2 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
13 * You should have received a copy of the GNU General Public Licens
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
19 #include <linux/swap.h>
20 #include <linux/bio.h>
21 #include <linux/blkdev.h>
22 #include <linux/uio.h>
23 #include <linux/iocontext.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/kernel.h>
27 #include <linux/export.h>
28 #include <linux/mempool.h>
29 #include <linux/workqueue.h>
30 #include <linux/cgroup.h>
32 #include <trace/events/block.h>
35 * Test patch to inline a certain number of bi_io_vec's inside the bio
36 * itself, to shrink a bio data allocation from two mempool calls to one
38 #define BIO_INLINE_VECS 4
41 * if you change this list, also change bvec_alloc or things will
42 * break badly! cannot be bigger than what you can fit into an
45 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
46 static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
47 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES),
52 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
53 * IO code that does not need private memory pools.
55 struct bio_set *fs_bio_set;
56 EXPORT_SYMBOL(fs_bio_set);
59 * Our slab pool management
62 struct kmem_cache *slab;
63 unsigned int slab_ref;
64 unsigned int slab_size;
67 static DEFINE_MUTEX(bio_slab_lock);
68 static struct bio_slab *bio_slabs;
69 static unsigned int bio_slab_nr, bio_slab_max;
71 static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size)
73 unsigned int sz = sizeof(struct bio) + extra_size;
74 struct kmem_cache *slab = NULL;
75 struct bio_slab *bslab, *new_bio_slabs;
76 unsigned int new_bio_slab_max;
77 unsigned int i, entry = -1;
79 mutex_lock(&bio_slab_lock);
82 while (i < bio_slab_nr) {
83 bslab = &bio_slabs[i];
85 if (!bslab->slab && entry == -1)
87 else if (bslab->slab_size == sz) {
98 if (bio_slab_nr == bio_slab_max && entry == -1) {
99 new_bio_slab_max = bio_slab_max << 1;
100 new_bio_slabs = krealloc(bio_slabs,
101 new_bio_slab_max * sizeof(struct bio_slab),
105 bio_slab_max = new_bio_slab_max;
106 bio_slabs = new_bio_slabs;
109 entry = bio_slab_nr++;
111 bslab = &bio_slabs[entry];
113 snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry);
114 slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN,
115 SLAB_HWCACHE_ALIGN, NULL);
121 bslab->slab_size = sz;
123 mutex_unlock(&bio_slab_lock);
127 static void bio_put_slab(struct bio_set *bs)
129 struct bio_slab *bslab = NULL;
132 mutex_lock(&bio_slab_lock);
134 for (i = 0; i < bio_slab_nr; i++) {
135 if (bs->bio_slab == bio_slabs[i].slab) {
136 bslab = &bio_slabs[i];
141 if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n"))
144 WARN_ON(!bslab->slab_ref);
146 if (--bslab->slab_ref)
149 kmem_cache_destroy(bslab->slab);
153 mutex_unlock(&bio_slab_lock);
156 unsigned int bvec_nr_vecs(unsigned short idx)
158 return bvec_slabs[idx].nr_vecs;
161 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx)
163 BIO_BUG_ON(idx >= BIOVEC_NR_POOLS);
165 if (idx == BIOVEC_MAX_IDX)
166 mempool_free(bv, pool);
168 struct biovec_slab *bvs = bvec_slabs + idx;
170 kmem_cache_free(bvs->slab, bv);
174 struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx,
180 * see comment near bvec_array define!
198 case 129 ... BIO_MAX_PAGES:
206 * idx now points to the pool we want to allocate from. only the
207 * 1-vec entry pool is mempool backed.
209 if (*idx == BIOVEC_MAX_IDX) {
211 bvl = mempool_alloc(pool, gfp_mask);
213 struct biovec_slab *bvs = bvec_slabs + *idx;
214 gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO);
217 * Make this allocation restricted and don't dump info on
218 * allocation failures, since we'll fallback to the mempool
219 * in case of failure.
221 __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN;
224 * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM
225 * is set, retry with the 1-entry mempool
227 bvl = kmem_cache_alloc(bvs->slab, __gfp_mask);
228 if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) {
229 *idx = BIOVEC_MAX_IDX;
237 static void __bio_free(struct bio *bio)
239 bio_disassociate_task(bio);
241 if (bio_integrity(bio))
242 bio_integrity_free(bio);
245 static void bio_free(struct bio *bio)
247 struct bio_set *bs = bio->bi_pool;
253 if (bio_flagged(bio, BIO_OWNS_VEC))
254 bvec_free(bs->bvec_pool, bio->bi_io_vec, BIO_POOL_IDX(bio));
257 * If we have front padding, adjust the bio pointer before freeing
262 mempool_free(p, bs->bio_pool);
264 /* Bio was allocated by bio_kmalloc() */
269 void bio_init(struct bio *bio)
271 memset(bio, 0, sizeof(*bio));
272 atomic_set(&bio->__bi_remaining, 1);
273 atomic_set(&bio->__bi_cnt, 1);
275 EXPORT_SYMBOL(bio_init);
278 * bio_reset - reinitialize a bio
282 * After calling bio_reset(), @bio will be in the same state as a freshly
283 * allocated bio returned bio bio_alloc_bioset() - the only fields that are
284 * preserved are the ones that are initialized by bio_alloc_bioset(). See
285 * comment in struct bio.
287 void bio_reset(struct bio *bio)
289 unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS);
293 memset(bio, 0, BIO_RESET_BYTES);
294 bio->bi_flags = flags;
295 atomic_set(&bio->__bi_remaining, 1);
297 EXPORT_SYMBOL(bio_reset);
299 static void bio_chain_endio(struct bio *bio)
301 struct bio *parent = bio->bi_private;
303 parent->bi_error = bio->bi_error;
309 * Increment chain count for the bio. Make sure the CHAIN flag update
310 * is visible before the raised count.
312 static inline void bio_inc_remaining(struct bio *bio)
314 bio_set_flag(bio, BIO_CHAIN);
315 smp_mb__before_atomic();
316 atomic_inc(&bio->__bi_remaining);
320 * bio_chain - chain bio completions
321 * @bio: the target bio
322 * @parent: the @bio's parent bio
324 * The caller won't have a bi_end_io called when @bio completes - instead,
325 * @parent's bi_end_io won't be called until both @parent and @bio have
326 * completed; the chained bio will also be freed when it completes.
328 * The caller must not set bi_private or bi_end_io in @bio.
330 void bio_chain(struct bio *bio, struct bio *parent)
332 BUG_ON(bio->bi_private || bio->bi_end_io);
334 bio->bi_private = parent;
335 bio->bi_end_io = bio_chain_endio;
336 bio_inc_remaining(parent);
338 EXPORT_SYMBOL(bio_chain);
340 static void bio_alloc_rescue(struct work_struct *work)
342 struct bio_set *bs = container_of(work, struct bio_set, rescue_work);
346 spin_lock(&bs->rescue_lock);
347 bio = bio_list_pop(&bs->rescue_list);
348 spin_unlock(&bs->rescue_lock);
353 generic_make_request(bio);
357 static void punt_bios_to_rescuer(struct bio_set *bs)
359 struct bio_list punt, nopunt;
363 * In order to guarantee forward progress we must punt only bios that
364 * were allocated from this bio_set; otherwise, if there was a bio on
365 * there for a stacking driver higher up in the stack, processing it
366 * could require allocating bios from this bio_set, and doing that from
367 * our own rescuer would be bad.
369 * Since bio lists are singly linked, pop them all instead of trying to
370 * remove from the middle of the list:
373 bio_list_init(&punt);
374 bio_list_init(&nopunt);
376 while ((bio = bio_list_pop(¤t->bio_list[0])))
377 bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio);
378 current->bio_list[0] = nopunt;
380 bio_list_init(&nopunt);
381 while ((bio = bio_list_pop(¤t->bio_list[1])))
382 bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio);
383 current->bio_list[1] = nopunt;
385 spin_lock(&bs->rescue_lock);
386 bio_list_merge(&bs->rescue_list, &punt);
387 spin_unlock(&bs->rescue_lock);
389 queue_work(bs->rescue_workqueue, &bs->rescue_work);
393 * bio_alloc_bioset - allocate a bio for I/O
394 * @gfp_mask: the GFP_ mask given to the slab allocator
395 * @nr_iovecs: number of iovecs to pre-allocate
396 * @bs: the bio_set to allocate from.
399 * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is
400 * backed by the @bs's mempool.
402 * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will
403 * always be able to allocate a bio. This is due to the mempool guarantees.
404 * To make this work, callers must never allocate more than 1 bio at a time
405 * from this pool. Callers that need to allocate more than 1 bio must always
406 * submit the previously allocated bio for IO before attempting to allocate
407 * a new one. Failure to do so can cause deadlocks under memory pressure.
409 * Note that when running under generic_make_request() (i.e. any block
410 * driver), bios are not submitted until after you return - see the code in
411 * generic_make_request() that converts recursion into iteration, to prevent
414 * This would normally mean allocating multiple bios under
415 * generic_make_request() would be susceptible to deadlocks, but we have
416 * deadlock avoidance code that resubmits any blocked bios from a rescuer
419 * However, we do not guarantee forward progress for allocations from other
420 * mempools. Doing multiple allocations from the same mempool under
421 * generic_make_request() should be avoided - instead, use bio_set's front_pad
422 * for per bio allocations.
425 * Pointer to new bio on success, NULL on failure.
427 struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
429 gfp_t saved_gfp = gfp_mask;
431 unsigned inline_vecs;
432 unsigned long idx = BIO_POOL_NONE;
433 struct bio_vec *bvl = NULL;
438 if (nr_iovecs > UIO_MAXIOV)
441 p = kmalloc(sizeof(struct bio) +
442 nr_iovecs * sizeof(struct bio_vec),
445 inline_vecs = nr_iovecs;
447 /* should not use nobvec bioset for nr_iovecs > 0 */
448 if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0))
451 * generic_make_request() converts recursion to iteration; this
452 * means if we're running beneath it, any bios we allocate and
453 * submit will not be submitted (and thus freed) until after we
456 * This exposes us to a potential deadlock if we allocate
457 * multiple bios from the same bio_set() while running
458 * underneath generic_make_request(). If we were to allocate
459 * multiple bios (say a stacking block driver that was splitting
460 * bios), we would deadlock if we exhausted the mempool's
463 * We solve this, and guarantee forward progress, with a rescuer
464 * workqueue per bio_set. If we go to allocate and there are
465 * bios on current->bio_list, we first try the allocation
466 * without __GFP_DIRECT_RECLAIM; if that fails, we punt those
467 * bios we would be blocking to the rescuer workqueue before
468 * we retry with the original gfp_flags.
471 if (current->bio_list &&
472 (!bio_list_empty(¤t->bio_list[0]) ||
473 !bio_list_empty(¤t->bio_list[1])))
474 gfp_mask &= ~__GFP_DIRECT_RECLAIM;
476 p = mempool_alloc(bs->bio_pool, gfp_mask);
477 if (!p && gfp_mask != saved_gfp) {
478 punt_bios_to_rescuer(bs);
479 gfp_mask = saved_gfp;
480 p = mempool_alloc(bs->bio_pool, gfp_mask);
483 front_pad = bs->front_pad;
484 inline_vecs = BIO_INLINE_VECS;
493 if (nr_iovecs > inline_vecs) {
494 bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool);
495 if (!bvl && gfp_mask != saved_gfp) {
496 punt_bios_to_rescuer(bs);
497 gfp_mask = saved_gfp;
498 bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool);
504 bio_set_flag(bio, BIO_OWNS_VEC);
505 } else if (nr_iovecs) {
506 bvl = bio->bi_inline_vecs;
510 bio->bi_flags |= idx << BIO_POOL_OFFSET;
511 bio->bi_max_vecs = nr_iovecs;
512 bio->bi_io_vec = bvl;
516 mempool_free(p, bs->bio_pool);
519 EXPORT_SYMBOL(bio_alloc_bioset);
521 void zero_fill_bio(struct bio *bio)
525 struct bvec_iter iter;
527 bio_for_each_segment(bv, bio, iter) {
528 char *data = bvec_kmap_irq(&bv, &flags);
529 memset(data, 0, bv.bv_len);
530 flush_dcache_page(bv.bv_page);
531 bvec_kunmap_irq(data, &flags);
534 EXPORT_SYMBOL(zero_fill_bio);
537 * bio_put - release a reference to a bio
538 * @bio: bio to release reference to
541 * Put a reference to a &struct bio, either one you have gotten with
542 * bio_alloc, bio_get or bio_clone. The last put of a bio will free it.
544 void bio_put(struct bio *bio)
546 if (!bio_flagged(bio, BIO_REFFED))
549 BIO_BUG_ON(!atomic_read(&bio->__bi_cnt));
554 if (atomic_dec_and_test(&bio->__bi_cnt))
558 EXPORT_SYMBOL(bio_put);
560 inline int bio_phys_segments(struct request_queue *q, struct bio *bio)
562 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
563 blk_recount_segments(q, bio);
565 return bio->bi_phys_segments;
567 EXPORT_SYMBOL(bio_phys_segments);
570 * __bio_clone_fast - clone a bio that shares the original bio's biovec
571 * @bio: destination bio
572 * @bio_src: bio to clone
574 * Clone a &bio. Caller will own the returned bio, but not
575 * the actual data it points to. Reference count of returned
578 * Caller must ensure that @bio_src is not freed before @bio.
580 void __bio_clone_fast(struct bio *bio, struct bio *bio_src)
582 BUG_ON(bio->bi_pool && BIO_POOL_IDX(bio) != BIO_POOL_NONE);
585 * most users will be overriding ->bi_bdev with a new target,
586 * so we don't set nor calculate new physical/hw segment counts here
588 bio->bi_bdev = bio_src->bi_bdev;
589 bio_set_flag(bio, BIO_CLONED);
590 bio->bi_rw = bio_src->bi_rw;
591 bio->bi_iter = bio_src->bi_iter;
592 bio->bi_io_vec = bio_src->bi_io_vec;
594 bio_clone_blkcg_association(bio, bio_src);
596 EXPORT_SYMBOL(__bio_clone_fast);
599 * bio_clone_fast - clone a bio that shares the original bio's biovec
601 * @gfp_mask: allocation priority
602 * @bs: bio_set to allocate from
604 * Like __bio_clone_fast, only also allocates the returned bio
606 struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs)
610 b = bio_alloc_bioset(gfp_mask, 0, bs);
614 __bio_clone_fast(b, bio);
616 if (bio_integrity(bio)) {
619 ret = bio_integrity_clone(b, bio, gfp_mask);
629 EXPORT_SYMBOL(bio_clone_fast);
632 * bio_clone_bioset - clone a bio
633 * @bio_src: bio to clone
634 * @gfp_mask: allocation priority
635 * @bs: bio_set to allocate from
637 * Clone bio. Caller will own the returned bio, but not the actual data it
638 * points to. Reference count of returned bio will be one.
640 struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask,
643 struct bvec_iter iter;
648 * Pre immutable biovecs, __bio_clone() used to just do a memcpy from
649 * bio_src->bi_io_vec to bio->bi_io_vec.
651 * We can't do that anymore, because:
653 * - The point of cloning the biovec is to produce a bio with a biovec
654 * the caller can modify: bi_idx and bi_bvec_done should be 0.
656 * - The original bio could've had more than BIO_MAX_PAGES biovecs; if
657 * we tried to clone the whole thing bio_alloc_bioset() would fail.
658 * But the clone should succeed as long as the number of biovecs we
659 * actually need to allocate is fewer than BIO_MAX_PAGES.
661 * - Lastly, bi_vcnt should not be looked at or relied upon by code
662 * that does not own the bio - reason being drivers don't use it for
663 * iterating over the biovec anymore, so expecting it to be kept up
664 * to date (i.e. for clones that share the parent biovec) is just
665 * asking for trouble and would force extra work on
666 * __bio_clone_fast() anyways.
669 bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs);
673 bio->bi_bdev = bio_src->bi_bdev;
674 bio->bi_rw = bio_src->bi_rw;
675 bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector;
676 bio->bi_iter.bi_size = bio_src->bi_iter.bi_size;
678 if (bio->bi_rw & REQ_DISCARD)
679 goto integrity_clone;
681 if (bio->bi_rw & REQ_WRITE_SAME) {
682 bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0];
683 goto integrity_clone;
686 bio_for_each_segment(bv, bio_src, iter)
687 bio->bi_io_vec[bio->bi_vcnt++] = bv;
690 if (bio_integrity(bio_src)) {
693 ret = bio_integrity_clone(bio, bio_src, gfp_mask);
700 bio_clone_blkcg_association(bio, bio_src);
704 EXPORT_SYMBOL(bio_clone_bioset);
707 * bio_add_pc_page - attempt to add page to bio
708 * @q: the target queue
709 * @bio: destination bio
711 * @len: vec entry length
712 * @offset: vec entry offset
714 * Attempt to add a page to the bio_vec maplist. This can fail for a
715 * number of reasons, such as the bio being full or target block device
716 * limitations. The target block device must allow bio's up to PAGE_SIZE,
717 * so it is always possible to add a single page to an empty bio.
719 * This should only be used by REQ_PC bios.
721 int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page
722 *page, unsigned int len, unsigned int offset)
724 int retried_segments = 0;
725 struct bio_vec *bvec;
728 * cloned bio must not modify vec list
730 if (unlikely(bio_flagged(bio, BIO_CLONED)))
733 if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q))
737 * For filesystems with a blocksize smaller than the pagesize
738 * we will often be called with the same page as last time and
739 * a consecutive offset. Optimize this special case.
741 if (bio->bi_vcnt > 0) {
742 struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1];
744 if (page == prev->bv_page &&
745 offset == prev->bv_offset + prev->bv_len) {
747 bio->bi_iter.bi_size += len;
752 * If the queue doesn't support SG gaps and adding this
753 * offset would create a gap, disallow it.
755 if (bvec_gap_to_prev(q, prev, offset))
759 if (bio->bi_vcnt >= bio->bi_max_vecs)
763 * setup the new entry, we might clear it again later if we
764 * cannot add the page
766 bvec = &bio->bi_io_vec[bio->bi_vcnt];
767 bvec->bv_page = page;
769 bvec->bv_offset = offset;
771 bio->bi_phys_segments++;
772 bio->bi_iter.bi_size += len;
775 * Perform a recount if the number of segments is greater
776 * than queue_max_segments(q).
779 while (bio->bi_phys_segments > queue_max_segments(q)) {
781 if (retried_segments)
784 retried_segments = 1;
785 blk_recount_segments(q, bio);
788 /* If we may be able to merge these biovecs, force a recount */
789 if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec)))
790 bio_clear_flag(bio, BIO_SEG_VALID);
796 bvec->bv_page = NULL;
800 bio->bi_iter.bi_size -= len;
801 blk_recount_segments(q, bio);
804 EXPORT_SYMBOL(bio_add_pc_page);
807 * bio_add_page - attempt to add page to bio
808 * @bio: destination bio
810 * @len: vec entry length
811 * @offset: vec entry offset
813 * Attempt to add a page to the bio_vec maplist. This will only fail
814 * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio.
816 int bio_add_page(struct bio *bio, struct page *page,
817 unsigned int len, unsigned int offset)
822 * cloned bio must not modify vec list
824 if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)))
828 * For filesystems with a blocksize smaller than the pagesize
829 * we will often be called with the same page as last time and
830 * a consecutive offset. Optimize this special case.
832 if (bio->bi_vcnt > 0) {
833 bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
835 if (page == bv->bv_page &&
836 offset == bv->bv_offset + bv->bv_len) {
842 if (bio->bi_vcnt >= bio->bi_max_vecs)
845 bv = &bio->bi_io_vec[bio->bi_vcnt];
848 bv->bv_offset = offset;
852 bio->bi_iter.bi_size += len;
855 EXPORT_SYMBOL(bio_add_page);
857 struct submit_bio_ret {
858 struct completion event;
862 static void submit_bio_wait_endio(struct bio *bio)
864 struct submit_bio_ret *ret = bio->bi_private;
866 ret->error = bio->bi_error;
867 complete(&ret->event);
871 * submit_bio_wait - submit a bio, and wait until it completes
872 * @rw: whether to %READ or %WRITE, or maybe to %READA (read ahead)
873 * @bio: The &struct bio which describes the I/O
875 * Simple wrapper around submit_bio(). Returns 0 on success, or the error from
876 * bio_endio() on failure.
878 int submit_bio_wait(int rw, struct bio *bio)
880 struct submit_bio_ret ret;
883 init_completion(&ret.event);
884 bio->bi_private = &ret;
885 bio->bi_end_io = submit_bio_wait_endio;
887 wait_for_completion(&ret.event);
891 EXPORT_SYMBOL(submit_bio_wait);
894 * bio_advance - increment/complete a bio by some number of bytes
895 * @bio: bio to advance
896 * @bytes: number of bytes to complete
898 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
899 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
900 * be updated on the last bvec as well.
902 * @bio will then represent the remaining, uncompleted portion of the io.
904 void bio_advance(struct bio *bio, unsigned bytes)
906 if (bio_integrity(bio))
907 bio_integrity_advance(bio, bytes);
909 bio_advance_iter(bio, &bio->bi_iter, bytes);
911 EXPORT_SYMBOL(bio_advance);
914 * bio_alloc_pages - allocates a single page for each bvec in a bio
915 * @bio: bio to allocate pages for
916 * @gfp_mask: flags for allocation
918 * Allocates pages up to @bio->bi_vcnt.
920 * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are
923 int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask)
928 bio_for_each_segment_all(bv, bio, i) {
929 bv->bv_page = alloc_page(gfp_mask);
931 while (--bv >= bio->bi_io_vec)
932 __free_page(bv->bv_page);
939 EXPORT_SYMBOL(bio_alloc_pages);
942 * bio_copy_data - copy contents of data buffers from one chain of bios to
944 * @src: source bio list
945 * @dst: destination bio list
947 * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats
948 * @src and @dst as linked lists of bios.
950 * Stops when it reaches the end of either @src or @dst - that is, copies
951 * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios).
953 void bio_copy_data(struct bio *dst, struct bio *src)
955 struct bvec_iter src_iter, dst_iter;
956 struct bio_vec src_bv, dst_bv;
960 src_iter = src->bi_iter;
961 dst_iter = dst->bi_iter;
964 if (!src_iter.bi_size) {
969 src_iter = src->bi_iter;
972 if (!dst_iter.bi_size) {
977 dst_iter = dst->bi_iter;
980 src_bv = bio_iter_iovec(src, src_iter);
981 dst_bv = bio_iter_iovec(dst, dst_iter);
983 bytes = min(src_bv.bv_len, dst_bv.bv_len);
985 src_p = kmap_atomic(src_bv.bv_page);
986 dst_p = kmap_atomic(dst_bv.bv_page);
988 memcpy(dst_p + dst_bv.bv_offset,
989 src_p + src_bv.bv_offset,
992 kunmap_atomic(dst_p);
993 kunmap_atomic(src_p);
995 bio_advance_iter(src, &src_iter, bytes);
996 bio_advance_iter(dst, &dst_iter, bytes);
999 EXPORT_SYMBOL(bio_copy_data);
1001 struct bio_map_data {
1003 struct iov_iter iter;
1007 static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count,
1010 if (iov_count > UIO_MAXIOV)
1013 return kmalloc(sizeof(struct bio_map_data) +
1014 sizeof(struct iovec) * iov_count, gfp_mask);
1018 * bio_copy_from_iter - copy all pages from iov_iter to bio
1019 * @bio: The &struct bio which describes the I/O as destination
1020 * @iter: iov_iter as source
1022 * Copy all pages from iov_iter to bio.
1023 * Returns 0 on success, or error on failure.
1025 static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter)
1028 struct bio_vec *bvec;
1030 bio_for_each_segment_all(bvec, bio, i) {
1033 ret = copy_page_from_iter(bvec->bv_page,
1038 if (!iov_iter_count(&iter))
1041 if (ret < bvec->bv_len)
1049 * bio_copy_to_iter - copy all pages from bio to iov_iter
1050 * @bio: The &struct bio which describes the I/O as source
1051 * @iter: iov_iter as destination
1053 * Copy all pages from bio to iov_iter.
1054 * Returns 0 on success, or error on failure.
1056 static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter)
1059 struct bio_vec *bvec;
1061 bio_for_each_segment_all(bvec, bio, i) {
1064 ret = copy_page_to_iter(bvec->bv_page,
1069 if (!iov_iter_count(&iter))
1072 if (ret < bvec->bv_len)
1079 static void bio_free_pages(struct bio *bio)
1081 struct bio_vec *bvec;
1084 bio_for_each_segment_all(bvec, bio, i)
1085 __free_page(bvec->bv_page);
1089 * bio_uncopy_user - finish previously mapped bio
1090 * @bio: bio being terminated
1092 * Free pages allocated from bio_copy_user_iov() and write back data
1093 * to user space in case of a read.
1095 int bio_uncopy_user(struct bio *bio)
1097 struct bio_map_data *bmd = bio->bi_private;
1100 if (!bio_flagged(bio, BIO_NULL_MAPPED)) {
1102 * if we're in a workqueue, the request is orphaned, so
1103 * don't copy into a random user address space, just free
1104 * and return -EINTR so user space doesn't expect any data.
1108 else if (bio_data_dir(bio) == READ)
1109 ret = bio_copy_to_iter(bio, bmd->iter);
1110 if (bmd->is_our_pages)
1111 bio_free_pages(bio);
1117 EXPORT_SYMBOL(bio_uncopy_user);
1120 * bio_copy_user_iov - copy user data to bio
1121 * @q: destination block queue
1122 * @map_data: pointer to the rq_map_data holding pages (if necessary)
1123 * @iter: iovec iterator
1124 * @gfp_mask: memory allocation flags
1126 * Prepares and returns a bio for indirect user io, bouncing data
1127 * to/from kernel pages as necessary. Must be paired with
1128 * call bio_uncopy_user() on io completion.
1130 struct bio *bio_copy_user_iov(struct request_queue *q,
1131 struct rq_map_data *map_data,
1132 const struct iov_iter *iter,
1135 struct bio_map_data *bmd;
1140 unsigned int len = iter->count;
1141 unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0;
1143 for (i = 0; i < iter->nr_segs; i++) {
1144 unsigned long uaddr;
1146 unsigned long start;
1148 uaddr = (unsigned long) iter->iov[i].iov_base;
1149 end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1)
1151 start = uaddr >> PAGE_SHIFT;
1157 return ERR_PTR(-EINVAL);
1159 nr_pages += end - start;
1165 bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask);
1167 return ERR_PTR(-ENOMEM);
1170 * We need to do a deep copy of the iov_iter including the iovecs.
1171 * The caller provided iov might point to an on-stack or otherwise
1174 bmd->is_our_pages = map_data ? 0 : 1;
1175 memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs);
1176 iov_iter_init(&bmd->iter, iter->type, bmd->iov,
1177 iter->nr_segs, iter->count);
1180 bio = bio_kmalloc(gfp_mask, nr_pages);
1184 if (iter->type & WRITE)
1185 bio->bi_rw |= REQ_WRITE;
1190 nr_pages = 1 << map_data->page_order;
1191 i = map_data->offset / PAGE_SIZE;
1194 unsigned int bytes = PAGE_SIZE;
1202 if (i == map_data->nr_entries * nr_pages) {
1207 page = map_data->pages[i / nr_pages];
1208 page += (i % nr_pages);
1212 page = alloc_page(q->bounce_gfp | gfp_mask);
1219 if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) {
1235 if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) ||
1236 (map_data && map_data->from_user)) {
1237 ret = bio_copy_from_iter(bio, *iter);
1242 bio->bi_private = bmd;
1246 bio_free_pages(bio);
1250 return ERR_PTR(ret);
1254 * bio_map_user_iov - map user iovec into bio
1255 * @q: the struct request_queue for the bio
1256 * @iter: iovec iterator
1257 * @gfp_mask: memory allocation flags
1259 * Map the user space address into a bio suitable for io to a block
1260 * device. Returns an error pointer in case of error.
1262 struct bio *bio_map_user_iov(struct request_queue *q,
1263 const struct iov_iter *iter,
1268 struct page **pages;
1274 struct bio_vec *bvec;
1276 iov_for_each(iov, i, *iter) {
1277 unsigned long uaddr = (unsigned long) iov.iov_base;
1278 unsigned long len = iov.iov_len;
1279 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1280 unsigned long start = uaddr >> PAGE_SHIFT;
1286 return ERR_PTR(-EINVAL);
1288 nr_pages += end - start;
1290 * buffer must be aligned to at least hardsector size for now
1292 if (uaddr & queue_dma_alignment(q))
1293 return ERR_PTR(-EINVAL);
1297 return ERR_PTR(-EINVAL);
1299 bio = bio_kmalloc(gfp_mask, nr_pages);
1301 return ERR_PTR(-ENOMEM);
1304 pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask);
1308 iov_for_each(iov, i, *iter) {
1309 unsigned long uaddr = (unsigned long) iov.iov_base;
1310 unsigned long len = iov.iov_len;
1311 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1312 unsigned long start = uaddr >> PAGE_SHIFT;
1313 const int local_nr_pages = end - start;
1314 const int page_limit = cur_page + local_nr_pages;
1316 ret = get_user_pages_fast(uaddr, local_nr_pages,
1317 (iter->type & WRITE) != WRITE,
1319 if (unlikely(ret < local_nr_pages)) {
1320 for (j = cur_page; j < page_limit; j++) {
1329 offset = uaddr & ~PAGE_MASK;
1330 for (j = cur_page; j < page_limit; j++) {
1331 unsigned int bytes = PAGE_SIZE - offset;
1332 unsigned short prev_bi_vcnt = bio->bi_vcnt;
1343 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) <
1348 * check if vector was merged with previous
1349 * drop page reference if needed
1351 if (bio->bi_vcnt == prev_bi_vcnt)
1360 * release the pages we didn't map into the bio, if any
1362 while (j < page_limit)
1363 page_cache_release(pages[j++]);
1369 * set data direction, and check if mapped pages need bouncing
1371 if (iter->type & WRITE)
1372 bio->bi_rw |= REQ_WRITE;
1374 bio_set_flag(bio, BIO_USER_MAPPED);
1377 * subtle -- if __bio_map_user() ended up bouncing a bio,
1378 * it would normally disappear when its bi_end_io is run.
1379 * however, we need it for the unmap, so grab an extra
1386 bio_for_each_segment_all(bvec, bio, j) {
1387 put_page(bvec->bv_page);
1392 return ERR_PTR(ret);
1395 static void __bio_unmap_user(struct bio *bio)
1397 struct bio_vec *bvec;
1401 * make sure we dirty pages we wrote to
1403 bio_for_each_segment_all(bvec, bio, i) {
1404 if (bio_data_dir(bio) == READ)
1405 set_page_dirty_lock(bvec->bv_page);
1407 page_cache_release(bvec->bv_page);
1414 * bio_unmap_user - unmap a bio
1415 * @bio: the bio being unmapped
1417 * Unmap a bio previously mapped by bio_map_user(). Must be called with
1418 * a process context.
1420 * bio_unmap_user() may sleep.
1422 void bio_unmap_user(struct bio *bio)
1424 __bio_unmap_user(bio);
1427 EXPORT_SYMBOL(bio_unmap_user);
1429 static void bio_map_kern_endio(struct bio *bio)
1435 * bio_map_kern - map kernel address into bio
1436 * @q: the struct request_queue for the bio
1437 * @data: pointer to buffer to map
1438 * @len: length in bytes
1439 * @gfp_mask: allocation flags for bio allocation
1441 * Map the kernel address into a bio suitable for io to a block
1442 * device. Returns an error pointer in case of error.
1444 struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len,
1447 unsigned long kaddr = (unsigned long)data;
1448 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1449 unsigned long start = kaddr >> PAGE_SHIFT;
1450 const int nr_pages = end - start;
1454 bio = bio_kmalloc(gfp_mask, nr_pages);
1456 return ERR_PTR(-ENOMEM);
1458 offset = offset_in_page(kaddr);
1459 for (i = 0; i < nr_pages; i++) {
1460 unsigned int bytes = PAGE_SIZE - offset;
1468 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes,
1470 /* we don't support partial mappings */
1472 return ERR_PTR(-EINVAL);
1480 bio->bi_end_io = bio_map_kern_endio;
1483 EXPORT_SYMBOL(bio_map_kern);
1485 static void bio_copy_kern_endio(struct bio *bio)
1487 bio_free_pages(bio);
1491 static void bio_copy_kern_endio_read(struct bio *bio)
1493 char *p = bio->bi_private;
1494 struct bio_vec *bvec;
1497 bio_for_each_segment_all(bvec, bio, i) {
1498 memcpy(p, page_address(bvec->bv_page), bvec->bv_len);
1502 bio_copy_kern_endio(bio);
1506 * bio_copy_kern - copy kernel address into bio
1507 * @q: the struct request_queue for the bio
1508 * @data: pointer to buffer to copy
1509 * @len: length in bytes
1510 * @gfp_mask: allocation flags for bio and page allocation
1511 * @reading: data direction is READ
1513 * copy the kernel address into a bio suitable for io to a block
1514 * device. Returns an error pointer in case of error.
1516 struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
1517 gfp_t gfp_mask, int reading)
1519 unsigned long kaddr = (unsigned long)data;
1520 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1521 unsigned long start = kaddr >> PAGE_SHIFT;
1530 return ERR_PTR(-EINVAL);
1532 nr_pages = end - start;
1533 bio = bio_kmalloc(gfp_mask, nr_pages);
1535 return ERR_PTR(-ENOMEM);
1539 unsigned int bytes = PAGE_SIZE;
1544 page = alloc_page(q->bounce_gfp | gfp_mask);
1549 memcpy(page_address(page), p, bytes);
1551 if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes)
1559 bio->bi_end_io = bio_copy_kern_endio_read;
1560 bio->bi_private = data;
1562 bio->bi_end_io = bio_copy_kern_endio;
1563 bio->bi_rw |= REQ_WRITE;
1569 bio_free_pages(bio);
1571 return ERR_PTR(-ENOMEM);
1573 EXPORT_SYMBOL(bio_copy_kern);
1576 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1577 * for performing direct-IO in BIOs.
1579 * The problem is that we cannot run set_page_dirty() from interrupt context
1580 * because the required locks are not interrupt-safe. So what we can do is to
1581 * mark the pages dirty _before_ performing IO. And in interrupt context,
1582 * check that the pages are still dirty. If so, fine. If not, redirty them
1583 * in process context.
1585 * We special-case compound pages here: normally this means reads into hugetlb
1586 * pages. The logic in here doesn't really work right for compound pages
1587 * because the VM does not uniformly chase down the head page in all cases.
1588 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1589 * handle them at all. So we skip compound pages here at an early stage.
1591 * Note that this code is very hard to test under normal circumstances because
1592 * direct-io pins the pages with get_user_pages(). This makes
1593 * is_page_cache_freeable return false, and the VM will not clean the pages.
1594 * But other code (eg, flusher threads) could clean the pages if they are mapped
1597 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1598 * deferred bio dirtying paths.
1602 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1604 void bio_set_pages_dirty(struct bio *bio)
1606 struct bio_vec *bvec;
1609 bio_for_each_segment_all(bvec, bio, i) {
1610 struct page *page = bvec->bv_page;
1612 if (page && !PageCompound(page))
1613 set_page_dirty_lock(page);
1617 static void bio_release_pages(struct bio *bio)
1619 struct bio_vec *bvec;
1622 bio_for_each_segment_all(bvec, bio, i) {
1623 struct page *page = bvec->bv_page;
1631 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1632 * If they are, then fine. If, however, some pages are clean then they must
1633 * have been written out during the direct-IO read. So we take another ref on
1634 * the BIO and the offending pages and re-dirty the pages in process context.
1636 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1637 * here on. It will run one page_cache_release() against each page and will
1638 * run one bio_put() against the BIO.
1641 static void bio_dirty_fn(struct work_struct *work);
1643 static DECLARE_WORK(bio_dirty_work, bio_dirty_fn);
1644 static DEFINE_SPINLOCK(bio_dirty_lock);
1645 static struct bio *bio_dirty_list;
1648 * This runs in process context
1650 static void bio_dirty_fn(struct work_struct *work)
1652 unsigned long flags;
1655 spin_lock_irqsave(&bio_dirty_lock, flags);
1656 bio = bio_dirty_list;
1657 bio_dirty_list = NULL;
1658 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1661 struct bio *next = bio->bi_private;
1663 bio_set_pages_dirty(bio);
1664 bio_release_pages(bio);
1670 void bio_check_pages_dirty(struct bio *bio)
1672 struct bio_vec *bvec;
1673 int nr_clean_pages = 0;
1676 bio_for_each_segment_all(bvec, bio, i) {
1677 struct page *page = bvec->bv_page;
1679 if (PageDirty(page) || PageCompound(page)) {
1680 page_cache_release(page);
1681 bvec->bv_page = NULL;
1687 if (nr_clean_pages) {
1688 unsigned long flags;
1690 spin_lock_irqsave(&bio_dirty_lock, flags);
1691 bio->bi_private = bio_dirty_list;
1692 bio_dirty_list = bio;
1693 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1694 schedule_work(&bio_dirty_work);
1700 void generic_start_io_acct(int rw, unsigned long sectors,
1701 struct hd_struct *part)
1703 int cpu = part_stat_lock();
1705 part_round_stats(cpu, part);
1706 part_stat_inc(cpu, part, ios[rw]);
1707 part_stat_add(cpu, part, sectors[rw], sectors);
1708 part_inc_in_flight(part, rw);
1712 EXPORT_SYMBOL(generic_start_io_acct);
1714 void generic_end_io_acct(int rw, struct hd_struct *part,
1715 unsigned long start_time)
1717 unsigned long duration = jiffies - start_time;
1718 int cpu = part_stat_lock();
1720 part_stat_add(cpu, part, ticks[rw], duration);
1721 part_round_stats(cpu, part);
1722 part_dec_in_flight(part, rw);
1726 EXPORT_SYMBOL(generic_end_io_acct);
1728 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1729 void bio_flush_dcache_pages(struct bio *bi)
1731 struct bio_vec bvec;
1732 struct bvec_iter iter;
1734 bio_for_each_segment(bvec, bi, iter)
1735 flush_dcache_page(bvec.bv_page);
1737 EXPORT_SYMBOL(bio_flush_dcache_pages);
1740 static inline bool bio_remaining_done(struct bio *bio)
1743 * If we're not chaining, then ->__bi_remaining is always 1 and
1744 * we always end io on the first invocation.
1746 if (!bio_flagged(bio, BIO_CHAIN))
1749 BUG_ON(atomic_read(&bio->__bi_remaining) <= 0);
1751 if (atomic_dec_and_test(&bio->__bi_remaining)) {
1752 bio_clear_flag(bio, BIO_CHAIN);
1760 * bio_endio - end I/O on a bio
1764 * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred
1765 * way to end I/O on a bio. No one should call bi_end_io() directly on a
1766 * bio unless they own it and thus know that it has an end_io function.
1768 void bio_endio(struct bio *bio)
1771 if (unlikely(!bio_remaining_done(bio)))
1775 * Need to have a real endio function for chained bios,
1776 * otherwise various corner cases will break (like stacking
1777 * block devices that save/restore bi_end_io) - however, we want
1778 * to avoid unbounded recursion and blowing the stack. Tail call
1779 * optimization would handle this, but compiling with frame
1780 * pointers also disables gcc's sibling call optimization.
1782 if (bio->bi_end_io == bio_chain_endio) {
1783 struct bio *parent = bio->bi_private;
1784 parent->bi_error = bio->bi_error;
1789 bio->bi_end_io(bio);
1794 EXPORT_SYMBOL(bio_endio);
1797 * bio_split - split a bio
1798 * @bio: bio to split
1799 * @sectors: number of sectors to split from the front of @bio
1801 * @bs: bio set to allocate from
1803 * Allocates and returns a new bio which represents @sectors from the start of
1804 * @bio, and updates @bio to represent the remaining sectors.
1806 * Unless this is a discard request the newly allocated bio will point
1807 * to @bio's bi_io_vec; it is the caller's responsibility to ensure that
1808 * @bio is not freed before the split.
1810 struct bio *bio_split(struct bio *bio, int sectors,
1811 gfp_t gfp, struct bio_set *bs)
1813 struct bio *split = NULL;
1815 BUG_ON(sectors <= 0);
1816 BUG_ON(sectors >= bio_sectors(bio));
1819 * Discards need a mutable bio_vec to accommodate the payload
1820 * required by the DSM TRIM and UNMAP commands.
1822 if (bio->bi_rw & REQ_DISCARD)
1823 split = bio_clone_bioset(bio, gfp, bs);
1825 split = bio_clone_fast(bio, gfp, bs);
1830 split->bi_iter.bi_size = sectors << 9;
1832 if (bio_integrity(split))
1833 bio_integrity_trim(split, 0, sectors);
1835 bio_advance(bio, split->bi_iter.bi_size);
1839 EXPORT_SYMBOL(bio_split);
1842 * bio_trim - trim a bio
1844 * @offset: number of sectors to trim from the front of @bio
1845 * @size: size we want to trim @bio to, in sectors
1847 void bio_trim(struct bio *bio, int offset, int size)
1849 /* 'bio' is a cloned bio which we need to trim to match
1850 * the given offset and size.
1854 if (offset == 0 && size == bio->bi_iter.bi_size)
1857 bio_clear_flag(bio, BIO_SEG_VALID);
1859 bio_advance(bio, offset << 9);
1861 bio->bi_iter.bi_size = size;
1863 EXPORT_SYMBOL_GPL(bio_trim);
1866 * create memory pools for biovec's in a bio_set.
1867 * use the global biovec slabs created for general use.
1869 mempool_t *biovec_create_pool(int pool_entries)
1871 struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX;
1873 return mempool_create_slab_pool(pool_entries, bp->slab);
1876 void bioset_free(struct bio_set *bs)
1878 if (bs->rescue_workqueue)
1879 destroy_workqueue(bs->rescue_workqueue);
1882 mempool_destroy(bs->bio_pool);
1885 mempool_destroy(bs->bvec_pool);
1887 bioset_integrity_free(bs);
1892 EXPORT_SYMBOL(bioset_free);
1894 static struct bio_set *__bioset_create(unsigned int pool_size,
1895 unsigned int front_pad,
1896 bool create_bvec_pool)
1898 unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec);
1901 bs = kzalloc(sizeof(*bs), GFP_KERNEL);
1905 bs->front_pad = front_pad;
1907 spin_lock_init(&bs->rescue_lock);
1908 bio_list_init(&bs->rescue_list);
1909 INIT_WORK(&bs->rescue_work, bio_alloc_rescue);
1911 bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad);
1912 if (!bs->bio_slab) {
1917 bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab);
1921 if (create_bvec_pool) {
1922 bs->bvec_pool = biovec_create_pool(pool_size);
1927 bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0);
1928 if (!bs->rescue_workqueue)
1938 * bioset_create - Create a bio_set
1939 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1940 * @front_pad: Number of bytes to allocate in front of the returned bio
1943 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
1944 * to ask for a number of bytes to be allocated in front of the bio.
1945 * Front pad allocation is useful for embedding the bio inside
1946 * another structure, to avoid allocating extra data to go with the bio.
1947 * Note that the bio must be embedded at the END of that structure always,
1948 * or things will break badly.
1950 struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad)
1952 return __bioset_create(pool_size, front_pad, true);
1954 EXPORT_SYMBOL(bioset_create);
1957 * bioset_create_nobvec - Create a bio_set without bio_vec mempool
1958 * @pool_size: Number of bio to cache in the mempool
1959 * @front_pad: Number of bytes to allocate in front of the returned bio
1962 * Same functionality as bioset_create() except that mempool is not
1963 * created for bio_vecs. Saving some memory for bio_clone_fast() users.
1965 struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad)
1967 return __bioset_create(pool_size, front_pad, false);
1969 EXPORT_SYMBOL(bioset_create_nobvec);
1971 #ifdef CONFIG_BLK_CGROUP
1974 * bio_associate_blkcg - associate a bio with the specified blkcg
1976 * @blkcg_css: css of the blkcg to associate
1978 * Associate @bio with the blkcg specified by @blkcg_css. Block layer will
1979 * treat @bio as if it were issued by a task which belongs to the blkcg.
1981 * This function takes an extra reference of @blkcg_css which will be put
1982 * when @bio is released. The caller must own @bio and is responsible for
1983 * synchronizing calls to this function.
1985 int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css)
1987 if (unlikely(bio->bi_css))
1990 bio->bi_css = blkcg_css;
1993 EXPORT_SYMBOL_GPL(bio_associate_blkcg);
1996 * bio_associate_current - associate a bio with %current
1999 * Associate @bio with %current if it hasn't been associated yet. Block
2000 * layer will treat @bio as if it were issued by %current no matter which
2001 * task actually issues it.
2003 * This function takes an extra reference of @task's io_context and blkcg
2004 * which will be put when @bio is released. The caller must own @bio,
2005 * ensure %current->io_context exists, and is responsible for synchronizing
2006 * calls to this function.
2008 int bio_associate_current(struct bio *bio)
2010 struct io_context *ioc;
2015 ioc = current->io_context;
2019 get_io_context_active(ioc);
2021 bio->bi_css = task_get_css(current, io_cgrp_id);
2024 EXPORT_SYMBOL_GPL(bio_associate_current);
2027 * bio_disassociate_task - undo bio_associate_current()
2030 void bio_disassociate_task(struct bio *bio)
2033 put_io_context(bio->bi_ioc);
2037 css_put(bio->bi_css);
2043 * bio_clone_blkcg_association - clone blkcg association from src to dst bio
2044 * @dst: destination bio
2047 void bio_clone_blkcg_association(struct bio *dst, struct bio *src)
2050 WARN_ON(bio_associate_blkcg(dst, src->bi_css));
2053 #endif /* CONFIG_BLK_CGROUP */
2055 static void __init biovec_init_slabs(void)
2059 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
2061 struct biovec_slab *bvs = bvec_slabs + i;
2063 if (bvs->nr_vecs <= BIO_INLINE_VECS) {
2068 size = bvs->nr_vecs * sizeof(struct bio_vec);
2069 bvs->slab = kmem_cache_create(bvs->name, size, 0,
2070 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
2074 static int __init init_bio(void)
2078 bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL);
2080 panic("bio: can't allocate bios\n");
2082 bio_integrity_init();
2083 biovec_init_slabs();
2085 fs_bio_set = bioset_create(BIO_POOL_SIZE, 0);
2087 panic("bio: can't allocate bios\n");
2089 if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE))
2090 panic("bio: can't create integrity pool\n");
2094 subsys_initcall(init_bio);