Btrfs: fix race between transaction commit and empty block group removal

[linux-2.6-block.git] / fs / btrfs / disk-io.c

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 0696a103751e7f8be5337469187faceb8b3e9283..41b320e235d7ccf3fda2f3b6894501abd68cb78b 100644 (file)
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1275,12 +1275,10 @@ static void __setup_root(u32 nodesize, u32 sectorsize, u32 stripesize,
        memset(&root->root_key, 0, sizeof(root->root_key));
        memset(&root->root_item, 0, sizeof(root->root_item));
        memset(&root->defrag_progress, 0, sizeof(root->defrag_progress));
-       memset(&root->root_kobj, 0, sizeof(root->root_kobj));
        if (fs_info)
                root->defrag_trans_start = fs_info->generation;
        else
                root->defrag_trans_start = 0;
-       init_completion(&root->kobj_unregister);
        root->root_key.objectid = objectid;
        root->anon_dev = 0;
 
@@ -2244,6 +2242,7 @@ int open_ctree(struct super_block *sb,
        spin_lock_init(&fs_info->qgroup_op_lock);
        spin_lock_init(&fs_info->buffer_lock);
        spin_lock_init(&fs_info->unused_bgs_lock);
+       mutex_init(&fs_info->unused_bg_unpin_mutex);
        rwlock_init(&fs_info->tree_mod_log_lock);
        mutex_init(&fs_info->reloc_mutex);
        mutex_init(&fs_info->delalloc_root_mutex);
@@ -3873,6 +3872,21 @@ static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
                printk(KERN_WARNING "BTRFS: log_root block unaligned: %llu\n",
                                btrfs_super_log_root(sb));
 
+       /*
+        * Check the lower bound, the alignment and other constraints are
+        * checked later.
+        */
+       if (btrfs_super_nodesize(sb) < 4096) {
+               printk(KERN_ERR "BTRFS: nodesize too small: %u < 4096\n",
+                               btrfs_super_nodesize(sb));
+               ret = -EINVAL;
+       }
+       if (btrfs_super_sectorsize(sb) < 4096) {
+               printk(KERN_ERR "BTRFS: sectorsize too small: %u < 4096\n",
+                               btrfs_super_sectorsize(sb));
+               ret = -EINVAL;
+       }
+
        if (memcmp(fs_info->fsid, sb->dev_item.fsid, BTRFS_UUID_SIZE) != 0) {
                printk(KERN_ERR "BTRFS: dev_item UUID does not match fsid: %pU != %pU\n",
                                fs_info->fsid, sb->dev_item.fsid);
@@ -3886,6 +3900,10 @@ static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
        if (btrfs_super_num_devices(sb) > (1UL << 31))
                printk(KERN_WARNING "BTRFS: suspicious number of devices: %llu\n",
                                btrfs_super_num_devices(sb));
+       if (btrfs_super_num_devices(sb) == 0) {
+               printk(KERN_ERR "BTRFS: number of devices is 0\n");
+               ret = -EINVAL;
+       }
 
        if (btrfs_super_bytenr(sb) != BTRFS_SUPER_INFO_OFFSET) {
                printk(KERN_ERR "BTRFS: super offset mismatch %llu != %u\n",
@@ -3893,6 +3911,25 @@ static int btrfs_check_super_valid(struct btrfs_fs_info *fs_info,
                ret = -EINVAL;
        }
 
+       /*
+        * Obvious sys_chunk_array corruptions, it must hold at least one key
+        * and one chunk
+        */
+       if (btrfs_super_sys_array_size(sb) > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE) {
+               printk(KERN_ERR "BTRFS: system chunk array too big %u > %u\n",
+                               btrfs_super_sys_array_size(sb),
+                               BTRFS_SYSTEM_CHUNK_ARRAY_SIZE);
+               ret = -EINVAL;
+       }
+       if (btrfs_super_sys_array_size(sb) < sizeof(struct btrfs_disk_key)
+                       + sizeof(struct btrfs_chunk)) {
+               printk(KERN_ERR "BTRFS: system chunk array too small %u < %lu\n",
+                               btrfs_super_sys_array_size(sb),
+                               sizeof(struct btrfs_disk_key)
+                               + sizeof(struct btrfs_chunk));
+               ret = -EINVAL;
+       }
+
        /*
         * The generation is a global counter, we'll trust it more than the others
         * but it's still possible that it's the one that's wrong.