1 files changed, 186 insertions, 33 deletions

diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index 9e0c1afac8bd..71fefb30e015 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -27,8 +27,61 @@
 #include "fanotify.h"
 
 #define FANOTIFY_DEFAULT_MAX_EVENTS	16384
-#define FANOTIFY_DEFAULT_MAX_MARKS	8192
-#define FANOTIFY_DEFAULT_MAX_LISTENERS	128
+#define FANOTIFY_OLD_DEFAULT_MAX_MARKS	8192
+#define FANOTIFY_DEFAULT_MAX_GROUPS	128
+
+/*
+ * Legacy fanotify marks limits (8192) is per group and we introduced a tunable
+ * limit of marks per user, similar to inotify.  Effectively, the legacy limit
+ * of fanotify marks per user is <max marks per group> * <max groups per user>.
+ * This default limit (1M) also happens to match the increased limit of inotify
+ * max_user_watches since v5.10.
+ */
+#define FANOTIFY_DEFAULT_MAX_USER_MARKS	\
+	(FANOTIFY_OLD_DEFAULT_MAX_MARKS * FANOTIFY_DEFAULT_MAX_GROUPS)
+
+/*
+ * Most of the memory cost of adding an inode mark is pinning the marked inode.
+ * The size of the filesystem inode struct is not uniform across filesystems,
+ * so double the size of a VFS inode is used as a conservative approximation.
+ */
+#define INODE_MARK_COST	(2 * sizeof(struct inode))
+
+/* configurable via /proc/sys/fs/fanotify/ */
+static int fanotify_max_queued_events __read_mostly;
+
+#ifdef CONFIG_SYSCTL
+
+#include <linux/sysctl.h>
+
+struct ctl_table fanotify_table[] = {
+	{
+		.procname	= "max_user_groups",
+		.data	= &init_user_ns.ucount_max[UCOUNT_FANOTIFY_GROUPS],
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= SYSCTL_ZERO,
+	},
+	{
+		.procname	= "max_user_marks",
+		.data	= &init_user_ns.ucount_max[UCOUNT_FANOTIFY_MARKS],
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= SYSCTL_ZERO,
+	},
+	{
+		.procname	= "max_queued_events",
+		.data		= &fanotify_max_queued_events,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= SYSCTL_ZERO
+	},
+	{ }
+};
+#endif /* CONFIG_SYSCTL */
 
 /*
  * All flags that may be specified in parameter event_f_flags of fanotify_init.
@@ -90,6 +143,23 @@ static int fanotify_event_info_len(unsigned int fid_mode,
 }
 
 /*
+ * Remove an hashed event from merge hash table.
+ */
+static void fanotify_unhash_event(struct fsnotify_group *group,
+				  struct fanotify_event *event)
+{
+	assert_spin_locked(&group->notification_lock);
+
+	pr_debug("%s: group=%p event=%p bucket=%u\n", __func__,
+		 group, event, fanotify_event_hash_bucket(group, event));
+
+	if (WARN_ON_ONCE(hlist_unhashed(&event->merge_list)))
+		return;
+
+	hlist_del_init(&event->merge_list);
+}
+
+/*
  * Get an fanotify notification event if one exists and is small
  * enough to fit in "count". Return an error pointer if the count
  * is not large enough. When permission event is dequeued, its state is
@@ -100,26 +170,34 @@ static struct fanotify_event *get_one_event(struct fsnotify_group *group,
 {
 	size_t event_size = FAN_EVENT_METADATA_LEN;
 	struct fanotify_event *event = NULL;
+	struct fsnotify_event *fsn_event;
 	unsigned int fid_mode = FAN_GROUP_FLAG(group, FANOTIFY_FID_BITS);
 
 	pr_debug("%s: group=%p count=%zd\n", __func__, group, count);
 
 	spin_lock(&group->notification_lock);
-	if (fsnotify_notify_queue_is_empty(group))
+	fsn_event = fsnotify_peek_first_event(group);
+	if (!fsn_event)
 		goto out;
 
-	if (fid_mode) {
-		event_size += fanotify_event_info_len(fid_mode,
-			FANOTIFY_E(fsnotify_peek_first_event(group)));
-	}
+	event = FANOTIFY_E(fsn_event);
+	if (fid_mode)
+		event_size += fanotify_event_info_len(fid_mode, event);
 
 	if (event_size > count) {
 		event = ERR_PTR(-EINVAL);
 		goto out;
 	}
-	event = FANOTIFY_E(fsnotify_remove_first_event(group));
+
+	/*
+	 * Held the notification_lock the whole time, so this is the
+	 * same event we peeked above.
+	 */
+	fsnotify_remove_first_event(group);
 	if (fanotify_is_perm_event(event->mask))
 		FANOTIFY_PERM(event)->state = FAN_EVENT_REPORTED;
+	if (fanotify_is_hashed_event(event->mask))
+		fanotify_unhash_event(group, event);
 out:
 	spin_unlock(&group->notification_lock);
 	return event;
@@ -341,6 +419,14 @@ static ssize_t copy_event_to_user(struct fsnotify_group *group,
 	metadata.reserved = 0;
 	metadata.mask = event->mask & FANOTIFY_OUTGOING_EVENTS;
 	metadata.pid = pid_vnr(event->pid);
+	/*
+	 * For an unprivileged listener, event->pid can be used to identify the
+	 * events generated by the listener process itself, without disclosing
+	 * the pids of other processes.
+	 */
+	if (!capable(CAP_SYS_ADMIN) &&
+	    task_tgid(current) != event->pid)
+		metadata.pid = 0;
 
 	if (path && path->mnt && path->dentry) {
 		fd = create_fd(group, path, &f);
@@ -573,6 +659,7 @@ static ssize_t fanotify_write(struct file *file, const char __user *buf, size_t
 static int fanotify_release(struct inode *ignored, struct file *file)
 {
 	struct fsnotify_group *group = file->private_data;
+	struct fsnotify_event *fsn_event;
 
 	/*
 	 * Stop new events from arriving in the notification queue. since
@@ -601,13 +688,12 @@ static int fanotify_release(struct inode *ignored, struct file *file)
 	 * dequeue them and set the response. They will be freed once the
 	 * response is consumed and fanotify_get_response() returns.
 	 */
-	while (!fsnotify_notify_queue_is_empty(group)) {
-		struct fanotify_event *event;
+	while ((fsn_event = fsnotify_remove_first_event(group))) {
+		struct fanotify_event *event = FANOTIFY_E(fsn_event);
 
-		event = FANOTIFY_E(fsnotify_remove_first_event(group));
 		if (!(event->mask & FANOTIFY_PERM_EVENTS)) {
 			spin_unlock(&group->notification_lock);
-			fsnotify_destroy_event(group, &event->fse);
+			fsnotify_destroy_event(group, fsn_event);
 		} else {
 			finish_permission_event(group, FANOTIFY_PERM(event),
 						FAN_ALLOW);
@@ -822,24 +908,38 @@ static struct fsnotify_mark *fanotify_add_new_mark(struct fsnotify_group *group,
 						   unsigned int type,
 						   __kernel_fsid_t *fsid)
 {
+	struct ucounts *ucounts = group->fanotify_data.ucounts;
 	struct fsnotify_mark *mark;
 	int ret;
 
-	if (atomic_read(&group->num_marks) > group->fanotify_data.max_marks)
+	/*
+	 * Enforce per user marks limits per user in all containing user ns.
+	 * A group with FAN_UNLIMITED_MARKS does not contribute to mark count
+	 * in the limited groups account.
+	 */
+	if (!FAN_GROUP_FLAG(group, FAN_UNLIMITED_MARKS) &&
+	    !inc_ucount(ucounts->ns, ucounts->uid, UCOUNT_FANOTIFY_MARKS))
 		return ERR_PTR(-ENOSPC);
 
 	mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
-	if (!mark)
-		return ERR_PTR(-ENOMEM);
+	if (!mark) {
+		ret = -ENOMEM;
+		goto out_dec_ucounts;
+	}
 
 	fsnotify_init_mark(mark, group);
 	ret = fsnotify_add_mark_locked(mark, connp, type, 0, fsid);
 	if (ret) {
 		fsnotify_put_mark(mark);
-		return ERR_PTR(ret);
+		goto out_dec_ucounts;
 	}
 
 	return mark;
+
+out_dec_ucounts:
+	if (!FAN_GROUP_FLAG(group, FAN_UNLIMITED_MARKS))
+		dec_ucount(ucounts, UCOUNT_FANOTIFY_MARKS);
+	return ERR_PTR(ret);
 }
 
 
@@ -919,20 +1019,41 @@ static struct fsnotify_event *fanotify_alloc_overflow_event(void)
 	return &oevent->fse;
 }
 
+static struct hlist_head *fanotify_alloc_merge_hash(void)
+{
+	struct hlist_head *hash;
+
+	hash = kmalloc(sizeof(struct hlist_head) << FANOTIFY_HTABLE_BITS,
+		       GFP_KERNEL_ACCOUNT);
+	if (!hash)
+		return NULL;
+
+	__hash_init(hash, FANOTIFY_HTABLE_SIZE);
+
+	return hash;
+}
+
 /* fanotify syscalls */
 SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 {
 	struct fsnotify_group *group;
 	int f_flags, fd;
-	struct user_struct *user;
 	unsigned int fid_mode = flags & FANOTIFY_FID_BITS;
 	unsigned int class = flags & FANOTIFY_CLASS_BITS;
 
 	pr_debug("%s: flags=%x event_f_flags=%x\n",
 		 __func__, flags, event_f_flags);
 
-	if (!capable(CAP_SYS_ADMIN))
-		return -EPERM;
+	if (!capable(CAP_SYS_ADMIN)) {
+		/*
+		 * An unprivileged user can setup an fanotify group with
+		 * limited functionality - an unprivileged group is limited to
+		 * notification events with file handles and it cannot use
+		 * unlimited queue/marks.
+		 */
+		if ((flags & FANOTIFY_ADMIN_INIT_FLAGS) || !fid_mode)
+			return -EPERM;
+	}
 
 #ifdef CONFIG_AUDITSYSCALL
 	if (flags & ~(FANOTIFY_INIT_FLAGS | FAN_ENABLE_AUDIT))
@@ -963,12 +1084,6 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 	if ((fid_mode & FAN_REPORT_NAME) && !(fid_mode & FAN_REPORT_DIR_FID))
 		return -EINVAL;
 
-	user = get_current_user();
-	if (atomic_read(&user->fanotify_listeners) > FANOTIFY_DEFAULT_MAX_LISTENERS) {
-		free_uid(user);
-		return -EMFILE;
-	}
-
 	f_flags = O_RDWR | FMODE_NONOTIFY;
 	if (flags & FAN_CLOEXEC)
 		f_flags |= O_CLOEXEC;
@@ -978,15 +1093,27 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 	/* fsnotify_alloc_group takes a ref.  Dropped in fanotify_release */
 	group = fsnotify_alloc_user_group(&fanotify_fsnotify_ops);
 	if (IS_ERR(group)) {
-		free_uid(user);
 		return PTR_ERR(group);
 	}
 
-	group->fanotify_data.user = user;
+	/* Enforce groups limits per user in all containing user ns */
+	group->fanotify_data.ucounts = inc_ucount(current_user_ns(),
+						  current_euid(),
+						  UCOUNT_FANOTIFY_GROUPS);
+	if (!group->fanotify_data.ucounts) {
+		fd = -EMFILE;
+		goto out_destroy_group;
+	}
+
 	group->fanotify_data.flags = flags;
-	atomic_inc(&user->fanotify_listeners);
 	group->memcg = get_mem_cgroup_from_mm(current->mm);
 
+	group->fanotify_data.merge_hash = fanotify_alloc_merge_hash();
+	if (!group->fanotify_data.merge_hash) {
+		fd = -ENOMEM;
+		goto out_destroy_group;
+	}
+
 	group->overflow_event = fanotify_alloc_overflow_event();
 	if (unlikely(!group->overflow_event)) {
 		fd = -ENOMEM;
@@ -1019,16 +1146,13 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 			goto out_destroy_group;
 		group->max_events = UINT_MAX;
 	} else {
-		group->max_events = FANOTIFY_DEFAULT_MAX_EVENTS;
+		group->max_events = fanotify_max_queued_events;
 	}
 
 	if (flags & FAN_UNLIMITED_MARKS) {
 		fd = -EPERM;
 		if (!capable(CAP_SYS_ADMIN))
 			goto out_destroy_group;
-		group->fanotify_data.max_marks = UINT_MAX;
-	} else {
-		group->fanotify_data.max_marks = FANOTIFY_DEFAULT_MAX_MARKS;
 	}
 
 	if (flags & FAN_ENABLE_AUDIT) {
@@ -1126,7 +1250,7 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask,
 		 __func__, fanotify_fd, flags, dfd, pathname, mask);
 
 	/* we only use the lower 32 bits as of right now. */
-	if (mask & ((__u64)0xffffffff << 32))
+	if (upper_32_bits(mask))
 		return -EINVAL;
 
 	if (flags & ~FANOTIFY_MARK_FLAGS)
@@ -1181,6 +1305,15 @@ static int do_fanotify_mark(int fanotify_fd, unsigned int flags, __u64 mask,
 	group = f.file->private_data;
 
 	/*
+	 * An unprivileged user is not allowed to watch a mount point nor
+	 * a filesystem.
+	 */
+	ret = -EPERM;
+	if (!capable(CAP_SYS_ADMIN) &&
+	    mark_type != FAN_MARK_INODE)
+		goto fput_and_out;
+
+	/*
 	 * group->priority == FS_PRIO_0 == FAN_CLASS_NOTIF.  These are not
 	 * allowed to set permissions events.
 	 */
@@ -1312,6 +1445,21 @@ SYSCALL32_DEFINE6(fanotify_mark,
  */
 static int __init fanotify_user_setup(void)
 {
+	struct sysinfo si;
+	int max_marks;
+
+	si_meminfo(&si);
+	/*
+	 * Allow up to 1% of addressable memory to be accounted for per user
+	 * marks limited to the range [8192, 1048576]. mount and sb marks are
+	 * a lot cheaper than inode marks, but there is no reason for a user
+	 * to have many of those, so calculate by the cost of inode marks.
+	 */
+	max_marks = (((si.totalram - si.totalhigh) / 100) << PAGE_SHIFT) /
+		    INODE_MARK_COST;
+	max_marks = clamp(max_marks, FANOTIFY_OLD_DEFAULT_MAX_MARKS,
+				     FANOTIFY_DEFAULT_MAX_USER_MARKS);
+
 	BUILD_BUG_ON(HWEIGHT32(FANOTIFY_INIT_FLAGS) != 10);
 	BUILD_BUG_ON(HWEIGHT32(FANOTIFY_MARK_FLAGS) != 9);
 
@@ -1326,6 +1474,11 @@ static int __init fanotify_user_setup(void)
 			KMEM_CACHE(fanotify_perm_event, SLAB_PANIC);
 	}
 
+	fanotify_max_queued_events = FANOTIFY_DEFAULT_MAX_EVENTS;
+	init_user_ns.ucount_max[UCOUNT_FANOTIFY_GROUPS] =
+					FANOTIFY_DEFAULT_MAX_GROUPS;
+	init_user_ns.ucount_max[UCOUNT_FANOTIFY_MARKS] = max_marks;
+
 	return 0;
 }
 device_initcall(fanotify_user_setup);