Skip to contentSkip to conversation

saner calling conventions for ->d_automount()

RFC PATCH In review Al Viro 6 participants 6 messages Apr 24, 2025 – Apr 25, 2025

Subsystems (10)
lore
Share & tools
lore

Al Viro

author maintainer
#0 lore raw

RFC PATCH 1/1 saner calling conventions for ->d_automount()

Currently the calling conventions for ->d_automount() instances have an odd wart - returned new mount to be attached is expected to have refcount 2.

That kludge is intended to make sure that mark_mounts_for_expiry() called before we get around to attaching that new mount to the tree won't decide to take it out. finish_automount() drops the extra reference after it's done with attaching mount to the tree - or drops the reference twice in case of error. ->d_automount() instances have rather counterintuitive boilerplate in them.

There's a much simpler approach: have mark_mounts_for_expiry() skip the mounts that are yet to be mounted. And to hell with grabbing/dropping those extra references. Makes for simpler correctness analysis, at that...

---

Documentation/filesystems/porting.rst +7 -0
index 767b2927c762..749637231773 100644
@@ -1203,3 +1203,10 @@ should use d_drop();d_splice_alias() and return the result of the latter.
 If a positive dentry cannot be returned for some reason, in-kernel
 clients such as cachefiles, nfsd, smb/server may not perform ideally but
 will fail-safe.
+
+---
+
+**mandatory**
+
+Calling conventions for ->d_automount() have changed; we should *not* grab
+an extra reference to new mount - it should be returned with refcount 1.
Documentation/filesystems/vfs.rst +1 -3
index ae79c30b6c0c..cc0a58e96770 100644
@@ -1411,9 +1411,7 @@ defined:
 
 	If a vfsmount is returned, the caller will attempt to mount it
 	on the mountpoint and will remove the vfsmount from its
-	expiration list in the case of failure.  The vfsmount should be
-	returned with 2 refs on it to prevent automatic expiration - the
-	caller will clean up the additional ref.
+	expiration list in the case of failure.
 
 	This function is only used if DCACHE_NEED_AUTOMOUNT is set on
 	the dentry.  This is set by __d_instantiate() if S_AUTOMOUNT is
fs/afs/mntpt.c +0 -1
index 45cee6534122..9434a5399f2b 100644
@@ -189,7 +189,6 @@ struct vfsmount *afs_d_automount(struct path *path)
 	if (IS_ERR(newmnt))
 		return newmnt;
 
-	mntget(newmnt); /* prevent immediate expiration */
 	mnt_set_expiry(newmnt, &afs_vfsmounts);
 	queue_delayed_work(afs_wq, &afs_mntpt_expiry_timer,
 			   afs_mntpt_expiry_timeout * HZ);
fs/fuse/dir.c +0 -3
index 83ac192e7fdd..05d8584fd3b9 100644
@@ -319,9 +319,6 @@ static struct vfsmount *fuse_dentry_automount(struct path *path)
 
 	/* Create the submount */
 	mnt = fc_mount(fsc);
-	if (!IS_ERR(mnt))
-		mntget(mnt);
-
 	put_fs_context(fsc);
 	return mnt;
 }
fs/namespace.c +3 -6
index bbda516444ff..1807ccb1a52d 100644
@@ -3903,10 +3903,6 @@ int finish_automount(struct vfsmount *m, const struct path *path)
 		return PTR_ERR(m);
 
 	mnt = real_mount(m);
-	/* The new mount record should have at least 2 refs to prevent it being
-	 * expired before we get a chance to add it
-	 */
-	BUG_ON(mnt_get_count(mnt) < 2);
 
 	if (m->mnt_sb == path->mnt->mnt_sb &&
 	    m->mnt_root == dentry) {
@@ -3939,7 +3935,6 @@ int finish_automount(struct vfsmount *m, const struct path *path)
 	unlock_mount(mp);
 	if (unlikely(err))
 		goto discard;
-	mntput(m);
 	return 0;
 
 discard_locked:
@@ -3953,7 +3948,6 @@ int finish_automount(struct vfsmount *m, const struct path *path)
 		namespace_unlock();
 	}
 	mntput(m);
-	mntput(m);
 	return err;
 }
 
@@ -3990,11 +3984,14 @@ void mark_mounts_for_expiry(struct list_head *mounts)
 
 	/* extract from the expiration list every vfsmount that matches the
 	 * following criteria:
+	 * - already mounted
 	 * - only referenced by its parent vfsmount
 	 * - still marked for expiry (marked on the last call here; marks are
 	 *   cleared by mntput())
 	 */
 	list_for_each_entry_safe(mnt, next, mounts, mnt_expire) {
+		if (!is_mounted(&mnt->mnt))
+			continue;
 		if (!xchg(&mnt->mnt_expiry_mark, 1) ||
 			propagate_mount_busy(mnt, 1))
 			continue;
fs/nfs/namespace.c +0 -1
index 973aed9cc5fe..7f1ec9c67ff2 100644
@@ -195,7 +195,6 @@ struct vfsmount *nfs_d_automount(struct path *path)
 	if (IS_ERR(mnt))
 		goto out_fc;
 
-	mntget(mnt); /* prevent immediate expiration */
 	if (timeout <= 0)
 		goto out_fc;
 
fs/smb/client/namespace.c +0 -1
index e3f9213131c4..778daf11f1db 100644
@@ -283,7 +283,6 @@ struct vfsmount *cifs_d_automount(struct path *path)
 		return newmnt;
 	}
 
-	mntget(newmnt); /* prevent immediate expiration */
 	mnt_set_expiry(newmnt, &cifs_automount_list);
 	schedule_delayed_work(&cifs_automount_task,
 			      cifs_mountpoint_expiry_timeout);
kernel/trace/trace.c +0 -2
index 8ddf6b17215c..fa488721019f 100644
@@ -10085,8 +10085,6 @@ static struct vfsmount *trace_automount(struct dentry *mntpt, void *ingore)
 	put_filesystem(type);
 	if (IS_ERR(mnt))
 		return NULL;
-	mntget(mnt);
-
 	return mnt;
 }
 
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 11 lines · Al Viro

Currently the calling conventions for ->d_automount() instances have an odd wart - returned new mount to be attached is expected to have refcount 2.

That kludge is intended to make sure that mark_mounts_for_expiry() called before we get around to attaching that new mount to the tree won't decide to take it out. finish_automount() drops the extra reference after it's done with attaching mount to the tree - or drops the reference twice in case of error. ->d_automount() instances have rather counterintuitive boilerplate in them.

Quoted 6 lines

There's a much simpler approach: have mark_mounts_for_expiry() skip the mounts that are yet to be mounted. And to hell with grabbing/dropping those extra references. Makes for simpler correctness analysis, at that...

---

Looks good!

maintainer Reviewed-by on v1 1/1
#2 lore raw
In reply to Al Viro
Quoted 149 lines · Al Viro

Currently the calling conventions for ->d_automount() instances have an odd wart - returned new mount to be attached is expected to have refcount 2.

That kludge is intended to make sure that mark_mounts_for_expiry() called before we get around to attaching that new mount to the tree won't decide to take it out. finish_automount() drops the extra reference after it's done with attaching mount to the tree - or drops the reference twice in case of error. ->d_automount() instances have rather counterintuitive boilerplate in them.

There's a much simpler approach: have mark_mounts_for_expiry() skip the mounts that are yet to be mounted. And to hell with grabbing/dropping those extra references. Makes for simpler correctness analysis, at that...

---

Documentation/filesystems/porting.rst +7 -0
index 767b2927c762..749637231773 100644
@@ -1203,3 +1203,10 @@ should use d_drop();d_splice_alias() and return the result of the latter.
 If a positive dentry cannot be returned for some reason, in-kernel
 clients such as cachefiles, nfsd, smb/server may not perform ideally but
 will fail-safe.
+
+---
+
+**mandatory**
+
+Calling conventions for ->d_automount() have changed; we should *not* grab
+an extra reference to new mount - it should be returned with refcount 1.
Documentation/filesystems/vfs.rst +1 -3
index ae79c30b6c0c..cc0a58e96770 100644
@@ -1411,9 +1411,7 @@ defined:
 
 	If a vfsmount is returned, the caller will attempt to mount it
 	on the mountpoint and will remove the vfsmount from its
-	expiration list in the case of failure.  The vfsmount should be
-	returned with 2 refs on it to prevent automatic expiration - the
-	caller will clean up the additional ref.
+	expiration list in the case of failure.
 
 	This function is only used if DCACHE_NEED_AUTOMOUNT is set on
 	the dentry.  This is set by __d_instantiate() if S_AUTOMOUNT is
fs/afs/mntpt.c +0 -1
index 45cee6534122..9434a5399f2b 100644
@@ -189,7 +189,6 @@ struct vfsmount *afs_d_automount(struct path *path)
 	if (IS_ERR(newmnt))
 		return newmnt;
 
-	mntget(newmnt); /* prevent immediate expiration */
 	mnt_set_expiry(newmnt, &afs_vfsmounts);
 	queue_delayed_work(afs_wq, &afs_mntpt_expiry_timer,
 			   afs_mntpt_expiry_timeout * HZ);
fs/fuse/dir.c +0 -3
index 83ac192e7fdd..05d8584fd3b9 100644
@@ -319,9 +319,6 @@ static struct vfsmount *fuse_dentry_automount(struct path *path)
 
 	/* Create the submount */
 	mnt = fc_mount(fsc);
-	if (!IS_ERR(mnt))
-		mntget(mnt);
-
 	put_fs_context(fsc);
 	return mnt;
 }
fs/namespace.c +3 -6
index bbda516444ff..1807ccb1a52d 100644
@@ -3903,10 +3903,6 @@ int finish_automount(struct vfsmount *m, const struct path *path)
 		return PTR_ERR(m);
 
 	mnt = real_mount(m);
-	/* The new mount record should have at least 2 refs to prevent it being
-	 * expired before we get a chance to add it
-	 */
-	BUG_ON(mnt_get_count(mnt) < 2);
 
 	if (m->mnt_sb == path->mnt->mnt_sb &&
 	    m->mnt_root == dentry) {
@@ -3939,7 +3935,6 @@ int finish_automount(struct vfsmount *m, const struct path *path)
 	unlock_mount(mp);
 	if (unlikely(err))
 		goto discard;
-	mntput(m);
 	return 0;
 
 discard_locked:
@@ -3953,7 +3948,6 @@ int finish_automount(struct vfsmount *m, const struct path *path)
 		namespace_unlock();
 	}
 	mntput(m);
-	mntput(m);
 	return err;
 }
 
@@ -3990,11 +3984,14 @@ void mark_mounts_for_expiry(struct list_head *mounts)
 
 	/* extract from the expiration list every vfsmount that matches the
 	 * following criteria:
+	 * - already mounted
 	 * - only referenced by its parent vfsmount
 	 * - still marked for expiry (marked on the last call here; marks are
 	 *   cleared by mntput())
 	 */
 	list_for_each_entry_safe(mnt, next, mounts, mnt_expire) {
+		if (!is_mounted(&mnt->mnt))
+			continue;
 		if (!xchg(&mnt->mnt_expiry_mark, 1) ||
 			propagate_mount_busy(mnt, 1))
 			continue;
fs/nfs/namespace.c +0 -1
index 973aed9cc5fe..7f1ec9c67ff2 100644
@@ -195,7 +195,6 @@ struct vfsmount *nfs_d_automount(struct path *path)
 	if (IS_ERR(mnt))
 		goto out_fc;
 
-	mntget(mnt); /* prevent immediate expiration */
 	if (timeout <= 0)
 		goto out_fc;
 
fs/smb/client/namespace.c +0 -1
index e3f9213131c4..778daf11f1db 100644
@@ -283,7 +283,6 @@ struct vfsmount *cifs_d_automount(struct path *path)
 		return newmnt;
 	}
 
-	mntget(newmnt); /* prevent immediate expiration */
 	mnt_set_expiry(newmnt, &cifs_automount_list);
 	schedule_delayed_work(&cifs_automount_task,
 			      cifs_mountpoint_expiry_timeout);
kernel/trace/trace.c +0 -0
index 8ddf6b17215c..fa488721019f 100644

kernel/trace/trace.c static struct vfsmount *trace_automount(struct dentry *mntpt, void *ingore) · line 10090

Quoted 8 lines
kernel/trace/trace.c +0 -2
@@ -10085,8 +10085,6 @@ static struct vfsmount *trace_automount(struct dentry *mntpt, void *ingore)
 	put_filesystem(type);
 	if (IS_ERR(mnt))
 		return NULL;
-	mntget(mnt);
-
 	return mnt;
 }
 

Much cleaner:

Reviewed-by on v1 1/1
#3 lore raw
In reply to Al Viro
Quoted 11 lines · Al Viro

Currently the calling conventions for ->d_automount() instances have an odd wart - returned new mount to be attached is expected to have refcount 2.

That kludge is intended to make sure that mark_mounts_for_expiry() called before we get around to attaching that new mount to the tree won't decide to take it out. finish_automount() drops the extra reference after it's done with attaching mount to the tree - or drops the reference twice in case of error. ->d_automount() instances have rather counterintuitive boilerplate in them.

Quoted 6 lines

There's a much simpler approach: have mark_mounts_for_expiry() skip the mounts that are yet to be mounted. And to hell with grabbing/dropping those extra references. Makes for simpler correctness analysis, at that...

---

maintainer Acked-by Tested-by on v1 1/1
#4 lore raw
In reply to Al Viro
Quoted 11 lines · Al Viro

Currently the calling conventions for ->d_automount() instances have an odd wart - returned new mount to be attached is expected to have refcount 2.

That kludge is intended to make sure that mark_mounts_for_expiry() called before we get around to attaching that new mount to the tree won't decide to take it out. finish_automount() drops the extra reference after it's done with attaching mount to the tree - or drops the reference twice in case of error. ->d_automount() instances have rather counterintuitive boilerplate in them.

Quoted 5 lines

There's a much simpler approach: have mark_mounts_for_expiry() skip the mounts that are yet to be mounted. And to hell with grabbing/dropping those extra references. Makes for simpler correctness analysis, at that...

You can test this by installing the kafs-client package, telling systemd to start afs.mount and then doing "ls /afs/openafs.org/doc", say.

maintainer Acked-by on v1 1/1
#5 lore raw
In reply to Al Viro

kernel/trace/trace.c static struct vfsmount *trace_automount(struct dentry *mntpt, void *ingore) · line 10089

Quoted 12 lines · On Thu, 24 Apr 2025 07:08:45 +0100 Al Viro <[email protected]> wrote:
kernel/trace/trace.c +0 -2
index 8ddf6b17215c..fa488721019f 100644
@@ -10085,8 +10085,6 @@ static struct vfsmount *trace_automount(struct dentry *mntpt, void *ingore)
 	put_filesystem(type);
 	if (IS_ERR(mnt))
 		return NULL;
-	mntget(mnt);
-
 	return mnt;
 }

-- Steve