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/slab.h>
23 #include <linux/init.h>
24 #include <linux/kernel.h>
25 #include <linux/module.h>
26 #include <linux/mempool.h>
27 #include <linux/workqueue.h>
28 #include <linux/blktrace_api.h>
29 #include <scsi/sg.h> /* for struct sg_iovec */
31 #define BIO_POOL_SIZE 256
33 static struct kmem_cache
*bio_slab __read_mostly
;
35 #define BIOVEC_NR_POOLS 6
38 * a small number of entries is fine, not going to be performance critical.
39 * basically we just need to survive
41 #define BIO_SPLIT_ENTRIES 8
42 mempool_t
*bio_split_pool __read_mostly
;
47 struct kmem_cache
*slab
;
51 * if you change this list, also change bvec_alloc or things will
52 * break badly! cannot be bigger than what you can fit into an
56 #define BV(x) { .nr_vecs = x, .name = "biovec-"__stringify(x) }
57 static struct biovec_slab bvec_slabs
[BIOVEC_NR_POOLS
] __read_mostly
= {
58 BV(1), BV(4), BV(16), BV(64), BV(128), BV(BIO_MAX_PAGES
),
63 * bio_set is used to allow other portions of the IO system to
64 * allocate their own private memory pools for bio and iovec structures.
65 * These memory pools in turn all allocate from the bio_slab
66 * and the bvec_slabs[].
70 mempool_t
*bvec_pools
[BIOVEC_NR_POOLS
];
74 * fs_bio_set is the bio_set containing bio and iovec memory pools used by
75 * IO code that does not need private memory pools.
77 static struct bio_set
*fs_bio_set
;
79 static inline struct bio_vec
*bvec_alloc_bs(gfp_t gfp_mask
, int nr
, unsigned long *idx
, struct bio_set
*bs
)
84 * see comment near bvec_array define!
87 case 1 : *idx
= 0; break;
88 case 2 ... 4: *idx
= 1; break;
89 case 5 ... 16: *idx
= 2; break;
90 case 17 ... 64: *idx
= 3; break;
91 case 65 ... 128: *idx
= 4; break;
92 case 129 ... BIO_MAX_PAGES
: *idx
= 5; break;
97 * idx now points to the pool we want to allocate from
100 bvl
= mempool_alloc(bs
->bvec_pools
[*idx
], gfp_mask
);
102 struct biovec_slab
*bp
= bvec_slabs
+ *idx
;
104 memset(bvl
, 0, bp
->nr_vecs
* sizeof(struct bio_vec
));
110 void bio_free(struct bio
*bio
, struct bio_set
*bio_set
)
112 const int pool_idx
= BIO_POOL_IDX(bio
);
114 BIO_BUG_ON(pool_idx
>= BIOVEC_NR_POOLS
);
116 mempool_free(bio
->bi_io_vec
, bio_set
->bvec_pools
[pool_idx
]);
117 mempool_free(bio
, bio_set
->bio_pool
);
121 * default destructor for a bio allocated with bio_alloc_bioset()
123 static void bio_fs_destructor(struct bio
*bio
)
125 bio_free(bio
, fs_bio_set
);
128 void bio_init(struct bio
*bio
)
132 bio
->bi_flags
= 1 << BIO_UPTODATE
;
136 bio
->bi_phys_segments
= 0;
137 bio
->bi_hw_segments
= 0;
138 bio
->bi_hw_front_size
= 0;
139 bio
->bi_hw_back_size
= 0;
141 bio
->bi_max_vecs
= 0;
142 bio
->bi_end_io
= NULL
;
143 atomic_set(&bio
->bi_cnt
, 1);
144 bio
->bi_private
= NULL
;
148 * bio_alloc_bioset - allocate a bio for I/O
149 * @gfp_mask: the GFP_ mask given to the slab allocator
150 * @nr_iovecs: number of iovecs to pre-allocate
151 * @bs: the bio_set to allocate from
154 * bio_alloc_bioset will first try it's on mempool to satisfy the allocation.
155 * If %__GFP_WAIT is set then we will block on the internal pool waiting
156 * for a &struct bio to become free.
158 * allocate bio and iovecs from the memory pools specified by the
161 struct bio
*bio_alloc_bioset(gfp_t gfp_mask
, int nr_iovecs
, struct bio_set
*bs
)
163 struct bio
*bio
= mempool_alloc(bs
->bio_pool
, gfp_mask
);
166 struct bio_vec
*bvl
= NULL
;
169 if (likely(nr_iovecs
)) {
170 unsigned long idx
= 0; /* shut up gcc */
172 bvl
= bvec_alloc_bs(gfp_mask
, nr_iovecs
, &idx
, bs
);
173 if (unlikely(!bvl
)) {
174 mempool_free(bio
, bs
->bio_pool
);
178 bio
->bi_flags
|= idx
<< BIO_POOL_OFFSET
;
179 bio
->bi_max_vecs
= bvec_slabs
[idx
].nr_vecs
;
181 bio
->bi_io_vec
= bvl
;
187 struct bio
*bio_alloc(gfp_t gfp_mask
, int nr_iovecs
)
189 struct bio
*bio
= bio_alloc_bioset(gfp_mask
, nr_iovecs
, fs_bio_set
);
192 bio
->bi_destructor
= bio_fs_destructor
;
197 void zero_fill_bio(struct bio
*bio
)
203 bio_for_each_segment(bv
, bio
, i
) {
204 char *data
= bvec_kmap_irq(bv
, &flags
);
205 memset(data
, 0, bv
->bv_len
);
206 flush_dcache_page(bv
->bv_page
);
207 bvec_kunmap_irq(data
, &flags
);
210 EXPORT_SYMBOL(zero_fill_bio
);
213 * bio_put - release a reference to a bio
214 * @bio: bio to release reference to
217 * Put a reference to a &struct bio, either one you have gotten with
218 * bio_alloc or bio_get. The last put of a bio will free it.
220 void bio_put(struct bio
*bio
)
222 BIO_BUG_ON(!atomic_read(&bio
->bi_cnt
));
227 if (atomic_dec_and_test(&bio
->bi_cnt
)) {
229 bio
->bi_destructor(bio
);
233 inline int bio_phys_segments(request_queue_t
*q
, struct bio
*bio
)
235 if (unlikely(!bio_flagged(bio
, BIO_SEG_VALID
)))
236 blk_recount_segments(q
, bio
);
238 return bio
->bi_phys_segments
;
241 inline int bio_hw_segments(request_queue_t
*q
, struct bio
*bio
)
243 if (unlikely(!bio_flagged(bio
, BIO_SEG_VALID
)))
244 blk_recount_segments(q
, bio
);
246 return bio
->bi_hw_segments
;
250 * __bio_clone - clone a bio
251 * @bio: destination bio
252 * @bio_src: bio to clone
254 * Clone a &bio. Caller will own the returned bio, but not
255 * the actual data it points to. Reference count of returned
258 void __bio_clone(struct bio
*bio
, struct bio
*bio_src
)
260 request_queue_t
*q
= bdev_get_queue(bio_src
->bi_bdev
);
262 memcpy(bio
->bi_io_vec
, bio_src
->bi_io_vec
,
263 bio_src
->bi_max_vecs
* sizeof(struct bio_vec
));
265 bio
->bi_sector
= bio_src
->bi_sector
;
266 bio
->bi_bdev
= bio_src
->bi_bdev
;
267 bio
->bi_flags
|= 1 << BIO_CLONED
;
268 bio
->bi_rw
= bio_src
->bi_rw
;
269 bio
->bi_vcnt
= bio_src
->bi_vcnt
;
270 bio
->bi_size
= bio_src
->bi_size
;
271 bio
->bi_idx
= bio_src
->bi_idx
;
272 bio_phys_segments(q
, bio
);
273 bio_hw_segments(q
, bio
);
277 * bio_clone - clone a bio
279 * @gfp_mask: allocation priority
281 * Like __bio_clone, only also allocates the returned bio
283 struct bio
*bio_clone(struct bio
*bio
, gfp_t gfp_mask
)
285 struct bio
*b
= bio_alloc_bioset(gfp_mask
, bio
->bi_max_vecs
, fs_bio_set
);
288 b
->bi_destructor
= bio_fs_destructor
;
296 * bio_get_nr_vecs - return approx number of vecs
299 * Return the approximate number of pages we can send to this target.
300 * There's no guarantee that you will be able to fit this number of pages
301 * into a bio, it does not account for dynamic restrictions that vary
304 int bio_get_nr_vecs(struct block_device
*bdev
)
306 request_queue_t
*q
= bdev_get_queue(bdev
);
309 nr_pages
= ((q
->max_sectors
<< 9) + PAGE_SIZE
- 1) >> PAGE_SHIFT
;
310 if (nr_pages
> q
->max_phys_segments
)
311 nr_pages
= q
->max_phys_segments
;
312 if (nr_pages
> q
->max_hw_segments
)
313 nr_pages
= q
->max_hw_segments
;
318 static int __bio_add_page(request_queue_t
*q
, struct bio
*bio
, struct page
319 *page
, unsigned int len
, unsigned int offset
,
320 unsigned short max_sectors
)
322 int retried_segments
= 0;
323 struct bio_vec
*bvec
;
326 * cloned bio must not modify vec list
328 if (unlikely(bio_flagged(bio
, BIO_CLONED
)))
331 if (((bio
->bi_size
+ len
) >> 9) > max_sectors
)
335 * For filesystems with a blocksize smaller than the pagesize
336 * we will often be called with the same page as last time and
337 * a consecutive offset. Optimize this special case.
339 if (bio
->bi_vcnt
> 0) {
340 struct bio_vec
*prev
= &bio
->bi_io_vec
[bio
->bi_vcnt
- 1];
342 if (page
== prev
->bv_page
&&
343 offset
== prev
->bv_offset
+ prev
->bv_len
) {
345 if (q
->merge_bvec_fn
&&
346 q
->merge_bvec_fn(q
, bio
, prev
) < len
) {
355 if (bio
->bi_vcnt
>= bio
->bi_max_vecs
)
359 * we might lose a segment or two here, but rather that than
360 * make this too complex.
363 while (bio
->bi_phys_segments
>= q
->max_phys_segments
364 || bio
->bi_hw_segments
>= q
->max_hw_segments
365 || BIOVEC_VIRT_OVERSIZE(bio
->bi_size
)) {
367 if (retried_segments
)
370 retried_segments
= 1;
371 blk_recount_segments(q
, bio
);
375 * setup the new entry, we might clear it again later if we
376 * cannot add the page
378 bvec
= &bio
->bi_io_vec
[bio
->bi_vcnt
];
379 bvec
->bv_page
= page
;
381 bvec
->bv_offset
= offset
;
384 * if queue has other restrictions (eg varying max sector size
385 * depending on offset), it can specify a merge_bvec_fn in the
386 * queue to get further control
388 if (q
->merge_bvec_fn
) {
390 * merge_bvec_fn() returns number of bytes it can accept
393 if (q
->merge_bvec_fn(q
, bio
, bvec
) < len
) {
394 bvec
->bv_page
= NULL
;
401 /* If we may be able to merge these biovecs, force a recount */
402 if (bio
->bi_vcnt
&& (BIOVEC_PHYS_MERGEABLE(bvec
-1, bvec
) ||
403 BIOVEC_VIRT_MERGEABLE(bvec
-1, bvec
)))
404 bio
->bi_flags
&= ~(1 << BIO_SEG_VALID
);
407 bio
->bi_phys_segments
++;
408 bio
->bi_hw_segments
++;
415 * bio_add_pc_page - attempt to add page to bio
416 * @q: the target queue
417 * @bio: destination bio
419 * @len: vec entry length
420 * @offset: vec entry offset
422 * Attempt to add a page to the bio_vec maplist. This can fail for a
423 * number of reasons, such as the bio being full or target block
424 * device limitations. The target block device must allow bio's
425 * smaller than PAGE_SIZE, so it is always possible to add a single
426 * page to an empty bio. This should only be used by REQ_PC bios.
428 int bio_add_pc_page(request_queue_t
*q
, struct bio
*bio
, struct page
*page
,
429 unsigned int len
, unsigned int offset
)
431 return __bio_add_page(q
, bio
, page
, len
, offset
, q
->max_hw_sectors
);
435 * bio_add_page - attempt to add page to bio
436 * @bio: destination bio
438 * @len: vec entry length
439 * @offset: vec entry offset
441 * Attempt to add a page to the bio_vec maplist. This can fail for a
442 * number of reasons, such as the bio being full or target block
443 * device limitations. The target block device must allow bio's
444 * smaller than PAGE_SIZE, so it is always possible to add a single
445 * page to an empty bio.
447 int bio_add_page(struct bio
*bio
, struct page
*page
, unsigned int len
,
450 struct request_queue
*q
= bdev_get_queue(bio
->bi_bdev
);
451 return __bio_add_page(q
, bio
, page
, len
, offset
, q
->max_sectors
);
454 struct bio_map_data
{
455 struct bio_vec
*iovecs
;
456 void __user
*userptr
;
459 static void bio_set_map_data(struct bio_map_data
*bmd
, struct bio
*bio
)
461 memcpy(bmd
->iovecs
, bio
->bi_io_vec
, sizeof(struct bio_vec
) * bio
->bi_vcnt
);
462 bio
->bi_private
= bmd
;
465 static void bio_free_map_data(struct bio_map_data
*bmd
)
471 static struct bio_map_data
*bio_alloc_map_data(int nr_segs
)
473 struct bio_map_data
*bmd
= kmalloc(sizeof(*bmd
), GFP_KERNEL
);
478 bmd
->iovecs
= kmalloc(sizeof(struct bio_vec
) * nr_segs
, GFP_KERNEL
);
487 * bio_uncopy_user - finish previously mapped bio
488 * @bio: bio being terminated
490 * Free pages allocated from bio_copy_user() and write back data
491 * to user space in case of a read.
493 int bio_uncopy_user(struct bio
*bio
)
495 struct bio_map_data
*bmd
= bio
->bi_private
;
496 const int read
= bio_data_dir(bio
) == READ
;
497 struct bio_vec
*bvec
;
500 __bio_for_each_segment(bvec
, bio
, i
, 0) {
501 char *addr
= page_address(bvec
->bv_page
);
502 unsigned int len
= bmd
->iovecs
[i
].bv_len
;
504 if (read
&& !ret
&& copy_to_user(bmd
->userptr
, addr
, len
))
507 __free_page(bvec
->bv_page
);
510 bio_free_map_data(bmd
);
516 * bio_copy_user - copy user data to bio
517 * @q: destination block queue
518 * @uaddr: start of user address
519 * @len: length in bytes
520 * @write_to_vm: bool indicating writing to pages or not
522 * Prepares and returns a bio for indirect user io, bouncing data
523 * to/from kernel pages as necessary. Must be paired with
524 * call bio_uncopy_user() on io completion.
526 struct bio
*bio_copy_user(request_queue_t
*q
, unsigned long uaddr
,
527 unsigned int len
, int write_to_vm
)
529 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
530 unsigned long start
= uaddr
>> PAGE_SHIFT
;
531 struct bio_map_data
*bmd
;
532 struct bio_vec
*bvec
;
537 bmd
= bio_alloc_map_data(end
- start
);
539 return ERR_PTR(-ENOMEM
);
541 bmd
->userptr
= (void __user
*) uaddr
;
544 bio
= bio_alloc(GFP_KERNEL
, end
- start
);
548 bio
->bi_rw
|= (!write_to_vm
<< BIO_RW
);
552 unsigned int bytes
= PAGE_SIZE
;
557 page
= alloc_page(q
->bounce_gfp
| GFP_KERNEL
);
563 if (bio_add_pc_page(q
, bio
, page
, bytes
, 0) < bytes
)
576 char __user
*p
= (char __user
*) uaddr
;
579 * for a write, copy in data to kernel pages
582 bio_for_each_segment(bvec
, bio
, i
) {
583 char *addr
= page_address(bvec
->bv_page
);
585 if (copy_from_user(addr
, p
, bvec
->bv_len
))
591 bio_set_map_data(bmd
, bio
);
594 bio_for_each_segment(bvec
, bio
, i
)
595 __free_page(bvec
->bv_page
);
599 bio_free_map_data(bmd
);
603 static struct bio
*__bio_map_user_iov(request_queue_t
*q
,
604 struct block_device
*bdev
,
605 struct sg_iovec
*iov
, int iov_count
,
615 for (i
= 0; i
< iov_count
; i
++) {
616 unsigned long uaddr
= (unsigned long)iov
[i
].iov_base
;
617 unsigned long len
= iov
[i
].iov_len
;
618 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
619 unsigned long start
= uaddr
>> PAGE_SHIFT
;
621 nr_pages
+= end
- start
;
623 * buffer must be aligned to at least hardsector size for now
625 if (uaddr
& queue_dma_alignment(q
))
626 return ERR_PTR(-EINVAL
);
630 return ERR_PTR(-EINVAL
);
632 bio
= bio_alloc(GFP_KERNEL
, nr_pages
);
634 return ERR_PTR(-ENOMEM
);
637 pages
= kcalloc(nr_pages
, sizeof(struct page
*), GFP_KERNEL
);
641 for (i
= 0; i
< iov_count
; i
++) {
642 unsigned long uaddr
= (unsigned long)iov
[i
].iov_base
;
643 unsigned long len
= iov
[i
].iov_len
;
644 unsigned long end
= (uaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
645 unsigned long start
= uaddr
>> PAGE_SHIFT
;
646 const int local_nr_pages
= end
- start
;
647 const int page_limit
= cur_page
+ local_nr_pages
;
649 down_read(¤t
->mm
->mmap_sem
);
650 ret
= get_user_pages(current
, current
->mm
, uaddr
,
652 write_to_vm
, 0, &pages
[cur_page
], NULL
);
653 up_read(¤t
->mm
->mmap_sem
);
655 if (ret
< local_nr_pages
) {
660 offset
= uaddr
& ~PAGE_MASK
;
661 for (j
= cur_page
; j
< page_limit
; j
++) {
662 unsigned int bytes
= PAGE_SIZE
- offset
;
673 if (bio_add_pc_page(q
, bio
, pages
[j
], bytes
, offset
) <
683 * release the pages we didn't map into the bio, if any
685 while (j
< page_limit
)
686 page_cache_release(pages
[j
++]);
692 * set data direction, and check if mapped pages need bouncing
695 bio
->bi_rw
|= (1 << BIO_RW
);
698 bio
->bi_flags
|= (1 << BIO_USER_MAPPED
);
702 for (i
= 0; i
< nr_pages
; i
++) {
705 page_cache_release(pages
[i
]);
714 * bio_map_user - map user address into bio
715 * @q: the request_queue_t for the bio
716 * @bdev: destination block device
717 * @uaddr: start of user address
718 * @len: length in bytes
719 * @write_to_vm: bool indicating writing to pages or not
721 * Map the user space address into a bio suitable for io to a block
722 * device. Returns an error pointer in case of error.
724 struct bio
*bio_map_user(request_queue_t
*q
, struct block_device
*bdev
,
725 unsigned long uaddr
, unsigned int len
, int write_to_vm
)
729 iov
.iov_base
= (void __user
*)uaddr
;
732 return bio_map_user_iov(q
, bdev
, &iov
, 1, write_to_vm
);
736 * bio_map_user_iov - map user sg_iovec table into bio
737 * @q: the request_queue_t for the bio
738 * @bdev: destination block device
740 * @iov_count: number of elements in the iovec
741 * @write_to_vm: bool indicating writing to pages or not
743 * Map the user space address into a bio suitable for io to a block
744 * device. Returns an error pointer in case of error.
746 struct bio
*bio_map_user_iov(request_queue_t
*q
, struct block_device
*bdev
,
747 struct sg_iovec
*iov
, int iov_count
,
752 bio
= __bio_map_user_iov(q
, bdev
, iov
, iov_count
, write_to_vm
);
758 * subtle -- if __bio_map_user() ended up bouncing a bio,
759 * it would normally disappear when its bi_end_io is run.
760 * however, we need it for the unmap, so grab an extra
768 static void __bio_unmap_user(struct bio
*bio
)
770 struct bio_vec
*bvec
;
774 * make sure we dirty pages we wrote to
776 __bio_for_each_segment(bvec
, bio
, i
, 0) {
777 if (bio_data_dir(bio
) == READ
)
778 set_page_dirty_lock(bvec
->bv_page
);
780 page_cache_release(bvec
->bv_page
);
787 * bio_unmap_user - unmap a bio
788 * @bio: the bio being unmapped
790 * Unmap a bio previously mapped by bio_map_user(). Must be called with
793 * bio_unmap_user() may sleep.
795 void bio_unmap_user(struct bio
*bio
)
797 __bio_unmap_user(bio
);
801 static int bio_map_kern_endio(struct bio
*bio
, unsigned int bytes_done
, int err
)
811 static struct bio
*__bio_map_kern(request_queue_t
*q
, void *data
,
812 unsigned int len
, gfp_t gfp_mask
)
814 unsigned long kaddr
= (unsigned long)data
;
815 unsigned long end
= (kaddr
+ len
+ PAGE_SIZE
- 1) >> PAGE_SHIFT
;
816 unsigned long start
= kaddr
>> PAGE_SHIFT
;
817 const int nr_pages
= end
- start
;
821 bio
= bio_alloc(gfp_mask
, nr_pages
);
823 return ERR_PTR(-ENOMEM
);
825 offset
= offset_in_page(kaddr
);
826 for (i
= 0; i
< nr_pages
; i
++) {
827 unsigned int bytes
= PAGE_SIZE
- offset
;
835 if (bio_add_pc_page(q
, bio
, virt_to_page(data
), bytes
,
844 bio
->bi_end_io
= bio_map_kern_endio
;
849 * bio_map_kern - map kernel address into bio
850 * @q: the request_queue_t for the bio
851 * @data: pointer to buffer to map
852 * @len: length in bytes
853 * @gfp_mask: allocation flags for bio allocation
855 * Map the kernel address into a bio suitable for io to a block
856 * device. Returns an error pointer in case of error.
858 struct bio
*bio_map_kern(request_queue_t
*q
, void *data
, unsigned int len
,
863 bio
= __bio_map_kern(q
, data
, len
, gfp_mask
);
867 if (bio
->bi_size
== len
)
871 * Don't support partial mappings.
874 return ERR_PTR(-EINVAL
);
878 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
879 * for performing direct-IO in BIOs.
881 * The problem is that we cannot run set_page_dirty() from interrupt context
882 * because the required locks are not interrupt-safe. So what we can do is to
883 * mark the pages dirty _before_ performing IO. And in interrupt context,
884 * check that the pages are still dirty. If so, fine. If not, redirty them
885 * in process context.
887 * We special-case compound pages here: normally this means reads into hugetlb
888 * pages. The logic in here doesn't really work right for compound pages
889 * because the VM does not uniformly chase down the head page in all cases.
890 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
891 * handle them at all. So we skip compound pages here at an early stage.
893 * Note that this code is very hard to test under normal circumstances because
894 * direct-io pins the pages with get_user_pages(). This makes
895 * is_page_cache_freeable return false, and the VM will not clean the pages.
896 * But other code (eg, pdflush) could clean the pages if they are mapped
899 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
900 * deferred bio dirtying paths.
904 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
906 void bio_set_pages_dirty(struct bio
*bio
)
908 struct bio_vec
*bvec
= bio
->bi_io_vec
;
911 for (i
= 0; i
< bio
->bi_vcnt
; i
++) {
912 struct page
*page
= bvec
[i
].bv_page
;
914 if (page
&& !PageCompound(page
))
915 set_page_dirty_lock(page
);
919 static void bio_release_pages(struct bio
*bio
)
921 struct bio_vec
*bvec
= bio
->bi_io_vec
;
924 for (i
= 0; i
< bio
->bi_vcnt
; i
++) {
925 struct page
*page
= bvec
[i
].bv_page
;
933 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
934 * If they are, then fine. If, however, some pages are clean then they must
935 * have been written out during the direct-IO read. So we take another ref on
936 * the BIO and the offending pages and re-dirty the pages in process context.
938 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
939 * here on. It will run one page_cache_release() against each page and will
940 * run one bio_put() against the BIO.
943 static void bio_dirty_fn(struct work_struct
*work
);
945 static DECLARE_WORK(bio_dirty_work
, bio_dirty_fn
);
946 static DEFINE_SPINLOCK(bio_dirty_lock
);
947 static struct bio
*bio_dirty_list
;
950 * This runs in process context
952 static void bio_dirty_fn(struct work_struct
*work
)
957 spin_lock_irqsave(&bio_dirty_lock
, flags
);
958 bio
= bio_dirty_list
;
959 bio_dirty_list
= NULL
;
960 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
963 struct bio
*next
= bio
->bi_private
;
965 bio_set_pages_dirty(bio
);
966 bio_release_pages(bio
);
972 void bio_check_pages_dirty(struct bio
*bio
)
974 struct bio_vec
*bvec
= bio
->bi_io_vec
;
975 int nr_clean_pages
= 0;
978 for (i
= 0; i
< bio
->bi_vcnt
; i
++) {
979 struct page
*page
= bvec
[i
].bv_page
;
981 if (PageDirty(page
) || PageCompound(page
)) {
982 page_cache_release(page
);
983 bvec
[i
].bv_page
= NULL
;
989 if (nr_clean_pages
) {
992 spin_lock_irqsave(&bio_dirty_lock
, flags
);
993 bio
->bi_private
= bio_dirty_list
;
994 bio_dirty_list
= bio
;
995 spin_unlock_irqrestore(&bio_dirty_lock
, flags
);
996 schedule_work(&bio_dirty_work
);
1003 * bio_endio - end I/O on a bio
1005 * @bytes_done: number of bytes completed
1006 * @error: error, if any
1009 * bio_endio() will end I/O on @bytes_done number of bytes. This may be
1010 * just a partial part of the bio, or it may be the whole bio. bio_endio()
1011 * is the preferred way to end I/O on a bio, it takes care of decrementing
1012 * bi_size and clearing BIO_UPTODATE on error. @error is 0 on success, and
1013 * and one of the established -Exxxx (-EIO, for instance) error values in
1014 * case something went wrong. Noone should call bi_end_io() directly on
1015 * a bio unless they own it and thus know that it has an end_io function.
1017 void bio_endio(struct bio
*bio
, unsigned int bytes_done
, int error
)
1020 clear_bit(BIO_UPTODATE
, &bio
->bi_flags
);
1022 if (unlikely(bytes_done
> bio
->bi_size
)) {
1023 printk("%s: want %u bytes done, only %u left\n", __FUNCTION__
,
1024 bytes_done
, bio
->bi_size
);
1025 bytes_done
= bio
->bi_size
;
1028 bio
->bi_size
-= bytes_done
;
1029 bio
->bi_sector
+= (bytes_done
>> 9);
1032 bio
->bi_end_io(bio
, bytes_done
, error
);
1035 void bio_pair_release(struct bio_pair
*bp
)
1037 if (atomic_dec_and_test(&bp
->cnt
)) {
1038 struct bio
*master
= bp
->bio1
.bi_private
;
1040 bio_endio(master
, master
->bi_size
, bp
->error
);
1041 mempool_free(bp
, bp
->bio2
.bi_private
);
1045 static int bio_pair_end_1(struct bio
* bi
, unsigned int done
, int err
)
1047 struct bio_pair
*bp
= container_of(bi
, struct bio_pair
, bio1
);
1055 bio_pair_release(bp
);
1059 static int bio_pair_end_2(struct bio
* bi
, unsigned int done
, int err
)
1061 struct bio_pair
*bp
= container_of(bi
, struct bio_pair
, bio2
);
1069 bio_pair_release(bp
);
1074 * split a bio - only worry about a bio with a single page
1077 struct bio_pair
*bio_split(struct bio
*bi
, mempool_t
*pool
, int first_sectors
)
1079 struct bio_pair
*bp
= mempool_alloc(pool
, GFP_NOIO
);
1084 blk_add_trace_pdu_int(bdev_get_queue(bi
->bi_bdev
), BLK_TA_SPLIT
, bi
,
1085 bi
->bi_sector
+ first_sectors
);
1087 BUG_ON(bi
->bi_vcnt
!= 1);
1088 BUG_ON(bi
->bi_idx
!= 0);
1089 atomic_set(&bp
->cnt
, 3);
1093 bp
->bio2
.bi_sector
+= first_sectors
;
1094 bp
->bio2
.bi_size
-= first_sectors
<< 9;
1095 bp
->bio1
.bi_size
= first_sectors
<< 9;
1097 bp
->bv1
= bi
->bi_io_vec
[0];
1098 bp
->bv2
= bi
->bi_io_vec
[0];
1099 bp
->bv2
.bv_offset
+= first_sectors
<< 9;
1100 bp
->bv2
.bv_len
-= first_sectors
<< 9;
1101 bp
->bv1
.bv_len
= first_sectors
<< 9;
1103 bp
->bio1
.bi_io_vec
= &bp
->bv1
;
1104 bp
->bio2
.bi_io_vec
= &bp
->bv2
;
1106 bp
->bio1
.bi_max_vecs
= 1;
1107 bp
->bio2
.bi_max_vecs
= 1;
1109 bp
->bio1
.bi_end_io
= bio_pair_end_1
;
1110 bp
->bio2
.bi_end_io
= bio_pair_end_2
;
1112 bp
->bio1
.bi_private
= bi
;
1113 bp
->bio2
.bi_private
= pool
;
1120 * create memory pools for biovec's in a bio_set.
1121 * use the global biovec slabs created for general use.
1123 static int biovec_create_pools(struct bio_set
*bs
, int pool_entries
, int scale
)
1127 for (i
= 0; i
< BIOVEC_NR_POOLS
; i
++) {
1128 struct biovec_slab
*bp
= bvec_slabs
+ i
;
1129 mempool_t
**bvp
= bs
->bvec_pools
+ i
;
1131 if (pool_entries
> 1 && i
>= scale
)
1134 *bvp
= mempool_create_slab_pool(pool_entries
, bp
->slab
);
1141 static void biovec_free_pools(struct bio_set
*bs
)
1145 for (i
= 0; i
< BIOVEC_NR_POOLS
; i
++) {
1146 mempool_t
*bvp
= bs
->bvec_pools
[i
];
1149 mempool_destroy(bvp
);
1154 void bioset_free(struct bio_set
*bs
)
1157 mempool_destroy(bs
->bio_pool
);
1159 biovec_free_pools(bs
);
1164 struct bio_set
*bioset_create(int bio_pool_size
, int bvec_pool_size
, int scale
)
1166 struct bio_set
*bs
= kzalloc(sizeof(*bs
), GFP_KERNEL
);
1171 bs
->bio_pool
= mempool_create_slab_pool(bio_pool_size
, bio_slab
);
1175 if (!biovec_create_pools(bs
, bvec_pool_size
, scale
))
1183 static void __init
biovec_init_slabs(void)
1187 for (i
= 0; i
< BIOVEC_NR_POOLS
; i
++) {
1189 struct biovec_slab
*bvs
= bvec_slabs
+ i
;
1191 size
= bvs
->nr_vecs
* sizeof(struct bio_vec
);
1192 bvs
->slab
= kmem_cache_create(bvs
->name
, size
, 0,
1193 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
, NULL
);
1197 static int __init
init_bio(void)
1199 int megabytes
, bvec_pool_entries
;
1200 int scale
= BIOVEC_NR_POOLS
;
1202 bio_slab
= kmem_cache_create("bio", sizeof(struct bio
), 0,
1203 SLAB_HWCACHE_ALIGN
|SLAB_PANIC
, NULL
, NULL
);
1205 biovec_init_slabs();
1207 megabytes
= nr_free_pages() >> (20 - PAGE_SHIFT
);
1210 * find out where to start scaling
1212 if (megabytes
<= 16)
1214 else if (megabytes
<= 32)
1216 else if (megabytes
<= 64)
1218 else if (megabytes
<= 96)
1220 else if (megabytes
<= 128)
1224 * Limit number of entries reserved -- mempools are only used when
1225 * the system is completely unable to allocate memory, so we only
1226 * need enough to make progress.
1228 bvec_pool_entries
= 1 + scale
;
1230 fs_bio_set
= bioset_create(BIO_POOL_SIZE
, bvec_pool_entries
, scale
);
1232 panic("bio: can't allocate bios\n");
1234 bio_split_pool
= mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES
,
1235 sizeof(struct bio_pair
));
1236 if (!bio_split_pool
)
1237 panic("bio: can't create split pool\n");
1242 subsys_initcall(init_bio
);
1244 EXPORT_SYMBOL(bio_alloc
);
1245 EXPORT_SYMBOL(bio_put
);
1246 EXPORT_SYMBOL(bio_free
);
1247 EXPORT_SYMBOL(bio_endio
);
1248 EXPORT_SYMBOL(bio_init
);
1249 EXPORT_SYMBOL(__bio_clone
);
1250 EXPORT_SYMBOL(bio_clone
);
1251 EXPORT_SYMBOL(bio_phys_segments
);
1252 EXPORT_SYMBOL(bio_hw_segments
);
1253 EXPORT_SYMBOL(bio_add_page
);
1254 EXPORT_SYMBOL(bio_add_pc_page
);
1255 EXPORT_SYMBOL(bio_get_nr_vecs
);
1256 EXPORT_SYMBOL(bio_map_user
);
1257 EXPORT_SYMBOL(bio_unmap_user
);
1258 EXPORT_SYMBOL(bio_map_kern
);
1259 EXPORT_SYMBOL(bio_pair_release
);
1260 EXPORT_SYMBOL(bio_split
);
1261 EXPORT_SYMBOL(bio_split_pool
);
1262 EXPORT_SYMBOL(bio_copy_user
);
1263 EXPORT_SYMBOL(bio_uncopy_user
);
1264 EXPORT_SYMBOL(bioset_create
);
1265 EXPORT_SYMBOL(bioset_free
);
1266 EXPORT_SYMBOL(bio_alloc_bioset
);