Merge branch 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6
[safe/jmp/linux-2.6] / fs / btrfs / super.c
index 6446ab7..d34b2df 100644 (file)
@@ -24,8 +24,8 @@
 #include <linux/highmem.h>
 #include <linux/time.h>
 #include <linux/init.h>
+#include <linux/seq_file.h>
 #include <linux/string.h>
-#include <linux/smp_lock.h>
 #include <linux/backing-dev.h>
 #include <linux/mount.h>
 #include <linux/mpage.h>
@@ -37,6 +37,9 @@
 #include <linux/ctype.h>
 #include <linux/namei.h>
 #include <linux/miscdevice.h>
+#include <linux/magic.h>
+#include <linux/slab.h>
+#include "compat.h"
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "volumes.h"
 #include "version.h"
 #include "export.h"
+#include "compression.h"
 
-#define BTRFS_SUPER_MAGIC 0x9123683E
+static const struct super_operations btrfs_super_ops;
 
-static struct super_operations btrfs_super_ops;
-
-static void btrfs_put_super (struct super_block * sb)
+static void btrfs_put_super(struct super_block *sb)
 {
        struct btrfs_root *root = btrfs_sb(sb);
-       struct btrfs_fs_info *fs = root->fs_info;
        int ret;
 
        ret = close_ctree(root);
-       if (ret) {
-               printk("close ctree returns %d\n", ret);
-       }
-       btrfs_sysfs_del_super(fs);
        sb->s_fs_info = NULL;
 }
 
 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_err,
+       Opt_degraded, Opt_subvol, Opt_subvolid, Opt_device, Opt_nodatasum,
+       Opt_nodatacow, Opt_max_inline, Opt_alloc_start, Opt_nobarrier, Opt_ssd,
+       Opt_nossd, Opt_ssd_spread, Opt_thread_pool, Opt_noacl, Opt_compress,
+       Opt_compress_force, Opt_notreelog, Opt_ratio, Opt_flushoncommit,
+       Opt_discard, Opt_err,
 };
 
 static match_table_t tokens = {
        {Opt_degraded, "degraded"},
        {Opt_subvol, "subvol=%s"},
+       {Opt_subvolid, "subvolid=%d"},
        {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_compress_force, "compress-force"},
        {Opt_ssd, "ssd"},
+       {Opt_ssd_spread, "ssd_spread"},
+       {Opt_nossd, "nossd"},
        {Opt_noacl, "noacl"},
+       {Opt_notreelog, "notreelog"},
+       {Opt_flushoncommit, "flushoncommit"},
+       {Opt_ratio, "metadata_ratio=%d"},
+       {Opt_discard, "discard"},
        {Opt_err, NULL},
 };
 
-u64 btrfs_parse_size(char *str)
-{
-       u64 res;
-       int mult = 1;
-       char *end;
-       char last;
-
-       res = simple_strtoul(str, &end, 10);
-
-       last = end[0];
-       if (isalpha(last)) {
-               last = tolower(last);
-               switch (last) {
-               case 'g':
-                       mult *= 1024;
-               case 'm':
-                       mult *= 1024;
-               case 'k':
-                       mult *= 1024;
-               }
-               res = res * mult;
-       }
-       return res;
-}
-
 /*
  * Regular mount options parser.  Everything that is needed only when
  * reading in a new superblock is parsed here.
@@ -121,8 +103,9 @@ 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;
+       char *p, *num, *orig;
        int intarg;
+       int ret = 0;
 
        if (!options)
                return 0;
@@ -135,6 +118,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
        if (!options)
                return -ENOMEM;
 
+       orig = options;
 
        while ((p = strsep(&options, ",")) != NULL) {
                int token;
@@ -148,6 +132,7 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                        btrfs_set_opt(info->mount_opt, DEGRADED);
                        break;
                case Opt_subvol:
+               case Opt_subvolid:
                case Opt_device:
                        /*
                         * These are parsed by btrfs_parse_early_options
@@ -155,7 +140,7 @@ 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:
@@ -163,10 +148,32 @@ 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_compress_force:
+                       printk(KERN_INFO "btrfs: forcing compression\n");
+                       btrfs_set_opt(info->mount_opt, FORCE_COMPRESS);
+                       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);
                        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");
+                       btrfs_set_opt(info->mount_opt, NOSSD);
+                       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");
                        btrfs_set_opt(info->mount_opt, NOBARRIER);
@@ -180,22 +187,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                                       info->thread_pool_size);
                        }
                        break;
-               case Opt_max_extent:
-                       num = match_strdup(&args[0]);
-                       if (num) {
-                               info->max_extent = btrfs_parse_size(num);
-                               kfree(num);
-
-                               info->max_extent = max_t(u64,
-                                       info->max_extent, root->sectorsize);
-                               printk(KERN_INFO "btrfs: max_extent at %llu\n",
-                                      info->max_extent);
-                       }
-                       break;
                case Opt_max_inline:
                        num = match_strdup(&args[0]);
                        if (num) {
-                               info->max_inline = btrfs_parse_size(num);
+                               info->max_inline = memparse(num, NULL);
                                kfree(num);
 
                                if (info->max_inline) {
@@ -204,28 +199,54 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
                                                root->sectorsize);
                                }
                                printk(KERN_INFO "btrfs: max_inline at %llu\n",
-                                       info->max_inline);
+                                       (unsigned long long)info->max_inline);
                        }
                        break;
                case Opt_alloc_start:
                        num = match_strdup(&args[0]);
                        if (num) {
-                               info->alloc_start = btrfs_parse_size(num);
+                               info->alloc_start = memparse(num, NULL);
                                kfree(num);
                                printk(KERN_INFO
                                        "btrfs: allocations start at %llu\n",
-                                       info->alloc_start);
+                                       (unsigned long long)info->alloc_start);
                        }
                        break;
                case Opt_noacl:
                        root->fs_info->sb->s_flags &= ~MS_POSIXACL;
                        break;
+               case Opt_notreelog:
+                       printk(KERN_INFO "btrfs: disabling tree log\n");
+                       btrfs_set_opt(info->mount_opt, NOTREELOG);
+                       break;
+               case Opt_flushoncommit:
+                       printk(KERN_INFO "btrfs: turning on flush-on-commit\n");
+                       btrfs_set_opt(info->mount_opt, FLUSHONCOMMIT);
+                       break;
+               case Opt_ratio:
+                       intarg = 0;
+                       match_int(&args[0], &intarg);
+                       if (intarg) {
+                               info->metadata_ratio = intarg;
+                               printk(KERN_INFO "btrfs: metadata ratio %d\n",
+                                      info->metadata_ratio);
+                       }
+                       break;
+               case Opt_discard:
+                       btrfs_set_opt(info->mount_opt, DISCARD);
+                       break;
+               case Opt_err:
+                       printk(KERN_INFO "btrfs: unrecognized mount option "
+                              "'%s'\n", p);
+                       ret = -EINVAL;
+                       goto out;
                default:
                        break;
                }
        }
-       kfree(options);
-       return 0;
+out:
+       kfree(orig);
+       return ret;
 }
 
 /*
@@ -234,13 +255,14 @@ 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, int flags,
-               void *holder, char **subvol_name,
+static int btrfs_parse_early_options(const char *options, fmode_t flags,
+               void *holder, char **subvol_name, u64 *subvol_objectid,
                struct btrfs_fs_devices **fs_devices)
 {
        substring_t args[MAX_OPT_ARGS];
        char *opts, *p;
        int error = 0;
+       int intarg;
 
        if (!options)
                goto out;
@@ -263,6 +285,18 @@ static int btrfs_parse_early_options(const char *options, int flags,
                case Opt_subvol:
                        *subvol_name = match_strdup(&args[0]);
                        break;
+               case Opt_subvolid:
+                       intarg = 0;
+                       error = match_int(&args[0], &intarg);
+                       if (!error) {
+                               /* we want the original fs_tree */
+                               if (!intarg)
+                                       *subvol_objectid =
+                                               BTRFS_FS_TREE_OBJECTID;
+                               else
+                                       *subvol_objectid = intarg;
+                       }
+                       break;
                case Opt_device:
                        error = btrfs_scan_one_device(match_strdup(&args[0]),
                                        flags, holder, fs_devices);
@@ -279,26 +313,130 @@ static int btrfs_parse_early_options(const char *options, int flags,
  out:
        /*
         * If no subvolume name is specified we use the default one.  Allocate
-        * a copy of the string "default" here so that code later in the
+        * a copy of the string "." here so that code later in the
         * mount path doesn't care if it's the default volume or another one.
         */
        if (!*subvol_name) {
-               *subvol_name = kstrdup("default", GFP_KERNEL);
+               *subvol_name = kstrdup(".", GFP_KERNEL);
                if (!*subvol_name)
                        return -ENOMEM;
        }
        return error;
 }
 
-static int btrfs_fill_super(struct super_block * sb,
+static struct dentry *get_default_root(struct super_block *sb,
+                                      u64 subvol_objectid)
+{
+       struct btrfs_root *root = sb->s_fs_info;
+       struct btrfs_root *new_root;
+       struct btrfs_dir_item *di;
+       struct btrfs_path *path;
+       struct btrfs_key location;
+       struct inode *inode;
+       struct dentry *dentry;
+       u64 dir_id;
+       int new = 0;
+
+       /*
+        * We have a specific subvol we want to mount, just setup location and
+        * go look up the root.
+        */
+       if (subvol_objectid) {
+               location.objectid = subvol_objectid;
+               location.type = BTRFS_ROOT_ITEM_KEY;
+               location.offset = (u64)-1;
+               goto find_root;
+       }
+
+       path = btrfs_alloc_path();
+       if (!path)
+               return ERR_PTR(-ENOMEM);
+       path->leave_spinning = 1;
+
+       /*
+        * Find the "default" dir item which points to the root item that we
+        * will mount by default if we haven't been given a specific subvolume
+        * to mount.
+        */
+       dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
+       di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
+       if (!di) {
+               /*
+                * Ok the default dir item isn't there.  This is weird since
+                * it's always been there, but don't freak out, just try and
+                * mount to root most subvolume.
+                */
+               btrfs_free_path(path);
+               dir_id = BTRFS_FIRST_FREE_OBJECTID;
+               new_root = root->fs_info->fs_root;
+               goto setup_root;
+       }
+
+       btrfs_dir_item_key_to_cpu(path->nodes[0], di, &location);
+       btrfs_free_path(path);
+
+find_root:
+       new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
+       if (IS_ERR(new_root))
+               return ERR_PTR(PTR_ERR(new_root));
+
+       if (btrfs_root_refs(&new_root->root_item) == 0)
+               return ERR_PTR(-ENOENT);
+
+       dir_id = btrfs_root_dirid(&new_root->root_item);
+setup_root:
+       location.objectid = dir_id;
+       location.type = BTRFS_INODE_ITEM_KEY;
+       location.offset = 0;
+
+       inode = btrfs_iget(sb, &location, new_root, &new);
+       if (!inode)
+               return ERR_PTR(-ENOMEM);
+
+       /*
+        * If we're just mounting the root most subvol put the inode and return
+        * a reference to the dentry.  We will have already gotten a reference
+        * to the inode in btrfs_fill_super so we're good to go.
+        */
+       if (!new && sb->s_root->d_inode == inode) {
+               iput(inode);
+               return dget(sb->s_root);
+       }
+
+       if (new) {
+               const struct qstr name = { .name = "/", .len = 1 };
+
+               /*
+                * New inode, we need to make the dentry a sibling of s_root so
+                * everything gets cleaned up properly on unmount.
+                */
+               dentry = d_alloc(sb->s_root, &name);
+               if (!dentry) {
+                       iput(inode);
+                       return ERR_PTR(-ENOMEM);
+               }
+               d_splice_alias(inode, dentry);
+       } else {
+               /*
+                * We found the inode in cache, just find a dentry for it and
+                * put the reference to the inode we just got.
+                */
+               dentry = d_find_alias(inode);
+               iput(inode);
+       }
+
+       return dentry;
+}
+
+static int btrfs_fill_super(struct super_block *sb,
                            struct btrfs_fs_devices *fs_devices,
-                           void * data, int silent)
+                           void *data, int silent)
 {
-       struct inode * inode;
-       struct dentry * root_dentry;
+       struct inode *inode;
+       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;
@@ -307,7 +445,9 @@ static int btrfs_fill_super(struct super_block * sb,
        sb->s_export_op = &btrfs_export_ops;
        sb->s_xattr = btrfs_xattr_handlers;
        sb->s_time_gran = 1;
+#ifdef CONFIG_BTRFS_FS_POSIX_ACL
        sb->s_flags |= MS_POSIXACL;
+#endif
 
        tree_root = open_ctree(sb, fs_devices, (char *)data);
 
@@ -317,23 +457,15 @@ 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_super_root_dir(disk_super),
-                                 tree_root);
-       bi = BTRFS_I(inode);
-       bi->location.objectid = inode->i_ino;
-       bi->location.offset = 0;
-       bi->root = tree_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, NULL);
+       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) {
@@ -342,17 +474,9 @@ static int btrfs_fill_super(struct super_block * sb,
                goto fail_close;
        }
 
-       /* this does the super kobj at the same time */
-       err = btrfs_sysfs_add_super(tree_root->fs_info);
-       if (err)
-               goto fail_close;
-
        sb->s_root = root_dentry;
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
        save_mount_options(sb, data);
-#endif
-
        return 0;
 
 fail_close:
@@ -363,25 +487,61 @@ fail_close:
 int btrfs_sync_fs(struct super_block *sb, int wait)
 {
        struct btrfs_trans_handle *trans;
-       struct btrfs_root *root;
+       struct btrfs_root *root = btrfs_sb(sb);
        int ret;
-       root = btrfs_sb(sb);
 
-       sb->s_dirt = 0;
        if (!wait) {
                filemap_flush(root->fs_info->btree_inode->i_mapping);
                return 0;
        }
-       btrfs_clean_old_snapshots(root);
-       trans = btrfs_start_transaction(root, 1);
+
+       btrfs_start_delalloc_inodes(root, 0);
+       btrfs_wait_ordered_extents(root, 0, 0);
+
+       trans = btrfs_start_transaction(root, 0);
        ret = btrfs_commit_transaction(trans, root);
-       sb->s_dirt = 0;
        return ret;
 }
 
-static void btrfs_write_super(struct super_block *sb)
+static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 {
-       sb->s_dirt = 0;
+       struct btrfs_root *root = btrfs_sb(vfs->mnt_sb);
+       struct btrfs_fs_info *info = root->fs_info;
+
+       if (btrfs_test_opt(root, DEGRADED))
+               seq_puts(seq, ",degraded");
+       if (btrfs_test_opt(root, NODATASUM))
+               seq_puts(seq, ",nodatasum");
+       if (btrfs_test_opt(root, NODATACOW))
+               seq_puts(seq, ",nodatacow");
+       if (btrfs_test_opt(root, NOBARRIER))
+               seq_puts(seq, ",nobarrier");
+       if (info->max_inline != 8192 * 1024)
+               seq_printf(seq, ",max_inline=%llu",
+                          (unsigned long long)info->max_inline);
+       if (info->alloc_start != 0)
+               seq_printf(seq, ",alloc_start=%llu",
+                          (unsigned long long)info->alloc_start);
+       if (info->thread_pool_size !=  min_t(unsigned long,
+                                            num_online_cpus() + 2, 8))
+               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))
+               seq_puts(seq, ",ssd");
+       if (btrfs_test_opt(root, NOTREELOG))
+               seq_puts(seq, ",notreelog");
+       if (btrfs_test_opt(root, FLUSHONCOMMIT))
+               seq_puts(seq, ",flushoncommit");
+       if (btrfs_test_opt(root, DISCARD))
+               seq_puts(seq, ",discard");
+       if (!(root->fs_info->sb->s_flags & MS_POSIXACL))
+               seq_puts(seq, ",noacl");
+       return 0;
 }
 
 static int btrfs_test_super(struct super_block *s, void *data)
@@ -401,26 +561,38 @@ static int btrfs_test_super(struct super_block *s, void *data)
 static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
                const char *dev_name, void *data, struct vfsmount *mnt)
 {
-       char *subvol_name = NULL;
        struct block_device *bdev = NULL;
        struct super_block *s;
        struct dentry *root;
        struct btrfs_fs_devices *fs_devices = NULL;
+       fmode_t mode = FMODE_READ;
+       char *subvol_name = NULL;
+       u64 subvol_objectid = 0;
        int error = 0;
+       int found = 0;
+
+       if (!(flags & MS_RDONLY))
+               mode |= FMODE_WRITE;
 
-       error = btrfs_parse_early_options(data, flags, fs_type,
-                                         &subvol_name, &fs_devices);
+       error = btrfs_parse_early_options(data, mode, fs_type,
+                                         &subvol_name, &subvol_objectid,
+                                         &fs_devices);
        if (error)
-               goto error;
+               return error;
 
-       error = btrfs_scan_one_device(dev_name, flags, fs_type, &fs_devices);
+       error = btrfs_scan_one_device(dev_name, mode, fs_type, &fs_devices);
        if (error)
                goto error_free_subvol_name;
 
-       error = btrfs_open_devices(fs_devices, flags, fs_type);
+       error = btrfs_open_devices(fs_devices, mode, fs_type);
        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;
        s = sget(fs_type, btrfs_test_super, set_anon_super, fs_devices);
        if (IS_ERR(s))
@@ -428,12 +600,13 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
 
        if (s->s_root) {
                if ((flags ^ s->s_flags) & MS_RDONLY) {
-                       up_write(&s->s_umount);
-                       deactivate_super(s);
+                       deactivate_locked_super(s);
                        error = -EBUSY;
-                       goto error_bdev;
+                       goto error_close_devices;
                }
 
+               found = 1;
+               btrfs_close_devices(fs_devices);
        } else {
                char b[BDEVNAME_SIZE];
 
@@ -442,30 +615,43 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
                error = btrfs_fill_super(s, fs_devices, data,
                                         flags & MS_SILENT ? 1 : 0);
                if (error) {
-                       up_write(&s->s_umount);
-                       deactivate_super(s);
-                       goto error;
+                       deactivate_locked_super(s);
+                       goto error_free_subvol_name;
                }
 
                btrfs_sb(s)->fs_info->bdev_holder = fs_type;
                s->s_flags |= MS_ACTIVE;
        }
 
-       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);
+       root = get_default_root(s, subvol_objectid);
        if (IS_ERR(root)) {
-               up_write(&s->s_umount);
-               deactivate_super(s);
                error = PTR_ERR(root);
+               deactivate_locked_super(s);
                goto error;
        }
-       if (!root->d_inode) {
+       /* if they gave us a subvolume name bind mount into that */
+       if (strcmp(subvol_name, ".")) {
+               struct dentry *new_root;
+               mutex_lock(&root->d_inode->i_mutex);
+               new_root = lookup_one_len(subvol_name, root,
+                                     strlen(subvol_name));
+               mutex_unlock(&root->d_inode->i_mutex);
+
+               if (IS_ERR(new_root)) {
+                       deactivate_locked_super(s);
+                       error = PTR_ERR(new_root);
+                       dput(root);
+                       goto error_close_devices;
+               }
+               if (!new_root->d_inode) {
+                       dput(root);
+                       dput(new_root);
+                       deactivate_locked_super(s);
+                       error = -ENXIO;
+                       goto error_close_devices;
+               }
                dput(root);
-               up_write(&s->s_umount);
-               deactivate_super(s);
-               error = -ENXIO;
-               goto error;
+               root = new_root;
        }
 
        mnt->mnt_sb = s;
@@ -476,7 +662,7 @@ 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);
@@ -484,25 +670,74 @@ error:
        return error;
 }
 
+static int btrfs_remount(struct super_block *sb, int *flags, char *data)
+{
+       struct btrfs_root *root = btrfs_sb(sb);
+       int ret;
+
+       ret = btrfs_parse_options(root, data);
+       if (ret)
+               return -EINVAL;
+
+       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_fs_roots(root->fs_info);
+               WARN_ON(ret);
+
+               /* recover relocation */
+               ret = btrfs_recover_relocation(root);
+               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;
+       struct list_head *head = &root->fs_info->space_info;
+       struct btrfs_space_info *found;
+       u64 total_used = 0;
        int bits = dentry->d_sb->s_blocksize_bits;
        __be32 *fsid = (__be32 *)root->fs_info->fsid;
 
+       rcu_read_lock();
+       list_for_each_entry_rcu(found, head, list)
+               total_used += found->disk_used;
+       rcu_read_unlock();
+
        buf->f_namelen = BTRFS_NAME_LEN;
        buf->f_blocks = btrfs_super_total_bytes(disk_super) >> bits;
-       buf->f_bfree = buf->f_blocks -
-               (btrfs_super_bytes_used(disk_super) >> bits);
+       buf->f_bfree = buf->f_blocks - (total_used >> bits);
        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 
+          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;
 }
 
@@ -514,62 +749,64 @@ 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)
 {
        struct btrfs_ioctl_vol_args *vol;
        struct btrfs_fs_devices *fs_devices;
-       int ret = 0;
-       int len;
+       int ret = -ENOTTY;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       vol = memdup_user((void __user *)arg, sizeof(*vol));
+       if (IS_ERR(vol))
+               return PTR_ERR(vol);
 
-       vol = kmalloc(sizeof(*vol), GFP_KERNEL);
-       if (copy_from_user(vol, (void __user *)arg, sizeof(*vol))) {
-               ret = -EFAULT;
-               goto out;
-       }
-       len = strnlen(vol->name, BTRFS_PATH_NAME_MAX);
        switch (cmd) {
        case BTRFS_IOC_SCAN_DEV:
-               ret = btrfs_scan_one_device(vol->name, MS_RDONLY,
+               ret = btrfs_scan_one_device(vol->name, FMODE_READ,
                                            &btrfs_fs_type, &fs_devices);
                break;
        }
-out:
+
        kfree(vol);
        return ret;
 }
 
-static void btrfs_write_super_lockfs(struct super_block *sb)
+static int btrfs_freeze(struct super_block *sb)
 {
        struct btrfs_root *root = btrfs_sb(sb);
        mutex_lock(&root->fs_info->transaction_kthread_mutex);
        mutex_lock(&root->fs_info->cleaner_mutex);
+       return 0;
 }
 
-static void btrfs_unlockfs(struct super_block *sb)
+static int btrfs_unfreeze(struct super_block *sb)
 {
        struct btrfs_root *root = btrfs_sb(sb);
        mutex_unlock(&root->fs_info->cleaner_mutex);
        mutex_unlock(&root->fs_info->transaction_kthread_mutex);
+       return 0;
 }
 
-static struct super_operations btrfs_super_ops = {
+static const struct super_operations btrfs_super_ops = {
+       .drop_inode     = btrfs_drop_inode,
        .delete_inode   = btrfs_delete_inode,
        .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
+       .show_options   = btrfs_show_options,
        .write_inode    = btrfs_write_inode,
        .dirty_inode    = btrfs_dirty_inode,
        .alloc_inode    = btrfs_alloc_inode,
        .destroy_inode  = btrfs_destroy_inode,
        .statfs         = btrfs_statfs,
-       .write_super_lockfs = btrfs_write_super_lockfs,
-       .unlockfs       = btrfs_unlockfs,
+       .remount_fs     = btrfs_remount,
+       .freeze_fs      = btrfs_freeze,
+       .unfreeze_fs    = btrfs_unfreeze,
 };
 
 static const struct file_operations btrfs_ctl_fops = {
@@ -579,20 +816,23 @@ static const struct file_operations btrfs_ctl_fops = {
 };
 
 static struct miscdevice btrfs_misc = {
-       .minor          = MISC_DYNAMIC_MINOR,
+       .minor          = BTRFS_MINOR,
        .name           = "btrfs-control",
        .fops           = &btrfs_ctl_fops
 };
 
+MODULE_ALIAS_MISCDEV(BTRFS_MINOR);
+MODULE_ALIAS("devname:btrfs-control");
+
 static int btrfs_interface_init(void)
 {
        return misc_register(&btrfs_misc);
 }
 
-void btrfs_interface_exit(void)
+static void btrfs_interface_exit(void)
 {
        if (misc_deregister(&btrfs_misc) < 0)
-               printk("misc_deregister failed for control device");
+               printk(KERN_INFO "misc_deregister failed for control device");
 }
 
 static int __init init_btrfs_fs(void)
@@ -618,6 +858,7 @@ 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;
@@ -647,6 +888,7 @@ 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)