aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--fs/btrfs/extent-tree.c6
-rw-r--r--fs/btrfs/extent_map.c13
-rw-r--r--fs/btrfs/extent_map.h1
-rw-r--r--fs/btrfs/file-item.c4
-rw-r--r--fs/btrfs/file.c10
-rw-r--r--fs/btrfs/free-space-cache.c20
-rw-r--r--fs/btrfs/inode.c137
-rw-r--r--fs/btrfs/ioctl.c129
-rw-r--r--fs/btrfs/qgroup.c20
-rw-r--r--fs/btrfs/send.c4
-rw-r--r--fs/btrfs/super.c2
-rw-r--r--fs/btrfs/transaction.c19
-rw-r--r--fs/btrfs/tree-log.c10
-rw-r--r--fs/btrfs/volumes.c23
14 files changed, 300 insertions, 98 deletions
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 521e9d4424f6..a8b8adc05070 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -3997,7 +3997,7 @@ again:
3997 * We make the other tasks wait for the flush only when we can flush 3997 * We make the other tasks wait for the flush only when we can flush
3998 * all things. 3998 * all things.
3999 */ 3999 */
4000 if (ret && flush == BTRFS_RESERVE_FLUSH_ALL) { 4000 if (ret && flush != BTRFS_RESERVE_NO_FLUSH) {
4001 flushing = true; 4001 flushing = true;
4002 space_info->flush = 1; 4002 space_info->flush = 1;
4003 } 4003 }
@@ -5560,7 +5560,7 @@ static noinline int find_free_extent(struct btrfs_trans_handle *trans,
5560 int empty_cluster = 2 * 1024 * 1024; 5560 int empty_cluster = 2 * 1024 * 1024;
5561 struct btrfs_space_info *space_info; 5561 struct btrfs_space_info *space_info;
5562 int loop = 0; 5562 int loop = 0;
5563 int index = 0; 5563 int index = __get_raid_index(data);
5564 int alloc_type = (data & BTRFS_BLOCK_GROUP_DATA) ? 5564 int alloc_type = (data & BTRFS_BLOCK_GROUP_DATA) ?
5565 RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC; 5565 RESERVE_ALLOC_NO_ACCOUNT : RESERVE_ALLOC;
5566 bool found_uncached_bg = false; 5566 bool found_uncached_bg = false;
@@ -6788,11 +6788,13 @@ static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
6788 &wc->flags[level]); 6788 &wc->flags[level]);
6789 if (ret < 0) { 6789 if (ret < 0) {
6790 btrfs_tree_unlock_rw(eb, path->locks[level]); 6790 btrfs_tree_unlock_rw(eb, path->locks[level]);
6791 path->locks[level] = 0;
6791 return ret; 6792 return ret;
6792 } 6793 }
6793 BUG_ON(wc->refs[level] == 0); 6794 BUG_ON(wc->refs[level] == 0);
6794 if (wc->refs[level] == 1) { 6795 if (wc->refs[level] == 1) {
6795 btrfs_tree_unlock_rw(eb, path->locks[level]); 6796 btrfs_tree_unlock_rw(eb, path->locks[level]);
6797 path->locks[level] = 0;
6796 return 1; 6798 return 1;
6797 } 6799 }
6798 } 6800 }
diff --git a/fs/btrfs/extent_map.c b/fs/btrfs/extent_map.c
index f169d6b11d7f..2e8cae63d247 100644
--- a/fs/btrfs/extent_map.c
+++ b/fs/btrfs/extent_map.c
@@ -171,6 +171,10 @@ static int mergable_maps(struct extent_map *prev, struct extent_map *next)
171 if (test_bit(EXTENT_FLAG_COMPRESSED, &prev->flags)) 171 if (test_bit(EXTENT_FLAG_COMPRESSED, &prev->flags))
172 return 0; 172 return 0;
173 173
174 if (test_bit(EXTENT_FLAG_LOGGING, &prev->flags) ||
175 test_bit(EXTENT_FLAG_LOGGING, &next->flags))
176 return 0;
177
174 if (extent_map_end(prev) == next->start && 178 if (extent_map_end(prev) == next->start &&
175 prev->flags == next->flags && 179 prev->flags == next->flags &&
176 prev->bdev == next->bdev && 180 prev->bdev == next->bdev &&
@@ -255,7 +259,8 @@ int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len,
255 if (!em) 259 if (!em)
256 goto out; 260 goto out;
257 261
258 list_move(&em->list, &tree->modified_extents); 262 if (!test_bit(EXTENT_FLAG_LOGGING, &em->flags))
263 list_move(&em->list, &tree->modified_extents);
259 em->generation = gen; 264 em->generation = gen;
260 clear_bit(EXTENT_FLAG_PINNED, &em->flags); 265 clear_bit(EXTENT_FLAG_PINNED, &em->flags);
261 em->mod_start = em->start; 266 em->mod_start = em->start;
@@ -280,6 +285,12 @@ out:
280 285
281} 286}
282 287
288void clear_em_logging(struct extent_map_tree *tree, struct extent_map *em)
289{
290 clear_bit(EXTENT_FLAG_LOGGING, &em->flags);
291 try_merge_map(tree, em);
292}
293
283/** 294/**
284 * add_extent_mapping - add new extent map to the extent tree 295 * add_extent_mapping - add new extent map to the extent tree
285 * @tree: tree to insert new map in 296 * @tree: tree to insert new map in
diff --git a/fs/btrfs/extent_map.h b/fs/btrfs/extent_map.h
index 922943ce29e8..c6598c89cff8 100644
--- a/fs/btrfs/extent_map.h
+++ b/fs/btrfs/extent_map.h
@@ -69,6 +69,7 @@ void free_extent_map(struct extent_map *em);
69int __init extent_map_init(void); 69int __init extent_map_init(void);
70void extent_map_exit(void); 70void extent_map_exit(void);
71int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len, u64 gen); 71int unpin_extent_cache(struct extent_map_tree *tree, u64 start, u64 len, u64 gen);
72void clear_em_logging(struct extent_map_tree *tree, struct extent_map *em);
72struct extent_map *search_extent_mapping(struct extent_map_tree *tree, 73struct extent_map *search_extent_mapping(struct extent_map_tree *tree,
73 u64 start, u64 len); 74 u64 start, u64 len);
74#endif 75#endif
diff --git a/fs/btrfs/file-item.c b/fs/btrfs/file-item.c
index bd38cef42358..94aa53b38721 100644
--- a/fs/btrfs/file-item.c
+++ b/fs/btrfs/file-item.c
@@ -460,8 +460,8 @@ int btrfs_csum_one_bio(struct btrfs_root *root, struct inode *inode,
460 if (!contig) 460 if (!contig)
461 offset = page_offset(bvec->bv_page) + bvec->bv_offset; 461 offset = page_offset(bvec->bv_page) + bvec->bv_offset;
462 462
463 if (!contig && (offset >= ordered->file_offset + ordered->len || 463 if (offset >= ordered->file_offset + ordered->len ||
464 offset < ordered->file_offset)) { 464 offset < ordered->file_offset) {
465 unsigned long bytes_left; 465 unsigned long bytes_left;
466 sums->len = this_sum_bytes; 466 sums->len = this_sum_bytes;
467 this_sum_bytes = 0; 467 this_sum_bytes = 0;
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 77061bf43edb..f76b1fd160d4 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -2241,6 +2241,7 @@ static int find_desired_extent(struct inode *inode, loff_t *offset, int whence)
2241 if (lockend <= lockstart) 2241 if (lockend <= lockstart)
2242 lockend = lockstart + root->sectorsize; 2242 lockend = lockstart + root->sectorsize;
2243 2243
2244 lockend--;
2244 len = lockend - lockstart + 1; 2245 len = lockend - lockstart + 1;
2245 2246
2246 len = max_t(u64, len, root->sectorsize); 2247 len = max_t(u64, len, root->sectorsize);
@@ -2307,9 +2308,12 @@ static int find_desired_extent(struct inode *inode, loff_t *offset, int whence)
2307 } 2308 }
2308 } 2309 }
2309 2310
2310 *offset = start; 2311 if (!test_bit(EXTENT_FLAG_PREALLOC,
2311 free_extent_map(em); 2312 &em->flags)) {
2312 break; 2313 *offset = start;
2314 free_extent_map(em);
2315 break;
2316 }
2313 } 2317 }
2314 } 2318 }
2315 2319
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 59ea2e4349c9..0be7a8742a43 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -1862,11 +1862,13 @@ int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
1862{ 1862{
1863 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl; 1863 struct btrfs_free_space_ctl *ctl = block_group->free_space_ctl;
1864 struct btrfs_free_space *info; 1864 struct btrfs_free_space *info;
1865 int ret = 0; 1865 int ret;
1866 bool re_search = false;
1866 1867
1867 spin_lock(&ctl->tree_lock); 1868 spin_lock(&ctl->tree_lock);
1868 1869
1869again: 1870again:
1871 ret = 0;
1870 if (!bytes) 1872 if (!bytes)
1871 goto out_lock; 1873 goto out_lock;
1872 1874
@@ -1879,17 +1881,17 @@ again:
1879 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset), 1881 info = tree_search_offset(ctl, offset_to_bitmap(ctl, offset),
1880 1, 0); 1882 1, 0);
1881 if (!info) { 1883 if (!info) {
1882 /* the tree logging code might be calling us before we 1884 /*
1883 * have fully loaded the free space rbtree for this 1885 * If we found a partial bit of our free space in a
1884 * block group. So it is possible the entry won't 1886 * bitmap but then couldn't find the other part this may
1885 * be in the rbtree yet at all. The caching code 1887 * be a problem, so WARN about it.
1886 * will make sure not to put it in the rbtree if
1887 * the logging code has pinned it.
1888 */ 1888 */
1889 WARN_ON(re_search);
1889 goto out_lock; 1890 goto out_lock;
1890 } 1891 }
1891 } 1892 }
1892 1893
1894 re_search = false;
1893 if (!info->bitmap) { 1895 if (!info->bitmap) {
1894 unlink_free_space(ctl, info); 1896 unlink_free_space(ctl, info);
1895 if (offset == info->offset) { 1897 if (offset == info->offset) {
@@ -1935,8 +1937,10 @@ again:
1935 } 1937 }
1936 1938
1937 ret = remove_from_bitmap(ctl, info, &offset, &bytes); 1939 ret = remove_from_bitmap(ctl, info, &offset, &bytes);
1938 if (ret == -EAGAIN) 1940 if (ret == -EAGAIN) {
1941 re_search = true;
1939 goto again; 1942 goto again;
1943 }
1940 BUG_ON(ret); /* logic error */ 1944 BUG_ON(ret); /* logic error */
1941out_lock: 1945out_lock:
1942 spin_unlock(&ctl->tree_lock); 1946 spin_unlock(&ctl->tree_lock);
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 16d9e8e191e6..cc93b23ca352 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -88,7 +88,7 @@ static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
88 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK, 88 [S_IFLNK >> S_SHIFT] = BTRFS_FT_SYMLINK,
89}; 89};
90 90
91static int btrfs_setsize(struct inode *inode, loff_t newsize); 91static int btrfs_setsize(struct inode *inode, struct iattr *attr);
92static int btrfs_truncate(struct inode *inode); 92static int btrfs_truncate(struct inode *inode);
93static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent); 93static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);
94static noinline int cow_file_range(struct inode *inode, 94static noinline int cow_file_range(struct inode *inode,
@@ -2478,6 +2478,18 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
2478 continue; 2478 continue;
2479 } 2479 }
2480 nr_truncate++; 2480 nr_truncate++;
2481
2482 /* 1 for the orphan item deletion. */
2483 trans = btrfs_start_transaction(root, 1);
2484 if (IS_ERR(trans)) {
2485 ret = PTR_ERR(trans);
2486 goto out;
2487 }
2488 ret = btrfs_orphan_add(trans, inode);
2489 btrfs_end_transaction(trans, root);
2490 if (ret)
2491 goto out;
2492
2481 ret = btrfs_truncate(inode); 2493 ret = btrfs_truncate(inode);
2482 } else { 2494 } else {
2483 nr_unlink++; 2495 nr_unlink++;
@@ -3665,6 +3677,7 @@ int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
3665 block_end - cur_offset, 0); 3677 block_end - cur_offset, 0);
3666 if (IS_ERR(em)) { 3678 if (IS_ERR(em)) {
3667 err = PTR_ERR(em); 3679 err = PTR_ERR(em);
3680 em = NULL;
3668 break; 3681 break;
3669 } 3682 }
3670 last_byte = min(extent_map_end(em), block_end); 3683 last_byte = min(extent_map_end(em), block_end);
@@ -3748,16 +3761,27 @@ next:
3748 return err; 3761 return err;
3749} 3762}
3750 3763
3751static int btrfs_setsize(struct inode *inode, loff_t newsize) 3764static int btrfs_setsize(struct inode *inode, struct iattr *attr)
3752{ 3765{
3753 struct btrfs_root *root = BTRFS_I(inode)->root; 3766 struct btrfs_root *root = BTRFS_I(inode)->root;
3754 struct btrfs_trans_handle *trans; 3767 struct btrfs_trans_handle *trans;
3755 loff_t oldsize = i_size_read(inode); 3768 loff_t oldsize = i_size_read(inode);
3769 loff_t newsize = attr->ia_size;
3770 int mask = attr->ia_valid;
3756 int ret; 3771 int ret;
3757 3772
3758 if (newsize == oldsize) 3773 if (newsize == oldsize)
3759 return 0; 3774 return 0;
3760 3775
3776 /*
3777 * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
3778 * special case where we need to update the times despite not having
3779 * these flags set. For all other operations the VFS set these flags
3780 * explicitly if it wants a timestamp update.
3781 */
3782 if (newsize != oldsize && (!(mask & (ATTR_CTIME | ATTR_MTIME))))
3783 inode->i_ctime = inode->i_mtime = current_fs_time(inode->i_sb);
3784
3761 if (newsize > oldsize) { 3785 if (newsize > oldsize) {
3762 truncate_pagecache(inode, oldsize, newsize); 3786 truncate_pagecache(inode, oldsize, newsize);
3763 ret = btrfs_cont_expand(inode, oldsize, newsize); 3787 ret = btrfs_cont_expand(inode, oldsize, newsize);
@@ -3783,9 +3807,34 @@ static int btrfs_setsize(struct inode *inode, loff_t newsize)
3783 set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE, 3807 set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
3784 &BTRFS_I(inode)->runtime_flags); 3808 &BTRFS_I(inode)->runtime_flags);
3785 3809
3810 /*
3811 * 1 for the orphan item we're going to add
3812 * 1 for the orphan item deletion.
3813 */
3814 trans = btrfs_start_transaction(root, 2);
3815 if (IS_ERR(trans))
3816 return PTR_ERR(trans);
3817
3818 /*
3819 * We need to do this in case we fail at _any_ point during the
3820 * actual truncate. Once we do the truncate_setsize we could
3821 * invalidate pages which forces any outstanding ordered io to
3822 * be instantly completed which will give us extents that need
3823 * to be truncated. If we fail to get an orphan inode down we
3824 * could have left over extents that were never meant to live,
3825 * so we need to garuntee from this point on that everything
3826 * will be consistent.
3827 */
3828 ret = btrfs_orphan_add(trans, inode);
3829 btrfs_end_transaction(trans, root);
3830 if (ret)
3831 return ret;
3832
3786 /* we don't support swapfiles, so vmtruncate shouldn't fail */ 3833 /* we don't support swapfiles, so vmtruncate shouldn't fail */
3787 truncate_setsize(inode, newsize); 3834 truncate_setsize(inode, newsize);
3788 ret = btrfs_truncate(inode); 3835 ret = btrfs_truncate(inode);
3836 if (ret && inode->i_nlink)
3837 btrfs_orphan_del(NULL, inode);
3789 } 3838 }
3790 3839
3791 return ret; 3840 return ret;
@@ -3805,7 +3854,7 @@ static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
3805 return err; 3854 return err;
3806 3855
3807 if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) { 3856 if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
3808 err = btrfs_setsize(inode, attr->ia_size); 3857 err = btrfs_setsize(inode, attr);
3809 if (err) 3858 if (err)
3810 return err; 3859 return err;
3811 } 3860 }
@@ -5572,10 +5621,13 @@ struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *pag
5572 return em; 5621 return em;
5573 if (em) { 5622 if (em) {
5574 /* 5623 /*
5575 * if our em maps to a hole, there might 5624 * if our em maps to
5576 * actually be delalloc bytes behind it 5625 * - a hole or
5626 * - a pre-alloc extent,
5627 * there might actually be delalloc bytes behind it.
5577 */ 5628 */
5578 if (em->block_start != EXTENT_MAP_HOLE) 5629 if (em->block_start != EXTENT_MAP_HOLE &&
5630 !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
5579 return em; 5631 return em;
5580 else 5632 else
5581 hole_em = em; 5633 hole_em = em;
@@ -5657,6 +5709,8 @@ struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *pag
5657 */ 5709 */
5658 em->block_start = hole_em->block_start; 5710 em->block_start = hole_em->block_start;
5659 em->block_len = hole_len; 5711 em->block_len = hole_len;
5712 if (test_bit(EXTENT_FLAG_PREALLOC, &hole_em->flags))
5713 set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
5660 } else { 5714 } else {
5661 em->start = range_start; 5715 em->start = range_start;
5662 em->len = found; 5716 em->len = found;
@@ -6915,11 +6969,9 @@ static int btrfs_truncate(struct inode *inode)
6915 6969
6916 /* 6970 /*
6917 * 1 for the truncate slack space 6971 * 1 for the truncate slack space
6918 * 1 for the orphan item we're going to add
6919 * 1 for the orphan item deletion
6920 * 1 for updating the inode. 6972 * 1 for updating the inode.
6921 */ 6973 */
6922 trans = btrfs_start_transaction(root, 4); 6974 trans = btrfs_start_transaction(root, 2);
6923 if (IS_ERR(trans)) { 6975 if (IS_ERR(trans)) {
6924 err = PTR_ERR(trans); 6976 err = PTR_ERR(trans);
6925 goto out; 6977 goto out;
@@ -6930,12 +6982,6 @@ static int btrfs_truncate(struct inode *inode)
6930 min_size); 6982 min_size);
6931 BUG_ON(ret); 6983 BUG_ON(ret);
6932 6984
6933 ret = btrfs_orphan_add(trans, inode);
6934 if (ret) {
6935 btrfs_end_transaction(trans, root);
6936 goto out;
6937 }
6938
6939 /* 6985 /*
6940 * setattr is responsible for setting the ordered_data_close flag, 6986 * setattr is responsible for setting the ordered_data_close flag,
6941 * but that is only tested during the last file release. That 6987 * but that is only tested during the last file release. That
@@ -7004,12 +7050,6 @@ static int btrfs_truncate(struct inode *inode)
7004 ret = btrfs_orphan_del(trans, inode); 7050 ret = btrfs_orphan_del(trans, inode);
7005 if (ret) 7051 if (ret)
7006 err = ret; 7052 err = ret;
7007 } else if (ret && inode->i_nlink > 0) {
7008 /*
7009 * Failed to do the truncate, remove us from the in memory
7010 * orphan list.
7011 */
7012 ret = btrfs_orphan_del(NULL, inode);
7013 } 7053 }
7014 7054
7015 if (trans) { 7055 if (trans) {
@@ -7531,41 +7571,61 @@ void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work)
7531 */ 7571 */
7532int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput) 7572int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
7533{ 7573{
7534 struct list_head *head = &root->fs_info->delalloc_inodes;
7535 struct btrfs_inode *binode; 7574 struct btrfs_inode *binode;
7536 struct inode *inode; 7575 struct inode *inode;
7537 struct btrfs_delalloc_work *work, *next; 7576 struct btrfs_delalloc_work *work, *next;
7538 struct list_head works; 7577 struct list_head works;
7578 struct list_head splice;
7539 int ret = 0; 7579 int ret = 0;
7540 7580
7541 if (root->fs_info->sb->s_flags & MS_RDONLY) 7581 if (root->fs_info->sb->s_flags & MS_RDONLY)
7542 return -EROFS; 7582 return -EROFS;
7543 7583
7544 INIT_LIST_HEAD(&works); 7584 INIT_LIST_HEAD(&works);
7545 7585 INIT_LIST_HEAD(&splice);
7586again:
7546 spin_lock(&root->fs_info->delalloc_lock); 7587 spin_lock(&root->fs_info->delalloc_lock);
7547 while (!list_empty(head)) { 7588 list_splice_init(&root->fs_info->delalloc_inodes, &splice);
7548 binode = list_entry(head->next, struct btrfs_inode, 7589 while (!list_empty(&splice)) {
7590 binode = list_entry(splice.next, struct btrfs_inode,
7549 delalloc_inodes); 7591 delalloc_inodes);
7592
7593 list_del_init(&binode->delalloc_inodes);
7594
7550 inode = igrab(&binode->vfs_inode); 7595 inode = igrab(&binode->vfs_inode);
7551 if (!inode) 7596 if (!inode)
7552 list_del_init(&binode->delalloc_inodes); 7597 continue;
7598
7599 list_add_tail(&binode->delalloc_inodes,
7600 &root->fs_info->delalloc_inodes);
7553 spin_unlock(&root->fs_info->delalloc_lock); 7601 spin_unlock(&root->fs_info->delalloc_lock);
7554 if (inode) { 7602
7555 work = btrfs_alloc_delalloc_work(inode, 0, delay_iput); 7603 work = btrfs_alloc_delalloc_work(inode, 0, delay_iput);
7556 if (!work) { 7604 if (unlikely(!work)) {
7557 ret = -ENOMEM; 7605 ret = -ENOMEM;
7558 goto out; 7606 goto out;
7559 }
7560 list_add_tail(&work->list, &works);
7561 btrfs_queue_worker(&root->fs_info->flush_workers,
7562 &work->work);
7563 } 7607 }
7608 list_add_tail(&work->list, &works);
7609 btrfs_queue_worker(&root->fs_info->flush_workers,
7610 &work->work);
7611
7564 cond_resched(); 7612 cond_resched();
7565 spin_lock(&root->fs_info->delalloc_lock); 7613 spin_lock(&root->fs_info->delalloc_lock);
7566 } 7614 }
7567 spin_unlock(&root->fs_info->delalloc_lock); 7615 spin_unlock(&root->fs_info->delalloc_lock);
7568 7616
7617 list_for_each_entry_safe(work, next, &works, list) {
7618 list_del_init(&work->list);
7619 btrfs_wait_and_free_delalloc_work(work);
7620 }
7621
7622 spin_lock(&root->fs_info->delalloc_lock);
7623 if (!list_empty(&root->fs_info->delalloc_inodes)) {
7624 spin_unlock(&root->fs_info->delalloc_lock);
7625 goto again;
7626 }
7627 spin_unlock(&root->fs_info->delalloc_lock);
7628
7569 /* the filemap_flush will queue IO into the worker threads, but 7629 /* the filemap_flush will queue IO into the worker threads, but
7570 * we have to make sure the IO is actually started and that 7630 * we have to make sure the IO is actually started and that
7571 * ordered extents get created before we return 7631 * ordered extents get created before we return
@@ -7578,11 +7638,18 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
7578 atomic_read(&root->fs_info->async_delalloc_pages) == 0)); 7638 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
7579 } 7639 }
7580 atomic_dec(&root->fs_info->async_submit_draining); 7640 atomic_dec(&root->fs_info->async_submit_draining);
7641 return 0;
7581out: 7642out:
7582 list_for_each_entry_safe(work, next, &works, list) { 7643 list_for_each_entry_safe(work, next, &works, list) {
7583 list_del_init(&work->list); 7644 list_del_init(&work->list);
7584 btrfs_wait_and_free_delalloc_work(work); 7645 btrfs_wait_and_free_delalloc_work(work);
7585 } 7646 }
7647
7648 if (!list_empty_careful(&splice)) {
7649 spin_lock(&root->fs_info->delalloc_lock);
7650 list_splice_tail(&splice, &root->fs_info->delalloc_inodes);
7651 spin_unlock(&root->fs_info->delalloc_lock);
7652 }
7586 return ret; 7653 return ret;
7587} 7654}
7588 7655
diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index 4b4516770f05..5b22d45d3c6a 100644
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -1339,7 +1339,8 @@ static noinline int btrfs_ioctl_resize(struct file *file,
1339 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 1339 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1340 1)) { 1340 1)) {
1341 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 1341 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1342 return -EINPROGRESS; 1342 mnt_drop_write_file(file);
1343 return -EINVAL;
1343 } 1344 }
1344 1345
1345 mutex_lock(&root->fs_info->volume_mutex); 1346 mutex_lock(&root->fs_info->volume_mutex);
@@ -1362,6 +1363,7 @@ static noinline int btrfs_ioctl_resize(struct file *file,
1362 printk(KERN_INFO "btrfs: resizing devid %llu\n", 1363 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1363 (unsigned long long)devid); 1364 (unsigned long long)devid);
1364 } 1365 }
1366
1365 device = btrfs_find_device(root->fs_info, devid, NULL, NULL); 1367 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1366 if (!device) { 1368 if (!device) {
1367 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n", 1369 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
@@ -1369,9 +1371,10 @@ static noinline int btrfs_ioctl_resize(struct file *file,
1369 ret = -EINVAL; 1371 ret = -EINVAL;
1370 goto out_free; 1372 goto out_free;
1371 } 1373 }
1372 if (device->fs_devices && device->fs_devices->seeding) { 1374
1375 if (!device->writeable) {
1373 printk(KERN_INFO "btrfs: resizer unable to apply on " 1376 printk(KERN_INFO "btrfs: resizer unable to apply on "
1374 "seeding device %llu\n", 1377 "readonly device %llu\n",
1375 (unsigned long long)devid); 1378 (unsigned long long)devid);
1376 ret = -EINVAL; 1379 ret = -EINVAL;
1377 goto out_free; 1380 goto out_free;
@@ -1443,8 +1446,8 @@ out_free:
1443 kfree(vol_args); 1446 kfree(vol_args);
1444out: 1447out:
1445 mutex_unlock(&root->fs_info->volume_mutex); 1448 mutex_unlock(&root->fs_info->volume_mutex);
1446 mnt_drop_write_file(file);
1447 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 1449 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1450 mnt_drop_write_file(file);
1448 return ret; 1451 return ret;
1449} 1452}
1450 1453
@@ -2095,13 +2098,13 @@ static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2095 err = inode_permission(inode, MAY_WRITE | MAY_EXEC); 2098 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2096 if (err) 2099 if (err)
2097 goto out_dput; 2100 goto out_dput;
2098
2099 /* check if subvolume may be deleted by a non-root user */
2100 err = btrfs_may_delete(dir, dentry, 1);
2101 if (err)
2102 goto out_dput;
2103 } 2101 }
2104 2102
2103 /* check if subvolume may be deleted by a user */
2104 err = btrfs_may_delete(dir, dentry, 1);
2105 if (err)
2106 goto out_dput;
2107
2105 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) { 2108 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2106 err = -EINVAL; 2109 err = -EINVAL;
2107 goto out_dput; 2110 goto out_dput;
@@ -2183,19 +2186,20 @@ static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2183 struct btrfs_ioctl_defrag_range_args *range; 2186 struct btrfs_ioctl_defrag_range_args *range;
2184 int ret; 2187 int ret;
2185 2188
2186 if (btrfs_root_readonly(root)) 2189 ret = mnt_want_write_file(file);
2187 return -EROFS; 2190 if (ret)
2191 return ret;
2188 2192
2189 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2193 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2190 1)) { 2194 1)) {
2191 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 2195 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2192 return -EINPROGRESS; 2196 mnt_drop_write_file(file);
2197 return -EINVAL;
2193 } 2198 }
2194 ret = mnt_want_write_file(file); 2199
2195 if (ret) { 2200 if (btrfs_root_readonly(root)) {
2196 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 2201 ret = -EROFS;
2197 0); 2202 goto out;
2198 return ret;
2199 } 2203 }
2200 2204
2201 switch (inode->i_mode & S_IFMT) { 2205 switch (inode->i_mode & S_IFMT) {
@@ -2247,8 +2251,8 @@ static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2247 ret = -EINVAL; 2251 ret = -EINVAL;
2248 } 2252 }
2249out: 2253out:
2250 mnt_drop_write_file(file);
2251 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2254 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2255 mnt_drop_write_file(file);
2252 return ret; 2256 return ret;
2253} 2257}
2254 2258
@@ -2263,7 +2267,7 @@ static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2263 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 2267 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2264 1)) { 2268 1)) {
2265 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 2269 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2266 return -EINPROGRESS; 2270 return -EINVAL;
2267 } 2271 }
2268 2272
2269 mutex_lock(&root->fs_info->volume_mutex); 2273 mutex_lock(&root->fs_info->volume_mutex);
@@ -2300,7 +2304,7 @@ static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2300 1)) { 2304 1)) {
2301 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n"); 2305 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2302 mnt_drop_write_file(file); 2306 mnt_drop_write_file(file);
2303 return -EINPROGRESS; 2307 return -EINVAL;
2304 } 2308 }
2305 2309
2306 mutex_lock(&root->fs_info->volume_mutex); 2310 mutex_lock(&root->fs_info->volume_mutex);
@@ -2316,8 +2320,8 @@ static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2316 kfree(vol_args); 2320 kfree(vol_args);
2317out: 2321out:
2318 mutex_unlock(&root->fs_info->volume_mutex); 2322 mutex_unlock(&root->fs_info->volume_mutex);
2319 mnt_drop_write_file(file);
2320 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0); 2323 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2324 mnt_drop_write_file(file);
2321 return ret; 2325 return ret;
2322} 2326}
2323 2327
@@ -3437,8 +3441,8 @@ static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3437 struct btrfs_fs_info *fs_info = root->fs_info; 3441 struct btrfs_fs_info *fs_info = root->fs_info;
3438 struct btrfs_ioctl_balance_args *bargs; 3442 struct btrfs_ioctl_balance_args *bargs;
3439 struct btrfs_balance_control *bctl; 3443 struct btrfs_balance_control *bctl;
3444 bool need_unlock; /* for mut. excl. ops lock */
3440 int ret; 3445 int ret;
3441 int need_to_clear_lock = 0;
3442 3446
3443 if (!capable(CAP_SYS_ADMIN)) 3447 if (!capable(CAP_SYS_ADMIN))
3444 return -EPERM; 3448 return -EPERM;
@@ -3447,14 +3451,61 @@ static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3447 if (ret) 3451 if (ret)
3448 return ret; 3452 return ret;
3449 3453
3450 mutex_lock(&fs_info->volume_mutex); 3454again:
3455 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3456 mutex_lock(&fs_info->volume_mutex);
3457 mutex_lock(&fs_info->balance_mutex);
3458 need_unlock = true;
3459 goto locked;
3460 }
3461
3462 /*
3463 * mut. excl. ops lock is locked. Three possibilites:
3464 * (1) some other op is running
3465 * (2) balance is running
3466 * (3) balance is paused -- special case (think resume)
3467 */
3451 mutex_lock(&fs_info->balance_mutex); 3468 mutex_lock(&fs_info->balance_mutex);
3469 if (fs_info->balance_ctl) {
3470 /* this is either (2) or (3) */
3471 if (!atomic_read(&fs_info->balance_running)) {
3472 mutex_unlock(&fs_info->balance_mutex);
3473 if (!mutex_trylock(&fs_info->volume_mutex))
3474 goto again;
3475 mutex_lock(&fs_info->balance_mutex);
3476
3477 if (fs_info->balance_ctl &&
3478 !atomic_read(&fs_info->balance_running)) {
3479 /* this is (3) */
3480 need_unlock = false;
3481 goto locked;
3482 }
3483
3484 mutex_unlock(&fs_info->balance_mutex);
3485 mutex_unlock(&fs_info->volume_mutex);
3486 goto again;
3487 } else {
3488 /* this is (2) */
3489 mutex_unlock(&fs_info->balance_mutex);
3490 ret = -EINPROGRESS;
3491 goto out;
3492 }
3493 } else {
3494 /* this is (1) */
3495 mutex_unlock(&fs_info->balance_mutex);
3496 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3497 ret = -EINVAL;
3498 goto out;
3499 }
3500
3501locked:
3502 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3452 3503
3453 if (arg) { 3504 if (arg) {
3454 bargs = memdup_user(arg, sizeof(*bargs)); 3505 bargs = memdup_user(arg, sizeof(*bargs));
3455 if (IS_ERR(bargs)) { 3506 if (IS_ERR(bargs)) {
3456 ret = PTR_ERR(bargs); 3507 ret = PTR_ERR(bargs);
3457 goto out; 3508 goto out_unlock;
3458 } 3509 }
3459 3510
3460 if (bargs->flags & BTRFS_BALANCE_RESUME) { 3511 if (bargs->flags & BTRFS_BALANCE_RESUME) {
@@ -3474,13 +3525,10 @@ static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3474 bargs = NULL; 3525 bargs = NULL;
3475 } 3526 }
3476 3527
3477 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running, 3528 if (fs_info->balance_ctl) {
3478 1)) {
3479 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3480 ret = -EINPROGRESS; 3529 ret = -EINPROGRESS;
3481 goto out_bargs; 3530 goto out_bargs;
3482 } 3531 }
3483 need_to_clear_lock = 1;
3484 3532
3485 bctl = kzalloc(sizeof(*bctl), GFP_NOFS); 3533 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3486 if (!bctl) { 3534 if (!bctl) {
@@ -3501,11 +3549,17 @@ static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3501 } 3549 }
3502 3550
3503do_balance: 3551do_balance:
3504 ret = btrfs_balance(bctl, bargs);
3505 /* 3552 /*
3506 * bctl is freed in __cancel_balance or in free_fs_info if 3553 * Ownership of bctl and mutually_exclusive_operation_running
3507 * restriper was paused all the way until unmount 3554 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3555 * or, if restriper was paused all the way until unmount, in
3556 * free_fs_info. mutually_exclusive_operation_running is
3557 * cleared in __cancel_balance.
3508 */ 3558 */
3559 need_unlock = false;
3560
3561 ret = btrfs_balance(bctl, bargs);
3562
3509 if (arg) { 3563 if (arg) {
3510 if (copy_to_user(arg, bargs, sizeof(*bargs))) 3564 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3511 ret = -EFAULT; 3565 ret = -EFAULT;
@@ -3513,12 +3567,12 @@ do_balance:
3513 3567
3514out_bargs: 3568out_bargs:
3515 kfree(bargs); 3569 kfree(bargs);
3516out: 3570out_unlock:
3517 if (need_to_clear_lock)
3518 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
3519 0);
3520 mutex_unlock(&fs_info->balance_mutex); 3571 mutex_unlock(&fs_info->balance_mutex);
3521 mutex_unlock(&fs_info->volume_mutex); 3572 mutex_unlock(&fs_info->volume_mutex);
3573 if (need_unlock)
3574 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3575out:
3522 mnt_drop_write_file(file); 3576 mnt_drop_write_file(file);
3523 return ret; 3577 return ret;
3524} 3578}
@@ -3698,6 +3752,11 @@ static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
3698 goto drop_write; 3752 goto drop_write;
3699 } 3753 }
3700 3754
3755 if (!sa->qgroupid) {
3756 ret = -EINVAL;
3757 goto out;
3758 }
3759
3701 trans = btrfs_join_transaction(root); 3760 trans = btrfs_join_transaction(root);
3702 if (IS_ERR(trans)) { 3761 if (IS_ERR(trans)) {
3703 ret = PTR_ERR(trans); 3762 ret = PTR_ERR(trans);
diff --git a/fs/btrfs/qgroup.c b/fs/btrfs/qgroup.c
index fe9d02c45f8e..a5c856234323 100644
--- a/fs/btrfs/qgroup.c
+++ b/fs/btrfs/qgroup.c
@@ -379,6 +379,13 @@ next1:
379 379
380 ret = add_relation_rb(fs_info, found_key.objectid, 380 ret = add_relation_rb(fs_info, found_key.objectid,
381 found_key.offset); 381 found_key.offset);
382 if (ret == -ENOENT) {
383 printk(KERN_WARNING
384 "btrfs: orphan qgroup relation 0x%llx->0x%llx\n",
385 (unsigned long long)found_key.objectid,
386 (unsigned long long)found_key.offset);
387 ret = 0; /* ignore the error */
388 }
382 if (ret) 389 if (ret)
383 goto out; 390 goto out;
384next2: 391next2:
@@ -956,17 +963,28 @@ int btrfs_remove_qgroup(struct btrfs_trans_handle *trans,
956 struct btrfs_fs_info *fs_info, u64 qgroupid) 963 struct btrfs_fs_info *fs_info, u64 qgroupid)
957{ 964{
958 struct btrfs_root *quota_root; 965 struct btrfs_root *quota_root;
966 struct btrfs_qgroup *qgroup;
959 int ret = 0; 967 int ret = 0;
960 968
961 quota_root = fs_info->quota_root; 969 quota_root = fs_info->quota_root;
962 if (!quota_root) 970 if (!quota_root)
963 return -EINVAL; 971 return -EINVAL;
964 972
973 /* check if there are no relations to this qgroup */
974 spin_lock(&fs_info->qgroup_lock);
975 qgroup = find_qgroup_rb(fs_info, qgroupid);
976 if (qgroup) {
977 if (!list_empty(&qgroup->groups) || !list_empty(&qgroup->members)) {
978 spin_unlock(&fs_info->qgroup_lock);
979 return -EBUSY;
980 }
981 }
982 spin_unlock(&fs_info->qgroup_lock);
983
965 ret = del_qgroup_item(trans, quota_root, qgroupid); 984 ret = del_qgroup_item(trans, quota_root, qgroupid);
966 985
967 spin_lock(&fs_info->qgroup_lock); 986 spin_lock(&fs_info->qgroup_lock);
968 del_qgroup_rb(quota_root->fs_info, qgroupid); 987 del_qgroup_rb(quota_root->fs_info, qgroupid);
969
970 spin_unlock(&fs_info->qgroup_lock); 988 spin_unlock(&fs_info->qgroup_lock);
971 989
972 return ret; 990 return ret;
diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c
index 54454542ad40..321b7fb4e441 100644
--- a/fs/btrfs/send.c
+++ b/fs/btrfs/send.c
@@ -1814,8 +1814,10 @@ static int name_cache_insert(struct send_ctx *sctx,
1814 (unsigned long)nce->ino); 1814 (unsigned long)nce->ino);
1815 if (!nce_head) { 1815 if (!nce_head) {
1816 nce_head = kmalloc(sizeof(*nce_head), GFP_NOFS); 1816 nce_head = kmalloc(sizeof(*nce_head), GFP_NOFS);
1817 if (!nce_head) 1817 if (!nce_head) {
1818 kfree(nce);
1818 return -ENOMEM; 1819 return -ENOMEM;
1820 }
1819 INIT_LIST_HEAD(nce_head); 1821 INIT_LIST_HEAD(nce_head);
1820 1822
1821 ret = radix_tree_insert(&sctx->name_cache, nce->ino, nce_head); 1823 ret = radix_tree_insert(&sctx->name_cache, nce->ino, nce_head);
diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 99545df1b86c..d8982e9601d3 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -267,7 +267,7 @@ void __btrfs_abort_transaction(struct btrfs_trans_handle *trans,
267 function, line, errstr); 267 function, line, errstr);
268 return; 268 return;
269 } 269 }
270 trans->transaction->aborted = errno; 270 ACCESS_ONCE(trans->transaction->aborted) = errno;
271 __btrfs_std_error(root->fs_info, function, line, errno, NULL); 271 __btrfs_std_error(root->fs_info, function, line, errno, NULL);
272} 272}
273/* 273/*
diff --git a/fs/btrfs/transaction.c b/fs/btrfs/transaction.c
index 87fac9a21ea5..f15494699f3b 100644
--- a/fs/btrfs/transaction.c
+++ b/fs/btrfs/transaction.c
@@ -1468,7 +1468,8 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
1468 goto cleanup_transaction; 1468 goto cleanup_transaction;
1469 } 1469 }
1470 1470
1471 if (cur_trans->aborted) { 1471 /* Stop the commit early if ->aborted is set */
1472 if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
1472 ret = cur_trans->aborted; 1473 ret = cur_trans->aborted;
1473 goto cleanup_transaction; 1474 goto cleanup_transaction;
1474 } 1475 }
@@ -1574,6 +1575,11 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
1574 wait_event(cur_trans->writer_wait, 1575 wait_event(cur_trans->writer_wait,
1575 atomic_read(&cur_trans->num_writers) == 1); 1576 atomic_read(&cur_trans->num_writers) == 1);
1576 1577
1578 /* ->aborted might be set after the previous check, so check it */
1579 if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
1580 ret = cur_trans->aborted;
1581 goto cleanup_transaction;
1582 }
1577 /* 1583 /*
1578 * the reloc mutex makes sure that we stop 1584 * the reloc mutex makes sure that we stop
1579 * the balancing code from coming in and moving 1585 * the balancing code from coming in and moving
@@ -1657,6 +1663,17 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
1657 goto cleanup_transaction; 1663 goto cleanup_transaction;
1658 } 1664 }
1659 1665
1666 /*
1667 * The tasks which save the space cache and inode cache may also
1668 * update ->aborted, check it.
1669 */
1670 if (unlikely(ACCESS_ONCE(cur_trans->aborted))) {
1671 ret = cur_trans->aborted;
1672 mutex_unlock(&root->fs_info->tree_log_mutex);
1673 mutex_unlock(&root->fs_info->reloc_mutex);
1674 goto cleanup_transaction;
1675 }
1676
1660 btrfs_prepare_extent_commit(trans, root); 1677 btrfs_prepare_extent_commit(trans, root);
1661 1678
1662 cur_trans = root->fs_info->running_transaction; 1679 cur_trans = root->fs_info->running_transaction;
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 83186c7e45d4..9027bb1e7466 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -3357,6 +3357,11 @@ static int log_one_extent(struct btrfs_trans_handle *trans,
3357 if (skip_csum) 3357 if (skip_csum)
3358 return 0; 3358 return 0;
3359 3359
3360 if (em->compress_type) {
3361 csum_offset = 0;
3362 csum_len = block_len;
3363 }
3364
3360 /* block start is already adjusted for the file extent offset. */ 3365 /* block start is already adjusted for the file extent offset. */
3361 ret = btrfs_lookup_csums_range(log->fs_info->csum_root, 3366 ret = btrfs_lookup_csums_range(log->fs_info->csum_root,
3362 em->block_start + csum_offset, 3367 em->block_start + csum_offset,
@@ -3410,13 +3415,13 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
3410 em = list_entry(extents.next, struct extent_map, list); 3415 em = list_entry(extents.next, struct extent_map, list);
3411 3416
3412 list_del_init(&em->list); 3417 list_del_init(&em->list);
3413 clear_bit(EXTENT_FLAG_LOGGING, &em->flags);
3414 3418
3415 /* 3419 /*
3416 * If we had an error we just need to delete everybody from our 3420 * If we had an error we just need to delete everybody from our
3417 * private list. 3421 * private list.
3418 */ 3422 */
3419 if (ret) { 3423 if (ret) {
3424 clear_em_logging(tree, em);
3420 free_extent_map(em); 3425 free_extent_map(em);
3421 continue; 3426 continue;
3422 } 3427 }
@@ -3424,8 +3429,9 @@ static int btrfs_log_changed_extents(struct btrfs_trans_handle *trans,
3424 write_unlock(&tree->lock); 3429 write_unlock(&tree->lock);
3425 3430
3426 ret = log_one_extent(trans, inode, root, em, path); 3431 ret = log_one_extent(trans, inode, root, em, path);
3427 free_extent_map(em);
3428 write_lock(&tree->lock); 3432 write_lock(&tree->lock);
3433 clear_em_logging(tree, em);
3434 free_extent_map(em);
3429 } 3435 }
3430 WARN_ON(!list_empty(&extents)); 3436 WARN_ON(!list_empty(&extents));
3431 write_unlock(&tree->lock); 3437 write_unlock(&tree->lock);
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 5cce6aa74012..15f6efdf6463 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1431,7 +1431,7 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
1431 } 1431 }
1432 } else { 1432 } else {
1433 ret = btrfs_get_bdev_and_sb(device_path, 1433 ret = btrfs_get_bdev_and_sb(device_path,
1434 FMODE_READ | FMODE_EXCL, 1434 FMODE_WRITE | FMODE_EXCL,
1435 root->fs_info->bdev_holder, 0, 1435 root->fs_info->bdev_holder, 0,
1436 &bdev, &bh); 1436 &bdev, &bh);
1437 if (ret) 1437 if (ret)
@@ -2614,7 +2614,14 @@ static int chunk_usage_filter(struct btrfs_fs_info *fs_info, u64 chunk_offset,
2614 cache = btrfs_lookup_block_group(fs_info, chunk_offset); 2614 cache = btrfs_lookup_block_group(fs_info, chunk_offset);
2615 chunk_used = btrfs_block_group_used(&cache->item); 2615 chunk_used = btrfs_block_group_used(&cache->item);
2616 2616
2617 user_thresh = div_factor_fine(cache->key.offset, bargs->usage); 2617 if (bargs->usage == 0)
2618 user_thresh = 0;
2619 else if (bargs->usage > 100)
2620 user_thresh = cache->key.offset;
2621 else
2622 user_thresh = div_factor_fine(cache->key.offset,
2623 bargs->usage);
2624
2618 if (chunk_used < user_thresh) 2625 if (chunk_used < user_thresh)
2619 ret = 0; 2626 ret = 0;
2620 2627
@@ -2959,6 +2966,8 @@ static void __cancel_balance(struct btrfs_fs_info *fs_info)
2959 unset_balance_control(fs_info); 2966 unset_balance_control(fs_info);
2960 ret = del_balance_item(fs_info->tree_root); 2967 ret = del_balance_item(fs_info->tree_root);
2961 BUG_ON(ret); 2968 BUG_ON(ret);
2969
2970 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
2962} 2971}
2963 2972
2964void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock, 2973void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
@@ -3138,8 +3147,10 @@ int btrfs_balance(struct btrfs_balance_control *bctl,
3138out: 3147out:
3139 if (bctl->flags & BTRFS_BALANCE_RESUME) 3148 if (bctl->flags & BTRFS_BALANCE_RESUME)
3140 __cancel_balance(fs_info); 3149 __cancel_balance(fs_info);
3141 else 3150 else {
3142 kfree(bctl); 3151 kfree(bctl);
3152 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3153 }
3143 return ret; 3154 return ret;
3144} 3155}
3145 3156
@@ -3156,7 +3167,6 @@ static int balance_kthread(void *data)
3156 ret = btrfs_balance(fs_info->balance_ctl, NULL); 3167 ret = btrfs_balance(fs_info->balance_ctl, NULL);
3157 } 3168 }
3158 3169
3159 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3160 mutex_unlock(&fs_info->balance_mutex); 3170 mutex_unlock(&fs_info->balance_mutex);
3161 mutex_unlock(&fs_info->volume_mutex); 3171 mutex_unlock(&fs_info->volume_mutex);
3162 3172
@@ -3179,7 +3189,6 @@ int btrfs_resume_balance_async(struct btrfs_fs_info *fs_info)
3179 return 0; 3189 return 0;
3180 } 3190 }
3181 3191
3182 WARN_ON(atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1));
3183 tsk = kthread_run(balance_kthread, fs_info, "btrfs-balance"); 3192 tsk = kthread_run(balance_kthread, fs_info, "btrfs-balance");
3184 if (IS_ERR(tsk)) 3193 if (IS_ERR(tsk))
3185 return PTR_ERR(tsk); 3194 return PTR_ERR(tsk);
@@ -3233,6 +3242,8 @@ int btrfs_recover_balance(struct btrfs_fs_info *fs_info)
3233 btrfs_balance_sys(leaf, item, &disk_bargs); 3242 btrfs_balance_sys(leaf, item, &disk_bargs);
3234 btrfs_disk_balance_args_to_cpu(&bctl->sys, &disk_bargs); 3243 btrfs_disk_balance_args_to_cpu(&bctl->sys, &disk_bargs);
3235 3244
3245 WARN_ON(atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1));
3246
3236 mutex_lock(&fs_info->volume_mutex); 3247 mutex_lock(&fs_info->volume_mutex);
3237 mutex_lock(&fs_info->balance_mutex); 3248 mutex_lock(&fs_info->balance_mutex);
3238 3249
@@ -3496,7 +3507,7 @@ struct btrfs_raid_attr btrfs_raid_array[BTRFS_NR_RAID_TYPES] = {
3496 { 1, 1, 2, 2, 2, 2 /* raid1 */ }, 3507 { 1, 1, 2, 2, 2, 2 /* raid1 */ },
3497 { 1, 2, 1, 1, 1, 2 /* dup */ }, 3508 { 1, 2, 1, 1, 1, 2 /* dup */ },
3498 { 1, 1, 0, 2, 1, 1 /* raid0 */ }, 3509 { 1, 1, 0, 2, 1, 1 /* raid0 */ },
3499 { 1, 1, 0, 1, 1, 1 /* single */ }, 3510 { 1, 1, 1, 1, 1, 1 /* single */ },
3500}; 3511};
3501 3512
3502static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans, 3513static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,