fanotify: implement "evictable" inode marks

When an inode mark is created with flag FAN_MARK_EVICTABLE, it will not
pin the marked inode to inode cache, so when inode is evicted from cache
due to memory pressure, the mark will be lost.

When an inode mark with flag FAN_MARK_EVICATBLE is updated without using
this flag, the marked inode is pinned to inode cache.

When an inode mark is updated with flag FAN_MARK_EVICTABLE but an
existing mark already has the inode pinned, the mark update fails with
error EEXIST.

Evictable inode marks can be used to setup inode marks with ignored mask
to suppress events from uninteresting files or directories in a lazy
manner, upon receiving the first event, without having to iterate all
the uninteresting files or directories before hand.

The evictbale inode mark feature allows performing this lazy marks setup
without exhausting the system memory with pinned inodes.

This change does not enable the feature yet.

Link: https://lore.kernel.org/linux-fsdevel/CAOQ4uxiRDpuS=2uA6+ZUM7yG9vVU-u212tkunBmSnP_u=mkv=Q@mail.gmail.com/
Link: https://lore.kernel.org/r/20220422120327.3459282-15-amir73il@gmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>

1 files changed, 36 insertions, 2 deletions

diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index 4005ee8e6e2c..ae36138afead 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -1084,6 +1084,7 @@ static int fanotify_remove_inode_mark(struct fsnotify_group *group,
 static bool fanotify_mark_update_flags(struct fsnotify_mark *fsn_mark,
 				       unsigned int fan_flags)
 {
+	bool want_iref = !(fan_flags & FAN_MARK_EVICTABLE);
 	bool recalc = false;
 
 	/*
@@ -1099,7 +1100,18 @@ static bool fanotify_mark_update_flags(struct fsnotify_mark *fsn_mark,
 			recalc = true;
 	}
 
-	return recalc;
+	if (fsn_mark->connector->type != FSNOTIFY_OBJ_TYPE_INODE ||
+	    want_iref == !(fsn_mark->flags & FSNOTIFY_MARK_FLAG_NO_IREF))
+		return recalc;
+
+	/*
+	 * NO_IREF may be removed from a mark, but not added.
+	 * When removed, fsnotify_recalc_mask() will take the inode ref.
+	 */
+	WARN_ON_ONCE(!want_iref);
+	fsn_mark->flags &= ~FSNOTIFY_MARK_FLAG_NO_IREF;
+
+	return true;
 }
 
 static bool fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark,
@@ -1125,6 +1137,7 @@ static bool fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark,
 static struct fsnotify_mark *fanotify_add_new_mark(struct fsnotify_group *group,
 						   fsnotify_connp_t *connp,
 						   unsigned int obj_type,
+						   unsigned int fan_flags,
 						   __kernel_fsid_t *fsid)
 {
 	struct ucounts *ucounts = group->fanotify_data.ucounts;
@@ -1147,6 +1160,9 @@ static struct fsnotify_mark *fanotify_add_new_mark(struct fsnotify_group *group,
 	}
 
 	fsnotify_init_mark(mark, group);
+	if (fan_flags & FAN_MARK_EVICTABLE)
+		mark->flags |= FSNOTIFY_MARK_FLAG_NO_IREF;
+
 	ret = fsnotify_add_mark_locked(mark, connp, obj_type, 0, fsid);
 	if (ret) {
 		fsnotify_put_mark(mark);
@@ -1183,7 +1199,8 @@ static int fanotify_add_mark(struct fsnotify_group *group,
 	mutex_lock(&group->mark_mutex);
 	fsn_mark = fsnotify_find_mark(connp, group);
 	if (!fsn_mark) {
-		fsn_mark = fanotify_add_new_mark(group, connp, obj_type, fsid);
+		fsn_mark = fanotify_add_new_mark(group, connp, obj_type,
+						 fan_flags, fsid);
 		if (IS_ERR(fsn_mark)) {
 			mutex_unlock(&group->mark_mutex);
 			return PTR_ERR(fsn_mark);
@@ -1191,6 +1208,15 @@ static int fanotify_add_mark(struct fsnotify_group *group,
 	}
 
 	/*
+	 * Non evictable mark cannot be downgraded to evictable mark.
+	 */
+	if (fan_flags & FAN_MARK_EVICTABLE &&
+	    !(fsn_mark->flags & FSNOTIFY_MARK_FLAG_NO_IREF)) {
+		ret = -EEXIST;
+		goto out;
+	}
+
+	/*
 	 * Error events are pre-allocated per group, only if strictly
 	 * needed (i.e. FAN_FS_ERROR was requested).
 	 */
@@ -1602,6 +1628,14 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask,
 		goto fput_and_out;
 
 	/*
+	 * Evictable is only relevant for inode marks, because only inode object
+	 * can be evicted on memory pressure.
+	 */
+	if (flags & FAN_MARK_EVICTABLE &&
+	     mark_type != FAN_MARK_INODE)
+		goto fput_and_out;
+
+	/*
 	 * Events that do not carry enough information to report
 	 * event->fd require a group that supports reporting fid.  Those
 	 * events are not supported on a mount mark, because they do not