aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--fs/btrfs/compression.c15
-rw-r--r--fs/btrfs/ctree.h6
-rw-r--r--fs/btrfs/disk-io.c38
-rw-r--r--fs/btrfs/export.c76
-rw-r--r--fs/btrfs/extent-tree.c2
-rw-r--r--fs/btrfs/extent_io.c77
-rw-r--r--fs/btrfs/extent_io.h3
-rw-r--r--fs/btrfs/file.c7
-rw-r--r--fs/btrfs/inode.c294
-rw-r--r--fs/btrfs/ioctl.c31
-rw-r--r--fs/btrfs/ordered-data.c67
-rw-r--r--fs/btrfs/ordered-data.h3
-rw-r--r--fs/btrfs/super.c41
-rw-r--r--fs/btrfs/transaction.c5
-rw-r--r--fs/btrfs/tree-log.c21
15 files changed, 572 insertions, 114 deletions
diff --git a/fs/btrfs/compression.c b/fs/btrfs/compression.c
index 7845d1f7d1d9..b50bc4bd5c56 100644
--- a/fs/btrfs/compression.c
+++ b/fs/btrfs/compression.c
@@ -91,23 +91,10 @@ static inline int compressed_bio_size(struct btrfs_root *root,
91static struct bio *compressed_bio_alloc(struct block_device *bdev, 91static struct bio *compressed_bio_alloc(struct block_device *bdev,
92 u64 first_byte, gfp_t gfp_flags) 92 u64 first_byte, gfp_t gfp_flags)
93{ 93{
94 struct bio *bio;
95 int nr_vecs; 94 int nr_vecs;
96 95
97 nr_vecs = bio_get_nr_vecs(bdev); 96 nr_vecs = bio_get_nr_vecs(bdev);
98 bio = bio_alloc(gfp_flags, nr_vecs); 97 return btrfs_bio_alloc(bdev, first_byte >> 9, nr_vecs, gfp_flags);
99
100 if (bio == NULL && (current->flags & PF_MEMALLOC)) {
101 while (!bio && (nr_vecs /= 2))
102 bio = bio_alloc(gfp_flags, nr_vecs);
103 }
104
105 if (bio) {
106 bio->bi_size = 0;
107 bio->bi_bdev = bdev;
108 bio->bi_sector = first_byte >> 9;
109 }
110 return bio;
111} 98}
112 99
113static int check_compressed_csum(struct inode *inode, 100static int check_compressed_csum(struct inode *inode,
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 8db9234f6b41..af52f6d7a4d8 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -808,9 +808,9 @@ struct btrfs_block_group_cache {
808 int extents_thresh; 808 int extents_thresh;
809 int free_extents; 809 int free_extents;
810 int total_bitmaps; 810 int total_bitmaps;
811 int ro:1; 811 unsigned int ro:1;
812 int dirty:1; 812 unsigned int dirty:1;
813 int iref:1; 813 unsigned int iref:1;
814 814
815 int disk_cache_state; 815 int disk_cache_state;
816 816
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index fb827d0d7181..c547cca26a26 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -28,6 +28,7 @@
28#include <linux/freezer.h> 28#include <linux/freezer.h>
29#include <linux/crc32c.h> 29#include <linux/crc32c.h>
30#include <linux/slab.h> 30#include <linux/slab.h>
31#include <linux/migrate.h>
31#include "compat.h" 32#include "compat.h"
32#include "ctree.h" 33#include "ctree.h"
33#include "disk-io.h" 34#include "disk-io.h"
@@ -355,6 +356,8 @@ static int csum_dirty_buffer(struct btrfs_root *root, struct page *page)
355 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE, 356 ret = btree_read_extent_buffer_pages(root, eb, start + PAGE_CACHE_SIZE,
356 btrfs_header_generation(eb)); 357 btrfs_header_generation(eb));
357 BUG_ON(ret); 358 BUG_ON(ret);
359 WARN_ON(!btrfs_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN));
360
358 found_start = btrfs_header_bytenr(eb); 361 found_start = btrfs_header_bytenr(eb);
359 if (found_start != start) { 362 if (found_start != start) {
360 WARN_ON(1); 363 WARN_ON(1);
@@ -693,6 +696,29 @@ static int btree_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
693 __btree_submit_bio_done); 696 __btree_submit_bio_done);
694} 697}
695 698
699static int btree_migratepage(struct address_space *mapping,
700 struct page *newpage, struct page *page)
701{
702 /*
703 * we can't safely write a btree page from here,
704 * we haven't done the locking hook
705 */
706 if (PageDirty(page))
707 return -EAGAIN;
708 /*
709 * Buffers may be managed in a filesystem specific way.
710 * We must have no buffers or drop them.
711 */
712 if (page_has_private(page) &&
713 !try_to_release_page(page, GFP_KERNEL))
714 return -EAGAIN;
715#ifdef CONFIG_MIGRATION
716 return migrate_page(mapping, newpage, page);
717#else
718 return -ENOSYS;
719#endif
720}
721
696static int btree_writepage(struct page *page, struct writeback_control *wbc) 722static int btree_writepage(struct page *page, struct writeback_control *wbc)
697{ 723{
698 struct extent_io_tree *tree; 724 struct extent_io_tree *tree;
@@ -707,8 +733,7 @@ static int btree_writepage(struct page *page, struct writeback_control *wbc)
707 } 733 }
708 734
709 redirty_page_for_writepage(wbc, page); 735 redirty_page_for_writepage(wbc, page);
710 eb = btrfs_find_tree_block(root, page_offset(page), 736 eb = btrfs_find_tree_block(root, page_offset(page), PAGE_CACHE_SIZE);
711 PAGE_CACHE_SIZE);
712 WARN_ON(!eb); 737 WARN_ON(!eb);
713 738
714 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags); 739 was_dirty = test_and_set_bit(EXTENT_BUFFER_DIRTY, &eb->bflags);
@@ -799,6 +824,9 @@ static const struct address_space_operations btree_aops = {
799 .releasepage = btree_releasepage, 824 .releasepage = btree_releasepage,
800 .invalidatepage = btree_invalidatepage, 825 .invalidatepage = btree_invalidatepage,
801 .sync_page = block_sync_page, 826 .sync_page = block_sync_page,
827#ifdef CONFIG_MIGRATION
828 .migratepage = btree_migratepage,
829#endif
802}; 830};
803 831
804int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize, 832int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
@@ -1538,10 +1566,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
1538 GFP_NOFS); 1566 GFP_NOFS);
1539 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root), 1567 struct btrfs_root *csum_root = kzalloc(sizeof(struct btrfs_root),
1540 GFP_NOFS); 1568 GFP_NOFS);
1541 struct btrfs_root *tree_root = kzalloc(sizeof(struct btrfs_root), 1569 struct btrfs_root *tree_root = btrfs_sb(sb);
1542 GFP_NOFS); 1570 struct btrfs_fs_info *fs_info = tree_root->fs_info;
1543 struct btrfs_fs_info *fs_info = kzalloc(sizeof(*fs_info),
1544 GFP_NOFS);
1545 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root), 1571 struct btrfs_root *chunk_root = kzalloc(sizeof(struct btrfs_root),
1546 GFP_NOFS); 1572 GFP_NOFS);
1547 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root), 1573 struct btrfs_root *dev_root = kzalloc(sizeof(struct btrfs_root),
diff --git a/fs/btrfs/export.c b/fs/btrfs/export.c
index 951ef09b82f4..6f0444473594 100644
--- a/fs/btrfs/export.c
+++ b/fs/btrfs/export.c
@@ -232,9 +232,85 @@ fail:
232 return ERR_PTR(ret); 232 return ERR_PTR(ret);
233} 233}
234 234
235static int btrfs_get_name(struct dentry *parent, char *name,
236 struct dentry *child)
237{
238 struct inode *inode = child->d_inode;
239 struct inode *dir = parent->d_inode;
240 struct btrfs_path *path;
241 struct btrfs_root *root = BTRFS_I(dir)->root;
242 struct btrfs_inode_ref *iref;
243 struct btrfs_root_ref *rref;
244 struct extent_buffer *leaf;
245 unsigned long name_ptr;
246 struct btrfs_key key;
247 int name_len;
248 int ret;
249
250 if (!dir || !inode)
251 return -EINVAL;
252
253 if (!S_ISDIR(dir->i_mode))
254 return -EINVAL;
255
256 path = btrfs_alloc_path();
257 if (!path)
258 return -ENOMEM;
259 path->leave_spinning = 1;
260
261 if (inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) {
262 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
263 key.type = BTRFS_ROOT_BACKREF_KEY;
264 key.offset = (u64)-1;
265 root = root->fs_info->tree_root;
266 } else {
267 key.objectid = inode->i_ino;
268 key.offset = dir->i_ino;
269 key.type = BTRFS_INODE_REF_KEY;
270 }
271
272 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
273 if (ret < 0) {
274 btrfs_free_path(path);
275 return ret;
276 } else if (ret > 0) {
277 if (inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) {
278 path->slots[0]--;
279 } else {
280 btrfs_free_path(path);
281 return -ENOENT;
282 }
283 }
284 leaf = path->nodes[0];
285
286 if (inode->i_ino == BTRFS_FIRST_FREE_OBJECTID) {
287 rref = btrfs_item_ptr(leaf, path->slots[0],
288 struct btrfs_root_ref);
289 name_ptr = (unsigned long)(rref + 1);
290 name_len = btrfs_root_ref_name_len(leaf, rref);
291 } else {
292 iref = btrfs_item_ptr(leaf, path->slots[0],
293 struct btrfs_inode_ref);
294 name_ptr = (unsigned long)(iref + 1);
295 name_len = btrfs_inode_ref_name_len(leaf, iref);
296 }
297
298 read_extent_buffer(leaf, name, name_ptr, name_len);
299 btrfs_free_path(path);
300
301 /*
302 * have to add the null termination to make sure that reconnect_path
303 * gets the right len for strlen
304 */
305 name[name_len] = '\0';
306
307 return 0;
308}
309
235const struct export_operations btrfs_export_ops = { 310const struct export_operations btrfs_export_ops = {
236 .encode_fh = btrfs_encode_fh, 311 .encode_fh = btrfs_encode_fh,
237 .fh_to_dentry = btrfs_fh_to_dentry, 312 .fh_to_dentry = btrfs_fh_to_dentry,
238 .fh_to_parent = btrfs_fh_to_parent, 313 .fh_to_parent = btrfs_fh_to_parent,
239 .get_parent = btrfs_get_parent, 314 .get_parent = btrfs_get_parent,
315 .get_name = btrfs_get_name,
240}; 316};
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 0c097f3aec41..bcd59c7dfb57 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3412,7 +3412,7 @@ again:
3412 * our reservation. 3412 * our reservation.
3413 */ 3413 */
3414 if (unused <= space_info->total_bytes) { 3414 if (unused <= space_info->total_bytes) {
3415 unused -= space_info->total_bytes; 3415 unused = space_info->total_bytes - unused;
3416 if (unused >= num_bytes) { 3416 if (unused >= num_bytes) {
3417 if (!reserved) 3417 if (!reserved)
3418 space_info->bytes_reserved += orig_bytes; 3418 space_info->bytes_reserved += orig_bytes;
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index eac10e3260a9..3e86b9f36507 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1828,9 +1828,9 @@ static void end_bio_extent_preparewrite(struct bio *bio, int err)
1828 bio_put(bio); 1828 bio_put(bio);
1829} 1829}
1830 1830
1831static struct bio * 1831struct bio *
1832extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs, 1832btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
1833 gfp_t gfp_flags) 1833 gfp_t gfp_flags)
1834{ 1834{
1835 struct bio *bio; 1835 struct bio *bio;
1836 1836
@@ -1919,7 +1919,7 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree,
1919 else 1919 else
1920 nr = bio_get_nr_vecs(bdev); 1920 nr = bio_get_nr_vecs(bdev);
1921 1921
1922 bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH); 1922 bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
1923 1923
1924 bio_add_page(bio, page, page_size, offset); 1924 bio_add_page(bio, page, page_size, offset);
1925 bio->bi_end_io = end_io_func; 1925 bio->bi_end_io = end_io_func;
@@ -2901,21 +2901,53 @@ out:
2901int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 2901int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2902 __u64 start, __u64 len, get_extent_t *get_extent) 2902 __u64 start, __u64 len, get_extent_t *get_extent)
2903{ 2903{
2904 int ret; 2904 int ret = 0;
2905 u64 off = start; 2905 u64 off = start;
2906 u64 max = start + len; 2906 u64 max = start + len;
2907 u32 flags = 0; 2907 u32 flags = 0;
2908 u32 found_type;
2909 u64 last;
2908 u64 disko = 0; 2910 u64 disko = 0;
2911 struct btrfs_key found_key;
2909 struct extent_map *em = NULL; 2912 struct extent_map *em = NULL;
2910 struct extent_state *cached_state = NULL; 2913 struct extent_state *cached_state = NULL;
2914 struct btrfs_path *path;
2915 struct btrfs_file_extent_item *item;
2911 int end = 0; 2916 int end = 0;
2912 u64 em_start = 0, em_len = 0; 2917 u64 em_start = 0, em_len = 0;
2913 unsigned long emflags; 2918 unsigned long emflags;
2914 ret = 0; 2919 int hole = 0;
2915 2920
2916 if (len == 0) 2921 if (len == 0)
2917 return -EINVAL; 2922 return -EINVAL;
2918 2923
2924 path = btrfs_alloc_path();
2925 if (!path)
2926 return -ENOMEM;
2927 path->leave_spinning = 1;
2928
2929 ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
2930 path, inode->i_ino, -1, 0);
2931 if (ret < 0) {
2932 btrfs_free_path(path);
2933 return ret;
2934 }
2935 WARN_ON(!ret);
2936 path->slots[0]--;
2937 item = btrfs_item_ptr(path->nodes[0], path->slots[0],
2938 struct btrfs_file_extent_item);
2939 btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
2940 found_type = btrfs_key_type(&found_key);
2941
2942 /* No extents, just return */
2943 if (found_key.objectid != inode->i_ino ||
2944 found_type != BTRFS_EXTENT_DATA_KEY) {
2945 btrfs_free_path(path);
2946 return 0;
2947 }
2948 last = found_key.offset;
2949 btrfs_free_path(path);
2950
2919 lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0, 2951 lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0,
2920 &cached_state, GFP_NOFS); 2952 &cached_state, GFP_NOFS);
2921 em = get_extent(inode, NULL, 0, off, max - off, 0); 2953 em = get_extent(inode, NULL, 0, off, max - off, 0);
@@ -2925,11 +2957,18 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2925 ret = PTR_ERR(em); 2957 ret = PTR_ERR(em);
2926 goto out; 2958 goto out;
2927 } 2959 }
2960
2928 while (!end) { 2961 while (!end) {
2962 hole = 0;
2929 off = em->start + em->len; 2963 off = em->start + em->len;
2930 if (off >= max) 2964 if (off >= max)
2931 end = 1; 2965 end = 1;
2932 2966
2967 if (em->block_start == EXTENT_MAP_HOLE) {
2968 hole = 1;
2969 goto next;
2970 }
2971
2933 em_start = em->start; 2972 em_start = em->start;
2934 em_len = em->len; 2973 em_len = em->len;
2935 2974
@@ -2939,8 +2978,6 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2939 if (em->block_start == EXTENT_MAP_LAST_BYTE) { 2978 if (em->block_start == EXTENT_MAP_LAST_BYTE) {
2940 end = 1; 2979 end = 1;
2941 flags |= FIEMAP_EXTENT_LAST; 2980 flags |= FIEMAP_EXTENT_LAST;
2942 } else if (em->block_start == EXTENT_MAP_HOLE) {
2943 flags |= FIEMAP_EXTENT_UNWRITTEN;
2944 } else if (em->block_start == EXTENT_MAP_INLINE) { 2981 } else if (em->block_start == EXTENT_MAP_INLINE) {
2945 flags |= (FIEMAP_EXTENT_DATA_INLINE | 2982 flags |= (FIEMAP_EXTENT_DATA_INLINE |
2946 FIEMAP_EXTENT_NOT_ALIGNED); 2983 FIEMAP_EXTENT_NOT_ALIGNED);
@@ -2953,10 +2990,10 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2953 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) 2990 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
2954 flags |= FIEMAP_EXTENT_ENCODED; 2991 flags |= FIEMAP_EXTENT_ENCODED;
2955 2992
2993next:
2956 emflags = em->flags; 2994 emflags = em->flags;
2957 free_extent_map(em); 2995 free_extent_map(em);
2958 em = NULL; 2996 em = NULL;
2959
2960 if (!end) { 2997 if (!end) {
2961 em = get_extent(inode, NULL, 0, off, max - off, 0); 2998 em = get_extent(inode, NULL, 0, off, max - off, 0);
2962 if (!em) 2999 if (!em)
@@ -2967,15 +3004,23 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
2967 } 3004 }
2968 emflags = em->flags; 3005 emflags = em->flags;
2969 } 3006 }
3007
2970 if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) { 3008 if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) {
2971 flags |= FIEMAP_EXTENT_LAST; 3009 flags |= FIEMAP_EXTENT_LAST;
2972 end = 1; 3010 end = 1;
2973 } 3011 }
2974 3012
2975 ret = fiemap_fill_next_extent(fieinfo, em_start, disko, 3013 if (em_start == last) {
2976 em_len, flags); 3014 flags |= FIEMAP_EXTENT_LAST;
2977 if (ret) 3015 end = 1;
2978 goto out_free; 3016 }
3017
3018 if (!hole) {
3019 ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
3020 em_len, flags);
3021 if (ret)
3022 goto out_free;
3023 }
2979 } 3024 }
2980out_free: 3025out_free:
2981 free_extent_map(em); 3026 free_extent_map(em);
@@ -3836,8 +3881,10 @@ int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page)
3836 3881
3837 spin_lock(&tree->buffer_lock); 3882 spin_lock(&tree->buffer_lock);
3838 eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT); 3883 eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
3839 if (!eb) 3884 if (!eb) {
3840 goto out; 3885 spin_unlock(&tree->buffer_lock);
3886 return ret;
3887 }
3841 3888
3842 if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) { 3889 if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
3843 ret = 0; 3890 ret = 0;
diff --git a/fs/btrfs/extent_io.h b/fs/btrfs/extent_io.h
index 1c6d4f342ef7..4183c8178f01 100644
--- a/fs/btrfs/extent_io.h
+++ b/fs/btrfs/extent_io.h
@@ -310,4 +310,7 @@ int extent_clear_unlock_delalloc(struct inode *inode,
310 struct extent_io_tree *tree, 310 struct extent_io_tree *tree,
311 u64 start, u64 end, struct page *locked_page, 311 u64 start, u64 end, struct page *locked_page,
312 unsigned long op); 312 unsigned long op);
313struct bio *
314btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
315 gfp_t gfp_flags);
313#endif 316#endif
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index e354c33df082..c1faded5fca0 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1047,8 +1047,14 @@ out:
1047 1047
1048 if ((file->f_flags & O_DSYNC) || IS_SYNC(inode)) { 1048 if ((file->f_flags & O_DSYNC) || IS_SYNC(inode)) {
1049 trans = btrfs_start_transaction(root, 0); 1049 trans = btrfs_start_transaction(root, 0);
1050 if (IS_ERR(trans)) {
1051 num_written = PTR_ERR(trans);
1052 goto done;
1053 }
1054 mutex_lock(&inode->i_mutex);
1050 ret = btrfs_log_dentry_safe(trans, root, 1055 ret = btrfs_log_dentry_safe(trans, root,
1051 file->f_dentry); 1056 file->f_dentry);
1057 mutex_unlock(&inode->i_mutex);
1052 if (ret == 0) { 1058 if (ret == 0) {
1053 ret = btrfs_sync_log(trans, root); 1059 ret = btrfs_sync_log(trans, root);
1054 if (ret == 0) 1060 if (ret == 0)
@@ -1067,6 +1073,7 @@ out:
1067 (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT); 1073 (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT);
1068 } 1074 }
1069 } 1075 }
1076done:
1070 current->backing_dev_info = NULL; 1077 current->backing_dev_info = NULL;
1071 return num_written ? num_written : err; 1078 return num_written ? num_written : err;
1072} 1079}
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 558cac2dfa54..8039390bd6a6 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -4501,6 +4501,7 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
4501 BTRFS_I(inode)->index_cnt = 2; 4501 BTRFS_I(inode)->index_cnt = 2;
4502 BTRFS_I(inode)->root = root; 4502 BTRFS_I(inode)->root = root;
4503 BTRFS_I(inode)->generation = trans->transid; 4503 BTRFS_I(inode)->generation = trans->transid;
4504 inode->i_generation = BTRFS_I(inode)->generation;
4504 btrfs_set_inode_space_info(root, inode); 4505 btrfs_set_inode_space_info(root, inode);
4505 4506
4506 if (mode & S_IFDIR) 4507 if (mode & S_IFDIR)
@@ -4622,12 +4623,12 @@ int btrfs_add_link(struct btrfs_trans_handle *trans,
4622} 4623}
4623 4624
4624static int btrfs_add_nondir(struct btrfs_trans_handle *trans, 4625static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
4625 struct dentry *dentry, struct inode *inode, 4626 struct inode *dir, struct dentry *dentry,
4626 int backref, u64 index) 4627 struct inode *inode, int backref, u64 index)
4627{ 4628{
4628 int err = btrfs_add_link(trans, dentry->d_parent->d_inode, 4629 int err = btrfs_add_link(trans, dir, inode,
4629 inode, dentry->d_name.name, 4630 dentry->d_name.name, dentry->d_name.len,
4630 dentry->d_name.len, backref, index); 4631 backref, index);
4631 if (!err) { 4632 if (!err) {
4632 d_instantiate(dentry, inode); 4633 d_instantiate(dentry, inode);
4633 return 0; 4634 return 0;
@@ -4668,8 +4669,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
4668 btrfs_set_trans_block_group(trans, dir); 4669 btrfs_set_trans_block_group(trans, dir);
4669 4670
4670 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, 4671 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4671 dentry->d_name.len, 4672 dentry->d_name.len, dir->i_ino, objectid,
4672 dentry->d_parent->d_inode->i_ino, objectid,
4673 BTRFS_I(dir)->block_group, mode, &index); 4673 BTRFS_I(dir)->block_group, mode, &index);
4674 err = PTR_ERR(inode); 4674 err = PTR_ERR(inode);
4675 if (IS_ERR(inode)) 4675 if (IS_ERR(inode))
@@ -4682,7 +4682,7 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
4682 } 4682 }
4683 4683
4684 btrfs_set_trans_block_group(trans, inode); 4684 btrfs_set_trans_block_group(trans, inode);
4685 err = btrfs_add_nondir(trans, dentry, inode, 0, index); 4685 err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
4686 if (err) 4686 if (err)
4687 drop_inode = 1; 4687 drop_inode = 1;
4688 else { 4688 else {
@@ -4730,10 +4730,8 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
4730 btrfs_set_trans_block_group(trans, dir); 4730 btrfs_set_trans_block_group(trans, dir);
4731 4731
4732 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, 4732 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4733 dentry->d_name.len, 4733 dentry->d_name.len, dir->i_ino, objectid,
4734 dentry->d_parent->d_inode->i_ino, 4734 BTRFS_I(dir)->block_group, mode, &index);
4735 objectid, BTRFS_I(dir)->block_group, mode,
4736 &index);
4737 err = PTR_ERR(inode); 4735 err = PTR_ERR(inode);
4738 if (IS_ERR(inode)) 4736 if (IS_ERR(inode))
4739 goto out_unlock; 4737 goto out_unlock;
@@ -4745,7 +4743,7 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry,
4745 } 4743 }
4746 4744
4747 btrfs_set_trans_block_group(trans, inode); 4745 btrfs_set_trans_block_group(trans, inode);
4748 err = btrfs_add_nondir(trans, dentry, inode, 0, index); 4746 err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
4749 if (err) 4747 if (err)
4750 drop_inode = 1; 4748 drop_inode = 1;
4751 else { 4749 else {
@@ -4787,6 +4785,7 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
4787 return -EPERM; 4785 return -EPERM;
4788 4786
4789 btrfs_inc_nlink(inode); 4787 btrfs_inc_nlink(inode);
4788 inode->i_ctime = CURRENT_TIME;
4790 4789
4791 err = btrfs_set_inode_index(dir, &index); 4790 err = btrfs_set_inode_index(dir, &index);
4792 if (err) 4791 if (err)
@@ -4805,15 +4804,17 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
4805 btrfs_set_trans_block_group(trans, dir); 4804 btrfs_set_trans_block_group(trans, dir);
4806 ihold(inode); 4805 ihold(inode);
4807 4806
4808 err = btrfs_add_nondir(trans, dentry, inode, 1, index); 4807 err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
4809 4808
4810 if (err) { 4809 if (err) {
4811 drop_inode = 1; 4810 drop_inode = 1;
4812 } else { 4811 } else {
4812 struct dentry *parent = dget_parent(dentry);
4813 btrfs_update_inode_block_group(trans, dir); 4813 btrfs_update_inode_block_group(trans, dir);
4814 err = btrfs_update_inode(trans, root, inode); 4814 err = btrfs_update_inode(trans, root, inode);
4815 BUG_ON(err); 4815 BUG_ON(err);
4816 btrfs_log_new_name(trans, inode, NULL, dentry->d_parent); 4816 btrfs_log_new_name(trans, inode, NULL, parent);
4817 dput(parent);
4817 } 4818 }
4818 4819
4819 nr = trans->blocks_used; 4820 nr = trans->blocks_used;
@@ -4853,8 +4854,7 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
4853 btrfs_set_trans_block_group(trans, dir); 4854 btrfs_set_trans_block_group(trans, dir);
4854 4855
4855 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, 4856 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
4856 dentry->d_name.len, 4857 dentry->d_name.len, dir->i_ino, objectid,
4857 dentry->d_parent->d_inode->i_ino, objectid,
4858 BTRFS_I(dir)->block_group, S_IFDIR | mode, 4858 BTRFS_I(dir)->block_group, S_IFDIR | mode,
4859 &index); 4859 &index);
4860 if (IS_ERR(inode)) { 4860 if (IS_ERR(inode)) {
@@ -4877,9 +4877,8 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode)
4877 if (err) 4877 if (err)
4878 goto out_fail; 4878 goto out_fail;
4879 4879
4880 err = btrfs_add_link(trans, dentry->d_parent->d_inode, 4880 err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
4881 inode, dentry->d_name.name, 4881 dentry->d_name.len, 0, index);
4882 dentry->d_name.len, 0, index);
4883 if (err) 4882 if (err)
4884 goto out_fail; 4883 goto out_fail;
4885 4884
@@ -5535,13 +5534,21 @@ struct btrfs_dio_private {
5535 u64 bytes; 5534 u64 bytes;
5536 u32 *csums; 5535 u32 *csums;
5537 void *private; 5536 void *private;
5537
5538 /* number of bios pending for this dio */
5539 atomic_t pending_bios;
5540
5541 /* IO errors */
5542 int errors;
5543
5544 struct bio *orig_bio;
5538}; 5545};
5539 5546
5540static void btrfs_endio_direct_read(struct bio *bio, int err) 5547static void btrfs_endio_direct_read(struct bio *bio, int err)
5541{ 5548{
5549 struct btrfs_dio_private *dip = bio->bi_private;
5542 struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1; 5550 struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
5543 struct bio_vec *bvec = bio->bi_io_vec; 5551 struct bio_vec *bvec = bio->bi_io_vec;
5544 struct btrfs_dio_private *dip = bio->bi_private;
5545 struct inode *inode = dip->inode; 5552 struct inode *inode = dip->inode;
5546 struct btrfs_root *root = BTRFS_I(inode)->root; 5553 struct btrfs_root *root = BTRFS_I(inode)->root;
5547 u64 start; 5554 u64 start;
@@ -5595,15 +5602,18 @@ static void btrfs_endio_direct_write(struct bio *bio, int err)
5595 struct btrfs_trans_handle *trans; 5602 struct btrfs_trans_handle *trans;
5596 struct btrfs_ordered_extent *ordered = NULL; 5603 struct btrfs_ordered_extent *ordered = NULL;
5597 struct extent_state *cached_state = NULL; 5604 struct extent_state *cached_state = NULL;
5605 u64 ordered_offset = dip->logical_offset;
5606 u64 ordered_bytes = dip->bytes;
5598 int ret; 5607 int ret;
5599 5608
5600 if (err) 5609 if (err)
5601 goto out_done; 5610 goto out_done;
5602 5611again:
5603 ret = btrfs_dec_test_ordered_pending(inode, &ordered, 5612 ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
5604 dip->logical_offset, dip->bytes); 5613 &ordered_offset,
5614 ordered_bytes);
5605 if (!ret) 5615 if (!ret)
5606 goto out_done; 5616 goto out_test;
5607 5617
5608 BUG_ON(!ordered); 5618 BUG_ON(!ordered);
5609 5619
@@ -5663,8 +5673,20 @@ out_unlock:
5663out: 5673out:
5664 btrfs_delalloc_release_metadata(inode, ordered->len); 5674 btrfs_delalloc_release_metadata(inode, ordered->len);
5665 btrfs_end_transaction(trans, root); 5675 btrfs_end_transaction(trans, root);
5676 ordered_offset = ordered->file_offset + ordered->len;
5666 btrfs_put_ordered_extent(ordered); 5677 btrfs_put_ordered_extent(ordered);
5667 btrfs_put_ordered_extent(ordered); 5678 btrfs_put_ordered_extent(ordered);
5679
5680out_test:
5681 /*
5682 * our bio might span multiple ordered extents. If we haven't
5683 * completed the accounting for the whole dio, go back and try again
5684 */
5685 if (ordered_offset < dip->logical_offset + dip->bytes) {
5686 ordered_bytes = dip->logical_offset + dip->bytes -
5687 ordered_offset;
5688 goto again;
5689 }
5668out_done: 5690out_done:
5669 bio->bi_private = dip->private; 5691 bio->bi_private = dip->private;
5670 5692
@@ -5684,6 +5706,176 @@ static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw,
5684 return 0; 5706 return 0;
5685} 5707}
5686 5708
5709static void btrfs_end_dio_bio(struct bio *bio, int err)
5710{
5711 struct btrfs_dio_private *dip = bio->bi_private;
5712
5713 if (err) {
5714 printk(KERN_ERR "btrfs direct IO failed ino %lu rw %lu "
5715 "disk_bytenr %lu len %u err no %d\n",
5716 dip->inode->i_ino, bio->bi_rw, bio->bi_sector,
5717 bio->bi_size, err);
5718 dip->errors = 1;
5719
5720 /*
5721 * before atomic variable goto zero, we must make sure
5722 * dip->errors is perceived to be set.
5723 */
5724 smp_mb__before_atomic_dec();
5725 }
5726
5727 /* if there are more bios still pending for this dio, just exit */
5728 if (!atomic_dec_and_test(&dip->pending_bios))
5729 goto out;
5730
5731 if (dip->errors)
5732 bio_io_error(dip->orig_bio);
5733 else {
5734 set_bit(BIO_UPTODATE, &dip->orig_bio->bi_flags);
5735 bio_endio(dip->orig_bio, 0);
5736 }
5737out:
5738 bio_put(bio);
5739}
5740
5741static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev,
5742 u64 first_sector, gfp_t gfp_flags)
5743{
5744 int nr_vecs = bio_get_nr_vecs(bdev);
5745 return btrfs_bio_alloc(bdev, first_sector, nr_vecs, gfp_flags);
5746}
5747
5748static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
5749 int rw, u64 file_offset, int skip_sum,
5750 u32 *csums)
5751{
5752 int write = rw & REQ_WRITE;
5753 struct btrfs_root *root = BTRFS_I(inode)->root;
5754 int ret;
5755
5756 bio_get(bio);
5757 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
5758 if (ret)
5759 goto err;
5760
5761 if (write && !skip_sum) {
5762 ret = btrfs_wq_submit_bio(root->fs_info,
5763 inode, rw, bio, 0, 0,
5764 file_offset,
5765 __btrfs_submit_bio_start_direct_io,
5766 __btrfs_submit_bio_done);
5767 goto err;
5768 } else if (!skip_sum)
5769 btrfs_lookup_bio_sums_dio(root, inode, bio,
5770 file_offset, csums);
5771
5772 ret = btrfs_map_bio(root, rw, bio, 0, 1);
5773err:
5774 bio_put(bio);
5775 return ret;
5776}
5777
5778static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
5779 int skip_sum)
5780{
5781 struct inode *inode = dip->inode;
5782 struct btrfs_root *root = BTRFS_I(inode)->root;
5783 struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
5784 struct bio *bio;
5785 struct bio *orig_bio = dip->orig_bio;
5786 struct bio_vec *bvec = orig_bio->bi_io_vec;
5787 u64 start_sector = orig_bio->bi_sector;
5788 u64 file_offset = dip->logical_offset;
5789 u64 submit_len = 0;
5790 u64 map_length;
5791 int nr_pages = 0;
5792 u32 *csums = dip->csums;
5793 int ret = 0;
5794
5795 bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
5796 if (!bio)
5797 return -ENOMEM;
5798 bio->bi_private = dip;
5799 bio->bi_end_io = btrfs_end_dio_bio;
5800 atomic_inc(&dip->pending_bios);
5801
5802 map_length = orig_bio->bi_size;
5803 ret = btrfs_map_block(map_tree, READ, start_sector << 9,
5804 &map_length, NULL, 0);
5805 if (ret) {
5806 bio_put(bio);
5807 return -EIO;
5808 }
5809
5810 while (bvec <= (orig_bio->bi_io_vec + orig_bio->bi_vcnt - 1)) {
5811 if (unlikely(map_length < submit_len + bvec->bv_len ||
5812 bio_add_page(bio, bvec->bv_page, bvec->bv_len,
5813 bvec->bv_offset) < bvec->bv_len)) {
5814 /*
5815 * inc the count before we submit the bio so
5816 * we know the end IO handler won't happen before
5817 * we inc the count. Otherwise, the dip might get freed
5818 * before we're done setting it up
5819 */
5820 atomic_inc(&dip->pending_bios);
5821 ret = __btrfs_submit_dio_bio(bio, inode, rw,
5822 file_offset, skip_sum,
5823 csums);
5824 if (ret) {
5825 bio_put(bio);
5826 atomic_dec(&dip->pending_bios);
5827 goto out_err;
5828 }
5829
5830 if (!skip_sum)
5831 csums = csums + nr_pages;
5832 start_sector += submit_len >> 9;
5833 file_offset += submit_len;
5834
5835 submit_len = 0;
5836 nr_pages = 0;
5837
5838 bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev,
5839 start_sector, GFP_NOFS);
5840 if (!bio)
5841 goto out_err;
5842 bio->bi_private = dip;
5843 bio->bi_end_io = btrfs_end_dio_bio;
5844
5845 map_length = orig_bio->bi_size;
5846 ret = btrfs_map_block(map_tree, READ, start_sector << 9,
5847 &map_length, NULL, 0);
5848 if (ret) {
5849 bio_put(bio);
5850 goto out_err;
5851 }
5852 } else {
5853 submit_len += bvec->bv_len;
5854 nr_pages ++;
5855 bvec++;
5856 }
5857 }
5858
5859 ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
5860 csums);
5861 if (!ret)
5862 return 0;
5863
5864 bio_put(bio);
5865out_err:
5866 dip->errors = 1;
5867 /*
5868 * before atomic variable goto zero, we must
5869 * make sure dip->errors is perceived to be set.
5870 */
5871 smp_mb__before_atomic_dec();
5872 if (atomic_dec_and_test(&dip->pending_bios))
5873 bio_io_error(dip->orig_bio);
5874
5875 /* bio_end_io() will handle error, so we needn't return it */
5876 return 0;
5877}
5878
5687static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode, 5879static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode,
5688 loff_t file_offset) 5880 loff_t file_offset)
5689{ 5881{
@@ -5723,36 +5915,18 @@ static void btrfs_submit_direct(int rw, struct bio *bio, struct inode *inode,
5723 5915
5724 dip->disk_bytenr = (u64)bio->bi_sector << 9; 5916 dip->disk_bytenr = (u64)bio->bi_sector << 9;
5725 bio->bi_private = dip; 5917 bio->bi_private = dip;
5918 dip->errors = 0;
5919 dip->orig_bio = bio;
5920 atomic_set(&dip->pending_bios, 0);
5726 5921
5727 if (write) 5922 if (write)
5728 bio->bi_end_io = btrfs_endio_direct_write; 5923 bio->bi_end_io = btrfs_endio_direct_write;
5729 else 5924 else
5730 bio->bi_end_io = btrfs_endio_direct_read; 5925 bio->bi_end_io = btrfs_endio_direct_read;
5731 5926
5732 ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0); 5927 ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
5733 if (ret) 5928 if (!ret)
5734 goto out_err;
5735
5736 if (write && !skip_sum) {
5737 ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
5738 inode, rw, bio, 0, 0,
5739 dip->logical_offset,
5740 __btrfs_submit_bio_start_direct_io,
5741 __btrfs_submit_bio_done);
5742 if (ret)
5743 goto out_err;
5744 return; 5929 return;
5745 } else if (!skip_sum)
5746 btrfs_lookup_bio_sums_dio(root, inode, bio,
5747 dip->logical_offset, dip->csums);
5748
5749 ret = btrfs_map_bio(root, rw, bio, 0, 1);
5750 if (ret)
5751 goto out_err;
5752 return;
5753out_err:
5754 kfree(dip->csums);
5755 kfree(dip);
5756free_ordered: 5930free_ordered:
5757 /* 5931 /*
5758 * If this is a write, we need to clean up the reserved space and kill 5932 * If this is a write, we need to clean up the reserved space and kill
@@ -6607,8 +6781,9 @@ static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
6607 BUG_ON(ret); 6781 BUG_ON(ret);
6608 6782
6609 if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) { 6783 if (old_inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
6610 btrfs_log_new_name(trans, old_inode, old_dir, 6784 struct dentry *parent = dget_parent(new_dentry);
6611 new_dentry->d_parent); 6785 btrfs_log_new_name(trans, old_inode, old_dir, parent);
6786 dput(parent);
6612 btrfs_end_log_trans(root); 6787 btrfs_end_log_trans(root);
6613 } 6788 }
6614out_fail: 6789out_fail:
@@ -6758,8 +6933,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
6758 btrfs_set_trans_block_group(trans, dir); 6933 btrfs_set_trans_block_group(trans, dir);
6759 6934
6760 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name, 6935 inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
6761 dentry->d_name.len, 6936 dentry->d_name.len, dir->i_ino, objectid,
6762 dentry->d_parent->d_inode->i_ino, objectid,
6763 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO, 6937 BTRFS_I(dir)->block_group, S_IFLNK|S_IRWXUGO,
6764 &index); 6938 &index);
6765 err = PTR_ERR(inode); 6939 err = PTR_ERR(inode);
@@ -6773,7 +6947,7 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
6773 } 6947 }
6774 6948
6775 btrfs_set_trans_block_group(trans, inode); 6949 btrfs_set_trans_block_group(trans, inode);
6776 err = btrfs_add_nondir(trans, dentry, inode, 0, index); 6950 err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
6777 if (err) 6951 if (err)
6778 drop_inode = 1; 6952 drop_inode = 1;
6779 else { 6953 else {
@@ -6844,6 +7018,7 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
6844 struct btrfs_root *root = BTRFS_I(inode)->root; 7018 struct btrfs_root *root = BTRFS_I(inode)->root;
6845 struct btrfs_key ins; 7019 struct btrfs_key ins;
6846 u64 cur_offset = start; 7020 u64 cur_offset = start;
7021 u64 i_size;
6847 int ret = 0; 7022 int ret = 0;
6848 bool own_trans = true; 7023 bool own_trans = true;
6849 7024
@@ -6885,11 +7060,11 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
6885 (actual_len > inode->i_size) && 7060 (actual_len > inode->i_size) &&
6886 (cur_offset > inode->i_size)) { 7061 (cur_offset > inode->i_size)) {
6887 if (cur_offset > actual_len) 7062 if (cur_offset > actual_len)
6888 i_size_write(inode, actual_len); 7063 i_size = actual_len;
6889 else 7064 else
6890 i_size_write(inode, cur_offset); 7065 i_size = cur_offset;
6891 i_size_write(inode, cur_offset); 7066 i_size_write(inode, i_size);
6892 btrfs_ordered_update_i_size(inode, cur_offset, NULL); 7067 btrfs_ordered_update_i_size(inode, i_size, NULL);
6893 } 7068 }
6894 7069
6895 ret = btrfs_update_inode(trans, root, inode); 7070 ret = btrfs_update_inode(trans, root, inode);
@@ -6943,6 +7118,10 @@ static long btrfs_fallocate(struct inode *inode, int mode,
6943 btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start); 7118 btrfs_wait_ordered_range(inode, alloc_start, alloc_end - alloc_start);
6944 7119
6945 mutex_lock(&inode->i_mutex); 7120 mutex_lock(&inode->i_mutex);
7121 ret = inode_newsize_ok(inode, alloc_end);
7122 if (ret)
7123 goto out;
7124
6946 if (alloc_start > inode->i_size) { 7125 if (alloc_start > inode->i_size) {
6947 ret = btrfs_cont_expand(inode, alloc_start); 7126 ret = btrfs_cont_expand(inode, alloc_start);
6948 if (ret) 7127 if (ret)
@@ -7139,6 +7318,7 @@ static const struct inode_operations btrfs_symlink_inode_operations = {
7139 .readlink = generic_readlink, 7318 .readlink = generic_readlink,
7140 .follow_link = page_follow_link_light, 7319 .follow_link = page_follow_link_light,
7141 .put_link = page_put_link, 7320 .put_link = page_put_link,
7321 .getattr = btrfs_getattr,
7142 .permission = btrfs_permission, 7322 .permission = btrfs_permission,
7143 .setxattr = btrfs_setxattr, 7323 .setxattr = btrfs_setxattr,
7144 .getxattr = btrfs_getxattr, 7324 .getxattr = btrfs_getxattr,
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 463d91b4dd3a..f1c9bb4079ed 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -233,7 +233,8 @@ static noinline int create_subvol(struct btrfs_root *root,
233 struct btrfs_inode_item *inode_item; 233 struct btrfs_inode_item *inode_item;
234 struct extent_buffer *leaf; 234 struct extent_buffer *leaf;
235 struct btrfs_root *new_root; 235 struct btrfs_root *new_root;
236 struct inode *dir = dentry->d_parent->d_inode; 236 struct dentry *parent = dget_parent(dentry);
237 struct inode *dir;
237 int ret; 238 int ret;
238 int err; 239 int err;
239 u64 objectid; 240 u64 objectid;
@@ -242,8 +243,13 @@ static noinline int create_subvol(struct btrfs_root *root,
242 243
243 ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root, 244 ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
244 0, &objectid); 245 0, &objectid);
245 if (ret) 246 if (ret) {
247 dput(parent);
246 return ret; 248 return ret;
249 }
250
251 dir = parent->d_inode;
252
247 /* 253 /*
248 * 1 - inode item 254 * 1 - inode item
249 * 2 - refs 255 * 2 - refs
@@ -251,8 +257,10 @@ static noinline int create_subvol(struct btrfs_root *root,
251 * 2 - dir items 257 * 2 - dir items
252 */ 258 */
253 trans = btrfs_start_transaction(root, 6); 259 trans = btrfs_start_transaction(root, 6);
254 if (IS_ERR(trans)) 260 if (IS_ERR(trans)) {
261 dput(parent);
255 return PTR_ERR(trans); 262 return PTR_ERR(trans);
263 }
256 264
257 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 265 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
258 0, objectid, NULL, 0, 0, 0); 266 0, objectid, NULL, 0, 0, 0);
@@ -339,6 +347,7 @@ static noinline int create_subvol(struct btrfs_root *root,
339 347
340 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry)); 348 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
341fail: 349fail:
350 dput(parent);
342 if (async_transid) { 351 if (async_transid) {
343 *async_transid = trans->transid; 352 *async_transid = trans->transid;
344 err = btrfs_commit_transaction_async(trans, root, 1); 353 err = btrfs_commit_transaction_async(trans, root, 1);
@@ -354,6 +363,7 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
354 char *name, int namelen, u64 *async_transid) 363 char *name, int namelen, u64 *async_transid)
355{ 364{
356 struct inode *inode; 365 struct inode *inode;
366 struct dentry *parent;
357 struct btrfs_pending_snapshot *pending_snapshot; 367 struct btrfs_pending_snapshot *pending_snapshot;
358 struct btrfs_trans_handle *trans; 368 struct btrfs_trans_handle *trans;
359 int ret; 369 int ret;
@@ -396,7 +406,9 @@ static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
396 406
397 btrfs_orphan_cleanup(pending_snapshot->snap); 407 btrfs_orphan_cleanup(pending_snapshot->snap);
398 408
399 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry); 409 parent = dget_parent(dentry);
410 inode = btrfs_lookup_dentry(parent->d_inode, dentry);
411 dput(parent);
400 if (IS_ERR(inode)) { 412 if (IS_ERR(inode)) {
401 ret = PTR_ERR(inode); 413 ret = PTR_ERR(inode);
402 goto fail; 414 goto fail;
@@ -1669,12 +1681,11 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1669 olen = len = src->i_size - off; 1681 olen = len = src->i_size - off;
1670 /* if we extend to eof, continue to block boundary */ 1682 /* if we extend to eof, continue to block boundary */
1671 if (off + len == src->i_size) 1683 if (off + len == src->i_size)
1672 len = ((src->i_size + bs-1) & ~(bs-1)) 1684 len = ALIGN(src->i_size, bs) - off;
1673 - off;
1674 1685
1675 /* verify the end result is block aligned */ 1686 /* verify the end result is block aligned */
1676 if ((off & (bs-1)) || 1687 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
1677 ((off + len) & (bs-1))) 1688 !IS_ALIGNED(destoff, bs))
1678 goto out_unlock; 1689 goto out_unlock;
1679 1690
1680 /* do any pending delalloc/csum calc on src, one way or 1691 /* do any pending delalloc/csum calc on src, one way or
@@ -1874,8 +1885,8 @@ static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1874 * but shouldn't round up the file size 1885 * but shouldn't round up the file size
1875 */ 1886 */
1876 endoff = new_key.offset + datal; 1887 endoff = new_key.offset + datal;
1877 if (endoff > off+olen) 1888 if (endoff > destoff+olen)
1878 endoff = off+olen; 1889 endoff = destoff+olen;
1879 if (endoff > inode->i_size) 1890 if (endoff > inode->i_size)
1880 btrfs_i_size_write(inode, endoff); 1891 btrfs_i_size_write(inode, endoff);
1881 1892
diff --git a/fs/btrfs/ordered-data.c b/fs/btrfs/ordered-data.c
index f4621f6deca1..ae7737e352c9 100644
--- a/fs/btrfs/ordered-data.c
+++ b/fs/btrfs/ordered-data.c
@@ -250,6 +250,73 @@ int btrfs_add_ordered_sum(struct inode *inode,
250 250
251/* 251/*
252 * this is used to account for finished IO across a given range 252 * this is used to account for finished IO across a given range
253 * of the file. The IO may span ordered extents. If
254 * a given ordered_extent is completely done, 1 is returned, otherwise
255 * 0.
256 *
257 * test_and_set_bit on a flag in the struct btrfs_ordered_extent is used
258 * to make sure this function only returns 1 once for a given ordered extent.
259 *
260 * file_offset is updated to one byte past the range that is recorded as
261 * complete. This allows you to walk forward in the file.
262 */
263int btrfs_dec_test_first_ordered_pending(struct inode *inode,
264 struct btrfs_ordered_extent **cached,
265 u64 *file_offset, u64 io_size)
266{
267 struct btrfs_ordered_inode_tree *tree;
268 struct rb_node *node;
269 struct btrfs_ordered_extent *entry = NULL;
270 int ret;
271 u64 dec_end;
272 u64 dec_start;
273 u64 to_dec;
274
275 tree = &BTRFS_I(inode)->ordered_tree;
276 spin_lock(&tree->lock);
277 node = tree_search(tree, *file_offset);
278 if (!node) {
279 ret = 1;
280 goto out;
281 }
282
283 entry = rb_entry(node, struct btrfs_ordered_extent, rb_node);
284 if (!offset_in_entry(entry, *file_offset)) {
285 ret = 1;
286 goto out;
287 }
288
289 dec_start = max(*file_offset, entry->file_offset);
290 dec_end = min(*file_offset + io_size, entry->file_offset +
291 entry->len);
292 *file_offset = dec_end;
293 if (dec_start > dec_end) {
294 printk(KERN_CRIT "bad ordering dec_start %llu end %llu\n",
295 (unsigned long long)dec_start,
296 (unsigned long long)dec_end);
297 }
298 to_dec = dec_end - dec_start;
299 if (to_dec > entry->bytes_left) {
300 printk(KERN_CRIT "bad ordered accounting left %llu size %llu\n",
301 (unsigned long long)entry->bytes_left,
302 (unsigned long long)to_dec);
303 }
304 entry->bytes_left -= to_dec;
305 if (entry->bytes_left == 0)
306 ret = test_and_set_bit(BTRFS_ORDERED_IO_DONE, &entry->flags);
307 else
308 ret = 1;
309out:
310 if (!ret && cached && entry) {
311 *cached = entry;
312 atomic_inc(&entry->refs);
313 }
314 spin_unlock(&tree->lock);
315 return ret == 0;
316}
317
318/*
319 * this is used to account for finished IO across a given range
253 * of the file. The IO should not span ordered extents. If 320 * of the file. The IO should not span ordered extents. If
254 * a given ordered_extent is completely done, 1 is returned, otherwise 321 * a given ordered_extent is completely done, 1 is returned, otherwise
255 * 0. 322 * 0.
diff --git a/fs/btrfs/ordered-data.h b/fs/btrfs/ordered-data.h
index 8ac365492a3f..61dca83119dd 100644
--- a/fs/btrfs/ordered-data.h
+++ b/fs/btrfs/ordered-data.h
@@ -141,6 +141,9 @@ int btrfs_remove_ordered_extent(struct inode *inode,
141int btrfs_dec_test_ordered_pending(struct inode *inode, 141int btrfs_dec_test_ordered_pending(struct inode *inode,
142 struct btrfs_ordered_extent **cached, 142 struct btrfs_ordered_extent **cached,
143 u64 file_offset, u64 io_size); 143 u64 file_offset, u64 io_size);
144int btrfs_dec_test_first_ordered_pending(struct inode *inode,
145 struct btrfs_ordered_extent **cached,
146 u64 *file_offset, u64 io_size);
144int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset, 147int btrfs_add_ordered_extent(struct inode *inode, u64 file_offset,
145 u64 start, u64 len, u64 disk_len, int type); 148 u64 start, u64 len, u64 disk_len, int type);
146int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset, 149int btrfs_add_ordered_extent_dio(struct inode *inode, u64 file_offset,
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 8299a25ffc8f..dbb51ea7a13c 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -244,6 +244,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
244 case Opt_space_cache: 244 case Opt_space_cache:
245 printk(KERN_INFO "btrfs: enabling disk space caching\n"); 245 printk(KERN_INFO "btrfs: enabling disk space caching\n");
246 btrfs_set_opt(info->mount_opt, SPACE_CACHE); 246 btrfs_set_opt(info->mount_opt, SPACE_CACHE);
247 break;
247 case Opt_clear_cache: 248 case Opt_clear_cache:
248 printk(KERN_INFO "btrfs: force clearing of disk cache\n"); 249 printk(KERN_INFO "btrfs: force clearing of disk cache\n");
249 btrfs_set_opt(info->mount_opt, CLEAR_CACHE); 250 btrfs_set_opt(info->mount_opt, CLEAR_CACHE);
@@ -562,12 +563,26 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
562 563
563static int btrfs_test_super(struct super_block *s, void *data) 564static int btrfs_test_super(struct super_block *s, void *data)
564{ 565{
565 struct btrfs_fs_devices *test_fs_devices = data; 566 struct btrfs_root *test_root = data;
566 struct btrfs_root *root = btrfs_sb(s); 567 struct btrfs_root *root = btrfs_sb(s);
567 568
568 return root->fs_info->fs_devices == test_fs_devices; 569 /*
570 * If this super block is going away, return false as it
571 * can't match as an existing super block.
572 */
573 if (!atomic_read(&s->s_active))
574 return 0;
575 return root->fs_info->fs_devices == test_root->fs_info->fs_devices;
576}
577
578static int btrfs_set_super(struct super_block *s, void *data)
579{
580 s->s_fs_info = data;
581
582 return set_anon_super(s, data);
569} 583}
570 584
585
571/* 586/*
572 * Find a superblock for the given device / mount point. 587 * Find a superblock for the given device / mount point.
573 * 588 *
@@ -581,6 +596,8 @@ static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
581 struct super_block *s; 596 struct super_block *s;
582 struct dentry *root; 597 struct dentry *root;
583 struct btrfs_fs_devices *fs_devices = NULL; 598 struct btrfs_fs_devices *fs_devices = NULL;
599 struct btrfs_root *tree_root = NULL;
600 struct btrfs_fs_info *fs_info = NULL;
584 fmode_t mode = FMODE_READ; 601 fmode_t mode = FMODE_READ;
585 char *subvol_name = NULL; 602 char *subvol_name = NULL;
586 u64 subvol_objectid = 0; 603 u64 subvol_objectid = 0;
@@ -608,8 +625,24 @@ static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
608 goto error_close_devices; 625 goto error_close_devices;
609 } 626 }
610 627
628 /*
629 * Setup a dummy root and fs_info for test/set super. This is because
630 * we don't actually fill this stuff out until open_ctree, but we need
631 * it for searching for existing supers, so this lets us do that and
632 * then open_ctree will properly initialize everything later.
633 */
634 fs_info = kzalloc(sizeof(struct btrfs_fs_info), GFP_NOFS);
635 tree_root = kzalloc(sizeof(struct btrfs_root), GFP_NOFS);
636 if (!fs_info || !tree_root) {
637 error = -ENOMEM;
638 goto error_close_devices;
639 }
640 fs_info->tree_root = tree_root;
641 fs_info->fs_devices = fs_devices;
642 tree_root->fs_info = fs_info;
643
611 bdev = fs_devices->latest_bdev; 644 bdev = fs_devices->latest_bdev;
612 s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices); 645 s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
613 if (IS_ERR(s)) 646 if (IS_ERR(s))
614 goto error_s; 647 goto error_s;
615 648
@@ -675,6 +708,8 @@ error_s:
675 error = PTR_ERR(s); 708 error = PTR_ERR(s);
676error_close_devices: 709error_close_devices:
677 btrfs_close_devices(fs_devices); 710 btrfs_close_devices(fs_devices);
711 kfree(fs_info);
712 kfree(tree_root);
678error_free_subvol_name: 713error_free_subvol_name:
679 kfree(subvol_name); 714 kfree(subvol_name);
680 return ERR_PTR(error); 715 return ERR_PTR(error);
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 1fffbc017bdf..f50e931fc217 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -902,6 +902,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
902 struct btrfs_root *root = pending->root; 902 struct btrfs_root *root = pending->root;
903 struct btrfs_root *parent_root; 903 struct btrfs_root *parent_root;
904 struct inode *parent_inode; 904 struct inode *parent_inode;
905 struct dentry *parent;
905 struct dentry *dentry; 906 struct dentry *dentry;
906 struct extent_buffer *tmp; 907 struct extent_buffer *tmp;
907 struct extent_buffer *old; 908 struct extent_buffer *old;
@@ -941,7 +942,8 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
941 trans->block_rsv = &pending->block_rsv; 942 trans->block_rsv = &pending->block_rsv;
942 943
943 dentry = pending->dentry; 944 dentry = pending->dentry;
944 parent_inode = dentry->d_parent->d_inode; 945 parent = dget_parent(dentry);
946 parent_inode = parent->d_inode;
945 parent_root = BTRFS_I(parent_inode)->root; 947 parent_root = BTRFS_I(parent_inode)->root;
946 record_root_in_trans(trans, parent_root); 948 record_root_in_trans(trans, parent_root);
947 949
@@ -989,6 +991,7 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
989 parent_inode->i_ino, index, 991 parent_inode->i_ino, index,
990 dentry->d_name.name, dentry->d_name.len); 992 dentry->d_name.name, dentry->d_name.len);
991 BUG_ON(ret); 993 BUG_ON(ret);
994 dput(parent);
992 995
993 key.offset = (u64)-1; 996 key.offset = (u64)-1;
994 pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key); 997 pending->snap = btrfs_read_fs_root_no_name(root->fs_info, &key);
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index a29f19384a27..054744ac5719 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -2869,6 +2869,7 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
2869{ 2869{
2870 int ret = 0; 2870 int ret = 0;
2871 struct btrfs_root *root; 2871 struct btrfs_root *root;
2872 struct dentry *old_parent = NULL;
2872 2873
2873 /* 2874 /*
2874 * for regular files, if its inode is already on disk, we don't 2875 * for regular files, if its inode is already on disk, we don't
@@ -2910,10 +2911,13 @@ static noinline int check_parent_dirs_for_sync(struct btrfs_trans_handle *trans,
2910 if (IS_ROOT(parent)) 2911 if (IS_ROOT(parent))
2911 break; 2912 break;
2912 2913
2913 parent = parent->d_parent; 2914 parent = dget_parent(parent);
2915 dput(old_parent);
2916 old_parent = parent;
2914 inode = parent->d_inode; 2917 inode = parent->d_inode;
2915 2918
2916 } 2919 }
2920 dput(old_parent);
2917out: 2921out:
2918 return ret; 2922 return ret;
2919} 2923}
@@ -2945,6 +2949,7 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
2945{ 2949{
2946 int inode_only = exists_only ? LOG_INODE_EXISTS : LOG_INODE_ALL; 2950 int inode_only = exists_only ? LOG_INODE_EXISTS : LOG_INODE_ALL;
2947 struct super_block *sb; 2951 struct super_block *sb;
2952 struct dentry *old_parent = NULL;
2948 int ret = 0; 2953 int ret = 0;
2949 u64 last_committed = root->fs_info->last_trans_committed; 2954 u64 last_committed = root->fs_info->last_trans_committed;
2950 2955
@@ -3016,10 +3021,13 @@ int btrfs_log_inode_parent(struct btrfs_trans_handle *trans,
3016 if (IS_ROOT(parent)) 3021 if (IS_ROOT(parent))
3017 break; 3022 break;
3018 3023
3019 parent = parent->d_parent; 3024 parent = dget_parent(parent);
3025 dput(old_parent);
3026 old_parent = parent;
3020 } 3027 }
3021 ret = 0; 3028 ret = 0;
3022end_trans: 3029end_trans:
3030 dput(old_parent);
3023 if (ret < 0) { 3031 if (ret < 0) {
3024 BUG_ON(ret != -ENOSPC); 3032 BUG_ON(ret != -ENOSPC);
3025 root->fs_info->last_trans_log_full_commit = trans->transid; 3033 root->fs_info->last_trans_log_full_commit = trans->transid;
@@ -3039,8 +3047,13 @@ end_no_trans:
3039int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans, 3047int btrfs_log_dentry_safe(struct btrfs_trans_handle *trans,
3040 struct btrfs_root *root, struct dentry *dentry) 3048 struct btrfs_root *root, struct dentry *dentry)
3041{ 3049{
3042 return btrfs_log_inode_parent(trans, root, dentry->d_inode, 3050 struct dentry *parent = dget_parent(dentry);
3043 dentry->d_parent, 0); 3051 int ret;
3052
3053 ret = btrfs_log_inode_parent(trans, root, dentry->d_inode, parent, 0);
3054 dput(parent);
3055
3056 return ret;
3044} 3057}
3045 3058
3046/* 3059/*