fanotify: permissions and blocking

This is the backend work needed for fanotify to support the new
FS_OPEN_PERM and FS_ACCESS_PERM fsnotify events.  This is done using the
new fsnotify secondary queue.  No userspace interface is provided actually
respond to or request these events.
Signed-off-by: Eric Paris <eparis@redhat.com>

fs/notify/fanotify/Kconfig

@@ -10,3 +10,17 @@ config FANOTIFY
 	   the event.
 
 	   If unsure, say Y.
+
+config FANOTIFY_ACCESS_PERMISSIONS
+	bool "fanotify permissions checking"
+	depends on FANOTIFY
+	depends on SECURITY
+	default n
+	---help---
+	   Say Y here is you want fanotify listeners to be able to make permissions
+	   decisions concerning filesystem events.  This is used by some fanotify
+	   listeners which need to scan files before allowing the system access to
+	   use those files.  This is used by some anti-malware vendors and by some
+	   hierarchical storage managent systems.
+
+	   If unsure, say N.

fs/notify/fanotify/fanotify.c

@@ -2,9 +2,12 @@
 #include <linux/fdtable.h>
 #include <linux/fsnotify_backend.h>
 #include <linux/init.h>
+#include <linux/jiffies.h>
 #include <linux/kernel.h> /* UINT_MAX */
 #include <linux/mount.h>
+#include <linux/sched.h>
 #include <linux/types.h>
+#include <linux/wait.h>
 
 static bool should_merge(struct fsnotify_event *old, struct fsnotify_event *new)
 {
@@ -88,10 +91,37 @@ static int fanotify_merge(struct list_head *list,
 	return ret;
 }
 
+#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
+static int fanotify_get_response_from_access(struct fsnotify_group *group,
+					     struct fsnotify_event *event)
+{
+	int ret;
+
+	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
+
+	wait_event(group->fanotify_data.access_waitq, event->response);
+
+	/* userspace responded, convert to something usable */
+	spin_lock(&event->lock);
+	switch (event->response) {
+	case FAN_ALLOW:
+		ret = 0;
+		break;
+	case FAN_DENY:
+	default:
+		ret = -EPERM;
+	}
+	event->response = 0;
+	spin_unlock(&event->lock);
+
+	return ret;
+}
+#endif
+
 static int fanotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
 {
 	int ret;
-	struct fsnotify_event *used_event;
+	struct fsnotify_event *notify_event = NULL;
 
 	BUILD_BUG_ON(FAN_ACCESS != FS_ACCESS);
 	BUILD_BUG_ON(FAN_MODIFY != FS_MODIFY);
@@ -100,15 +130,31 @@ static int fanotify_handle_event(struct fsnotify_group *group, struct fsnotify_e
 	BUILD_BUG_ON(FAN_OPEN != FS_OPEN);
 	BUILD_BUG_ON(FAN_EVENT_ON_CHILD != FS_EVENT_ON_CHILD);
 	BUILD_BUG_ON(FAN_Q_OVERFLOW != FS_Q_OVERFLOW);
+	BUILD_BUG_ON(FAN_OPEN_PERM != FS_OPEN_PERM);
+	BUILD_BUG_ON(FAN_ACCESS_PERM != FS_ACCESS_PERM);
 
 	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
 
-	ret = fsnotify_add_notify_event(group, event, NULL, fanotify_merge, (void **)&used_event);
+	ret = fsnotify_add_notify_event(group, event, NULL, fanotify_merge,
+					(void **)&notify_event);
 	/* -EEXIST means this event was merged with another, not that it was an error */
 	if (ret == -EEXIST)
 		ret = 0;
-	if (used_event)
-		fsnotify_put_event(used_event);
+	if (ret)
+		goto out;
+
+#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
+	if (event->mask & FAN_ALL_PERM_EVENTS) {
+		/* if we merged we need to wait on the new event */
+		if (notify_event)
+			event = notify_event;
+		ret = fanotify_get_response_from_access(group, event);
+	}
+#endif
+
+out:
+	if (notify_event)
+		fsnotify_put_event(notify_event);
 	return ret;
 }
 

fs/notify/fanotify/fanotify_user.c

@@ -482,6 +482,11 @@ SYSCALL_DEFINE3(fanotify_init, unsigned int, flags, unsigned int, event_f_flags,
 		return PTR_ERR(group);
 
 	group->priority = priority;
+#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
+	mutex_init(&group->fanotify_data.access_mutex);
+	init_waitqueue_head(&group->fanotify_data.access_waitq);
+	INIT_LIST_HEAD(&group->fanotify_data.access_list);
+#endif
 
 	fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags);
 	if (fd < 0)

include/linux/fanotify.h

@@ -15,6 +15,9 @@
 /* FIXME currently Q's have no limit.... */
 #define FAN_Q_OVERFLOW		0x00004000	/* Event queued overflowed */
 
+#define FAN_OPEN_PERM		0x00010000	/* File open in perm check */
+#define FAN_ACCESS_PERM		0x00020000	/* File accessed in perm check */
+
 /* helper events */
 #define FAN_CLOSE		(FAN_CLOSE_WRITE | FAN_CLOSE_NOWRITE) /* close */
 
@@ -52,7 +55,14 @@
 			FAN_CLOSE |\
 			FAN_OPEN)
 
+/*
+ * All events which require a permission response from userspace
+ */
+#define FAN_ALL_PERM_EVENTS (FAN_OPEN_PERM |\
+			     FAN_ACCESS_PERM)
+
 #define FAN_ALL_OUTGOING_EVENTS	(FAN_ALL_EVENTS |\
+				 FAN_ALL_PERM_EVENTS |\
 				 FAN_Q_OVERFLOW)
 
 #define FANOTIFY_METADATA_VERSION	1
@@ -65,6 +75,10 @@ struct fanotify_event_metadata {
 	__s64 pid;
 } __attribute__ ((packed));
 
+/* Legit userspace responses to a _PERM event */
+#define FAN_ALLOW	0x01
+#define FAN_DENY	0x02
+
 /* Helper functions to deal with fanotify_event_metadata buffers */
 #define FAN_EVENT_METADATA_LEN (sizeof(struct fanotify_event_metadata))
 
@@ -78,5 +92,9 @@ struct fanotify_event_metadata {
 
 #ifdef __KERNEL__
 
+struct fanotify_wait {
+	struct fsnotify_event *event;
+	__s32 fd;
+};
 #endif /* __KERNEL__ */
 #endif /* _LINUX_FANOTIFY_H */

include/linux/fsnotify_backend.h

@@ -159,6 +159,14 @@ struct fsnotify_group {
 			struct fasync_struct    *fa;    /* async notification */
 			struct user_struct      *user;
 		} inotify_data;
+#endif
+#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
+		struct fanotify_group_private_data {
+			/* allows a group to block waiting for a userspace response */
+			struct mutex access_mutex;
+			struct list_head access_list;
+			wait_queue_head_t access_waitq;
+		} fanotify_data;
 #endif
 	};
 };
@@ -227,6 +235,10 @@ struct fsnotify_event {
 	size_t name_len;
 	struct pid *tgid;
 
+#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
+	__u32 response;	/* userspace answer to question */
+#endif /* CONFIG_FANOTIFY_ACCESS_PERMISSIONS */
+
 	struct list_head private_data_list;	/* groups can store private data here */
 };