2 * Copyright (C) 2001 Jens Axboe <axboe@kernel.dk>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License version 2 as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
13 * You should have received a copy of the GNU General Public Licens
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
19 #include <linux/swap.h>
20 #include <linux/bio.h>
21 #include <linux/blkdev.h>
22 #include <linux/uio.h>
23 #include <linux/iocontext.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/kernel.h>
27 #include <linux/export.h>
28 #include <linux/mempool.h>
29 #include <linux/workqueue.h>
30 #include <linux/cgroup.h>
32 #include <trace/events/block.h>
35 * Test patch to inline a certain number of bi_io_vec's inside the bio
36 * itself, to shrink a bio data allocation from two mempool calls to one
38 #define BIO_INLINE_VECS 4
41 * if you change this list, also change bvec_alloc or things will
42 * break badly! cannot be bigger than what you can fit into an
45 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
46 static struct biovec_slab bvec_slabs
[BIOVEC_NR_POOLS
] __read_mostly
= {
47 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES
),
52 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
53 * IO code that does not need private memory pools.
55 struct bio_set
*fs_bio_set
;
56 EXPORT_SYMBOL(fs_bio_set
);
59 * Our slab pool management
62 struct kmem_cache
*slab
;
63 unsigned int slab_ref
;
64 unsigned int slab_size
;
67 static DEFINE_MUTEX(bio_slab_lock
);
68 static struct bio_slab
*bio_slabs
;
69 static unsigned int bio_slab_nr
, bio_slab_max
;
71 static struct kmem_cache
*bio_find_or_create_slab(unsigned int extra_size
)
73 unsigned int sz
= sizeof(struct bio
) + extra_size
;
74 struct kmem_cache
*slab
= NULL
;
75 struct bio_slab
*bslab
, *new_bio_slabs
;
76 unsigned int new_bio_slab_max
;
77 unsigned int i
, entry
= -1;
79 mutex_lock(&bio_slab_lock
);
82 while (i
< bio_slab_nr
) {
83 bslab
= &bio_slabs
[i
];
85 if (!bslab
->slab
&& entry
== -1)
87 else if (bslab
->slab_size
== sz
) {
98 if (bio_slab_nr
== bio_slab_max
&& entry
== -1) {
99 new_bio_slab_max
= bio_slab_max
<< 1;
100 new_bio_slabs
= krealloc(bio_slabs
,
101 new_bio_slab_max
* sizeof(struct bio_slab
),
105 bio_slab_max
= new_bio_slab_max
;
106 bio_slabs
= new_bio_slabs
;
109 entry
= bio_slab_nr
++;
111 bslab
= &bio_slabs
[entry
];
113 snprintf(bslab
->name
, sizeof(bslab
->name
), "bio-%d", entry
);
114 slab
= kmem_cache_create(bslab
->name
, sz
, ARCH_KMALLOC_MINALIGN
,
115 SLAB_HWCACHE_ALIGN
, NULL
);
121 bslab
->slab_size
= sz
;
123 mutex_unlock(&bio_slab_lock
);
127 static void bio_put_slab(struct bio_set
*bs
)
129 struct bio_slab
*bslab
= NULL
;
132 mutex_lock(&bio_slab_lock
);
134 for (i
= 0; i
< bio_slab_nr
; i
++) {
135 if (bs
->bio_slab
== bio_slabs
[i
].slab
) {
136 bslab
= &bio_slabs
[i
];
141 if (WARN(!bslab
, KERN_ERR
"bio: unable to find slab!\n"))
144 WARN_ON(!bslab
->slab_ref
);
146 if (--bslab
->slab_ref
)
149 kmem_cache_destroy(bslab
->slab
);
153 mutex_unlock(&bio_slab_lock
);
156 unsigned int bvec_nr_vecs(unsigned short idx
)
158 return bvec_slabs
[idx
].nr_vecs
;
161 void bvec_free(mempool_t
*pool
, struct bio_vec
*bv
, unsigned int idx
)
163 BIO_BUG_ON(idx
>= BIOVEC_NR_POOLS
);
165 if (idx
== BIOVEC_MAX_IDX
)
166 mempool_free(bv
, pool
);
168 struct biovec_slab
*bvs
= bvec_slabs
+ idx
;
170 kmem_cache_free(bvs
->slab
, bv
);
174 struct bio_vec
*bvec_alloc(gfp_t gfp_mask
, int nr
, unsigned long *idx
,
180 * see comment near bvec_array define!
198 case 129 ... BIO_MAX_PAGES
:
206 * idx now points to the pool we want to allocate from. only the
207 * 1-vec entry pool is mempool backed.
209 if (*idx
== BIOVEC_MAX_IDX
) {
211 bvl
= mempool_alloc(pool
, gfp_mask
);
213 struct biovec_slab
*bvs
= bvec_slabs
+ *idx
;
214 gfp_t __gfp_mask
= gfp_mask
& ~(__GFP_DIRECT_RECLAIM
| __GFP_IO
);
217 * Make this allocation restricted and don't dump info on
218 * allocation failures, since we'll fallback to the mempool
219 * in case of failure.
221 __gfp_mask
|= __GFP_NOMEMALLOC
| __GFP_NORETRY
| __GFP_NOWARN
;
224 * Try a slab allocation. If this fails and __GFP_DIRECT_RECLAIM
225 * is set, retry with the 1-entry mempool
227 bvl
= kmem_cache_alloc(bvs
->slab
, __gfp_mask
);
228 if (unlikely(!bvl
&& (gfp_mask
& __GFP_DIRECT_RECLAIM
))) {
229 *idx
= BIOVEC_MAX_IDX
;
237 static void __bio_free(struct bio
*bio
)
239 bio_disassociate_task(bio
);
241 if (bio_integrity(bio
))
242 bio_integrity_free(bio
);
245 static void bio_free(struct bio
*bio
)
247 struct bio_set
*bs
= bio
->bi_pool
;
253 if (bio_flagged(bio
, BIO_OWNS_VEC
))
254 bvec_free(bs
->bvec_pool
, bio
->bi_io_vec
, BIO_POOL_IDX(bio
));
257 * If we have front padding, adjust the bio pointer before freeing
262 mempool_free(p
, bs
->bio_pool
);
264 /* Bio was allocated by bio_kmalloc() */
269 void bio_init(struct bio
*bio
)
271 memset(bio
, 0, sizeof(*bio
));
272 atomic_set(&bio
->__bi_remaining
, 1);
273 atomic_set(&bio
->__bi_cnt
, 1);
275 EXPORT_SYMBOL(bio_init
);
278 * bio_reset - reinitialize a bio
282 * After calling bio_reset(), @bio will be in the same state as a freshly
283 * allocated bio returned bio bio_alloc_bioset() - the only fields that are
284 * preserved are the ones that are initialized by bio_alloc_bioset(). See
285 * comment in struct bio.
287 void bio_reset(struct bio
*bio
)
289 unsigned long flags
= bio
->bi_flags
& (~0UL << BIO_RESET_BITS
);
293 memset(bio
, 0, BIO_RESET_BYTES
);
294 bio
->bi_flags
= flags
;
295 atomic_set(&bio
->__bi_remaining
, 1);
297 EXPORT_SYMBOL(bio_reset
);
299 static struct bio
*__bio_chain_endio(struct bio
*bio
)
301 struct bio
*parent
= bio
->bi_private
;
303 if (!parent
->bi_error
)
304 parent
->bi_error
= bio
->bi_error
;
309 static void bio_chain_endio(struct bio
*bio
)
311 bio_endio(__bio_chain_endio(bio
));
315 * bio_chain - chain bio completions
316 * @bio: the target bio
317 * @parent: the @bio's parent bio
319 * The caller won't have a bi_end_io called when @bio completes - instead,
320 * @parent's bi_end_io won't be called until both @parent and @bio have
321 * completed; the chained bio will also be freed when it completes.
323 * The caller must not set bi_private or bi_end_io in @bio.
325 void bio_chain(struct bio
*bio
, struct bio
*parent
)
327 BUG_ON(bio
->bi_private
|| bio
->bi_end_io
);
329 bio
->bi_private
= parent
;
330 bio
->bi_end_io
= bio_chain_endio
;
331 bio_inc_remaining(parent
);
333 EXPORT_SYMBOL(bio_chain
);
335 static void bio_alloc_rescue(struct work_struct
*work
)
337 struct bio_set
*bs
= container_of(work
, struct bio_set
, rescue_work
);
341 spin_lock(&bs
->rescue_lock
);
342 bio
= bio_list_pop(&bs
->rescue_list
);
343 spin_unlock(&bs
->rescue_lock
);
348 generic_make_request(bio
);
352 static void punt_bios_to_rescuer(struct bio_set
*bs
)
354 struct bio_list punt
, nopunt
;
358 * In order to guarantee forward progress we must punt only bios that
359 * were allocated from this bio_set; otherwise, if there was a bio on
360 * there for a stacking driver higher up in the stack, processing it
361 * could require allocating bios from this bio_set, and doing that from
362 * our own rescuer would be bad.
364 * Since bio lists are singly linked, pop them all instead of trying to
365 * remove from the middle of the list:
368 bio_list_init(&punt
);
369 bio_list_init(&nopunt
);
371 while ((bio
= bio_list_pop(current
->bio_list
)))
372 bio_list_add(bio
->bi_pool
== bs
? &punt
: &nopunt
, bio
);
374 *current
->bio_list
= nopunt
;
376 spin_lock(&bs
->rescue_lock
);
377 bio_list_merge(&bs
->rescue_list
, &punt
);
378 spin_unlock(&bs
->rescue_lock
);
380 queue_work(bs
->rescue_workqueue
, &bs
->rescue_work
);
384 * bio_alloc_bioset - allocate a bio for I/O
385 * @gfp_mask: the GFP_ mask given to the slab allocator
386 * @nr_iovecs: number of iovecs to pre-allocate
387 * @bs: the bio_set to allocate from.
390 * If @bs is NULL, uses kmalloc() to allocate the bio; else the allocation is
391 * backed by the @bs's mempool.
393 * When @bs is not NULL, if %__GFP_DIRECT_RECLAIM is set then bio_alloc will
394 * always be able to allocate a bio. This is due to the mempool guarantees.
395 * To make this work, callers must never allocate more than 1 bio at a time
396 * from this pool. Callers that need to allocate more than 1 bio must always
397 * submit the previously allocated bio for IO before attempting to allocate
398 * a new one. Failure to do so can cause deadlocks under memory pressure.
400 * Note that when running under generic_make_request() (i.e. any block
401 * driver), bios are not submitted until after you return - see the code in
402 * generic_make_request() that converts recursion into iteration, to prevent
405 * This would normally mean allocating multiple bios under
406 * generic_make_request() would be susceptible to deadlocks, but we have
407 * deadlock avoidance code that resubmits any blocked bios from a rescuer
410 * However, we do not guarantee forward progress for allocations from other
411 * mempools. Doing multiple allocations from the same mempool under
412 * generic_make_request() should be avoided - instead, use bio_set's front_pad
413 * for per bio allocations.
416 * Pointer to new bio on success, NULL on failure.
418 struct bio
*bio_alloc_bioset(gfp_t gfp_mask
, int nr_iovecs
, struct bio_set
*bs
)
420 gfp_t saved_gfp
= gfp_mask
;
422 unsigned inline_vecs
;
423 unsigned long idx
= BIO_POOL_NONE
;
424 struct bio_vec
*bvl
= NULL
;
429 if (nr_iovecs
> UIO_MAXIOV
)
432 p
= kmalloc(sizeof(struct bio
) +
433 nr_iovecs
* sizeof(struct bio_vec
),
436 inline_vecs
= nr_iovecs
;
438 /* should not use nobvec bioset for nr_iovecs > 0 */
439 if (WARN_ON_ONCE(!bs
->bvec_pool
&& nr_iovecs
> 0))
442 * generic_make_request() converts recursion to iteration; this
443 * means if we're running beneath it, any bios we allocate and
444 * submit will not be submitted (and thus freed) until after we
447 * This exposes us to a potential deadlock if we allocate
448 * multiple bios from the same bio_set() while running
449 * underneath generic_make_request(). If we were to allocate
450 * multiple bios (say a stacking block driver that was splitting
451 * bios), we would deadlock if we exhausted the mempool's
454 * We solve this, and guarantee forward progress, with a rescuer
455 * workqueue per bio_set. If we go to allocate and there are
456 * bios on current->bio_list, we first try the allocation
457 * without __GFP_DIRECT_RECLAIM; if that fails, we punt those
458 * bios we would be blocking to the rescuer workqueue before
459 * we retry with the original gfp_flags.
462 if (current
->bio_list
&& !bio_list_empty(current
->bio_list
))
463 gfp_mask
&= ~__GFP_DIRECT_RECLAIM
;
465 p
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
466 if (!p
&& gfp_mask
!= saved_gfp
) {
467 punt_bios_to_rescuer(bs
);
468 gfp_mask
= saved_gfp
;
469 p
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
472 front_pad
= bs
->front_pad
;
473 inline_vecs
= BIO_INLINE_VECS
;
482 if (nr_iovecs
> inline_vecs
) {
483 bvl
= bvec_alloc(gfp_mask
, nr_iovecs
, &idx
, bs
->bvec_pool
);
484 if (!bvl
&& gfp_mask
!= saved_gfp
) {
485 punt_bios_to_rescuer(bs
);
486 gfp_mask
= saved_gfp
;
487 bvl
= bvec_alloc(gfp_mask
, nr_iovecs
, &idx
, bs
->bvec_pool
);
493 bio_set_flag(bio
, BIO_OWNS_VEC
);
494 } else if (nr_iovecs
) {
495 bvl
= bio
->bi_inline_vecs
;
499 bio
->bi_flags
|= idx
<< BIO_POOL_OFFSET
;
500 bio
->bi_max_vecs
= nr_iovecs
;
501 bio
->bi_io_vec
= bvl
;
505 mempool_free(p
, bs
->bio_pool
);
508 EXPORT_SYMBOL(bio_alloc_bioset
);
510 void zero_fill_bio(struct bio
*bio
)
514 struct bvec_iter iter
;
516 bio_for_each_segment(bv
, bio
, iter
) {
517 char *data
= bvec_kmap_irq(&bv
, &flags
);
518 memset(data
, 0, bv
.bv_len
);
519 flush_dcache_page(bv
.bv_page
);
520 bvec_kunmap_irq(data
, &flags
);
523 EXPORT_SYMBOL(zero_fill_bio
);
526 * bio_put - release a reference to a bio
527 * @bio: bio to release reference to
530 * Put a reference to a &struct bio, either one you have gotten with
531 * bio_alloc, bio_get or bio_clone. The last put of a bio will free it.
533 void bio_put(struct bio
*bio
)
535 if (!bio_flagged(bio
, BIO_REFFED
))
538 BIO_BUG_ON(!atomic_read(&bio
->__bi_cnt
));
543 if (atomic_dec_and_test(&bio
->__bi_cnt
))
547 EXPORT_SYMBOL(bio_put
);
549 inline int bio_phys_segments(struct request_queue
*q
, struct bio
*bio
)
551 if (unlikely(!bio_flagged(bio
, BIO_SEG_VALID
)))
552 blk_recount_segments(q
, bio
);
554 return bio
->bi_phys_segments
;
556 EXPORT_SYMBOL(bio_phys_segments
);
559 * __bio_clone_fast - clone a bio that shares the original bio's biovec
560 * @bio: destination bio
561 * @bio_src: bio to clone
563 * Clone a &bio. Caller will own the returned bio, but not
564 * the actual data it points to. Reference count of returned
567 * Caller must ensure that @bio_src is not freed before @bio.
569 void __bio_clone_fast(struct bio
*bio
, struct bio
*bio_src
)
571 BUG_ON(bio
->bi_pool
&& BIO_POOL_IDX(bio
) != BIO_POOL_NONE
);
574 * most users will be overriding ->bi_bdev with a new target,
575 * so we don't set nor calculate new physical/hw segment counts here
577 bio
->bi_bdev
= bio_src
->bi_bdev
;
578 bio_set_flag(bio
, BIO_CLONED
);
579 bio
->bi_rw
= bio_src
->bi_rw
;
580 bio
->bi_iter
= bio_src
->bi_iter
;
581 bio
->bi_io_vec
= bio_src
->bi_io_vec
;
583 EXPORT_SYMBOL(__bio_clone_fast
);
586 * bio_clone_fast - clone a bio that shares the original bio's biovec
588 * @gfp_mask: allocation priority
589 * @bs: bio_set to allocate from
591 * Like __bio_clone_fast, only also allocates the returned bio
593 struct bio
*bio_clone_fast(struct bio
*bio
, gfp_t gfp_mask
, struct bio_set
*bs
)
597 b
= bio_alloc_bioset(gfp_mask
, 0, bs
);
601 __bio_clone_fast(b
, bio
);
603 if (bio_integrity(bio
)) {
606 ret
= bio_integrity_clone(b
, bio
, gfp_mask
);
616 EXPORT_SYMBOL(bio_clone_fast
);
619 * bio_clone_bioset - clone a bio
620 * @bio_src: bio to clone
621 * @gfp_mask: allocation priority
622 * @bs: bio_set to allocate from
624 * Clone bio. Caller will own the returned bio, but not the actual data it
625 * points to. Reference count of returned bio will be one.
627 struct bio
*bio_clone_bioset(struct bio
*bio_src
, gfp_t gfp_mask
,
630 struct bvec_iter iter
;
635 * Pre immutable biovecs, __bio_clone() used to just do a memcpy from
636 * bio_src->bi_io_vec to bio->bi_io_vec.
638 * We can't do that anymore, because:
640 * - The point of cloning the biovec is to produce a bio with a biovec
641 * the caller can modify: bi_idx and bi_bvec_done should be 0.
643 * - The original bio could've had more than BIO_MAX_PAGES biovecs; if
644 * we tried to clone the whole thing bio_alloc_bioset() would fail.
645 * But the clone should succeed as long as the number of biovecs we
646 * actually need to allocate is fewer than BIO_MAX_PAGES.
648 * - Lastly, bi_vcnt should not be looked at or relied upon by code
649 * that does not own the bio - reason being drivers don't use it for
650 * iterating over the biovec anymore, so expecting it to be kept up
651 * to date (i.e. for clones that share the parent biovec) is just
652 * asking for trouble and would force extra work on
653 * __bio_clone_fast() anyways.
656 bio
= bio_alloc_bioset(gfp_mask
, bio_segments(bio_src
), bs
);
659 bio
->bi_bdev
= bio_src
->bi_bdev
;
660 bio
->bi_rw
= bio_src
->bi_rw
;
661 bio
->bi_iter
.bi_sector
= bio_src
->bi_iter
.bi_sector
;
662 bio
->bi_iter
.bi_size
= bio_src
->bi_iter
.bi_size
;
664 if (bio_op(bio
) == REQ_OP_DISCARD
)
665 goto integrity_clone
;
667 if (bio_op(bio
) == REQ_OP_WRITE_SAME
) {
668 bio
->bi_io_vec
[bio
->bi_vcnt
++] = bio_src
->bi_io_vec
[0];
669 goto integrity_clone
;
672 bio_for_each_segment(bv
, bio_src
, iter
)
673 bio
->bi_io_vec
[bio
->bi_vcnt
++] = bv
;
676 if (bio_integrity(bio_src
)) {
679 ret
= bio_integrity_clone(bio
, bio_src
, gfp_mask
);
688 EXPORT_SYMBOL(bio_clone_bioset
);
691 * bio_add_pc_page - attempt to add page to bio
692 * @q: the target queue
693 * @bio: destination bio
695 * @len: vec entry length
696 * @offset: vec entry offset
698 * Attempt to add a page to the bio_vec maplist. This can fail for a
699 * number of reasons, such as the bio being full or target block device
700 * limitations. The target block device must allow bio's up to PAGE_SIZE,
701 * so it is always possible to add a single page to an empty bio.
703 * This should only be used by REQ_PC bios.
705 int bio_add_pc_page(struct request_queue
*q
, struct bio
*bio
, struct page
706 *page
, unsigned int len
, unsigned int offset
)
708 int retried_segments
= 0;
709 struct bio_vec
*bvec
;
712 * cloned bio must not modify vec list
714 if (unlikely(bio_flagged(bio
, BIO_CLONED
)))
717 if (((bio
->bi_iter
.bi_size
+ len
) >> 9) > queue_max_hw_sectors(q
))
721 * For filesystems with a blocksize smaller than the pagesize
722 * we will often be called with the same page as last time and
723 * a consecutive offset. Optimize this special case.
725 if (bio
->bi_vcnt
> 0) {
726 struct bio_vec
*prev
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
728 if (page
== prev
->bv_page
&&
729 offset
== prev
->bv_offset
+ prev
->bv_len
) {
731 bio
->bi_iter
.bi_size
+= len
;
736 * If the queue doesn't support SG gaps and adding this
737 * offset would create a gap, disallow it.
739 if (bvec_gap_to_prev(q
, prev
, offset
))
743 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
747 * setup the new entry, we might clear it again later if we
748 * cannot add the page
750 bvec
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
751 bvec
->bv_page
= page
;
753 bvec
->bv_offset
= offset
;
755 bio
->bi_phys_segments
++;
756 bio
->bi_iter
.bi_size
+= len
;
759 * Perform a recount if the number of segments is greater
760 * than queue_max_segments(q).
763 while (bio
->bi_phys_segments
> queue_max_segments(q
)) {
765 if (retried_segments
)
768 retried_segments
= 1;
769 blk_recount_segments(q
, bio
);
772 /* If we may be able to merge these biovecs, force a recount */
773 if (bio
->bi_vcnt
> 1 && (BIOVEC_PHYS_MERGEABLE(bvec
-1, bvec
)))
774 bio_clear_flag(bio
, BIO_SEG_VALID
);
780 bvec
->bv_page
= NULL
;
784 bio
->bi_iter
.bi_size
-= len
;
785 blk_recount_segments(q
, bio
);
788 EXPORT_SYMBOL(bio_add_pc_page
);
791 * bio_add_page - attempt to add page to bio
792 * @bio: destination bio
794 * @len: vec entry length
795 * @offset: vec entry offset
797 * Attempt to add a page to the bio_vec maplist. This will only fail
798 * if either bio->bi_vcnt == bio->bi_max_vecs or it's a cloned bio.
800 int bio_add_page(struct bio
*bio
, struct page
*page
,
801 unsigned int len
, unsigned int offset
)
806 * cloned bio must not modify vec list
808 if (WARN_ON_ONCE(bio_flagged(bio
, BIO_CLONED
)))
812 * For filesystems with a blocksize smaller than the pagesize
813 * we will often be called with the same page as last time and
814 * a consecutive offset. Optimize this special case.
816 if (bio
->bi_vcnt
> 0) {
817 bv
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
819 if (page
== bv
->bv_page
&&
820 offset
== bv
->bv_offset
+ bv
->bv_len
) {
826 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
829 bv
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
832 bv
->bv_offset
= offset
;
836 bio
->bi_iter
.bi_size
+= len
;
839 EXPORT_SYMBOL(bio_add_page
);
841 struct submit_bio_ret
{
842 struct completion event
;
846 static void submit_bio_wait_endio(struct bio
*bio
)
848 struct submit_bio_ret
*ret
= bio
->bi_private
;
850 ret
->error
= bio
->bi_error
;
851 complete(&ret
->event
);
855 * submit_bio_wait - submit a bio, and wait until it completes
856 * @bio: The &struct bio which describes the I/O
858 * Simple wrapper around submit_bio(). Returns 0 on success, or the error from
859 * bio_endio() on failure.
861 int submit_bio_wait(struct bio
*bio
)
863 struct submit_bio_ret ret
;
865 init_completion(&ret
.event
);
866 bio
->bi_private
= &ret
;
867 bio
->bi_end_io
= submit_bio_wait_endio
;
868 bio
->bi_rw
|= REQ_SYNC
;
870 wait_for_completion_io(&ret
.event
);
874 EXPORT_SYMBOL(submit_bio_wait
);
877 * bio_advance - increment/complete a bio by some number of bytes
878 * @bio: bio to advance
879 * @bytes: number of bytes to complete
881 * This updates bi_sector, bi_size and bi_idx; if the number of bytes to
882 * complete doesn't align with a bvec boundary, then bv_len and bv_offset will
883 * be updated on the last bvec as well.
885 * @bio will then represent the remaining, uncompleted portion of the io.
887 void bio_advance(struct bio
*bio
, unsigned bytes
)
889 if (bio_integrity(bio
))
890 bio_integrity_advance(bio
, bytes
);
892 bio_advance_iter(bio
, &bio
->bi_iter
, bytes
);
894 EXPORT_SYMBOL(bio_advance
);
897 * bio_alloc_pages - allocates a single page for each bvec in a bio
898 * @bio: bio to allocate pages for
899 * @gfp_mask: flags for allocation
901 * Allocates pages up to @bio->bi_vcnt.
903 * Returns 0 on success, -ENOMEM on failure. On failure, any allocated pages are
906 int bio_alloc_pages(struct bio
*bio
, gfp_t gfp_mask
)
911 bio_for_each_segment_all(bv
, bio
, i
) {
912 bv
->bv_page
= alloc_page(gfp_mask
);
914 while (--bv
>= bio
->bi_io_vec
)
915 __free_page(bv
->bv_page
);
922 EXPORT_SYMBOL(bio_alloc_pages
);
925 * bio_copy_data - copy contents of data buffers from one chain of bios to
927 * @src: source bio list
928 * @dst: destination bio list
930 * If @src and @dst are single bios, bi_next must be NULL - otherwise, treats
931 * @src and @dst as linked lists of bios.
933 * Stops when it reaches the end of either @src or @dst - that is, copies
934 * min(src->bi_size, dst->bi_size) bytes (or the equivalent for lists of bios).
936 void bio_copy_data(struct bio
*dst
, struct bio
*src
)
938 struct bvec_iter src_iter
, dst_iter
;
939 struct bio_vec src_bv
, dst_bv
;
943 src_iter
= src
->bi_iter
;
944 dst_iter
= dst
->bi_iter
;
947 if (!src_iter
.bi_size
) {
952 src_iter
= src
->bi_iter
;
955 if (!dst_iter
.bi_size
) {
960 dst_iter
= dst
->bi_iter
;
963 src_bv
= bio_iter_iovec(src
, src_iter
);
964 dst_bv
= bio_iter_iovec(dst
, dst_iter
);
966 bytes
= min(src_bv
.bv_len
, dst_bv
.bv_len
);
968 src_p
= kmap_atomic(src_bv
.bv_page
);
969 dst_p
= kmap_atomic(dst_bv
.bv_page
);
971 memcpy(dst_p
+ dst_bv
.bv_offset
,
972 src_p
+ src_bv
.bv_offset
,
975 kunmap_atomic(dst_p
);
976 kunmap_atomic(src_p
);
978 bio_advance_iter(src
, &src_iter
, bytes
);
979 bio_advance_iter(dst
, &dst_iter
, bytes
);
982 EXPORT_SYMBOL(bio_copy_data
);
984 struct bio_map_data
{
986 struct iov_iter iter
;
990 static struct bio_map_data
*bio_alloc_map_data(unsigned int iov_count
,
993 if (iov_count
> UIO_MAXIOV
)
996 return kmalloc(sizeof(struct bio_map_data
) +
997 sizeof(struct iovec
) * iov_count
, gfp_mask
);
1001 * bio_copy_from_iter - copy all pages from iov_iter to bio
1002 * @bio: The &struct bio which describes the I/O as destination
1003 * @iter: iov_iter as source
1005 * Copy all pages from iov_iter to bio.
1006 * Returns 0 on success, or error on failure.
1008 static int bio_copy_from_iter(struct bio
*bio
, struct iov_iter iter
)
1011 struct bio_vec
*bvec
;
1013 bio_for_each_segment_all(bvec
, bio
, i
) {
1016 ret
= copy_page_from_iter(bvec
->bv_page
,
1021 if (!iov_iter_count(&iter
))
1024 if (ret
< bvec
->bv_len
)
1032 * bio_copy_to_iter - copy all pages from bio to iov_iter
1033 * @bio: The &struct bio which describes the I/O as source
1034 * @iter: iov_iter as destination
1036 * Copy all pages from bio to iov_iter.
1037 * Returns 0 on success, or error on failure.
1039 static int bio_copy_to_iter(struct bio
*bio
, struct iov_iter iter
)
1042 struct bio_vec
*bvec
;
1044 bio_for_each_segment_all(bvec
, bio
, i
) {
1047 ret
= copy_page_to_iter(bvec
->bv_page
,
1052 if (!iov_iter_count(&iter
))
1055 if (ret
< bvec
->bv_len
)
1062 static void bio_free_pages(struct bio
*bio
)
1064 struct bio_vec
*bvec
;
1067 bio_for_each_segment_all(bvec
, bio
, i
)
1068 __free_page(bvec
->bv_page
);
1072 * bio_uncopy_user - finish previously mapped bio
1073 * @bio: bio being terminated
1075 * Free pages allocated from bio_copy_user_iov() and write back data
1076 * to user space in case of a read.
1078 int bio_uncopy_user(struct bio
*bio
)
1080 struct bio_map_data
*bmd
= bio
->bi_private
;
1083 if (!bio_flagged(bio
, BIO_NULL_MAPPED
)) {
1085 * if we're in a workqueue, the request is orphaned, so
1086 * don't copy into a random user address space, just free
1087 * and return -EINTR so user space doesn't expect any data.
1091 else if (bio_data_dir(bio
) == READ
)
1092 ret
= bio_copy_to_iter(bio
, bmd
->iter
);
1093 if (bmd
->is_our_pages
)
1094 bio_free_pages(bio
);
1100 EXPORT_SYMBOL(bio_uncopy_user
);
1103 * bio_copy_user_iov - copy user data to bio
1104 * @q: destination block queue
1105 * @map_data: pointer to the rq_map_data holding pages (if necessary)
1106 * @iter: iovec iterator
1107 * @gfp_mask: memory allocation flags
1109 * Prepares and returns a bio for indirect user io, bouncing data
1110 * to/from kernel pages as necessary. Must be paired with
1111 * call bio_uncopy_user() on io completion.
1113 struct bio
*bio_copy_user_iov(struct request_queue
*q
,
1114 struct rq_map_data
*map_data
,
1115 const struct iov_iter
*iter
,
1118 struct bio_map_data
*bmd
;
1123 unsigned int len
= iter
->count
;
1124 unsigned int offset
= map_data
? offset_in_page(map_data
->offset
) : 0;
1126 for (i
= 0; i
< iter
->nr_segs
; i
++) {
1127 unsigned long uaddr
;
1129 unsigned long start
;
1131 uaddr
= (unsigned long) iter
->iov
[i
].iov_base
;
1132 end
= (uaddr
+ iter
->iov
[i
].iov_len
+ PAGE_SIZE
- 1)
1134 start
= uaddr
>> PAGE_SHIFT
;
1140 return ERR_PTR(-EINVAL
);
1142 nr_pages
+= end
- start
;
1148 bmd
= bio_alloc_map_data(iter
->nr_segs
, gfp_mask
);
1150 return ERR_PTR(-ENOMEM
);
1153 * We need to do a deep copy of the iov_iter including the iovecs.
1154 * The caller provided iov might point to an on-stack or otherwise
1157 bmd
->is_our_pages
= map_data
? 0 : 1;
1158 memcpy(bmd
->iov
, iter
->iov
, sizeof(struct iovec
) * iter
->nr_segs
);
1159 iov_iter_init(&bmd
->iter
, iter
->type
, bmd
->iov
,
1160 iter
->nr_segs
, iter
->count
);
1163 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1167 if (iter
->type
& WRITE
)
1168 bio_set_op_attrs(bio
, REQ_OP_WRITE
, 0);
1173 nr_pages
= 1 << map_data
->page_order
;
1174 i
= map_data
->offset
/ PAGE_SIZE
;
1177 unsigned int bytes
= PAGE_SIZE
;
1185 if (i
== map_data
->nr_entries
* nr_pages
) {
1190 page
= map_data
->pages
[i
/ nr_pages
];
1191 page
+= (i
% nr_pages
);
1195 page
= alloc_page(q
->bounce_gfp
| gfp_mask
);
1202 if (bio_add_pc_page(q
, bio
, page
, bytes
, offset
) < bytes
)
1215 if (((iter
->type
& WRITE
) && (!map_data
|| !map_data
->null_mapped
)) ||
1216 (map_data
&& map_data
->from_user
)) {
1217 ret
= bio_copy_from_iter(bio
, *iter
);
1222 bio
->bi_private
= bmd
;
1226 bio_free_pages(bio
);
1230 return ERR_PTR(ret
);
1234 * bio_map_user_iov - map user iovec into bio
1235 * @q: the struct request_queue for the bio
1236 * @iter: iovec iterator
1237 * @gfp_mask: memory allocation flags
1239 * Map the user space address into a bio suitable for io to a block
1240 * device. Returns an error pointer in case of error.
1242 struct bio
*bio_map_user_iov(struct request_queue
*q
,
1243 const struct iov_iter
*iter
,
1248 struct page
**pages
;
1255 iov_for_each(iov
, i
, *iter
) {
1256 unsigned long uaddr
= (unsigned long) iov
.iov_base
;
1257 unsigned long len
= iov
.iov_len
;
1258 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1259 unsigned long start
= uaddr
>> PAGE_SHIFT
;
1265 return ERR_PTR(-EINVAL
);
1267 nr_pages
+= end
- start
;
1269 * buffer must be aligned to at least hardsector size for now
1271 if (uaddr
& queue_dma_alignment(q
))
1272 return ERR_PTR(-EINVAL
);
1276 return ERR_PTR(-EINVAL
);
1278 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1280 return ERR_PTR(-ENOMEM
);
1283 pages
= kcalloc(nr_pages
, sizeof(struct page
*), gfp_mask
);
1287 iov_for_each(iov
, i
, *iter
) {
1288 unsigned long uaddr
= (unsigned long) iov
.iov_base
;
1289 unsigned long len
= iov
.iov_len
;
1290 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1291 unsigned long start
= uaddr
>> PAGE_SHIFT
;
1292 const int local_nr_pages
= end
- start
;
1293 const int page_limit
= cur_page
+ local_nr_pages
;
1295 ret
= get_user_pages_fast(uaddr
, local_nr_pages
,
1296 (iter
->type
& WRITE
) != WRITE
,
1298 if (ret
< local_nr_pages
) {
1303 offset
= offset_in_page(uaddr
);
1304 for (j
= cur_page
; j
< page_limit
; j
++) {
1305 unsigned int bytes
= PAGE_SIZE
- offset
;
1316 if (bio_add_pc_page(q
, bio
, pages
[j
], bytes
, offset
) <
1326 * release the pages we didn't map into the bio, if any
1328 while (j
< page_limit
)
1329 put_page(pages
[j
++]);
1335 * set data direction, and check if mapped pages need bouncing
1337 if (iter
->type
& WRITE
)
1338 bio_set_op_attrs(bio
, REQ_OP_WRITE
, 0);
1340 bio_set_flag(bio
, BIO_USER_MAPPED
);
1343 * subtle -- if __bio_map_user() ended up bouncing a bio,
1344 * it would normally disappear when its bi_end_io is run.
1345 * however, we need it for the unmap, so grab an extra
1352 for (j
= 0; j
< nr_pages
; j
++) {
1360 return ERR_PTR(ret
);
1363 static void __bio_unmap_user(struct bio
*bio
)
1365 struct bio_vec
*bvec
;
1369 * make sure we dirty pages we wrote to
1371 bio_for_each_segment_all(bvec
, bio
, i
) {
1372 if (bio_data_dir(bio
) == READ
)
1373 set_page_dirty_lock(bvec
->bv_page
);
1375 put_page(bvec
->bv_page
);
1382 * bio_unmap_user - unmap a bio
1383 * @bio: the bio being unmapped
1385 * Unmap a bio previously mapped by bio_map_user(). Must be called with
1386 * a process context.
1388 * bio_unmap_user() may sleep.
1390 void bio_unmap_user(struct bio
*bio
)
1392 __bio_unmap_user(bio
);
1395 EXPORT_SYMBOL(bio_unmap_user
);
1397 static void bio_map_kern_endio(struct bio
*bio
)
1403 * bio_map_kern - map kernel address into bio
1404 * @q: the struct request_queue for the bio
1405 * @data: pointer to buffer to map
1406 * @len: length in bytes
1407 * @gfp_mask: allocation flags for bio allocation
1409 * Map the kernel address into a bio suitable for io to a block
1410 * device. Returns an error pointer in case of error.
1412 struct bio
*bio_map_kern(struct request_queue
*q
, void *data
, unsigned int len
,
1415 unsigned long kaddr
= (unsigned long)data
;
1416 unsigned long end
= (kaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
1417 unsigned long start
= kaddr
>> PAGE_SHIFT
;
1418 const int nr_pages
= end
- start
;
1422 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1424 return ERR_PTR(-ENOMEM
);
1426 offset
= offset_in_page(kaddr
);
1427 for (i
= 0; i
< nr_pages
; i
++) {
1428 unsigned int bytes
= PAGE_SIZE
- offset
;
1436 if (bio_add_pc_page(q
, bio
, virt_to_page(data
), bytes
,
1438 /* we don't support partial mappings */
1440 return ERR_PTR(-EINVAL
);
1448 bio
->bi_end_io
= bio_map_kern_endio
;
1451 EXPORT_SYMBOL(bio_map_kern
);
1453 static void bio_copy_kern_endio(struct bio
*bio
)
1455 bio_free_pages(bio
);
1459 static void bio_copy_kern_endio_read(struct bio
*bio
)
1461 char *p
= bio
->bi_private
;
1462 struct bio_vec
*bvec
;
1465 bio_for_each_segment_all(bvec
, bio
, i
) {
1466 memcpy(p
, page_address(bvec
->bv_page
), bvec
->bv_len
);
1470 bio_copy_kern_endio(bio
);
1474 * bio_copy_kern - copy kernel address into bio
1475 * @q: the struct request_queue for the bio
1476 * @data: pointer to buffer to copy
1477 * @len: length in bytes
1478 * @gfp_mask: allocation flags for bio and page allocation
1479 * @reading: data direction is READ
1481 * copy 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_copy_kern(struct request_queue
*q
, void *data
, unsigned int len
,
1485 gfp_t gfp_mask
, int reading
)
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
;
1498 return ERR_PTR(-EINVAL
);
1500 nr_pages
= end
- start
;
1501 bio
= bio_kmalloc(gfp_mask
, nr_pages
);
1503 return ERR_PTR(-ENOMEM
);
1507 unsigned int bytes
= PAGE_SIZE
;
1512 page
= alloc_page(q
->bounce_gfp
| gfp_mask
);
1517 memcpy(page_address(page
), p
, bytes
);
1519 if (bio_add_pc_page(q
, bio
, page
, bytes
, 0) < bytes
)
1527 bio
->bi_end_io
= bio_copy_kern_endio_read
;
1528 bio
->bi_private
= data
;
1530 bio
->bi_end_io
= bio_copy_kern_endio
;
1531 bio_set_op_attrs(bio
, REQ_OP_WRITE
, 0);
1537 bio_free_pages(bio
);
1539 return ERR_PTR(-ENOMEM
);
1541 EXPORT_SYMBOL(bio_copy_kern
);
1544 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
1545 * for performing direct-IO in BIOs.
1547 * The problem is that we cannot run set_page_dirty() from interrupt context
1548 * because the required locks are not interrupt-safe. So what we can do is to
1549 * mark the pages dirty _before_ performing IO. And in interrupt context,
1550 * check that the pages are still dirty. If so, fine. If not, redirty them
1551 * in process context.
1553 * We special-case compound pages here: normally this means reads into hugetlb
1554 * pages. The logic in here doesn't really work right for compound pages
1555 * because the VM does not uniformly chase down the head page in all cases.
1556 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
1557 * handle them at all. So we skip compound pages here at an early stage.
1559 * Note that this code is very hard to test under normal circumstances because
1560 * direct-io pins the pages with get_user_pages(). This makes
1561 * is_page_cache_freeable return false, and the VM will not clean the pages.
1562 * But other code (eg, flusher threads) could clean the pages if they are mapped
1565 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
1566 * deferred bio dirtying paths.
1570 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
1572 void bio_set_pages_dirty(struct bio
*bio
)
1574 struct bio_vec
*bvec
;
1577 bio_for_each_segment_all(bvec
, bio
, i
) {
1578 struct page
*page
= bvec
->bv_page
;
1580 if (page
&& !PageCompound(page
))
1581 set_page_dirty_lock(page
);
1585 static void bio_release_pages(struct bio
*bio
)
1587 struct bio_vec
*bvec
;
1590 bio_for_each_segment_all(bvec
, bio
, i
) {
1591 struct page
*page
= bvec
->bv_page
;
1599 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
1600 * If they are, then fine. If, however, some pages are clean then they must
1601 * have been written out during the direct-IO read. So we take another ref on
1602 * the BIO and the offending pages and re-dirty the pages in process context.
1604 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
1605 * here on. It will run one put_page() against each page and will run one
1606 * bio_put() against the BIO.
1609 static void bio_dirty_fn(struct work_struct
*work
);
1611 static DECLARE_WORK(bio_dirty_work
, bio_dirty_fn
);
1612 static DEFINE_SPINLOCK(bio_dirty_lock
);
1613 static struct bio
*bio_dirty_list
;
1616 * This runs in process context
1618 static void bio_dirty_fn(struct work_struct
*work
)
1620 unsigned long flags
;
1623 spin_lock_irqsave(&bio_dirty_lock
, flags
);
1624 bio
= bio_dirty_list
;
1625 bio_dirty_list
= NULL
;
1626 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
1629 struct bio
*next
= bio
->bi_private
;
1631 bio_set_pages_dirty(bio
);
1632 bio_release_pages(bio
);
1638 void bio_check_pages_dirty(struct bio
*bio
)
1640 struct bio_vec
*bvec
;
1641 int nr_clean_pages
= 0;
1644 bio_for_each_segment_all(bvec
, bio
, i
) {
1645 struct page
*page
= bvec
->bv_page
;
1647 if (PageDirty(page
) || PageCompound(page
)) {
1649 bvec
->bv_page
= NULL
;
1655 if (nr_clean_pages
) {
1656 unsigned long flags
;
1658 spin_lock_irqsave(&bio_dirty_lock
, flags
);
1659 bio
->bi_private
= bio_dirty_list
;
1660 bio_dirty_list
= bio
;
1661 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
1662 schedule_work(&bio_dirty_work
);
1668 void generic_start_io_acct(int rw
, unsigned long sectors
,
1669 struct hd_struct
*part
)
1671 int cpu
= part_stat_lock();
1673 part_round_stats(cpu
, part
);
1674 part_stat_inc(cpu
, part
, ios
[rw
]);
1675 part_stat_add(cpu
, part
, sectors
[rw
], sectors
);
1676 part_inc_in_flight(part
, rw
);
1680 EXPORT_SYMBOL(generic_start_io_acct
);
1682 void generic_end_io_acct(int rw
, struct hd_struct
*part
,
1683 unsigned long start_time
)
1685 unsigned long duration
= jiffies
- start_time
;
1686 int cpu
= part_stat_lock();
1688 part_stat_add(cpu
, part
, ticks
[rw
], duration
);
1689 part_round_stats(cpu
, part
);
1690 part_dec_in_flight(part
, rw
);
1694 EXPORT_SYMBOL(generic_end_io_acct
);
1696 #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1697 void bio_flush_dcache_pages(struct bio
*bi
)
1699 struct bio_vec bvec
;
1700 struct bvec_iter iter
;
1702 bio_for_each_segment(bvec
, bi
, iter
)
1703 flush_dcache_page(bvec
.bv_page
);
1705 EXPORT_SYMBOL(bio_flush_dcache_pages
);
1708 static inline bool bio_remaining_done(struct bio
*bio
)
1711 * If we're not chaining, then ->__bi_remaining is always 1 and
1712 * we always end io on the first invocation.
1714 if (!bio_flagged(bio
, BIO_CHAIN
))
1717 BUG_ON(atomic_read(&bio
->__bi_remaining
) <= 0);
1719 if (atomic_dec_and_test(&bio
->__bi_remaining
)) {
1720 bio_clear_flag(bio
, BIO_CHAIN
);
1728 * bio_endio - end I/O on a bio
1732 * bio_endio() will end I/O on the whole bio. bio_endio() is the preferred
1733 * way to end I/O on a bio. No one should call bi_end_io() directly on a
1734 * bio unless they own it and thus know that it has an end_io function.
1736 void bio_endio(struct bio
*bio
)
1739 if (!bio_remaining_done(bio
))
1743 * Need to have a real endio function for chained bios, otherwise
1744 * various corner cases will break (like stacking block devices that
1745 * save/restore bi_end_io) - however, we want to avoid unbounded
1746 * recursion and blowing the stack. Tail call optimization would
1747 * handle this, but compiling with frame pointers also disables
1748 * gcc's sibling call optimization.
1750 if (bio
->bi_end_io
== bio_chain_endio
) {
1751 bio
= __bio_chain_endio(bio
);
1756 bio
->bi_end_io(bio
);
1758 EXPORT_SYMBOL(bio_endio
);
1761 * bio_split - split a bio
1762 * @bio: bio to split
1763 * @sectors: number of sectors to split from the front of @bio
1765 * @bs: bio set to allocate from
1767 * Allocates and returns a new bio which represents @sectors from the start of
1768 * @bio, and updates @bio to represent the remaining sectors.
1770 * Unless this is a discard request the newly allocated bio will point
1771 * to @bio's bi_io_vec; it is the caller's responsibility to ensure that
1772 * @bio is not freed before the split.
1774 struct bio
*bio_split(struct bio
*bio
, int sectors
,
1775 gfp_t gfp
, struct bio_set
*bs
)
1777 struct bio
*split
= NULL
;
1779 BUG_ON(sectors
<= 0);
1780 BUG_ON(sectors
>= bio_sectors(bio
));
1783 * Discards need a mutable bio_vec to accommodate the payload
1784 * required by the DSM TRIM and UNMAP commands.
1786 if (bio_op(bio
) == REQ_OP_DISCARD
)
1787 split
= bio_clone_bioset(bio
, gfp
, bs
);
1789 split
= bio_clone_fast(bio
, gfp
, bs
);
1794 split
->bi_iter
.bi_size
= sectors
<< 9;
1796 if (bio_integrity(split
))
1797 bio_integrity_trim(split
, 0, sectors
);
1799 bio_advance(bio
, split
->bi_iter
.bi_size
);
1803 EXPORT_SYMBOL(bio_split
);
1806 * bio_trim - trim a bio
1808 * @offset: number of sectors to trim from the front of @bio
1809 * @size: size we want to trim @bio to, in sectors
1811 void bio_trim(struct bio
*bio
, int offset
, int size
)
1813 /* 'bio' is a cloned bio which we need to trim to match
1814 * the given offset and size.
1818 if (offset
== 0 && size
== bio
->bi_iter
.bi_size
)
1821 bio_clear_flag(bio
, BIO_SEG_VALID
);
1823 bio_advance(bio
, offset
<< 9);
1825 bio
->bi_iter
.bi_size
= size
;
1827 EXPORT_SYMBOL_GPL(bio_trim
);
1830 * create memory pools for biovec's in a bio_set.
1831 * use the global biovec slabs created for general use.
1833 mempool_t
*biovec_create_pool(int pool_entries
)
1835 struct biovec_slab
*bp
= bvec_slabs
+ BIOVEC_MAX_IDX
;
1837 return mempool_create_slab_pool(pool_entries
, bp
->slab
);
1840 void bioset_free(struct bio_set
*bs
)
1842 if (bs
->rescue_workqueue
)
1843 destroy_workqueue(bs
->rescue_workqueue
);
1846 mempool_destroy(bs
->bio_pool
);
1849 mempool_destroy(bs
->bvec_pool
);
1851 bioset_integrity_free(bs
);
1856 EXPORT_SYMBOL(bioset_free
);
1858 static struct bio_set
*__bioset_create(unsigned int pool_size
,
1859 unsigned int front_pad
,
1860 bool create_bvec_pool
)
1862 unsigned int back_pad
= BIO_INLINE_VECS
* sizeof(struct bio_vec
);
1865 bs
= kzalloc(sizeof(*bs
), GFP_KERNEL
);
1869 bs
->front_pad
= front_pad
;
1871 spin_lock_init(&bs
->rescue_lock
);
1872 bio_list_init(&bs
->rescue_list
);
1873 INIT_WORK(&bs
->rescue_work
, bio_alloc_rescue
);
1875 bs
->bio_slab
= bio_find_or_create_slab(front_pad
+ back_pad
);
1876 if (!bs
->bio_slab
) {
1881 bs
->bio_pool
= mempool_create_slab_pool(pool_size
, bs
->bio_slab
);
1885 if (create_bvec_pool
) {
1886 bs
->bvec_pool
= biovec_create_pool(pool_size
);
1891 bs
->rescue_workqueue
= alloc_workqueue("bioset", WQ_MEM_RECLAIM
, 0);
1892 if (!bs
->rescue_workqueue
)
1902 * bioset_create - Create a bio_set
1903 * @pool_size: Number of bio and bio_vecs to cache in the mempool
1904 * @front_pad: Number of bytes to allocate in front of the returned bio
1907 * Set up a bio_set to be used with @bio_alloc_bioset. Allows the caller
1908 * to ask for a number of bytes to be allocated in front of the bio.
1909 * Front pad allocation is useful for embedding the bio inside
1910 * another structure, to avoid allocating extra data to go with the bio.
1911 * Note that the bio must be embedded at the END of that structure always,
1912 * or things will break badly.
1914 struct bio_set
*bioset_create(unsigned int pool_size
, unsigned int front_pad
)
1916 return __bioset_create(pool_size
, front_pad
, true);
1918 EXPORT_SYMBOL(bioset_create
);
1921 * bioset_create_nobvec - Create a bio_set without bio_vec mempool
1922 * @pool_size: Number of bio to cache in the mempool
1923 * @front_pad: Number of bytes to allocate in front of the returned bio
1926 * Same functionality as bioset_create() except that mempool is not
1927 * created for bio_vecs. Saving some memory for bio_clone_fast() users.
1929 struct bio_set
*bioset_create_nobvec(unsigned int pool_size
, unsigned int front_pad
)
1931 return __bioset_create(pool_size
, front_pad
, false);
1933 EXPORT_SYMBOL(bioset_create_nobvec
);
1935 #ifdef CONFIG_BLK_CGROUP
1938 * bio_associate_blkcg - associate a bio with the specified blkcg
1940 * @blkcg_css: css of the blkcg to associate
1942 * Associate @bio with the blkcg specified by @blkcg_css. Block layer will
1943 * treat @bio as if it were issued by a task which belongs to the blkcg.
1945 * This function takes an extra reference of @blkcg_css which will be put
1946 * when @bio is released. The caller must own @bio and is responsible for
1947 * synchronizing calls to this function.
1949 int bio_associate_blkcg(struct bio
*bio
, struct cgroup_subsys_state
*blkcg_css
)
1951 if (unlikely(bio
->bi_css
))
1954 bio
->bi_css
= blkcg_css
;
1957 EXPORT_SYMBOL_GPL(bio_associate_blkcg
);
1960 * bio_associate_current - associate a bio with %current
1963 * Associate @bio with %current if it hasn't been associated yet. Block
1964 * layer will treat @bio as if it were issued by %current no matter which
1965 * task actually issues it.
1967 * This function takes an extra reference of @task's io_context and blkcg
1968 * which will be put when @bio is released. The caller must own @bio,
1969 * ensure %current->io_context exists, and is responsible for synchronizing
1970 * calls to this function.
1972 int bio_associate_current(struct bio
*bio
)
1974 struct io_context
*ioc
;
1979 ioc
= current
->io_context
;
1983 get_io_context_active(ioc
);
1985 bio
->bi_css
= task_get_css(current
, io_cgrp_id
);
1988 EXPORT_SYMBOL_GPL(bio_associate_current
);
1991 * bio_disassociate_task - undo bio_associate_current()
1994 void bio_disassociate_task(struct bio
*bio
)
1997 put_io_context(bio
->bi_ioc
);
2001 css_put(bio
->bi_css
);
2006 #endif /* CONFIG_BLK_CGROUP */
2008 static void __init
biovec_init_slabs(void)
2012 for (i
= 0; i
< BIOVEC_NR_POOLS
; i
++) {
2014 struct biovec_slab
*bvs
= bvec_slabs
+ i
;
2016 if (bvs
->nr_vecs
<= BIO_INLINE_VECS
) {
2021 size
= bvs
->nr_vecs
* sizeof(struct bio_vec
);
2022 bvs
->slab
= kmem_cache_create(bvs
->name
, size
, 0,
2023 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
);
2027 static int __init
init_bio(void)
2031 bio_slabs
= kzalloc(bio_slab_max
* sizeof(struct bio_slab
), GFP_KERNEL
);
2033 panic("bio: can't allocate bios\n");
2035 bio_integrity_init();
2036 biovec_init_slabs();
2038 fs_bio_set
= bioset_create(BIO_POOL_SIZE
, 0);
2040 panic("bio: can't allocate bios\n");
2042 if (bioset_integrity_create(fs_bio_set
, BIO_POOL_SIZE
))
2043 panic("bio: can't create integrity pool\n");
2047 subsys_initcall(init_bio
);
This page took 0.073478 seconds and 5 git commands to generate.