Btrfs: Fix infinite loop in btrfs_extent_post_op
[safe/jmp/linux-2.6] / fs / btrfs / super.c
index 196d0e2..4c0b756 100644 (file)
@@ -37,6 +37,7 @@
 #include <linux/ctype.h>
 #include <linux/namei.h>
 #include <linux/miscdevice.h>
+#include "compat.h"
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
 #include "print-tree.h"
 #include "xattr.h"
 #include "volumes.h"
+#include "version.h"
+#include "export.h"
+#include "compression.h"
 
 #define BTRFS_SUPER_MAGIC 0x9123683E
 
 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_err,
+       Opt_ssd, Opt_thread_pool, Opt_noacl,  Opt_compress, Opt_err,
 };
 
 static match_table_t tokens = {
@@ -81,8 +80,10 @@ static match_table_t tokens = {
        {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)
@@ -160,6 +161,10 @@ 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);
@@ -214,6 +219,9 @@ 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;
                }
@@ -228,7 +236,7 @@ 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,
+static int btrfs_parse_early_options(const char *options, fmode_t flags,
                void *holder, char **subvol_name,
                struct btrfs_fs_devices **fs_devices)
 {
@@ -273,23 +281,23 @@ 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 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;
@@ -298,8 +306,10 @@ 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);
 
@@ -309,12 +319,12 @@ 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);
+       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;
+       bi->root = tree_root->fs_info->fs_root;
 
        btrfs_set_key_type(&bi->location, BTRFS_INODE_ITEM_KEY);
 
@@ -333,19 +343,16 @@ static int btrfs_fill_super(struct super_block * sb,
                err = -ENOMEM;
                goto fail_close;
        }
-
+#if 0
        /* this does the super kobj at the same time */
        err = btrfs_sysfs_add_super(tree_root->fs_info);
        if (err)
                goto fail_close;
+#endif
 
        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:
@@ -360,18 +367,22 @@ 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;
 }
 
@@ -402,25 +413,32 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
        struct super_block *s;
        struct dentry *root;
        struct btrfs_fs_devices *fs_devices = NULL;
+       fmode_t mode = FMODE_READ;
        int error = 0;
 
-       error = btrfs_parse_early_options(data, flags, fs_type,
+       if (!(flags & MS_RDONLY))
+               mode |= FMODE_WRITE;
+
+       error = btrfs_parse_early_options(data, mode, fs_type,
                                          &subvol_name, &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;
-       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;
 
@@ -429,9 +447,10 @@ 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];
 
@@ -442,26 +461,34 @@ static int btrfs_get_sb(struct file_system_type *fs_type, int flags,
                if (error) {
                        up_write(&s->s_umount);
                        deactivate_super(s);
-                       goto error;
+                       goto error_free_subvol_name;
                }
 
                btrfs_sb(s)->fs_info->bdev_holder = fs_type;
                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_free_subvol_name;
+               }
+               if (!root->d_inode) {
+                       dput(root);
+                       up_write(&s->s_umount);
+                       deactivate_super(s);
+                       error = -ENXIO;
+                       goto error_free_subvol_name;
+               }
        }
 
        mnt->mnt_sb = s;
@@ -472,19 +499,51 @@ 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);
-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;
@@ -493,6 +552,16 @@ 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;
 }
 
@@ -504,23 +573,31 @@ 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 = kmalloc(sizeof(*vol), GFP_KERNEL);
+       if (!vol)
+               return -ENOMEM;
+
        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;
        }
@@ -532,13 +609,15 @@ 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 = {
@@ -546,16 +625,13 @@ 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,
 };
@@ -577,10 +653,10 @@ 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)
@@ -591,10 +667,9 @@ 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)
@@ -607,9 +682,12 @@ 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:
@@ -620,15 +698,13 @@ 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();
@@ -636,6 +712,7 @@ 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)