fs: Provide helpers for manipulating sb->s_readonly_remount

Provide helpers to set and clear sb->s_readonly_remount including
appropriate memory barriers. Also use this opportunity to document what
the barriers pair with and why they are needed.

Suggested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Message-Id: <20230620112832.5158-1-jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>

4 files changed, 64 insertions, 21 deletions

diff --git a/fs/internal.h b/fs/internal.h
index bd3b2810a36b..b916b84809f3 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -121,6 +121,47 @@ extern bool mount_capable(struct fs_context *);
 int sb_init_dio_done_wq(struct super_block *sb);
 
 /*
+ * Prepare superblock for changing its read-only state (i.e., either remount
+ * read-write superblock read-only or vice versa). After this function returns
+ * mnt_is_readonly() will return true for any mount of the superblock if its
+ * caller is able to observe any changes done by the remount. This holds until
+ * sb_end_ro_state_change() is called.
+ */
+static inline void sb_start_ro_state_change(struct super_block *sb)
+{
+	WRITE_ONCE(sb->s_readonly_remount, 1);
+	/*
+	 * For RO->RW transition, the barrier pairs with the barrier in
+	 * mnt_is_readonly() making sure if mnt_is_readonly() sees SB_RDONLY
+	 * cleared, it will see s_readonly_remount set.
+	 * For RW->RO transition, the barrier pairs with the barrier in
+	 * __mnt_want_write() before the mnt_is_readonly() check. The barrier
+	 * makes sure if __mnt_want_write() sees MNT_WRITE_HOLD already
+	 * cleared, it will see s_readonly_remount set.
+	 */
+	smp_wmb();
+}
+
+/*
+ * Ends section changing read-only state of the superblock. After this function
+ * returns if mnt_is_readonly() returns false, the caller will be able to
+ * observe all the changes remount did to the superblock.
+ */
+static inline void sb_end_ro_state_change(struct super_block *sb)
+{
+	/*
+	 * This barrier provides release semantics that pairs with
+	 * the smp_rmb() acquire semantics in mnt_is_readonly().
+	 * This barrier pair ensure that when mnt_is_readonly() sees
+	 * 0 for sb->s_readonly_remount, it will also see all the
+	 * preceding flag changes that were made during the RO state
+	 * change.
+	 */
+	smp_wmb();
+	WRITE_ONCE(sb->s_readonly_remount, 0);
+}
+
+/*
  * open.c
  */
 struct open_flags {
diff --git a/fs/namespace.c b/fs/namespace.c
index 54847db5b819..5ba1eca6f720 100644
--- a/fs/namespace.c
+++ b/fs/namespace.c
@@ -309,9 +309,16 @@ static unsigned int mnt_get_writers(struct mount *mnt)
 
 static int mnt_is_readonly(struct vfsmount *mnt)
 {
-	if (mnt->mnt_sb->s_readonly_remount)
+	if (READ_ONCE(mnt->mnt_sb->s_readonly_remount))
 		return 1;
-	/* Order wrt setting s_flags/s_readonly_remount in do_remount() */
+	/*
+	 * The barrier pairs with the barrier in sb_start_ro_state_change()
+	 * making sure if we don't see s_readonly_remount set yet, we also will
+	 * not see any superblock / mount flag changes done by remount.
+	 * It also pairs with the barrier in sb_end_ro_state_change()
+	 * assuring that if we see s_readonly_remount already cleared, we will
+	 * see the values of superblock / mount flags updated by remount.
+	 */
 	smp_rmb();
 	return __mnt_is_readonly(mnt);
 }
@@ -364,9 +371,11 @@ int __mnt_want_write(struct vfsmount *m)
 		}
 	}
 	/*
-	 * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will
-	 * be set to match its requirements. So we must not load that until
-	 * MNT_WRITE_HOLD is cleared.
+	 * The barrier pairs with the barrier sb_start_ro_state_change() making
+	 * sure that if we see MNT_WRITE_HOLD cleared, we will also see
+	 * s_readonly_remount set (or even SB_RDONLY / MNT_READONLY flags) in
+	 * mnt_is_readonly() and bail in case we are racing with remount
+	 * read-only.
 	 */
 	smp_rmb();
 	if (mnt_is_readonly(m)) {
@@ -588,10 +597,8 @@ int sb_prepare_remount_readonly(struct super_block *sb)
 	if (!err && atomic_long_read(&sb->s_remove_count))
 		err = -EBUSY;
 
-	if (!err) {
-		sb->s_readonly_remount = 1;
-		smp_wmb();
-	}
+	if (!err)
+		sb_start_ro_state_change(sb);
 	list_for_each_entry(mnt, &sb->s_mounts, mnt_instance) {
 		if (mnt->mnt.mnt_flags & MNT_WRITE_HOLD)
 			mnt->mnt.mnt_flags &= ~MNT_WRITE_HOLD;
diff --git a/fs/super.c b/fs/super.c
index 5bf056087acc..5ee5da1fd498 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -944,8 +944,7 @@ int reconfigure_super(struct fs_context *fc)
 	 */
 	if (remount_ro) {
 		if (force) {
-			sb->s_readonly_remount = 1;
-			smp_wmb();
+			sb_start_ro_state_change(sb);
 		} else {
 			retval = sb_prepare_remount_readonly(sb);
 			if (retval)
@@ -953,12 +952,10 @@ int reconfigure_super(struct fs_context *fc)
 		}
 	} else if (remount_rw) {
 		/*
-		 * We set s_readonly_remount here to protect filesystem's
-		 * reconfigure code from writes from userspace until
-		 * reconfigure finishes.
+		 * Protect filesystem's reconfigure code from writes from
+		 * userspace until reconfigure finishes.
 		 */
-		sb->s_readonly_remount = 1;
-		smp_wmb();
+		sb_start_ro_state_change(sb);
 	}
 
 	if (fc->ops->reconfigure) {
@@ -974,9 +971,7 @@ int reconfigure_super(struct fs_context *fc)
 
 	WRITE_ONCE(sb->s_flags, ((sb->s_flags & ~fc->sb_flags_mask) |
 				 (fc->sb_flags & fc->sb_flags_mask)));
-	/* Needs to be ordered wrt mnt_is_readonly() */
-	smp_wmb();
-	sb->s_readonly_remount = 0;
+	sb_end_ro_state_change(sb);
 
 	/*
 	 * Some filesystems modify their metadata via some other path than the
@@ -991,7 +986,7 @@ int reconfigure_super(struct fs_context *fc)
 	return 0;
 
 cancel_readonly:
-	sb->s_readonly_remount = 0;
+	sb_end_ro_state_change(sb);
 	return retval;
 }
 
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 40bef9bf8749..4caac7fdc5d3 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1248,7 +1248,7 @@ struct super_block {
 	 */
 	atomic_long_t s_fsnotify_connectors;
 
-	/* Being remounted read-only */
+	/* Read-only state of the superblock is being changed */
 	int s_readonly_remount;
 
 	/* per-sb errseq_t for reporting writeback errors via syncfs */