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>
37 * Test patch to inline a certain number of bi_io_vec's inside the bio
38 * itself, to shrink a bio data allocation from two mempool calls to one
40 #define BIO_INLINE_VECS 4
43 * if you change this list, also change bvec_alloc or things will
44 * break badly! cannot be bigger than what you can fit into an
47 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
48 static struct biovec_slab bvec_slabs[BIOVEC_NR_POOLS] __read_mostly = {
49 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES),
54 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
55 * IO code that does not need private memory pools.
57 struct bio_set *fs_bio_set;
58 EXPORT_SYMBOL(fs_bio_set);
61 * Our slab pool management
64 struct kmem_cache *slab;
65 unsigned int slab_ref;
66 unsigned int slab_size;
69 static DEFINE_MUTEX(bio_slab_lock);
70 static struct bio_slab *bio_slabs;
71 static unsigned int bio_slab_nr, bio_slab_max;
73 static struct kmem_cache *bio_find_or_create_slab(unsigned int extra_size)
75 unsigned int sz = sizeof(struct bio) + extra_size;
76 struct kmem_cache *slab = NULL;
77 struct bio_slab *bslab, *new_bio_slabs;
78 unsigned int new_bio_slab_max;
79 unsigned int i, entry = -1;
81 mutex_lock(&bio_slab_lock);
84 while (i < bio_slab_nr) {
85 bslab = &bio_slabs[i];
87 if (!bslab->slab && entry == -1)
89 else if (bslab->slab_size == sz) {
100 if (bio_slab_nr == bio_slab_max && entry == -1) {
101 new_bio_slab_max = bio_slab_max << 1;
102 new_bio_slabs = krealloc(bio_slabs,
103 new_bio_slab_max * sizeof(struct bio_slab),
107 bio_slab_max = new_bio_slab_max;
108 bio_slabs = new_bio_slabs;
111 entry = bio_slab_nr++;
113 bslab = &bio_slabs[entry];
115 snprintf(bslab->name, sizeof(bslab->name), "bio-%d", entry);
116 slab = kmem_cache_create(bslab->name, sz, ARCH_KMALLOC_MINALIGN,
117 SLAB_HWCACHE_ALIGN, NULL);
123 bslab->slab_size = sz;
125 mutex_unlock(&bio_slab_lock);
129 static void bio_put_slab(struct bio_set *bs)
131 struct bio_slab *bslab = NULL;
134 mutex_lock(&bio_slab_lock);
136 for (i = 0; i < bio_slab_nr; i++) {
137 if (bs->bio_slab == bio_slabs[i].slab) {
138 bslab = &bio_slabs[i];
143 if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n"))
146 WARN_ON(!bslab->slab_ref);
148 if (--bslab->slab_ref)
151 kmem_cache_destroy(bslab->slab);
155 mutex_unlock(&bio_slab_lock);
158 unsigned int bvec_nr_vecs(unsigned short idx)
160 return bvec_slabs[idx].nr_vecs;
163 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx)
165 BIO_BUG_ON(idx >= BIOVEC_NR_POOLS);
167 if (idx == BIOVEC_MAX_IDX)
168 mempool_free(bv, pool);
170 struct biovec_slab *bvs = bvec_slabs + idx;
172 kmem_cache_free(bvs->slab, bv);
176 struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx,
182 * see comment near bvec_array define!
200 case 129 ... BIO_MAX_PAGES:
208 * idx now points to the pool we want to allocate from. only the
209 * 1-vec entry pool is mempool backed.
211 if (*idx == BIOVEC_MAX_IDX) {
213 bvl = mempool_alloc(pool, gfp_mask);
215 struct biovec_slab *bvs = bvec_slabs + *idx;
216 gfp_t __gfp_mask = gfp_mask & ~(__GFP_DIRECT_RECLAIM | __GFP_IO);
219 * Make this allocation restricted and don't dump info on
220 * allocation failures, since we'll fallback to the mempool
221 * in case of failure.
223 __gfp_mask |= __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN;
226 * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM
227 * is set, retry with the 1-entry mempool
229 bvl = kmem_cache_alloc(bvs->slab, __gfp_mask);
230 if (unlikely(!bvl && (gfp_mask & __GFP_DIRECT_RECLAIM))) {
231 *idx = BIOVEC_MAX_IDX;
239 static void __bio_free(struct bio *bio)
241 bio_disassociate_task(bio);
243 if (bio_integrity(bio))
244 bio_integrity_free(bio);
247 static void bio_free(struct bio *bio)
249 struct bio_set *bs = bio->bi_pool;
255 if (bio_flagged(bio, BIO_OWNS_VEC))
256 bvec_free(bs->bvec_pool, bio->bi_io_vec, BIO_POOL_IDX(bio));
259 * If we have front padding, adjust the bio pointer before freeing
264 mempool_free(p, bs->bio_pool);
266 /* Bio was allocated by bio_kmalloc() */
271 void bio_init(struct bio *bio)
273 memset(bio, 0, sizeof(*bio));
274 atomic_set(&bio->__bi_remaining, 1);
275 atomic_set(&bio->__bi_cnt, 1);
277 EXPORT_SYMBOL(bio_init);
280 * bio_reset - reinitialize a bio
284 * After calling bio_reset(), @bio will be in the same state as a freshly
285 * allocated bio returned bio bio_alloc_bioset() - the only fields that are
286 * preserved are the ones that are initialized by bio_alloc_bioset(). See
287 * comment in struct bio.
289 void bio_reset(struct bio *bio)
291 unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS);
295 memset(bio, 0, BIO_RESET_BYTES);
296 bio->bi_flags = flags;
297 atomic_set(&bio->__bi_remaining, 1);
299 EXPORT_SYMBOL(bio_reset);
301 static void bio_chain_endio(struct bio *bio)
303 struct bio *parent = bio->bi_private;
305 parent->bi_error = bio->bi_error;
311 * Increment chain count for the bio. Make sure the CHAIN flag update
312 * is visible before the raised count.
314 static inline void bio_inc_remaining(struct bio *bio)
316 bio_set_flag(bio, BIO_CHAIN);
317 smp_mb__before_atomic();
318 atomic_inc(&bio->__bi_remaining);
322 * bio_chain - chain bio completions
323 * @bio: the target bio
324 * @parent: the @bio's parent bio
326 * The caller won't have a bi_end_io called when @bio completes - instead,
327 * @parent's bi_end_io won't be called until both @parent and @bio have
328 * completed; the chained bio will also be freed when it completes.
330 * The caller must not set bi_private or bi_end_io in @bio.
332 void bio_chain(struct bio *bio, struct bio *parent)
334 BUG_ON(bio->bi_private || bio->bi_end_io);
336 bio->bi_private = parent;
337 bio->bi_end_io = bio_chain_endio;
338 bio_inc_remaining(parent);
340 EXPORT_SYMBOL(bio_chain);
342 static void bio_alloc_rescue(struct work_struct *work)
344 struct bio_set *bs = container_of(work, struct bio_set, rescue_work);
348 spin_lock(&bs->rescue_lock);
349 bio = bio_list_pop(&bs->rescue_list);
350 spin_unlock(&bs->rescue_lock);
355 generic_make_request(bio);
359 static void punt_bios_to_rescuer(struct bio_set *bs)
361 struct bio_list punt, nopunt;
365 * In order to guarantee forward progress we must punt only bios that
366 * were allocated from this bio_set; otherwise, if there was a bio on
367 * there for a stacking driver higher up in the stack, processing it
368 * could require allocating bios from this bio_set, and doing that from
369 * our own rescuer would be bad.
371 * Since bio lists are singly linked, pop them all instead of trying to
372 * remove from the middle of the list:
375 bio_list_init(&punt);
376 bio_list_init(&nopunt);
378 while ((bio = bio_list_pop(¤t->bio_list[0])))
379 bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio);
380 current->bio_list[0] = nopunt;
382 bio_list_init(&nopunt);
383 while ((bio = bio_list_pop(¤t->bio_list[1])))
384 bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio);
385 current->bio_list[1] = nopunt;
387 spin_lock(&bs->rescue_lock);
388 bio_list_merge(&bs->rescue_list, &punt);
389 spin_unlock(&bs->rescue_lock);
391 queue_work(bs->rescue_workqueue, &bs->rescue_work);
395 * bio_alloc_bioset - allocate a bio for I/O
396 * @gfp_mask: the GFP_ mask given to the slab allocator
397 * @nr_iovecs: number of iovecs to pre-allocate
398 * @bs: the bio_set to allocate from.
401 * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is
402 * backed by the @bs's mempool.
404 * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will
405 * always be able to allocate a bio. This is due to the mempool guarantees.
406 * To make this work, callers must never allocate more than 1 bio at a time
407 * from this pool. Callers that need to allocate more than 1 bio must always
408 * submit the previously allocated bio for IO before attempting to allocate
409 * a new one. Failure to do so can cause deadlocks under memory pressure.
411 * Note that when running under generic_make_request() (i.e. any block
412 * driver), bios are not submitted until after you return - see the code in
413 * generic_make_request() that converts recursion into iteration, to prevent
416 * This would normally mean allocating multiple bios under
417 * generic_make_request() would be susceptible to deadlocks, but we have
418 * deadlock avoidance code that resubmits any blocked bios from a rescuer
421 * However, we do not guarantee forward progress for allocations from other
422 * mempools. Doing multiple allocations from the same mempool under
423 * generic_make_request() should be avoided - instead, use bio_set's front_pad
424 * for per bio allocations.
427 * Pointer to new bio on success, NULL on failure.
429 struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
431 gfp_t saved_gfp = gfp_mask;
433 unsigned inline_vecs;
434 unsigned long idx = BIO_POOL_NONE;
435 struct bio_vec *bvl = NULL;
440 if (nr_iovecs > UIO_MAXIOV)
443 p = kmalloc(sizeof(struct bio) +
444 nr_iovecs * sizeof(struct bio_vec),
447 inline_vecs = nr_iovecs;
449 /* should not use nobvec bioset for nr_iovecs > 0 */
450 if (WARN_ON_ONCE(!bs->bvec_pool && nr_iovecs > 0))
453 * generic_make_request() converts recursion to iteration; this
454 * means if we're running beneath it, any bios we allocate and
455 * submit will not be submitted (and thus freed) until after we
458 * This exposes us to a potential deadlock if we allocate
459 * multiple bios from the same bio_set() while running
460 * underneath generic_make_request(). If we were to allocate
461 * multiple bios (say a stacking block driver that was splitting
462 * bios), we would deadlock if we exhausted the mempool's
465 * We solve this, and guarantee forward progress, with a rescuer
466 * workqueue per bio_set. If we go to allocate and there are
467 * bios on current->bio_list, we first try the allocation
468 * without __GFP_DIRECT_RECLAIM; if that fails, we punt those
469 * bios we would be blocking to the rescuer workqueue before
470 * we retry with the original gfp_flags.
473 if (current->bio_list &&
474 (!bio_list_empty(¤t->bio_list[0]) ||
475 !bio_list_empty(¤t->bio_list[1])))
476 gfp_mask &= ~__GFP_DIRECT_RECLAIM;
478 p = mempool_alloc(bs->bio_pool, gfp_mask);
479 if (!p && gfp_mask != saved_gfp) {
480 punt_bios_to_rescuer(bs);
481 gfp_mask = saved_gfp;
482 p = mempool_alloc(bs->bio_pool, gfp_mask);
485 front_pad = bs->front_pad;
486 inline_vecs = BIO_INLINE_VECS;
495 if (nr_iovecs > inline_vecs) {
496 bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool);
497 if (!bvl && gfp_mask != saved_gfp) {
498 punt_bios_to_rescuer(bs);
499 gfp_mask = saved_gfp;
500 bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, bs->bvec_pool);
506 bio_set_flag(bio, BIO_OWNS_VEC);
507 } else if (nr_iovecs) {
508 bvl = bio->bi_inline_vecs;
512 bio->bi_flags |= idx << BIO_POOL_OFFSET;
513 bio->bi_max_vecs = nr_iovecs;
514 bio->bi_io_vec = bvl;
518 mempool_free(p, bs->bio_pool);
521 EXPORT_SYMBOL(bio_alloc_bioset);
523 void zero_fill_bio(struct bio *bio)
527 struct bvec_iter iter;
529 bio_for_each_segment(bv, bio, iter) {
530 char *data = bvec_kmap_irq(&bv, &flags);
531 memset(data, 0, bv.bv_len);
532 flush_dcache_page(bv.bv_page);
533 bvec_kunmap_irq(data, &flags);
536 EXPORT_SYMBOL(zero_fill_bio);
539 * bio_put - release a reference to a bio
540 * @bio: bio to release reference to
543 * Put a reference to a &struct bio, either one you have gotten with
544 * bio_alloc, bio_get or bio_clone. The last put of a bio will free it.
546 void bio_put(struct bio *bio)
548 if (!bio_flagged(bio, BIO_REFFED))
551 BIO_BUG_ON(!atomic_read(&bio->__bi_cnt));
556 if (atomic_dec_and_test(&bio->__bi_cnt))
560 EXPORT_SYMBOL(bio_put);
562 inline int bio_phys_segments(struct request_queue *q, struct bio *bio)
564 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
565 blk_recount_segments(q, bio);
567 return bio->bi_phys_segments;
569 EXPORT_SYMBOL(bio_phys_segments);
572 * __bio_clone_fast - clone a bio that shares the original bio's biovec
573 * @bio: destination bio
574 * @bio_src: bio to clone
576 * Clone a &bio. Caller will own the returned bio, but not
577 * the actual data it points to. Reference count of returned
580 * Caller must ensure that @bio_src is not freed before @bio.
582 void __bio_clone_fast(struct bio *bio, struct bio *bio_src)
584 BUG_ON(bio->bi_pool && BIO_POOL_IDX(bio) != BIO_POOL_NONE);
587 * most users will be overriding ->bi_bdev with a new target,
588 * so we don't set nor calculate new physical/hw segment counts here
590 bio->bi_bdev = bio_src->bi_bdev;
591 bio_set_flag(bio, BIO_CLONED);
592 bio->bi_rw = bio_src->bi_rw;
593 bio->bi_iter = bio_src->bi_iter;
594 bio->bi_io_vec = bio_src->bi_io_vec;
595 bio->bi_dio_inode = bio_src->bi_dio_inode;
597 bio_clone_blkcg_association(bio, bio_src);
599 EXPORT_SYMBOL(__bio_clone_fast);
602 * bio_clone_fast - clone a bio that shares the original bio's biovec
604 * @gfp_mask: allocation priority
605 * @bs: bio_set to allocate from
607 * Like __bio_clone_fast, only also allocates the returned bio
609 struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs)
613 b = bio_alloc_bioset(gfp_mask, 0, bs);
617 __bio_clone_fast(b, bio);
619 if (bio_integrity(bio)) {
622 ret = bio_integrity_clone(b, bio, gfp_mask);
632 EXPORT_SYMBOL(bio_clone_fast);
635 * bio_clone_bioset - clone a bio
636 * @bio_src: bio to clone
637 * @gfp_mask: allocation priority
638 * @bs: bio_set to allocate from
640 * Clone bio. Caller will own the returned bio, but not the actual data it
641 * points to. Reference count of returned bio will be one.
643 struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask,
646 struct bvec_iter iter;
651 * Pre immutable biovecs, __bio_clone() used to just do a memcpy from
652 * bio_src->bi_io_vec to bio->bi_io_vec.
654 * We can't do that anymore, because:
656 * - The point of cloning the biovec is to produce a bio with a biovec
657 * the caller can modify: bi_idx and bi_bvec_done should be 0.
659 * - The original bio could've had more than BIO_MAX_PAGES biovecs; if
660 * we tried to clone the whole thing bio_alloc_bioset() would fail.
661 * But the clone should succeed as long as the number of biovecs we
662 * actually need to allocate is fewer than BIO_MAX_PAGES.
664 * - Lastly, bi_vcnt should not be looked at or relied upon by code
665 * that does not own the bio - reason being drivers don't use it for
666 * iterating over the biovec anymore, so expecting it to be kept up
667 * to date (i.e. for clones that share the parent biovec) is just
668 * asking for trouble and would force extra work on
669 * __bio_clone_fast() anyways.
672 bio = bio_alloc_bioset(gfp_mask, bio_segments(bio_src), bs);
676 bio->bi_bdev = bio_src->bi_bdev;
677 bio->bi_rw = bio_src->bi_rw;
678 bio->bi_iter.bi_sector = bio_src->bi_iter.bi_sector;
679 bio->bi_iter.bi_size = bio_src->bi_iter.bi_size;
681 if (bio->bi_rw & REQ_DISCARD)
682 goto integrity_clone;
684 if (bio->bi_rw & REQ_WRITE_SAME) {
685 bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0];
686 goto integrity_clone;
689 bio_for_each_segment(bv, bio_src, iter)
690 bio->bi_io_vec[bio->bi_vcnt++] = bv;
693 if (bio_integrity(bio_src)) {
696 ret = bio_integrity_clone(bio, bio_src, gfp_mask);
703 bio_clone_blkcg_association(bio, bio_src);
707 EXPORT_SYMBOL(bio_clone_bioset);
710 * bio_add_pc_page - attempt to add page to bio
711 * @q: the target queue
712 * @bio: destination bio
714 * @len: vec entry length
715 * @offset: vec entry offset
717 * Attempt to add a page to the bio_vec maplist. This can fail for a
718 * number of reasons, such as the bio being full or target block device
719 * limitations. The target block device must allow bio's up to PAGE_SIZE,
720 * so it is always possible to add a single page to an empty bio.
722 * This should only be used by REQ_PC bios.
724 int bio_add_pc_page(struct request_queue *q, struct bio *bio, struct page
725 *page, unsigned int len, unsigned int offset)
727 int retried_segments = 0;
728 struct bio_vec *bvec;
731 * cloned bio must not modify vec list
733 if (unlikely(bio_flagged(bio, BIO_CLONED)))
736 if (((bio->bi_iter.bi_size + len) >> 9) > queue_max_hw_sectors(q))
740 * For filesystems with a blocksize smaller than the pagesize
741 * we will often be called with the same page as last time and
742 * a consecutive offset. Optimize this special case.
744 if (bio->bi_vcnt > 0) {
745 struct bio_vec *prev = &bio->bi_io_vec[bio->bi_vcnt - 1];
747 if (page == prev->bv_page &&
748 offset == prev->bv_offset + prev->bv_len) {
750 bio->bi_iter.bi_size += len;
755 * If the queue doesn't support SG gaps and adding this
756 * offset would create a gap, disallow it.
758 if (bvec_gap_to_prev(q, prev, offset))
762 if (bio->bi_vcnt >= bio->bi_max_vecs)
766 * setup the new entry, we might clear it again later if we
767 * cannot add the page
769 bvec = &bio->bi_io_vec[bio->bi_vcnt];
770 bvec->bv_page = page;
772 bvec->bv_offset = offset;
774 bio->bi_phys_segments++;
775 bio->bi_iter.bi_size += len;
778 * Perform a recount if the number of segments is greater
779 * than queue_max_segments(q).
782 while (bio->bi_phys_segments > queue_max_segments(q)) {
784 if (retried_segments)
787 retried_segments = 1;
788 blk_recount_segments(q, bio);
791 /* If we may be able to merge these biovecs, force a recount */
792 if (bio->bi_vcnt > 1 && (BIOVEC_PHYS_MERGEABLE(bvec-1, bvec)))
793 bio_clear_flag(bio, BIO_SEG_VALID);
799 bvec->bv_page = NULL;
803 bio->bi_iter.bi_size -= len;
804 blk_recount_segments(q, bio);
807 EXPORT_SYMBOL(bio_add_pc_page);
810 * bio_add_page - attempt to add page to bio
811 * @bio: destination bio
813 * @len: vec entry length
814 * @offset: vec entry offset
816 * Attempt to add a page to the bio_vec maplist. This will only fail
817 * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio.
819 int bio_add_page(struct bio *bio, struct page *page,
820 unsigned int len, unsigned int offset)
825 * cloned bio must not modify vec list
827 if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)))
831 * For filesystems with a blocksize smaller than the pagesize
832 * we will often be called with the same page as last time and
833 * a consecutive offset. Optimize this special case.
835 if (bio->bi_vcnt > 0) {
836 bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
838 if (page == bv->bv_page &&
839 offset == bv->bv_offset + bv->bv_len) {
845 if (bio->bi_vcnt >= bio->bi_max_vecs)
848 bv = &bio->bi_io_vec[bio->bi_vcnt];
851 bv->bv_offset = offset;
855 bio->bi_iter.bi_size += len;
858 EXPORT_SYMBOL(bio_add_page);
860 struct submit_bio_ret {
861 struct completion event;
865 static void submit_bio_wait_endio(struct bio *bio)
867 struct submit_bio_ret *ret = bio->bi_private;
869 ret->error = bio->bi_error;
870 complete(&ret->event);
874 * submit_bio_wait - submit a bio, and wait until it completes
875 * @rw: whether to %READ or %WRITE, or maybe to %READA (read ahead)
876 * @bio: The &struct bio which describes the I/O
878 * Simple wrapper around submit_bio(). Returns 0 on success, or the error from
879 * bio_endio() on failure.
881 int submit_bio_wait(int rw, struct bio *bio)
883 struct submit_bio_ret ret;
886 init_completion(&ret.event);
887 bio->bi_private = &ret;
888 bio->bi_end_io = submit_bio_wait_endio;
890 wait_for_completion(&ret.event);
894 EXPORT_SYMBOL(submit_bio_wait);
897 * bio_advance - increment/complete a bio by some number of bytes
898 * @bio: bio to advance
899 * @bytes: number of bytes to complete
901 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
902 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
903 * be updated on the last bvec as well.
905 * @bio will then represent the remaining, uncompleted portion of the io.
907 void bio_advance(struct bio *bio, unsigned bytes)
909 if (bio_integrity(bio))
910 bio_integrity_advance(bio, bytes);
912 bio_advance_iter(bio, &bio->bi_iter, bytes);
914 EXPORT_SYMBOL(bio_advance);
917 * bio_alloc_pages - allocates a single page for each bvec in a bio
918 * @bio: bio to allocate pages for
919 * @gfp_mask: flags for allocation
921 * Allocates pages up to @bio->bi_vcnt.
923 * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are
926 int bio_alloc_pages(struct bio *bio, gfp_t gfp_mask)
931 bio_for_each_segment_all(bv, bio, i) {
932 bv->bv_page = alloc_page(gfp_mask);
934 while (--bv >= bio->bi_io_vec)
935 __free_page(bv->bv_page);
942 EXPORT_SYMBOL(bio_alloc_pages);
945 * bio_copy_data - copy contents of data buffers from one chain of bios to
947 * @src: source bio list
948 * @dst: destination bio list
950 * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats
951 * @src and @dst as linked lists of bios.
953 * Stops when it reaches the end of either @src or @dst - that is, copies
954 * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios).
956 void bio_copy_data(struct bio *dst, struct bio *src)
958 struct bvec_iter src_iter, dst_iter;
959 struct bio_vec src_bv, dst_bv;
963 src_iter = src->bi_iter;
964 dst_iter = dst->bi_iter;
967 if (!src_iter.bi_size) {
972 src_iter = src->bi_iter;
975 if (!dst_iter.bi_size) {
980 dst_iter = dst->bi_iter;
983 src_bv = bio_iter_iovec(src, src_iter);
984 dst_bv = bio_iter_iovec(dst, dst_iter);
986 bytes = min(src_bv.bv_len, dst_bv.bv_len);
988 src_p = kmap_atomic(src_bv.bv_page);
989 dst_p = kmap_atomic(dst_bv.bv_page);
991 memcpy(dst_p + dst_bv.bv_offset,
992 src_p + src_bv.bv_offset,
995 kunmap_atomic(dst_p);
996 kunmap_atomic(src_p);
998 bio_advance_iter(src, &src_iter, bytes);
999 bio_advance_iter(dst, &dst_iter, bytes);
1002 EXPORT_SYMBOL(bio_copy_data);
1004 struct bio_map_data {
1006 struct iov_iter iter;
1010 static struct bio_map_data *bio_alloc_map_data(unsigned int iov_count,
1013 if (iov_count > UIO_MAXIOV)
1016 return kmalloc(sizeof(struct bio_map_data) +
1017 sizeof(struct iovec) * iov_count, gfp_mask);
1021 * bio_copy_from_iter - copy all pages from iov_iter to bio
1022 * @bio: The &struct bio which describes the I/O as destination
1023 * @iter: iov_iter as source
1025 * Copy all pages from iov_iter to bio.
1026 * Returns 0 on success, or error on failure.
1028 static int bio_copy_from_iter(struct bio *bio, struct iov_iter iter)
1031 struct bio_vec *bvec;
1033 bio_for_each_segment_all(bvec, bio, i) {
1036 ret = copy_page_from_iter(bvec->bv_page,
1041 if (!iov_iter_count(&iter))
1044 if (ret < bvec->bv_len)
1052 * bio_copy_to_iter - copy all pages from bio to iov_iter
1053 * @bio: The &struct bio which describes the I/O as source
1054 * @iter: iov_iter as destination
1056 * Copy all pages from bio to iov_iter.
1057 * Returns 0 on success, or error on failure.
1059 static int bio_copy_to_iter(struct bio *bio, struct iov_iter iter)
1062 struct bio_vec *bvec;
1064 bio_for_each_segment_all(bvec, bio, i) {
1067 ret = copy_page_to_iter(bvec->bv_page,
1072 if (!iov_iter_count(&iter))
1075 if (ret < bvec->bv_len)
1082 static void bio_free_pages(struct bio *bio)
1084 struct bio_vec *bvec;
1087 bio_for_each_segment_all(bvec, bio, i)
1088 __free_page(bvec->bv_page);
1092 * bio_uncopy_user - finish previously mapped bio
1093 * @bio: bio being terminated
1095 * Free pages allocated from bio_copy_user_iov() and write back data
1096 * to user space in case of a read.
1098 int bio_uncopy_user(struct bio *bio)
1100 struct bio_map_data *bmd = bio->bi_private;
1103 if (!bio_flagged(bio, BIO_NULL_MAPPED)) {
1105 * if we're in a workqueue, the request is orphaned, so
1106 * don't copy into a random user address space, just free
1107 * and return -EINTR so user space doesn't expect any data.
1111 else if (bio_data_dir(bio) == READ)
1112 ret = bio_copy_to_iter(bio, bmd->iter);
1113 if (bmd->is_our_pages)
1114 bio_free_pages(bio);
1120 EXPORT_SYMBOL(bio_uncopy_user);
1123 * bio_copy_user_iov - copy user data to bio
1124 * @q: destination block queue
1125 * @map_data: pointer to the rq_map_data holding pages (if necessary)
1126 * @iter: iovec iterator
1127 * @gfp_mask: memory allocation flags
1129 * Prepares and returns a bio for indirect user io, bouncing data
1130 * to/from kernel pages as necessary. Must be paired with
1131 * call bio_uncopy_user() on io completion.
1133 struct bio *bio_copy_user_iov(struct request_queue *q,
1134 struct rq_map_data *map_data,
1135 const struct iov_iter *iter,
1138 struct bio_map_data *bmd;
1143 unsigned int len = iter->count;
1144 unsigned int offset = map_data ? map_data->offset & ~PAGE_MASK : 0;
1146 for (i = 0; i < iter->nr_segs; i++) {
1147 unsigned long uaddr;
1149 unsigned long start;
1151 uaddr = (unsigned long) iter->iov[i].iov_base;
1152 end = (uaddr + iter->iov[i].iov_len + PAGE_SIZE - 1)
1154 start = uaddr >> PAGE_SHIFT;
1160 return ERR_PTR(-EINVAL);
1162 nr_pages += end - start;
1168 bmd = bio_alloc_map_data(iter->nr_segs, gfp_mask);
1170 return ERR_PTR(-ENOMEM);
1173 * We need to do a deep copy of the iov_iter including the iovecs.
1174 * The caller provided iov might point to an on-stack or otherwise
1177 bmd->is_our_pages = map_data ? 0 : 1;
1178 memcpy(bmd->iov, iter->iov, sizeof(struct iovec) * iter->nr_segs);
1179 iov_iter_init(&bmd->iter, iter->type, bmd->iov,
1180 iter->nr_segs, iter->count);
1183 bio = bio_kmalloc(gfp_mask, nr_pages);
1187 if (iter->type & WRITE)
1188 bio->bi_rw |= REQ_WRITE;
1193 nr_pages = 1 << map_data->page_order;
1194 i = map_data->offset / PAGE_SIZE;
1197 unsigned int bytes = PAGE_SIZE;
1205 if (i == map_data->nr_entries * nr_pages) {
1210 page = map_data->pages[i / nr_pages];
1211 page += (i % nr_pages);
1215 page = alloc_page(q->bounce_gfp | gfp_mask);
1222 if (bio_add_pc_page(q, bio, page, bytes, offset) < bytes) {
1238 if (((iter->type & WRITE) && (!map_data || !map_data->null_mapped)) ||
1239 (map_data && map_data->from_user)) {
1240 ret = bio_copy_from_iter(bio, *iter);
1245 bio->bi_private = bmd;
1249 bio_free_pages(bio);
1253 return ERR_PTR(ret);
1257 * bio_map_user_iov - map user iovec into bio
1258 * @q: the struct request_queue for the bio
1259 * @iter: iovec iterator
1260 * @gfp_mask: memory allocation flags
1262 * Map the user space address into a bio suitable for io to a block
1263 * device. Returns an error pointer in case of error.
1265 struct bio *bio_map_user_iov(struct request_queue *q,
1266 const struct iov_iter *iter,
1271 struct page **pages;
1277 struct bio_vec *bvec;
1279 iov_for_each(iov, i, *iter) {
1280 unsigned long uaddr = (unsigned long) iov.iov_base;
1281 unsigned long len = iov.iov_len;
1282 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1283 unsigned long start = uaddr >> PAGE_SHIFT;
1289 return ERR_PTR(-EINVAL);
1291 nr_pages += end - start;
1293 * buffer must be aligned to at least hardsector size for now
1295 if (uaddr & queue_dma_alignment(q))
1296 return ERR_PTR(-EINVAL);
1300 return ERR_PTR(-EINVAL);
1302 bio = bio_kmalloc(gfp_mask, nr_pages);
1304 return ERR_PTR(-ENOMEM);
1307 pages = kcalloc(nr_pages, sizeof(struct page *), gfp_mask);
1311 iov_for_each(iov, i, *iter) {
1312 unsigned long uaddr = (unsigned long) iov.iov_base;
1313 unsigned long len = iov.iov_len;
1314 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1315 unsigned long start = uaddr >> PAGE_SHIFT;
1316 const int local_nr_pages = end - start;
1317 const int page_limit = cur_page + local_nr_pages;
1319 ret = get_user_pages_fast(uaddr, local_nr_pages,
1320 (iter->type & WRITE) != WRITE,
1322 if (unlikely(ret < local_nr_pages)) {
1323 for (j = cur_page; j < page_limit; j++) {
1332 offset = uaddr & ~PAGE_MASK;
1333 for (j = cur_page; j < page_limit; j++) {
1334 unsigned int bytes = PAGE_SIZE - offset;
1335 unsigned short prev_bi_vcnt = bio->bi_vcnt;
1346 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) <
1351 * check if vector was merged with previous
1352 * drop page reference if needed
1354 if (bio->bi_vcnt == prev_bi_vcnt)
1363 * release the pages we didn't map into the bio, if any
1365 while (j < page_limit)
1366 page_cache_release(pages[j++]);
1372 * set data direction, and check if mapped pages need bouncing
1374 if (iter->type & WRITE)
1375 bio->bi_rw |= REQ_WRITE;
1377 bio_set_flag(bio, BIO_USER_MAPPED);
1380 * subtle -- if __bio_map_user() ended up bouncing a bio,
1381 * it would normally disappear when its bi_end_io is run.
1382 * however, we need it for the unmap, so grab an extra
1389 bio_for_each_segment_all(bvec, bio, j) {
1390 put_page(bvec->bv_page);
1395 return ERR_PTR(ret);
1398 static void __bio_unmap_user(struct bio *bio)
1400 struct bio_vec *bvec;
1404 * make sure we dirty pages we wrote to
1406 bio_for_each_segment_all(bvec, bio, i) {
1407 if (bio_data_dir(bio) == READ)
1408 set_page_dirty_lock(bvec->bv_page);
1410 page_cache_release(bvec->bv_page);
1417 * bio_unmap_user - unmap a bio
1418 * @bio: the bio being unmapped
1420 * Unmap a bio previously mapped by bio_map_user(). Must be called with
1421 * a process context.
1423 * bio_unmap_user() may sleep.
1425 void bio_unmap_user(struct bio *bio)
1427 __bio_unmap_user(bio);
1430 EXPORT_SYMBOL(bio_unmap_user);
1432 static void bio_map_kern_endio(struct bio *bio)
1438 * bio_map_kern - map kernel address into bio
1439 * @q: the struct request_queue for the bio
1440 * @data: pointer to buffer to map
1441 * @len: length in bytes
1442 * @gfp_mask: allocation flags for bio allocation
1444 * Map the kernel address into a bio suitable for io to a block
1445 * device. Returns an error pointer in case of error.
1447 struct bio *bio_map_kern(struct request_queue *q, void *data, unsigned int len,
1450 unsigned long kaddr = (unsigned long)data;
1451 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1452 unsigned long start = kaddr >> PAGE_SHIFT;
1453 const int nr_pages = end - start;
1457 bio = bio_kmalloc(gfp_mask, nr_pages);
1459 return ERR_PTR(-ENOMEM);
1461 offset = offset_in_page(kaddr);
1462 for (i = 0; i < nr_pages; i++) {
1463 unsigned int bytes = PAGE_SIZE - offset;
1471 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes,
1473 /* we don't support partial mappings */
1475 return ERR_PTR(-EINVAL);
1483 bio->bi_end_io = bio_map_kern_endio;
1486 EXPORT_SYMBOL(bio_map_kern);
1488 static void bio_copy_kern_endio(struct bio *bio)
1490 bio_free_pages(bio);
1494 static void bio_copy_kern_endio_read(struct bio *bio)
1496 char *p = bio->bi_private;
1497 struct bio_vec *bvec;
1500 bio_for_each_segment_all(bvec, bio, i) {
1501 memcpy(p, page_address(bvec->bv_page), bvec->bv_len);
1505 bio_copy_kern_endio(bio);
1509 * bio_copy_kern - copy kernel address into bio
1510 * @q: the struct request_queue for the bio
1511 * @data: pointer to buffer to copy
1512 * @len: length in bytes
1513 * @gfp_mask: allocation flags for bio and page allocation
1514 * @reading: data direction is READ
1516 * copy the kernel address into a bio suitable for io to a block
1517 * device. Returns an error pointer in case of error.
1519 struct bio *bio_copy_kern(struct request_queue *q, void *data, unsigned int len,
1520 gfp_t gfp_mask, int reading)
1522 unsigned long kaddr = (unsigned long)data;
1523 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
1524 unsigned long start = kaddr >> PAGE_SHIFT;
1533 return ERR_PTR(-EINVAL);
1535 nr_pages = end - start;
1536 bio = bio_kmalloc(gfp_mask, nr_pages);
1538 return ERR_PTR(-ENOMEM);
1542 unsigned int bytes = PAGE_SIZE;
1547 page = alloc_page(q->bounce_gfp | gfp_mask);
1552 memcpy(page_address(page), p, bytes);
1554 if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes)
1562 bio->bi_end_io = bio_copy_kern_endio_read;
1563 bio->bi_private = data;
1565 bio->bi_end_io = bio_copy_kern_endio;
1566 bio->bi_rw |= REQ_WRITE;
1572 bio_free_pages(bio);
1574 return ERR_PTR(-ENOMEM);
1576 EXPORT_SYMBOL(bio_copy_kern);
1579 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1580 * for performing direct-IO in BIOs.
1582 * The problem is that we cannot run set_page_dirty() from interrupt context
1583 * because the required locks are not interrupt-safe. So what we can do is to
1584 * mark the pages dirty _before_ performing IO. And in interrupt context,
1585 * check that the pages are still dirty. If so, fine. If not, redirty them
1586 * in process context.
1588 * We special-case compound pages here: normally this means reads into hugetlb
1589 * pages. The logic in here doesn't really work right for compound pages
1590 * because the VM does not uniformly chase down the head page in all cases.
1591 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1592 * handle them at all. So we skip compound pages here at an early stage.
1594 * Note that this code is very hard to test under normal circumstances because
1595 * direct-io pins the pages with get_user_pages(). This makes
1596 * is_page_cache_freeable return false, and the VM will not clean the pages.
1597 * But other code (eg, flusher threads) could clean the pages if they are mapped
1600 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1601 * deferred bio dirtying paths.
1605 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1607 void bio_set_pages_dirty(struct bio *bio)
1609 struct bio_vec *bvec;
1612 bio_for_each_segment_all(bvec, bio, i) {
1613 struct page *page = bvec->bv_page;
1615 if (page && !PageCompound(page))
1616 set_page_dirty_lock(page);
1620 static void bio_release_pages(struct bio *bio)
1622 struct bio_vec *bvec;
1625 bio_for_each_segment_all(bvec, bio, i) {
1626 struct page *page = bvec->bv_page;
1634 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1635 * If they are, then fine. If, however, some pages are clean then they must
1636 * have been written out during the direct-IO read. So we take another ref on
1637 * the BIO and the offending pages and re-dirty the pages in process context.
1639 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1640 * here on. It will run one page_cache_release() against each page and will
1641 * run one bio_put() against the BIO.
1644 static void bio_dirty_fn(struct work_struct *work);
1646 static DECLARE_WORK(bio_dirty_work, bio_dirty_fn);
1647 static DEFINE_SPINLOCK(bio_dirty_lock);
1648 static struct bio *bio_dirty_list;
1651 * This runs in process context
1653 static void bio_dirty_fn(struct work_struct *work)
1655 unsigned long flags;
1658 spin_lock_irqsave(&bio_dirty_lock, flags);
1659 bio = bio_dirty_list;
1660 bio_dirty_list = NULL;
1661 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1664 struct bio *next = bio->bi_private;
1666 bio_set_pages_dirty(bio);
1667 bio_release_pages(bio);
1673 void bio_check_pages_dirty(struct bio *bio)
1675 struct bio_vec *bvec;
1676 int nr_clean_pages = 0;
1679 bio_for_each_segment_all(bvec, bio, i) {
1680 struct page *page = bvec->bv_page;
1682 if (PageDirty(page) || PageCompound(page)) {
1683 page_cache_release(page);
1684 bvec->bv_page = NULL;
1690 if (nr_clean_pages) {
1691 unsigned long flags;
1693 spin_lock_irqsave(&bio_dirty_lock, flags);
1694 bio->bi_private = bio_dirty_list;
1695 bio_dirty_list = bio;
1696 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1697 schedule_work(&bio_dirty_work);
1703 void generic_start_io_acct(int rw, unsigned long sectors,
1704 struct hd_struct *part)
1706 int cpu = part_stat_lock();
1708 part_round_stats(cpu, part);
1709 part_stat_inc(cpu, part, ios[rw]);
1710 part_stat_add(cpu, part, sectors[rw], sectors);
1711 part_inc_in_flight(part, rw);
1715 EXPORT_SYMBOL(generic_start_io_acct);
1717 void generic_end_io_acct(int rw, struct hd_struct *part,
1718 unsigned long start_time)
1720 unsigned long duration = jiffies - start_time;
1721 int cpu = part_stat_lock();
1723 part_stat_add(cpu, part, ticks[rw], duration);
1724 part_round_stats(cpu, part);
1725 part_dec_in_flight(part, rw);
1729 EXPORT_SYMBOL(generic_end_io_acct);
1731 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1732 void bio_flush_dcache_pages(struct bio *bi)
1734 struct bio_vec bvec;
1735 struct bvec_iter iter;
1737 bio_for_each_segment(bvec, bi, iter)
1738 flush_dcache_page(bvec.bv_page);
1740 EXPORT_SYMBOL(bio_flush_dcache_pages);
1743 static inline bool bio_remaining_done(struct bio *bio)
1746 * If we're not chaining, then ->__bi_remaining is always 1 and
1747 * we always end io on the first invocation.
1749 if (!bio_flagged(bio, BIO_CHAIN))
1752 BUG_ON(atomic_read(&bio->__bi_remaining) <= 0);
1754 if (atomic_dec_and_test(&bio->__bi_remaining)) {
1755 bio_clear_flag(bio, BIO_CHAIN);
1763 * bio_endio - end I/O on a bio
1767 * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred
1768 * way to end I/O on a bio. No one should call bi_end_io() directly on a
1769 * bio unless they own it and thus know that it has an end_io function.
1771 void bio_endio(struct bio *bio)
1774 if (unlikely(!bio_remaining_done(bio)))
1778 * Need to have a real endio function for chained bios,
1779 * otherwise various corner cases will break (like stacking
1780 * block devices that save/restore bi_end_io) - however, we want
1781 * to avoid unbounded recursion and blowing the stack. Tail call
1782 * optimization would handle this, but compiling with frame
1783 * pointers also disables gcc's sibling call optimization.
1785 if (bio->bi_end_io == bio_chain_endio) {
1786 struct bio *parent = bio->bi_private;
1787 parent->bi_error = bio->bi_error;
1791 if (bio->bi_end_io) {
1792 blk_update_perf_stats(bio);
1793 bio->bi_end_io(bio);
1799 EXPORT_SYMBOL(bio_endio);
1802 * bio_split - split a bio
1803 * @bio: bio to split
1804 * @sectors: number of sectors to split from the front of @bio
1806 * @bs: bio set to allocate from
1808 * Allocates and returns a new bio which represents @sectors from the start of
1809 * @bio, and updates @bio to represent the remaining sectors.
1811 * Unless this is a discard request the newly allocated bio will point
1812 * to @bio's bi_io_vec; it is the caller's responsibility to ensure that
1813 * @bio is not freed before the split.
1815 struct bio *bio_split(struct bio *bio, int sectors,
1816 gfp_t gfp, struct bio_set *bs)
1818 struct bio *split = NULL;
1820 BUG_ON(sectors <= 0);
1821 BUG_ON(sectors >= bio_sectors(bio));
1824 * Discards need a mutable bio_vec to accommodate the payload
1825 * required by the DSM TRIM and UNMAP commands.
1827 if (bio->bi_rw & REQ_DISCARD)
1828 split = bio_clone_bioset(bio, gfp, bs);
1830 split = bio_clone_fast(bio, gfp, bs);
1835 split->bi_iter.bi_size = sectors << 9;
1837 if (bio_integrity(split))
1838 bio_integrity_trim(split, 0, sectors);
1840 bio_advance(bio, split->bi_iter.bi_size);
1844 EXPORT_SYMBOL(bio_split);
1847 * bio_trim - trim a bio
1849 * @offset: number of sectors to trim from the front of @bio
1850 * @size: size we want to trim @bio to, in sectors
1852 void bio_trim(struct bio *bio, int offset, int size)
1854 /* 'bio' is a cloned bio which we need to trim to match
1855 * the given offset and size.
1859 if (offset == 0 && size == bio->bi_iter.bi_size)
1862 bio_clear_flag(bio, BIO_SEG_VALID);
1864 bio_advance(bio, offset << 9);
1866 bio->bi_iter.bi_size = size;
1868 EXPORT_SYMBOL_GPL(bio_trim);
1871 * create memory pools for biovec's in a bio_set.
1872 * use the global biovec slabs created for general use.
1874 mempool_t *biovec_create_pool(int pool_entries)
1876 struct biovec_slab *bp = bvec_slabs + BIOVEC_MAX_IDX;
1878 return mempool_create_slab_pool(pool_entries, bp->slab);
1881 void bioset_free(struct bio_set *bs)
1883 if (bs->rescue_workqueue)
1884 destroy_workqueue(bs->rescue_workqueue);
1887 mempool_destroy(bs->bio_pool);
1890 mempool_destroy(bs->bvec_pool);
1892 bioset_integrity_free(bs);
1897 EXPORT_SYMBOL(bioset_free);
1899 static struct bio_set *__bioset_create(unsigned int pool_size,
1900 unsigned int front_pad,
1901 bool create_bvec_pool)
1903 unsigned int back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec);
1906 bs = kzalloc(sizeof(*bs), GFP_KERNEL);
1910 bs->front_pad = front_pad;
1912 spin_lock_init(&bs->rescue_lock);
1913 bio_list_init(&bs->rescue_list);
1914 INIT_WORK(&bs->rescue_work, bio_alloc_rescue);
1916 bs->bio_slab = bio_find_or_create_slab(front_pad + back_pad);
1917 if (!bs->bio_slab) {
1922 bs->bio_pool = mempool_create_slab_pool(pool_size, bs->bio_slab);
1926 if (create_bvec_pool) {
1927 bs->bvec_pool = biovec_create_pool(pool_size);
1932 bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0);
1933 if (!bs->rescue_workqueue)
1943 * bioset_create - Create a bio_set
1944 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1945 * @front_pad: Number of bytes to allocate in front of the returned bio
1948 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
1949 * to ask for a number of bytes to be allocated in front of the bio.
1950 * Front pad allocation is useful for embedding the bio inside
1951 * another structure, to avoid allocating extra data to go with the bio.
1952 * Note that the bio must be embedded at the END of that structure always,
1953 * or things will break badly.
1955 struct bio_set *bioset_create(unsigned int pool_size, unsigned int front_pad)
1957 return __bioset_create(pool_size, front_pad, true);
1959 EXPORT_SYMBOL(bioset_create);
1962 * bioset_create_nobvec - Create a bio_set without bio_vec mempool
1963 * @pool_size: Number of bio to cache in the mempool
1964 * @front_pad: Number of bytes to allocate in front of the returned bio
1967 * Same functionality as bioset_create() except that mempool is not
1968 * created for bio_vecs. Saving some memory for bio_clone_fast() users.
1970 struct bio_set *bioset_create_nobvec(unsigned int pool_size, unsigned int front_pad)
1972 return __bioset_create(pool_size, front_pad, false);
1974 EXPORT_SYMBOL(bioset_create_nobvec);
1976 #ifdef CONFIG_BLK_CGROUP
1979 * bio_associate_blkcg - associate a bio with the specified blkcg
1981 * @blkcg_css: css of the blkcg to associate
1983 * Associate @bio with the blkcg specified by @blkcg_css. Block layer will
1984 * treat @bio as if it were issued by a task which belongs to the blkcg.
1986 * This function takes an extra reference of @blkcg_css which will be put
1987 * when @bio is released. The caller must own @bio and is responsible for
1988 * synchronizing calls to this function.
1990 int bio_associate_blkcg(struct bio *bio, struct cgroup_subsys_state *blkcg_css)
1992 if (unlikely(bio->bi_css))
1995 bio->bi_css = blkcg_css;
1998 EXPORT_SYMBOL_GPL(bio_associate_blkcg);
2001 * bio_associate_current - associate a bio with %current
2004 * Associate @bio with %current if it hasn't been associated yet. Block
2005 * layer will treat @bio as if it were issued by %current no matter which
2006 * task actually issues it.
2008 * This function takes an extra reference of @task's io_context and blkcg
2009 * which will be put when @bio is released. The caller must own @bio,
2010 * ensure %current->io_context exists, and is responsible for synchronizing
2011 * calls to this function.
2013 int bio_associate_current(struct bio *bio)
2015 struct io_context *ioc;
2020 ioc = current->io_context;
2024 get_io_context_active(ioc);
2026 bio->bi_css = task_get_css(current, io_cgrp_id);
2029 EXPORT_SYMBOL_GPL(bio_associate_current);
2032 * bio_disassociate_task - undo bio_associate_current()
2035 void bio_disassociate_task(struct bio *bio)
2038 put_io_context(bio->bi_ioc);
2042 css_put(bio->bi_css);
2048 * bio_clone_blkcg_association - clone blkcg association from src to dst bio
2049 * @dst: destination bio
2052 void bio_clone_blkcg_association(struct bio *dst, struct bio *src)
2055 WARN_ON(bio_associate_blkcg(dst, src->bi_css));
2058 #endif /* CONFIG_BLK_CGROUP */
2060 static void __init biovec_init_slabs(void)
2064 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
2066 struct biovec_slab *bvs = bvec_slabs + i;
2068 if (bvs->nr_vecs <= BIO_INLINE_VECS) {
2073 size = bvs->nr_vecs * sizeof(struct bio_vec);
2074 bvs->slab = kmem_cache_create(bvs->name, size, 0,
2075 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL);
2079 static int __init init_bio(void)
2083 bio_slabs = kzalloc(bio_slab_max * sizeof(struct bio_slab), GFP_KERNEL);
2085 panic("bio: can't allocate bios\n");
2087 bio_integrity_init();
2088 biovec_init_slabs();
2090 fs_bio_set = bioset_create(BIO_POOL_SIZE, 0);
2092 panic("bio: can't allocate bios\n");
2094 if (bioset_integrity_create(fs_bio_set, BIO_POOL_SIZE))
2095 panic("bio: can't create integrity pool\n");
2099 subsys_initcall(init_bio);