]>
git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - block/bio.c
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>
36 * Test patch to inline a certain number of bi_io_vec's inside the bio
37 * itself, to shrink a bio data allocation from two mempool calls to one
39 #define BIO_INLINE_VECS 4
42 * if you change this list, also change bvec_alloc or things will
43 * break badly! cannot be bigger than what you can fit into an
46 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
47 static struct biovec_slab bvec_slabs
[BVEC_POOL_NR
] __read_mostly
= {
48 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES
),
53 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
54 * IO code that does not need private memory pools.
56 struct bio_set
*fs_bio_set
;
57 EXPORT_SYMBOL(fs_bio_set
);
60 * Our slab pool management
63 struct kmem_cache
*slab
;
64 unsigned int slab_ref
;
65 unsigned int slab_size
;
68 static DEFINE_MUTEX(bio_slab_lock
);
69 static struct bio_slab
*bio_slabs
;
70 static unsigned int bio_slab_nr
, bio_slab_max
;
72 static struct kmem_cache
*bio_find_or_create_slab(unsigned int extra_size
)
74 unsigned int sz
= sizeof(struct bio
) + extra_size
;
75 struct kmem_cache
*slab
= NULL
;
76 struct bio_slab
*bslab
, *new_bio_slabs
;
77 unsigned int new_bio_slab_max
;
78 unsigned int i
, entry
= -1;
80 mutex_lock(&bio_slab_lock
);
83 while (i
< bio_slab_nr
) {
84 bslab
= &bio_slabs
[i
];
86 if (!bslab
->slab
&& entry
== -1)
88 else if (bslab
->slab_size
== sz
) {
99 if (bio_slab_nr
== bio_slab_max
&& entry
== -1) {
100 new_bio_slab_max
= bio_slab_max
<< 1;
101 new_bio_slabs
= krealloc(bio_slabs
,
102 new_bio_slab_max
* sizeof(struct bio_slab
),
106 bio_slab_max
= new_bio_slab_max
;
107 bio_slabs
= new_bio_slabs
;
110 entry
= bio_slab_nr
++;
112 bslab
= &bio_slabs
[entry
];
114 snprintf(bslab
->name
, sizeof(bslab
->name
), "bio-%d", entry
);
115 slab
= kmem_cache_create(bslab
->name
, sz
, ARCH_KMALLOC_MINALIGN
,
116 SLAB_HWCACHE_ALIGN
, NULL
);
122 bslab
->slab_size
= sz
;
124 mutex_unlock(&bio_slab_lock
);
128 static void bio_put_slab(struct bio_set
*bs
)
130 struct bio_slab
*bslab
= NULL
;
133 mutex_lock(&bio_slab_lock
);
135 for (i
= 0; i
< bio_slab_nr
; i
++) {
136 if (bs
->bio_slab
== bio_slabs
[i
].slab
) {
137 bslab
= &bio_slabs
[i
];
142 if (WARN(!bslab
, KERN_ERR
"bio: unable to find slab!\n"))
145 WARN_ON(!bslab
->slab_ref
);
147 if (--bslab
->slab_ref
)
150 kmem_cache_destroy(bslab
->slab
);
154 mutex_unlock(&bio_slab_lock
);
157 unsigned int bvec_nr_vecs(unsigned short idx
)
159 return bvec_slabs
[idx
].nr_vecs
;
162 void bvec_free(mempool_t
*pool
, struct bio_vec
*bv
, unsigned int idx
)
168 BIO_BUG_ON(idx
>= BVEC_POOL_NR
);
170 if (idx
== BVEC_POOL_MAX
) {
171 mempool_free(bv
, pool
);
173 struct biovec_slab
*bvs
= bvec_slabs
+ idx
;
175 kmem_cache_free(bvs
->slab
, bv
);
179 struct bio_vec
*bvec_alloc(gfp_t gfp_mask
, int nr
, unsigned long *idx
,
185 * see comment near bvec_array define!
203 case 129 ... BIO_MAX_PAGES
:
211 * idx now points to the pool we want to allocate from. only the
212 * 1-vec entry pool is mempool backed.
214 if (*idx
== BVEC_POOL_MAX
) {
216 bvl
= mempool_alloc(pool
, gfp_mask
);
218 struct biovec_slab
*bvs
= bvec_slabs
+ *idx
;
219 gfp_t __gfp_mask
= gfp_mask
& ~(__GFP_DIRECT_RECLAIM
| __GFP_IO
);
222 * Make this allocation restricted and don't dump info on
223 * allocation failures, since we'll fallback to the mempool
224 * in case of failure.
226 __gfp_mask
|= __GFP_NOMEMALLOC
| __GFP_NORETRY
| __GFP_NOWARN
;
229 * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM
230 * is set, retry with the 1-entry mempool
232 bvl
= kmem_cache_alloc(bvs
->slab
, __gfp_mask
);
233 if (unlikely(!bvl
&& (gfp_mask
& __GFP_DIRECT_RECLAIM
))) {
234 *idx
= BVEC_POOL_MAX
;
243 void bio_uninit(struct bio
*bio
)
245 bio_disassociate_task(bio
);
247 if (bio_integrity(bio
))
248 bio_integrity_free(bio
);
250 EXPORT_SYMBOL(bio_uninit
);
252 static void bio_free(struct bio
*bio
)
254 struct bio_set
*bs
= bio
->bi_pool
;
260 bvec_free(bs
->bvec_pool
, bio
->bi_io_vec
, BVEC_POOL_IDX(bio
));
263 * If we have front padding, adjust the bio pointer before freeing
268 mempool_free(p
, bs
->bio_pool
);
270 /* Bio was allocated by bio_kmalloc() */
276 * Users of this function have their own bio allocation. Subsequently,
277 * they must remember to pair any call to bio_init() with bio_uninit()
278 * when IO has completed, or when the bio is released.
280 void bio_init(struct bio
*bio
, struct bio_vec
*table
,
281 unsigned short max_vecs
)
283 memset(bio
, 0, sizeof(*bio
));
284 atomic_set(&bio
->__bi_remaining
, 1);
285 atomic_set(&bio
->__bi_cnt
, 1);
287 bio
->bi_io_vec
= table
;
288 bio
->bi_max_vecs
= max_vecs
;
290 EXPORT_SYMBOL(bio_init
);
293 * bio_reset - reinitialize a bio
297 * After calling bio_reset(), @bio will be in the same state as a freshly
298 * allocated bio returned bio bio_alloc_bioset() - the only fields that are
299 * preserved are the ones that are initialized by bio_alloc_bioset(). See
300 * comment in struct bio.
302 void bio_reset(struct bio
*bio
)
304 unsigned long flags
= bio
->bi_flags
& (~0UL << BIO_RESET_BITS
);
308 memset(bio
, 0, BIO_RESET_BYTES
);
309 bio
->bi_flags
= flags
;
310 atomic_set(&bio
->__bi_remaining
, 1);
312 EXPORT_SYMBOL(bio_reset
);
314 static struct bio
*__bio_chain_endio(struct bio
*bio
)
316 struct bio
*parent
= bio
->bi_private
;
318 if (!parent
->bi_status
)
319 parent
->bi_status
= bio
->bi_status
;
324 static void bio_chain_endio(struct bio
*bio
)
326 bio_endio(__bio_chain_endio(bio
));
330 * bio_chain - chain bio completions
331 * @bio: the target bio
332 * @parent: the @bio's parent bio
334 * The caller won't have a bi_end_io called when @bio completes - instead,
335 * @parent's bi_end_io won't be called until both @parent and @bio have
336 * completed; the chained bio will also be freed when it completes.
338 * The caller must not set bi_private or bi_end_io in @bio.
340 void bio_chain(struct bio
*bio
, struct bio
*parent
)
342 BUG_ON(bio
->bi_private
|| bio
->bi_end_io
);
344 bio
->bi_private
= parent
;
345 bio
->bi_end_io
= bio_chain_endio
;
346 bio_inc_remaining(parent
);
348 EXPORT_SYMBOL(bio_chain
);
350 static void bio_alloc_rescue(struct work_struct
*work
)
352 struct bio_set
*bs
= container_of(work
, struct bio_set
, rescue_work
);
356 spin_lock(&bs
->rescue_lock
);
357 bio
= bio_list_pop(&bs
->rescue_list
);
358 spin_unlock(&bs
->rescue_lock
);
363 generic_make_request(bio
);
367 static void punt_bios_to_rescuer(struct bio_set
*bs
)
369 struct bio_list punt
, nopunt
;
372 if (WARN_ON_ONCE(!bs
->rescue_workqueue
))
375 * In order to guarantee forward progress we must punt only bios that
376 * were allocated from this bio_set; otherwise, if there was a bio on
377 * there for a stacking driver higher up in the stack, processing it
378 * could require allocating bios from this bio_set, and doing that from
379 * our own rescuer would be bad.
381 * Since bio lists are singly linked, pop them all instead of trying to
382 * remove from the middle of the list:
385 bio_list_init(&punt
);
386 bio_list_init(&nopunt
);
388 while ((bio
= bio_list_pop(¤t
->bio_list
[0])))
389 bio_list_add(bio
->bi_pool
== bs
? &punt
: &nopunt
, bio
);
390 current
->bio_list
[0] = nopunt
;
392 bio_list_init(&nopunt
);
393 while ((bio
= bio_list_pop(¤t
->bio_list
[1])))
394 bio_list_add(bio
->bi_pool
== bs
? &punt
: &nopunt
, bio
);
395 current
->bio_list
[1] = nopunt
;
397 spin_lock(&bs
->rescue_lock
);
398 bio_list_merge(&bs
->rescue_list
, &punt
);
399 spin_unlock(&bs
->rescue_lock
);
401 queue_work(bs
->rescue_workqueue
, &bs
->rescue_work
);
405 * bio_alloc_bioset - allocate a bio for I/O
406 * @gfp_mask: the GFP_ mask given to the slab allocator
407 * @nr_iovecs: number of iovecs to pre-allocate
408 * @bs: the bio_set to allocate from.
411 * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is
412 * backed by the @bs's mempool.
414 * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will
415 * always be able to allocate a bio. This is due to the mempool guarantees.
416 * To make this work, callers must never allocate more than 1 bio at a time
417 * from this pool. Callers that need to allocate more than 1 bio must always
418 * submit the previously allocated bio for IO before attempting to allocate
419 * a new one. Failure to do so can cause deadlocks under memory pressure.
421 * Note that when running under generic_make_request() (i.e. any block
422 * driver), bios are not submitted until after you return - see the code in
423 * generic_make_request() that converts recursion into iteration, to prevent
426 * This would normally mean allocating multiple bios under
427 * generic_make_request() would be susceptible to deadlocks, but we have
428 * deadlock avoidance code that resubmits any blocked bios from a rescuer
431 * However, we do not guarantee forward progress for allocations from other
432 * mempools. Doing multiple allocations from the same mempool under
433 * generic_make_request() should be avoided - instead, use bio_set's front_pad
434 * for per bio allocations.
437 * Pointer to new bio on success, NULL on failure.
439 struct bio
*bio_alloc_bioset(gfp_t gfp_mask
, unsigned int nr_iovecs
,
442 gfp_t saved_gfp
= gfp_mask
;
444 unsigned inline_vecs
;
445 struct bio_vec
*bvl
= NULL
;
450 if (nr_iovecs
> UIO_MAXIOV
)
453 p
= kmalloc(sizeof(struct bio
) +
454 nr_iovecs
* sizeof(struct bio_vec
),
457 inline_vecs
= nr_iovecs
;
459 /* should not use nobvec bioset for nr_iovecs > 0 */
460 if (WARN_ON_ONCE(!bs
->bvec_pool
&& nr_iovecs
> 0))
463 * generic_make_request() converts recursion to iteration; this
464 * means if we're running beneath it, any bios we allocate and
465 * submit will not be submitted (and thus freed) until after we
468 * This exposes us to a potential deadlock if we allocate
469 * multiple bios from the same bio_set() while running
470 * underneath generic_make_request(). If we were to allocate
471 * multiple bios (say a stacking block driver that was splitting
472 * bios), we would deadlock if we exhausted the mempool's
475 * We solve this, and guarantee forward progress, with a rescuer
476 * workqueue per bio_set. If we go to allocate and there are
477 * bios on current->bio_list, we first try the allocation
478 * without __GFP_DIRECT_RECLAIM; if that fails, we punt those
479 * bios we would be blocking to the rescuer workqueue before
480 * we retry with the original gfp_flags.
483 if (current
->bio_list
&&
484 (!bio_list_empty(¤t
->bio_list
[0]) ||
485 !bio_list_empty(¤t
->bio_list
[1])) &&
486 bs
->rescue_workqueue
)
487 gfp_mask
&= ~__GFP_DIRECT_RECLAIM
;
489 p
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
490 if (!p
&& gfp_mask
!= saved_gfp
) {
491 punt_bios_to_rescuer(bs
);
492 gfp_mask
= saved_gfp
;
493 p
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
496 front_pad
= bs
->front_pad
;
497 inline_vecs
= BIO_INLINE_VECS
;
504 bio_init(bio
, NULL
, 0);
506 if (nr_iovecs
> inline_vecs
) {
507 unsigned long idx
= 0;
509 bvl
= bvec_alloc(gfp_mask
, nr_iovecs
, &idx
, bs
->bvec_pool
);
510 if (!bvl
&& gfp_mask
!= saved_gfp
) {
511 punt_bios_to_rescuer(bs
);
512 gfp_mask
= saved_gfp
;
513 bvl
= bvec_alloc(gfp_mask
, nr_iovecs
, &idx
, bs
->bvec_pool
);
519 bio
->bi_flags
|= idx
<< BVEC_POOL_OFFSET
;
520 } else if (nr_iovecs
) {
521 bvl
= bio
->bi_inline_vecs
;
525 bio
->bi_max_vecs
= nr_iovecs
;
526 bio
->bi_io_vec
= bvl
;
530 mempool_free(p
, bs
->bio_pool
);
533 EXPORT_SYMBOL(bio_alloc_bioset
);
535 void zero_fill_bio(struct bio
*bio
)
539 struct bvec_iter iter
;
541 bio_for_each_segment(bv
, bio
, iter
) {
542 char *data
= bvec_kmap_irq(&bv
, &flags
);
543 memset(data
, 0, bv
.bv_len
);
544 flush_dcache_page(bv
.bv_page
);
545 bvec_kunmap_irq(data
, &flags
);
548 EXPORT_SYMBOL(zero_fill_bio
);
551 * bio_put - release a reference to a bio
552 * @bio: bio to release reference to
555 * Put a reference to a &struct bio, either one you have gotten with
556 * bio_alloc, bio_get or bio_clone_*. The last put of a bio will free it.
558 void bio_put(struct bio
*bio
)
560 if (!bio_flagged(bio
, BIO_REFFED
))
563 BIO_BUG_ON(!atomic_read(&bio
->__bi_cnt
));
568 if (atomic_dec_and_test(&bio
->__bi_cnt
))
572 EXPORT_SYMBOL(bio_put
);
574 inline int bio_phys_segments(struct request_queue
*q
, struct bio
*bio
)
576 if (unlikely(!bio_flagged(bio
, BIO_SEG_VALID
)))
577 blk_recount_segments(q
, bio
);
579 return bio
->bi_phys_segments
;
581 EXPORT_SYMBOL(bio_phys_segments
);
584 * __bio_clone_fast - clone a bio that shares the original bio's biovec
585 * @bio: destination bio
586 * @bio_src: bio to clone
588 * Clone a &bio. Caller will own the returned bio, but not
589 * the actual data it points to. Reference count of returned
592 * Caller must ensure that @bio_src is not freed before @bio.
594 void __bio_clone_fast(struct bio
*bio
, struct bio
*bio_src
)
596 BUG_ON(bio
->bi_pool
&& BVEC_POOL_IDX(bio
));
599 * most users will be overriding ->bi_bdev with a new target,
600 * so we don't set nor calculate new physical/hw segment counts here
602 bio
->bi_bdev
= bio_src
->bi_bdev
;
603 bio_set_flag(bio
, BIO_CLONED
);
604 bio
->bi_opf
= bio_src
->bi_opf
;
605 bio
->bi_write_hint
= bio_src
->bi_write_hint
;
606 bio
->bi_iter
= bio_src
->bi_iter
;
607 bio
->bi_io_vec
= bio_src
->bi_io_vec
;
609 bio_clone_blkcg_association(bio
, bio_src
);
611 EXPORT_SYMBOL(__bio_clone_fast
);
614 * bio_clone_fast - clone a bio that shares the original bio's biovec
616 * @gfp_mask: allocation priority
617 * @bs: bio_set to allocate from
619 * Like __bio_clone_fast, only also allocates the returned bio
621 struct bio
*bio_clone_fast(struct bio
*bio
, gfp_t gfp_mask
, struct bio_set
*bs
)
625 b
= bio_alloc_bioset(gfp_mask
, 0, bs
);
629 __bio_clone_fast(b
, bio
);
631 if (bio_integrity(bio
)) {
634 ret
= bio_integrity_clone(b
, bio
, gfp_mask
);
644 EXPORT_SYMBOL(bio_clone_fast
);
647 * bio_clone_bioset - clone a bio
648 * @bio_src: bio to clone
649 * @gfp_mask: allocation priority
650 * @bs: bio_set to allocate from
652 * Clone bio. Caller will own the returned bio, but not the actual data it
653 * points to. Reference count of returned bio will be one.
655 struct bio
*bio_clone_bioset(struct bio
*bio_src
, gfp_t gfp_mask
,
658 struct bvec_iter iter
;
663 * Pre immutable biovecs, __bio_clone() used to just do a memcpy from
664 * bio_src->bi_io_vec to bio->bi_io_vec.
666 * We can't do that anymore, because:
668 * - The point of cloning the biovec is to produce a bio with a biovec
669 * the caller can modify: bi_idx and bi_bvec_done should be 0.
671 * - The original bio could've had more than BIO_MAX_PAGES biovecs; if
672 * we tried to clone the whole thing bio_alloc_bioset() would fail.
673 * But the clone should succeed as long as the number of biovecs we
674 * actually need to allocate is fewer than BIO_MAX_PAGES.
676 * - Lastly, bi_vcnt should not be looked at or relied upon by code
677 * that does not own the bio - reason being drivers don't use it for
678 * iterating over the biovec anymore, so expecting it to be kept up
679 * to date (i.e. for clones that share the parent biovec) is just
680 * asking for trouble and would force extra work on
681 * __bio_clone_fast() anyways.
684 bio
= bio_alloc_bioset(gfp_mask
, bio_segments(bio_src
), bs
);
687 bio
->bi_bdev
= bio_src
->bi_bdev
;
688 bio
->bi_opf
= bio_src
->bi_opf
;
689 bio
->bi_write_hint
= bio_src
->bi_write_hint
;
690 bio
->bi_iter
.bi_sector
= bio_src
->bi_iter
.bi_sector
;
691 bio
->bi_iter
.bi_size
= bio_src
->bi_iter
.bi_size
;
693 switch (bio_op(bio
)) {
695 case REQ_OP_SECURE_ERASE
:
696 case REQ_OP_WRITE_ZEROES
:
698 case REQ_OP_WRITE_SAME
:
699 bio
->bi_io_vec
[bio
->bi_vcnt
++] = bio_src
->bi_io_vec
[0];
702 bio_for_each_segment(bv
, bio_src
, iter
)
703 bio
->bi_io_vec
[bio
->bi_vcnt
++] = bv
;
707 if (bio_integrity(bio_src
)) {
710 ret
= bio_integrity_clone(bio
, bio_src
, gfp_mask
);
717 bio_clone_blkcg_association(bio
, bio_src
);
721 EXPORT_SYMBOL(bio_clone_bioset
);
724 * bio_add_pc_page - attempt to add page to bio
725 * @q: the target queue
726 * @bio: destination bio
728 * @len: vec entry length
729 * @offset: vec entry offset
731 * Attempt to add a page to the bio_vec maplist. This can fail for a
732 * number of reasons, such as the bio being full or target block device
733 * limitations. The target block device must allow bio's up to PAGE_SIZE,
734 * so it is always possible to add a single page to an empty bio.
736 * This should only be used by REQ_PC bios.
738 int bio_add_pc_page(struct request_queue
*q
, struct bio
*bio
, struct page
739 *page
, unsigned int len
, unsigned int offset
)
741 int retried_segments
= 0;
742 struct bio_vec
*bvec
;
745 * cloned bio must not modify vec list
747 if (unlikely(bio_flagged(bio
, BIO_CLONED
)))
750 if (((bio
->bi_iter
.bi_size
+ len
) >> 9) > queue_max_hw_sectors(q
))
754 * For filesystems with a blocksize smaller than the pagesize
755 * we will often be called with the same page as last time and
756 * a consecutive offset. Optimize this special case.
758 if (bio
->bi_vcnt
> 0) {
759 struct bio_vec
*prev
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
761 if (page
== prev
->bv_page
&&
762 offset
== prev
->bv_offset
+ prev
->bv_len
) {
764 bio
->bi_iter
.bi_size
+= len
;
769 * If the queue doesn't support SG gaps and adding this
770 * offset would create a gap, disallow it.
772 if (bvec_gap_to_prev(q
, prev
, offset
))
776 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
780 * setup the new entry, we might clear it again later if we
781 * cannot add the page
783 bvec
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
784 bvec
->bv_page
= page
;
786 bvec
->bv_offset
= offset
;
788 bio
->bi_phys_segments
++;
789 bio
->bi_iter
.bi_size
+= len
;
792 * Perform a recount if the number of segments is greater
793 * than queue_max_segments(q).
796 while (bio
->bi_phys_segments
> queue_max_segments(q
)) {
798 if (retried_segments
)
801 retried_segments
= 1;
802 blk_recount_segments(q
, bio
);
805 /* If we may be able to merge these biovecs, force a recount */
806 if (bio
->bi_vcnt
> 1 && (BIOVEC_PHYS_MERGEABLE(bvec
-1, bvec
)))
807 bio_clear_flag(bio
, BIO_SEG_VALID
);
813 bvec
->bv_page
= NULL
;
817 bio
->bi_iter
.bi_size
-= len
;
818 blk_recount_segments(q
, bio
);
821 EXPORT_SYMBOL(bio_add_pc_page
);
824 * bio_add_page - attempt to add page to bio
825 * @bio: destination bio
827 * @len: vec entry length
828 * @offset: vec entry offset
830 * Attempt to add a page to the bio_vec maplist. This will only fail
831 * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio.
833 int bio_add_page(struct bio
*bio
, struct page
*page
,
834 unsigned int len
, unsigned int offset
)
839 * cloned bio must not modify vec list
841 if (WARN_ON_ONCE(bio_flagged(bio
, BIO_CLONED
)))
845 * For filesystems with a blocksize smaller than the pagesize
846 * we will often be called with the same page as last time and
847 * a consecutive offset. Optimize this special case.
849 if (bio
->bi_vcnt
> 0) {
850 bv
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
852 if (page
== bv
->bv_page
&&
853 offset
== bv
->bv_offset
+ bv
->bv_len
) {
859 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
862 bv
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
865 bv
->bv_offset
= offset
;
869 bio
->bi_iter
.bi_size
+= len
;
872 EXPORT_SYMBOL(bio_add_page
);
875 * bio_iov_iter_get_pages - pin user or kernel pages and add them to a bio
876 * @bio: bio to add pages to
877 * @iter: iov iterator describing the region to be mapped
879 * Pins as many pages from *iter and appends them to @bio's bvec array. The
880 * pages will have to be released using put_page() when done.
882 int bio_iov_iter_get_pages(struct bio
*bio
, struct iov_iter
*iter
)
884 unsigned short nr_pages
= bio
->bi_max_vecs
- bio
->bi_vcnt
;
885 struct bio_vec
*bv
= bio
->bi_io_vec
+ bio
->bi_vcnt
;
886 struct page
**pages
= (struct page
**)bv
;
890 size
= iov_iter_get_pages(iter
, pages
, LONG_MAX
, nr_pages
, &offset
);
891 if (unlikely(size
<= 0))
892 return size
? size
: -EFAULT
;
893 nr_pages
= (size
+ offset
+ PAGE_SIZE
- 1) / PAGE_SIZE
;
896 * Deep magic below: We need to walk the pinned pages backwards
897 * because we are abusing the space allocated for the bio_vecs
898 * for the page array. Because the bio_vecs are larger than the
899 * page pointers by definition this will always work. But it also
900 * means we can't use bio_add_page, so any changes to it's semantics
901 * need to be reflected here as well.
903 bio
->bi_iter
.bi_size
+= size
;
904 bio
->bi_vcnt
+= nr_pages
;
906 diff
= (nr_pages
* PAGE_SIZE
- offset
) - size
;
908 bv
[nr_pages
].bv_page
= pages
[nr_pages
];
909 bv
[nr_pages
].bv_len
= PAGE_SIZE
;
910 bv
[nr_pages
].bv_offset
= 0;
913 bv
[0].bv_offset
+= offset
;
914 bv
[0].bv_len
-= offset
;
916 bv
[bio
->bi_vcnt
- 1].bv_len
-= diff
;
918 iov_iter_advance(iter
, size
);
921 EXPORT_SYMBOL_GPL(bio_iov_iter_get_pages
);
923 struct submit_bio_ret
{
924 struct completion event
;
928 static void submit_bio_wait_endio(struct bio
*bio
)
930 struct submit_bio_ret
*ret
= bio
->bi_private
;
932 ret
->error
= blk_status_to_errno(bio
->bi_status
);
933 complete(&ret
->event
);
937 * submit_bio_wait - submit a bio, and wait until it completes
938 * @bio: The &struct bio which describes the I/O
940 * Simple wrapper around submit_bio(). Returns 0 on success, or the error from
941 * bio_endio() on failure.
943 int submit_bio_wait(struct bio
*bio
)
945 struct submit_bio_ret ret
;
947 init_completion(&ret
.event
);
948 bio
->bi_private
= &ret
;
949 bio
->bi_end_io
= submit_bio_wait_endio
;
950 bio
->bi_opf
|= REQ_SYNC
;
952 wait_for_completion_io(&ret
.event
);
956 EXPORT_SYMBOL(submit_bio_wait
);
959 * bio_advance - increment/complete a bio by some number of bytes
960 * @bio: bio to advance
961 * @bytes: number of bytes to complete
963 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
964 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
965 * be updated on the last bvec as well.
967 * @bio will then represent the remaining, uncompleted portion of the io.
969 void bio_advance(struct bio
*bio
, unsigned bytes
)
971 if (bio_integrity(bio
))
972 bio_integrity_advance(bio
, bytes
);
974 bio_advance_iter(bio
, &bio
->bi_iter
, bytes
);
976 EXPORT_SYMBOL(bio_advance
);
979 * bio_alloc_pages - allocates a single page for each bvec in a bio
980 * @bio: bio to allocate pages for
981 * @gfp_mask: flags for allocation
983 * Allocates pages up to @bio->bi_vcnt.
985 * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are
988 int bio_alloc_pages(struct bio
*bio
, gfp_t gfp_mask
)
993 bio_for_each_segment_all(bv
, bio
, i
) {
994 bv
->bv_page
= alloc_page(gfp_mask
);
996 while (--bv
>= bio
->bi_io_vec
)
997 __free_page(bv
->bv_page
);
1004 EXPORT_SYMBOL(bio_alloc_pages
);
1007 * bio_copy_data - copy contents of data buffers from one chain of bios to
1009 * @src: source bio list
1010 * @dst: destination bio list
1012 * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats
1013 * @src and @dst as linked lists of bios.
1015 * Stops when it reaches the end of either @src or @dst - that is, copies
1016 * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios).
1018 void bio_copy_data(struct bio
*dst
, struct bio
*src
)
1020 struct bvec_iter src_iter
, dst_iter
;
1021 struct bio_vec src_bv
, dst_bv
;
1022 void *src_p
, *dst_p
;
1025 src_iter
= src
->bi_iter
;
1026 dst_iter
= dst
->bi_iter
;
1029 if (!src_iter
.bi_size
) {
1034 src_iter
= src
->bi_iter
;
1037 if (!dst_iter
.bi_size
) {
1042 dst_iter
= dst
->bi_iter
;
1045 src_bv
= bio_iter_iovec(src
, src_iter
);
1046 dst_bv
= bio_iter_iovec(dst
, dst_iter
);
1048 bytes
= min(src_bv
.bv_len
, dst_bv
.bv_len
);
1050 src_p
= kmap_atomic(src_bv
.bv_page
);
1051 dst_p
= kmap_atomic(dst_bv
.bv_page
);
1053 memcpy(dst_p
+ dst_bv
.bv_offset
,
1054 src_p
+ src_bv
.bv_offset
,
1057 kunmap_atomic(dst_p
);
1058 kunmap_atomic(src_p
);
1060 bio_advance_iter(src
, &src_iter
, bytes
);
1061 bio_advance_iter(dst
, &dst_iter
, bytes
);
1064 EXPORT_SYMBOL(bio_copy_data
);
1066 struct bio_map_data
{
1068 struct iov_iter iter
;
1072 static struct bio_map_data
*bio_alloc_map_data(unsigned int iov_count
,
1075 if (iov_count
> UIO_MAXIOV
)
1078 return kmalloc(sizeof(struct bio_map_data
) +
1079 sizeof(struct iovec
) * iov_count
, gfp_mask
);
1083 * bio_copy_from_iter - copy all pages from iov_iter to bio
1084 * @bio: The &struct bio which describes the I/O as destination
1085 * @iter: iov_iter as source
1087 * Copy all pages from iov_iter to bio.
1088 * Returns 0 on success, or error on failure.
1090 static int bio_copy_from_iter(struct bio
*bio
, struct iov_iter iter
)
1093 struct bio_vec
*bvec
;
1095 bio_for_each_segment_all(bvec
, bio
, i
) {
1098 ret
= copy_page_from_iter(bvec
->bv_page
,
1103 if (!iov_iter_count(&iter
))
1106 if (ret
< bvec
->bv_len
)
1114 * bio_copy_to_iter - copy all pages from bio to iov_iter
1115 * @bio: The &struct bio which describes the I/O as source
1116 * @iter: iov_iter as destination
1118 * Copy all pages from bio to iov_iter.
1119 * Returns 0 on success, or error on failure.
1121 static int bio_copy_to_iter(struct bio
*bio
, struct iov_iter iter
)
1124 struct bio_vec
*bvec
;
1126 bio_for_each_segment_all(bvec
, bio
, i
) {
1129 ret
= copy_page_to_iter(bvec
->bv_page
,
1134 if (!iov_iter_count(&iter
))
1137 if (ret
< bvec
->bv_len
)
1144 void bio_free_pages(struct bio
*bio
)
1146 struct bio_vec
*bvec
;
1149 bio_for_each_segment_all(bvec
, bio
, i
)
1150 __free_page(bvec
->bv_page
);
1152 EXPORT_SYMBOL(bio_free_pages
);
1155 * bio_uncopy_user - finish previously mapped bio
1156 * @bio: bio being terminated
1158 * Free pages allocated from bio_copy_user_iov() and write back data
1159 * to user space in case of a read.
1161 int bio_uncopy_user(struct bio
*bio
)
1163 struct bio_map_data
*bmd
= bio
->bi_private
;
1166 if (!bio_flagged(bio
, BIO_NULL_MAPPED
)) {
1168 * if we're in a workqueue, the request is orphaned, so
1169 * don't copy into a random user address space, just free
1170 * and return -EINTR so user space doesn't expect any data.
1174 else if (bio_data_dir(bio
) == READ
)
1175 ret
= bio_copy_to_iter(bio
, bmd
->iter
);
1176 if (bmd
->is_our_pages
)
1177 bio_free_pages(bio
);
1185 * bio_copy_user_iov - copy user data to bio
1186 * @q: destination block queue
1187 * @map_data: pointer to the rq_map_data holding pages (if necessary)
1188 * @iter: iovec iterator
1189 * @gfp_mask: memory allocation flags
1191 * Prepares and returns a bio for indirect user io, bouncing data
1192 * to/from kernel pages as necessary. Must be paired with
1193 * call bio_uncopy_user() on io completion.
1195 struct bio
*bio_copy_user_iov(struct request_queue
*q
,
1196 struct rq_map_data
*map_data
,
1197 const struct iov_iter
*iter
,
1200 struct bio_map_data
*bmd
;
1205 unsigned int len
= iter
->count
;
1206 unsigned int offset
= map_data
? offset_in_page(map_data
->offset
) : 0;
1208 for (i
= 0; i
< iter
->nr_segs
; i
++) {
1209 unsigned long uaddr
;
1211 unsigned long start
;
1213 uaddr
= (unsigned long) iter
->iov
[i
].iov_base
;
1214 end
= (uaddr
+ iter
->iov
[i
].iov_len
+ PAGE_SIZE
- 1)
1216 start
= uaddr
>> PAGE_SHIFT
;
1222 return ERR_PTR(-EINVAL
);
1224 nr_pages
+= end
- start
;
1230 bmd
= bio_alloc_map_data(iter
->nr_segs
, gfp_mask
);
1232 return ERR_PTR(-ENOMEM
);
1235 * We need to do a deep copy of the iov_iter including the iovecs.
1236 * The caller provided iov might point to an on-stack or otherwise
1239 bmd
->is_our_pages
= map_data
? 0 : 1;
1240 memcpy(bmd
->iov
, iter
->iov
, sizeof(struct iovec
) * iter
->nr_segs
);
1241 iov_iter_init(&bmd
->iter
, iter
->type
, bmd
->iov
,
1242 iter
->nr_segs
, iter
->count
);
1245 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1252 nr_pages
= 1 << map_data
->page_order
;
1253 i
= map_data
->offset
/ PAGE_SIZE
;
1256 unsigned int bytes
= PAGE_SIZE
;
1264 if (i
== map_data
->nr_entries
* nr_pages
) {
1269 page
= map_data
->pages
[i
/ nr_pages
];
1270 page
+= (i
% nr_pages
);
1274 page
= alloc_page(q
->bounce_gfp
| gfp_mask
);
1281 if (bio_add_pc_page(q
, bio
, page
, bytes
, offset
) < bytes
)
1294 if (((iter
->type
& WRITE
) && (!map_data
|| !map_data
->null_mapped
)) ||
1295 (map_data
&& map_data
->from_user
)) {
1296 ret
= bio_copy_from_iter(bio
, *iter
);
1301 bio
->bi_private
= bmd
;
1305 bio_free_pages(bio
);
1309 return ERR_PTR(ret
);
1313 * bio_map_user_iov - map user iovec into bio
1314 * @q: the struct request_queue for the bio
1315 * @iter: iovec iterator
1316 * @gfp_mask: memory allocation flags
1318 * Map the user space address into a bio suitable for io to a block
1319 * device. Returns an error pointer in case of error.
1321 struct bio
*bio_map_user_iov(struct request_queue
*q
,
1322 const struct iov_iter
*iter
,
1327 struct page
**pages
;
1334 iov_for_each(iov
, i
, *iter
) {
1335 unsigned long uaddr
= (unsigned long) iov
.iov_base
;
1336 unsigned long len
= iov
.iov_len
;
1337 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1338 unsigned long start
= uaddr
>> PAGE_SHIFT
;
1344 return ERR_PTR(-EINVAL
);
1346 nr_pages
+= end
- start
;
1348 * buffer must be aligned to at least logical block size for now
1350 if (uaddr
& queue_dma_alignment(q
))
1351 return ERR_PTR(-EINVAL
);
1355 return ERR_PTR(-EINVAL
);
1357 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1359 return ERR_PTR(-ENOMEM
);
1362 pages
= kcalloc(nr_pages
, sizeof(struct page
*), gfp_mask
);
1366 iov_for_each(iov
, i
, *iter
) {
1367 unsigned long uaddr
= (unsigned long) iov
.iov_base
;
1368 unsigned long len
= iov
.iov_len
;
1369 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1370 unsigned long start
= uaddr
>> PAGE_SHIFT
;
1371 const int local_nr_pages
= end
- start
;
1372 const int page_limit
= cur_page
+ local_nr_pages
;
1374 ret
= get_user_pages_fast(uaddr
, local_nr_pages
,
1375 (iter
->type
& WRITE
) != WRITE
,
1377 if (ret
< local_nr_pages
) {
1382 offset
= offset_in_page(uaddr
);
1383 for (j
= cur_page
; j
< page_limit
; j
++) {
1384 unsigned int bytes
= PAGE_SIZE
- offset
;
1395 if (bio_add_pc_page(q
, bio
, pages
[j
], bytes
, offset
) <
1405 * release the pages we didn't map into the bio, if any
1407 while (j
< page_limit
)
1408 put_page(pages
[j
++]);
1413 bio_set_flag(bio
, BIO_USER_MAPPED
);
1416 * subtle -- if bio_map_user_iov() ended up bouncing a bio,
1417 * it would normally disappear when its bi_end_io is run.
1418 * however, we need it for the unmap, so grab an extra
1425 for (j
= 0; j
< nr_pages
; j
++) {
1433 return ERR_PTR(ret
);
1436 static void __bio_unmap_user(struct bio
*bio
)
1438 struct bio_vec
*bvec
;
1442 * make sure we dirty pages we wrote to
1444 bio_for_each_segment_all(bvec
, bio
, i
) {
1445 if (bio_data_dir(bio
) == READ
)
1446 set_page_dirty_lock(bvec
->bv_page
);
1448 put_page(bvec
->bv_page
);
1455 * bio_unmap_user - unmap a bio
1456 * @bio: the bio being unmapped
1458 * Unmap a bio previously mapped by bio_map_user_iov(). Must be called from
1461 * bio_unmap_user() may sleep.
1463 void bio_unmap_user(struct bio
*bio
)
1465 __bio_unmap_user(bio
);
1469 static void bio_map_kern_endio(struct bio
*bio
)
1475 * bio_map_kern - map kernel address into bio
1476 * @q: the struct request_queue for the bio
1477 * @data: pointer to buffer to map
1478 * @len: length in bytes
1479 * @gfp_mask: allocation flags for bio allocation
1481 * Map the kernel address into a bio suitable for io to a block
1482 * device. Returns an error pointer in case of error.
1484 struct bio
*bio_map_kern(struct request_queue
*q
, void *data
, unsigned int len
,
1487 unsigned long kaddr
= (unsigned long)data
;
1488 unsigned long end
= (kaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1489 unsigned long start
= kaddr
>> PAGE_SHIFT
;
1490 const int nr_pages
= end
- start
;
1494 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1496 return ERR_PTR(-ENOMEM
);
1498 offset
= offset_in_page(kaddr
);
1499 for (i
= 0; i
< nr_pages
; i
++) {
1500 unsigned int bytes
= PAGE_SIZE
- offset
;
1508 if (bio_add_pc_page(q
, bio
, virt_to_page(data
), bytes
,
1510 /* we don't support partial mappings */
1512 return ERR_PTR(-EINVAL
);
1520 bio
->bi_end_io
= bio_map_kern_endio
;
1523 EXPORT_SYMBOL(bio_map_kern
);
1525 static void bio_copy_kern_endio(struct bio
*bio
)
1527 bio_free_pages(bio
);
1531 static void bio_copy_kern_endio_read(struct bio
*bio
)
1533 char *p
= bio
->bi_private
;
1534 struct bio_vec
*bvec
;
1537 bio_for_each_segment_all(bvec
, bio
, i
) {
1538 memcpy(p
, page_address(bvec
->bv_page
), bvec
->bv_len
);
1542 bio_copy_kern_endio(bio
);
1546 * bio_copy_kern - copy kernel address into bio
1547 * @q: the struct request_queue for the bio
1548 * @data: pointer to buffer to copy
1549 * @len: length in bytes
1550 * @gfp_mask: allocation flags for bio and page allocation
1551 * @reading: data direction is READ
1553 * copy the kernel address into a bio suitable for io to a block
1554 * device. Returns an error pointer in case of error.
1556 struct bio
*bio_copy_kern(struct request_queue
*q
, void *data
, unsigned int len
,
1557 gfp_t gfp_mask
, int reading
)
1559 unsigned long kaddr
= (unsigned long)data
;
1560 unsigned long end
= (kaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1561 unsigned long start
= kaddr
>> PAGE_SHIFT
;
1570 return ERR_PTR(-EINVAL
);
1572 nr_pages
= end
- start
;
1573 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1575 return ERR_PTR(-ENOMEM
);
1579 unsigned int bytes
= PAGE_SIZE
;
1584 page
= alloc_page(q
->bounce_gfp
| gfp_mask
);
1589 memcpy(page_address(page
), p
, bytes
);
1591 if (bio_add_pc_page(q
, bio
, page
, bytes
, 0) < bytes
)
1599 bio
->bi_end_io
= bio_copy_kern_endio_read
;
1600 bio
->bi_private
= data
;
1602 bio
->bi_end_io
= bio_copy_kern_endio
;
1608 bio_free_pages(bio
);
1610 return ERR_PTR(-ENOMEM
);
1614 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1615 * for performing direct-IO in BIOs.
1617 * The problem is that we cannot run set_page_dirty() from interrupt context
1618 * because the required locks are not interrupt-safe. So what we can do is to
1619 * mark the pages dirty _before_ performing IO. And in interrupt context,
1620 * check that the pages are still dirty. If so, fine. If not, redirty them
1621 * in process context.
1623 * We special-case compound pages here: normally this means reads into hugetlb
1624 * pages. The logic in here doesn't really work right for compound pages
1625 * because the VM does not uniformly chase down the head page in all cases.
1626 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1627 * handle them at all. So we skip compound pages here at an early stage.
1629 * Note that this code is very hard to test under normal circumstances because
1630 * direct-io pins the pages with get_user_pages(). This makes
1631 * is_page_cache_freeable return false, and the VM will not clean the pages.
1632 * But other code (eg, flusher threads) could clean the pages if they are mapped
1635 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1636 * deferred bio dirtying paths.
1640 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1642 void bio_set_pages_dirty(struct bio
*bio
)
1644 struct bio_vec
*bvec
;
1647 bio_for_each_segment_all(bvec
, bio
, i
) {
1648 struct page
*page
= bvec
->bv_page
;
1650 if (page
&& !PageCompound(page
))
1651 set_page_dirty_lock(page
);
1655 static void bio_release_pages(struct bio
*bio
)
1657 struct bio_vec
*bvec
;
1660 bio_for_each_segment_all(bvec
, bio
, i
) {
1661 struct page
*page
= bvec
->bv_page
;
1669 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1670 * If they are, then fine. If, however, some pages are clean then they must
1671 * have been written out during the direct-IO read. So we take another ref on
1672 * the BIO and the offending pages and re-dirty the pages in process context.
1674 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1675 * here on. It will run one put_page() against each page and will run one
1676 * bio_put() against the BIO.
1679 static void bio_dirty_fn(struct work_struct
*work
);
1681 static DECLARE_WORK(bio_dirty_work
, bio_dirty_fn
);
1682 static DEFINE_SPINLOCK(bio_dirty_lock
);
1683 static struct bio
*bio_dirty_list
;
1686 * This runs in process context
1688 static void bio_dirty_fn(struct work_struct
*work
)
1690 unsigned long flags
;
1693 spin_lock_irqsave(&bio_dirty_lock
, flags
);
1694 bio
= bio_dirty_list
;
1695 bio_dirty_list
= NULL
;
1696 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
1699 struct bio
*next
= bio
->bi_private
;
1701 bio_set_pages_dirty(bio
);
1702 bio_release_pages(bio
);
1708 void bio_check_pages_dirty(struct bio
*bio
)
1710 struct bio_vec
*bvec
;
1711 int nr_clean_pages
= 0;
1714 bio_for_each_segment_all(bvec
, bio
, i
) {
1715 struct page
*page
= bvec
->bv_page
;
1717 if (PageDirty(page
) || PageCompound(page
)) {
1719 bvec
->bv_page
= NULL
;
1725 if (nr_clean_pages
) {
1726 unsigned long flags
;
1728 spin_lock_irqsave(&bio_dirty_lock
, flags
);
1729 bio
->bi_private
= bio_dirty_list
;
1730 bio_dirty_list
= bio
;
1731 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
1732 schedule_work(&bio_dirty_work
);
1738 void generic_start_io_acct(int rw
, unsigned long sectors
,
1739 struct hd_struct
*part
)
1741 int cpu
= part_stat_lock();
1743 part_round_stats(cpu
, part
);
1744 part_stat_inc(cpu
, part
, ios
[rw
]);
1745 part_stat_add(cpu
, part
, sectors
[rw
], sectors
);
1746 part_inc_in_flight(part
, rw
);
1750 EXPORT_SYMBOL(generic_start_io_acct
);
1752 void generic_end_io_acct(int rw
, struct hd_struct
*part
,
1753 unsigned long start_time
)
1755 unsigned long duration
= jiffies
- start_time
;
1756 int cpu
= part_stat_lock();
1758 part_stat_add(cpu
, part
, ticks
[rw
], duration
);
1759 part_round_stats(cpu
, part
);
1760 part_dec_in_flight(part
, rw
);
1764 EXPORT_SYMBOL(generic_end_io_acct
);
1766 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1767 void bio_flush_dcache_pages(struct bio
*bi
)
1769 struct bio_vec bvec
;
1770 struct bvec_iter iter
;
1772 bio_for_each_segment(bvec
, bi
, iter
)
1773 flush_dcache_page(bvec
.bv_page
);
1775 EXPORT_SYMBOL(bio_flush_dcache_pages
);
1778 static inline bool bio_remaining_done(struct bio
*bio
)
1781 * If we're not chaining, then ->__bi_remaining is always 1 and
1782 * we always end io on the first invocation.
1784 if (!bio_flagged(bio
, BIO_CHAIN
))
1787 BUG_ON(atomic_read(&bio
->__bi_remaining
) <= 0);
1789 if (atomic_dec_and_test(&bio
->__bi_remaining
)) {
1790 bio_clear_flag(bio
, BIO_CHAIN
);
1798 * bio_endio - end I/O on a bio
1802 * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred
1803 * way to end I/O on a bio. No one should call bi_end_io() directly on a
1804 * bio unless they own it and thus know that it has an end_io function.
1806 * bio_endio() can be called several times on a bio that has been chained
1807 * using bio_chain(). The ->bi_end_io() function will only be called the
1808 * last time. At this point the BLK_TA_COMPLETE tracing event will be
1809 * generated if BIO_TRACE_COMPLETION is set.
1811 void bio_endio(struct bio
*bio
)
1814 if (!bio_remaining_done(bio
))
1818 * Need to have a real endio function for chained bios, otherwise
1819 * various corner cases will break (like stacking block devices that
1820 * save/restore bi_end_io) - however, we want to avoid unbounded
1821 * recursion and blowing the stack. Tail call optimization would
1822 * handle this, but compiling with frame pointers also disables
1823 * gcc's sibling call optimization.
1825 if (bio
->bi_end_io
== bio_chain_endio
) {
1826 bio
= __bio_chain_endio(bio
);
1830 if (bio
->bi_bdev
&& bio_flagged(bio
, BIO_TRACE_COMPLETION
)) {
1831 trace_block_bio_complete(bdev_get_queue(bio
->bi_bdev
), bio
,
1832 blk_status_to_errno(bio
->bi_status
));
1833 bio_clear_flag(bio
, BIO_TRACE_COMPLETION
);
1836 blk_throtl_bio_endio(bio
);
1838 bio
->bi_end_io(bio
);
1840 EXPORT_SYMBOL(bio_endio
);
1843 * bio_split - split a bio
1844 * @bio: bio to split
1845 * @sectors: number of sectors to split from the front of @bio
1847 * @bs: bio set to allocate from
1849 * Allocates and returns a new bio which represents @sectors from the start of
1850 * @bio, and updates @bio to represent the remaining sectors.
1852 * Unless this is a discard request the newly allocated bio will point
1853 * to @bio's bi_io_vec; it is the caller's responsibility to ensure that
1854 * @bio is not freed before the split.
1856 struct bio
*bio_split(struct bio
*bio
, int sectors
,
1857 gfp_t gfp
, struct bio_set
*bs
)
1859 struct bio
*split
= NULL
;
1861 BUG_ON(sectors
<= 0);
1862 BUG_ON(sectors
>= bio_sectors(bio
));
1864 split
= bio_clone_fast(bio
, gfp
, bs
);
1868 split
->bi_iter
.bi_size
= sectors
<< 9;
1870 if (bio_integrity(split
))
1871 bio_integrity_trim(split
, 0, sectors
);
1873 bio_advance(bio
, split
->bi_iter
.bi_size
);
1875 if (bio_flagged(bio
, BIO_TRACE_COMPLETION
))
1876 bio_set_flag(bio
, BIO_TRACE_COMPLETION
);
1880 EXPORT_SYMBOL(bio_split
);
1883 * bio_trim - trim a bio
1885 * @offset: number of sectors to trim from the front of @bio
1886 * @size: size we want to trim @bio to, in sectors
1888 void bio_trim(struct bio
*bio
, int offset
, int size
)
1890 /* 'bio' is a cloned bio which we need to trim to match
1891 * the given offset and size.
1895 if (offset
== 0 && size
== bio
->bi_iter
.bi_size
)
1898 bio_clear_flag(bio
, BIO_SEG_VALID
);
1900 bio_advance(bio
, offset
<< 9);
1902 bio
->bi_iter
.bi_size
= size
;
1904 EXPORT_SYMBOL_GPL(bio_trim
);
1907 * create memory pools for biovec's in a bio_set.
1908 * use the global biovec slabs created for general use.
1910 mempool_t
*biovec_create_pool(int pool_entries
)
1912 struct biovec_slab
*bp
= bvec_slabs
+ BVEC_POOL_MAX
;
1914 return mempool_create_slab_pool(pool_entries
, bp
->slab
);
1917 void bioset_free(struct bio_set
*bs
)
1919 if (bs
->rescue_workqueue
)
1920 destroy_workqueue(bs
->rescue_workqueue
);
1923 mempool_destroy(bs
->bio_pool
);
1926 mempool_destroy(bs
->bvec_pool
);
1928 bioset_integrity_free(bs
);
1933 EXPORT_SYMBOL(bioset_free
);
1936 * bioset_create - Create a bio_set
1937 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1938 * @front_pad: Number of bytes to allocate in front of the returned bio
1939 * @flags: Flags to modify behavior, currently %BIOSET_NEED_BVECS
1940 * and %BIOSET_NEED_RESCUER
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.
1949 * If %BIOSET_NEED_BVECS is set in @flags, a separate pool will be allocated
1950 * for allocating iovecs. This pool is not needed e.g. for bio_clone_fast().
1951 * If %BIOSET_NEED_RESCUER is set, a workqueue is created which can be used to
1952 * dispatch queued requests when the mempool runs out of space.
1955 struct bio_set
*bioset_create(unsigned int pool_size
,
1956 unsigned int front_pad
,
1959 unsigned int back_pad
= BIO_INLINE_VECS
* sizeof(struct bio_vec
);
1962 bs
= kzalloc(sizeof(*bs
), GFP_KERNEL
);
1966 bs
->front_pad
= front_pad
;
1968 spin_lock_init(&bs
->rescue_lock
);
1969 bio_list_init(&bs
->rescue_list
);
1970 INIT_WORK(&bs
->rescue_work
, bio_alloc_rescue
);
1972 bs
->bio_slab
= bio_find_or_create_slab(front_pad
+ back_pad
);
1973 if (!bs
->bio_slab
) {
1978 bs
->bio_pool
= mempool_create_slab_pool(pool_size
, bs
->bio_slab
);
1982 if (flags
& BIOSET_NEED_BVECS
) {
1983 bs
->bvec_pool
= biovec_create_pool(pool_size
);
1988 if (!(flags
& BIOSET_NEED_RESCUER
))
1991 bs
->rescue_workqueue
= alloc_workqueue("bioset", WQ_MEM_RECLAIM
, 0);
1992 if (!bs
->rescue_workqueue
)
2000 EXPORT_SYMBOL(bioset_create
);
2002 #ifdef CONFIG_BLK_CGROUP
2005 * bio_associate_blkcg - associate a bio with the specified blkcg
2007 * @blkcg_css: css of the blkcg to associate
2009 * Associate @bio with the blkcg specified by @blkcg_css. Block layer will
2010 * treat @bio as if it were issued by a task which belongs to the blkcg.
2012 * This function takes an extra reference of @blkcg_css which will be put
2013 * when @bio is released. The caller must own @bio and is responsible for
2014 * synchronizing calls to this function.
2016 int bio_associate_blkcg(struct bio
*bio
, struct cgroup_subsys_state
*blkcg_css
)
2018 if (unlikely(bio
->bi_css
))
2021 bio
->bi_css
= blkcg_css
;
2024 EXPORT_SYMBOL_GPL(bio_associate_blkcg
);
2027 * bio_associate_current - associate a bio with %current
2030 * Associate @bio with %current if it hasn't been associated yet. Block
2031 * layer will treat @bio as if it were issued by %current no matter which
2032 * task actually issues it.
2034 * This function takes an extra reference of @task's io_context and blkcg
2035 * which will be put when @bio is released. The caller must own @bio,
2036 * ensure %current->io_context exists, and is responsible for synchronizing
2037 * calls to this function.
2039 int bio_associate_current(struct bio
*bio
)
2041 struct io_context
*ioc
;
2046 ioc
= current
->io_context
;
2050 get_io_context_active(ioc
);
2052 bio
->bi_css
= task_get_css(current
, io_cgrp_id
);
2055 EXPORT_SYMBOL_GPL(bio_associate_current
);
2058 * bio_disassociate_task - undo bio_associate_current()
2061 void bio_disassociate_task(struct bio
*bio
)
2064 put_io_context(bio
->bi_ioc
);
2068 css_put(bio
->bi_css
);
2074 * bio_clone_blkcg_association - clone blkcg association from src to dst bio
2075 * @dst: destination bio
2078 void bio_clone_blkcg_association(struct bio
*dst
, struct bio
*src
)
2081 WARN_ON(bio_associate_blkcg(dst
, src
->bi_css
));
2084 #endif /* CONFIG_BLK_CGROUP */
2086 static void __init
biovec_init_slabs(void)
2090 for (i
= 0; i
< BVEC_POOL_NR
; i
++) {
2092 struct biovec_slab
*bvs
= bvec_slabs
+ i
;
2094 if (bvs
->nr_vecs
<= BIO_INLINE_VECS
) {
2099 size
= bvs
->nr_vecs
* sizeof(struct bio_vec
);
2100 bvs
->slab
= kmem_cache_create(bvs
->name
, size
, 0,
2101 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2105 static int __init
init_bio(void)
2109 bio_slabs
= kzalloc(bio_slab_max
* sizeof(struct bio_slab
), GFP_KERNEL
);
2111 panic("bio: can't allocate bios\n");
2113 bio_integrity_init();
2114 biovec_init_slabs();
2116 fs_bio_set
= bioset_create(BIO_POOL_SIZE
, 0, BIOSET_NEED_BVECS
);
2118 panic("bio: can't allocate bios\n");
2120 if (bioset_integrity_create(fs_bio_set
, BIO_POOL_SIZE
))
2121 panic("bio: can't create integrity pool\n");
2125 subsys_initcall(init_bio
);