ia64/linux-2.6.18-xen.hg

view fs/bio.c @ 452:c7ed6fe5dca0

kexec: dont initialise regions in reserve_memory()

There is no need to initialise efi_memmap_res and boot_param_res in
reserve_memory() for the initial xen domain as it is done in
machine_kexec_setup_resources() using values from the kexec hypercall.

Signed-off-by: Simon Horman <horms@verge.net.au>
author Keir Fraser <keir.fraser@citrix.com>
date Thu Feb 28 10:55:18 2008 +0000 (2008-02-28)
parents 7128fe32720e
children
line source
1 /*
2 * Copyright (C) 2001 Jens Axboe <axboe@suse.de>
3 *
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.
7 *
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.
12 *
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-
16 *
17 */
18 #include <linux/mm.h>
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 kmem_cache_t *bio_slab __read_mostly;
35 #define BIOVEC_NR_POOLS 6
37 /*
38 * a small number of entries is fine, not going to be performance critical.
39 * basically we just need to survive
40 */
41 #define BIO_SPLIT_ENTRIES 8
42 mempool_t *bio_split_pool __read_mostly;
44 struct biovec_slab {
45 int nr_vecs;
46 char *name;
47 kmem_cache_t *slab;
48 };
50 /*
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
53 * unsigned short
54 */
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),
59 };
60 #undef BV
62 /*
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[].
67 */
68 struct bio_set {
69 mempool_t *bio_pool;
70 mempool_t *bvec_pools[BIOVEC_NR_POOLS];
71 };
73 /*
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.
76 */
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)
80 {
81 struct bio_vec *bvl;
82 struct biovec_slab *bp;
84 /*
85 * see comment near bvec_array define!
86 */
87 switch (nr) {
88 case 1 : *idx = 0; break;
89 case 2 ... 4: *idx = 1; break;
90 case 5 ... 16: *idx = 2; break;
91 case 17 ... 64: *idx = 3; break;
92 case 65 ... 128: *idx = 4; break;
93 case 129 ... BIO_MAX_PAGES: *idx = 5; break;
94 default:
95 return NULL;
96 }
97 /*
98 * idx now points to the pool we want to allocate from
99 */
101 bp = bvec_slabs + *idx;
102 bvl = mempool_alloc(bs->bvec_pools[*idx], gfp_mask);
103 if (bvl)
104 memset(bvl, 0, bp->nr_vecs * sizeof(struct bio_vec));
106 return bvl;
107 }
109 void bio_free(struct bio *bio, struct bio_set *bio_set)
110 {
111 const int pool_idx = BIO_POOL_IDX(bio);
113 BIO_BUG_ON(pool_idx >= BIOVEC_NR_POOLS);
115 if (bio->bi_io_vec)
116 mempool_free(bio->bi_io_vec, bio_set->bvec_pools[pool_idx]);
117 mempool_free(bio, bio_set->bio_pool);
118 }
120 /*
121 * default destructor for a bio allocated with bio_alloc_bioset()
122 */
123 static void bio_fs_destructor(struct bio *bio)
124 {
125 bio_free(bio, fs_bio_set);
126 }
128 void bio_init(struct bio *bio)
129 {
130 bio->bi_next = NULL;
131 bio->bi_bdev = NULL;
132 bio->bi_flags = 1 << BIO_UPTODATE;
133 bio->bi_rw = 0;
134 bio->bi_vcnt = 0;
135 bio->bi_idx = 0;
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;
140 bio->bi_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;
145 }
147 /**
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
152 *
153 * Description:
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.
157 *
158 * allocate bio and iovecs from the memory pools specified by the
159 * bio_set structure.
160 **/
161 struct bio *bio_alloc_bioset(gfp_t gfp_mask, int nr_iovecs, struct bio_set *bs)
162 {
163 struct bio *bio = mempool_alloc(bs->bio_pool, gfp_mask);
165 if (likely(bio)) {
166 struct bio_vec *bvl = NULL;
168 bio_init(bio);
169 if (likely(nr_iovecs)) {
170 unsigned long idx;
172 bvl = bvec_alloc_bs(gfp_mask, nr_iovecs, &idx, bs);
173 if (unlikely(!bvl)) {
174 mempool_free(bio, bs->bio_pool);
175 bio = NULL;
176 goto out;
177 }
178 bio->bi_flags |= idx << BIO_POOL_OFFSET;
179 bio->bi_max_vecs = bvec_slabs[idx].nr_vecs;
180 }
181 bio->bi_io_vec = bvl;
182 }
183 out:
184 return bio;
185 }
187 struct bio *bio_alloc(gfp_t gfp_mask, int nr_iovecs)
188 {
189 struct bio *bio = bio_alloc_bioset(gfp_mask, nr_iovecs, fs_bio_set);
191 if (bio)
192 bio->bi_destructor = bio_fs_destructor;
194 return bio;
195 }
197 void zero_fill_bio(struct bio *bio)
198 {
199 unsigned long flags;
200 struct bio_vec *bv;
201 int i;
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);
208 }
209 }
210 EXPORT_SYMBOL(zero_fill_bio);
212 /**
213 * bio_put - release a reference to a bio
214 * @bio: bio to release reference to
215 *
216 * Description:
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.
219 **/
220 void bio_put(struct bio *bio)
221 {
222 BIO_BUG_ON(!atomic_read(&bio->bi_cnt));
224 /*
225 * last put frees it
226 */
227 if (atomic_dec_and_test(&bio->bi_cnt)) {
228 bio->bi_next = NULL;
229 bio->bi_destructor(bio);
230 }
231 }
233 inline int bio_phys_segments(request_queue_t *q, struct bio *bio)
234 {
235 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
236 blk_recount_segments(q, bio);
238 return bio->bi_phys_segments;
239 }
241 inline int bio_hw_segments(request_queue_t *q, struct bio *bio)
242 {
243 if (unlikely(!bio_flagged(bio, BIO_SEG_VALID)))
244 blk_recount_segments(q, bio);
246 return bio->bi_hw_segments;
247 }
249 /**
250 * __bio_clone - clone a bio
251 * @bio: destination bio
252 * @bio_src: bio to clone
253 *
254 * Clone a &bio. Caller will own the returned bio, but not
255 * the actual data it points to. Reference count of returned
256 * bio will be one.
257 */
258 void __bio_clone(struct bio *bio, struct bio *bio_src)
259 {
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);
274 }
276 /**
277 * bio_clone - clone a bio
278 * @bio: bio to clone
279 * @gfp_mask: allocation priority
280 *
281 * Like __bio_clone, only also allocates the returned bio
282 */
283 struct bio *bio_clone(struct bio *bio, gfp_t gfp_mask)
284 {
285 struct bio *b = bio_alloc_bioset(gfp_mask, bio->bi_max_vecs, fs_bio_set);
287 if (b) {
288 b->bi_destructor = bio_fs_destructor;
289 __bio_clone(b, bio);
290 }
292 return b;
293 }
295 /**
296 * bio_get_nr_vecs - return approx number of vecs
297 * @bdev: I/O target
298 *
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
302 * on offset.
303 */
304 int bio_get_nr_vecs(struct block_device *bdev)
305 {
306 request_queue_t *q = bdev_get_queue(bdev);
307 int nr_pages;
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;
315 return nr_pages;
316 }
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)
321 {
322 int retried_segments = 0;
323 struct bio_vec *bvec;
325 /*
326 * cloned bio must not modify vec list
327 */
328 if (unlikely(bio_flagged(bio, BIO_CLONED)))
329 return 0;
331 if (((bio->bi_size + len) >> 9) > max_sectors)
332 return 0;
334 /*
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.
338 */
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) {
344 prev->bv_len += len;
345 if (q->merge_bvec_fn &&
346 q->merge_bvec_fn(q, bio, prev) < len) {
347 prev->bv_len -= len;
348 return 0;
349 }
351 goto done;
352 }
353 }
355 if (bio->bi_vcnt >= bio->bi_max_vecs)
356 return 0;
358 /*
359 * we might lose a segment or two here, but rather that than
360 * make this too complex.
361 */
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)
368 return 0;
370 retried_segments = 1;
371 blk_recount_segments(q, bio);
372 }
374 /*
375 * setup the new entry, we might clear it again later if we
376 * cannot add the page
377 */
378 bvec = &bio->bi_io_vec[bio->bi_vcnt];
379 bvec->bv_page = page;
380 bvec->bv_len = len;
381 bvec->bv_offset = offset;
383 /*
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
387 */
388 if (q->merge_bvec_fn) {
389 /*
390 * merge_bvec_fn() returns number of bytes it can accept
391 * at this offset
392 */
393 if (q->merge_bvec_fn(q, bio, bvec) < len) {
394 bvec->bv_page = NULL;
395 bvec->bv_len = 0;
396 bvec->bv_offset = 0;
397 return 0;
398 }
399 }
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);
406 bio->bi_vcnt++;
407 bio->bi_phys_segments++;
408 bio->bi_hw_segments++;
409 done:
410 bio->bi_size += len;
411 return len;
412 }
414 /**
415 * bio_add_pc_page - attempt to add page to bio
416 * @q: the target queue
417 * @bio: destination bio
418 * @page: page to add
419 * @len: vec entry length
420 * @offset: vec entry offset
421 *
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.
427 */
428 int bio_add_pc_page(request_queue_t *q, struct bio *bio, struct page *page,
429 unsigned int len, unsigned int offset)
430 {
431 return __bio_add_page(q, bio, page, len, offset, q->max_hw_sectors);
432 }
434 /**
435 * bio_add_page - attempt to add page to bio
436 * @bio: destination bio
437 * @page: page to add
438 * @len: vec entry length
439 * @offset: vec entry offset
440 *
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.
446 */
447 int bio_add_page(struct bio *bio, struct page *page, unsigned int len,
448 unsigned int offset)
449 {
450 struct request_queue *q = bdev_get_queue(bio->bi_bdev);
451 return __bio_add_page(q, bio, page, len, offset, q->max_sectors);
452 }
454 struct bio_map_data {
455 struct bio_vec *iovecs;
456 void __user *userptr;
457 };
459 static void bio_set_map_data(struct bio_map_data *bmd, struct bio *bio)
460 {
461 memcpy(bmd->iovecs, bio->bi_io_vec, sizeof(struct bio_vec) * bio->bi_vcnt);
462 bio->bi_private = bmd;
463 }
465 static void bio_free_map_data(struct bio_map_data *bmd)
466 {
467 kfree(bmd->iovecs);
468 kfree(bmd);
469 }
471 static struct bio_map_data *bio_alloc_map_data(int nr_segs)
472 {
473 struct bio_map_data *bmd = kmalloc(sizeof(*bmd), GFP_KERNEL);
475 if (!bmd)
476 return NULL;
478 bmd->iovecs = kmalloc(sizeof(struct bio_vec) * nr_segs, GFP_KERNEL);
479 if (bmd->iovecs)
480 return bmd;
482 kfree(bmd);
483 return NULL;
484 }
486 /**
487 * bio_uncopy_user - finish previously mapped bio
488 * @bio: bio being terminated
489 *
490 * Free pages allocated from bio_copy_user() and write back data
491 * to user space in case of a read.
492 */
493 int bio_uncopy_user(struct bio *bio)
494 {
495 struct bio_map_data *bmd = bio->bi_private;
496 const int read = bio_data_dir(bio) == READ;
497 struct bio_vec *bvec;
498 int i, ret = 0;
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))
505 ret = -EFAULT;
507 __free_page(bvec->bv_page);
508 bmd->userptr += len;
509 }
510 bio_free_map_data(bmd);
511 bio_put(bio);
512 return ret;
513 }
515 /**
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
521 *
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.
525 */
526 struct bio *bio_copy_user(request_queue_t *q, unsigned long uaddr,
527 unsigned int len, int write_to_vm)
528 {
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;
533 struct page *page;
534 struct bio *bio;
535 int i, ret;
537 bmd = bio_alloc_map_data(end - start);
538 if (!bmd)
539 return ERR_PTR(-ENOMEM);
541 bmd->userptr = (void __user *) uaddr;
543 ret = -ENOMEM;
544 bio = bio_alloc(GFP_KERNEL, end - start);
545 if (!bio)
546 goto out_bmd;
548 bio->bi_rw |= (!write_to_vm << BIO_RW);
550 ret = 0;
551 while (len) {
552 unsigned int bytes = PAGE_SIZE;
554 if (bytes > len)
555 bytes = len;
557 page = alloc_page(q->bounce_gfp | GFP_KERNEL);
558 if (!page) {
559 ret = -ENOMEM;
560 break;
561 }
563 if (bio_add_pc_page(q, bio, page, bytes, 0) < bytes) {
564 ret = -EINVAL;
565 break;
566 }
568 len -= bytes;
569 }
571 if (ret)
572 goto cleanup;
574 /*
575 * success
576 */
577 if (!write_to_vm) {
578 char __user *p = (char __user *) uaddr;
580 /*
581 * for a write, copy in data to kernel pages
582 */
583 ret = -EFAULT;
584 bio_for_each_segment(bvec, bio, i) {
585 char *addr = page_address(bvec->bv_page);
587 if (copy_from_user(addr, p, bvec->bv_len))
588 goto cleanup;
589 p += bvec->bv_len;
590 }
591 }
593 bio_set_map_data(bmd, bio);
594 return bio;
595 cleanup:
596 bio_for_each_segment(bvec, bio, i)
597 __free_page(bvec->bv_page);
599 bio_put(bio);
600 out_bmd:
601 bio_free_map_data(bmd);
602 return ERR_PTR(ret);
603 }
605 static struct bio *__bio_map_user_iov(request_queue_t *q,
606 struct block_device *bdev,
607 struct sg_iovec *iov, int iov_count,
608 int write_to_vm)
609 {
610 int i, j;
611 int nr_pages = 0;
612 struct page **pages;
613 struct bio *bio;
614 int cur_page = 0;
615 int ret, offset;
617 for (i = 0; i < iov_count; i++) {
618 unsigned long uaddr = (unsigned long)iov[i].iov_base;
619 unsigned long len = iov[i].iov_len;
620 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
621 unsigned long start = uaddr >> PAGE_SHIFT;
623 nr_pages += end - start;
624 /*
625 * transfer and buffer must be aligned to at least hardsector
626 * size for now, in the future we can relax this restriction
627 */
628 if ((uaddr & queue_dma_alignment(q)) || (len & queue_dma_alignment(q)))
629 return ERR_PTR(-EINVAL);
630 }
632 if (!nr_pages)
633 return ERR_PTR(-EINVAL);
635 bio = bio_alloc(GFP_KERNEL, nr_pages);
636 if (!bio)
637 return ERR_PTR(-ENOMEM);
639 ret = -ENOMEM;
640 pages = kcalloc(nr_pages, sizeof(struct page *), GFP_KERNEL);
641 if (!pages)
642 goto out;
644 for (i = 0; i < iov_count; i++) {
645 unsigned long uaddr = (unsigned long)iov[i].iov_base;
646 unsigned long len = iov[i].iov_len;
647 unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
648 unsigned long start = uaddr >> PAGE_SHIFT;
649 const int local_nr_pages = end - start;
650 const int page_limit = cur_page + local_nr_pages;
652 down_read(&current->mm->mmap_sem);
653 ret = get_user_pages(current, current->mm, uaddr,
654 local_nr_pages,
655 write_to_vm, 0, &pages[cur_page], NULL);
656 up_read(&current->mm->mmap_sem);
658 if (ret < local_nr_pages) {
659 ret = -EFAULT;
660 goto out_unmap;
661 }
663 offset = uaddr & ~PAGE_MASK;
664 for (j = cur_page; j < page_limit; j++) {
665 unsigned int bytes = PAGE_SIZE - offset;
667 if (len <= 0)
668 break;
670 if (bytes > len)
671 bytes = len;
673 /*
674 * sorry...
675 */
676 if (bio_add_pc_page(q, bio, pages[j], bytes, offset) <
677 bytes)
678 break;
680 len -= bytes;
681 offset = 0;
682 }
684 cur_page = j;
685 /*
686 * release the pages we didn't map into the bio, if any
687 */
688 while (j < page_limit)
689 page_cache_release(pages[j++]);
690 }
692 kfree(pages);
694 /*
695 * set data direction, and check if mapped pages need bouncing
696 */
697 if (!write_to_vm)
698 bio->bi_rw |= (1 << BIO_RW);
700 bio->bi_bdev = bdev;
701 bio->bi_flags |= (1 << BIO_USER_MAPPED);
702 return bio;
704 out_unmap:
705 for (i = 0; i < nr_pages; i++) {
706 if(!pages[i])
707 break;
708 page_cache_release(pages[i]);
709 }
710 out:
711 kfree(pages);
712 bio_put(bio);
713 return ERR_PTR(ret);
714 }
716 /**
717 * bio_map_user - map user address into bio
718 * @q: the request_queue_t for the bio
719 * @bdev: destination block device
720 * @uaddr: start of user address
721 * @len: length in bytes
722 * @write_to_vm: bool indicating writing to pages or not
723 *
724 * Map the user space address into a bio suitable for io to a block
725 * device. Returns an error pointer in case of error.
726 */
727 struct bio *bio_map_user(request_queue_t *q, struct block_device *bdev,
728 unsigned long uaddr, unsigned int len, int write_to_vm)
729 {
730 struct sg_iovec iov;
732 iov.iov_base = (void __user *)uaddr;
733 iov.iov_len = len;
735 return bio_map_user_iov(q, bdev, &iov, 1, write_to_vm);
736 }
738 /**
739 * bio_map_user_iov - map user sg_iovec table into bio
740 * @q: the request_queue_t for the bio
741 * @bdev: destination block device
742 * @iov: the iovec.
743 * @iov_count: number of elements in the iovec
744 * @write_to_vm: bool indicating writing to pages or not
745 *
746 * Map the user space address into a bio suitable for io to a block
747 * device. Returns an error pointer in case of error.
748 */
749 struct bio *bio_map_user_iov(request_queue_t *q, struct block_device *bdev,
750 struct sg_iovec *iov, int iov_count,
751 int write_to_vm)
752 {
753 struct bio *bio;
754 int len = 0, i;
756 bio = __bio_map_user_iov(q, bdev, iov, iov_count, write_to_vm);
758 if (IS_ERR(bio))
759 return bio;
761 /*
762 * subtle -- if __bio_map_user() ended up bouncing a bio,
763 * it would normally disappear when its bi_end_io is run.
764 * however, we need it for the unmap, so grab an extra
765 * reference to it
766 */
767 bio_get(bio);
769 for (i = 0; i < iov_count; i++)
770 len += iov[i].iov_len;
772 if (bio->bi_size == len)
773 return bio;
775 /*
776 * don't support partial mappings
777 */
778 bio_endio(bio, bio->bi_size, 0);
779 bio_unmap_user(bio);
780 return ERR_PTR(-EINVAL);
781 }
783 static void __bio_unmap_user(struct bio *bio)
784 {
785 struct bio_vec *bvec;
786 int i;
788 /*
789 * make sure we dirty pages we wrote to
790 */
791 __bio_for_each_segment(bvec, bio, i, 0) {
792 if (bio_data_dir(bio) == READ)
793 set_page_dirty_lock(bvec->bv_page);
795 page_cache_release(bvec->bv_page);
796 }
798 bio_put(bio);
799 }
801 /**
802 * bio_unmap_user - unmap a bio
803 * @bio: the bio being unmapped
804 *
805 * Unmap a bio previously mapped by bio_map_user(). Must be called with
806 * a process context.
807 *
808 * bio_unmap_user() may sleep.
809 */
810 void bio_unmap_user(struct bio *bio)
811 {
812 __bio_unmap_user(bio);
813 bio_put(bio);
814 }
816 static int bio_map_kern_endio(struct bio *bio, unsigned int bytes_done, int err)
817 {
818 if (bio->bi_size)
819 return 1;
821 bio_put(bio);
822 return 0;
823 }
826 static struct bio *__bio_map_kern(request_queue_t *q, void *data,
827 unsigned int len, gfp_t gfp_mask)
828 {
829 unsigned long kaddr = (unsigned long)data;
830 unsigned long end = (kaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
831 unsigned long start = kaddr >> PAGE_SHIFT;
832 const int nr_pages = end - start;
833 int offset, i;
834 struct bio *bio;
836 bio = bio_alloc(gfp_mask, nr_pages);
837 if (!bio)
838 return ERR_PTR(-ENOMEM);
840 offset = offset_in_page(kaddr);
841 for (i = 0; i < nr_pages; i++) {
842 unsigned int bytes = PAGE_SIZE - offset;
844 if (len <= 0)
845 break;
847 if (bytes > len)
848 bytes = len;
850 if (bio_add_pc_page(q, bio, virt_to_page(data), bytes,
851 offset) < bytes)
852 break;
854 data += bytes;
855 len -= bytes;
856 offset = 0;
857 }
859 bio->bi_end_io = bio_map_kern_endio;
860 return bio;
861 }
863 /**
864 * bio_map_kern - map kernel address into bio
865 * @q: the request_queue_t for the bio
866 * @data: pointer to buffer to map
867 * @len: length in bytes
868 * @gfp_mask: allocation flags for bio allocation
869 *
870 * Map the kernel address into a bio suitable for io to a block
871 * device. Returns an error pointer in case of error.
872 */
873 struct bio *bio_map_kern(request_queue_t *q, void *data, unsigned int len,
874 gfp_t gfp_mask)
875 {
876 struct bio *bio;
878 bio = __bio_map_kern(q, data, len, gfp_mask);
879 if (IS_ERR(bio))
880 return bio;
882 if (bio->bi_size == len)
883 return bio;
885 /*
886 * Don't support partial mappings.
887 */
888 bio_put(bio);
889 return ERR_PTR(-EINVAL);
890 }
892 /*
893 * bio_set_pages_dirty() and bio_check_pages_dirty() are support functions
894 * for performing direct-IO in BIOs.
895 *
896 * The problem is that we cannot run set_page_dirty() from interrupt context
897 * because the required locks are not interrupt-safe. So what we can do is to
898 * mark the pages dirty _before_ performing IO. And in interrupt context,
899 * check that the pages are still dirty. If so, fine. If not, redirty them
900 * in process context.
901 *
902 * We special-case compound pages here: normally this means reads into hugetlb
903 * pages. The logic in here doesn't really work right for compound pages
904 * because the VM does not uniformly chase down the head page in all cases.
905 * But dirtiness of compound pages is pretty meaningless anyway: the VM doesn't
906 * handle them at all. So we skip compound pages here at an early stage.
907 *
908 * Note that this code is very hard to test under normal circumstances because
909 * direct-io pins the pages with get_user_pages(). This makes
910 * is_page_cache_freeable return false, and the VM will not clean the pages.
911 * But other code (eg, pdflush) could clean the pages if they are mapped
912 * pagecache.
913 *
914 * Simply disabling the call to bio_set_pages_dirty() is a good way to test the
915 * deferred bio dirtying paths.
916 */
918 /*
919 * bio_set_pages_dirty() will mark all the bio's pages as dirty.
920 */
921 void bio_set_pages_dirty(struct bio *bio)
922 {
923 struct bio_vec *bvec = bio->bi_io_vec;
924 int i;
926 for (i = 0; i < bio->bi_vcnt; i++) {
927 struct page *page = bvec[i].bv_page;
929 if (page && !PageCompound(page))
930 set_page_dirty_lock(page);
931 }
932 }
934 static void bio_release_pages(struct bio *bio)
935 {
936 struct bio_vec *bvec = bio->bi_io_vec;
937 int i;
939 for (i = 0; i < bio->bi_vcnt; i++) {
940 struct page *page = bvec[i].bv_page;
942 if (page)
943 put_page(page);
944 }
945 }
947 /*
948 * bio_check_pages_dirty() will check that all the BIO's pages are still dirty.
949 * If they are, then fine. If, however, some pages are clean then they must
950 * have been written out during the direct-IO read. So we take another ref on
951 * the BIO and the offending pages and re-dirty the pages in process context.
952 *
953 * It is expected that bio_check_pages_dirty() will wholly own the BIO from
954 * here on. It will run one page_cache_release() against each page and will
955 * run one bio_put() against the BIO.
956 */
958 static void bio_dirty_fn(void *data);
960 static DECLARE_WORK(bio_dirty_work, bio_dirty_fn, NULL);
961 static DEFINE_SPINLOCK(bio_dirty_lock);
962 static struct bio *bio_dirty_list;
964 /*
965 * This runs in process context
966 */
967 static void bio_dirty_fn(void *data)
968 {
969 unsigned long flags;
970 struct bio *bio;
972 spin_lock_irqsave(&bio_dirty_lock, flags);
973 bio = bio_dirty_list;
974 bio_dirty_list = NULL;
975 spin_unlock_irqrestore(&bio_dirty_lock, flags);
977 while (bio) {
978 struct bio *next = bio->bi_private;
980 bio_set_pages_dirty(bio);
981 bio_release_pages(bio);
982 bio_put(bio);
983 bio = next;
984 }
985 }
987 void bio_check_pages_dirty(struct bio *bio)
988 {
989 struct bio_vec *bvec = bio->bi_io_vec;
990 int nr_clean_pages = 0;
991 int i;
993 for (i = 0; i < bio->bi_vcnt; i++) {
994 struct page *page = bvec[i].bv_page;
996 if (PageDirty(page) || PageCompound(page)) {
997 page_cache_release(page);
998 bvec[i].bv_page = NULL;
999 } else {
1000 nr_clean_pages++;
1004 if (nr_clean_pages) {
1005 unsigned long flags;
1007 spin_lock_irqsave(&bio_dirty_lock, flags);
1008 bio->bi_private = bio_dirty_list;
1009 bio_dirty_list = bio;
1010 spin_unlock_irqrestore(&bio_dirty_lock, flags);
1011 schedule_work(&bio_dirty_work);
1012 } else {
1013 bio_put(bio);
1017 /**
1018 * bio_endio - end I/O on a bio
1019 * @bio: bio
1020 * @bytes_done: number of bytes completed
1021 * @error: error, if any
1023 * Description:
1024 * bio_endio() will end I/O on @bytes_done number of bytes. This may be
1025 * just a partial part of the bio, or it may be the whole bio. bio_endio()
1026 * is the preferred way to end I/O on a bio, it takes care of decrementing
1027 * bi_size and clearing BIO_UPTODATE on error. @error is 0 on success, and
1028 * and one of the established -Exxxx (-EIO, for instance) error values in
1029 * case something went wrong. Noone should call bi_end_io() directly on
1030 * a bio unless they own it and thus know that it has an end_io function.
1031 **/
1032 void bio_endio(struct bio *bio, unsigned int bytes_done, int error)
1034 if (error)
1035 clear_bit(BIO_UPTODATE, &bio->bi_flags);
1037 if (unlikely(bytes_done > bio->bi_size)) {
1038 printk("%s: want %u bytes done, only %u left\n", __FUNCTION__,
1039 bytes_done, bio->bi_size);
1040 bytes_done = bio->bi_size;
1043 bio->bi_size -= bytes_done;
1044 bio->bi_sector += (bytes_done >> 9);
1046 if (bio->bi_end_io)
1047 bio->bi_end_io(bio, bytes_done, error);
1050 void bio_pair_release(struct bio_pair *bp)
1052 if (atomic_dec_and_test(&bp->cnt)) {
1053 struct bio *master = bp->bio1.bi_private;
1055 bio_endio(master, master->bi_size, bp->error);
1056 mempool_free(bp, bp->bio2.bi_private);
1060 static int bio_pair_end_1(struct bio * bi, unsigned int done, int err)
1062 struct bio_pair *bp = container_of(bi, struct bio_pair, bio1);
1064 if (err)
1065 bp->error = err;
1067 if (bi->bi_size)
1068 return 1;
1070 bio_pair_release(bp);
1071 return 0;
1074 static int bio_pair_end_2(struct bio * bi, unsigned int done, int err)
1076 struct bio_pair *bp = container_of(bi, struct bio_pair, bio2);
1078 if (err)
1079 bp->error = err;
1081 if (bi->bi_size)
1082 return 1;
1084 bio_pair_release(bp);
1085 return 0;
1088 /*
1089 * split a bio - only worry about a bio with a single page
1090 * in it's iovec
1091 */
1092 struct bio_pair *bio_split(struct bio *bi, mempool_t *pool, int first_sectors)
1094 struct bio_pair *bp = mempool_alloc(pool, GFP_NOIO);
1096 if (!bp)
1097 return bp;
1099 blk_add_trace_pdu_int(bdev_get_queue(bi->bi_bdev), BLK_TA_SPLIT, bi,
1100 bi->bi_sector + first_sectors);
1102 BUG_ON(bi->bi_vcnt != 1);
1103 BUG_ON(bi->bi_idx != 0);
1104 atomic_set(&bp->cnt, 3);
1105 bp->error = 0;
1106 bp->bio1 = *bi;
1107 bp->bio2 = *bi;
1108 bp->bio2.bi_sector += first_sectors;
1109 bp->bio2.bi_size -= first_sectors << 9;
1110 bp->bio1.bi_size = first_sectors << 9;
1112 bp->bv1 = bi->bi_io_vec[0];
1113 bp->bv2 = bi->bi_io_vec[0];
1114 bp->bv2.bv_offset += first_sectors << 9;
1115 bp->bv2.bv_len -= first_sectors << 9;
1116 bp->bv1.bv_len = first_sectors << 9;
1118 bp->bio1.bi_io_vec = &bp->bv1;
1119 bp->bio2.bi_io_vec = &bp->bv2;
1121 bp->bio1.bi_max_vecs = 1;
1122 bp->bio2.bi_max_vecs = 1;
1124 bp->bio1.bi_end_io = bio_pair_end_1;
1125 bp->bio2.bi_end_io = bio_pair_end_2;
1127 bp->bio1.bi_private = bi;
1128 bp->bio2.bi_private = pool;
1130 return bp;
1134 /*
1135 * create memory pools for biovec's in a bio_set.
1136 * use the global biovec slabs created for general use.
1137 */
1138 static int biovec_create_pools(struct bio_set *bs, int pool_entries, int scale)
1140 int i;
1142 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1143 struct biovec_slab *bp = bvec_slabs + i;
1144 mempool_t **bvp = bs->bvec_pools + i;
1146 if (i >= scale)
1147 pool_entries >>= 1;
1149 *bvp = mempool_create_slab_pool(pool_entries, bp->slab);
1150 if (!*bvp)
1151 return -ENOMEM;
1153 return 0;
1156 static void biovec_free_pools(struct bio_set *bs)
1158 int i;
1160 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1161 mempool_t *bvp = bs->bvec_pools[i];
1163 if (bvp)
1164 mempool_destroy(bvp);
1169 void bioset_free(struct bio_set *bs)
1171 if (bs->bio_pool)
1172 mempool_destroy(bs->bio_pool);
1174 biovec_free_pools(bs);
1176 kfree(bs);
1179 struct bio_set *bioset_create(int bio_pool_size, int bvec_pool_size, int scale)
1181 struct bio_set *bs = kzalloc(sizeof(*bs), GFP_KERNEL);
1183 if (!bs)
1184 return NULL;
1186 bs->bio_pool = mempool_create_slab_pool(bio_pool_size, bio_slab);
1187 if (!bs->bio_pool)
1188 goto bad;
1190 if (!biovec_create_pools(bs, bvec_pool_size, scale))
1191 return bs;
1193 bad:
1194 bioset_free(bs);
1195 return NULL;
1198 static void __init biovec_init_slabs(void)
1200 int i;
1202 for (i = 0; i < BIOVEC_NR_POOLS; i++) {
1203 int size;
1204 struct biovec_slab *bvs = bvec_slabs + i;
1206 size = bvs->nr_vecs * sizeof(struct bio_vec);
1207 bvs->slab = kmem_cache_create(bvs->name, size, 0,
1208 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1212 static int __init init_bio(void)
1214 int megabytes, bvec_pool_entries;
1215 int scale = BIOVEC_NR_POOLS;
1217 bio_slab = kmem_cache_create("bio", sizeof(struct bio), 0,
1218 SLAB_HWCACHE_ALIGN|SLAB_PANIC, NULL, NULL);
1220 biovec_init_slabs();
1222 megabytes = nr_free_pages() >> (20 - PAGE_SHIFT);
1224 /*
1225 * find out where to start scaling
1226 */
1227 if (megabytes <= 16)
1228 scale = 0;
1229 else if (megabytes <= 32)
1230 scale = 1;
1231 else if (megabytes <= 64)
1232 scale = 2;
1233 else if (megabytes <= 96)
1234 scale = 3;
1235 else if (megabytes <= 128)
1236 scale = 4;
1238 /*
1239 * Limit number of entries reserved -- mempools are only used when
1240 * the system is completely unable to allocate memory, so we only
1241 * need enough to make progress.
1242 */
1243 bvec_pool_entries = 1 + scale;
1245 fs_bio_set = bioset_create(BIO_POOL_SIZE, bvec_pool_entries, scale);
1246 if (!fs_bio_set)
1247 panic("bio: can't allocate bios\n");
1249 bio_split_pool = mempool_create_kmalloc_pool(BIO_SPLIT_ENTRIES,
1250 sizeof(struct bio_pair));
1251 if (!bio_split_pool)
1252 panic("bio: can't create split pool\n");
1254 return 0;
1257 subsys_initcall(init_bio);
1259 EXPORT_SYMBOL(bio_alloc);
1260 EXPORT_SYMBOL(bio_put);
1261 EXPORT_SYMBOL(bio_free);
1262 EXPORT_SYMBOL(bio_endio);
1263 EXPORT_SYMBOL(bio_init);
1264 EXPORT_SYMBOL(__bio_clone);
1265 EXPORT_SYMBOL(bio_clone);
1266 EXPORT_SYMBOL(bio_phys_segments);
1267 EXPORT_SYMBOL(bio_hw_segments);
1268 EXPORT_SYMBOL(bio_add_page);
1269 EXPORT_SYMBOL(bio_add_pc_page);
1270 EXPORT_SYMBOL(bio_get_nr_vecs);
1271 EXPORT_SYMBOL(bio_map_user);
1272 EXPORT_SYMBOL(bio_unmap_user);
1273 EXPORT_SYMBOL(bio_map_kern);
1274 EXPORT_SYMBOL(bio_pair_release);
1275 EXPORT_SYMBOL(bio_split);
1276 EXPORT_SYMBOL(bio_split_pool);
1277 EXPORT_SYMBOL(bio_copy_user);
1278 EXPORT_SYMBOL(bio_uncopy_user);
1279 EXPORT_SYMBOL(bioset_create);
1280 EXPORT_SYMBOL(bioset_free);
1281 EXPORT_SYMBOL(bio_alloc_bioset);