1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk>
6 #include <linux/swap.h>
8 #include <linux/blkdev.h>
10 #include <linux/iocontext.h>
11 #include <linux/slab.h>
12 #include <linux/init.h>
13 #include <linux/kernel.h>
14 #include <linux/export.h>
15 #include <linux/mempool.h>
16 #include <linux/workqueue.h>
17 #include <linux/cgroup.h>
18 #include <linux/blk-cgroup.h>
19 #include <linux/highmem.h>
20 #include <linux/sched/sysctl.h>
21 #include <linux/blk-crypto.h>
22 #include <linux/xarray.h>
24 #include <trace/events/block.h>
26 #include "blk-rq-qos.h"
28 static struct biovec_slab {
31 struct kmem_cache *slab;
32 } bvec_slabs[] __read_mostly = {
33 { .nr_vecs = 16, .name = "biovec-16" },
34 { .nr_vecs = 64, .name = "biovec-64" },
35 { .nr_vecs = 128, .name = "biovec-128" },
36 { .nr_vecs = BIO_MAX_PAGES, .name = "biovec-max" },
40 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
41 * IO code that does not need private memory pools.
43 struct bio_set fs_bio_set;
44 EXPORT_SYMBOL(fs_bio_set);
47 * Our slab pool management
50 struct kmem_cache *slab;
51 unsigned int slab_ref;
52 unsigned int slab_size;
55 static DEFINE_MUTEX(bio_slab_lock);
56 static DEFINE_XARRAY(bio_slabs);
58 static struct bio_slab *create_bio_slab(unsigned int size)
60 struct bio_slab *bslab = kzalloc(sizeof(*bslab), GFP_KERNEL);
65 snprintf(bslab->name, sizeof(bslab->name), "bio-%d", size);
66 bslab->slab = kmem_cache_create(bslab->name, size,
67 ARCH_KMALLOC_MINALIGN, SLAB_HWCACHE_ALIGN, NULL);
72 bslab->slab_size = size;
74 if (!xa_err(xa_store(&bio_slabs, size, bslab, GFP_KERNEL)))
77 kmem_cache_destroy(bslab->slab);
84 static inline unsigned int bs_bio_slab_size(struct bio_set *bs)
86 return bs->front_pad + sizeof(struct bio) + bs->back_pad;
89 static struct kmem_cache *bio_find_or_create_slab(struct bio_set *bs)
91 unsigned int size = bs_bio_slab_size(bs);
92 struct bio_slab *bslab;
94 mutex_lock(&bio_slab_lock);
95 bslab = xa_load(&bio_slabs, size);
99 bslab = create_bio_slab(size);
100 mutex_unlock(&bio_slab_lock);
107 static void bio_put_slab(struct bio_set *bs)
109 struct bio_slab *bslab = NULL;
110 unsigned int slab_size = bs_bio_slab_size(bs);
112 mutex_lock(&bio_slab_lock);
114 bslab = xa_load(&bio_slabs, slab_size);
115 if (WARN(!bslab, KERN_ERR "bio: unable to find slab!\n"))
118 WARN_ON_ONCE(bslab->slab != bs->bio_slab);
120 WARN_ON(!bslab->slab_ref);
122 if (--bslab->slab_ref)
125 xa_erase(&bio_slabs, slab_size);
127 kmem_cache_destroy(bslab->slab);
131 mutex_unlock(&bio_slab_lock);
134 unsigned int bvec_nr_vecs(unsigned short idx)
136 return bvec_slabs[--idx].nr_vecs;
139 void bvec_free(mempool_t *pool, struct bio_vec *bv, unsigned int idx)
145 BIO_BUG_ON(idx >= BVEC_POOL_NR);
147 if (idx == BVEC_POOL_MAX) {
148 mempool_free(bv, pool);
150 struct biovec_slab *bvs = bvec_slabs + idx;
152 kmem_cache_free(bvs->slab, bv);
157 * Make the first allocation restricted and don't dump info on allocation
158 * failures, since we'll fall back to the mempool in case of failure.
160 static inline gfp_t bvec_alloc_gfp(gfp_t gfp)
162 return (gfp & ~(__GFP_DIRECT_RECLAIM | __GFP_IO)) |
163 __GFP_NOMEMALLOC | __GFP_NORETRY | __GFP_NOWARN;
166 struct bio_vec *bvec_alloc(gfp_t gfp_mask, int nr, unsigned long *idx,
170 * see comment near bvec_array define!
173 /* smaller bios use inline vecs */
183 case 129 ... BIO_MAX_PAGES:
191 * Try a slab allocation first for all smaller allocations. If that
192 * fails and __GFP_DIRECT_RECLAIM is set retry with the mempool.
193 * The mempool is sized to handle up to BIO_MAX_PAGES entries.
195 if (*idx < BVEC_POOL_MAX) {
196 struct biovec_slab *bvs = bvec_slabs + *idx;
199 bvl = kmem_cache_alloc(bvs->slab, bvec_alloc_gfp(gfp_mask));
200 if (likely(bvl) || !(gfp_mask & __GFP_DIRECT_RECLAIM)) {
204 *idx = BVEC_POOL_MAX;
208 return mempool_alloc(pool, gfp_mask);
211 void bio_uninit(struct bio *bio)
213 #ifdef CONFIG_BLK_CGROUP
215 blkg_put(bio->bi_blkg);
219 if (bio_integrity(bio))
220 bio_integrity_free(bio);
222 bio_crypt_free_ctx(bio);
224 EXPORT_SYMBOL(bio_uninit);
226 static void bio_free(struct bio *bio)
228 struct bio_set *bs = bio->bi_pool;
234 bvec_free(&bs->bvec_pool, bio->bi_io_vec, BVEC_POOL_IDX(bio));
237 * If we have front padding, adjust the bio pointer before freeing
242 mempool_free(p, &bs->bio_pool);
244 /* Bio was allocated by bio_kmalloc() */
250 * Users of this function have their own bio allocation. Subsequently,
251 * they must remember to pair any call to bio_init() with bio_uninit()
252 * when IO has completed, or when the bio is released.
254 void bio_init(struct bio *bio, struct bio_vec *table,
255 unsigned short max_vecs)
257 memset(bio, 0, sizeof(*bio));
258 atomic_set(&bio->__bi_remaining, 1);
259 atomic_set(&bio->__bi_cnt, 1);
261 bio->bi_io_vec = table;
262 bio->bi_max_vecs = max_vecs;
264 EXPORT_SYMBOL(bio_init);
267 * bio_reset - reinitialize a bio
271 * After calling bio_reset(), @bio will be in the same state as a freshly
272 * allocated bio returned bio bio_alloc_bioset() - the only fields that are
273 * preserved are the ones that are initialized by bio_alloc_bioset(). See
274 * comment in struct bio.
276 void bio_reset(struct bio *bio)
278 unsigned long flags = bio->bi_flags & (~0UL << BIO_RESET_BITS);
282 memset(bio, 0, BIO_RESET_BYTES);
283 bio->bi_flags = flags;
284 atomic_set(&bio->__bi_remaining, 1);
286 EXPORT_SYMBOL(bio_reset);
288 static struct bio *__bio_chain_endio(struct bio *bio)
290 struct bio *parent = bio->bi_private;
292 if (!parent->bi_status)
293 parent->bi_status = bio->bi_status;
298 static void bio_chain_endio(struct bio *bio)
300 bio_endio(__bio_chain_endio(bio));
304 * bio_chain - chain bio completions
305 * @bio: the target bio
306 * @parent: the parent bio of @bio
308 * The caller won't have a bi_end_io called when @bio completes - instead,
309 * @parent's bi_end_io won't be called until both @parent and @bio have
310 * completed; the chained bio will also be freed when it completes.
312 * The caller must not set bi_private or bi_end_io in @bio.
314 void bio_chain(struct bio *bio, struct bio *parent)
316 BUG_ON(bio->bi_private || bio->bi_end_io);
318 bio->bi_private = parent;
319 bio->bi_end_io = bio_chain_endio;
320 bio_inc_remaining(parent);
322 EXPORT_SYMBOL(bio_chain);
324 static void bio_alloc_rescue(struct work_struct *work)
326 struct bio_set *bs = container_of(work, struct bio_set, rescue_work);
330 spin_lock(&bs->rescue_lock);
331 bio = bio_list_pop(&bs->rescue_list);
332 spin_unlock(&bs->rescue_lock);
337 submit_bio_noacct(bio);
341 static void punt_bios_to_rescuer(struct bio_set *bs)
343 struct bio_list punt, nopunt;
346 if (WARN_ON_ONCE(!bs->rescue_workqueue))
349 * In order to guarantee forward progress we must punt only bios that
350 * were allocated from this bio_set; otherwise, if there was a bio on
351 * there for a stacking driver higher up in the stack, processing it
352 * could require allocating bios from this bio_set, and doing that from
353 * our own rescuer would be bad.
355 * Since bio lists are singly linked, pop them all instead of trying to
356 * remove from the middle of the list:
359 bio_list_init(&punt);
360 bio_list_init(&nopunt);
362 while ((bio = bio_list_pop(¤t->bio_list[0])))
363 bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio);
364 current->bio_list[0] = nopunt;
366 bio_list_init(&nopunt);
367 while ((bio = bio_list_pop(¤t->bio_list[1])))
368 bio_list_add(bio->bi_pool == bs ? &punt : &nopunt, bio);
369 current->bio_list[1] = nopunt;
371 spin_lock(&bs->rescue_lock);
372 bio_list_merge(&bs->rescue_list, &punt);
373 spin_unlock(&bs->rescue_lock);
375 queue_work(bs->rescue_workqueue, &bs->rescue_work);
379 * bio_alloc_bioset - allocate a bio for I/O
380 * @gfp_mask: the GFP_* mask given to the slab allocator
381 * @nr_iovecs: number of iovecs to pre-allocate
382 * @bs: the bio_set to allocate from.
384 * Allocate a bio from the mempools in @bs.
386 * If %__GFP_DIRECT_RECLAIM is set then bio_alloc will always be able to
387 * allocate a bio. This is due to the mempool guarantees. To make this work,
388 * callers must never allocate more than 1 bio at a time from the general pool.
389 * Callers that need to allocate more than 1 bio must always submit the
390 * previously allocated bio for IO before attempting to allocate a new one.
391 * Failure to do so can cause deadlocks under memory pressure.
393 * Note that when running under submit_bio_noacct() (i.e. any block driver),
394 * bios are not submitted until after you return - see the code in
395 * submit_bio_noacct() that converts recursion into iteration, to prevent
398 * This would normally mean allocating multiple bios under submit_bio_noacct()
399 * would be susceptible to deadlocks, but we have
400 * deadlock avoidance code that resubmits any blocked bios from a rescuer
403 * However, we do not guarantee forward progress for allocations from other
404 * mempools. Doing multiple allocations from the same mempool under
405 * submit_bio_noacct() should be avoided - instead, use bio_set's front_pad
406 * for per bio allocations.
408 * Returns: Pointer to new bio on success, NULL on failure.
410 struct bio *bio_alloc_bioset(gfp_t gfp_mask, unsigned short nr_iovecs,
413 gfp_t saved_gfp = gfp_mask;
417 /* should not use nobvec bioset for nr_iovecs > 0 */
418 if (WARN_ON_ONCE(!mempool_initialized(&bs->bvec_pool) && nr_iovecs > 0))
422 * submit_bio_noacct() converts recursion to iteration; this means if
423 * we're running beneath it, any bios we allocate and submit will not be
424 * submitted (and thus freed) until after we return.
426 * This exposes us to a potential deadlock if we allocate multiple bios
427 * from the same bio_set() while running underneath submit_bio_noacct().
428 * If we were to allocate multiple bios (say a stacking block driver
429 * that was splitting bios), we would deadlock if we exhausted the
432 * We solve this, and guarantee forward progress, with a rescuer
433 * workqueue per bio_set. If we go to allocate and there are bios on
434 * current->bio_list, we first try the allocation without
435 * __GFP_DIRECT_RECLAIM; if that fails, we punt those bios we would be
436 * blocking to the rescuer workqueue before we retry with the original
439 if (current->bio_list &&
440 (!bio_list_empty(¤t->bio_list[0]) ||
441 !bio_list_empty(¤t->bio_list[1])) &&
442 bs->rescue_workqueue)
443 gfp_mask &= ~__GFP_DIRECT_RECLAIM;
445 p = mempool_alloc(&bs->bio_pool, gfp_mask);
446 if (!p && gfp_mask != saved_gfp) {
447 punt_bios_to_rescuer(bs);
448 gfp_mask = saved_gfp;
449 p = mempool_alloc(&bs->bio_pool, gfp_mask);
454 bio = p + bs->front_pad;
455 if (nr_iovecs > BIO_INLINE_VECS) {
456 unsigned long idx = 0;
457 struct bio_vec *bvl = NULL;
459 bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx, &bs->bvec_pool);
460 if (!bvl && gfp_mask != saved_gfp) {
461 punt_bios_to_rescuer(bs);
462 gfp_mask = saved_gfp;
463 bvl = bvec_alloc(gfp_mask, nr_iovecs, &idx,
470 bio_init(bio, bvl, bvec_nr_vecs(idx));
471 bio->bi_flags |= idx << BVEC_POOL_OFFSET;
472 } else if (nr_iovecs) {
473 bio_init(bio, bio->bi_inline_vecs, BIO_INLINE_VECS);
475 bio_init(bio, NULL, 0);
482 mempool_free(p, &bs->bio_pool);
485 EXPORT_SYMBOL(bio_alloc_bioset);
488 * bio_kmalloc - kmalloc a bio for I/O
489 * @gfp_mask: the GFP_* mask given to the slab allocator
490 * @nr_iovecs: number of iovecs to pre-allocate
492 * Use kmalloc to allocate and initialize a bio.
494 * Returns: Pointer to new bio on success, NULL on failure.
496 struct bio *bio_kmalloc(gfp_t gfp_mask, unsigned short nr_iovecs)
500 if (nr_iovecs > UIO_MAXIOV)
503 bio = kmalloc(struct_size(bio, bi_inline_vecs, nr_iovecs), gfp_mask);
506 bio_init(bio, nr_iovecs ? bio->bi_inline_vecs : NULL, nr_iovecs);
510 EXPORT_SYMBOL(bio_kmalloc);
512 void zero_fill_bio_iter(struct bio *bio, struct bvec_iter start)
516 struct bvec_iter iter;
518 __bio_for_each_segment(bv, bio, iter, start) {
519 char *data = bvec_kmap_irq(&bv, &flags);
520 memset(data, 0, bv.bv_len);
521 flush_dcache_page(bv.bv_page);
522 bvec_kunmap_irq(data, &flags);
525 EXPORT_SYMBOL(zero_fill_bio_iter);
528 * bio_truncate - truncate the bio to small size of @new_size
529 * @bio: the bio to be truncated
530 * @new_size: new size for truncating the bio
533 * Truncate the bio to new size of @new_size. If bio_op(bio) is
534 * REQ_OP_READ, zero the truncated part. This function should only
535 * be used for handling corner cases, such as bio eod.
537 void bio_truncate(struct bio *bio, unsigned new_size)
540 struct bvec_iter iter;
541 unsigned int done = 0;
542 bool truncated = false;
544 if (new_size >= bio->bi_iter.bi_size)
547 if (bio_op(bio) != REQ_OP_READ)
550 bio_for_each_segment(bv, bio, iter) {
551 if (done + bv.bv_len > new_size) {
555 offset = new_size - done;
558 zero_user(bv.bv_page, offset, bv.bv_len - offset);
566 * Don't touch bvec table here and make it really immutable, since
567 * fs bio user has to retrieve all pages via bio_for_each_segment_all
568 * in its .end_bio() callback.
570 * It is enough to truncate bio by updating .bi_size since we can make
571 * correct bvec with the updated .bi_size for drivers.
573 bio->bi_iter.bi_size = new_size;
577 * guard_bio_eod - truncate a BIO to fit the block device
578 * @bio: bio to truncate
580 * This allows us to do IO even on the odd last sectors of a device, even if the
581 * block size is some multiple of the physical sector size.
583 * We'll just truncate the bio to the size of the device, and clear the end of
584 * the buffer head manually. Truly out-of-range accesses will turn into actual
585 * I/O errors, this only handles the "we need to be able to do I/O at the final
588 void guard_bio_eod(struct bio *bio)
590 sector_t maxsector = bdev_nr_sectors(bio->bi_bdev);
596 * If the *whole* IO is past the end of the device,
597 * let it through, and the IO layer will turn it into
600 if (unlikely(bio->bi_iter.bi_sector >= maxsector))
603 maxsector -= bio->bi_iter.bi_sector;
604 if (likely((bio->bi_iter.bi_size >> 9) <= maxsector))
607 bio_truncate(bio, maxsector << 9);
611 * bio_put - release a reference to a bio
612 * @bio: bio to release reference to
615 * Put a reference to a &struct bio, either one you have gotten with
616 * bio_alloc, bio_get or bio_clone_*. The last put of a bio will free it.
618 void bio_put(struct bio *bio)
620 if (!bio_flagged(bio, BIO_REFFED))
623 BIO_BUG_ON(!atomic_read(&bio->__bi_cnt));
628 if (atomic_dec_and_test(&bio->__bi_cnt))
632 EXPORT_SYMBOL(bio_put);
635 * __bio_clone_fast - clone a bio that shares the original bio's biovec
636 * @bio: destination bio
637 * @bio_src: bio to clone
639 * Clone a &bio. Caller will own the returned bio, but not
640 * the actual data it points to. Reference count of returned
643 * Caller must ensure that @bio_src is not freed before @bio.
645 void __bio_clone_fast(struct bio *bio, struct bio *bio_src)
647 BUG_ON(bio->bi_pool && BVEC_POOL_IDX(bio));
650 * most users will be overriding ->bi_bdev with a new target,
651 * so we don't set nor calculate new physical/hw segment counts here
653 bio->bi_bdev = bio_src->bi_bdev;
654 bio_set_flag(bio, BIO_CLONED);
655 if (bio_flagged(bio_src, BIO_THROTTLED))
656 bio_set_flag(bio, BIO_THROTTLED);
657 if (bio_flagged(bio_src, BIO_REMAPPED))
658 bio_set_flag(bio, BIO_REMAPPED);
659 bio->bi_opf = bio_src->bi_opf;
660 bio->bi_ioprio = bio_src->bi_ioprio;
661 bio->bi_write_hint = bio_src->bi_write_hint;
662 bio->bi_iter = bio_src->bi_iter;
663 bio->bi_io_vec = bio_src->bi_io_vec;
665 bio_clone_blkg_association(bio, bio_src);
666 blkcg_bio_issue_init(bio);
668 EXPORT_SYMBOL(__bio_clone_fast);
671 * bio_clone_fast - clone a bio that shares the original bio's biovec
673 * @gfp_mask: allocation priority
674 * @bs: bio_set to allocate from
676 * Like __bio_clone_fast, only also allocates the returned bio
678 struct bio *bio_clone_fast(struct bio *bio, gfp_t gfp_mask, struct bio_set *bs)
682 b = bio_alloc_bioset(gfp_mask, 0, bs);
686 __bio_clone_fast(b, bio);
688 if (bio_crypt_clone(b, bio, gfp_mask) < 0)
691 if (bio_integrity(bio) &&
692 bio_integrity_clone(b, bio, gfp_mask) < 0)
701 EXPORT_SYMBOL(bio_clone_fast);
703 const char *bio_devname(struct bio *bio, char *buf)
705 return bdevname(bio->bi_bdev, buf);
707 EXPORT_SYMBOL(bio_devname);
709 static inline bool page_is_mergeable(const struct bio_vec *bv,
710 struct page *page, unsigned int len, unsigned int off,
713 size_t bv_end = bv->bv_offset + bv->bv_len;
714 phys_addr_t vec_end_addr = page_to_phys(bv->bv_page) + bv_end - 1;
715 phys_addr_t page_addr = page_to_phys(page);
717 if (vec_end_addr + 1 != page_addr + off)
719 if (xen_domain() && !xen_biovec_phys_mergeable(bv, page))
722 *same_page = ((vec_end_addr & PAGE_MASK) == page_addr);
725 return (bv->bv_page + bv_end / PAGE_SIZE) == (page + off / PAGE_SIZE);
729 * Try to merge a page into a segment, while obeying the hardware segment
730 * size limit. This is not for normal read/write bios, but for passthrough
731 * or Zone Append operations that we can't split.
733 static bool bio_try_merge_hw_seg(struct request_queue *q, struct bio *bio,
734 struct page *page, unsigned len,
735 unsigned offset, bool *same_page)
737 struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
738 unsigned long mask = queue_segment_boundary(q);
739 phys_addr_t addr1 = page_to_phys(bv->bv_page) + bv->bv_offset;
740 phys_addr_t addr2 = page_to_phys(page) + offset + len - 1;
742 if ((addr1 | mask) != (addr2 | mask))
744 if (bv->bv_len + len > queue_max_segment_size(q))
746 return __bio_try_merge_page(bio, page, len, offset, same_page);
750 * bio_add_hw_page - attempt to add a page to a bio with hw constraints
751 * @q: the target queue
752 * @bio: destination bio
754 * @len: vec entry length
755 * @offset: vec entry offset
756 * @max_sectors: maximum number of sectors that can be added
757 * @same_page: return if the segment has been merged inside the same page
759 * Add a page to a bio while respecting the hardware max_sectors, max_segment
760 * and gap limitations.
762 int bio_add_hw_page(struct request_queue *q, struct bio *bio,
763 struct page *page, unsigned int len, unsigned int offset,
764 unsigned int max_sectors, bool *same_page)
766 struct bio_vec *bvec;
768 if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)))
771 if (((bio->bi_iter.bi_size + len) >> 9) > max_sectors)
774 if (bio->bi_vcnt > 0) {
775 if (bio_try_merge_hw_seg(q, bio, page, len, offset, same_page))
779 * If the queue doesn't support SG gaps and adding this segment
780 * would create a gap, disallow it.
782 bvec = &bio->bi_io_vec[bio->bi_vcnt - 1];
783 if (bvec_gap_to_prev(q, bvec, offset))
787 if (bio_full(bio, len))
790 if (bio->bi_vcnt >= queue_max_segments(q))
793 bvec = &bio->bi_io_vec[bio->bi_vcnt];
794 bvec->bv_page = page;
796 bvec->bv_offset = offset;
798 bio->bi_iter.bi_size += len;
803 * bio_add_pc_page - attempt to add page to passthrough bio
804 * @q: the target queue
805 * @bio: destination bio
807 * @len: vec entry length
808 * @offset: vec entry offset
810 * Attempt to add a page to the bio_vec maplist. This can fail for a
811 * number of reasons, such as the bio being full or target block device
812 * limitations. The target block device must allow bio's up to PAGE_SIZE,
813 * so it is always possible to add a single page to an empty bio.
815 * This should only be used by passthrough bios.
817 int bio_add_pc_page(struct request_queue *q, struct bio *bio,
818 struct page *page, unsigned int len, unsigned int offset)
820 bool same_page = false;
821 return bio_add_hw_page(q, bio, page, len, offset,
822 queue_max_hw_sectors(q), &same_page);
824 EXPORT_SYMBOL(bio_add_pc_page);
827 * __bio_try_merge_page - try appending data to an existing bvec.
828 * @bio: destination bio
829 * @page: start page to add
830 * @len: length of the data to add
831 * @off: offset of the data relative to @page
832 * @same_page: return if the segment has been merged inside the same page
834 * Try to add the data at @page + @off to the last bvec of @bio. This is a
835 * useful optimisation for file systems with a block size smaller than the
838 * Warn if (@len, @off) crosses pages in case that @same_page is true.
840 * Return %true on success or %false on failure.
842 bool __bio_try_merge_page(struct bio *bio, struct page *page,
843 unsigned int len, unsigned int off, bool *same_page)
845 if (WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED)))
848 if (bio->bi_vcnt > 0) {
849 struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
851 if (page_is_mergeable(bv, page, len, off, same_page)) {
852 if (bio->bi_iter.bi_size > UINT_MAX - len) {
857 bio->bi_iter.bi_size += len;
863 EXPORT_SYMBOL_GPL(__bio_try_merge_page);
866 * __bio_add_page - add page(s) to a bio in a new segment
867 * @bio: destination bio
868 * @page: start page to add
869 * @len: length of the data to add, may cross pages
870 * @off: offset of the data relative to @page, may cross pages
872 * Add the data at @page + @off to @bio as a new bvec. The caller must ensure
873 * that @bio has space for another bvec.
875 void __bio_add_page(struct bio *bio, struct page *page,
876 unsigned int len, unsigned int off)
878 struct bio_vec *bv = &bio->bi_io_vec[bio->bi_vcnt];
880 WARN_ON_ONCE(bio_flagged(bio, BIO_CLONED));
881 WARN_ON_ONCE(bio_full(bio, len));
887 bio->bi_iter.bi_size += len;
890 if (!bio_flagged(bio, BIO_WORKINGSET) && unlikely(PageWorkingset(page)))
891 bio_set_flag(bio, BIO_WORKINGSET);
893 EXPORT_SYMBOL_GPL(__bio_add_page);
896 * bio_add_page - attempt to add page(s) to bio
897 * @bio: destination bio
898 * @page: start page to add
899 * @len: vec entry length, may cross pages
900 * @offset: vec entry offset relative to @page, may cross pages
902 * Attempt to add page(s) to the bio_vec maplist. This will only fail
903 * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio.
905 int bio_add_page(struct bio *bio, struct page *page,
906 unsigned int len, unsigned int offset)
908 bool same_page = false;
910 if (!__bio_try_merge_page(bio, page, len, offset, &same_page)) {
911 if (bio_full(bio, len))
913 __bio_add_page(bio, page, len, offset);
917 EXPORT_SYMBOL(bio_add_page);
919 void bio_release_pages(struct bio *bio, bool mark_dirty)
921 struct bvec_iter_all iter_all;
922 struct bio_vec *bvec;
924 if (bio_flagged(bio, BIO_NO_PAGE_REF))
927 bio_for_each_segment_all(bvec, bio, iter_all) {
928 if (mark_dirty && !PageCompound(bvec->bv_page))
929 set_page_dirty_lock(bvec->bv_page);
930 put_page(bvec->bv_page);
933 EXPORT_SYMBOL_GPL(bio_release_pages);
935 static int bio_iov_bvec_set(struct bio *bio, struct iov_iter *iter)
937 WARN_ON_ONCE(BVEC_POOL_IDX(bio) != 0);
939 bio->bi_vcnt = iter->nr_segs;
940 bio->bi_max_vecs = iter->nr_segs;
941 bio->bi_io_vec = (struct bio_vec *)iter->bvec;
942 bio->bi_iter.bi_bvec_done = iter->iov_offset;
943 bio->bi_iter.bi_size = iter->count;
945 iov_iter_advance(iter, iter->count);
949 #define PAGE_PTRS_PER_BVEC (sizeof(struct bio_vec) / sizeof(struct page *))
952 * __bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio
953 * @bio: bio to add pages to
954 * @iter: iov iterator describing the region to be mapped
956 * Pins pages from *iter and appends them to @bio's bvec array. The
957 * pages will have to be released using put_page() when done.
958 * For multi-segment *iter, this function only adds pages from the
959 * next non-empty segment of the iov iterator.
961 static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
963 unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt;
964 unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt;
965 struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt;
966 struct page **pages = (struct page **)bv;
967 bool same_page = false;
973 * Move page array up in the allocated memory for the bio vecs as far as
974 * possible so that we can start filling biovecs from the beginning
975 * without overwriting the temporary page array.
977 BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
978 pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
980 size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
981 if (unlikely(size <= 0))
982 return size ? size : -EFAULT;
984 for (left = size, i = 0; left > 0; left -= len, i++) {
985 struct page *page = pages[i];
987 len = min_t(size_t, PAGE_SIZE - offset, left);
989 if (__bio_try_merge_page(bio, page, len, offset, &same_page)) {
993 if (WARN_ON_ONCE(bio_full(bio, len)))
995 __bio_add_page(bio, page, len, offset);
1000 iov_iter_advance(iter, size);
1004 static int __bio_iov_append_get_pages(struct bio *bio, struct iov_iter *iter)
1006 unsigned short nr_pages = bio->bi_max_vecs - bio->bi_vcnt;
1007 unsigned short entries_left = bio->bi_max_vecs - bio->bi_vcnt;
1008 struct request_queue *q = bio->bi_bdev->bd_disk->queue;
1009 unsigned int max_append_sectors = queue_max_zone_append_sectors(q);
1010 struct bio_vec *bv = bio->bi_io_vec + bio->bi_vcnt;
1011 struct page **pages = (struct page **)bv;
1017 if (WARN_ON_ONCE(!max_append_sectors))
1021 * Move page array up in the allocated memory for the bio vecs as far as
1022 * possible so that we can start filling biovecs from the beginning
1023 * without overwriting the temporary page array.
1025 BUILD_BUG_ON(PAGE_PTRS_PER_BVEC < 2);
1026 pages += entries_left * (PAGE_PTRS_PER_BVEC - 1);
1028 size = iov_iter_get_pages(iter, pages, LONG_MAX, nr_pages, &offset);
1029 if (unlikely(size <= 0))
1030 return size ? size : -EFAULT;
1032 for (left = size, i = 0; left > 0; left -= len, i++) {
1033 struct page *page = pages[i];
1034 bool same_page = false;
1036 len = min_t(size_t, PAGE_SIZE - offset, left);
1037 if (bio_add_hw_page(q, bio, page, len, offset,
1038 max_append_sectors, &same_page) != len) {
1047 iov_iter_advance(iter, size - left);
1052 * bio_iov_iter_get_pages - add user or kernel pages to a bio
1053 * @bio: bio to add pages to
1054 * @iter: iov iterator describing the region to be added
1056 * This takes either an iterator pointing to user memory, or one pointing to
1057 * kernel pages (BVEC iterator). If we're adding user pages, we pin them and
1058 * map them into the kernel. On IO completion, the caller should put those
1059 * pages. For bvec based iterators bio_iov_iter_get_pages() uses the provided
1060 * bvecs rather than copying them. Hence anyone issuing kiocb based IO needs
1061 * to ensure the bvecs and pages stay referenced until the submitted I/O is
1062 * completed by a call to ->ki_complete() or returns with an error other than
1063 * -EIOCBQUEUED. The caller needs to check if the bio is flagged BIO_NO_PAGE_REF
1064 * on IO completion. If it isn't, then pages should be released.
1066 * The function tries, but does not guarantee, to pin as many pages as
1067 * fit into the bio, or are requested in @iter, whatever is smaller. If
1068 * MM encounters an error pinning the requested pages, it stops. Error
1069 * is returned only if 0 pages could be pinned.
1071 * It's intended for direct IO, so doesn't do PSI tracking, the caller is
1072 * responsible for setting BIO_WORKINGSET if necessary.
1074 int bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
1078 if (iov_iter_is_bvec(iter)) {
1079 if (WARN_ON_ONCE(bio_op(bio) == REQ_OP_ZONE_APPEND))
1081 bio_iov_bvec_set(bio, iter);
1082 bio_set_flag(bio, BIO_NO_PAGE_REF);
1087 if (bio_op(bio) == REQ_OP_ZONE_APPEND)
1088 ret = __bio_iov_append_get_pages(bio, iter);
1090 ret = __bio_iov_iter_get_pages(bio, iter);
1091 } while (!ret && iov_iter_count(iter) && !bio_full(bio, 0));
1093 /* don't account direct I/O as memory stall */
1094 bio_clear_flag(bio, BIO_WORKINGSET);
1095 return bio->bi_vcnt ? 0 : ret;
1097 EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages);
1099 static void submit_bio_wait_endio(struct bio *bio)
1101 complete(bio->bi_private);
1105 * submit_bio_wait - submit a bio, and wait until it completes
1106 * @bio: The &struct bio which describes the I/O
1108 * Simple wrapper around submit_bio(). Returns 0 on success, or the error from
1109 * bio_endio() on failure.
1111 * WARNING: Unlike to how submit_bio() is usually used, this function does not
1112 * result in bio reference to be consumed. The caller must drop the reference
1115 int submit_bio_wait(struct bio *bio)
1117 DECLARE_COMPLETION_ONSTACK_MAP(done,
1118 bio->bi_bdev->bd_disk->lockdep_map);
1119 unsigned long hang_check;
1121 bio->bi_private = &done;
1122 bio->bi_end_io = submit_bio_wait_endio;
1123 bio->bi_opf |= REQ_SYNC;
1126 /* Prevent hang_check timer from firing at us during very long I/O */
1127 hang_check = sysctl_hung_task_timeout_secs;
1129 while (!wait_for_completion_io_timeout(&done,
1130 hang_check * (HZ/2)))
1133 wait_for_completion_io(&done);
1135 return blk_status_to_errno(bio->bi_status);
1137 EXPORT_SYMBOL(submit_bio_wait);
1140 * bio_advance - increment/complete a bio by some number of bytes
1141 * @bio: bio to advance
1142 * @bytes: number of bytes to complete
1144 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
1145 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
1146 * be updated on the last bvec as well.
1148 * @bio will then represent the remaining, uncompleted portion of the io.
1150 void bio_advance(struct bio *bio, unsigned bytes)
1152 if (bio_integrity(bio))
1153 bio_integrity_advance(bio, bytes);
1155 bio_crypt_advance(bio, bytes);
1156 bio_advance_iter(bio, &bio->bi_iter, bytes);
1158 EXPORT_SYMBOL(bio_advance);
1160 void bio_copy_data_iter(struct bio *dst, struct bvec_iter *dst_iter,
1161 struct bio *src, struct bvec_iter *src_iter)
1163 struct bio_vec src_bv, dst_bv;
1164 void *src_p, *dst_p;
1167 while (src_iter->bi_size && dst_iter->bi_size) {
1168 src_bv = bio_iter_iovec(src, *src_iter);
1169 dst_bv = bio_iter_iovec(dst, *dst_iter);
1171 bytes = min(src_bv.bv_len, dst_bv.bv_len);
1173 src_p = kmap_atomic(src_bv.bv_page);
1174 dst_p = kmap_atomic(dst_bv.bv_page);
1176 memcpy(dst_p + dst_bv.bv_offset,
1177 src_p + src_bv.bv_offset,
1180 kunmap_atomic(dst_p);
1181 kunmap_atomic(src_p);
1183 flush_dcache_page(dst_bv.bv_page);
1185 bio_advance_iter_single(src, src_iter, bytes);
1186 bio_advance_iter_single(dst, dst_iter, bytes);
1189 EXPORT_SYMBOL(bio_copy_data_iter);
1192 * bio_copy_data - copy contents of data buffers from one bio to another
1194 * @dst: destination bio
1196 * Stops when it reaches the end of either @src or @dst - that is, copies
1197 * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios).
1199 void bio_copy_data(struct bio *dst, struct bio *src)
1201 struct bvec_iter src_iter = src->bi_iter;
1202 struct bvec_iter dst_iter = dst->bi_iter;
1204 bio_copy_data_iter(dst, &dst_iter, src, &src_iter);
1206 EXPORT_SYMBOL(bio_copy_data);
1209 * bio_list_copy_data - copy contents of data buffers from one chain of bios to
1211 * @src: source bio list
1212 * @dst: destination bio list
1214 * Stops when it reaches the end of either the @src list or @dst list - that is,
1215 * copies min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of
1218 void bio_list_copy_data(struct bio *dst, struct bio *src)
1220 struct bvec_iter src_iter = src->bi_iter;
1221 struct bvec_iter dst_iter = dst->bi_iter;
1224 if (!src_iter.bi_size) {
1229 src_iter = src->bi_iter;
1232 if (!dst_iter.bi_size) {
1237 dst_iter = dst->bi_iter;
1240 bio_copy_data_iter(dst, &dst_iter, src, &src_iter);
1243 EXPORT_SYMBOL(bio_list_copy_data);
1245 void bio_free_pages(struct bio *bio)
1247 struct bio_vec *bvec;
1248 struct bvec_iter_all iter_all;
1250 bio_for_each_segment_all(bvec, bio, iter_all)
1251 __free_page(bvec->bv_page);
1253 EXPORT_SYMBOL(bio_free_pages);
1256 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1257 * for performing direct-IO in BIOs.
1259 * The problem is that we cannot run set_page_dirty() from interrupt context
1260 * because the required locks are not interrupt-safe. So what we can do is to
1261 * mark the pages dirty _before_ performing IO. And in interrupt context,
1262 * check that the pages are still dirty. If so, fine. If not, redirty them
1263 * in process context.
1265 * We special-case compound pages here: normally this means reads into hugetlb
1266 * pages. The logic in here doesn't really work right for compound pages
1267 * because the VM does not uniformly chase down the head page in all cases.
1268 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1269 * handle them at all. So we skip compound pages here at an early stage.
1271 * Note that this code is very hard to test under normal circumstances because
1272 * direct-io pins the pages with get_user_pages(). This makes
1273 * is_page_cache_freeable return false, and the VM will not clean the pages.
1274 * But other code (eg, flusher threads) could clean the pages if they are mapped
1277 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1278 * deferred bio dirtying paths.
1282 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1284 void bio_set_pages_dirty(struct bio *bio)
1286 struct bio_vec *bvec;
1287 struct bvec_iter_all iter_all;
1289 bio_for_each_segment_all(bvec, bio, iter_all) {
1290 if (!PageCompound(bvec->bv_page))
1291 set_page_dirty_lock(bvec->bv_page);
1296 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1297 * If they are, then fine. If, however, some pages are clean then they must
1298 * have been written out during the direct-IO read. So we take another ref on
1299 * the BIO and re-dirty the pages in process context.
1301 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1302 * here on. It will run one put_page() against each page and will run one
1303 * bio_put() against the BIO.
1306 static void bio_dirty_fn(struct work_struct *work);
1308 static DECLARE_WORK(bio_dirty_work, bio_dirty_fn);
1309 static DEFINE_SPINLOCK(bio_dirty_lock);
1310 static struct bio *bio_dirty_list;
1313 * This runs in process context
1315 static void bio_dirty_fn(struct work_struct *work)
1317 struct bio *bio, *next;
1319 spin_lock_irq(&bio_dirty_lock);
1320 next = bio_dirty_list;
1321 bio_dirty_list = NULL;
1322 spin_unlock_irq(&bio_dirty_lock);
1324 while ((bio = next) != NULL) {
1325 next = bio->bi_private;
1327 bio_release_pages(bio, true);
1332 void bio_check_pages_dirty(struct bio *bio)
1334 struct bio_vec *bvec;
1335 unsigned long flags;
1336 struct bvec_iter_all iter_all;
1338 bio_for_each_segment_all(bvec, bio, iter_all) {
1339 if (!PageDirty(bvec->bv_page) && !PageCompound(bvec->bv_page))
1343 bio_release_pages(bio, false);
1347 spin_lock_irqsave(&bio_dirty_lock, flags);
1348 bio->bi_private = bio_dirty_list;
1349 bio_dirty_list = bio;
1350 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1351 schedule_work(&bio_dirty_work);
1354 static inline bool bio_remaining_done(struct bio *bio)
1357 * If we're not chaining, then ->__bi_remaining is always 1 and
1358 * we always end io on the first invocation.
1360 if (!bio_flagged(bio, BIO_CHAIN))
1363 BUG_ON(atomic_read(&bio->__bi_remaining) <= 0);
1365 if (atomic_dec_and_test(&bio->__bi_remaining)) {
1366 bio_clear_flag(bio, BIO_CHAIN);
1374 * bio_endio - end I/O on a bio
1378 * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred
1379 * way to end I/O on a bio. No one should call bi_end_io() directly on a
1380 * bio unless they own it and thus know that it has an end_io function.
1382 * bio_endio() can be called several times on a bio that has been chained
1383 * using bio_chain(). The ->bi_end_io() function will only be called the
1384 * last time. At this point the BLK_TA_COMPLETE tracing event will be
1385 * generated if BIO_TRACE_COMPLETION is set.
1387 void bio_endio(struct bio *bio)
1390 if (!bio_remaining_done(bio))
1392 if (!bio_integrity_endio(bio))
1396 rq_qos_done_bio(bio->bi_bdev->bd_disk->queue, bio);
1399 * Need to have a real endio function for chained bios, otherwise
1400 * various corner cases will break (like stacking block devices that
1401 * save/restore bi_end_io) - however, we want to avoid unbounded
1402 * recursion and blowing the stack. Tail call optimization would
1403 * handle this, but compiling with frame pointers also disables
1404 * gcc's sibling call optimization.
1406 if (bio->bi_end_io == bio_chain_endio) {
1407 bio = __bio_chain_endio(bio);
1411 if (bio->bi_bdev && bio_flagged(bio, BIO_TRACE_COMPLETION)) {
1412 trace_block_bio_complete(bio->bi_bdev->bd_disk->queue, bio);
1413 bio_clear_flag(bio, BIO_TRACE_COMPLETION);
1416 blk_throtl_bio_endio(bio);
1417 /* release cgroup info */
1420 bio->bi_end_io(bio);
1422 EXPORT_SYMBOL(bio_endio);
1425 * bio_split - split a bio
1426 * @bio: bio to split
1427 * @sectors: number of sectors to split from the front of @bio
1429 * @bs: bio set to allocate from
1431 * Allocates and returns a new bio which represents @sectors from the start of
1432 * @bio, and updates @bio to represent the remaining sectors.
1434 * Unless this is a discard request the newly allocated bio will point
1435 * to @bio's bi_io_vec. It is the caller's responsibility to ensure that
1436 * neither @bio nor @bs are freed before the split bio.
1438 struct bio *bio_split(struct bio *bio, int sectors,
1439 gfp_t gfp, struct bio_set *bs)
1443 BUG_ON(sectors <= 0);
1444 BUG_ON(sectors >= bio_sectors(bio));
1446 /* Zone append commands cannot be split */
1447 if (WARN_ON_ONCE(bio_op(bio) == REQ_OP_ZONE_APPEND))
1450 split = bio_clone_fast(bio, gfp, bs);
1454 split->bi_iter.bi_size = sectors << 9;
1456 if (bio_integrity(split))
1457 bio_integrity_trim(split);
1459 bio_advance(bio, split->bi_iter.bi_size);
1461 if (bio_flagged(bio, BIO_TRACE_COMPLETION))
1462 bio_set_flag(split, BIO_TRACE_COMPLETION);
1466 EXPORT_SYMBOL(bio_split);
1469 * bio_trim - trim a bio
1471 * @offset: number of sectors to trim from the front of @bio
1472 * @size: size we want to trim @bio to, in sectors
1474 void bio_trim(struct bio *bio, int offset, int size)
1476 /* 'bio' is a cloned bio which we need to trim to match
1477 * the given offset and size.
1481 if (offset == 0 && size == bio->bi_iter.bi_size)
1484 bio_advance(bio, offset << 9);
1485 bio->bi_iter.bi_size = size;
1487 if (bio_integrity(bio))
1488 bio_integrity_trim(bio);
1491 EXPORT_SYMBOL_GPL(bio_trim);
1494 * create memory pools for biovec's in a bio_set.
1495 * use the global biovec slabs created for general use.
1497 int biovec_init_pool(mempool_t *pool, int pool_entries)
1499 struct biovec_slab *bp = bvec_slabs + BVEC_POOL_MAX;
1501 return mempool_init_slab_pool(pool, pool_entries, bp->slab);
1505 * bioset_exit - exit a bioset initialized with bioset_init()
1507 * May be called on a zeroed but uninitialized bioset (i.e. allocated with
1510 void bioset_exit(struct bio_set *bs)
1512 if (bs->rescue_workqueue)
1513 destroy_workqueue(bs->rescue_workqueue);
1514 bs->rescue_workqueue = NULL;
1516 mempool_exit(&bs->bio_pool);
1517 mempool_exit(&bs->bvec_pool);
1519 bioset_integrity_free(bs);
1522 bs->bio_slab = NULL;
1524 EXPORT_SYMBOL(bioset_exit);
1527 * bioset_init - Initialize a bio_set
1528 * @bs: pool to initialize
1529 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1530 * @front_pad: Number of bytes to allocate in front of the returned bio
1531 * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS
1532 * and %BIOSET_NEED_RESCUER
1535 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
1536 * to ask for a number of bytes to be allocated in front of the bio.
1537 * Front pad allocation is useful for embedding the bio inside
1538 * another structure, to avoid allocating extra data to go with the bio.
1539 * Note that the bio must be embedded at the END of that structure always,
1540 * or things will break badly.
1541 * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated
1542 * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast().
1543 * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to
1544 * dispatch queued requests when the mempool runs out of space.
1547 int bioset_init(struct bio_set *bs,
1548 unsigned int pool_size,
1549 unsigned int front_pad,
1552 bs->front_pad = front_pad;
1553 if (flags & BIOSET_NEED_BVECS)
1554 bs->back_pad = BIO_INLINE_VECS * sizeof(struct bio_vec);
1558 spin_lock_init(&bs->rescue_lock);
1559 bio_list_init(&bs->rescue_list);
1560 INIT_WORK(&bs->rescue_work, bio_alloc_rescue);
1562 bs->bio_slab = bio_find_or_create_slab(bs);
1566 if (mempool_init_slab_pool(&bs->bio_pool, pool_size, bs->bio_slab))
1569 if ((flags & BIOSET_NEED_BVECS) &&
1570 biovec_init_pool(&bs->bvec_pool, pool_size))
1573 if (!(flags & BIOSET_NEED_RESCUER))
1576 bs->rescue_workqueue = alloc_workqueue("bioset", WQ_MEM_RECLAIM, 0);
1577 if (!bs->rescue_workqueue)
1585 EXPORT_SYMBOL(bioset_init);
1588 * Initialize and setup a new bio_set, based on the settings from
1591 int bioset_init_from_src(struct bio_set *bs, struct bio_set *src)
1596 if (src->bvec_pool.min_nr)
1597 flags |= BIOSET_NEED_BVECS;
1598 if (src->rescue_workqueue)
1599 flags |= BIOSET_NEED_RESCUER;
1601 return bioset_init(bs, src->bio_pool.min_nr, src->front_pad, flags);
1603 EXPORT_SYMBOL(bioset_init_from_src);
1605 static int __init init_bio(void)
1609 BUILD_BUG_ON(BIO_FLAG_LAST > BVEC_POOL_OFFSET);
1611 bio_integrity_init();
1613 for (i = 0; i < ARRAY_SIZE(bvec_slabs); i++) {
1614 struct biovec_slab *bvs = bvec_slabs + i;
1616 bvs->slab = kmem_cache_create(bvs->name,
1617 bvs->nr_vecs * sizeof(struct bio_vec), 0,
1618 SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
1621 if (bioset_init(&fs_bio_set, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS))
1622 panic("bio: can't allocate bios\n");
1624 if (bioset_integrity_create(&fs_bio_set, BIO_POOL_SIZE))
1625 panic("bio: can't create integrity pool\n");
1629 subsys_initcall(init_bio);