epoll: Always use epoll_ref for the epoll data variable

epoll_ref contains a variety of information useful when handling epoll
events on our sockets, and we place it in the epoll_event data field
returned by epoll.  However, for a few other things we use the 'fd' field
in the standard union of types for that data field.

This actually introduces a bug which is vanishingly unlikely to hit in
practice, but very nasty if it ever did: theoretically if we had a very
large file descriptor number for fd_tap or fd_tap_listen it could overflow
into bits that overlap with the 'proto' field in epoll_ref.  With some
very bad luck this could mean that we mistakenly think an event on a
regular socket is an event on fd_tap or fd_tap_listen.

More practically, using different (but overlapping) fields of the
epoll_data means we can't unify dispatch for the various different objects
in the epoll.  Therefore use the same epoll_ref as the data for the tap
fds and the netns quit fd, adding new fd type values to describe them.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>

passt.c

@@ -60,6 +60,8 @@ char *epoll_type_str[EPOLL_TYPE_MAX + 1] = {
 	[EPOLL_TYPE_UDP]	= "UDP socket",
 	[EPOLL_TYPE_ICMP]	= "ICMP socket",
 	[EPOLL_TYPE_ICMPV6]	= "ICMPv6 socket",
+	[EPOLL_TYPE_NSQUIT]	= "namespace inotify",
+	[EPOLL_TYPE_TAP]	= "tap device",
 };
 
 /**
@@ -328,12 +330,11 @@ loop:
 
 	for (i = 0; i < nfds; i++) {
 		union epoll_ref ref = *((union epoll_ref *)&events[i].data.u64);
-		int fd = events[i].data.fd;
 
-		if (fd == c.fd_tap || fd == c.fd_tap_listen)
+		if (ref.type == EPOLL_TYPE_TAP)
-			tap_handler(&c, fd, events[i].events, &now);
+			tap_handler(&c, ref.fd, events[i].events, &now);
-		else if (fd == quit_fd)
+		else if (ref.type == EPOLL_TYPE_NSQUIT)
-			pasta_netns_quit_handler(&c, fd);
+			pasta_netns_quit_handler(&c, quit_fd);
 		else
 			sock_handler(&c, ref, events[i].events, &now);
 	}

passt.h

@@ -55,8 +55,12 @@ enum epoll_type {
 	EPOLL_TYPE_ICMP,
 	/* ICMPv6 sockets */
 	EPOLL_TYPE_ICMPV6,
+	/* inotify fd watching for end of netns (pasta) */
+	EPOLL_TYPE_NSQUIT,
+	/* tap char device, or qemu socket fd */
+	EPOLL_TYPE_TAP,
 
-	EPOLL_TYPE_MAX = EPOLL_TYPE_ICMPV6,
+	EPOLL_TYPE_MAX = EPOLL_TYPE_TAP,
 };
 
 /**

pasta.c

@@ -365,7 +365,10 @@ void pasta_ns_conf(struct ctx *c)
 int pasta_netns_quit_init(struct ctx *c)
 {
 	int flags = O_NONBLOCK | O_CLOEXEC;
-	struct epoll_event ev = { .events = EPOLLIN };
+	union epoll_ref ref = { .type = EPOLL_TYPE_NSQUIT };
+	struct epoll_event ev = {
+		.events = EPOLLIN
+	};
 	int inotify_fd;
 
 	if (c->mode != MODE_PASTA || c->no_netns_quit || !*c->netns_base)
@@ -381,7 +384,8 @@ int pasta_netns_quit_init(struct ctx *c)
 		return -1;
 	}
 
-	ev.data.fd = inotify_fd;
+	ref.fd = inotify_fd;
+	ev.data.u64 = ref.u64;
 	epoll_ctl(c->epollfd, EPOLL_CTL_ADD, inotify_fd, &ev);
 
 	return inotify_fd;

tap.c

@@ -1071,6 +1071,7 @@ restart:
 static void tap_sock_unix_init(struct ctx *c)
 {
 	int fd = socket(AF_UNIX, SOCK_STREAM, 0);
+	union epoll_ref ref = { .type = EPOLL_TYPE_TAP };
 	struct epoll_event ev = { 0 };
 	struct sockaddr_un addr = {
 		.sun_family = AF_UNIX,
@@ -1123,8 +1124,9 @@ static void tap_sock_unix_init(struct ctx *c)
 
 	listen(fd, 0);
 
-	ev.data.fd = c->fd_tap_listen = fd;
+	ref.fd = c->fd_tap_listen = fd;
 	ev.events = EPOLLIN | EPOLLET;
+	ev.data.u64 = ref.u64;
 	epoll_ctl(c->epollfd, EPOLL_CTL_ADD, c->fd_tap_listen, &ev);
 
 	info("You can now start qemu (>= 7.2, with commit 13c6be96618c):");
@@ -1141,6 +1143,7 @@ static void tap_sock_unix_init(struct ctx *c)
  */
 static void tap_sock_unix_new(struct ctx *c, uint32_t events)
 {
+	union epoll_ref ref = { .type = EPOLL_TYPE_TAP };
 	struct epoll_event ev = { 0 };
 	int v = INT_MAX / 2;
 	struct ucred ucred;
@@ -1180,8 +1183,9 @@ static void tap_sock_unix_new(struct ctx *c, uint32_t events)
 	    setsockopt(c->fd_tap, SOL_SOCKET, SO_SNDBUF, &v, sizeof(v)))
 		trace("tap: failed to set SO_SNDBUF to %i", v);
 
-	ev.data.fd = c->fd_tap;
+	ref.fd = c->fd_tap;
 	ev.events = EPOLLIN | EPOLLET | EPOLLRDHUP;
+	ev.data.u64 = ref.u64;
 	epoll_ctl(c->epollfd, EPOLL_CTL_ADD, c->fd_tap, &ev);
 }
 
@@ -1226,6 +1230,7 @@ static int tap_ns_tun(void *arg)
  */
 static void tap_sock_tun_init(struct ctx *c)
 {
+	union epoll_ref ref = { .type = EPOLL_TYPE_TAP };
 	struct epoll_event ev = { 0 };
 
 	NS_CALL(tap_ns_tun, c);
@@ -1234,8 +1239,9 @@ static void tap_sock_tun_init(struct ctx *c)
 
 	pasta_ns_conf(c);
 
-	ev.data.fd = c->fd_tap;
+	ref.fd = c->fd_tap;
 	ev.events = EPOLLIN | EPOLLRDHUP;
+	ev.data.u64 = ref.u64;
 	epoll_ctl(c->epollfd, EPOLL_CTL_ADD, c->fd_tap, &ev);
 }
 
@@ -1257,11 +1263,13 @@ void tap_sock_init(struct ctx *c)
 	}
 
 	if (c->fd_tap != -1) { /* Passed as --fd */
+		union epoll_ref ref = { .type = EPOLL_TYPE_TAP };
 		struct epoll_event ev = { 0 };
 		ASSERT(c->one_off);
 
-		ev.data.fd = c->fd_tap;
+		ref.fd = c->fd_tap;
 		ev.events = EPOLLIN | EPOLLET | EPOLLRDHUP;
+		ev.data.u64 = ref.u64;
 		epoll_ctl(c->epollfd, EPOLL_CTL_ADD, c->fd_tap, &ev);
 		return;
 	}