ read_more: (unsigned long long)r10_bio->sector); bio = bio_clone_mddev(r10_bio->master_bio, GFP_NOIO, mddev); - md_trim_bio(bio, - r10_bio->sector - bio->bi_sector, - max_sectors); + bio_trim(bio, r10_bio->sector - bio->bi_sector, max_sectors); r10_bio->devs[slot].bio = bio; r10_bio->devs[slot].rdev = rdev; bio->bi_sector = r10_bio->devs[slot].addr diff --git a/fs/bio.c b/fs/bio.c index ea5035da4d9a..2bdb4e25ee77 100644 --- a/fs/bio.c +++ b/fs/bio.c @@ -1804,6 +1804,52 @@ struct bio_pair *bio_split(struct bio *bi, int first_sectors) } EXPORT_SYMBOL(bio_split); +/** + * bio_trim - trim a bio + * @bio: bio to trim + * @offset: number of sectors to trim from the front of @bio + * @size: size we want to trim @bio to, in sectors + */ +void bio_trim(struct bio *bio, int offset, int size) +{ + /* 'bio' is a cloned bio which we need to trim to match + * the given offset and size. + * This requires adjusting bi_sector, bi_size, and bi_io_vec + */ + int i; + struct bio_vec *bvec; + int sofar = 0; + + size <<= 9; + if (offset == 0 && size == bio->bi_size) + return; + + clear_bit(BIO_SEG_VALID, &bio->bi_flags); + + bio_advance(bio, offset << 9); + + bio->bi_size = size; + + /* avoid any complications with bi_idx being non-zero*/ + if (bio->bi_idx) { + memmove(bio->bi_io_vec, bio->bi_io_vec+bio->bi_idx, + (bio->bi_vcnt - bio->bi_idx) * sizeof(struct bio_vec)); + bio->bi_vcnt -= bio->bi_idx; + bio->bi_idx = 0; + } + /* Make sure vcnt and last bv are not too big */ + bio_for_each_segment(bvec, bio, i) { + if (sofar + bvec->bv_len > size) + bvec->bv_len = size - sofar; + if (bvec->bv_len == 0) { + bio->bi_vcnt = i; + break; + } + sofar += bvec->bv_len; + } +} +EXPORT_SYMBOL_GPL(bio_trim); + /** * bio_sector_offset - Find hardware sector offset in bio * @bio: bio to inspect diff --git a/include/linux/bio.h b/include/linux/bio.h index ec48bac5b039..162036aca741 100644 --- a/include/linux/bio.h +++ b/include/linux/bio.h @@ -218,6 +218,7 @@ struct bio_pair { }; extern struct bio_pair *bio_split(struct bio *bi, int first_sectors); extern void bio_pair_release(struct bio_pair *dbio); +extern void bio_trim(struct bio *bio, int offset, int size); extern struct bio_set *bioset_create(unsigned int, unsigned int); extern void bioset_free(struct bio_set *); -- cgit v1.2.3