From 5d4f98a28c7d334091c1b7744f48a1acdd2a4ae0 Mon Sep 17 00:00:00 2001 From: Yan Zheng Date: Wed, 10 Jun 2009 10:45:14 -0400 Subject: Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng Signed-off-by: Chris Mason --- fs/btrfs/super.c | 26 +++++++++----------------- 1 file changed, 9 insertions(+), 17 deletions(-) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 2ff7cd2db25f..e9ef8c3307fe 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -52,7 +52,6 @@ #include "export.h" #include "compression.h" - static struct super_operations btrfs_super_ops; static void btrfs_put_super(struct super_block *sb) @@ -322,7 +321,7 @@ static int btrfs_fill_super(struct super_block *sb, struct dentry *root_dentry; struct btrfs_super_block *disk_super; struct btrfs_root *tree_root; - struct btrfs_inode *bi; + struct btrfs_key key; int err; sb->s_maxbytes = MAX_LFS_FILESIZE; @@ -341,23 +340,15 @@ static int btrfs_fill_super(struct super_block *sb, } sb->s_fs_info = tree_root; disk_super = &tree_root->fs_info->super_copy; - inode = btrfs_iget_locked(sb, BTRFS_FIRST_FREE_OBJECTID, - tree_root->fs_info->fs_root); - bi = BTRFS_I(inode); - bi->location.objectid = inode->i_ino; - bi->location.offset = 0; - bi->root = tree_root->fs_info->fs_root; - - btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY); - if (!inode) { - err = -ENOMEM; + key.objectid = BTRFS_FIRST_FREE_OBJECTID; + key.type = BTRFS_INODE_ITEM_KEY; + key.offset = 0; + inode = btrfs_iget(sb, &key, tree_root->fs_info->fs_root); + if (IS_ERR(inode)) { + err = PTR_ERR(inode); goto fail_close; } - if (inode->i_state & I_NEW) { - btrfs_read_locked_inode(inode); - unlock_new_inode(inode); - } root_dentry = d_alloc_root(inode); if (!root_dentry) { @@ -584,7 +575,8 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data) if (btrfs_super_log_root(&root->fs_info->super_copy) != 0) return -EINVAL; - ret = btrfs_cleanup_reloc_trees(root); + /* recover relocation */ + ret = btrfs_recover_relocation(root); WARN_ON(ret); ret = btrfs_cleanup_fs_roots(root->fs_info); -- cgit v1.2.3 From 3b30c22f64a6bb297719c60e494af1d26563f584 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 9 Jun 2009 16:42:22 -0400 Subject: Btrfs: Add mount -o nossd This allows you to turn off the ssd mode via remount. Signed-off-by: Chris Mason --- fs/btrfs/super.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index e9ef8c3307fe..22855a18eb48 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -66,8 +66,8 @@ static void btrfs_put_super(struct super_block *sb) enum { Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, - Opt_ssd, Opt_thread_pool, Opt_noacl, Opt_compress, Opt_notreelog, - Opt_ratio, Opt_flushoncommit, Opt_err, + Opt_ssd, Opt_nossd, Opt_thread_pool, Opt_noacl, Opt_compress, + Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err, }; static match_table_t tokens = { @@ -83,6 +83,7 @@ static match_table_t tokens = { {Opt_thread_pool, "thread_pool=%d"}, {Opt_compress, "compress"}, {Opt_ssd, "ssd"}, + {Opt_nossd, "nossd"}, {Opt_noacl, "noacl"}, {Opt_notreelog, "notreelog"}, {Opt_flushoncommit, "flushoncommit"}, @@ -173,6 +174,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); btrfs_set_opt(info->mount_opt, SSD); break; + case Opt_nossd: + printk(KERN_INFO "btrfs: not using ssd allocation scheme\n"); + btrfs_clear_opt(info->mount_opt, SSD); + break; case Opt_nobarrier: printk(KERN_INFO "btrfs: turning off barriers\n"); btrfs_set_opt(info->mount_opt, NOBARRIER); -- cgit v1.2.3 From 451d7585a8bb1b9bec0d676ce3dece1923164e55 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Tue, 9 Jun 2009 20:28:34 -0400 Subject: Btrfs: add mount -o ssd_spread to spread allocations out Some SSDs perform best when reusing block numbers often, while others perform much better when clustering strictly allocates big chunks of unused space. The default mount -o ssd will find rough groupings of blocks where there are a bunch of free blocks that might have some allocated blocks mixed in. mount -o ssd_spread will make sure there are no allocated blocks mixed in. It should perform better on lower end SSDs. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/extent-tree.c | 2 +- fs/btrfs/free-space-cache.c | 5 ++++- fs/btrfs/free-space-cache.h | 1 + fs/btrfs/super.c | 19 +++++++++++++++---- 5 files changed, 22 insertions(+), 6 deletions(-) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index ce3ab4e13064..b9d8788b299e 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1100,6 +1100,7 @@ struct btrfs_root { #define BTRFS_MOUNT_COMPRESS (1 << 5) #define BTRFS_MOUNT_NOTREELOG (1 << 6) #define BTRFS_MOUNT_FLUSHONCOMMIT (1 << 7) +#define BTRFS_MOUNT_SSD_SPREAD (1 << 8) #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index a42419c276e2..3355d7ea8308 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -3607,7 +3607,7 @@ refill_cluster: last_ptr_loop = 0; /* allocate a cluster in this block group */ - ret = btrfs_find_space_cluster(trans, + ret = btrfs_find_space_cluster(trans, root, block_group, last_ptr, offset, num_bytes, empty_cluster + empty_size); diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c index ac23476beb6e..4538e48581a5 100644 --- a/fs/btrfs/free-space-cache.c +++ b/fs/btrfs/free-space-cache.c @@ -579,6 +579,7 @@ out: * it returns -enospc */ int btrfs_find_space_cluster(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_block_group_cache *block_group, struct btrfs_free_cluster *cluster, u64 offset, u64 bytes, u64 empty_size) @@ -595,7 +596,9 @@ int btrfs_find_space_cluster(struct btrfs_trans_handle *trans, int ret; /* for metadata, allow allocates with more holes */ - if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) { + if (btrfs_test_opt(root, SSD_SPREAD)) { + min_bytes = bytes + empty_size; + } else if (block_group->flags & BTRFS_BLOCK_GROUP_METADATA) { /* * we want to do larger allocations when we are * flushing out the delayed refs, it helps prevent diff --git a/fs/btrfs/free-space-cache.h b/fs/btrfs/free-space-cache.h index ab0bdc0a63ce..266fb8764054 100644 --- a/fs/btrfs/free-space-cache.h +++ b/fs/btrfs/free-space-cache.h @@ -31,6 +31,7 @@ void btrfs_dump_free_space(struct btrfs_block_group_cache *block_group, u64 bytes); u64 btrfs_block_group_free_space(struct btrfs_block_group_cache *block_group); int btrfs_find_space_cluster(struct btrfs_trans_handle *trans, + struct btrfs_root *root, struct btrfs_block_group_cache *block_group, struct btrfs_free_cluster *cluster, u64 offset, u64 bytes, u64 empty_size); diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 22855a18eb48..7f5b2889949a 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -66,8 +66,8 @@ static void btrfs_put_super(struct super_block *sb) enum { Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow, Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, - Opt_ssd, Opt_nossd, Opt_thread_pool, Opt_noacl, Opt_compress, - Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err, + Opt_ssd, Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, + Opt_compress, Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_err, }; static match_table_t tokens = { @@ -83,6 +83,7 @@ static match_table_t tokens = { {Opt_thread_pool, "thread_pool=%d"}, {Opt_compress, "compress"}, {Opt_ssd, "ssd"}, + {Opt_ssd_spread, "ssd_spread"}, {Opt_nossd, "nossd"}, {Opt_noacl, "noacl"}, {Opt_notreelog, "notreelog"}, @@ -174,9 +175,17 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) printk(KERN_INFO "btrfs: use ssd allocation scheme\n"); btrfs_set_opt(info->mount_opt, SSD); break; + case Opt_ssd_spread: + printk(KERN_INFO "btrfs: use spread ssd " + "allocation scheme\n"); + btrfs_set_opt(info->mount_opt, SSD); + btrfs_set_opt(info->mount_opt, SSD_SPREAD); + break; case Opt_nossd: - printk(KERN_INFO "btrfs: not using ssd allocation scheme\n"); + printk(KERN_INFO "btrfs: not using ssd allocation " + "scheme\n"); btrfs_clear_opt(info->mount_opt, SSD); + btrfs_clear_opt(info->mount_opt, SSD_SPREAD); break; case Opt_nobarrier: printk(KERN_INFO "btrfs: turning off barriers\n"); @@ -429,7 +438,9 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs) seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); if (btrfs_test_opt(root, COMPRESS)) seq_puts(seq, ",compress"); - if (btrfs_test_opt(root, SSD)) + if (btrfs_test_opt(root, SSD_SPREAD)) + seq_puts(seq, ",ssd_spread"); + else if (btrfs_test_opt(root, SSD)) seq_puts(seq, ",ssd"); if (btrfs_test_opt(root, NOTREELOG)) seq_puts(seq, ",notreelog"); -- cgit v1.2.3 From c289811cc096c57ff35550ee8132793a4f9b5b59 Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Wed, 10 Jun 2009 09:51:32 -0400 Subject: Btrfs: autodetect SSD devices During mount, btrfs will check the queue nonrot flag for all the devices found in the FS. If they are all non-rotating, SSD mode is enabled by default. If the FS was mounted with -o nossd, the non-rotating flag is ignored. Signed-off-by: Chris Mason --- fs/btrfs/ctree.h | 1 + fs/btrfs/disk-io.c | 9 +++++++++ fs/btrfs/super.c | 3 +++ fs/btrfs/volumes.c | 6 ++++++ fs/btrfs/volumes.h | 5 +++++ 5 files changed, 24 insertions(+) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index b9d8788b299e..5fa7d7d287a4 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -1101,6 +1101,7 @@ struct btrfs_root { #define BTRFS_MOUNT_NOTREELOG (1 << 6) #define BTRFS_MOUNT_FLUSHONCOMMIT (1 << 7) #define BTRFS_MOUNT_SSD_SPREAD (1 << 8) +#define BTRFS_MOUNT_NOSSD (1 << 9) #define btrfs_clear_opt(o, opt) ((o) &= ~BTRFS_MOUNT_##opt) #define btrfs_set_opt(o, opt) ((o) |= BTRFS_MOUNT_##opt) diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c index e572cf478a5d..f4dfbb7ab496 100644 --- a/fs/btrfs/disk-io.c +++ b/fs/btrfs/disk-io.c @@ -1850,6 +1850,14 @@ struct btrfs_root *open_ctree(struct super_block *sb, if (IS_ERR(fs_info->transaction_kthread)) goto fail_cleaner; + if (!btrfs_test_opt(tree_root, SSD) && + !btrfs_test_opt(tree_root, NOSSD) && + !fs_info->fs_devices->rotating) { + printk(KERN_INFO "Btrfs detected SSD devices, enabling SSD " + "mode\n"); + btrfs_set_opt(fs_info->mount_opt, SSD); + } + if (btrfs_super_log_root(disk_super) != 0) { u64 bytenr = btrfs_super_log_root(disk_super); @@ -1893,6 +1901,7 @@ struct btrfs_root *open_ctree(struct super_block *sb, fs_info->fs_root = btrfs_read_fs_root_no_name(fs_info, &location); if (!fs_info->fs_root) goto fail_trans_kthread; + return tree_root; fail_trans_kthread: diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 7f5b2889949a..3427db28f6fe 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -184,6 +184,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) case Opt_nossd: printk(KERN_INFO "btrfs: not using ssd allocation " "scheme\n"); + btrfs_set_opt(info->mount_opt, NOSSD); btrfs_clear_opt(info->mount_opt, SSD); btrfs_clear_opt(info->mount_opt, SSD_SPREAD); break; @@ -438,6 +439,8 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs) seq_printf(seq, ",thread_pool=%d", info->thread_pool_size); if (btrfs_test_opt(root, COMPRESS)) seq_puts(seq, ",compress"); + if (btrfs_test_opt(root, NOSSD)) + seq_puts(seq, ",nossd"); if (btrfs_test_opt(root, SSD_SPREAD)) seq_puts(seq, ",ssd_spread"); else if (btrfs_test_opt(root, SSD)) diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c index 27d5f37b845f..3f4a5932eac9 100644 --- a/fs/btrfs/volumes.c +++ b/fs/btrfs/volumes.c @@ -605,6 +605,9 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices, device->in_fs_metadata = 0; device->mode = flags; + if (!blk_queue_nonrot(bdev_get_queue(bdev))) + fs_devices->rotating = 1; + fs_devices->open_devices++; if (device->writeable) { fs_devices->rw_devices++; @@ -1473,6 +1476,9 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path) root->fs_info->fs_devices->rw_devices++; root->fs_info->fs_devices->total_rw_bytes += device->total_bytes; + if (!blk_queue_nonrot(bdev_get_queue(bdev))) + root->fs_info->fs_devices->rotating = 1; + total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy); btrfs_set_super_total_bytes(&root->fs_info->super_copy, total_bytes + device->total_bytes); diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h index 5c3ff6d02fd7..3c1f7310421e 100644 --- a/fs/btrfs/volumes.h +++ b/fs/btrfs/volumes.h @@ -107,6 +107,11 @@ struct btrfs_fs_devices { int seeding; int opened; + + /* set when we find or add a device that doesn't have the + * nonrot flag set + */ + int rotating; }; struct btrfs_bio_stripe { -- cgit v1.2.3 From 067c28adc53807514ac0c6ebb6af3243cbd071fa Mon Sep 17 00:00:00 2001 From: Chris Mason Date: Thu, 11 Jun 2009 09:30:13 -0400 Subject: Btrfs: fix -o nodatasum printk spelling It was printing nodatacsum, which was not the correct option name. Signed-off-by: Chris Mason --- fs/btrfs/super.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 3427db28f6fe..708ac06b953b 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -159,7 +159,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options) */ break; case Opt_nodatasum: - printk(KERN_INFO "btrfs: setting nodatacsum\n"); + printk(KERN_INFO "btrfs: setting nodatasum\n"); btrfs_set_opt(info->mount_opt, NODATASUM); break; case Opt_nodatacow: -- cgit v1.2.3 From 59d697b70285c348c01cfc2695c3469ba71d7539 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Mon, 27 Apr 2009 09:46:41 -0400 Subject: btrfs: remove ->write_super and stop maintaining ->s_dirt Signed-off-by: Christoph Hellwig Signed-off-by: Al Viro --- fs/btrfs/inode.c | 7 ------- fs/btrfs/super.c | 8 -------- 2 files changed, 15 deletions(-) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 5b68330f8585..8612b3a09811 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -2322,7 +2322,6 @@ err: btrfs_update_inode(trans, root, dir); btrfs_drop_nlink(inode); ret = btrfs_update_inode(trans, root, inode); - dir->i_sb->s_dirt = 1; out: return ret; } @@ -2806,7 +2805,6 @@ error: pending_del_nr); } btrfs_free_path(path); - inode->i_sb->s_dirt = 1; return ret; } @@ -3768,7 +3766,6 @@ static int btrfs_mknod(struct inode *dir, struct dentry *dentry, init_special_inode(inode, inode->i_mode, rdev); btrfs_update_inode(trans, root, inode); } - dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); btrfs_update_inode_block_group(trans, dir); out_unlock: @@ -3833,7 +3830,6 @@ static int btrfs_create(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } - dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); btrfs_update_inode_block_group(trans, dir); out_unlock: @@ -3880,7 +3876,6 @@ static int btrfs_link(struct dentry *old_dentry, struct inode *dir, if (err) drop_inode = 1; - dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, dir); err = btrfs_update_inode(trans, root, inode); @@ -3962,7 +3957,6 @@ static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, int mode) d_instantiate(dentry, inode); drop_on_err = 0; - dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); btrfs_update_inode_block_group(trans, dir); @@ -4991,7 +4985,6 @@ static int btrfs_symlink(struct inode *dir, struct dentry *dentry, inode->i_op = &btrfs_file_inode_operations; BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops; } - dir->i_sb->s_dirt = 1; btrfs_update_inode_block_group(trans, inode); btrfs_update_inode_block_group(trans, dir); if (drop_inode) diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 708ac06b953b..52d84522c2c2 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -397,7 +397,6 @@ int btrfs_sync_fs(struct super_block *sb, int wait) if (sb->s_flags & MS_RDONLY) return 0; - sb->s_dirt = 0; if (!wait) { filemap_flush(root->fs_info->btree_inode->i_mapping); return 0; @@ -408,7 +407,6 @@ int btrfs_sync_fs(struct super_block *sb, int wait) trans = btrfs_start_transaction(root, 1); ret = btrfs_commit_transaction(trans, root); - sb->s_dirt = 0; return ret; } @@ -454,11 +452,6 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs) return 0; } -static void btrfs_write_super(struct super_block *sb) -{ - sb->s_dirt = 0; -} - static int btrfs_test_super(struct super_block *s, void *data) { struct btrfs_fs_devices *test_fs_devices = data; @@ -689,7 +682,6 @@ static int btrfs_unfreeze(struct super_block *sb) static struct super_operations btrfs_super_ops = { .delete_inode = btrfs_delete_inode, .put_super = btrfs_put_super, - .write_super = btrfs_write_super, .sync_fs = btrfs_sync_fs, .show_options = btrfs_show_options, .write_inode = btrfs_write_inode, -- cgit v1.2.3 From 5af7926ff33b68b3ba46531471c6e0564b285efc Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Tue, 5 May 2009 15:41:25 +0200 Subject: enforce ->sync_fs is only called for rw superblock Make sure a superblock really is writeable by checking MS_RDONLY under s_umount. sync_filesystems needed some re-arragement for that, but all but one sync_filesystem caller had the correct locking already so that we could add that check there. cachefiles grew s_umount locking. I've also added a WARN_ON to sync_filesystem to assert this for future callers. Signed-off-by: Christoph Hellwig Signed-off-by: Al Viro --- fs/btrfs/super.c | 3 --- fs/cachefiles/interface.c | 2 ++ fs/reiserfs/super.c | 21 +++++++++------------ fs/sync.c | 23 ++++++++++++++++------- fs/ubifs/super.c | 3 --- 5 files changed, 27 insertions(+), 25 deletions(-) (limited to 'fs/btrfs/super.c') diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c index 52d84522c2c2..9f179d4832d5 100644 --- a/fs/btrfs/super.c +++ b/fs/btrfs/super.c @@ -394,9 +394,6 @@ int btrfs_sync_fs(struct super_block *sb, int wait) struct btrfs_root *root = btrfs_sb(sb); int ret; - if (sb->s_flags & MS_RDONLY) - return 0; - if (!wait) { filemap_flush(root->fs_info->btree_inode->i_mapping); return 0; diff --git a/fs/cachefiles/interface.c b/fs/cachefiles/interface.c index dafd484d7bda..431accd475a7 100644 --- a/fs/cachefiles/interface.c +++ b/fs/cachefiles/interface.c @@ -354,7 +354,9 @@ static void cachefiles_sync_cache(struct fscache_cache *_cache) /* make sure all pages pinned by operations on behalf of the netfs are * written to disc */ cachefiles_begin_secure(cache, &saved_cred); + down_read(&cache->mnt->mnt_sb->s_umount); ret = sync_filesystem(cache->mnt->mnt_sb); + up_read(&cache->mnt->mnt_sb->s_umount); cachefiles_end_secure(cache, saved_cred); if (ret == -EIO) diff --git a/fs/reiserfs/super.c b/fs/reiserfs/super.c index 1b52daa351c5..3da0401c0a96 100644 --- a/fs/reiserfs/super.c +++ b/fs/reiserfs/super.c @@ -64,18 +64,15 @@ static int reiserfs_statfs(struct dentry *dentry, struct kstatfs *buf); static int reiserfs_sync_fs(struct super_block *s, int wait) { - if (!(s->s_flags & MS_RDONLY)) { - struct reiserfs_transaction_handle th; - reiserfs_write_lock(s); - if (!journal_begin(&th, s, 1)) - if (!journal_end_sync(&th, s, 1)) - reiserfs_flush_old_commits(s); - s->s_dirt = 0; /* Even if it's not true. - * We'll loop forever in sync_supers otherwise */ - reiserfs_write_unlock(s); - } else { - s->s_dirt = 0; - } + struct reiserfs_transaction_handle th; + + reiserfs_write_lock(s); + if (!journal_begin(&th, s, 1)) + if (!journal_end_sync(&th, s, 1)) + reiserfs_flush_old_commits(s); + s->s_dirt = 0; /* Even if it's not true. + * We'll loop forever in sync_supers otherwise */ + reiserfs_write_unlock(s); return 0; } diff --git a/fs/sync.c b/fs/sync.c index 4487b5560dc8..89c37f732afa 100644 --- a/fs/sync.c +++ b/fs/sync.c @@ -51,6 +51,18 @@ int sync_filesystem(struct super_block *sb) { int ret; + /* + * We need to be protected against the filesystem going from + * r/o to r/w or vice versa. + */ + WARN_ON(!rwsem_is_locked(&sb->s_umount)); + + /* + * No point in syncing out anything if the filesystem is read-only. + */ + if (sb->s_flags & MS_RDONLY) + return 0; + ret = __sync_filesystem(sb, 0); if (ret < 0) return ret; @@ -79,25 +91,22 @@ static void sync_filesystems(int wait) mutex_lock(&mutex); /* Could be down_interruptible */ spin_lock(&sb_lock); - list_for_each_entry(sb, &super_blocks, s_list) { - if (sb->s_flags & MS_RDONLY) - continue; + list_for_each_entry(sb, &super_blocks, s_list) sb->s_need_sync = 1; - } restart: list_for_each_entry(sb, &super_blocks, s_list) { if (!sb->s_need_sync) continue; sb->s_need_sync = 0; - if (sb->s_flags & MS_RDONLY) - continue; /* hm. Was remounted r/o meanwhile */ sb->s_count++; spin_unlock(&sb_lock); + down_read(&sb->s_umount); - if (sb->s_root) + if (!(sb->s_flags & MS_RDONLY) && sb->s_root) __sync_filesystem(sb, wait); up_read(&sb->s_umount); + /* restart only when sb is no longer on the list */ spin_lock(&sb_lock); if (__put_super_and_need_restart(sb)) diff --git a/fs/ubifs/super.c b/fs/ubifs/super.c index e9f7a754c4f7..84f3c7fd1552 100644 --- a/fs/ubifs/super.c +++ b/fs/ubifs/super.c @@ -447,9 +447,6 @@ static int ubifs_sync_fs(struct super_block *sb, int wait) if (!wait) return 0; - if (sb->s_flags & MS_RDONLY) - return 0; - /* * VFS calls '->sync_fs()' before synchronizing all dirty inodes and * pages, so synchronize them first, then commit the journal. Strictly -- cgit v1.2.3