Merge tag 'v6.6-vfs.super' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

Pull superblock updates from Christian Brauner:
 "This contains the super rework that was ready for this cycle. The
  first part changes the order of how we open block devices and allocate
  superblocks, contains various cleanups, simplifications, and a new
  mechanism to wait on superblock state changes.

  This unblocks work to ultimately limit the number of writers to a
  block device. Jan has already scheduled follow-up work that will be
  ready for v6.7 and allows us to restrict the number of writers to a
  given block device. That series builds on this work right here.

  The second part contains filesystem freezing updates.

  Overview:

  The generic superblock changes are rougly organized as follows
  (ignoring additional minor cleanups):

   (1) Removal of the bd_super member from struct block_device.

       This was a very odd back pointer to struct super_block with
       unclear rules. For all relevant places we have other means to get
       the same information so just get rid of this.

   (2) Simplify rules for superblock cleanup.

       Roughly, everything that is allocated during fs_context
       initialization and that's stored in fs_context->s_fs_info needs
       to be cleaned up by the fs_context->free() implementation before
       the superblock allocation function has been called successfully.

       After sget_fc() returned fs_context->s_fs_info has been
       transferred to sb->s_fs_info at which point sb->kill_sb() if
       fully responsible for cleanup. Adhering to these rules means that
       cleanup of sb->s_fs_info in fill_super() is to be avoided as it's
       brittle and inconsistent.

       Cleanup shouldn't be duplicated between sb->put_super() as
       sb->put_super() is only called if sb->s_root has been set aka
       when the filesystem has been successfully born (SB_BORN). That
       complexity should be avoided.

       This also means that block devices are to be closed in
       sb->kill_sb() instead of sb->put_super(). More details in the
       lower section.

   (3) Make it possible to lookup or create a superblock before opening
       block devices

       There's a subtle dependency on (2) as some filesystems did rely
       on fill_super() to be called in order to correctly clean up
       sb->s_fs_info. All these filesystems have been fixed.

   (4) Switch most filesystem to follow the same logic as the generic
       mount code now does as outlined in (3).

   (5) Use the superblock as the holder of the block device. We can now
       easily go back from block device to owning superblock.

   (6) Export and extend the generic fs_holder_ops and use them as
       holder ops everywhere and remove the filesystem specific holder
       ops.

   (7) Call from the block layer up into the filesystem layer when the
       block device is removed, allowing to shut down the filesystem
       without risk of deadlocks.

   (8) Get rid of get_super().

       We can now easily go back from the block device to owning
       superblock and can call up from the block layer into the
       filesystem layer when the device is removed. So no need to wade
       through all registered superblock to find the owning superblock
       anymore"

Link: https://lore.kernel.org/lkml/20230824-prall-intakt-95dbffdee4a0@brauner/

* tag 'v6.6-vfs.super' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (47 commits)
  super: use higher-level helper for {freeze,thaw}
  super: wait until we passed kill super
  super: wait for nascent superblocks
  super: make locking naming consistent
  super: use locking helpers
  fs: simplify invalidate_inodes
  fs: remove get_super
  block: call into the file system for ioctl BLKFLSBUF
  block: call into the file system for bdev_mark_dead
  block: consolidate __invalidate_device and fsync_bdev
  block: drop the "busy inodes on changed media" log message
  dasd: also call __invalidate_device when setting the device offline
  amiflop: don't call fsync_bdev in FDFMTBEG
  floppy: call disk_force_media_change when changing the format
  block: simplify the disk_force_media_change interface
  nbd: call blk_mark_disk_dead in nbd_clear_sock_ioctl
  xfs use fs_holder_ops for the log and RT devices
  xfs: drop s_umount over opening the log and RT devices
  ext4: use fs_holder_ops for the log device
  ext4: drop s_umount over opening the log device
  ...

25 files changed, 935 insertions, 520 deletions

diff --git a/fs/buffer.c b/fs/buffer.c
index 3aa2380a75e1..084a6ade108a 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -1226,19 +1226,14 @@ EXPORT_SYMBOL(mark_buffer_dirty);
 
 void mark_buffer_write_io_error(struct buffer_head *bh)
 {
-	struct super_block *sb;
-
 	set_buffer_write_io_error(bh);
 	/* FIXME: do we need to set this in both places? */
 	if (bh->b_folio && bh->b_folio->mapping)
 		mapping_set_error(bh->b_folio->mapping, -EIO);
-	if (bh->b_assoc_map)
+	if (bh->b_assoc_map) {
 		mapping_set_error(bh->b_assoc_map, -EIO);
-	rcu_read_lock();
-	sb = READ_ONCE(bh->b_bdev->bd_super);
-	if (sb)
-		errseq_set(&sb->s_wb_err, -EIO);
-	rcu_read_unlock();
+		errseq_set(&bh->b_assoc_map->host->i_sb->s_wb_err, -EIO);
+	}
 }
 EXPORT_SYMBOL(mark_buffer_write_io_error);
 
diff --git a/fs/cramfs/inode.c b/fs/cramfs/inode.c
index e755b2223936..5ee7d7bbb361 100644
--- a/fs/cramfs/inode.c
+++ b/fs/cramfs/inode.c
@@ -486,12 +486,16 @@ static void cramfs_kill_sb(struct super_block *sb)
 {
 	struct cramfs_sb_info *sbi = CRAMFS_SB(sb);
 
+	generic_shutdown_super(sb);
+
 	if (IS_ENABLED(CONFIG_CRAMFS_MTD) && sb->s_mtd) {
 		if (sbi && sbi->mtd_point_size)
 			mtd_unpoint(sb->s_mtd, 0, sbi->mtd_point_size);
-		kill_mtd_super(sb);
+		put_mtd_device(sb->s_mtd);
+		sb->s_mtd = NULL;
 	} else if (IS_ENABLED(CONFIG_CRAMFS_BLOCKDEV) && sb->s_bdev) {
-		kill_block_super(sb);
+		sync_blockdev(sb->s_bdev);
+		blkdev_put(sb->s_bdev, sb);
 	}
 	kfree(sbi);
 }
diff --git a/fs/exfat/exfat_fs.h b/fs/exfat/exfat_fs.h
index 729ada9e26e8..f55498e5c23d 100644
--- a/fs/exfat/exfat_fs.h
+++ b/fs/exfat/exfat_fs.h
@@ -273,8 +273,6 @@ struct exfat_sb_info {
 
 	spinlock_t inode_hash_lock;
 	struct hlist_head inode_hashtable[EXFAT_HASH_SIZE];
-
-	struct rcu_head rcu;
 };
 
 #define EXFAT_CACHE_VALID	0
diff --git a/fs/exfat/super.c b/fs/exfat/super.c
index 8e0ca4b0fc8f..2778bd9b631e 100644
--- a/fs/exfat/super.c
+++ b/fs/exfat/super.c
@@ -31,16 +31,6 @@ static void exfat_free_iocharset(struct exfat_sb_info *sbi)
 		kfree(sbi->options.iocharset);
 }
 
-static void exfat_delayed_free(struct rcu_head *p)
-{
-	struct exfat_sb_info *sbi = container_of(p, struct exfat_sb_info, rcu);
-
-	unload_nls(sbi->nls_io);
-	exfat_free_iocharset(sbi);
-	exfat_free_upcase_table(sbi);
-	kfree(sbi);
-}
-
 static void exfat_put_super(struct super_block *sb)
 {
 	struct exfat_sb_info *sbi = EXFAT_SB(sb);
@@ -50,7 +40,8 @@ static void exfat_put_super(struct super_block *sb)
 	brelse(sbi->boot_bh);
 	mutex_unlock(&sbi->s_lock);
 
-	call_rcu(&sbi->rcu, exfat_delayed_free);
+	unload_nls(sbi->nls_io);
+	exfat_free_upcase_table(sbi);
 }
 
 static int exfat_sync_fs(struct super_block *sb, int wait)
@@ -709,9 +700,6 @@ free_table:
 
 check_nls_io:
 	unload_nls(sbi->nls_io);
-	exfat_free_iocharset(sbi);
-	sb->s_fs_info = NULL;
-	kfree(sbi);
 	return err;
 }
 
@@ -720,14 +708,18 @@ static int exfat_get_tree(struct fs_context *fc)
 	return get_tree_bdev(fc, exfat_fill_super);
 }
 
+static void exfat_free_sbi(struct exfat_sb_info *sbi)
+{
+	exfat_free_iocharset(sbi);
+	kfree(sbi);
+}
+
 static void exfat_free(struct fs_context *fc)
 {
 	struct exfat_sb_info *sbi = fc->s_fs_info;
 
-	if (sbi) {
-		exfat_free_iocharset(sbi);
-		kfree(sbi);
-	}
+	if (sbi)
+		exfat_free_sbi(sbi);
 }
 
 static int exfat_reconfigure(struct fs_context *fc)
@@ -772,12 +764,21 @@ static int exfat_init_fs_context(struct fs_context *fc)
 	return 0;
 }
 
+static void exfat_kill_sb(struct super_block *sb)
+{
+	struct exfat_sb_info *sbi = sb->s_fs_info;
+
+	kill_block_super(sb);
+	if (sbi)
+		exfat_free_sbi(sbi);
+}
+
 static struct file_system_type exfat_fs_type = {
 	.owner			= THIS_MODULE,
 	.name			= "exfat",
 	.init_fs_context	= exfat_init_fs_context,
 	.parameters		= exfat_parameters,
-	.kill_sb		= kill_block_super,
+	.kill_sb		= exfat_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV,
 };
 
diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c
index 77f318ec8abb..b38d59581411 100644
--- a/fs/ext4/ext4_jbd2.c
+++ b/fs/ext4/ext4_jbd2.c
@@ -234,8 +234,7 @@ int __ext4_journal_get_write_access(const char *where, unsigned int line,
 
 	might_sleep();
 
-	if (bh->b_bdev->bd_super)
-		ext4_check_bdev_write_error(bh->b_bdev->bd_super);
+	ext4_check_bdev_write_error(sb);
 
 	if (ext4_handle_valid(handle)) {
 		err = jbd2_journal_get_write_access(handle, bh);
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index cb1ff47af156..73547d2334fd 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -93,6 +93,7 @@ static int ext4_get_tree(struct fs_context *fc);
 static int ext4_reconfigure(struct fs_context *fc);
 static void ext4_fc_free(struct fs_context *fc);
 static int ext4_init_fs_context(struct fs_context *fc);
+static void ext4_kill_sb(struct super_block *sb);
 static const struct fs_parameter_spec ext4_param_specs[];
 
 /*
@@ -135,12 +136,12 @@ static struct file_system_type ext2_fs_type = {
 	.name			= "ext2",
 	.init_fs_context	= ext4_init_fs_context,
 	.parameters		= ext4_param_specs,
-	.kill_sb		= kill_block_super,
+	.kill_sb		= ext4_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV,
 };
 MODULE_ALIAS_FS("ext2");
 MODULE_ALIAS("ext2");
-#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
+#define IS_EXT2_SB(sb) ((sb)->s_type == &ext2_fs_type)
 #else
 #define IS_EXT2_SB(sb) (0)
 #endif
@@ -151,12 +152,12 @@ static struct file_system_type ext3_fs_type = {
 	.name			= "ext3",
 	.init_fs_context	= ext4_init_fs_context,
 	.parameters		= ext4_param_specs,
-	.kill_sb		= kill_block_super,
+	.kill_sb		= ext4_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV,
 };
 MODULE_ALIAS_FS("ext3");
 MODULE_ALIAS("ext3");
-#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
+#define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type)
 
 
 static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
@@ -1096,15 +1097,6 @@ void ext4_update_dynamic_rev(struct super_block *sb)
 	 */
 }
 
-static void ext4_bdev_mark_dead(struct block_device *bdev)
-{
-	ext4_force_shutdown(bdev->bd_holder, EXT4_GOING_FLAGS_NOLOGFLUSH);
-}
-
-static const struct blk_holder_ops ext4_holder_ops = {
-	.mark_dead		= ext4_bdev_mark_dead,
-};
-
 /*
  * Open the external journal device
  */
@@ -1113,7 +1105,7 @@ static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
 	struct block_device *bdev;
 
 	bdev = blkdev_get_by_dev(dev, BLK_OPEN_READ | BLK_OPEN_WRITE, sb,
-				 &ext4_holder_ops);
+				 &fs_holder_ops);
 	if (IS_ERR(bdev))
 		goto fail;
 	return bdev;
@@ -1125,25 +1117,6 @@ fail:
 	return NULL;
 }
 
-/*
- * Release the journal device
- */
-static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
-{
-	struct block_device *bdev;
-	bdev = sbi->s_journal_bdev;
-	if (bdev) {
-		/*
-		 * Invalidate the journal device's buffers.  We don't want them
-		 * floating about in memory - the physical journal device may
-		 * hotswapped, and it breaks the `ro-after' testing code.
-		 */
-		invalidate_bdev(bdev);
-		blkdev_put(bdev, sbi->s_sb);
-		sbi->s_journal_bdev = NULL;
-	}
-}
-
 static inline struct inode *orphan_list_entry(struct list_head *l)
 {
 	return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
@@ -1339,8 +1312,13 @@ static void ext4_put_super(struct super_block *sb)
 	sync_blockdev(sb->s_bdev);
 	invalidate_bdev(sb->s_bdev);
 	if (sbi->s_journal_bdev) {
+		/*
+		 * Invalidate the journal device's buffers.  We don't want them
+		 * floating about in memory - the physical journal device may
+		 * hotswapped, and it breaks the `ro-after' testing code.
+		 */
 		sync_blockdev(sbi->s_journal_bdev);
-		ext4_blkdev_remove(sbi);
+		invalidate_bdev(sbi->s_journal_bdev);
 	}
 
 	ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
@@ -5572,7 +5550,6 @@ static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
 	spin_lock_init(&sbi->s_bdev_wb_lock);
 	errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
 				 &sbi->s_bdev_wb_err);
-	sb->s_bdev->bd_super = sb;
 	EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
 	ext4_orphan_cleanup(sb, es);
 	EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
@@ -5664,9 +5641,11 @@ failed_mount:
 		kfree(get_qf_name(sb, sbi, i));
 #endif
 	fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
-	/* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
 	brelse(sbi->s_sbh);
-	ext4_blkdev_remove(sbi);
+	if (sbi->s_journal_bdev) {
+		invalidate_bdev(sbi->s_journal_bdev);
+		blkdev_put(sbi->s_journal_bdev, sb);
+	}
 out_fail:
 	invalidate_bdev(sb->s_bdev);
 	sb->s_fs_info = NULL;
@@ -5854,7 +5833,10 @@ static journal_t *ext4_get_dev_journal(struct super_block *sb,
 	if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
 		return NULL;
 
+	/* see get_tree_bdev why this is needed and safe */
+	up_write(&sb->s_umount);
 	bdev = ext4_blkdev_get(j_dev, sb);
+	down_write(&sb->s_umount);
 	if (bdev == NULL)
 		return NULL;
 
@@ -7273,12 +7255,23 @@ static inline int ext3_feature_set_ok(struct super_block *sb)
 	return 1;
 }
 
+static void ext4_kill_sb(struct super_block *sb)
+{
+	struct ext4_sb_info *sbi = EXT4_SB(sb);
+	struct block_device *journal_bdev = sbi ? sbi->s_journal_bdev : NULL;
+
+	kill_block_super(sb);
+
+	if (journal_bdev)
+		blkdev_put(journal_bdev, sb);
+}
+
 static struct file_system_type ext4_fs_type = {
 	.owner			= THIS_MODULE,
 	.name			= "ext4",
 	.init_fs_context	= ext4_init_fs_context,
 	.parameters		= ext4_param_specs,
-	.kill_sb		= kill_block_super,
+	.kill_sb		= ext4_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP | FS_MGTIME,
 };
 MODULE_ALIAS_FS("ext4");
diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 01effd3fcb6c..a1ca394bc327 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -2181,12 +2181,14 @@ out_drop_write:
 	if (err)
 		return err;
 
-	err = freeze_super(sbi->sb);
+	err = freeze_super(sbi->sb, FREEZE_HOLDER_USERSPACE);
 	if (err)
 		return err;
 
 	if (f2fs_readonly(sbi->sb)) {
-		thaw_super(sbi->sb);
+		err = thaw_super(sbi->sb, FREEZE_HOLDER_USERSPACE);
+		if (err)
+			return err;
 		return -EROFS;
 	}
 
@@ -2240,6 +2242,6 @@ recover_out:
 out_err:
 	f2fs_up_write(&sbi->cp_global_sem);
 	f2fs_up_write(&sbi->gc_lock);
-	thaw_super(sbi->sb);
+	thaw_super(sbi->sb, FREEZE_HOLDER_USERSPACE);
 	return err;
 }
diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
index 28aeffc0048c..aa1f9a3a8037 100644
--- a/fs/f2fs/super.c
+++ b/fs/f2fs/super.c
@@ -1561,7 +1561,7 @@ static void destroy_device_list(struct f2fs_sb_info *sbi)
 	int i;
 
 	for (i = 0; i < sbi->s_ndevs; i++) {
-		blkdev_put(FDEV(i).bdev, sbi->sb->s_type);
+		blkdev_put(FDEV(i).bdev, sbi->sb);
 #ifdef CONFIG_BLK_DEV_ZONED
 		kvfree(FDEV(i).blkz_seq);
 #endif
@@ -4198,7 +4198,7 @@ static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
 			/* Single zoned block device mount */
 			FDEV(0).bdev =
 				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, mode,
-						  sbi->sb->s_type, NULL);
+						  sbi->sb, NULL);
 		} else {
 			/* Multi-device mount */
 			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
@@ -4217,8 +4217,7 @@ static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
 					sbi->log_blocks_per_seg) - 1;
 			}
 			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, mode,
-							  sbi->sb->s_type,
-							  NULL);
+							  sbi->sb, NULL);
 		}
 		if (IS_ERR(FDEV(i).bdev))
 			return PTR_ERR(FDEV(i).bdev);
diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index aca4b4811394..969ce991b0b0 100644
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -1953,9 +1953,9 @@ static long __writeback_inodes_wb(struct bdi_writeback *wb,
 		struct inode *inode = wb_inode(wb->b_io.prev);
 		struct super_block *sb = inode->i_sb;
 
-		if (!trylock_super(sb)) {
+		if (!super_trylock_shared(sb)) {
 			/*
-			 * trylock_super() may fail consistently due to
+			 * super_trylock_shared() may fail consistently due to
 			 * s_umount being grabbed by someone else. Don't use
 			 * requeue_io() to avoid busy retrying the inode/sb.
 			 */
diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c
index ec0296b35dfe..2f701335e8ee 100644
--- a/fs/gfs2/super.c
+++ b/fs/gfs2/super.c
@@ -689,7 +689,7 @@ static int gfs2_freeze_locally(struct gfs2_sbd *sdp)
 	struct super_block *sb = sdp->sd_vfs;
 	int error;
 
-	error = freeze_super(sb);
+	error = freeze_super(sb, FREEZE_HOLDER_USERSPACE);
 	if (error)
 		return error;
 
@@ -697,7 +697,9 @@ static int gfs2_freeze_locally(struct gfs2_sbd *sdp)
 		gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_FREEZE |
 			       GFS2_LFC_FREEZE_GO_SYNC);
 		if (gfs2_withdrawn(sdp)) {
-			thaw_super(sb);
+			error = thaw_super(sb, FREEZE_HOLDER_USERSPACE);
+			if (error)
+				return error;
 			return -EIO;
 		}
 	}
@@ -712,7 +714,7 @@ static int gfs2_do_thaw(struct gfs2_sbd *sdp)
 	error = gfs2_freeze_lock_shared(sdp);
 	if (error)
 		goto fail;
-	error = thaw_super(sb);
+	error = thaw_super(sb, FREEZE_HOLDER_USERSPACE);
 	if (!error)
 		return 0;
 
@@ -761,7 +763,7 @@ out:
  *
  */
 
-static int gfs2_freeze_super(struct super_block *sb)
+static int gfs2_freeze_super(struct super_block *sb, enum freeze_holder who)
 {
 	struct gfs2_sbd *sdp = sb->s_fs_info;
 	int error;
@@ -816,7 +818,7 @@ out:
  *
  */
 
-static int gfs2_thaw_super(struct super_block *sb)
+static int gfs2_thaw_super(struct super_block *sb, enum freeze_holder who)
 {
 	struct gfs2_sbd *sdp = sb->s_fs_info;
 	int error;
diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c
index 2dfbe2f188dd..c60bc7f628e1 100644
--- a/fs/gfs2/sys.c
+++ b/fs/gfs2/sys.c
@@ -168,10 +168,10 @@ static ssize_t freeze_store(struct gfs2_sbd *sdp, const char *buf, size_t len)
 
 	switch (n) {
 	case 0:
-		error = thaw_super(sdp->sd_vfs);
+		error = thaw_super(sdp->sd_vfs, FREEZE_HOLDER_USERSPACE);
 		break;
 	case 1:
-		error = freeze_super(sdp->sd_vfs);
+		error = freeze_super(sdp->sd_vfs, FREEZE_HOLDER_USERSPACE);
 		break;
 	default:
 		return -EINVAL;
diff --git a/fs/inode.c b/fs/inode.c
index 0ad4fb4b8fe9..35fd688168c5 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -751,16 +751,11 @@ EXPORT_SYMBOL_GPL(evict_inodes);
 /**
  * invalidate_inodes	- attempt to free all inodes on a superblock
  * @sb:		superblock to operate on
- * @kill_dirty: flag to guide handling of dirty inodes
  *
- * Attempts to free all inodes for a given superblock.  If there were any
- * busy inodes return a non-zero value, else zero.
- * If @kill_dirty is set, discard dirty inodes too, otherwise treat
- * them as busy.
+ * Attempts to free all inodes (including dirty inodes) for a given superblock.
  */
-int invalidate_inodes(struct super_block *sb, bool kill_dirty)
+void invalidate_inodes(struct super_block *sb)
 {
-	int busy = 0;
 	struct inode *inode, *next;
 	LIST_HEAD(dispose);
 
@@ -772,14 +767,8 @@ again:
 			spin_unlock(&inode->i_lock);
 			continue;
 		}
-		if (inode->i_state & I_DIRTY_ALL && !kill_dirty) {
-			spin_unlock(&inode->i_lock);
-			busy = 1;
-			continue;
-		}
 		if (atomic_read(&inode->i_count)) {
 			spin_unlock(&inode->i_lock);
-			busy = 1;
 			continue;
 		}
 
@@ -797,8 +786,6 @@ again:
 	spin_unlock(&sb->s_inode_list_lock);
 
 	dispose_list(&dispose);
-
-	return busy;
 }
 
 /*
diff --git a/fs/internal.h b/fs/internal.h
index f7a3dc111026..74d3b161dd2c 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -115,7 +115,7 @@ static inline void put_file_access(struct file *file)
  * super.c
  */
 extern int reconfigure_super(struct fs_context *);
-extern bool trylock_super(struct super_block *sb);
+extern bool super_trylock_shared(struct super_block *sb);
 struct super_block *user_get_super(dev_t, bool excl);
 void put_super(struct super_block *sb);
 extern bool mount_capable(struct fs_context *);
@@ -201,7 +201,7 @@ void lock_two_inodes(struct inode *inode1, struct inode *inode2,
  * fs-writeback.c
  */
 extern long get_nr_dirty_inodes(void);
-extern int invalidate_inodes(struct super_block *, bool);
+void invalidate_inodes(struct super_block *sb);
 
 /*
  * dcache.c
diff --git a/fs/ioctl.c b/fs/ioctl.c
index d413e0b8f6c2..f5fd99d6b0d4 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -397,8 +397,8 @@ static int ioctl_fsfreeze(struct file *filp)
 
 	/* Freeze */
 	if (sb->s_op->freeze_super)
-		return sb->s_op->freeze_super(sb);
-	return freeze_super(sb);
+		return sb->s_op->freeze_super(sb, FREEZE_HOLDER_USERSPACE);
+	return freeze_super(sb, FREEZE_HOLDER_USERSPACE);
 }
 
 static int ioctl_fsthaw(struct file *filp)
@@ -410,8 +410,8 @@ static int ioctl_fsthaw(struct file *filp)
 
 	/* Thaw */
 	if (sb->s_op->thaw_super)
-		return sb->s_op->thaw_super(sb);
-	return thaw_super(sb);
+		return sb->s_op->thaw_super(sb, FREEZE_HOLDER_USERSPACE);
+	return thaw_super(sb, FREEZE_HOLDER_USERSPACE);
 }
 
 static int ioctl_file_dedupe_range(struct file *file,
diff --git a/fs/nilfs2/super.c b/fs/nilfs2/super.c
index 0ef8c71bde8e..a5d1fa4e7552 100644
--- a/fs/nilfs2/super.c
+++ b/fs/nilfs2/super.c
@@ -35,6 +35,7 @@
 #include <linux/writeback.h>
 #include <linux/seq_file.h>
 #include <linux/mount.h>
+#include <linux/fs_context.h>
 #include "nilfs.h"
 #include "export.h"
 #include "mdt.h"
@@ -1216,7 +1217,6 @@ static int nilfs_remount(struct super_block *sb, int *flags, char *data)
 }
 
 struct nilfs_super_data {
-	struct block_device *bdev;
 	__u64 cno;
 	int flags;
 };
@@ -1283,64 +1283,49 @@ static int nilfs_identify(char *data, struct nilfs_super_data *sd)
 
 static int nilfs_set_bdev_super(struct super_block *s, void *data)
 {
-	s->s_bdev = data;
-	s->s_dev = s->s_bdev->bd_dev;
+	s->s_dev = *(dev_t *)data;
 	return 0;
 }
 
 static int nilfs_test_bdev_super(struct super_block *s, void *data)
 {
-	return (void *)s->s_bdev == data;
+	return !(s->s_iflags & SB_I_RETIRED) && s->s_dev == *(dev_t *)data;
 }
 
 static struct dentry *
 nilfs_mount(struct file_system_type *fs_type, int flags,
 	     const char *dev_name, void *data)
 {
-	struct nilfs_super_data sd;
+	struct nilfs_super_data sd = { .flags = flags };
 	struct super_block *s;
-	struct dentry *root_dentry;
-	int err, s_new = false;
+	dev_t dev;
+	int err;
 
-	sd.bdev = blkdev_get_by_path(dev_name, sb_open_mode(flags), fs_type,
-				     NULL);
-	if (IS_ERR(sd.bdev))
-		return ERR_CAST(sd.bdev);
+	if (nilfs_identify(data, &sd))
+		return ERR_PTR(-EINVAL);
 
-	sd.cno = 0;
-	sd.flags = flags;
-	if (nilfs_identify((char *)data, &sd)) {
-		err = -EINVAL;
-		goto failed;
-	}
+	err = lookup_bdev(dev_name, &dev);
+	if (err)
+		return ERR_PTR(err);
 
-	/*
-	 * once the super is inserted into the list by sget, s_umount
-	 * will protect the lockfs code from trying to start a snapshot
-	 * while we are mounting
-	 */
-	mutex_lock(&sd.bdev->bd_fsfreeze_mutex);
-	if (sd.bdev->bd_fsfreeze_count > 0) {
-		mutex_unlock(&sd.bdev->bd_fsfreeze_mutex);
-		err = -EBUSY;
-		goto failed;
-	}
 	s = sget(fs_type, nilfs_test_bdev_super, nilfs_set_bdev_super, flags,
-		 sd.bdev);
-	mutex_unlock(&sd.bdev->bd_fsfreeze_mutex);
-	if (IS_ERR(s)) {
-		err = PTR_ERR(s);
-		goto failed;
-	}
+		 &dev);
+	if (IS_ERR(s))
+		return ERR_CAST(s);
 
 	if (!s->s_root) {
-		s_new = true;
-
-		/* New superblock instance created */
-		snprintf(s->s_id, sizeof(s->s_id), "%pg", sd.bdev);
-		sb_set_blocksize(s, block_size(sd.bdev));
-
-		err = nilfs_fill_super(s, data, flags & SB_SILENT ? 1 : 0);
+		/*
+		 * We drop s_umount here because we need to open the bdev and
+		 * bdev->open_mutex ranks above s_umount (blkdev_put() ->
+		 * __invalidate_device()). It is safe because we have active sb
+		 * reference and SB_BORN is not set yet.
+		 */
+		up_write(&s->s_umount);
+		err = setup_bdev_super(s, flags, NULL);
+		down_write(&s->s_umount);
+		if (!err)
+			err = nilfs_fill_super(s, data,
+					       flags & SB_SILENT ? 1 : 0);
 		if (err)
 			goto failed_super;
 
@@ -1366,24 +1351,18 @@ nilfs_mount(struct file_system_type *fs_type, int flags,
 	}
 
 	if (sd.cno) {
+		struct dentry *root_dentry;
+
 		err = nilfs_attach_snapshot(s, sd.cno, &root_dentry);
 		if (err)
 			goto failed_super;
-	} else {
-		root_dentry = dget(s->s_root);
+		return root_dentry;
 	}
 
-	if (!s_new)
-		blkdev_put(sd.bdev, fs_type);
-
-	return root_dentry;
+	return dget(s->s_root);
 
  failed_super:
 	deactivate_locked_super(s);
-
- failed:
-	if (!s_new)
-		blkdev_put(sd.bdev, fs_type);
 	return ERR_PTR(err);
 }
 
diff --git a/fs/ntfs3/super.c b/fs/ntfs3/super.c
index 1a02072b6b0e..5fffddea554f 100644
--- a/fs/ntfs3/super.c
+++ b/fs/ntfs3/super.c
@@ -569,9 +569,9 @@ static void init_once(void *foo)
 }
 
 /*
- * put_ntfs - Noinline to reduce binary size.
+ * Noinline to reduce binary size.
  */
-static noinline void put_ntfs(struct ntfs_sb_info *sbi)
+static noinline void ntfs3_free_sbi(struct ntfs_sb_info *sbi)
 {
 	kfree(sbi->new_rec);
 	kvfree(ntfs_put_shared(sbi->upcase));
@@ -625,12 +625,6 @@ static void ntfs_put_super(struct super_block *sb)
 
 	/* Mark rw ntfs as clear, if possible. */
 	ntfs_set_state(sbi, NTFS_DIRTY_CLEAR);
-
-	put_mount_options(sbi->options);
-	put_ntfs(sbi);
-	sb->s_fs_info = NULL;
-
-	sync_blockdev(sb->s_bdev);
 }
 
 static int ntfs_statfs(struct dentry *dentry, struct kstatfs *buf)
@@ -1564,15 +1558,7 @@ load_root:
 put_inode_out:
 	iput(inode);
 out:
-	/*
-	 * Free resources here.
-	 * ntfs_fs_free will be called with fc->s_fs_info = NULL
-	 */
-	put_mount_options(sbi->options);
-	put_ntfs(sbi);
-	sb->s_fs_info = NULL;
 	kfree(boot2);
-
 	return err;
 }
 
@@ -1659,7 +1645,7 @@ static void ntfs_fs_free(struct fs_context *fc)
 	struct ntfs_sb_info *sbi = fc->s_fs_info;
 
 	if (sbi)
-		put_ntfs(sbi);
+		ntfs3_free_sbi(sbi);
 
 	if (opts)
 		put_mount_options(opts);
@@ -1728,13 +1714,24 @@ free_opts:
 	return -ENOMEM;
 }
 
+static void ntfs3_kill_sb(struct super_block *sb)
+{
+	struct ntfs_sb_info *sbi = sb->s_fs_info;
+
+	kill_block_super(sb);
+
+	if (sbi->options)
+		put_mount_options(sbi->options);
+	ntfs3_free_sbi(sbi);
+}
+
 // clang-format off
 static struct file_system_type ntfs_fs_type = {
 	.owner			= THIS_MODULE,
 	.name			= "ntfs3",
 	.init_fs_context	= ntfs_init_fs_context,
 	.parameters		= ntfs_fs_parameters,
-	.kill_sb		= kill_block_super,
+	.kill_sb		= ntfs3_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 };
 // clang-format on
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
index 25d8072ccfce..c19c730c26e2 100644
--- a/fs/ocfs2/journal.c
+++ b/fs/ocfs2/journal.c
@@ -557,7 +557,7 @@ static void ocfs2_abort_trigger(struct jbd2_buffer_trigger_type *triggers,
 	     (unsigned long)bh,
 	     (unsigned long long)bh->b_blocknr);
 
-	ocfs2_error(bh->b_bdev->bd_super,
+	ocfs2_error(bh->b_assoc_map->host->i_sb,
 		    "JBD2 has aborted our journal, ocfs2 cannot continue\n");
 }
 
@@ -780,14 +780,14 @@ void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh)
 		mlog_errno(status);
 		if (!is_handle_aborted(handle)) {
 			journal_t *journal = handle->h_transaction->t_journal;
-			struct super_block *sb = bh->b_bdev->bd_super;
 
 			mlog(ML_ERROR, "jbd2_journal_dirty_metadata failed. "
 					"Aborting transaction and journal.\n");
 			handle->h_err = status;
 			jbd2_journal_abort_handle(handle);
 			jbd2_journal_abort(journal, status);
-			ocfs2_abort(sb, "Journal already aborted.\n");
+			ocfs2_abort(bh->b_assoc_map->host->i_sb,
+				    "Journal already aborted.\n");
 		}
 	}
 }
diff --git a/fs/romfs/super.c b/fs/romfs/super.c
index d89739655f9e..5c35f6c76037 100644
--- a/fs/romfs/super.c
+++ b/fs/romfs/super.c
@@ -582,16 +582,18 @@ static int romfs_init_fs_context(struct fs_context *fc)
  */
 static void romfs_kill_sb(struct super_block *sb)
 {
+	generic_shutdown_super(sb);
+
 #ifdef CONFIG_ROMFS_ON_MTD
 	if (sb->s_mtd) {
-		kill_mtd_super(sb);
-		return;
+		put_mtd_device(sb->s_mtd);
+		sb->s_mtd = NULL;
 	}
 #endif
 #ifdef CONFIG_ROMFS_ON_BLOCK
 	if (sb->s_bdev) {
-		kill_block_super(sb);
-		return;
+		sync_blockdev(sb->s_bdev);
+		blkdev_put(sb->s_bdev, sb);
 	}
 #endif
 }
diff --git a/fs/super.c b/fs/super.c
index e04a403dbffd..692654c2af36 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -39,7 +39,7 @@
 #include <uapi/linux/mount.h>
 #include "internal.h"
 
-static int thaw_super_locked(struct super_block *sb);
+static int thaw_super_locked(struct super_block *sb, enum freeze_holder who);
 
 static LIST_HEAD(super_blocks);
 static DEFINE_SPINLOCK(sb_lock);
@@ -50,6 +50,130 @@ static char *sb_writers_name[SB_FREEZE_LEVELS] = {
 	"sb_internal",
 };
 
+static inline void __super_lock(struct super_block *sb, bool excl)
+{
+	if (excl)
+		down_write(&sb->s_umount);
+	else
+		down_read(&sb->s_umount);
+}
+
+static inline void super_unlock(struct super_block *sb, bool excl)
+{
+	if (excl)
+		up_write(&sb->s_umount);
+	else
+		up_read(&sb->s_umount);
+}
+
+static inline void __super_lock_excl(struct super_block *sb)
+{
+	__super_lock(sb, true);
+}
+
+static inline void super_unlock_excl(struct super_block *sb)
+{
+	super_unlock(sb, true);
+}
+
+static inline void super_unlock_shared(struct super_block *sb)
+{
+	super_unlock(sb, false);
+}
+
+static inline bool wait_born(struct super_block *sb)
+{
+	unsigned int flags;
+
+	/*
+	 * Pairs with smp_store_release() in super_wake() and ensures
+	 * that we see SB_BORN or SB_DYING after we're woken.
+	 */
+	flags = smp_load_acquire(&sb->s_flags);
+	return flags & (SB_BORN | SB_DYING);
+}
+
+/**
+ * super_lock - wait for superblock to become ready and lock it
+ * @sb: superblock to wait for
+ * @excl: whether exclusive access is required
+ *
+ * If the superblock has neither passed through vfs_get_tree() or
+ * generic_shutdown_super() yet wait for it to happen. Either superblock
+ * creation will succeed and SB_BORN is set by vfs_get_tree() or we're
+ * woken and we'll see SB_DYING.
+ *
+ * The caller must have acquired a temporary reference on @sb->s_count.
+ *
+ * Return: This returns true if SB_BORN was set, false if SB_DYING was
+ *         set. The function acquires s_umount and returns with it held.
+ */
+static __must_check bool super_lock(struct super_block *sb, bool excl)
+{
+
+	lockdep_assert_not_held(&sb->s_umount);
+
+relock:
+	__super_lock(sb, excl);
+
+	/*
+	 * Has gone through generic_shutdown_super() in the meantime.
+	 * @sb->s_root is NULL and @sb->s_active is 0. No one needs to
+	 * grab a reference to this. Tell them so.
+	 */
+	if (sb->s_flags & SB_DYING)
+		return false;
+
+	/* Has called ->get_tree() successfully. */
+	if (sb->s_flags & SB_BORN)
+		return true;
+
+	super_unlock(sb, excl);
+
+	/* wait until the superblock is ready or dying */
+	wait_var_event(&sb->s_flags, wait_born(sb));
+
+	/*
+	 * Neither SB_BORN nor SB_DYING are ever unset so we never loop.
+	 * Just reacquire @sb->s_umount for the caller.
+	 */
+	goto relock;
+}
+
+/* wait and acquire read-side of @sb->s_umount */
+static inline bool super_lock_shared(struct super_block *sb)
+{
+	return super_lock(sb, false);
+}
+
+/* wait and acquire write-side of @sb->s_umount */
+static inline bool super_lock_excl(struct super_block *sb)
+{
+	return super_lock(sb, true);
+}
+
+/* wake waiters */
+#define SUPER_WAKE_FLAGS (SB_BORN | SB_DYING | SB_DEAD)
+static void super_wake(struct super_block *sb, unsigned int flag)
+{
+	WARN_ON_ONCE((flag & ~SUPER_WAKE_FLAGS));
+	WARN_ON_ONCE(hweight32(flag & SUPER_WAKE_FLAGS) > 1);
+
+	/*
+	 * Pairs with smp_load_acquire() in super_lock() to make sure
+	 * all initializations in the superblock are seen by the user
+	 * seeing SB_BORN sent.
+	 */
+	smp_store_release(&sb->s_flags, sb->s_flags | flag);
+	/*
+	 * Pairs with the barrier in prepare_to_wait_event() to make sure
+	 * ___wait_var_event() either sees SB_BORN set or
+	 * waitqueue_active() check in wake_up_var() sees the waiter.
+	 */
+	smp_mb();
+	wake_up_var(&sb->s_flags);
+}
+
 /*
  * One thing we have to be careful of with a per-sb shrinker is that we don't
  * drop the last active reference to the superblock from within the shrinker.
@@ -76,7 +200,7 @@ static unsigned long super_cache_scan(struct shrinker *shrink,
 	if (!(sc->gfp_mask & __GFP_FS))
 		return SHRINK_STOP;
 
-	if (!trylock_super(sb))
+	if (!super_trylock_shared(sb))
 		return SHRINK_STOP;
 
 	if (sb->s_op->nr_cached_objects)
@@ -110,7 +234,7 @@ static unsigned long super_cache_scan(struct shrinker *shrink,
 		freed += sb->s_op->free_cached_objects(sb, sc);
 	}
 
-	up_read(&sb->s_umount);
+	super_unlock_shared(sb);
 	return freed;
 }
 
@@ -123,17 +247,17 @@ static unsigned long super_cache_count(struct shrinker *shrink,
 	sb = container_of(shrink, struct super_block, s_shrink);
 
 	/*
-	 * We don't call trylock_super() here as it is a scalability bottleneck,
-	 * so we're exposed to partial setup state. The shrinker rwsem does not
-	 * protect filesystem operations backing list_lru_shrink_count() or
-	 * s_op->nr_cached_objects(). Counts can change between
-	 * super_cache_count and super_cache_scan, so we really don't need locks
-	 * here.
+	 * We don't call super_trylock_shared() here as it is a scalability
+	 * bottleneck, so we're exposed to partial setup state. The shrinker
+	 * rwsem does not protect filesystem operations backing
+	 * list_lru_shrink_count() or s_op->nr_cached_objects(). Counts can
+	 * change between super_cache_count and super_cache_scan, so we really
+	 * don't need locks here.
 	 *
 	 * However, if we are currently mounting the superblock, the underlying
 	 * filesystem might be in a state of partial construction and hence it
-	 * is dangerous to access it.  trylock_super() uses a SB_BORN check to
-	 * avoid this situation, so do the same here. The memory barrier is
+	 * is dangerous to access it.  super_trylock_shared() uses a SB_BORN check
+	 * to avoid this situation, so do the same here. The memory barrier is
 	 * matched with the one in mount_fs() as we don't hold locks here.
 	 */
 	if (!(sb->s_flags & SB_BORN))
@@ -176,7 +300,7 @@ static void destroy_unused_super(struct super_block *s)
 {
 	if (!s)
 		return;
-	up_write(&s->s_umount);
+	super_unlock_excl(s);
 	list_lru_destroy(&s->s_dentry_lru);
 	list_lru_destroy(&s->s_inode_lru);
 	security_sb_free(s);
@@ -337,10 +461,29 @@ void deactivate_locked_super(struct super_block *s)
 		list_lru_destroy(&s->s_dentry_lru);
 		list_lru_destroy(&s->s_inode_lru);
 
+		/*
+		 * Remove it from @fs_supers so it isn't found by new
+		 * sget{_fc}() walkers anymore. Any concurrent mounter still
+		 * managing to grab a temporary reference is guaranteed to
+		 * already see SB_DYING and will wait until we notify them about
+		 * SB_DEAD.
+		 */
+		spin_lock(&sb_lock);
+		hlist_del_init(&s->s_instances);
+		spin_unlock(&sb_lock);
+
+		/*
+		 * Let concurrent mounts know that this thing is really dead.
+		 * We don't need @sb->s_umount here as every concurrent caller
+		 * will see SB_DYING and either discard the superblock or wait
+		 * for SB_DEAD.
+		 */
+		super_wake(s, SB_DEAD);
+
 		put_filesystem(fs);
 		put_super(s);
 	} else {
-		up_write(&s->s_umount);
+		super_unlock_excl(s);
 	}
 }
 
@@ -357,7 +500,7 @@ EXPORT_SYMBOL(deactivate_locked_super);
 void deactivate_super(struct super_block *s)
 {
 	if (!atomic_add_unless(&s->s_active, -1, 1)) {
-		down_write(&s->s_umount);
+		__super_lock_excl(s);
 		deactivate_locked_super(s);
 	}
 }
@@ -379,20 +522,61 @@ EXPORT_SYMBOL(deactivate_super);
  */
 static int grab_super(struct super_block *s) __releases(sb_lock)
 {
+	bool born;
+
 	s->s_count++;
 	spin_unlock(&sb_lock);
-	down_write(&s->s_umount);
-	if ((s->s_flags & SB_BORN) && atomic_inc_not_zero(&s->s_active)) {
+	born = super_lock_excl(s);
+	if (born && atomic_inc_not_zero(&s->s_active)) {
 		put_super(s);
 		return 1;
 	}
-	up_write(&s->s_umount);
+	super_unlock_excl(s);
 	put_super(s);
 	return 0;
 }
 
+static inline bool wait_dead(struct super_block *sb)
+{
+	unsigned int flags;
+
+	/*
+	 * Pairs with memory barrier in super_wake() and ensures
+	 * that we see SB_DEAD after we're woken.
+	 */
+	flags = smp_load_acquire(&sb->s_flags);
+	return flags & SB_DEAD;
+}
+
+/**
+ * grab_super_dead - acquire an active reference to a superblock
+ * @sb: superblock to acquire
+ *
+ * Acquire a temporary reference on a superblock and try to trade it for
+ * an active reference. This is used in sget{_fc}() to wait for a
+ * superblock to either become SB_BORN or for it to pass through
+ * sb->kill() and be marked as SB_DEAD.
+ *
+ * Return: This returns true if an active reference could be acquired,
+ *         false if not.
+ */
+static bool grab_super_dead(struct super_block *sb)
+{
+
+	sb->s_count++;
+	if (grab_super(sb)) {
+		put_super(sb);
+		lockdep_assert_held(&sb->s_umount);
+		return true;
+	}
+	wait_var_event(&sb->s_flags, wait_dead(sb));
+	put_super(sb);
+	lockdep_assert_not_held(&sb->s_umount);
+	return false;
+}
+
 /*
- *	trylock_super - try to grab ->s_umount shared
+ *	super_trylock_shared - try to grab ->s_umount shared
  *	@sb: reference we are trying to grab
  *
  *	Try to prevent fs shutdown.  This is used in places where we
@@ -408,13 +592,13 @@ static int grab_super(struct super_block *s) __releases(sb_lock)
  *	of down_read().  There's a couple of places that are OK with that, but
  *	it's very much not a general-purpose interface.
  */
-bool trylock_super(struct super_block *sb)
+bool super_trylock_shared(struct super_block *sb)
 {
 	if (down_read_trylock(&sb->s_umount)) {
-		if (!hlist_unhashed(&sb->s_instances) &&
-		    sb->s_root && (sb->s_flags & SB_BORN))
+		if (!(sb->s_flags & SB_DYING) && sb->s_root &&
+		    (sb->s_flags & SB_BORN))
 			return true;
-		up_read(&sb->s_umount);
+		super_unlock_shared(sb);
 	}
 
 	return false;
@@ -439,13 +623,13 @@ bool trylock_super(struct super_block *sb)
 void retire_super(struct super_block *sb)
 {
 	WARN_ON(!sb->s_bdev);
-	down_write(&sb->s_umount);
+	__super_lock_excl(sb);
 	if (sb->s_iflags & SB_I_PERSB_BDI) {
 		bdi_unregister(sb->s_bdi);
 		sb->s_iflags &= ~SB_I_PERSB_BDI;
 	}
 	sb->s_iflags |= SB_I_RETIRED;
-	up_write(&sb->s_umount);
+	super_unlock_excl(sb);
 }
 EXPORT_SYMBOL(retire_super);
 
@@ -517,11 +701,17 @@ void generic_shutdown_super(struct super_block *sb)
 			spin_unlock(&sb->s_inode_list_lock);
 		}
 	}
-	spin_lock(&sb_lock);
-	/* should be initialized for __put_super_and_need_restart() */
-	hlist_del_init(&sb->s_instances);
-	spin_unlock(&sb_lock);
-	up_write(&sb->s_umount);
+	/*
+	 * Broadcast to everyone that grabbed a temporary reference to this
+	 * superblock before we removed it from @fs_supers that the superblock
+	 * is dying. Every walker of @fs_supers outside of sget{_fc}() will now
+	 * discard this superblock and treat it as dead.
+	 *
+	 * We leave the superblock on @fs_supers so it can be found by
+	 * sget{_fc}() until we passed sb->kill_sb().
+	 */
+	super_wake(sb, SB_DYING);
+	super_unlock_excl(sb);
 	if (sb->s_bdi != &noop_backing_dev_info) {
 		if (sb->s_iflags & SB_I_PERSB_BDI)
 			bdi_unregister(sb->s_bdi);
@@ -609,6 +799,11 @@ retry:
 	s->s_type = fc->fs_type;
 	s->s_iflags |= fc->s_iflags;
 	strscpy(s->s_id, s->s_type->name, sizeof(s->s_id));
+	/*
+	 * Make the superblock visible on @super_blocks and @fs_supers.
+	 * It's in a nascent state and users should wait on SB_BORN or
+	 * SB_DYING to be set.
+	 */
 	list_add_tail(&s->s_list, &super_blocks);
 	hlist_add_head(&s->s_instances, &s->s_type->fs_supers);
 	spin_unlock(&sb_lock);
@@ -626,7 +821,7 @@ share_extant_sb:
 			warnfc(fc, "reusing existing filesystem in another namespace not allowed");
 		return ERR_PTR(-EBUSY);
 	}
-	if (!grab_super(old))
+	if (!grab_super_dead(old))
 		goto retry;
 	destroy_unused_super(s);
 	return old;
@@ -670,7 +865,7 @@ retry:
 				destroy_unused_super(s);
 				return ERR_PTR(-EBUSY);
 			}
-			if (!grab_super(old))
+			if (!grab_super_dead(old))
 				goto retry;
 			destroy_unused_super(s);
 			return old;
@@ -703,7 +898,7 @@ EXPORT_SYMBOL(sget);
 
 void drop_super(struct super_block *sb)
 {
-	up_read(&sb->s_umount);
+	super_unlock_shared(sb);
 	put_super(sb);
 }
 
@@ -711,7 +906,7 @@ EXPORT_SYMBOL(drop_super);
 
 void drop_super_exclusive(struct super_block *sb)
 {
-	up_write(&sb->s_umount);
+	super_unlock_excl(sb);
 	put_super(sb);
 }
 EXPORT_SYMBOL(drop_super_exclusive);
@@ -722,7 +917,8 @@ static void __iterate_supers(void (*f)(struct super_block *))
 
 	spin_lock(&sb_lock);
 	list_for_each_entry(sb, &super_blocks, s_list) {
-		if (hlist_unhashed(&sb->s_instances))
+		/* Pairs with memory marrier in super_wake(). */
+		if (smp_load_acquire(&sb->s_flags) & SB_DYING)
 			continue;
 		sb->s_count++;
 		spin_unlock(&sb_lock);
@@ -752,15 +948,15 @@ void iterate_supers(void (*f)(struct super_block *, void *), void *arg)
 
 	spin_lock(&sb_lock);
 	list_for_each_entry(sb, &super_blocks, s_list) {
-		if (hlist_unhashed(&sb->s_instances))
-			continue;
+		bool born;
+
 		sb->s_count++;
 		spin_unlock(&sb_lock);
 
-		down_read(&sb->s_umount);
-		if (sb->s_root && (sb->s_flags & SB_BORN))
+		born = super_lock_shared(sb);
+		if (born && sb->s_root)
 			f(sb, arg);
-		up_read(&sb->s_umount);
+		super_unlock_shared(sb);
 
 		spin_lock(&sb_lock);
 		if (p)
@@ -788,13 +984,15 @@ void iterate_supers_type(struct file_system_type *type,
 
 	spin_lock(&sb_lock);
 	hlist_for_each_entry(sb, &type->fs_supers, s_instances) {
+		bool born;
+
 		sb->s_count++;
 		spin_unlock(&sb_lock);
 
-		down_read(&sb->s_umount);
-		if (sb->s_root && (sb->s_flags & SB_BORN))
+		born = super_lock_shared(sb);
+		if (born && sb->s_root)
 			f(sb, arg);
-		up_read(&sb->s_umount);
+		super_unlock_shared(sb);
 
 		spin_lock(&sb_lock);
 		if (p)
@@ -809,43 +1007,6 @@ void iterate_supers_type(struct file_system_type *type,
 EXPORT_SYMBOL(iterate_supers_type);
 
 /**
- * get_super - get the superblock of a device
- * @bdev: device to get the superblock for
- *
- * Scans the superblock list and finds the superblock of the file system
- * mounted on the device given. %NULL is returned if no match is found.
- */
-struct super_block *get_super(struct block_device *bdev)
-{
-	struct super_block *sb;
-
-	if (!bdev)
-		return NULL;
-
-	spin_lock(&sb_lock);
-rescan:
-	list_for_each_entry(sb, &super_blocks, s_list) {
-		if (hlist_unhashed(&sb->s_instances))
-			continue;
-		if (sb->s_bdev == bdev) {
-			sb->s_count++;
-			spin_unlock(&sb_lock);
-			down_read(&sb->s_umount);
-			/* still alive? */
-			if (sb->s_root && (sb->s_flags & SB_BORN))
-				return sb;
-			up_read(&sb->s_umount);
-			/* nope, got unmounted */
-			spin_lock(&sb_lock);
-			__put_super(sb);
-			goto rescan;
-		}
-	}
-	spin_unlock(&sb_lock);
-	return NULL;
-}
-
-/**
  * get_active_super - get an active reference to the superblock of a device
  * @bdev: device to get the superblock for
  *
@@ -860,15 +1021,12 @@ struct super_block *get_active_super(struct block_device *bdev)
 	if (!bdev)
 		return NULL;
 
-restart:
 	spin_lock(&sb_lock);
 	list_for_each_entry(sb, &super_blocks, s_list) {
-		if (hlist_unhashed(&sb->s_instances))
-			continue;
 		if (sb->s_bdev == bdev) {
 			if (!grab_super(sb))
-				goto restart;
-			up_write(&sb->s_umount);
+				return NULL;
+			super_unlock_excl(sb);
 			return sb;
 		}
 	}
@@ -881,28 +1039,21 @@ struct super_block *user_get_super(dev_t dev, bool excl)
 	struct super_block *sb;
 
 	spin_lock(&sb_lock);
-rescan:
 	list_for_each_entry(sb, &super_blocks, s_list) {
-		if (hlist_unhashed(&sb->s_instances))
-			continue;
 		if (sb->s_dev ==  dev) {
+			bool born;
+
 			sb->s_count++;
 			spin_unlock(&sb_lock);
-			if (excl)
-				down_write(&sb->s_umount);
-			else
-				down_read(&sb->s_umount);
 			/* still alive? */
-			if (sb->s_root && (sb->s_flags & SB_BORN))
+			born = super_lock(sb, excl);
+			if (born && sb->s_root)
 				return sb;
-			if (excl)
-				up_write(&sb->s_umount);
-			else
-				up_read(&sb->s_umount);
+			super_unlock(sb, excl);
 			/* nope, got unmounted */
 			spin_lock(&sb_lock);
 			__put_super(sb);
-			goto rescan;
+			break;
 		}
 	}
 	spin_unlock(&sb_lock);
@@ -944,9 +1095,9 @@ int reconfigure_super(struct fs_context *fc)
 
 	if (remount_ro) {
 		if (!hlist_empty(&sb->s_pins)) {
-			up_write(&sb->s_umount);
+			super_unlock_excl(sb);
 			group_pin_kill(&sb->s_pins);
-			down_write(&sb->s_umount);
+			__super_lock_excl(sb);
 			if (!sb->s_root)
 				return 0;
 			if (sb->s_writers.frozen != SB_UNFROZEN)
@@ -1009,9 +1160,9 @@ cancel_readonly:
 
 static void do_emergency_remount_callback(struct super_block *sb)
 {
-	down_write(&sb->s_umount);
-	if (sb->s_root && sb->s_bdev && (sb->s_flags & SB_BORN) &&
-	    !sb_rdonly(sb)) {
+	bool born = super_lock_excl(sb);
+
+	if (born && sb->s_root && sb->s_bdev && !sb_rdonly(sb)) {
 		struct fs_context *fc;
 
 		fc = fs_context_for_reconfigure(sb->s_root,
@@ -1022,7 +1173,7 @@ static void do_emergency_remount_callback(struct super_block *sb)
 			put_fs_context(fc);
 		}
 	}
-	up_write(&sb->s_umount);
+	super_unlock_excl(sb);
 }
 
 static void do_emergency_remount(struct work_struct *work)
@@ -1045,12 +1196,13 @@ void emergency_remount(void)
 
 static void do_thaw_all_callback(struct super_block *sb)
 {
-	down_write(&sb->s_umount);
-	if (sb->s_root && sb->s_flags & SB_BORN) {
+	bool born = super_lock_excl(sb);
+
+	if (born && sb->s_root) {
 		emergency_thaw_bdev(sb);
-		thaw_super_locked(sb);
+		thaw_super_locked(sb, FREEZE_HOLDER_USERSPACE);
 	} else {
-		up_write(&sb->s_umount);
+		super_unlock_excl(sb);
 	}
 }
 
@@ -1209,31 +1361,66 @@ int get_tree_keyed(struct fs_context *fc,
 EXPORT_SYMBOL(get_tree_keyed);
 
 #ifdef CONFIG_BLOCK
-static void fs_mark_dead(struct block_device *bdev)
+/*
+ * Lock a super block that the callers holds a reference to.
+ *
+ * The caller needs to ensure that the super_block isn't being freed while
+ * calling this function, e.g. by holding a lock over the call to this function
+ * and the place that clears the pointer to the superblock used by this function
+ * before freeing the superblock.
+ */
+static bool super_lock_shared_active(struct super_block *sb)
 {
-	struct super_block *sb;
+	bool born = super_lock_shared(sb);
 
-	sb = get_super(bdev);
-	if (!sb)
+	if (!born || !sb->s_root || !(sb->s_flags & SB_ACTIVE)) {
+		super_unlock_shared(sb);
+		return false;
+	}
+	return true;
+}
+
+static void fs_bdev_mark_dead(struct block_device *bdev, bool surprise)
+{
+	struct super_block *sb = bdev->bd_holder;
+
+	/* bd_holder_lock ensures that the sb isn't freed */
+	lockdep_assert_held(&bdev->bd_holder_lock);
+
+	if (!super_lock_shared_active(sb))
 		return;
 
+	if (!surprise)
+		sync_filesystem(sb);
+	shrink_dcache_sb(sb);
+	invalidate_inodes(sb);
 	if (sb->s_op->shutdown)
 		sb->s_op->shutdown(sb);
-	drop_super(sb);
+
+	super_unlock_shared(sb);
 }
 
-static const struct blk_holder_ops fs_holder_ops = {
-	.mark_dead		= fs_mark_dead,
+static void fs_bdev_sync(struct block_device *bdev)
+{
+	struct super_block *sb = bdev->bd_holder;
+
+	lockdep_assert_held(&bdev->bd_holder_lock);
+
+	if (!super_lock_shared_active(sb))
+		return;
+	sync_filesystem(sb);
+	super_unlock_shared(sb);
+}
+
+const struct blk_holder_ops fs_holder_ops = {
+	.mark_dead		= fs_bdev_mark_dead,
+	.sync			= fs_bdev_sync,
 };
+EXPORT_SYMBOL_GPL(fs_holder_ops);
 
 static int set_bdev_super(struct super_block *s, void *data)
 {
-	s->s_bdev = data;
-	s->s_dev = s->s_bdev->bd_dev;
-	s->s_bdi = bdi_get(s->s_bdev->bd_disk->bdi);
-
-	if (bdev_stable_writes(s->s_bdev))
-		s->s_iflags |= SB_I_STABLE_WRITES;
+	s->s_dev = *(dev_t *)data;
 	return 0;
 }
 
@@ -1244,8 +1431,63 @@ static int set_bdev_super_fc(struct super_block *s, struct fs_context *fc)
 
 static int test_bdev_super_fc(struct super_block *s, struct fs_context *fc)
 {
-	return !(s->s_iflags & SB_I_RETIRED) && s->s_bdev == fc->sget_key;
+	return !(s->s_iflags & SB_I_RETIRED) &&
+		s->s_dev == *(dev_t *)fc->sget_key;
+}
+
+int setup_bdev_super(struct super_block *sb, int sb_flags,
+		struct fs_context *fc)
+{
+	blk_mode_t mode = sb_open_mode(sb_flags);
+	struct block_device *bdev;
+
+	bdev = blkdev_get_by_dev(sb->s_dev, mode, sb, &fs_holder_ops);
+	if (IS_ERR(bdev)) {
+		if (fc)
+			errorf(fc, "%s: Can't open blockdev", fc->source);
+		return PTR_ERR(bdev);
+	}
+
+	/*
+	 * This really should be in blkdev_get_by_dev, but right now can't due
+	 * to legacy issues that require us to allow opening a block device node
+	 * writable from userspace even for a read-only block device.
+	 */
+	if ((mode & BLK_OPEN_WRITE) && bdev_read_only(bdev)) {
+		blkdev_put(bdev, sb);
+		return -EACCES;
+	}
+
+	/*
+	 * Until SB_BORN flag is set, there can be no active superblock
+	 * references and thus no filesystem freezing. get_active_super() will
+	 * just loop waiting for SB_BORN so even freeze_bdev() cannot proceed.
+	 *
+	 * It is enough to check bdev was not frozen before we set s_bdev.
+	 */
+	mutex_lock(&bdev->bd_fsfreeze_mutex);
+	if (bdev->bd_fsfreeze_count > 0) {
+		mutex_unlock(&bdev->bd_fsfreeze_mutex);
+		if (fc)
+			warnf(fc, "%pg: Can't mount, blockdev is frozen", bdev);
+		blkdev_put(bdev, sb);
+		return -EBUSY;
+	}
+	spin_lock(&sb_lock);
+	sb->s_bdev = bdev;
+	sb->s_bdi = bdi_get(bdev->bd_disk->bdi);
+	if (bdev_stable_writes(bdev))
+		sb->s_iflags |= SB_I_STABLE_WRITES;
+	spin_unlock(&sb_lock);
+	mutex_unlock(&bdev->bd_fsfreeze_mutex);
+
+	snprintf(sb->s_id, sizeof(sb->s_id), "%pg", bdev);
+	shrinker_debugfs_rename(&sb->s_shrink, "sb-%s:%s", sb->s_type->name,
+				sb->s_id);
+	sb_set_blocksize(sb, block_size(bdev));
+	return 0;
 }
+EXPORT_SYMBOL_GPL(setup_bdev_super);
 
 /**
  * get_tree_bdev - Get a superblock based on a single block device
@@ -1256,73 +1498,49 @@ int get_tree_bdev(struct fs_context *fc,
 		int (*fill_super)(struct super_block *,
 				  struct fs_context *))
 {
-	struct block_device *bdev;
 	struct super_block *s;
 	int error = 0;
+	dev_t dev;
 
 	if (!fc->source)
 		return invalf(fc, "No source specified");
 
-	bdev = blkdev_get_by_path(fc->source, sb_open_mode(fc->sb_flags),
-				  fc->fs_type, &fs_holder_ops);
-	if (IS_ERR(bdev)) {
-		errorf(fc, "%s: Can't open blockdev", fc->source);
-		return PTR_ERR(bdev);
-	}
-
-	/* Once the superblock is inserted into the list by sget_fc(), s_umount
-	 * will protect the lockfs code from trying to start a snapshot while
-	 * we are mounting
-	 */
-	mutex_lock(&bdev->bd_fsfreeze_mutex);
-	if (bdev->bd_fsfreeze_count > 0) {
-		mutex_unlock(&bdev->bd_fsfreeze_mutex);
-		warnf(fc, "%pg: Can't mount, blockdev is frozen", bdev);
-		blkdev_put(bdev, fc->fs_type);
-		return -EBUSY;
+	error = lookup_bdev(fc->source, &dev);
+	if (error) {
+		errorf(fc, "%s: Can't lookup blockdev", fc->source);
+		return error;
 	}
 
 	fc->sb_flags |= SB_NOSEC;
-	fc->sget_key = bdev;
+	fc->sget_key = &dev;
 	s = sget_fc(fc, test_bdev_super_fc, set_bdev_super_fc);
-	mutex_unlock(&bdev->bd_fsfreeze_mutex);
-	if (IS_ERR(s)) {
-		blkdev_put(bdev, fc->fs_type);
+	if (IS_ERR(s))
 		return PTR_ERR(s);
-	}
 
 	if (s->s_root) {
 		/* Don't summarily change the RO/RW state. */
 		if ((fc->sb_flags ^ s->s_flags) & SB_RDONLY) {
-			warnf(fc, "%pg: Can't mount, would change RO state", bdev);
+			warnf(fc, "%pg: Can't mount, would change RO state", s->s_bdev);
 			deactivate_locked_super(s);
-			blkdev_put(bdev, fc->fs_type);
 			return -EBUSY;
 		}
-
+	} else {
 		/*
-		 * s_umount nests inside open_mutex during
-		 * __invalidate_device().  blkdev_put() acquires
-		 * open_mutex and can't be called under s_umount.  Drop
-		 * s_umount temporarily.  This is safe as we're
-		 * holding an active reference.
+		 * We drop s_umount here because we need to open the bdev and
+		 * bdev->open_mutex ranks above s_umount (blkdev_put() ->
+		 * bdev_mark_dead()). It is safe because we have active sb
+		 * reference and SB_BORN is not set yet.
 		 */
-		up_write(&s->s_umount);
-		blkdev_put(bdev, fc->fs_type);
-		down_write(&s->s_umount);
-	} else {
-		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
-		shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s",
-					fc->fs_type->name, s->s_id);
-		sb_set_blocksize(s, block_size(bdev));
-		error = fill_super(s, fc);
+		super_unlock_excl(s);
+		error = setup_bdev_super(s, fc->sb_flags, fc);
+		__super_lock_excl(s);
+		if (!error)
+			error = fill_super(s, fc);
 		if (error) {
 			deactivate_locked_super(s);
 			return error;
 		}
-
 		s->s_flags |= SB_ACTIVE;
-		bdev->bd_super = s;
 	}
 
 	BUG_ON(fc->root);
@@ -1333,79 +1551,52 @@ EXPORT_SYMBOL(get_tree_bdev);
 
 static int test_bdev_super(struct super_block *s, void *data)
 {
-	return !(s->s_iflags & SB_I_RETIRED) && (void *)s->s_bdev == data;
+	return !(s->s_iflags & SB_I_RETIRED) && s->s_dev == *(dev_t *)data;
 }
 
 struct dentry *mount_bdev(struct file_system_type *fs_type,
 	int flags, const char *dev_name, void *data,
 	int (*fill_super)(struct super_block *, void *, int))
 {
-	struct block_device *bdev;
 	struct super_block *s;
-	int error = 0;
+	int error;
+	dev_t dev;
 
-	bdev = blkdev_get_by_path(dev_name, sb_open_mode(flags), fs_type,
-				  &fs_holder_ops);
-	if (IS_ERR(bdev))
-		return ERR_CAST(bdev);
+	error = lookup_bdev(dev_name, &dev);
+	if (error)
+		return ERR_PTR(error);
 
-	/*
-	 * once the super is inserted into the list by sget, s_umount
-	 * will protect the lockfs code from trying to start a snapshot
-	 * while we are mounting
-	 */
-	mutex_lock(&bdev->bd_fsfreeze_mutex);
-	if (bdev->bd_fsfreeze_count > 0) {
-		mutex_unlock(&bdev->bd_fsfreeze_mutex);
-		error = -EBUSY;
-		goto error_bdev;
-	}
-	s = sget(fs_type, test_bdev_super, set_bdev_super, flags | SB_NOSEC,
-		 bdev);
-	mutex_unlock(&bdev->bd_fsfreeze_mutex);
+	flags |= SB_NOSEC;
+	s = sget(fs_type, test_bdev_super, set_bdev_super, flags, &dev);
 	if (IS_ERR(s))
-		goto error_s;
+		return ERR_CAST(s);
 
 	if (s->s_root) {
 		if ((flags ^ s->s_flags) & SB_RDONLY) {
 			deactivate_locked_super(s);
-			error = -EBUSY;
-			goto error_bdev;
+			return ERR_PTR(-EBUSY);
 		}
-
+	} else {
 		/*
-		 * s_umount nests inside open_mutex during
-		 * __invalidate_device().  blkdev_put() acquires
-		 * open_mutex and can't be called under s_umount.  Drop
-		 * s_umount temporarily.  This is safe as we're
-		 * holding an active reference.
+		 * We drop s_umount here because we need to open the bdev and
+		 * bdev->open_mutex ranks above s_umount (blkdev_put() ->
+		 * bdev_mark_dead()). It is safe because we have active sb
+		 * reference and SB_BORN is not set yet.
 		 */
-		up_write(&s->s_umount);
-		blkdev_put(bdev, fs_type);
-		down_write(&s->s_umount);
-	} else {
-		snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
-		shrinker_debugfs_rename(&s->s_shrink, "sb-%s:%s",
-					fs_type->name, s->s_id);
-		sb_set_blocksize(s, block_size(bdev));
-		error = fill_super(s, data, flags & SB_SILENT ? 1 : 0);
+		super_unlock_excl(s);
+		error = setup_bdev_super(s, flags, NULL);
+		__super_lock_excl(s);
+		if (!error)
+			error = fill_super(s, data, flags & SB_SILENT ? 1 : 0);
 		if (error) {
 			deactivate_locked_super(s);
-			goto error;
+			return ERR_PTR(error);
 		}
 
 		s->s_flags |= SB_ACTIVE;
-		bdev->bd_super = s;
 	}
 
 	return dget(s->s_root);
-
-error_s:
-	error = PTR_ERR(s);
-error_bdev:
-	blkdev_put(bdev, fs_type);
-error:
-	return ERR_PTR(error);
 }
 EXPORT_SYMBOL(mount_bdev);
 
@@ -1413,10 +1604,11 @@ void kill_block_super(struct super_block *sb)
 {
 	struct block_device *bdev = sb->s_bdev;
 
-	bdev->bd_super = NULL;
 	generic_shutdown_super(sb);
-	sync_blockdev(bdev);
-	blkdev_put(bdev, sb->s_type);
+	if (bdev) {
+		sync_blockdev(bdev);
+		blkdev_put(bdev, sb);
+	}
 }
 
 EXPORT_SYMBOL(kill_block_super);
@@ -1533,13 +1725,13 @@ int vfs_get_tree(struct fs_context *fc)
 	WARN_ON(!sb->s_bdi);
 
 	/*
-	 * Write barrier is for super_cache_count(). We place it before setting
-	 * SB_BORN as the data dependency between the two functions is the
-	 * superblock structure contents that we just set up, not the SB_BORN
-	 * flag.
+	 * super_wake() contains a memory barrier which also care of
+	 * ordering for super_cache_count(). We place it before setting
+	 * SB_BORN as the data dependency between the two functions is
+	 * the superblock structure contents that we just set up, not
+	 * the SB_BORN flag.
 	 */
-	smp_wmb();
-	sb->s_flags |= SB_BORN;
+	super_wake(sb, SB_BORN);
 
 	error = security_sb_set_mnt_opts(sb, fc->security, 0, NULL);
 	if (unlikely(error)) {
@@ -1644,14 +1836,43 @@ static void sb_freeze_unlock(struct super_block *sb, int level)
 		percpu_up_write(sb->s_writers.rw_sem + level);
 }
 
+static int wait_for_partially_frozen(struct super_block *sb)
+{
+	int ret = 0;
+
+	do {
+		unsigned short old = sb->s_writers.frozen;
+
+		up_write(&sb->s_umount);
+		ret = wait_var_event_killable(&sb->s_writers.frozen,
+					       sb->s_writers.frozen != old);
+		down_write(&sb->s_umount);
+	} while (ret == 0 &&
+		 sb->s_writers.frozen != SB_UNFROZEN &&
+		 sb->s_writers.frozen != SB_FREEZE_COMPLETE);
+
+	return ret;
+}
+
 /**
  * freeze_super - lock the filesystem and force it into a consistent state
  * @sb: the super to lock
+ * @who: context that wants to freeze
  *
  * Syncs the super to make sure the filesystem is consistent and calls the fs's
- * freeze_fs.  Subsequent calls to this without first thawing the fs will return
+ * freeze_fs.  Subsequent calls to this without first thawing the fs may return
  * -EBUSY.
  *
+ * @who should be:
+ * * %FREEZE_HOLDER_USERSPACE if userspace wants to freeze the fs;
+ * * %FREEZE_HOLDER_KERNEL if the kernel wants to freeze the fs.
+ *
+ * The @who argument distinguishes between the kernel and userspace trying to
+ * freeze the filesystem.  Although there cannot be multiple kernel freezes or
+ * multiple userspace freezes in effect at any given time, the kernel and
+ * userspace can both hold a filesystem frozen.  The filesystem remains frozen
+ * until there are no kernel or userspace freezes in effect.
+ *
  * During this function, sb->s_writers.frozen goes through these values:
  *
  * SB_UNFROZEN: File system is normal, all writes progress as usual.
@@ -1677,34 +1898,62 @@ static void sb_freeze_unlock(struct super_block *sb, int level)
  *
  * sb->s_writers.frozen is protected by sb->s_umount.
  */
-int freeze_super(struct super_block *sb)
+int freeze_super(struct super_block *sb, enum freeze_holder who)
 {
 	int ret;
 
 	atomic_inc(&sb->s_active);
-	down_write(&sb->s_umount);
+	if (!super_lock_excl(sb))
+		WARN(1, "Dying superblock while freezing!");
+
+retry:
+	if (sb->s_writers.frozen == SB_FREEZE_COMPLETE) {
+		if (sb->s_writers.freeze_holders & who) {
+			deactivate_locked_super(sb);
+			return -EBUSY;
+		}
+
+		WARN_ON(sb->s_writers.freeze_holders == 0);
+
+		/*
+		 * Someone else already holds this type of freeze; share the
+		 * freeze and assign the active ref to the freeze.
+		 */
+		sb->s_writers.freeze_holders |= who;
+		super_unlock_excl(sb);
+		return 0;
+	}
+
 	if (sb->s_writers.frozen != SB_UNFROZEN) {
-		deactivate_locked_super(sb);
-		return -EBUSY;
+		ret = wait_for_partially_frozen(sb);
+		if (ret) {
+			deactivate_locked_super(sb);
+			return ret;
+		}
+
+		goto retry;
 	}
 
 	if (!(sb->s_flags & SB_BORN)) {
-		up_write(&sb->s_umount);
+		super_unlock_excl(sb);
 		return 0;	/* sic - it's "nothing to do" */
 	}
 
 	if (sb_rdonly(sb)) {
 		/* Nothing to do really... */
+		sb->s_writers.freeze_holders |= who;
 		sb->s_writers.frozen = SB_FREEZE_COMPLETE;
-		up_write(&sb->s_umount);
+		wake_up_var(&sb->s_writers.frozen);
+		super_unlock_excl(sb);
 		return 0;
 	}
 
 	sb->s_writers.frozen = SB_FREEZE_WRITE;
 	/* Release s_umount to preserve sb_start_write -> s_umount ordering */
-	up_write(&sb->s_umount);
+	super_unlock_excl(sb);
 	sb_wait_write(sb, SB_FREEZE_WRITE);
-	down_write(&sb->s_umount);
+	if (!super_lock_excl(sb))
+		WARN(1, "Dying superblock while freezing!");
 
 	/* Now we go and block page faults... */
 	sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
@@ -1715,6 +1964,7 @@ int freeze_super(struct super_block *sb)
 	if (ret) {
 		sb->s_writers.frozen = SB_UNFROZEN;
 		sb_freeze_unlock(sb, SB_FREEZE_PAGEFAULT);
+		wake_up_var(&sb->s_writers.frozen);
 		deactivate_locked_super(sb);
 		return ret;
 	}
@@ -1730,6 +1980,7 @@ int freeze_super(struct super_block *sb)
 				"VFS:Filesystem freeze failed\n");
 			sb->s_writers.frozen = SB_UNFROZEN;
 			sb_freeze_unlock(sb, SB_FREEZE_FS);
+			wake_up_var(&sb->s_writers.frozen);
 			deactivate_locked_super(sb);
 			return ret;
 		}
@@ -1738,24 +1989,50 @@ int freeze_super(struct super_block *sb)
 	 * For debugging purposes so that fs can warn if it sees write activity
 	 * when frozen is set to SB_FREEZE_COMPLETE, and for thaw_super().
 	 */
+	sb->s_writers.freeze_holders |= who;
 	sb->s_writers.frozen = SB_FREEZE_COMPLETE;
+	wake_up_var(&sb->s_writers.frozen);
 	lockdep_sb_freeze_release(sb);
-	up_write(&sb->s_umount);
+	super_unlock_excl(sb);
 	return 0;
 }
 EXPORT_SYMBOL(freeze_super);
 
-static int thaw_super_locked(struct super_block *sb)
+/*
+ * Undoes the effect of a freeze_super_locked call.  If the filesystem is
+ * frozen both by userspace and the kernel, a thaw call from either source
+ * removes that state without releasing the other state or unlocking the
+ * filesystem.
+ */
+static int thaw_super_locked(struct super_block *sb, enum freeze_holder who)
 {
 	int error;
 
-	if (sb->s_writers.frozen != SB_FREEZE_COMPLETE) {
-		up_write(&sb->s_umount);
+	if (sb->s_writers.frozen == SB_FREEZE_COMPLETE) {
+		if (!(sb->s_writers.freeze_holders & who)) {
+			super_unlock_excl(sb);
+			return -EINVAL;
+		}
+
+		/*
+		 * Freeze is shared with someone else.  Release our hold and
+		 * drop the active ref that freeze_super assigned to the
+		 * freezer.
+		 */
+		if (sb->s_writers.freeze_holders & ~who) {
+			sb->s_writers.freeze_holders &= ~who;
+			deactivate_locked_super(sb);
+			return 0;
+		}
+	} else {
+		super_unlock_excl(sb);
 		return -EINVAL;
 	}
 
 	if (sb_rdonly(sb)) {
+		sb->s_writers.freeze_holders &= ~who;
 		sb->s_writers.frozen = SB_UNFROZEN;
+		wake_up_var(&sb->s_writers.frozen);
 		goto out;
 	}
 
@@ -1764,15 +2041,16 @@ static int thaw_super_locked(struct super_block *sb)
 	if (sb->s_op->unfreeze_fs) {
 		error = sb->s_op->unfreeze_fs(sb);
 		if (error) {
-			printk(KERN_ERR
-				"VFS:Filesystem thaw failed\n");
+			printk(KERN_ERR "VFS:Filesystem thaw failed\n");
 			lockdep_sb_freeze_release(sb);
-			up_write(&sb->s_umount);
+			super_unlock_excl(sb);
 			return error;
 		}
 	}
 
+	sb->s_writers.freeze_holders &= ~who;
 	sb->s_writers.frozen = SB_UNFROZEN;
+	wake_up_var(&sb->s_writers.frozen);
 	sb_freeze_unlock(sb, SB_FREEZE_FS);
 out:
 	deactivate_locked_super(sb);
@@ -1782,13 +2060,20 @@ out:
 /**
  * thaw_super -- unlock filesystem
  * @sb: the super to thaw
+ * @who: context that wants to freeze
+ *
+ * Unlocks the filesystem and marks it writeable again after freeze_super()
+ * if there are no remaining freezes on the filesystem.
  *
- * Unlocks the filesystem and marks it writeable again after freeze_super().
+ * @who should be:
+ * * %FREEZE_HOLDER_USERSPACE if userspace wants to thaw the fs;
+ * * %FREEZE_HOLDER_KERNEL if the kernel wants to thaw the fs.
  */
-int thaw_super(struct super_block *sb)
+int thaw_super(struct super_block *sb, enum freeze_holder who)
 {
-	down_write(&sb->s_umount);
-	return thaw_super_locked(sb);
+	if (!super_lock_excl(sb))
+		WARN(1, "Dying superblock while thawing!");
+	return thaw_super_locked(sb, who);
 }
 EXPORT_SYMBOL(thaw_super);
 
diff --git a/fs/xfs/scrub/fscounters.c b/fs/xfs/scrub/fscounters.c
index e382a35e98d8..05be757668bb 100644
--- a/fs/xfs/scrub/fscounters.c
+++ b/fs/xfs/scrub/fscounters.c
@@ -1,4 +1,4 @@
-// SPDX-License-Identifier: GPL-2.0+
+// SPDX-License-Identifier: GPL-2.0-or-later
 /*
  * Copyright (C) 2019-2023 Oracle.  All Rights Reserved.
  * Author: Darrick J. Wong <djwong@kernel.org>
@@ -8,6 +8,8 @@
 #include "xfs_shared.h"
 #include "xfs_format.h"
 #include "xfs_trans_resv.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
 #include "xfs_mount.h"
 #include "xfs_alloc.h"
 #include "xfs_ialloc.h"
@@ -16,6 +18,7 @@
 #include "xfs_ag.h"
 #include "xfs_rtalloc.h"
 #include "xfs_inode.h"
+#include "xfs_icache.h"
 #include "scrub/scrub.h"
 #include "scrub/common.h"
 #include "scrub/trace.h"
@@ -53,6 +56,7 @@ struct xchk_fscounters {
 	uint64_t		frextents;
 	unsigned long long	icount_min;
 	unsigned long long	icount_max;
+	bool			frozen;
 };
 
 /*
@@ -123,6 +127,82 @@ xchk_fscount_warmup(
 	return error;
 }
 
+static inline int
+xchk_fsfreeze(
+	struct xfs_scrub	*sc)
+{
+	int			error;
+
+	error = freeze_super(sc->mp->m_super, FREEZE_HOLDER_KERNEL);
+	trace_xchk_fsfreeze(sc, error);
+	return error;
+}
+
+static inline int
+xchk_fsthaw(
+	struct xfs_scrub	*sc)
+{
+	int			error;
+
+	/* This should always succeed, we have a kernel freeze */
+	error = thaw_super(sc->mp->m_super, FREEZE_HOLDER_KERNEL);
+	trace_xchk_fsthaw(sc, error);
+	return error;
+}
+
+/*
+ * We couldn't stabilize the filesystem long enough to sample all the variables
+ * that comprise the summary counters and compare them to the percpu counters.
+ * We need to disable all writer threads, which means taking the first two
+ * freeze levels to put userspace to sleep, and the third freeze level to
+ * prevent background threads from starting new transactions.  Take one level
+ * more to prevent other callers from unfreezing the filesystem while we run.
+ */
+STATIC int
+xchk_fscounters_freeze(
+	struct xfs_scrub	*sc)
+{
+	struct xchk_fscounters	*fsc = sc->buf;
+	int			error = 0;
+
+	if (sc->flags & XCHK_HAVE_FREEZE_PROT) {
+		sc->flags &= ~XCHK_HAVE_FREEZE_PROT;
+		mnt_drop_write_file(sc->file);
+	}
+
+	/* Try to grab a kernel freeze. */
+	while ((error = xchk_fsfreeze(sc)) == -EBUSY) {
+		if (xchk_should_terminate(sc, &error))
+			return error;
+
+		delay(HZ / 10);
+	}
+	if (error)
+		return error;
+
+	fsc->frozen = true;
+	return 0;
+}
+
+/* Thaw the filesystem after checking or repairing fscounters. */
+STATIC void
+xchk_fscounters_cleanup(
+	void			*buf)
+{
+	struct xchk_fscounters	*fsc = buf;
+	struct xfs_scrub	*sc = fsc->sc;
+	int			error;
+
+	if (!fsc->frozen)
+		return;
+
+	error = xchk_fsthaw(sc);
+	if (error)
+		xfs_emerg(sc->mp, "still frozen after scrub, err=%d", error);
+	else
+		fsc->frozen = false;
+}
+
 int
 xchk_setup_fscounters(
 	struct xfs_scrub	*sc)
@@ -140,6 +220,7 @@ xchk_setup_fscounters(
 	sc->buf = kzalloc(sizeof(struct xchk_fscounters), XCHK_GFP_FLAGS);
 	if (!sc->buf)
 		return -ENOMEM;
+	sc->buf_cleanup = xchk_fscounters_cleanup;
 	fsc = sc->buf;
 	fsc->sc = sc;
 
@@ -150,7 +231,18 @@ xchk_setup_fscounters(
 	if (error)
 		return error;
 
-	return xchk_trans_alloc(sc, 0);
+	/*
+	 * Pause all writer activity in the filesystem while we're scrubbing to
+	 * reduce the likelihood of background perturbations to the counters
+	 * throwing off our calculations.
+	 */
+	if (sc->flags & XCHK_TRY_HARDER) {
+		error = xchk_fscounters_freeze(sc);
+		if (error)
+			return error;
+	}
+
+	return xfs_trans_alloc_empty(sc->mp, &sc->tp);
 }
 
 /*
@@ -290,8 +382,7 @@ retry:
 	if (fsc->ifree > fsc->icount) {
 		if (tries--)
 			goto retry;
-		xchk_set_incomplete(sc);
-		return 0;
+		return -EDEADLOCK;
 	}
 
 	return 0;
@@ -367,6 +458,8 @@ xchk_fscount_count_frextents(
  * Otherwise, we /might/ have a problem.  If the change in the summations is
  * more than we want to tolerate, the filesystem is probably busy and we should
  * just send back INCOMPLETE and see if userspace will try again.
+ *
+ * If we're repairing then we require an exact match.
  */
 static inline bool
 xchk_fscount_within_range(
@@ -396,21 +489,7 @@ xchk_fscount_within_range(
 	if (expected >= min_value && expected <= max_value)
 		return true;
 
-	/*
-	 * If the difference between the two summations is too large, the fs
-	 * might just be busy and so we'll mark the scrub incomplete.  Return
-	 * true here so that we don't mark the counter corrupt.
-	 *
-	 * XXX: In the future when userspace can grant scrub permission to
-	 * quiesce the filesystem to solve the outsized variance problem, this
-	 * check should be moved up and the return code changed to signal to
-	 * userspace that we need quiesce permission.
-	 */
-	if (max_value - min_value >= XCHK_FSCOUNT_MIN_VARIANCE) {
-		xchk_set_incomplete(sc);
-		return true;
-	}
-
+	/* Everything else is bad. */
 	return false;
 }
 
@@ -422,6 +501,7 @@ xchk_fscounters(
 	struct xfs_mount	*mp = sc->mp;
 	struct xchk_fscounters	*fsc = sc->buf;
 	int64_t			icount, ifree, fdblocks, frextents;
+	bool			try_again = false;
 	int			error;
 
 	/* Snapshot the percpu counters. */
@@ -431,9 +511,26 @@ xchk_fscounters(
 	frextents = percpu_counter_sum(&mp->m_frextents);
 
 	/* No negative values, please! */
-	if (icount < 0 || ifree < 0 || fdblocks < 0 || frextents < 0)
+	if (icount < 0 || ifree < 0)
 		xchk_set_corrupt(sc);
 
+	/*
+	 * If the filesystem is not frozen, the counter summation calls above
+	 * can race with xfs_mod_freecounter, which subtracts a requested space
+	 * reservation from the counter and undoes the subtraction if that made
+	 * the counter go negative.  Therefore, it's possible to see negative
+	 * values here, and we should only flag that as a corruption if we
+	 * froze the fs.  This is much more likely to happen with frextents
+	 * since there are no reserved pools.
+	 */
+	if (fdblocks < 0 || frextents < 0) {
+		if (!fsc->frozen)
+			return -EDEADLOCK;
+
+		xchk_set_corrupt(sc);
+		return 0;
+	}
+
 	/* See if icount is obviously wrong. */
 	if (icount < fsc->icount_min || icount > fsc->icount_max)
 		xchk_set_corrupt(sc);
@@ -447,12 +544,6 @@ xchk_fscounters(
 		xchk_set_corrupt(sc);
 
 	/*
-	 * XXX: We can't quiesce percpu counter updates, so exit early.
-	 * This can be re-enabled when we gain exclusive freeze functionality.
-	 */
-	return 0;
-
-	/*
 	 * If ifree exceeds icount by more than the minimum variance then
 	 * something's probably wrong with the counters.
 	 */
@@ -463,8 +554,6 @@ xchk_fscounters(
 	error = xchk_fscount_aggregate_agcounts(sc, fsc);
 	if (!xchk_process_error(sc, 0, XFS_SB_BLOCK(mp), &error))
 		return error;
-	if (sc->sm->sm_flags & XFS_SCRUB_OFLAG_INCOMPLETE)
-		return 0;
 
 	/* Count the free extents counter for rt volumes. */
 	error = xchk_fscount_count_frextents(sc, fsc);
@@ -473,20 +562,45 @@ xchk_fscounters(
 	if (sc->sm->sm_flags & XFS_SCRUB_OFLAG_INCOMPLETE)
 		return 0;
 
-	/* Compare the in-core counters with whatever we counted. */
-	if (!xchk_fscount_within_range(sc, icount, &mp->m_icount, fsc->icount))
-		xchk_set_corrupt(sc);
+	/*
+	 * Compare the in-core counters with whatever we counted.  If the fs is
+	 * frozen, we treat the discrepancy as a corruption because the freeze
+	 * should have stabilized the counter values.  Otherwise, we need
+	 * userspace to call us back having granted us freeze permission.
+	 */
+	if (!xchk_fscount_within_range(sc, icount, &mp->m_icount,
+				fsc->icount)) {
+		if (fsc->frozen)
+			xchk_set_corrupt(sc);
+		else
+			try_again = true;
+	}
 
-	if (!xchk_fscount_within_range(sc, ifree, &mp->m_ifree, fsc->ifree))
-		xchk_set_corrupt(sc);
+	if (!xchk_fscount_within_range(sc, ifree, &mp->m_ifree, fsc->ifree)) {
+		if (fsc->frozen)
+			xchk_set_corrupt(sc);
+		else
+			try_again = true;
+	}
 
 	if (!xchk_fscount_within_range(sc, fdblocks, &mp->m_fdblocks,
-			fsc->fdblocks))
-		xchk_set_corrupt(sc);
+			fsc->fdblocks)) {
+		if (fsc->frozen)
+			xchk_set_corrupt(sc);
+		else
+			try_again = true;
+	}
 
 	if (!xchk_fscount_within_range(sc, frextents, &mp->m_frextents,
-			fsc->frextents))
-		xchk_set_corrupt(sc);
+			fsc->frextents)) {
+		if (fsc->frozen)
+			xchk_set_corrupt(sc);
+		else
+			try_again = true;
+	}
+
+	if (try_again)
+		return -EDEADLOCK;
 
 	return 0;
 }
diff --git a/fs/xfs/scrub/scrub.c b/fs/xfs/scrub/scrub.c
index 3d98f604765e..a0fffbcd022b 100644
--- a/fs/xfs/scrub/scrub.c
+++ b/fs/xfs/scrub/scrub.c
@@ -184,8 +184,10 @@ xchk_teardown(
 			xchk_irele(sc, sc->ip);
 		sc->ip = NULL;
 	}
-	if (sc->sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR)
+	if (sc->flags & XCHK_HAVE_FREEZE_PROT) {
+		sc->flags &= ~XCHK_HAVE_FREEZE_PROT;
 		mnt_drop_write_file(sc->file);
+	}
 	if (sc->buf) {
 		if (sc->buf_cleanup)
 			sc->buf_cleanup(sc->buf);
@@ -505,6 +507,8 @@ retry_op:
 		error = mnt_want_write_file(sc->file);
 		if (error)
 			goto out_sc;
+
+		sc->flags |= XCHK_HAVE_FREEZE_PROT;
 	}
 
 	/* Set up for the operation. */
diff --git a/fs/xfs/scrub/scrub.h b/fs/xfs/scrub/scrub.h
index e113f2f5c254..f8ba00e51ca9 100644
--- a/fs/xfs/scrub/scrub.h
+++ b/fs/xfs/scrub/scrub.h
@@ -106,6 +106,7 @@ struct xfs_scrub {
 
 /* XCHK state flags grow up from zero, XREP state flags grown down from 2^31 */
 #define XCHK_TRY_HARDER		(1U << 0)  /* can't get resources, try again */
+#define XCHK_HAVE_FREEZE_PROT	(1U << 1)  /* do we have freeze protection? */
 #define XCHK_FSGATES_DRAIN	(1U << 2)  /* defer ops draining enabled */
 #define XCHK_NEED_DRAIN		(1U << 3)  /* scrub needs to drain defer ops */
 #define XREP_ALREADY_FIXED	(1U << 31) /* checking our repair work */
diff --git a/fs/xfs/scrub/trace.h b/fs/xfs/scrub/trace.h
index b3894daeb86a..0b54f1a1cf0c 100644
--- a/fs/xfs/scrub/trace.h
+++ b/fs/xfs/scrub/trace.h
@@ -98,6 +98,7 @@ TRACE_DEFINE_ENUM(XFS_SCRUB_TYPE_FSCOUNTERS);
 
 #define XFS_SCRUB_STATE_STRINGS \
 	{ XCHK_TRY_HARDER,			"try_harder" }, \
+	{ XCHK_HAVE_FREEZE_PROT,		"nofreeze" }, \
 	{ XCHK_FSGATES_DRAIN,			"fsgates_drain" }, \
 	{ XCHK_NEED_DRAIN,			"need_drain" }, \
 	{ XREP_ALREADY_FIXED,			"already_fixed" }
@@ -693,6 +694,31 @@ TRACE_EVENT(xchk_fscounters_within_range,
 		  __entry->old_value)
 )
 
+DECLARE_EVENT_CLASS(xchk_fsfreeze_class,
+	TP_PROTO(struct xfs_scrub *sc, int error),
+	TP_ARGS(sc, error),
+	TP_STRUCT__entry(
+		__field(dev_t, dev)
+		__field(unsigned int, type)
+		__field(int, error)
+	),
+	TP_fast_assign(
+		__entry->dev = sc->mp->m_super->s_dev;
+		__entry->type = sc->sm->sm_type;
+		__entry->error = error;
+	),
+	TP_printk("dev %d:%d type %s error %d",
+		  MAJOR(__entry->dev), MINOR(__entry->dev),
+		  __print_symbolic(__entry->type, XFS_SCRUB_TYPE_STRINGS),
+		  __entry->error)
+);
+#define DEFINE_XCHK_FSFREEZE_EVENT(name) \
+DEFINE_EVENT(xchk_fsfreeze_class, name, \
+	TP_PROTO(struct xfs_scrub *sc, int error), \
+	TP_ARGS(sc, error))
+DEFINE_XCHK_FSFREEZE_EVENT(xchk_fsfreeze);
+DEFINE_XCHK_FSFREEZE_EVENT(xchk_fsthaw);
+
 TRACE_EVENT(xchk_refcount_incorrect,
 	TP_PROTO(struct xfs_perag *pag, const struct xfs_refcount_irec *irec,
 		 xfs_nlink_t seen),
diff --git a/fs/xfs/xfs_buf.c b/fs/xfs/xfs_buf.c
index 15d1e5a7c2d3..3b903f6bce98 100644
--- a/fs/xfs/xfs_buf.c
+++ b/fs/xfs/xfs_buf.c
@@ -1938,14 +1938,17 @@ void
 xfs_free_buftarg(
 	struct xfs_buftarg	*btp)
 {
+	struct block_device	*bdev = btp->bt_bdev;
+
 	unregister_shrinker(&btp->bt_shrinker);
 	ASSERT(percpu_counter_sum(&btp->bt_io_count) == 0);
 	percpu_counter_destroy(&btp->bt_io_count);
 	list_lru_destroy(&btp->bt_lru);
 
-	blkdev_issue_flush(btp->bt_bdev);
-	invalidate_bdev(btp->bt_bdev);
 	fs_put_dax(btp->bt_daxdev, btp->bt_mount);
+	/* the main block device is closed by kill_block_super */
+	if (bdev != btp->bt_mount->m_super->s_bdev)
+		blkdev_put(bdev, btp->bt_mount->m_super);
 
 	kmem_free(btp);
 }
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index 4b10edb2c972..c79eac048456 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -377,17 +377,6 @@ disable_dax:
 	return 0;
 }
 
-static void
-xfs_bdev_mark_dead(
-	struct block_device	*bdev)
-{
-	xfs_force_shutdown(bdev->bd_holder, SHUTDOWN_DEVICE_REMOVED);
-}
-
-static const struct blk_holder_ops xfs_holder_ops = {
-	.mark_dead		= xfs_bdev_mark_dead,
-};
-
 STATIC int
 xfs_blkdev_get(
 	xfs_mount_t		*mp,
@@ -396,8 +385,8 @@ xfs_blkdev_get(
 {
 	int			error = 0;
 
-	*bdevp = blkdev_get_by_path(name, BLK_OPEN_READ | BLK_OPEN_WRITE, mp,
-				    &xfs_holder_ops);
+	*bdevp = blkdev_get_by_path(name, BLK_OPEN_READ | BLK_OPEN_WRITE,
+				    mp->m_super, &fs_holder_ops);
 	if (IS_ERR(*bdevp)) {
 		error = PTR_ERR(*bdevp);
 		xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
@@ -407,31 +396,45 @@ xfs_blkdev_get(
 }
 
 STATIC void
-xfs_blkdev_put(
-	struct xfs_mount	*mp,
-	struct block_device	*bdev)
-{
-	if (bdev)
-		blkdev_put(bdev, mp);
-}
-
-STATIC void
-xfs_close_devices(
+xfs_shutdown_devices(
 	struct xfs_mount	*mp)
 {
+	/*
+	 * Udev is triggered whenever anyone closes a block device or unmounts
+	 * a file systemm on a block device.
+	 * The default udev rules invoke blkid to read the fs super and create
+	 * symlinks to the bdev under /dev/disk.  For this, it uses buffered
+	 * reads through the page cache.
+	 *
+	 * xfs_db also uses buffered reads to examine metadata.  There is no
+	 * coordination between xfs_db and udev, which means that they can run
+	 * concurrently.  Note there is no coordination between the kernel and
+	 * blkid either.
+	 *
+	 * On a system with 64k pages, the page cache can cache the superblock
+	 * and the root inode (and hence the root directory) with the same 64k
+	 * page.  If udev spawns blkid after the mkfs and the system is busy
+	 * enough that it is still running when xfs_db starts up, they'll both
+	 * read from the same page in the pagecache.
+	 *
+	 * The unmount writes updated inode metadata to disk directly.  The XFS
+	 * buffer cache does not use the bdev pagecache, so it needs to
+	 * invalidate that pagecache on unmount.  If the above scenario occurs,
+	 * the pagecache no longer reflects what's on disk, xfs_db reads the
+	 * stale metadata, and fails to find /a.  Most of the time this succeeds
+	 * because closing a bdev invalidates the page cache, but when processes
+	 * race, everyone loses.
+	 */
 	if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
-		struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
-
-		xfs_free_buftarg(mp->m_logdev_targp);
-		xfs_blkdev_put(mp, logdev);
+		blkdev_issue_flush(mp->m_logdev_targp->bt_bdev);
+		invalidate_bdev(mp->m_logdev_targp->bt_bdev);
 	}
 	if (mp->m_rtdev_targp) {
-		struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
-
-		xfs_free_buftarg(mp->m_rtdev_targp);
-		xfs_blkdev_put(mp, rtdev);
+		blkdev_issue_flush(mp->m_rtdev_targp->bt_bdev);
+		invalidate_bdev(mp->m_rtdev_targp->bt_bdev);
 	}
-	xfs_free_buftarg(mp->m_ddev_targp);
+	blkdev_issue_flush(mp->m_ddev_targp->bt_bdev);
+	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
 }
 
 /*
@@ -448,17 +451,24 @@ STATIC int
 xfs_open_devices(
 	struct xfs_mount	*mp)
 {
-	struct block_device	*ddev = mp->m_super->s_bdev;
+	struct super_block	*sb = mp->m_super;
+	struct block_device	*ddev = sb->s_bdev;
 	struct block_device	*logdev = NULL, *rtdev = NULL;
 	int			error;
 
 	/*
+	 * blkdev_put() can't be called under s_umount, see the comment
+	 * in get_tree_bdev() for more details
+	 */
+	up_write(&sb->s_umount);
+
+	/*
 	 * Open real time and log devices - order is important.
 	 */
 	if (mp->m_logname) {
 		error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
 		if (error)
-			return error;
+			goto out_relock;
 	}
 
 	if (mp->m_rtname) {
@@ -496,7 +506,10 @@ xfs_open_devices(
 		mp->m_logdev_targp = mp->m_ddev_targp;
 	}
 
-	return 0;
+	error = 0;
+out_relock:
+	down_write(&sb->s_umount);
+	return error;
 
  out_free_rtdev_targ:
 	if (mp->m_rtdev_targp)
@@ -504,11 +517,12 @@ xfs_open_devices(
  out_free_ddev_targ:
 	xfs_free_buftarg(mp->m_ddev_targp);
  out_close_rtdev:
-	xfs_blkdev_put(mp, rtdev);
+	 if (rtdev)
+		 blkdev_put(rtdev, sb);
  out_close_logdev:
 	if (logdev && logdev != ddev)
-		xfs_blkdev_put(mp, logdev);
-	return error;
+		blkdev_put(logdev, sb);
+	goto out_relock;
 }
 
 /*
@@ -758,6 +772,17 @@ static void
 xfs_mount_free(
 	struct xfs_mount	*mp)
 {
+	/*
+	 * Free the buftargs here because blkdev_put needs to be called outside
+	 * of sb->s_umount, which is held around the call to ->put_super.
+	 */
+	if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp)
+		xfs_free_buftarg(mp->m_logdev_targp);
+	if (mp->m_rtdev_targp)
+		xfs_free_buftarg(mp->m_rtdev_targp);
+	if (mp->m_ddev_targp)
+		xfs_free_buftarg(mp->m_ddev_targp);
+
 	kfree(mp->m_rtname);
 	kfree(mp->m_logname);
 	kmem_free(mp);
@@ -1133,10 +1158,6 @@ xfs_fs_put_super(
 {
 	struct xfs_mount	*mp = XFS_M(sb);
 
-	/* if ->fill_super failed, we have no mount to tear down */
-	if (!sb->s_fs_info)
-		return;
-
 	xfs_notice(mp, "Unmounting Filesystem %pU", &mp->m_sb.sb_uuid);
 	xfs_filestream_unmount(mp);
 	xfs_unmountfs(mp);
@@ -1147,10 +1168,7 @@ xfs_fs_put_super(
 	xfs_inodegc_free_percpu(mp);
 	xfs_destroy_percpu_counters(mp);
 	xfs_destroy_mount_workqueues(mp);
-	xfs_close_devices(mp);
-
-	sb->s_fs_info = NULL;
-	xfs_mount_free(mp);
+	xfs_shutdown_devices(mp);
 }
 
 static long
@@ -1492,7 +1510,7 @@ xfs_fs_fill_super(
 
 	error = xfs_fs_validate_params(mp);
 	if (error)
-		goto out_free_names;
+		return error;
 
 	sb_min_blocksize(sb, BBSIZE);
 	sb->s_xattr = xfs_xattr_handlers;
@@ -1519,11 +1537,11 @@ xfs_fs_fill_super(
 
 	error = xfs_open_devices(mp);
 	if (error)
-		goto out_free_names;
+		return error;
 
 	error = xfs_init_mount_workqueues(mp);
 	if (error)
-		goto out_close_devices;
+		goto out_shutdown_devices;
 
 	error = xfs_init_percpu_counters(mp);
 	if (error)
@@ -1737,11 +1755,8 @@ xfs_fs_fill_super(
 	xfs_destroy_percpu_counters(mp);
  out_destroy_workqueues:
 	xfs_destroy_mount_workqueues(mp);
- out_close_devices:
-	xfs_close_devices(mp);
- out_free_names:
-	sb->s_fs_info = NULL;
-	xfs_mount_free(mp);
+ out_shutdown_devices:
+	xfs_shutdown_devices(mp);
 	return error;
 
  out_unmount:
@@ -1934,7 +1949,8 @@ xfs_fs_reconfigure(
 	return 0;
 }
 
-static void xfs_fs_free(
+static void
+xfs_fs_free(
 	struct fs_context	*fc)
 {
 	struct xfs_mount	*mp = fc->s_fs_info;
@@ -2003,12 +2019,20 @@ static int xfs_init_fs_context(
 	return 0;
 }
 
+static void
+xfs_kill_sb(
+	struct super_block		*sb)
+{
+	kill_block_super(sb);
+	xfs_mount_free(XFS_M(sb));
+}
+
 static struct file_system_type xfs_fs_type = {
 	.owner			= THIS_MODULE,
 	.name			= "xfs",
 	.init_fs_context	= xfs_init_fs_context,
 	.parameters		= xfs_fs_parameters,
-	.kill_sb		= kill_block_super,
+	.kill_sb		= xfs_kill_sb,
 	.fs_flags		= FS_REQUIRES_DEV | FS_ALLOW_IDMAP | FS_MGTIME,
 };
 MODULE_ALIAS_FS("xfs");