Btrfs: Seed device support
[linux-2.6-block.git] / fs / btrfs / super.c
index 288300fa5848108673d5da56b7e0bbebf7b88787..92393cc60d086a9888088e83f868b180a7c1bac4 100644 (file)
@@ -45,6 +45,9 @@
 #include "print-tree.h"
 #include "xattr.h"
 #include "volumes.h"
+#include "version.h"
+#include "export.h"
+#include "compression.h"
 
 #define BTRFS_SUPER_MAGIC 0x9123683E
 
@@ -65,22 +68,26 @@ static void btrfs_put_super (struct super_block * sb)
 }
 
 enum {
-       Opt_degraded, Opt_subvol, Opt_nodatasum, Opt_nodatacow,
+       Opt_degraded, Opt_subvol, Opt_device, Opt_nodatasum, Opt_nodatacow,
        Opt_max_extent, Opt_max_inline, Opt_alloc_start, Opt_nobarrier,
-       Opt_ssd, Opt_err,
+       Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_err,
 };
 
 static match_table_t tokens = {
        {Opt_degraded, "degraded"},
        {Opt_subvol, "subvol=%s"},
+       {Opt_device, "device=%s"},
        {Opt_nodatasum, "nodatasum"},
        {Opt_nodatacow, "nodatacow"},
        {Opt_nobarrier, "nobarrier"},
        {Opt_max_extent, "max_extent=%s"},
        {Opt_max_inline, "max_inline=%s"},
        {Opt_alloc_start, "alloc_start=%s"},
+       {Opt_thread_pool, "thread_pool=%d"},
+       {Opt_compress, "compress"},
        {Opt_ssd, "ssd"},
-       {Opt_err, NULL}
+       {Opt_noacl, "noacl"},
+       {Opt_err, NULL},
 };
 
 u64 btrfs_parse_size(char *str)
@@ -117,6 +124,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
        struct btrfs_fs_info *info = root->fs_info;
        substring_t args[MAX_OPT_ARGS];
        char *p, *num;
+       int intarg;
 
        if (!options)
                return 0;
@@ -142,8 +150,9 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                        btrfs_set_opt(info->mount_opt, DEGRADED);
                        break;
                case Opt_subvol:
+               case Opt_device:
                        /*
-                        * This one is parsed by btrfs_parse_early_options
+                        * These are parsed by btrfs_parse_early_options
                         * and can be happily ignored here.
                         */
                        break;
@@ -156,6 +165,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                        btrfs_set_opt(info->mount_opt, NODATACOW);
                        btrfs_set_opt(info->mount_opt, NODATASUM);
                        break;
+               case Opt_compress:
+                       printk(KERN_INFO "btrfs: use compression\n");
+                       btrfs_set_opt(info->mount_opt, COMPRESS);
+                       break;
                case Opt_ssd:
                        printk(KERN_INFO "btrfs: use ssd allocation scheme\n");
                        btrfs_set_opt(info->mount_opt, SSD);
@@ -164,6 +177,15 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                        printk(KERN_INFO "btrfs: turning off barriers\n");
                        btrfs_set_opt(info->mount_opt, NOBARRIER);
                        break;
+               case Opt_thread_pool:
+                       intarg = 0;
+                       match_int(&args[0], &intarg);
+                       if (intarg) {
+                               info->thread_pool_size = intarg;
+                               printk(KERN_INFO "btrfs: thread pool %d\n",
+                                      info->thread_pool_size);
+                       }
+                       break;
                case Opt_max_extent:
                        num = match_strdup(&args[0]);
                        if (num) {
@@ -182,8 +204,11 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                                info->max_inline = btrfs_parse_size(num);
                                kfree(num);
 
-                               info->max_inline = max_t(u64,
-                                       info->max_inline, root->sectorsize);
+                               if (info->max_inline) {
+                                       info->max_inline = max_t(u64,
+                                               info->max_inline,
+                                               root->sectorsize);
+                               }
                                printk(KERN_INFO "btrfs: max_inline at %llu\n",
                                        info->max_inline);
                        }
@@ -198,6 +223,9 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                                        info->alloc_start);
                        }
                        break;
+               case Opt_noacl:
+                       root->fs_info->sb->s_flags &= ~MS_POSIXACL;
+                       break;
                default:
                        break;
                }
@@ -212,8 +240,9 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
  * All other options will be parsed on much later in the mount process and
  * only when we need to allocate a new super block.
  */
-static int btrfs_parse_early_options(const char *options,
-                       char **subvol_name)
+static int btrfs_parse_early_options(const char *options, int flags,
+               void *holder, char **subvol_name,
+               struct btrfs_fs_devices **fs_devices)
 {
        substring_t args[MAX_OPT_ARGS];
        char *opts, *p;
@@ -240,11 +269,18 @@ static int btrfs_parse_early_options(const char *options,
                case Opt_subvol:
                        *subvol_name = match_strdup(&args[0]);
                        break;
+               case Opt_device:
+                       error = btrfs_scan_one_device(match_strdup(&args[0]),
+                                       flags, holder, fs_devices);
+                       if (error)
+                               goto out_free_opts;
+                       break;
                default:
                        break;
                }
        }
 
+ out_free_opts:
        kfree(opts);
  out:
        /*
@@ -274,8 +310,10 @@ static int btrfs_fill_super(struct super_block * sb,
        sb->s_maxbytes = MAX_LFS_FILESIZE;
        sb->s_magic = BTRFS_SUPER_MAGIC;
        sb->s_op = &btrfs_super_ops;
+       sb->s_export_op = &btrfs_export_ops;
        sb->s_xattr = btrfs_xattr_handlers;
        sb->s_time_gran = 1;
+       sb->s_flags |= MS_POSIXACL;
 
        tree_root = open_ctree(sb, fs_devices, (char *)data);
 
@@ -316,12 +354,8 @@ static int btrfs_fill_super(struct super_block * sb,
                goto fail_close;
 
        sb->s_root = root_dentry;
-       btrfs_transaction_queue_work(tree_root, HZ * 30);
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
        save_mount_options(sb, data);
-#endif
-
        return 0;
 
 fail_close:
@@ -336,18 +370,22 @@ int btrfs_sync_fs(struct super_block *sb, int wait)
        int ret;
        root = btrfs_sb(sb);
 
+       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;
        }
+
+       btrfs_start_delalloc_inodes(root);
+       btrfs_wait_ordered_extents(root, 0);
+
        btrfs_clean_old_snapshots(root);
-       mutex_lock(&root->fs_info->fs_mutex);
-       btrfs_defrag_dirty_roots(root->fs_info);
        trans = btrfs_start_transaction(root, 1);
        ret = btrfs_commit_transaction(trans, root);
        sb->s_dirt = 0;
-       mutex_unlock(&root->fs_info->fs_mutex);
        return ret;
 }
 
@@ -380,7 +418,8 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
        struct btrfs_fs_devices *fs_devices = NULL;
        int error = 0;
 
-       error = btrfs_parse_early_options(data, &subvol_name);
+       error = btrfs_parse_early_options(data, flags, fs_type,
+                                         &subvol_name, &fs_devices);
        if (error)
                goto error;
 
@@ -392,10 +431,13 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
        if (error)
                goto error_free_subvol_name;
 
+       if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
+               error = -EACCES;
+               goto error_close_devices;
+       }
+
        bdev = fs_devices->latest_bdev;
-       btrfs_lock_volumes();
        s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
-       btrfs_unlock_volumes();
        if (IS_ERR(s))
                goto error_s;
 
@@ -404,9 +446,10 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
                        up_write(&s->s_umount);
                        deactivate_super(s);
                        error = -EBUSY;
-                       goto error_bdev;
+                       goto error_close_devices;
                }
 
+               btrfs_close_devices(fs_devices);
        } else {
                char b[BDEVNAME_SIZE];
 
@@ -424,19 +467,25 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
                s->s_flags |= MS_ACTIVE;
        }
 
-       root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name));
-       if (IS_ERR(root)) {
-               up_write(&s->s_umount);
-               deactivate_super(s);
-               error = PTR_ERR(root);
-               goto error;
-       }
-       if (!root->d_inode) {
-               dput(root);
-               up_write(&s->s_umount);
-               deactivate_super(s);
-               error = -ENXIO;
-               goto error;
+       if (!strcmp(subvol_name, "."))
+               root = dget(s->s_root);
+       else {
+               mutex_lock(&s->s_root->d_inode->i_mutex);
+               root = lookup_one_len(subvol_name, s->s_root, strlen(subvol_name));
+               mutex_unlock(&s->s_root->d_inode->i_mutex);
+               if (IS_ERR(root)) {
+                       up_write(&s->s_umount);
+                       deactivate_super(s);
+                       error = PTR_ERR(root);
+                       goto error;
+               }
+               if (!root->d_inode) {
+                       dput(root);
+                       up_write(&s->s_umount);
+                       deactivate_super(s);
+                       error = -ENXIO;
+                       goto error;
+               }
        }
 
        mnt->mnt_sb = s;
@@ -447,7 +496,7 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 
 error_s:
        error = PTR_ERR(s);
-error_bdev:
+error_close_devices:
        btrfs_close_devices(fs_devices);
 error_free_subvol_name:
        kfree(subvol_name);
@@ -455,11 +504,44 @@ error:
        return error;
 }
 
+static int btrfs_remount(struct super_block *sb, int *flags, char *data)
+{
+       struct btrfs_root *root = btrfs_sb(sb);
+       int ret;
+
+       if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY))
+               return 0;
+
+       if (*flags & MS_RDONLY) {
+               sb->s_flags |= MS_RDONLY;
+
+               ret =  btrfs_commit_super(root);
+               WARN_ON(ret);
+       } else {
+               if (root->fs_info->fs_devices->rw_devices == 0)
+                       return -EACCES;
+
+               if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
+                       return -EINVAL;
+
+               ret = btrfs_cleanup_reloc_trees(root);
+               WARN_ON(ret);
+
+               ret = btrfs_cleanup_fs_roots(root->fs_info);
+               WARN_ON(ret);
+
+               sb->s_flags &= ~MS_RDONLY;
+       }
+
+       return 0;
+}
+
 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
        struct btrfs_root *root = btrfs_sb(dentry->d_sb);
        struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
        int bits = dentry->d_sb->s_blocksize_bits;
+       __be32 *fsid = (__be32 *)root->fs_info->fsid;
 
        buf->f_namelen = BTRFS_NAME_LEN;
        buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
@@ -468,6 +550,15 @@ static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
        buf->f_bavail = buf->f_bfree;
        buf->f_bsize = dentry->d_sb->s_blocksize;
        buf->f_type = BTRFS_SUPER_MAGIC;
+       /* We treat it as constant endianness (it doesn't matter _which_)
+          because we want the fsid to come out the same whether mounted 
+          on a big-endian or little-endian host */
+       buf->f_fsid.val[0] = be32_to_cpu(fsid[0]) ^ be32_to_cpu(fsid[2]);
+       buf->f_fsid.val[1] = be32_to_cpu(fsid[1]) ^ be32_to_cpu(fsid[3]);
+       /* Mask in the root object ID too, to disambiguate subvols */
+       buf->f_fsid.val[0] ^= BTRFS_I(dentry->d_inode)->root->objectid >> 32;
+       buf->f_fsid.val[1] ^= BTRFS_I(dentry->d_inode)->root->objectid;
+
        return 0;
 }
 
@@ -479,6 +570,9 @@ static struct file_system_type btrfs_fs_type = {
        .fs_flags       = FS_REQUIRES_DEV,
 };
 
+/*
+ * used by btrfsctl to scan devices when no FS is mounted
+ */
 static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
                                unsigned long arg)
 {
@@ -507,13 +601,15 @@ out:
 static void btrfs_write_super_lockfs(struct super_block *sb)
 {
        struct btrfs_root *root = btrfs_sb(sb);
-       btrfs_transaction_flush_work(root);
+       mutex_lock(&root->fs_info->transaction_kthread_mutex);
+       mutex_lock(&root->fs_info->cleaner_mutex);
 }
 
 static void btrfs_unlockfs(struct super_block *sb)
 {
        struct btrfs_root *root = btrfs_sb(sb);
-       btrfs_transaction_queue_work(root, HZ * 30);
+       mutex_unlock(&root->fs_info->cleaner_mutex);
+       mutex_unlock(&root->fs_info->transaction_kthread_mutex);
 }
 
 static struct super_operations btrfs_super_ops = {
@@ -521,16 +617,13 @@ static struct super_operations btrfs_super_ops = {
        .put_super      = btrfs_put_super,
        .write_super    = btrfs_write_super,
        .sync_fs        = btrfs_sync_fs,
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)
-       .read_inode     = btrfs_read_locked_inode,
-#else
        .show_options   = generic_show_options,
-#endif
        .write_inode    = btrfs_write_inode,
        .dirty_inode    = btrfs_dirty_inode,
        .alloc_inode    = btrfs_alloc_inode,
        .destroy_inode  = btrfs_destroy_inode,
        .statfs         = btrfs_statfs,
+       .remount_fs     = btrfs_remount,
        .write_super_lockfs = btrfs_write_super_lockfs,
        .unlockfs       = btrfs_unlockfs,
 };
@@ -566,10 +659,9 @@ static int __init init_btrfs_fs(void)
        if (err)
                return err;
 
-       btrfs_init_transaction_sys();
        err = btrfs_init_cachep();
        if (err)
-               goto free_transaction_sys;
+               goto free_sysfs;
 
        err = extent_io_init();
        if (err)
@@ -582,9 +674,12 @@ static int __init init_btrfs_fs(void)
        err = btrfs_interface_init();
        if (err)
                goto free_extent_map;
+
        err = register_filesystem(&btrfs_fs_type);
        if (err)
                goto unregister_ioctl;
+
+       printk(KERN_INFO "%s loaded\n", BTRFS_BUILD_VERSION);
        return 0;
 
 unregister_ioctl:
@@ -595,15 +690,13 @@ free_extent_io:
        extent_io_exit();
 free_cachep:
        btrfs_destroy_cachep();
-free_transaction_sys:
-       btrfs_exit_transaction_sys();
+free_sysfs:
        btrfs_exit_sysfs();
        return err;
 }
 
 static void __exit exit_btrfs_fs(void)
 {
-       btrfs_exit_transaction_sys();
        btrfs_destroy_cachep();
        extent_map_exit();
        extent_io_exit();
@@ -611,6 +704,7 @@ static void __exit exit_btrfs_fs(void)
        unregister_filesystem(&btrfs_fs_type);
        btrfs_exit_sysfs();
        btrfs_cleanup_fs_uuids();
+       btrfs_zlib_exit();
 }
 
 module_init(init_btrfs_fs)