Btrfs: using rcu lock in the reader side of devices list

fs_devices->devices is only updated on remove and add device paths, so we can
use rcu to protect it in the reader side

Signed-off-by: Xiao Guangrong <xiaoguangrong@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 4e53a4f..deba3d9 100644 (file)
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -1410,8 +1410,8 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
        struct btrfs_device *device;
        struct backing_dev_info *bdi;
 
-       mutex_lock(&info->fs_devices->device_list_mutex);
-       list_for_each_entry(device, &info->fs_devices->devices, dev_list) {
+       rcu_read_lock();
+       list_for_each_entry_rcu(device, &info->fs_devices->devices, dev_list) {
                if (!device->bdev)
                        continue;
                bdi = blk_get_backing_dev_info(device->bdev);
@@ -1420,7 +1420,7 @@ static int btrfs_congested_fn(void *congested_data, int bdi_bits)
                        break;
                }
        }
-       mutex_unlock(&info->fs_devices->device_list_mutex);
+       rcu_read_unlock();
        return ret;
 }
 
@@ -2332,9 +2332,9 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
        sb = &root->fs_info->super_for_commit;
        dev_item = &sb->dev_item;
 
-       mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
+       rcu_read_lock();
        head = &root->fs_info->fs_devices->devices;
-       list_for_each_entry(dev, head, dev_list) {
+       list_for_each_entry_rcu(dev, head, dev_list) {
                if (!dev->bdev) {
                        total_errors++;
                        continue;
@@ -2367,7 +2367,7 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
        }
 
        total_errors = 0;
-       list_for_each_entry(dev, head, dev_list) {
+       list_for_each_entry_rcu(dev, head, dev_list) {
                if (!dev->bdev)
                        continue;
                if (!dev->in_fs_metadata || !dev->writeable)
@@ -2377,7 +2377,7 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
                if (ret)
                        total_errors++;
        }
-       mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+       rcu_read_unlock();
        if (total_errors > max_errors) {
                printk(KERN_ERR "btrfs: %d errors while writing supers\n",
                       total_errors);

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index ce773fb..0de71fe 100644 (file)
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -281,8 +281,9 @@ static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
        if (!capable(CAP_SYS_ADMIN))
                return -EPERM;
 
-       mutex_lock(&fs_info->fs_devices->device_list_mutex);
-       list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
+       rcu_read_lock();
+       list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
+                               dev_list) {
                if (!device->bdev)
                        continue;
                q = bdev_get_queue(device->bdev);
@@ -292,7 +293,7 @@ static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
                                     minlen);
                }
        }
-       mutex_unlock(&fs_info->fs_devices->device_list_mutex);
+       rcu_read_unlock();
        if (!num_devices)
                return -EOPNOTSUPP;
 

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 0b5ca27..e7844f8 100644 (file)
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -363,7 +363,7 @@ static noinline int device_list_add(const char *path,
                INIT_LIST_HEAD(&device->dev_alloc_list);
 
                mutex_lock(&fs_devices->device_list_mutex);
-               list_add(&device->dev_list, &fs_devices->devices);
+               list_add_rcu(&device->dev_list, &fs_devices->devices);
                mutex_unlock(&fs_devices->device_list_mutex);
 
                device->fs_devices = fs_devices;
@@ -471,6 +471,29 @@ again:
        return 0;
 }
 
+static void __free_device(struct work_struct *work)
+{
+       struct btrfs_device *device;
+
+       device = container_of(work, struct btrfs_device, rcu_work);
+
+       if (device->bdev)
+               blkdev_put(device->bdev, device->mode);
+
+       kfree(device->name);
+       kfree(device);
+}
+
+static void free_device(struct rcu_head *head)
+{
+       struct btrfs_device *device;
+
+       device = container_of(head, struct btrfs_device, rcu);
+
+       INIT_WORK(&device->rcu_work, __free_device);
+       schedule_work(&device->rcu_work);
+}
+
 static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 {
        struct btrfs_device *device;
@@ -480,18 +503,27 @@ static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
 
        mutex_lock(&fs_devices->device_list_mutex);
        list_for_each_entry(device, &fs_devices->devices, dev_list) {
-               if (device->bdev) {
-                       blkdev_put(device->bdev, device->mode);
+               struct btrfs_device *new_device;
+
+               if (device->bdev)
                        fs_devices->open_devices--;
-               }
+
                if (device->writeable) {
                        list_del_init(&device->dev_alloc_list);
                        fs_devices->rw_devices--;
                }
 
-               device->bdev = NULL;
-               device->writeable = 0;
-               device->in_fs_metadata = 0;
+               new_device = kmalloc(sizeof(*new_device), GFP_NOFS);
+               BUG_ON(!new_device);
+               memcpy(new_device, device, sizeof(*new_device));
+               new_device->name = kstrdup(device->name, GFP_NOFS);
+               BUG_ON(!new_device->name);
+               new_device->bdev = NULL;
+               new_device->writeable = 0;
+               new_device->in_fs_metadata = 0;
+               list_replace_rcu(&device->dev_list, &new_device->dev_list);
+
+               call_rcu(&device->rcu, free_device);
        }
        mutex_unlock(&fs_devices->device_list_mutex);
 
@@ -1204,11 +1236,13 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
        struct block_device *bdev;
        struct buffer_head *bh = NULL;
        struct btrfs_super_block *disk_super;
+       struct btrfs_fs_devices *cur_devices;
        u64 all_avail;
        u64 devid;
        u64 num_devices;
        u8 *dev_uuid;
        int ret = 0;
+       bool clear_super = false;
 
        mutex_lock(&uuid_mutex);
        mutex_lock(&root->fs_info->volume_mutex);
@@ -1294,6 +1328,7 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
                list_del_init(&device->dev_alloc_list);
                unlock_chunks(root);
                root->fs_info->fs_devices->rw_devices--;
+               clear_super = true;
        }
 
        ret = btrfs_shrink_device(device, 0);
@@ -1304,16 +1339,15 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
        if (ret)
                goto error_undo;
 
-       device->in_fs_metadata = 0;
-
        /*
         * the device list mutex makes sure that we don't change
         * the device list while someone else is writing out all
         * the device supers.
         */
+
+       cur_devices = device->fs_devices;
        mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
-       list_del_init(&device->dev_list);
-       mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
+       list_del_rcu(&device->dev_list);
 
        device->fs_devices->num_devices--;
 
@@ -1327,36 +1361,36 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
        if (device->bdev == root->fs_info->fs_devices->latest_bdev)
                root->fs_info->fs_devices->latest_bdev = next_device->bdev;
 
-       if (device->bdev) {
-               blkdev_put(device->bdev, device->mode);
-               device->bdev = NULL;
+       if (device->bdev)
                device->fs_devices->open_devices--;
-       }
+
+       call_rcu(&device->rcu, free_device);
+       mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
        num_devices = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
        btrfs_set_super_num_devices(&root->fs_info->super_copy, num_devices);
 
-       if (device->fs_devices->open_devices == 0) {
+       if (cur_devices->open_devices == 0) {
                struct btrfs_fs_devices *fs_devices;
                fs_devices = root->fs_info->fs_devices;
                while (fs_devices) {
-                       if (fs_devices->seed == device->fs_devices)
+                       if (fs_devices->seed == cur_devices)
                                break;
                        fs_devices = fs_devices->seed;
                }
-               fs_devices->seed = device->fs_devices->seed;
-               device->fs_devices->seed = NULL;
+               fs_devices->seed = cur_devices->seed;
+               cur_devices->seed = NULL;
                lock_chunks(root);
-               __btrfs_close_devices(device->fs_devices);
+               __btrfs_close_devices(cur_devices);
                unlock_chunks(root);
-               free_fs_devices(device->fs_devices);
+               free_fs_devices(cur_devices);
        }
 
        /*
         * at this point, the device is zero sized.  We want to
         * remove it from the devices list and zero out the old super
         */
-       if (device->writeable) {
+       if (clear_super) {
                /* make sure this device isn't detected as part of
                 * the FS anymore
                 */
@@ -1365,8 +1399,6 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
                sync_dirty_buffer(bh);
        }
 
-       kfree(device->name);
-       kfree(device);
        ret = 0;
 
 error_brelse:
@@ -1425,7 +1457,8 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
        mutex_init(&seed_devices->device_list_mutex);
 
        mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
-       list_splice_init(&fs_devices->devices, &seed_devices->devices);
+       list_splice_init_rcu(&fs_devices->devices, &seed_devices->devices,
+                             synchronize_rcu);
        mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
        list_splice_init(&fs_devices->alloc_list, &seed_devices->alloc_list);
@@ -1624,7 +1657,7 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
         * half setup
         */
        mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
-       list_add(&device->dev_list, &root->fs_info->fs_devices->devices);
+       list_add_rcu(&device->dev_list, &root->fs_info->fs_devices->devices);
        list_add(&device->dev_alloc_list,
                 &root->fs_info->fs_devices->alloc_list);
        root->fs_info->fs_devices->num_devices++;

diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index cc2eada..f1b2e4f 100644 (file)
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -86,6 +86,8 @@ struct btrfs_device {
        u8 uuid[BTRFS_UUID_SIZE];
 
        struct btrfs_work work;
+       struct rcu_head rcu;
+       struct work_struct rcu_work;
 };
 
 struct btrfs_fs_devices {