fanotify: encode file identifier for FAN_REPORT_FID

When user requests the flag FAN_REPORT_FID in fanotify_init(),
a unique file identifier of the event target object will be reported
with the event.

The file identifier includes the filesystem's fsid (i.e. from statfs(2))
and an NFS file handle of the file (i.e. from name_to_handle_at(2)).

The file identifier makes holding the path reference and passing a file
descriptor to user redundant, so those are disabled in a group with
FAN_REPORT_FID.

Encode fid and store it in event for a group with FAN_REPORT_FID.
Up to 12 bytes of file handle on 32bit arch (16 bytes on 64bit arch)
are stored inline in fanotify_event struct. Larger file handles are
stored in an external allocated buffer.

On failure to encode fid, we print a warning and queue the event
without the fid information.

[JK: Fold part of later patched into this one to use
exportfs_encode_inode_fh() right away]

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>

3 files changed, 155 insertions, 13 deletions

diff --git a/fs/notify/fanotify/fanotify.c b/fs/notify/fanotify/fanotify.c
index d8e3b6e50844..dd33227e518a 100644
--- a/fs/notify/fanotify/fanotify.c
+++ b/fs/notify/fanotify/fanotify.c
@@ -13,6 +13,7 @@
 #include <linux/wait.h>
 #include <linux/audit.h>
 #include <linux/sched/mm.h>
+#include <linux/statfs.h>
 
 #include "fanotify.h"
 
@@ -25,10 +26,18 @@ static bool should_merge(struct fsnotify_event *old_fsn,
 	old = FANOTIFY_E(old_fsn);
 	new = FANOTIFY_E(new_fsn);
 
-	if (old_fsn->inode == new_fsn->inode && old->pid == new->pid &&
-	    old->path.mnt == new->path.mnt &&
-	    old->path.dentry == new->path.dentry)
-		return true;
+	if (old_fsn->inode != new_fsn->inode || old->pid != new->pid ||
+	    old->fh_type != new->fh_type || old->fh_len != new->fh_len)
+		return false;
+
+	if (fanotify_event_has_path(old)) {
+		return old->path.mnt == new->path.mnt &&
+			old->path.dentry == new->path.dentry;
+	} else if (fanotify_event_has_fid(old)) {
+		return fanotify_fid_equal(&old->fid, &new->fid, old->fh_len);
+	}
+
+	/* Do not merge events if we failed to encode fid */
 	return false;
 }
 
@@ -143,6 +152,60 @@ static u32 fanotify_group_event_mask(struct fsnotify_iter_info *iter_info,
 		~marks_ignored_mask;
 }
 
+static int fanotify_encode_fid(struct fanotify_event *event,
+			       const struct path *path, gfp_t gfp)
+{
+	struct fanotify_fid *fid = &event->fid;
+	int dwords, bytes = 0;
+	struct kstatfs stat;
+	int err, type;
+
+	stat.f_fsid.val[0] = stat.f_fsid.val[1] = 0;
+	fid->ext_fh = NULL;
+	dwords = 0;
+	err = -ENOENT;
+	type = exportfs_encode_inode_fh(d_inode(path->dentry), NULL, &dwords,
+					NULL);
+	if (!dwords)
+		goto out_err;
+
+	err = vfs_statfs(path, &stat);
+	if (err)
+		goto out_err;
+
+	bytes = dwords << 2;
+	if (bytes > FANOTIFY_INLINE_FH_LEN) {
+		/* Treat failure to allocate fh as failure to allocate event */
+		err = -ENOMEM;
+		fid->ext_fh = kmalloc(bytes, gfp);
+		if (!fid->ext_fh)
+			goto out_err;
+	}
+
+	type = exportfs_encode_inode_fh(d_inode(path->dentry),
+					fanotify_fid_fh(fid, bytes), &dwords,
+					NULL);
+	err = -EINVAL;
+	if (!type || type == FILEID_INVALID || bytes != dwords << 2)
+		goto out_err;
+
+	fid->fsid = stat.f_fsid;
+	event->fh_len = bytes;
+
+	return type;
+
+out_err:
+	pr_warn_ratelimited("fanotify: failed to encode fid (fsid=%x.%x, "
+			    "type=%d, bytes=%d, err=%i)\n",
+			    stat.f_fsid.val[0], stat.f_fsid.val[1],
+			    type, bytes, err);
+	kfree(fid->ext_fh);
+	fid->ext_fh = NULL;
+	event->fh_len = 0;
+
+	return FILEID_INVALID;
+}
+
 struct fanotify_event *fanotify_alloc_event(struct fsnotify_group *group,
 						 struct inode *inode, u32 mask,
 						 const struct path *path)
@@ -181,10 +244,16 @@ init: __maybe_unused
 		event->pid = get_pid(task_pid(current));
 	else
 		event->pid = get_pid(task_tgid(current));
-	if (path) {
+	event->fh_len = 0;
+	if (path && FAN_GROUP_FLAG(group, FAN_REPORT_FID)) {
+		/* Report the event without a file identifier on encode error */
+		event->fh_type = fanotify_encode_fid(event, path, gfp);
+	} else if (path) {
+		event->fh_type = FILEID_ROOT;
 		event->path = *path;
 		path_get(&event->path);
 	} else {
+		event->fh_type = FILEID_INVALID;
 		event->path.mnt = NULL;
 		event->path.dentry = NULL;
 	}
@@ -281,7 +350,10 @@ static void fanotify_free_event(struct fsnotify_event *fsn_event)
 	struct fanotify_event *event;
 
 	event = FANOTIFY_E(fsn_event);
-	path_put(&event->path);
+	if (fanotify_event_has_path(event))
+		path_put(&event->path);
+	else if (fanotify_event_has_ext_fh(event))
+		kfree(event->fid.ext_fh);
 	put_pid(event->pid);
 	if (fanotify_is_perm_event(event->mask)) {
 		kmem_cache_free(fanotify_perm_event_cachep,
diff --git a/fs/notify/fanotify/fanotify.h b/fs/notify/fanotify/fanotify.h
index 898b5b2bc1c7..271482fb9611 100644
--- a/fs/notify/fanotify/fanotify.h
+++ b/fs/notify/fanotify/fanotify.h
@@ -2,12 +2,50 @@
 #include <linux/fsnotify_backend.h>
 #include <linux/path.h>
 #include <linux/slab.h>
+#include <linux/exportfs.h>
 
 extern struct kmem_cache *fanotify_mark_cache;
 extern struct kmem_cache *fanotify_event_cachep;
 extern struct kmem_cache *fanotify_perm_event_cachep;
 
 /*
+ * 3 dwords are sufficient for most local fs (64bit ino, 32bit generation).
+ * For 32bit arch, fid increases the size of fanotify_event by 12 bytes and
+ * fh_* fields increase the size of fanotify_event by another 4 bytes.
+ * For 64bit arch, fid increases the size of fanotify_fid by 8 bytes and
+ * fh_* fields are packed in a hole after mask.
+ */
+#if BITS_PER_LONG == 32
+#define FANOTIFY_INLINE_FH_LEN	(3 << 2)
+#else
+#define FANOTIFY_INLINE_FH_LEN	(4 << 2)
+#endif
+
+struct fanotify_fid {
+	__kernel_fsid_t fsid;
+	union {
+		unsigned char fh[FANOTIFY_INLINE_FH_LEN];
+		unsigned char *ext_fh;
+	};
+};
+
+static inline void *fanotify_fid_fh(struct fanotify_fid *fid,
+				    unsigned int fh_len)
+{
+	return fh_len <= FANOTIFY_INLINE_FH_LEN ? fid->fh : fid->ext_fh;
+}
+
+static inline bool fanotify_fid_equal(struct fanotify_fid *fid1,
+				      struct fanotify_fid *fid2,
+				      unsigned int fh_len)
+{
+	return fid1->fsid.val[0] == fid2->fsid.val[0] &&
+		fid1->fsid.val[1] == fid2->fsid.val[1] &&
+		!memcmp(fanotify_fid_fh(fid1, fh_len),
+			fanotify_fid_fh(fid2, fh_len), fh_len);
+}
+
+/*
  * Structure for normal fanotify events. It gets allocated in
  * fanotify_handle_event() and freed when the information is retrieved by
  * userspace
@@ -16,13 +54,47 @@ struct fanotify_event {
 	struct fsnotify_event fse;
 	u32 mask;
 	/*
-	 * We hold ref to this path so it may be dereferenced at any point
-	 * during this object's lifetime
+	 * Those fields are outside fanotify_fid to pack fanotify_event nicely
+	 * on 64bit arch and to use fh_type as an indication of whether path
+	 * or fid are used in the union:
+	 * FILEID_ROOT (0) for path, > 0 for fid, FILEID_INVALID for neither.
 	 */
-	struct path path;
+	u8 fh_type;
+	u8 fh_len;
+	u16 pad;
+	union {
+		/*
+		 * We hold ref to this path so it may be dereferenced at any
+		 * point during this object's lifetime
+		 */
+		struct path path;
+		/*
+		 * With FAN_REPORT_FID, we do not hold any reference on the
+		 * victim object. Instead we store its NFS file handle and its
+		 * filesystem's fsid as a unique identifier.
+		 */
+		struct fanotify_fid fid;
+	};
 	struct pid *pid;
 };
 
+static inline bool fanotify_event_has_path(struct fanotify_event *event)
+{
+	return event->fh_type == FILEID_ROOT;
+}
+
+static inline bool fanotify_event_has_fid(struct fanotify_event *event)
+{
+	return event->fh_type != FILEID_ROOT &&
+		event->fh_type != FILEID_INVALID;
+}
+
+static inline bool fanotify_event_has_ext_fh(struct fanotify_event *event)
+{
+	return fanotify_event_has_fid(event) &&
+		event->fh_len > FANOTIFY_INLINE_FH_LEN;
+}
+
 /*
  * Structure for permission fanotify events. It gets allocated and freed in
  * fanotify_handle_event() since we wait there for user response. When the
diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
index 096503bd0edb..c965fcf4979e 100644
--- a/fs/notify/fanotify/fanotify_user.c
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -181,7 +181,7 @@ static ssize_t copy_event_to_user(struct fsnotify_group *group,
 	struct fanotify_event_metadata metadata;
 	struct fanotify_event *event;
 	struct file *f = NULL;
-	int fd, ret;
+	int ret, fd = FAN_NOFD;
 
 	pr_debug("%s: group=%p event=%p\n", __func__, group, fsn_event);
 
@@ -193,9 +193,7 @@ static ssize_t copy_event_to_user(struct fsnotify_group *group,
 	metadata.mask = event->mask & FANOTIFY_OUTGOING_EVENTS;
 	metadata.pid = pid_vnr(event->pid);
 
-	if (unlikely(event->mask & FAN_Q_OVERFLOW)) {
-		fd = FAN_NOFD;
-	} else {
+	if (fanotify_event_has_path(event)) {
 		fd = create_fd(group, event, &f);
 		if (fd < 0)
 			return fd;