aboutsummaryrefslogtreecommitdiffstats
path: root/fs/btrfs
diff options
context:
space:
mode:
Diffstat (limited to 'fs/btrfs')
-rw-r--r--fs/btrfs/compression.c124
-rw-r--r--fs/btrfs/ctree.h30
-rw-r--r--fs/btrfs/disk-io.c45
-rw-r--r--fs/btrfs/extent_io.c5
-rw-r--r--fs/btrfs/file-item.c185
-rw-r--r--fs/btrfs/inode.c45
-rw-r--r--fs/btrfs/ioctl.c55
-rw-r--r--fs/btrfs/ordered-data.c7
-rw-r--r--fs/btrfs/ordered-data.h10
-rw-r--r--fs/btrfs/tree-log.c121
-rw-r--r--fs/btrfs/volumes.c1
11 files changed, 387 insertions, 241 deletions
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 4febe2eb0b83..ad7274137309 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -69,11 +69,27 @@ struct compressed_bio {
69 69
70 /* IO errors */ 70 /* IO errors */
71 int errors; 71 int errors;
72 int mirror_num;
72 73
73 /* for reads, this is the bio we are copying the data into */ 74 /* for reads, this is the bio we are copying the data into */
74 struct bio *orig_bio; 75 struct bio *orig_bio;
76
77 /*
78 * the start of a variable length array of checksums only
79 * used by reads
80 */
81 u32 sums;
75}; 82};
76 83
84static inline int compressed_bio_size(struct btrfs_root *root,
85 unsigned long disk_size)
86{
87 u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);
88 return sizeof(struct compressed_bio) +
89 ((disk_size + root->sectorsize - 1) / root->sectorsize) *
90 csum_size;
91}
92
77static struct bio *compressed_bio_alloc(struct block_device *bdev, 93static struct bio *compressed_bio_alloc(struct block_device *bdev,
78 u64 first_byte, gfp_t gfp_flags) 94 u64 first_byte, gfp_t gfp_flags)
79{ 95{
@@ -96,6 +112,47 @@ static struct bio *compressed_bio_alloc(struct block_device *bdev,
96 return bio; 112 return bio;
97} 113}
98 114
115static int check_compressed_csum(struct inode *inode,
116 struct compressed_bio *cb,
117 u64 disk_start)
118{
119 int ret;
120 struct btrfs_root *root = BTRFS_I(inode)->root;
121 struct page *page;
122 unsigned long i;
123 char *kaddr;
124 u32 csum;
125 u32 *cb_sum = &cb->sums;
126
127 if (btrfs_test_opt(root, NODATASUM) ||
128 btrfs_test_flag(inode, NODATASUM))
129 return 0;
130
131 for (i = 0; i < cb->nr_pages; i++) {
132 page = cb->compressed_pages[i];
133 csum = ~(u32)0;
134
135 kaddr = kmap_atomic(page, KM_USER0);
136 csum = btrfs_csum_data(root, kaddr, csum, PAGE_CACHE_SIZE);
137 btrfs_csum_final(csum, (char *)&csum);
138 kunmap_atomic(kaddr, KM_USER0);
139
140 if (csum != *cb_sum) {
141 printk("btrfs csum failed ino %lu extent %llu csum %u "
142 "wanted %u mirror %d\n", inode->i_ino,
143 (unsigned long long)disk_start,
144 csum, *cb_sum, cb->mirror_num);
145 ret = -EIO;
146 goto fail;
147 }
148 cb_sum++;
149
150 }
151 ret = 0;
152fail:
153 return ret;
154}
155
99/* when we finish reading compressed pages from the disk, we 156/* when we finish reading compressed pages from the disk, we
100 * decompress them and then run the bio end_io routines on the 157 * decompress them and then run the bio end_io routines on the
101 * decompressed pages (in the inode address space). 158 * decompressed pages (in the inode address space).
@@ -124,16 +181,21 @@ static void end_compressed_bio_read(struct bio *bio, int err)
124 if (!atomic_dec_and_test(&cb->pending_bios)) 181 if (!atomic_dec_and_test(&cb->pending_bios))
125 goto out; 182 goto out;
126 183
184 inode = cb->inode;
185 ret = check_compressed_csum(inode, cb, (u64)bio->bi_sector << 9);
186 if (ret)
187 goto csum_failed;
188
127 /* ok, we're the last bio for this extent, lets start 189 /* ok, we're the last bio for this extent, lets start
128 * the decompression. 190 * the decompression.
129 */ 191 */
130 inode = cb->inode;
131 tree = &BTRFS_I(inode)->io_tree; 192 tree = &BTRFS_I(inode)->io_tree;
132 ret = btrfs_zlib_decompress_biovec(cb->compressed_pages, 193 ret = btrfs_zlib_decompress_biovec(cb->compressed_pages,
133 cb->start, 194 cb->start,
134 cb->orig_bio->bi_io_vec, 195 cb->orig_bio->bi_io_vec,
135 cb->orig_bio->bi_vcnt, 196 cb->orig_bio->bi_vcnt,
136 cb->compressed_len); 197 cb->compressed_len);
198csum_failed:
137 if (ret) 199 if (ret)
138 cb->errors = 1; 200 cb->errors = 1;
139 201
@@ -148,8 +210,21 @@ static void end_compressed_bio_read(struct bio *bio, int err)
148 /* do io completion on the original bio */ 210 /* do io completion on the original bio */
149 if (cb->errors) { 211 if (cb->errors) {
150 bio_io_error(cb->orig_bio); 212 bio_io_error(cb->orig_bio);
151 } else 213 } else {
214 int bio_index = 0;
215 struct bio_vec *bvec = cb->orig_bio->bi_io_vec;
216
217 /*
218 * we have verified the checksum already, set page
219 * checked so the end_io handlers know about it
220 */
221 while(bio_index < cb->orig_bio->bi_vcnt) {
222 SetPageChecked(bvec->bv_page);
223 bvec++;
224 bio_index++;
225 }
152 bio_endio(cb->orig_bio, 0); 226 bio_endio(cb->orig_bio, 0);
227 }
153 228
154 /* finally free the cb struct */ 229 /* finally free the cb struct */
155 kfree(cb->compressed_pages); 230 kfree(cb->compressed_pages);
@@ -277,12 +352,13 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
277 int ret; 352 int ret;
278 353
279 WARN_ON(start & ((u64)PAGE_CACHE_SIZE - 1)); 354 WARN_ON(start & ((u64)PAGE_CACHE_SIZE - 1));
280 cb = kmalloc(sizeof(*cb), GFP_NOFS); 355 cb = kmalloc(compressed_bio_size(root, compressed_len), GFP_NOFS);
281 atomic_set(&cb->pending_bios, 0); 356 atomic_set(&cb->pending_bios, 0);
282 cb->errors = 0; 357 cb->errors = 0;
283 cb->inode = inode; 358 cb->inode = inode;
284 cb->start = start; 359 cb->start = start;
285 cb->len = len; 360 cb->len = len;
361 cb->mirror_num = 0;
286 cb->compressed_pages = compressed_pages; 362 cb->compressed_pages = compressed_pages;
287 cb->compressed_len = compressed_len; 363 cb->compressed_len = compressed_len;
288 cb->orig_bio = NULL; 364 cb->orig_bio = NULL;
@@ -290,9 +366,6 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
290 366
291 bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev; 367 bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
292 368
293 ret = btrfs_csum_file_bytes(root, inode, start, len);
294 BUG_ON(ret);
295
296 bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS); 369 bio = compressed_bio_alloc(bdev, first_byte, GFP_NOFS);
297 bio->bi_private = cb; 370 bio->bi_private = cb;
298 bio->bi_end_io = end_compressed_bio_write; 371 bio->bi_end_io = end_compressed_bio_write;
@@ -325,6 +398,9 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
325 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); 398 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
326 BUG_ON(ret); 399 BUG_ON(ret);
327 400
401 ret = btrfs_csum_one_bio(root, inode, bio, start, 1);
402 BUG_ON(ret);
403
328 ret = btrfs_map_bio(root, WRITE, bio, 0, 1); 404 ret = btrfs_map_bio(root, WRITE, bio, 0, 1);
329 BUG_ON(ret); 405 BUG_ON(ret);
330 406
@@ -348,6 +424,9 @@ int btrfs_submit_compressed_write(struct inode *inode, u64 start,
348 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); 424 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
349 BUG_ON(ret); 425 BUG_ON(ret);
350 426
427 ret = btrfs_csum_one_bio(root, inode, bio, start, 1);
428 BUG_ON(ret);
429
351 ret = btrfs_map_bio(root, WRITE, bio, 0, 1); 430 ret = btrfs_map_bio(root, WRITE, bio, 0, 1);
352 BUG_ON(ret); 431 BUG_ON(ret);
353 432
@@ -510,6 +589,7 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
510 u64 em_start; 589 u64 em_start;
511 struct extent_map *em; 590 struct extent_map *em;
512 int ret; 591 int ret;
592 u32 *sums;
513 593
514 tree = &BTRFS_I(inode)->io_tree; 594 tree = &BTRFS_I(inode)->io_tree;
515 em_tree = &BTRFS_I(inode)->extent_tree; 595 em_tree = &BTRFS_I(inode)->extent_tree;
@@ -521,15 +601,18 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
521 PAGE_CACHE_SIZE); 601 PAGE_CACHE_SIZE);
522 spin_unlock(&em_tree->lock); 602 spin_unlock(&em_tree->lock);
523 603
524 cb = kmalloc(sizeof(*cb), GFP_NOFS); 604 compressed_len = em->block_len;
605 cb = kmalloc(compressed_bio_size(root, compressed_len), GFP_NOFS);
525 atomic_set(&cb->pending_bios, 0); 606 atomic_set(&cb->pending_bios, 0);
526 cb->errors = 0; 607 cb->errors = 0;
527 cb->inode = inode; 608 cb->inode = inode;
609 cb->mirror_num = mirror_num;
610 sums = &cb->sums;
528 611
529 cb->start = em->orig_start; 612 cb->start = em->orig_start;
530 compressed_len = em->block_len;
531 em_len = em->len; 613 em_len = em->len;
532 em_start = em->start; 614 em_start = em->start;
615
533 free_extent_map(em); 616 free_extent_map(em);
534 em = NULL; 617 em = NULL;
535 618
@@ -551,11 +634,6 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
551 634
552 add_ra_bio_pages(inode, em_start + em_len, cb); 635 add_ra_bio_pages(inode, em_start + em_len, cb);
553 636
554 if (!btrfs_test_opt(root, NODATASUM) &&
555 !btrfs_test_flag(inode, NODATASUM)) {
556 btrfs_lookup_bio_sums(root, inode, cb->orig_bio);
557 }
558
559 /* include any pages we added in add_ra-bio_pages */ 637 /* include any pages we added in add_ra-bio_pages */
560 uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE; 638 uncompressed_len = bio->bi_vcnt * PAGE_CACHE_SIZE;
561 cb->len = uncompressed_len; 639 cb->len = uncompressed_len;
@@ -568,6 +646,8 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
568 for (page_index = 0; page_index < nr_pages; page_index++) { 646 for (page_index = 0; page_index < nr_pages; page_index++) {
569 page = cb->compressed_pages[page_index]; 647 page = cb->compressed_pages[page_index];
570 page->mapping = inode->i_mapping; 648 page->mapping = inode->i_mapping;
649 page->index = em_start >> PAGE_CACHE_SHIFT;
650
571 if (comp_bio->bi_size) 651 if (comp_bio->bi_size)
572 ret = tree->ops->merge_bio_hook(page, 0, 652 ret = tree->ops->merge_bio_hook(page, 0,
573 PAGE_CACHE_SIZE, 653 PAGE_CACHE_SIZE,
@@ -591,7 +671,16 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
591 */ 671 */
592 atomic_inc(&cb->pending_bios); 672 atomic_inc(&cb->pending_bios);
593 673
594 ret = btrfs_map_bio(root, READ, comp_bio, 0, 0); 674 if (!btrfs_test_opt(root, NODATASUM) &&
675 !btrfs_test_flag(inode, NODATASUM)) {
676 btrfs_lookup_bio_sums(root, inode, comp_bio,
677 sums);
678 }
679 sums += (comp_bio->bi_size + root->sectorsize - 1) /
680 root->sectorsize;
681
682 ret = btrfs_map_bio(root, READ, comp_bio,
683 mirror_num, 0);
595 BUG_ON(ret); 684 BUG_ON(ret);
596 685
597 bio_put(comp_bio); 686 bio_put(comp_bio);
@@ -610,7 +699,12 @@ int btrfs_submit_compressed_read(struct inode *inode, struct bio *bio,
610 ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0); 699 ret = btrfs_bio_wq_end_io(root->fs_info, comp_bio, 0);
611 BUG_ON(ret); 700 BUG_ON(ret);
612 701
613 ret = btrfs_map_bio(root, READ, comp_bio, 0, 0); 702 if (!btrfs_test_opt(root, NODATASUM) &&
703 !btrfs_test_flag(inode, NODATASUM)) {
704 btrfs_lookup_bio_sums(root, inode, comp_bio, sums);
705 }
706
707 ret = btrfs_map_bio(root, READ, comp_bio, mirror_num, 0);
614 BUG_ON(ret); 708 BUG_ON(ret);
615 709
616 bio_put(comp_bio); 710 bio_put(comp_bio);
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 96f2ec7ad5bd..242b961ae6de 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -73,6 +73,9 @@ struct btrfs_ordered_sum;
73/* directory objectid inside the root tree */ 73/* directory objectid inside the root tree */
74#define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL 74#define BTRFS_ROOT_TREE_DIR_OBJECTID 6ULL
75 75
76/* holds checksums of all the data extents */
77#define BTRFS_CSUM_TREE_OBJECTID 7ULL
78
76/* orhpan objectid for tracking unlinked/truncated files */ 79/* orhpan objectid for tracking unlinked/truncated files */
77#define BTRFS_ORPHAN_OBJECTID -5ULL 80#define BTRFS_ORPHAN_OBJECTID -5ULL
78 81
@@ -84,6 +87,13 @@ struct btrfs_ordered_sum;
84#define BTRFS_TREE_RELOC_OBJECTID -8ULL 87#define BTRFS_TREE_RELOC_OBJECTID -8ULL
85#define BTRFS_DATA_RELOC_TREE_OBJECTID -9ULL 88#define BTRFS_DATA_RELOC_TREE_OBJECTID -9ULL
86 89
90/*
91 * extent checksums all have this objectid
92 * this allows them to share the logging tree
93 * for fsyncs
94 */
95#define BTRFS_EXTENT_CSUM_OBJECTID -10ULL
96
87/* dummy objectid represents multiple objectids */ 97/* dummy objectid represents multiple objectids */
88#define BTRFS_MULTIPLE_OBJECTIDS -255ULL 98#define BTRFS_MULTIPLE_OBJECTIDS -255ULL
89 99
@@ -634,6 +644,7 @@ struct btrfs_fs_info {
634 struct btrfs_root *chunk_root; 644 struct btrfs_root *chunk_root;
635 struct btrfs_root *dev_root; 645 struct btrfs_root *dev_root;
636 struct btrfs_root *fs_root; 646 struct btrfs_root *fs_root;
647 struct btrfs_root *csum_root;
637 648
638 /* the log root tree is a directory of all the other log roots */ 649 /* the log root tree is a directory of all the other log roots */
639 struct btrfs_root *log_root_tree; 650 struct btrfs_root *log_root_tree;
@@ -716,6 +727,7 @@ struct btrfs_fs_info {
716 struct btrfs_workers workers; 727 struct btrfs_workers workers;
717 struct btrfs_workers delalloc_workers; 728 struct btrfs_workers delalloc_workers;
718 struct btrfs_workers endio_workers; 729 struct btrfs_workers endio_workers;
730 struct btrfs_workers endio_meta_workers;
719 struct btrfs_workers endio_write_workers; 731 struct btrfs_workers endio_write_workers;
720 struct btrfs_workers submit_workers; 732 struct btrfs_workers submit_workers;
721 /* 733 /*
@@ -858,13 +870,12 @@ struct btrfs_root {
858 * extent data is for file data 870 * extent data is for file data
859 */ 871 */
860#define BTRFS_EXTENT_DATA_KEY 108 872#define BTRFS_EXTENT_DATA_KEY 108
873
861/* 874/*
862 * csum items have the checksums for data in the extents 875 * extent csums are stored in a separate tree and hold csums for
876 * an entire extent on disk.
863 */ 877 */
864#define BTRFS_CSUM_ITEM_KEY 120 878#define BTRFS_EXTENT_CSUM_KEY 128
865
866
867/* reserve 21-31 for other file/dir stuff */
868 879
869/* 880/*
870 * root items point to tree roots. There are typically in the root 881 * root items point to tree roots. There are typically in the root
@@ -1917,7 +1928,7 @@ int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
1917 1928
1918/* file-item.c */ 1929/* file-item.c */
1919int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, 1930int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
1920 struct bio *bio); 1931 struct bio *bio, u32 *dst);
1921int btrfs_insert_file_extent(struct btrfs_trans_handle *trans, 1932int btrfs_insert_file_extent(struct btrfs_trans_handle *trans,
1922 struct btrfs_root *root, 1933 struct btrfs_root *root,
1923 u64 objectid, u64 pos, 1934 u64 objectid, u64 pos,
@@ -1929,17 +1940,16 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
1929 struct btrfs_path *path, u64 objectid, 1940 struct btrfs_path *path, u64 objectid,
1930 u64 bytenr, int mod); 1941 u64 bytenr, int mod);
1931int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, 1942int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
1932 struct btrfs_root *root, struct inode *inode, 1943 struct btrfs_root *root,
1933 struct btrfs_ordered_sum *sums); 1944 struct btrfs_ordered_sum *sums);
1934int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, 1945int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
1935 struct bio *bio); 1946 struct bio *bio, u64 file_start, int contig);
1936int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode, 1947int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode,
1937 u64 start, unsigned long len); 1948 u64 start, unsigned long len);
1938struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, 1949struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
1939 struct btrfs_root *root, 1950 struct btrfs_root *root,
1940 struct btrfs_path *path, 1951 struct btrfs_path *path,
1941 u64 objectid, u64 offset, 1952 u64 bytenr, int cow);
1942 int cow);
1943int btrfs_csum_truncate(struct btrfs_trans_handle *trans, 1953int btrfs_csum_truncate(struct btrfs_trans_handle *trans,
1944 struct btrfs_root *root, struct btrfs_path *path, 1954 struct btrfs_root *root, struct btrfs_path *path,
1945 u64 isize); 1955 u64 isize);
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 3eb7c2576fe5..61dc3b2c834b 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -445,11 +445,18 @@ static void end_workqueue_bio(struct bio *bio, int err)
445 end_io_wq->error = err; 445 end_io_wq->error = err;
446 end_io_wq->work.func = end_workqueue_fn; 446 end_io_wq->work.func = end_workqueue_fn;
447 end_io_wq->work.flags = 0; 447 end_io_wq->work.flags = 0;
448 if (bio->bi_rw & (1 << BIO_RW)) 448
449 if (bio->bi_rw & (1 << BIO_RW)) {
449 btrfs_queue_worker(&fs_info->endio_write_workers, 450 btrfs_queue_worker(&fs_info->endio_write_workers,
450 &end_io_wq->work); 451 &end_io_wq->work);
451 else 452 } else {
452 btrfs_queue_worker(&fs_info->endio_workers, &end_io_wq->work); 453 if (end_io_wq->metadata)
454 btrfs_queue_worker(&fs_info->endio_meta_workers,
455 &end_io_wq->work);
456 else
457 btrfs_queue_worker(&fs_info->endio_workers,
458 &end_io_wq->work);
459 }
453} 460}
454 461
455int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio, 462int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
@@ -1208,6 +1215,9 @@ static void __unplug_io_fn(struct backing_dev_info *bdi, struct page *page)
1208 info = (struct btrfs_fs_info *)bdi->unplug_io_data; 1215 info = (struct btrfs_fs_info *)bdi->unplug_io_data;
1209 list_for_each(cur, &info->fs_devices->devices) { 1216 list_for_each(cur, &info->fs_devices->devices) {
1210 device = list_entry(cur, struct btrfs_device, dev_list); 1217 device = list_entry(cur, struct btrfs_device, dev_list);
1218 if (!device->bdev)
1219 continue;
1220
1211 bdi = blk_get_backing_dev_info(device->bdev); 1221 bdi = blk_get_backing_dev_info(device->bdev);
1212 if (bdi->unplug_io_fn) { 1222 if (bdi->unplug_io_fn) {
1213 bdi->unplug_io_fn(bdi, page); 1223 bdi->unplug_io_fn(bdi, page);
@@ -1344,7 +1354,7 @@ static void end_workqueue_fn(struct btrfs_work *work)
1344 * blocksize <= pagesize, it is basically a noop 1354 * blocksize <= pagesize, it is basically a noop
1345 */ 1355 */
1346 if (end_io_wq->metadata && !bio_ready_for_csum(bio)) { 1356 if (end_io_wq->metadata && !bio_ready_for_csum(bio)) {
1347 btrfs_queue_worker(&fs_info->endio_workers, 1357 btrfs_queue_worker(&fs_info->endio_meta_workers,
1348 &end_io_wq->work); 1358 &end_io_wq->work);
1349 return; 1359 return;
1350 } 1360 }
@@ -1454,6 +1464,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1454 struct buffer_head *bh; 1464 struct buffer_head *bh;
1455 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root), 1465 struct btrfs_root *extent_root = kzalloc(sizeof(struct btrfs_root),
1456 GFP_NOFS); 1466 GFP_NOFS);
1467 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1468 GFP_NOFS);
1457 struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), 1469 struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root),
1458 GFP_NOFS); 1470 GFP_NOFS);
1459 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info), 1471 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
@@ -1470,7 +1482,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1470 struct btrfs_super_block *disk_super; 1482 struct btrfs_super_block *disk_super;
1471 1483
1472 if (!extent_root || !tree_root || !fs_info || 1484 if (!extent_root || !tree_root || !fs_info ||
1473 !chunk_root || !dev_root) { 1485 !chunk_root || !dev_root || !csum_root) {
1474 err = -ENOMEM; 1486 err = -ENOMEM;
1475 goto fail; 1487 goto fail;
1476 } 1488 }
@@ -1487,6 +1499,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1487 init_completion(&fs_info->kobj_unregister); 1499 init_completion(&fs_info->kobj_unregister);
1488 fs_info->tree_root = tree_root; 1500 fs_info->tree_root = tree_root;
1489 fs_info->extent_root = extent_root; 1501 fs_info->extent_root = extent_root;
1502 fs_info->csum_root = csum_root;
1490 fs_info->chunk_root = chunk_root; 1503 fs_info->chunk_root = chunk_root;
1491 fs_info->dev_root = dev_root; 1504 fs_info->dev_root = dev_root;
1492 fs_info->fs_devices = fs_devices; 1505 fs_info->fs_devices = fs_devices;
@@ -1652,6 +1665,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1652 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1); 1665 btrfs_init_workers(&fs_info->fixup_workers, "fixup", 1);
1653 btrfs_init_workers(&fs_info->endio_workers, "endio", 1666 btrfs_init_workers(&fs_info->endio_workers, "endio",
1654 fs_info->thread_pool_size); 1667 fs_info->thread_pool_size);
1668 btrfs_init_workers(&fs_info->endio_meta_workers, "endio-meta",
1669 fs_info->thread_pool_size);
1655 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write", 1670 btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
1656 fs_info->thread_pool_size); 1671 fs_info->thread_pool_size);
1657 1672
@@ -1667,6 +1682,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1667 btrfs_start_workers(&fs_info->delalloc_workers, 1); 1682 btrfs_start_workers(&fs_info->delalloc_workers, 1);
1668 btrfs_start_workers(&fs_info->fixup_workers, 1); 1683 btrfs_start_workers(&fs_info->fixup_workers, 1);
1669 btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size); 1684 btrfs_start_workers(&fs_info->endio_workers, fs_info->thread_pool_size);
1685 btrfs_start_workers(&fs_info->endio_meta_workers,
1686 fs_info->thread_pool_size);
1670 btrfs_start_workers(&fs_info->endio_write_workers, 1687 btrfs_start_workers(&fs_info->endio_write_workers,
1671 fs_info->thread_pool_size); 1688 fs_info->thread_pool_size);
1672 1689
@@ -1751,6 +1768,13 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1751 if (ret) 1768 if (ret)
1752 goto fail_extent_root; 1769 goto fail_extent_root;
1753 1770
1771 ret = find_and_setup_root(tree_root, fs_info,
1772 BTRFS_CSUM_TREE_OBJECTID, csum_root);
1773 if (ret)
1774 goto fail_extent_root;
1775
1776 csum_root->track_dirty = 1;
1777
1754 btrfs_read_block_groups(extent_root); 1778 btrfs_read_block_groups(extent_root);
1755 1779
1756 fs_info->generation = generation + 1; 1780 fs_info->generation = generation + 1;
@@ -1761,7 +1785,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1761 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root, 1785 fs_info->cleaner_kthread = kthread_run(cleaner_kthread, tree_root,
1762 "btrfs-cleaner"); 1786 "btrfs-cleaner");
1763 if (!fs_info->cleaner_kthread) 1787 if (!fs_info->cleaner_kthread)
1764 goto fail_extent_root; 1788 goto fail_csum_root;
1765 1789
1766 fs_info->transaction_kthread = kthread_run(transaction_kthread, 1790 fs_info->transaction_kthread = kthread_run(transaction_kthread,
1767 tree_root, 1791 tree_root,
@@ -1825,6 +1849,8 @@ fail_cleaner:
1825 filemap_write_and_wait(fs_info->btree_inode->i_mapping); 1849 filemap_write_and_wait(fs_info->btree_inode->i_mapping);
1826 invalidate_inode_pages2(fs_info->btree_inode->i_mapping); 1850 invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
1827 1851
1852fail_csum_root:
1853 free_extent_buffer(csum_root->node);
1828fail_extent_root: 1854fail_extent_root:
1829 free_extent_buffer(extent_root->node); 1855 free_extent_buffer(extent_root->node);
1830fail_tree_root: 1856fail_tree_root:
@@ -1838,6 +1864,7 @@ fail_sb_buffer:
1838 btrfs_stop_workers(&fs_info->delalloc_workers); 1864 btrfs_stop_workers(&fs_info->delalloc_workers);
1839 btrfs_stop_workers(&fs_info->workers); 1865 btrfs_stop_workers(&fs_info->workers);
1840 btrfs_stop_workers(&fs_info->endio_workers); 1866 btrfs_stop_workers(&fs_info->endio_workers);
1867 btrfs_stop_workers(&fs_info->endio_meta_workers);
1841 btrfs_stop_workers(&fs_info->endio_write_workers); 1868 btrfs_stop_workers(&fs_info->endio_write_workers);
1842 btrfs_stop_workers(&fs_info->submit_workers); 1869 btrfs_stop_workers(&fs_info->submit_workers);
1843fail_iput: 1870fail_iput:
@@ -1853,6 +1880,7 @@ fail:
1853 kfree(fs_info); 1880 kfree(fs_info);
1854 kfree(chunk_root); 1881 kfree(chunk_root);
1855 kfree(dev_root); 1882 kfree(dev_root);
1883 kfree(csum_root);
1856 return ERR_PTR(err); 1884 return ERR_PTR(err);
1857} 1885}
1858 1886
@@ -2131,6 +2159,9 @@ int close_ctree(struct btrfs_root *root)
2131 if (root->fs_info->dev_root->node); 2159 if (root->fs_info->dev_root->node);
2132 free_extent_buffer(root->fs_info->dev_root->node); 2160 free_extent_buffer(root->fs_info->dev_root->node);
2133 2161
2162 if (root->fs_info->csum_root->node);
2163 free_extent_buffer(root->fs_info->csum_root->node);
2164
2134 btrfs_free_block_groups(root->fs_info); 2165 btrfs_free_block_groups(root->fs_info);
2135 2166
2136 del_fs_roots(fs_info); 2167 del_fs_roots(fs_info);
@@ -2141,6 +2172,7 @@ int close_ctree(struct btrfs_root *root)
2141 btrfs_stop_workers(&fs_info->delalloc_workers); 2172 btrfs_stop_workers(&fs_info->delalloc_workers);
2142 btrfs_stop_workers(&fs_info->workers); 2173 btrfs_stop_workers(&fs_info->workers);
2143 btrfs_stop_workers(&fs_info->endio_workers); 2174 btrfs_stop_workers(&fs_info->endio_workers);
2175 btrfs_stop_workers(&fs_info->endio_meta_workers);
2144 btrfs_stop_workers(&fs_info->endio_write_workers); 2176 btrfs_stop_workers(&fs_info->endio_write_workers);
2145 btrfs_stop_workers(&fs_info->submit_workers); 2177 btrfs_stop_workers(&fs_info->submit_workers);
2146 2178
@@ -2163,6 +2195,7 @@ int close_ctree(struct btrfs_root *root)
2163 kfree(fs_info->tree_root); 2195 kfree(fs_info->tree_root);
2164 kfree(fs_info->chunk_root); 2196 kfree(fs_info->chunk_root);
2165 kfree(fs_info->dev_root); 2197 kfree(fs_info->dev_root);
2198 kfree(fs_info->csum_root);
2166 return 0; 2199 return 0;
2167} 2200}
2168 2201
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index c3dfe2a0ec85..7449ecf32c50 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1732,6 +1732,9 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
1732 int whole_page; 1732 int whole_page;
1733 int ret; 1733 int ret;
1734 1734
1735 if (err)
1736 uptodate = 0;
1737
1735 do { 1738 do {
1736 struct page *page = bvec->bv_page; 1739 struct page *page = bvec->bv_page;
1737 tree = &BTRFS_I(page->mapping->host)->io_tree; 1740 tree = &BTRFS_I(page->mapping->host)->io_tree;
@@ -1761,6 +1764,8 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
1761 if (ret == 0) { 1764 if (ret == 0) {
1762 uptodate = 1765 uptodate =
1763 test_bit(BIO_UPTODATE, &bio->bi_flags); 1766 test_bit(BIO_UPTODATE, &bio->bi_flags);
1767 if (err)
1768 uptodate = 0;
1764 continue; 1769 continue;
1765 } 1770 }
1766 } 1771 }
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index 234ed441736c..a3ad2ce00116 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -74,8 +74,7 @@ out:
74struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans, 74struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
75 struct btrfs_root *root, 75 struct btrfs_root *root,
76 struct btrfs_path *path, 76 struct btrfs_path *path,
77 u64 objectid, u64 offset, 77 u64 bytenr, int cow)
78 int cow)
79{ 78{
80 int ret; 79 int ret;
81 struct btrfs_key file_key; 80 struct btrfs_key file_key;
@@ -87,9 +86,9 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
87 btrfs_super_csum_size(&root->fs_info->super_copy); 86 btrfs_super_csum_size(&root->fs_info->super_copy);
88 int csums_in_item; 87 int csums_in_item;
89 88
90 file_key.objectid = objectid; 89 file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
91 file_key.offset = offset; 90 file_key.offset = bytenr;
92 btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); 91 btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY);
93 ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow); 92 ret = btrfs_search_slot(trans, root, &file_key, path, 0, cow);
94 if (ret < 0) 93 if (ret < 0)
95 goto fail; 94 goto fail;
@@ -100,11 +99,10 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
100 goto fail; 99 goto fail;
101 path->slots[0]--; 100 path->slots[0]--;
102 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 101 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
103 if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY || 102 if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY)
104 found_key.objectid != objectid) {
105 goto fail; 103 goto fail;
106 } 104
107 csum_offset = (offset - found_key.offset) >> 105 csum_offset = (bytenr - found_key.offset) >>
108 root->fs_info->sb->s_blocksize_bits; 106 root->fs_info->sb->s_blocksize_bits;
109 csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]); 107 csums_in_item = btrfs_item_size_nr(leaf, path->slots[0]);
110 csums_in_item /= csum_size; 108 csums_in_item /= csum_size;
@@ -143,7 +141,7 @@ int btrfs_lookup_file_extent(struct btrfs_trans_handle *trans,
143} 141}
144 142
145int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode, 143int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
146 struct bio *bio) 144 struct bio *bio, u32 *dst)
147{ 145{
148 u32 sum; 146 u32 sum;
149 struct bio_vec *bvec = bio->bi_io_vec; 147 struct bio_vec *bvec = bio->bi_io_vec;
@@ -151,6 +149,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
151 u64 offset; 149 u64 offset;
152 u64 item_start_offset = 0; 150 u64 item_start_offset = 0;
153 u64 item_last_offset = 0; 151 u64 item_last_offset = 0;
152 u64 disk_bytenr;
154 u32 diff; 153 u32 diff;
155 u16 csum_size = 154 u16 csum_size =
156 btrfs_super_csum_size(&root->fs_info->super_copy); 155 btrfs_super_csum_size(&root->fs_info->super_copy);
@@ -165,21 +164,22 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
165 164
166 WARN_ON(bio->bi_vcnt <= 0); 165 WARN_ON(bio->bi_vcnt <= 0);
167 166
167 disk_bytenr = (u64)bio->bi_sector << 9;
168 while(bio_index < bio->bi_vcnt) { 168 while(bio_index < bio->bi_vcnt) {
169 offset = page_offset(bvec->bv_page) + bvec->bv_offset; 169 offset = page_offset(bvec->bv_page) + bvec->bv_offset;
170 ret = btrfs_find_ordered_sum(inode, offset, &sum); 170 ret = btrfs_find_ordered_sum(inode, offset, disk_bytenr, &sum);
171 if (ret == 0) 171 if (ret == 0)
172 goto found; 172 goto found;
173 173
174 if (!item || offset < item_start_offset || 174 if (!item || disk_bytenr < item_start_offset ||
175 offset >= item_last_offset) { 175 disk_bytenr >= item_last_offset) {
176 struct btrfs_key found_key; 176 struct btrfs_key found_key;
177 u32 item_size; 177 u32 item_size;
178 178
179 if (item) 179 if (item)
180 btrfs_release_path(root, path); 180 btrfs_release_path(root, path);
181 item = btrfs_lookup_csum(NULL, root, path, 181 item = btrfs_lookup_csum(NULL, root->fs_info->csum_root,
182 inode->i_ino, offset, 0); 182 path, disk_bytenr, 0);
183 if (IS_ERR(item)) { 183 if (IS_ERR(item)) {
184 ret = PTR_ERR(item); 184 ret = PTR_ERR(item);
185 if (ret == -ENOENT || ret == -EFBIG) 185 if (ret == -ENOENT || ret == -EFBIG)
@@ -208,7 +208,7 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
208 * this byte range must be able to fit inside 208 * this byte range must be able to fit inside
209 * a single leaf so it will also fit inside a u32 209 * a single leaf so it will also fit inside a u32
210 */ 210 */
211 diff = offset - item_start_offset; 211 diff = disk_bytenr - item_start_offset;
212 diff = diff / root->sectorsize; 212 diff = diff / root->sectorsize;
213 diff = diff * csum_size; 213 diff = diff * csum_size;
214 214
@@ -216,7 +216,11 @@ int btrfs_lookup_bio_sums(struct btrfs_root *root, struct inode *inode,
216 ((unsigned long)item) + diff, 216 ((unsigned long)item) + diff,
217 csum_size); 217 csum_size);
218found: 218found:
219 set_state_private(io_tree, offset, sum); 219 if (dst)
220 *dst++ = sum;
221 else
222 set_state_private(io_tree, offset, sum);
223 disk_bytenr += bvec->bv_len;
220 bio_index++; 224 bio_index++;
221 bvec++; 225 bvec++;
222 } 226 }
@@ -224,75 +228,8 @@ found:
224 return 0; 228 return 0;
225} 229}
226 230
227int btrfs_csum_file_bytes(struct btrfs_root *root, struct inode *inode,
228 u64 start, unsigned long len)
229{
230 struct btrfs_ordered_sum *sums;
231 struct btrfs_sector_sum *sector_sum;
232 struct btrfs_ordered_extent *ordered;
233 char *data;
234 struct page *page;
235 unsigned long total_bytes = 0;
236 unsigned long this_sum_bytes = 0;
237
238 sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS);
239 if (!sums)
240 return -ENOMEM;
241
242 sector_sum = sums->sums;
243 sums->file_offset = start;
244 sums->len = len;
245 INIT_LIST_HEAD(&sums->list);
246 ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset);
247 BUG_ON(!ordered);
248
249 while(len > 0) {
250 if (start >= ordered->file_offset + ordered->len ||
251 start < ordered->file_offset) {
252 sums->len = this_sum_bytes;
253 this_sum_bytes = 0;
254 btrfs_add_ordered_sum(inode, ordered, sums);
255 btrfs_put_ordered_extent(ordered);
256
257 sums = kzalloc(btrfs_ordered_sum_size(root, len),
258 GFP_NOFS);
259 BUG_ON(!sums);
260 sector_sum = sums->sums;
261 sums->len = len;
262 sums->file_offset = start;
263 ordered = btrfs_lookup_ordered_extent(inode,
264 sums->file_offset);
265 BUG_ON(!ordered);
266 }
267
268 page = find_get_page(inode->i_mapping,
269 start >> PAGE_CACHE_SHIFT);
270
271 data = kmap_atomic(page, KM_USER0);
272 sector_sum->sum = ~(u32)0;
273 sector_sum->sum = btrfs_csum_data(root, data, sector_sum->sum,
274 PAGE_CACHE_SIZE);
275 kunmap_atomic(data, KM_USER0);
276 btrfs_csum_final(sector_sum->sum,
277 (char *)&sector_sum->sum);
278 sector_sum->offset = page_offset(page);
279 page_cache_release(page);
280
281 sector_sum++;
282 total_bytes += PAGE_CACHE_SIZE;
283 this_sum_bytes += PAGE_CACHE_SIZE;
284 start += PAGE_CACHE_SIZE;
285
286 WARN_ON(len < PAGE_CACHE_SIZE);
287 len -= PAGE_CACHE_SIZE;
288 }
289 btrfs_add_ordered_sum(inode, ordered, sums);
290 btrfs_put_ordered_extent(ordered);
291 return 0;
292}
293
294int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode, 231int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
295 struct bio *bio) 232 struct bio *bio, u64 file_start, int contig)
296{ 233{
297 struct btrfs_ordered_sum *sums; 234 struct btrfs_ordered_sum *sums;
298 struct btrfs_sector_sum *sector_sum; 235 struct btrfs_sector_sum *sector_sum;
@@ -303,6 +240,7 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
303 unsigned long total_bytes = 0; 240 unsigned long total_bytes = 0;
304 unsigned long this_sum_bytes = 0; 241 unsigned long this_sum_bytes = 0;
305 u64 offset; 242 u64 offset;
243 u64 disk_bytenr;
306 244
307 WARN_ON(bio->bi_vcnt <= 0); 245 WARN_ON(bio->bi_vcnt <= 0);
308 sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS); 246 sums = kzalloc(btrfs_ordered_sum_size(root, bio->bi_size), GFP_NOFS);
@@ -310,16 +248,25 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
310 return -ENOMEM; 248 return -ENOMEM;
311 249
312 sector_sum = sums->sums; 250 sector_sum = sums->sums;
313 sums->file_offset = page_offset(bvec->bv_page) + bvec->bv_offset; 251 disk_bytenr = (u64)bio->bi_sector << 9;
314 sums->len = bio->bi_size; 252 sums->len = bio->bi_size;
315 INIT_LIST_HEAD(&sums->list); 253 INIT_LIST_HEAD(&sums->list);
316 ordered = btrfs_lookup_ordered_extent(inode, sums->file_offset); 254
255 if (contig)
256 offset = file_start;
257 else
258 offset = page_offset(bvec->bv_page) + bvec->bv_offset;
259
260 ordered = btrfs_lookup_ordered_extent(inode, offset);
317 BUG_ON(!ordered); 261 BUG_ON(!ordered);
262 sums->bytenr = ordered->start;
318 263
319 while(bio_index < bio->bi_vcnt) { 264 while(bio_index < bio->bi_vcnt) {
320 offset = page_offset(bvec->bv_page) + bvec->bv_offset; 265 if (!contig)
321 if (offset >= ordered->file_offset + ordered->len || 266 offset = page_offset(bvec->bv_page) + bvec->bv_offset;
322 offset < ordered->file_offset) { 267
268 if (!contig && (offset >= ordered->file_offset + ordered->len ||
269 offset < ordered->file_offset)) {
323 unsigned long bytes_left; 270 unsigned long bytes_left;
324 sums->len = this_sum_bytes; 271 sums->len = this_sum_bytes;
325 this_sum_bytes = 0; 272 this_sum_bytes = 0;
@@ -333,10 +280,9 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
333 BUG_ON(!sums); 280 BUG_ON(!sums);
334 sector_sum = sums->sums; 281 sector_sum = sums->sums;
335 sums->len = bytes_left; 282 sums->len = bytes_left;
336 sums->file_offset = offset; 283 ordered = btrfs_lookup_ordered_extent(inode, offset);
337 ordered = btrfs_lookup_ordered_extent(inode,
338 sums->file_offset);
339 BUG_ON(!ordered); 284 BUG_ON(!ordered);
285 sums->bytenr = ordered->start;
340 } 286 }
341 287
342 data = kmap_atomic(bvec->bv_page, KM_USER0); 288 data = kmap_atomic(bvec->bv_page, KM_USER0);
@@ -348,13 +294,14 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
348 kunmap_atomic(data, KM_USER0); 294 kunmap_atomic(data, KM_USER0);
349 btrfs_csum_final(sector_sum->sum, 295 btrfs_csum_final(sector_sum->sum,
350 (char *)&sector_sum->sum); 296 (char *)&sector_sum->sum);
351 sector_sum->offset = page_offset(bvec->bv_page) + 297 sector_sum->bytenr = disk_bytenr;
352 bvec->bv_offset;
353 298
354 sector_sum++; 299 sector_sum++;
355 bio_index++; 300 bio_index++;
356 total_bytes += bvec->bv_len; 301 total_bytes += bvec->bv_len;
357 this_sum_bytes += bvec->bv_len; 302 this_sum_bytes += bvec->bv_len;
303 disk_bytenr += bvec->bv_len;
304 offset += bvec->bv_len;
358 bvec++; 305 bvec++;
359 } 306 }
360 this_sum_bytes = 0; 307 this_sum_bytes = 0;
@@ -364,11 +311,10 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
364} 311}
365 312
366int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans, 313int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
367 struct btrfs_root *root, struct inode *inode, 314 struct btrfs_root *root,
368 struct btrfs_ordered_sum *sums) 315 struct btrfs_ordered_sum *sums)
369{ 316{
370 u64 objectid = inode->i_ino; 317 u64 bytenr;
371 u64 offset;
372 int ret; 318 int ret;
373 struct btrfs_key file_key; 319 struct btrfs_key file_key;
374 struct btrfs_key found_key; 320 struct btrfs_key found_key;
@@ -396,13 +342,12 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
396again: 342again:
397 next_offset = (u64)-1; 343 next_offset = (u64)-1;
398 found_next = 0; 344 found_next = 0;
399 offset = sector_sum->offset; 345 file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
400 file_key.objectid = objectid; 346 file_key.offset = sector_sum->bytenr;
401 file_key.offset = offset; 347 bytenr = sector_sum->bytenr;
402 btrfs_set_key_type(&file_key, BTRFS_CSUM_ITEM_KEY); 348 btrfs_set_key_type(&file_key, BTRFS_EXTENT_CSUM_KEY);
403 349
404 mutex_lock(&BTRFS_I(inode)->csum_mutex); 350 item = btrfs_lookup_csum(trans, root, path, sector_sum->bytenr, 1);
405 item = btrfs_lookup_csum(trans, root, path, objectid, offset, 1);
406 if (!IS_ERR(item)) { 351 if (!IS_ERR(item)) {
407 leaf = path->nodes[0]; 352 leaf = path->nodes[0];
408 ret = 0; 353 ret = 0;
@@ -432,8 +377,8 @@ again:
432 slot = 0; 377 slot = 0;
433 } 378 }
434 btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot); 379 btrfs_item_key_to_cpu(path->nodes[0], &found_key, slot);
435 if (found_key.objectid != objectid || 380 if (found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
436 found_key.type != BTRFS_CSUM_ITEM_KEY) { 381 found_key.type != BTRFS_EXTENT_CSUM_KEY) {
437 found_next = 1; 382 found_next = 1;
438 goto insert; 383 goto insert;
439 } 384 }
@@ -460,10 +405,10 @@ again:
460 path->slots[0]--; 405 path->slots[0]--;
461 leaf = path->nodes[0]; 406 leaf = path->nodes[0];
462 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]); 407 btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
463 csum_offset = (offset - found_key.offset) >> 408 csum_offset = (bytenr - found_key.offset) >>
464 root->fs_info->sb->s_blocksize_bits; 409 root->fs_info->sb->s_blocksize_bits;
465 if (btrfs_key_type(&found_key) != BTRFS_CSUM_ITEM_KEY || 410 if (btrfs_key_type(&found_key) != BTRFS_EXTENT_CSUM_KEY ||
466 found_key.objectid != objectid || 411 found_key.objectid != BTRFS_EXTENT_CSUM_OBJECTID ||
467 csum_offset >= MAX_CSUM_ITEMS(root, csum_size)) { 412 csum_offset >= MAX_CSUM_ITEMS(root, csum_size)) {
468 goto insert; 413 goto insert;
469 } 414 }
@@ -482,8 +427,18 @@ insert:
482 btrfs_release_path(root, path); 427 btrfs_release_path(root, path);
483 csum_offset = 0; 428 csum_offset = 0;
484 if (found_next) { 429 if (found_next) {
485 u64 tmp = min((u64)i_size_read(inode), next_offset); 430 u64 tmp = total_bytes + root->sectorsize;
486 tmp -= offset & ~((u64)root->sectorsize -1); 431 u64 next_sector = sector_sum->bytenr;
432 struct btrfs_sector_sum *next = sector_sum + 1;
433
434 while(tmp < sums->len) {
435 if (next_sector + root->sectorsize != next->bytenr)
436 break;
437 tmp += root->sectorsize;
438 next_sector = next->bytenr;
439 next++;
440 }
441 tmp = min(tmp, next_offset - file_key.offset);
487 tmp >>= root->fs_info->sb->s_blocksize_bits; 442 tmp >>= root->fs_info->sb->s_blocksize_bits;
488 tmp = max((u64)1, tmp); 443 tmp = max((u64)1, tmp);
489 tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size)); 444 tmp = min(tmp, (u64)MAX_CSUM_ITEMS(root, csum_size));
@@ -510,7 +465,6 @@ found:
510 item_end = (struct btrfs_csum_item *)((unsigned char *)item_end + 465 item_end = (struct btrfs_csum_item *)((unsigned char *)item_end +
511 btrfs_item_size_nr(leaf, path->slots[0])); 466 btrfs_item_size_nr(leaf, path->slots[0]));
512 eb_token = NULL; 467 eb_token = NULL;
513 mutex_unlock(&BTRFS_I(inode)->csum_mutex);
514 cond_resched(); 468 cond_resched();
515next_sector: 469next_sector:
516 470
@@ -541,9 +495,9 @@ next_sector:
541 if (total_bytes < sums->len) { 495 if (total_bytes < sums->len) {
542 item = (struct btrfs_csum_item *)((char *)item + 496 item = (struct btrfs_csum_item *)((char *)item +
543 csum_size); 497 csum_size);
544 if (item < item_end && offset + PAGE_CACHE_SIZE == 498 if (item < item_end && bytenr + PAGE_CACHE_SIZE ==
545 sector_sum->offset) { 499 sector_sum->bytenr) {
546 offset = sector_sum->offset; 500 bytenr = sector_sum->bytenr;
547 goto next_sector; 501 goto next_sector;
548 } 502 }
549 } 503 }
@@ -562,7 +516,6 @@ out:
562 return ret; 516 return ret;
563 517
564fail_unlock: 518fail_unlock:
565 mutex_unlock(&BTRFS_I(inode)->csum_mutex);
566 goto out; 519 goto out;
567} 520}
568 521
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 09efc9473a3d..c03d847b8c4e 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -1221,7 +1221,7 @@ static int __btrfs_submit_bio_start(struct inode *inode, int rw, struct bio *bio
1221 struct btrfs_root *root = BTRFS_I(inode)->root; 1221 struct btrfs_root *root = BTRFS_I(inode)->root;
1222 int ret = 0; 1222 int ret = 0;
1223 1223
1224 ret = btrfs_csum_one_bio(root, inode, bio); 1224 ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1225 BUG_ON(ret); 1225 BUG_ON(ret);
1226 return 0; 1226 return 0;
1227} 1227}
@@ -1259,12 +1259,11 @@ static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
1259 btrfs_test_flag(inode, NODATASUM); 1259 btrfs_test_flag(inode, NODATASUM);
1260 1260
1261 if (!(rw & (1 << BIO_RW))) { 1261 if (!(rw & (1 << BIO_RW))) {
1262 1262 if (bio_flags & EXTENT_BIO_COMPRESSED) {
1263 if (bio_flags & EXTENT_BIO_COMPRESSED)
1264 return btrfs_submit_compressed_read(inode, bio, 1263 return btrfs_submit_compressed_read(inode, bio,
1265 mirror_num, bio_flags); 1264 mirror_num, bio_flags);
1266 else if (!skip_sum) 1265 } else if (!skip_sum)
1267 btrfs_lookup_bio_sums(root, inode, bio); 1266 btrfs_lookup_bio_sums(root, inode, bio, NULL);
1268 goto mapit; 1267 goto mapit;
1269 } else if (!skip_sum) { 1268 } else if (!skip_sum) {
1270 /* we're doing a write, do the async checksumming */ 1269 /* we're doing a write, do the async checksumming */
@@ -1292,8 +1291,8 @@ static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1292 btrfs_set_trans_block_group(trans, inode); 1291 btrfs_set_trans_block_group(trans, inode);
1293 list_for_each(cur, list) { 1292 list_for_each(cur, list) {
1294 sum = list_entry(cur, struct btrfs_ordered_sum, list); 1293 sum = list_entry(cur, struct btrfs_ordered_sum, list);
1295 btrfs_csum_file_blocks(trans, BTRFS_I(inode)->root, 1294 btrfs_csum_file_blocks(trans,
1296 inode, sum); 1295 BTRFS_I(inode)->root->fs_info->csum_root, sum);
1297 } 1296 }
1298 return 0; 1297 return 0;
1299} 1298}
@@ -1545,6 +1544,7 @@ struct io_failure_record {
1545 u64 start; 1544 u64 start;
1546 u64 len; 1545 u64 len;
1547 u64 logical; 1546 u64 logical;
1547 unsigned long bio_flags;
1548 int last_mirror; 1548 int last_mirror;
1549}; 1549};
1550 1550
@@ -1563,7 +1563,6 @@ static int btrfs_io_failed_hook(struct bio *failed_bio,
1563 int ret; 1563 int ret;
1564 int rw; 1564 int rw;
1565 u64 logical; 1565 u64 logical;
1566 unsigned long bio_flags = 0;
1567 1566
1568 ret = get_state_private(failure_tree, start, &private); 1567 ret = get_state_private(failure_tree, start, &private);
1569 if (ret) { 1568 if (ret) {
@@ -1573,6 +1572,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio,
1573 failrec->start = start; 1572 failrec->start = start;
1574 failrec->len = end - start + 1; 1573 failrec->len = end - start + 1;
1575 failrec->last_mirror = 0; 1574 failrec->last_mirror = 0;
1575 failrec->bio_flags = 0;
1576 1576
1577 spin_lock(&em_tree->lock); 1577 spin_lock(&em_tree->lock);
1578 em = lookup_extent_mapping(em_tree, start, failrec->len); 1578 em = lookup_extent_mapping(em_tree, start, failrec->len);
@@ -1588,8 +1588,10 @@ static int btrfs_io_failed_hook(struct bio *failed_bio,
1588 } 1588 }
1589 logical = start - em->start; 1589 logical = start - em->start;
1590 logical = em->block_start + logical; 1590 logical = em->block_start + logical;
1591 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) 1591 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
1592 bio_flags = EXTENT_BIO_COMPRESSED; 1592 logical = em->block_start;
1593 failrec->bio_flags = EXTENT_BIO_COMPRESSED;
1594 }
1593 failrec->logical = logical; 1595 failrec->logical = logical;
1594 free_extent_map(em); 1596 free_extent_map(em);
1595 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED | 1597 set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
@@ -1626,6 +1628,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio,
1626 bio->bi_sector = failrec->logical >> 9; 1628 bio->bi_sector = failrec->logical >> 9;
1627 bio->bi_bdev = failed_bio->bi_bdev; 1629 bio->bi_bdev = failed_bio->bi_bdev;
1628 bio->bi_size = 0; 1630 bio->bi_size = 0;
1631
1629 bio_add_page(bio, page, failrec->len, start - page_offset(page)); 1632 bio_add_page(bio, page, failrec->len, start - page_offset(page));
1630 if (failed_bio->bi_rw & (1 << BIO_RW)) 1633 if (failed_bio->bi_rw & (1 << BIO_RW))
1631 rw = WRITE; 1634 rw = WRITE;
@@ -1634,7 +1637,7 @@ static int btrfs_io_failed_hook(struct bio *failed_bio,
1634 1637
1635 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio, 1638 BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
1636 failrec->last_mirror, 1639 failrec->last_mirror,
1637 bio_flags); 1640 failrec->bio_flags);
1638 return 0; 1641 return 0;
1639} 1642}
1640 1643
@@ -1688,9 +1691,14 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
1688 u32 csum = ~(u32)0; 1691 u32 csum = ~(u32)0;
1689 unsigned long flags; 1692 unsigned long flags;
1690 1693
1694 if (PageChecked(page)) {
1695 ClearPageChecked(page);
1696 goto good;
1697 }
1691 if (btrfs_test_opt(root, NODATASUM) || 1698 if (btrfs_test_opt(root, NODATASUM) ||
1692 btrfs_test_flag(inode, NODATASUM)) 1699 btrfs_test_flag(inode, NODATASUM))
1693 return 0; 1700 return 0;
1701
1694 if (state && state->start == start) { 1702 if (state && state->start == start) {
1695 private = state->private; 1703 private = state->private;
1696 ret = 0; 1704 ret = 0;
@@ -1709,7 +1717,7 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
1709 } 1717 }
1710 kunmap_atomic(kaddr, KM_IRQ0); 1718 kunmap_atomic(kaddr, KM_IRQ0);
1711 local_irq_restore(flags); 1719 local_irq_restore(flags);
1712 1720good:
1713 /* if the io failure tree for this inode is non-empty, 1721 /* if the io failure tree for this inode is non-empty,
1714 * check to see if we've recovered from a failed IO 1722 * check to see if we've recovered from a failed IO
1715 */ 1723 */
@@ -2243,6 +2251,7 @@ fail:
2243 return err; 2251 return err;
2244} 2252}
2245 2253
2254#if 0
2246/* 2255/*
2247 * when truncating bytes in a file, it is possible to avoid reading 2256 * when truncating bytes in a file, it is possible to avoid reading
2248 * the leaves that contain only checksum items. This can be the 2257 * the leaves that contain only checksum items. This can be the
@@ -2410,6 +2419,8 @@ out:
2410 return ret; 2419 return ret;
2411} 2420}
2412 2421
2422#endif
2423
2413/* 2424/*
2414 * this can truncate away extent items, csum items and directory items. 2425 * this can truncate away extent items, csum items and directory items.
2415 * It starts at a high offset and removes keys until it can't find 2426 * It starts at a high offset and removes keys until it can't find
@@ -2459,9 +2470,6 @@ noinline int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
2459 2470
2460 btrfs_init_path(path); 2471 btrfs_init_path(path);
2461 2472
2462 ret = drop_csum_leaves(trans, root, path, inode, new_size);
2463 BUG_ON(ret);
2464
2465search_again: 2473search_again:
2466 ret = btrfs_search_slot(trans, root, &key, path, -1, 1); 2474 ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
2467 if (ret < 0) { 2475 if (ret < 0) {
@@ -2509,16 +2517,11 @@ search_again:
2509 } 2517 }
2510 item_end--; 2518 item_end--;
2511 } 2519 }
2512 if (found_type == BTRFS_CSUM_ITEM_KEY) {
2513 ret = btrfs_csum_truncate(trans, root, path,
2514 new_size);
2515 BUG_ON(ret);
2516 }
2517 if (item_end < new_size) { 2520 if (item_end < new_size) {
2518 if (found_type == BTRFS_DIR_ITEM_KEY) { 2521 if (found_type == BTRFS_DIR_ITEM_KEY) {
2519 found_type = BTRFS_INODE_ITEM_KEY; 2522 found_type = BTRFS_INODE_ITEM_KEY;
2520 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) { 2523 } else if (found_type == BTRFS_EXTENT_ITEM_KEY) {
2521 found_type = BTRFS_CSUM_ITEM_KEY; 2524 found_type = BTRFS_EXTENT_DATA_KEY;
2522 } else if (found_type == BTRFS_EXTENT_DATA_KEY) { 2525 } else if (found_type == BTRFS_EXTENT_DATA_KEY) {
2523 found_type = BTRFS_XATTR_ITEM_KEY; 2526 found_type = BTRFS_XATTR_ITEM_KEY;
2524 } else if (found_type == BTRFS_XATTR_ITEM_KEY) { 2527 } else if (found_type == BTRFS_XATTR_ITEM_KEY) {
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index b4da53d55c82..6228b69c2b93 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -714,8 +714,7 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
714 u64 len = olen; 714 u64 len = olen;
715 u64 bs = root->fs_info->sb->s_blocksize; 715 u64 bs = root->fs_info->sb->s_blocksize;
716 u64 hint_byte; 716 u64 hint_byte;
717 u16 csum_size = 717
718 btrfs_super_csum_size(&root->fs_info->super_copy);
719 /* 718 /*
720 * TODO: 719 * TODO:
721 * - split compressed inline extents. annoying: we need to 720 * - split compressed inline extents. annoying: we need to
@@ -833,7 +832,7 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
833 slot = path->slots[0]; 832 slot = path->slots[0];
834 833
835 btrfs_item_key_to_cpu(leaf, &key, slot); 834 btrfs_item_key_to_cpu(leaf, &key, slot);
836 if (btrfs_key_type(&key) > BTRFS_CSUM_ITEM_KEY || 835 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
837 key.objectid != src->i_ino) 836 key.objectid != src->i_ino)
838 break; 837 break;
839 838
@@ -958,56 +957,6 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
958 btrfs_mark_buffer_dirty(leaf); 957 btrfs_mark_buffer_dirty(leaf);
959 } 958 }
960 959
961 if (btrfs_key_type(&key) == BTRFS_CSUM_ITEM_KEY) {
962 u32 size;
963 struct btrfs_key new_key;
964 u64 coverslen;
965 int coff, clen;
966
967 size = btrfs_item_size_nr(leaf, slot);
968 coverslen = (size / csum_size) <<
969 root->fs_info->sb->s_blocksize_bits;
970 printk("csums for %llu~%llu\n",
971 key.offset, coverslen);
972 if (key.offset + coverslen < off ||
973 key.offset >= off+len)
974 goto next;
975
976 read_extent_buffer(leaf, buf,
977 btrfs_item_ptr_offset(leaf, slot),
978 size);
979 btrfs_release_path(root, path);
980
981 coff = 0;
982 if (off > key.offset)
983 coff = ((off - key.offset) >>
984 root->fs_info->sb->s_blocksize_bits) *
985 csum_size;
986 clen = size - coff;
987 if (key.offset + coverslen > off+len)
988 clen -= ((key.offset+coverslen-off-len) >>
989 root->fs_info->sb->s_blocksize_bits) *
990 csum_size;
991 printk(" will dup %d~%d of %d\n",
992 coff, clen, size);
993
994 memcpy(&new_key, &key, sizeof(new_key));
995 new_key.objectid = inode->i_ino;
996 new_key.offset = key.offset + destoff - off;
997
998 ret = btrfs_insert_empty_item(trans, root, path,
999 &new_key, clen);
1000 if (ret)
1001 goto out;
1002
1003 leaf = path->nodes[0];
1004 slot = path->slots[0];
1005 write_extent_buffer(leaf, buf + coff,
1006 btrfs_item_ptr_offset(leaf, slot),
1007 clen);
1008 btrfs_mark_buffer_dirty(leaf);
1009 }
1010
1011 next: 960 next:
1012 btrfs_release_path(root, path); 961 btrfs_release_path(root, path);
1013 key.offset++; 962 key.offset++;
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index 027ad6b3839e..d9e232227da4 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -610,7 +610,8 @@ out:
610 * try to find a checksum. This is used because we allow pages to 610 * try to find a checksum. This is used because we allow pages to
611 * be reclaimed before their checksum is actually put into the btree 611 * be reclaimed before their checksum is actually put into the btree
612 */ 612 */
613int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum) 613int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr,
614 u32 *sum)
614{ 615{
615 struct btrfs_ordered_sum *ordered_sum; 616 struct btrfs_ordered_sum *ordered_sum;
616 struct btrfs_sector_sum *sector_sums; 617 struct btrfs_sector_sum *sector_sums;
@@ -629,11 +630,11 @@ int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum)
629 mutex_lock(&tree->mutex); 630 mutex_lock(&tree->mutex);
630 list_for_each_prev(cur, &ordered->list) { 631 list_for_each_prev(cur, &ordered->list) {
631 ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list); 632 ordered_sum = list_entry(cur, struct btrfs_ordered_sum, list);
632 if (offset >= ordered_sum->file_offset) { 633 if (disk_bytenr >= ordered_sum->bytenr) {
633 num_sectors = ordered_sum->len / sectorsize; 634 num_sectors = ordered_sum->len / sectorsize;
634 sector_sums = ordered_sum->sums; 635 sector_sums = ordered_sum->sums;
635 for (i = 0; i < num_sectors; i++) { 636 for (i = 0; i < num_sectors; i++) {
636 if (sector_sums[i].offset == offset) { 637 if (sector_sums[i].bytenr == disk_bytenr) {
637 *sum = sector_sums[i].sum; 638 *sum = sector_sums[i].sum;
638 ret = 0; 639 ret = 0;
639 goto out; 640 goto out;
diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h
index 260bf95dfe0c..ab66d5e8d6d6 100644
--- a/fs/btrfs/ordered-data.h
+++ b/fs/btrfs/ordered-data.h
@@ -33,15 +33,17 @@ struct btrfs_ordered_inode_tree {
33 * the ordered extent are on disk 33 * the ordered extent are on disk
34 */ 34 */
35struct btrfs_sector_sum { 35struct btrfs_sector_sum {
36 u64 offset; 36 /* bytenr on disk */
37 u64 bytenr;
37 u32 sum; 38 u32 sum;
38}; 39};
39 40
40struct btrfs_ordered_sum { 41struct btrfs_ordered_sum {
41 u64 file_offset; 42 /* bytenr is the start of this extent on disk */
43 u64 bytenr;
44
42 /* 45 /*
43 * this is the length in bytes covered by the sums array below. 46 * this is the length in bytes covered by the sums array below.
44 * But, the sums array may not be contiguous in the file.
45 */ 47 */
46 unsigned long len; 48 unsigned long len;
47 struct list_head list; 49 struct list_head list;
@@ -147,7 +149,7 @@ struct btrfs_ordered_extent *
147btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset); 149btrfs_lookup_first_ordered_extent(struct inode * inode, u64 file_offset);
148int btrfs_ordered_update_i_size(struct inode *inode, 150int btrfs_ordered_update_i_size(struct inode *inode,
149 struct btrfs_ordered_extent *ordered); 151 struct btrfs_ordered_extent *ordered);
150int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u32 *sum); 152int btrfs_find_ordered_sum(struct inode *inode, u64 offset, u64 disk_bytenr, u32 *sum);
151int btrfs_wait_on_page_writeback_range(struct address_space *mapping, 153int btrfs_wait_on_page_writeback_range(struct address_space *mapping,
152 pgoff_t start, pgoff_t end); 154 pgoff_t start, pgoff_t end);
153int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start, 155int btrfs_fdatawrite_range(struct address_space *mapping, loff_t start,
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index c766649ad453..08469ec05850 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -934,24 +934,17 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans,
934 unsigned long file_bytes; 934 unsigned long file_bytes;
935 struct btrfs_ordered_sum *sums; 935 struct btrfs_ordered_sum *sums;
936 struct btrfs_sector_sum *sector_sum; 936 struct btrfs_sector_sum *sector_sum;
937 struct inode *inode;
938 unsigned long ptr; 937 unsigned long ptr;
939 938
940 file_bytes = (item_size / csum_size) * root->sectorsize; 939 file_bytes = (item_size / csum_size) * root->sectorsize;
941 inode = read_one_inode(root, key->objectid);
942 if (!inode) {
943 return -EIO;
944 }
945
946 sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS); 940 sums = kzalloc(btrfs_ordered_sum_size(root, file_bytes), GFP_NOFS);
947 if (!sums) { 941 if (!sums) {
948 iput(inode);
949 return -ENOMEM; 942 return -ENOMEM;
950 } 943 }
951 944
952 INIT_LIST_HEAD(&sums->list); 945 INIT_LIST_HEAD(&sums->list);
953 sums->len = file_bytes; 946 sums->len = file_bytes;
954 sums->file_offset = key->offset; 947 sums->bytenr = key->offset;
955 948
956 /* 949 /*
957 * copy all the sums into the ordered sum struct 950 * copy all the sums into the ordered sum struct
@@ -960,7 +953,7 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans,
960 cur_offset = key->offset; 953 cur_offset = key->offset;
961 ptr = btrfs_item_ptr_offset(eb, slot); 954 ptr = btrfs_item_ptr_offset(eb, slot);
962 while(item_size > 0) { 955 while(item_size > 0) {
963 sector_sum->offset = cur_offset; 956 sector_sum->bytenr = cur_offset;
964 read_extent_buffer(eb, &sector_sum->sum, ptr, csum_size); 957 read_extent_buffer(eb, &sector_sum->sum, ptr, csum_size);
965 sector_sum++; 958 sector_sum++;
966 item_size -= csum_size; 959 item_size -= csum_size;
@@ -969,11 +962,9 @@ static noinline int replay_one_csum(struct btrfs_trans_handle *trans,
969 } 962 }
970 963
971 /* let btrfs_csum_file_blocks add them into the file */ 964 /* let btrfs_csum_file_blocks add them into the file */
972 ret = btrfs_csum_file_blocks(trans, root, inode, sums); 965 ret = btrfs_csum_file_blocks(trans, root->fs_info->csum_root, sums);
973 BUG_ON(ret); 966 BUG_ON(ret);
974 kfree(sums); 967 kfree(sums);
975 iput(inode);
976
977 return 0; 968 return 0;
978} 969}
979/* 970/*
@@ -1670,7 +1661,7 @@ static int replay_one_buffer(struct btrfs_root *log, struct extent_buffer *eb,
1670 ret = replay_one_extent(wc->trans, root, path, 1661 ret = replay_one_extent(wc->trans, root, path,
1671 eb, i, &key); 1662 eb, i, &key);
1672 BUG_ON(ret); 1663 BUG_ON(ret);
1673 } else if (key.type == BTRFS_CSUM_ITEM_KEY) { 1664 } else if (key.type == BTRFS_EXTENT_CSUM_KEY) {
1674 ret = replay_one_csum(wc->trans, root, path, 1665 ret = replay_one_csum(wc->trans, root, path,
1675 eb, i, &key); 1666 eb, i, &key);
1676 BUG_ON(ret); 1667 BUG_ON(ret);
@@ -2466,6 +2457,85 @@ static int drop_objectid_items(struct btrfs_trans_handle *trans,
2466 return 0; 2457 return 0;
2467} 2458}
2468 2459
2460static noinline int copy_extent_csums(struct btrfs_trans_handle *trans,
2461 struct list_head *list,
2462 struct btrfs_root *root,
2463 u64 disk_bytenr, u64 len)
2464{
2465 struct btrfs_ordered_sum *sums;
2466 struct btrfs_sector_sum *sector_sum;
2467 int ret;
2468 struct btrfs_path *path;
2469 struct btrfs_csum_item *item = NULL;
2470 u64 end = disk_bytenr + len;
2471 u64 item_start_offset = 0;
2472 u64 item_last_offset = 0;
2473 u32 diff;
2474 u32 sum;
2475 u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);
2476
2477 sums = kzalloc(btrfs_ordered_sum_size(root, len), GFP_NOFS);
2478
2479 sector_sum = sums->sums;
2480 sums->bytenr = disk_bytenr;
2481 sums->len = len;
2482 list_add_tail(&sums->list, list);
2483
2484 path = btrfs_alloc_path();
2485 while(disk_bytenr < end) {
2486 if (!item || disk_bytenr < item_start_offset ||
2487 disk_bytenr >= item_last_offset) {
2488 struct btrfs_key found_key;
2489 u32 item_size;
2490
2491 if (item)
2492 btrfs_release_path(root, path);
2493 item = btrfs_lookup_csum(NULL, root, path,
2494 disk_bytenr, 0);
2495 if (IS_ERR(item)) {
2496 ret = PTR_ERR(item);
2497 if (ret == -ENOENT || ret == -EFBIG)
2498 ret = 0;
2499 sum = 0;
2500 printk("log no csum found for byte %llu\n",
2501 (unsigned long long)disk_bytenr);
2502 item = NULL;
2503 btrfs_release_path(root, path);
2504 goto found;
2505 }
2506 btrfs_item_key_to_cpu(path->nodes[0], &found_key,
2507 path->slots[0]);
2508
2509 item_start_offset = found_key.offset;
2510 item_size = btrfs_item_size_nr(path->nodes[0],
2511 path->slots[0]);
2512 item_last_offset = item_start_offset +
2513 (item_size / csum_size) *
2514 root->sectorsize;
2515 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2516 struct btrfs_csum_item);
2517 }
2518 /*
2519 * this byte range must be able to fit inside
2520 * a single leaf so it will also fit inside a u32
2521 */
2522 diff = disk_bytenr - item_start_offset;
2523 diff = diff / root->sectorsize;
2524 diff = diff * csum_size;
2525
2526 read_extent_buffer(path->nodes[0], &sum,
2527 ((unsigned long)item) + diff,
2528 csum_size);
2529found:
2530 sector_sum->bytenr = disk_bytenr;
2531 sector_sum->sum = sum;
2532 disk_bytenr += root->sectorsize;
2533 sector_sum++;
2534 }
2535 btrfs_free_path(path);
2536 return 0;
2537}
2538
2469static noinline int copy_items(struct btrfs_trans_handle *trans, 2539static noinline int copy_items(struct btrfs_trans_handle *trans,
2470 struct btrfs_root *log, 2540 struct btrfs_root *log,
2471 struct btrfs_path *dst_path, 2541 struct btrfs_path *dst_path,
@@ -2481,6 +2551,9 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
2481 u32 *ins_sizes; 2551 u32 *ins_sizes;
2482 char *ins_data; 2552 char *ins_data;
2483 int i; 2553 int i;
2554 struct list_head ordered_sums;
2555
2556 INIT_LIST_HEAD(&ordered_sums);
2484 2557
2485 ins_data = kmalloc(nr * sizeof(struct btrfs_key) + 2558 ins_data = kmalloc(nr * sizeof(struct btrfs_key) +
2486 nr * sizeof(u32), GFP_NOFS); 2559 nr * sizeof(u32), GFP_NOFS);
@@ -2535,6 +2608,9 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
2535 extent); 2608 extent);
2536 u64 dl = btrfs_file_extent_disk_num_bytes(src, 2609 u64 dl = btrfs_file_extent_disk_num_bytes(src,
2537 extent); 2610 extent);
2611 u64 cs = btrfs_file_extent_offset(src, extent);
2612 u64 cl = btrfs_file_extent_num_bytes(src,
2613 extent);;
2538 /* ds == 0 is a hole */ 2614 /* ds == 0 is a hole */
2539 if (ds != 0) { 2615 if (ds != 0) {
2540 ret = btrfs_inc_extent_ref(trans, log, 2616 ret = btrfs_inc_extent_ref(trans, log,
@@ -2544,6 +2620,11 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
2544 trans->transid, 2620 trans->transid,
2545 ins_keys[i].objectid); 2621 ins_keys[i].objectid);
2546 BUG_ON(ret); 2622 BUG_ON(ret);
2623 ret = copy_extent_csums(trans,
2624 &ordered_sums,
2625 log->fs_info->csum_root,
2626 ds + cs, cl);
2627 BUG_ON(ret);
2547 } 2628 }
2548 } 2629 }
2549 } 2630 }
@@ -2553,6 +2634,20 @@ static noinline int copy_items(struct btrfs_trans_handle *trans,
2553 btrfs_mark_buffer_dirty(dst_path->nodes[0]); 2634 btrfs_mark_buffer_dirty(dst_path->nodes[0]);
2554 btrfs_release_path(log, dst_path); 2635 btrfs_release_path(log, dst_path);
2555 kfree(ins_data); 2636 kfree(ins_data);
2637
2638 /*
2639 * we have to do this after the loop above to avoid changing the
2640 * log tree while trying to change the log tree.
2641 */
2642 while(!list_empty(&ordered_sums)) {
2643 struct btrfs_ordered_sum *sums = list_entry(ordered_sums.next,
2644 struct btrfs_ordered_sum,
2645 list);
2646 ret = btrfs_csum_file_blocks(trans, log, sums);
2647 BUG_ON(ret);
2648 list_del(&sums->list);
2649 kfree(sums);
2650 }
2556 return 0; 2651 return 0;
2557} 2652}
2558 2653
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 6c523b3360f6..2049d179ccd5 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -2771,6 +2771,7 @@ static struct btrfs_device *add_missing_dev(struct btrfs_root *root,
2771 device->work.func = pending_bios_fn; 2771 device->work.func = pending_bios_fn;
2772 fs_devices->num_devices++; 2772 fs_devices->num_devices++;
2773 spin_lock_init(&device->io_lock); 2773 spin_lock_init(&device->io_lock);
2774 INIT_LIST_HEAD(&device->dev_alloc_list);
2774 memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE); 2775 memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE);
2775 return device; 2776 return device;
2776} 2777}
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-04 22:56:59 -0500