net: sockets: Cleanup socket properly if POSIX API is enabled

The sock_obj_core_dealloc() was not called if close() is called
instead of zsock_close(). This happens if POSIX API is enabled.

Fix this by calling zvfs_close() from zsock_close() and then
pass the socket number to zsock_close_ctx() so that the cleanup
can be done properly.

Reported-by: Andreas Ålgård <aal@ixys.no>
Signed-off-by: Jukka Rissanen <jukka.rissanen@nordicsemi.no>

include/zephyr/sys/fdtable.h

@@ -64,7 +64,10 @@ struct fd_op_vtable {
 		ssize_t (*write)(void *obj, const void *buf, size_t sz);
 		ssize_t (*write_offs)(void *obj, const void *buf, size_t sz, size_t offset);
 	};
-	int (*close)(void *obj);
+	union {
+		int (*close)(void *obj);
+		int (*close2)(void *obj, int fd);
+	};
 	int (*ioctl)(void *obj, unsigned int request, va_list args);
 };
 

lib/os/fdtable.c

@@ -389,7 +389,14 @@ int zvfs_close(int fd)
 	(void)k_mutex_lock(&fdtable[fd].lock, K_FOREVER);
 	if (fdtable[fd].vtable->close != NULL) {
 		/* close() is optional - e.g. stdinout_fd_op_vtable */
-		res = fdtable[fd].vtable->close(fdtable[fd].obj);
+		if (fdtable[fd].mode & ZVFS_MODE_IFSOCK) {
+			/* Network socket needs to know socket number so pass
+			 * it via close2() call.
+			 */
+			res = fdtable[fd].vtable->close2(fdtable[fd].obj, fd);
+		} else {
+			res = fdtable[fd].vtable->close(fdtable[fd].obj);
+		}
 	}
 	k_mutex_unlock(&fdtable[fd].lock);
 

subsys/net/lib/sockets/sockets.c

@@ -147,42 +147,26 @@ static inline int z_vrfy_zsock_socket(int family, int type, int proto)
 #include <zephyr/syscalls/zsock_socket_mrsh.c>
 #endif /* CONFIG_USERSPACE */
 
+extern int zvfs_close(int fd);
+
 int z_impl_zsock_close(int sock)
+{
+	return zvfs_close(sock);
+}
+
+#ifdef CONFIG_USERSPACE
+static inline int z_vrfy_zsock_close(int sock)
 {
 	const struct socket_op_vtable *vtable;
 	struct k_mutex *lock;
 	void *ctx;
-	int ret;
-
-	SYS_PORT_TRACING_OBJ_FUNC_ENTER(socket, close, sock);
 
 	ctx = get_sock_vtable(sock, &vtable, &lock);
 	if (ctx == NULL) {
 		errno = EBADF;
-		SYS_PORT_TRACING_OBJ_FUNC_EXIT(socket, close, sock, -errno);
 		return -1;
 	}
 
-	(void)k_mutex_lock(lock, K_FOREVER);
-
-	NET_DBG("close: ctx=%p, fd=%d", ctx, sock);
-
-	ret = vtable->fd_vtable.close(ctx);
-
-	k_mutex_unlock(lock);
-
-	SYS_PORT_TRACING_OBJ_FUNC_EXIT(socket, close, sock, ret < 0 ? -errno : ret);
-
-	zvfs_free_fd(sock);
-
-	(void)sock_obj_core_dealloc(sock);
-
-	return ret;
-}
-
-#ifdef CONFIG_USERSPACE
-static inline int z_vrfy_zsock_close(int sock)
-{
 	return z_impl_zsock_close(sock);
 }
 #include <zephyr/syscalls/zsock_close_mrsh.c>

subsys/net/lib/sockets/sockets_inet.c

@@ -137,10 +137,14 @@ static int zsock_socket_internal(int family, int type, int proto)
 	return fd;
 }
 
-int zsock_close_ctx(struct net_context *ctx)
+int zsock_close_ctx(struct net_context *ctx, int sock)
 {
 	int ret;
 
+	SYS_PORT_TRACING_OBJ_FUNC_ENTER(socket, close, sock);
+
+	NET_DBG("close: ctx=%p, fd=%d", ctx, sock);
+
 	/* Reset callbacks to avoid any race conditions while
 	 * flushing queues. No need to check return values here,
 	 * as these are fail-free operations and we're closing
@@ -160,10 +164,16 @@ int zsock_close_ctx(struct net_context *ctx)
 	ret = net_context_put(ctx);
 	if (ret < 0) {
 		errno = -ret;
-		return -1;
+		ret = -1;
 	}
 
-	return 0;
+	SYS_PORT_TRACING_OBJ_FUNC_EXIT(socket, close, sock, ret < 0 ? -errno : ret);
+
+	if (ret == 0) {
+		(void)sock_obj_core_dealloc(sock);
+	}
+
+	return ret;
 }
 
 static void zsock_accepted_cb(struct net_context *new_ctx,
@@ -2771,9 +2781,9 @@ static int sock_setsockopt_vmeth(void *obj, int level, int optname,
 	return zsock_setsockopt_ctx(obj, level, optname, optval, optlen);
 }
 
-static int sock_close_vmeth(void *obj)
+static int sock_close2_vmeth(void *obj, int fd)
 {
-	return zsock_close_ctx(obj);
+	return zsock_close_ctx(obj, fd);
 }
 static int sock_getpeername_vmeth(void *obj, struct sockaddr *addr,
 				  socklen_t *addrlen)
@@ -2791,7 +2801,7 @@ const struct socket_op_vtable sock_fd_op_vtable = {
 	.fd_vtable = {
 		.read = sock_read_vmeth,
 		.write = sock_write_vmeth,
-		.close = sock_close_vmeth,
+		.close2 = sock_close2_vmeth,
 		.ioctl = sock_ioctl_vmeth,
 	},
 	.shutdown = sock_shutdown_vmeth,

subsys/net/lib/sockets/sockets_internal.h

@@ -15,7 +15,7 @@
 #define SOCK_NONBLOCK 2
 #define SOCK_ERROR 4
 
-int zsock_close_ctx(struct net_context *ctx);
+int zsock_close_ctx(struct net_context *ctx, int sock);
 int zsock_poll_internal(struct zsock_pollfd *fds, int nfds, k_timeout_t timeout);
 
 int zsock_wait_data(struct net_context *ctx, k_timeout_t *timeout);

subsys/net/lib/sockets/sockets_packet.c

@@ -462,16 +462,16 @@ static int packet_sock_setsockopt_vmeth(void *obj, int level, int optname,
 	return zpacket_setsockopt_ctx(obj, level, optname, optval, optlen);
 }
 
-static int packet_sock_close_vmeth(void *obj)
+static int packet_sock_close2_vmeth(void *obj, int fd)
 {
-	return zsock_close_ctx(obj);
+	return zsock_close_ctx(obj, fd);
 }
 
 static const struct socket_op_vtable packet_sock_fd_op_vtable = {
 	.fd_vtable = {
 		.read = packet_sock_read_vmeth,
 		.write = packet_sock_write_vmeth,
-		.close = packet_sock_close_vmeth,
+		.close2 = packet_sock_close2_vmeth,
 		.ioctl = packet_sock_ioctl_vmeth,
 	},
 	.bind = packet_sock_bind_vmeth,

subsys/net/lib/sockets/sockets_tls.c

@@ -2108,7 +2108,7 @@ static int ztls_socket(int family, int type, int proto)
 	return -1;
 }
 
-int ztls_close_ctx(struct tls_context *ctx)
+int ztls_close_ctx(struct tls_context *ctx, int sock)
 {
 	int ret, err = 0;
 
@@ -2120,6 +2120,10 @@ int ztls_close_ctx(struct tls_context *ctx)
 	err = tls_release(ctx);
 	ret = zsock_close(ctx->sock);
 
+	if (ret == 0) {
+		(void)sock_obj_core_dealloc(sock);
+	}
+
 	/* In case close fails, we propagate errno value set by close.
 	 * In case close succeeds, but tls_release fails, set errno
 	 * according to tls_release return value.
@@ -3826,9 +3830,9 @@ static int tls_sock_setsockopt_vmeth(void *obj, int level, int optname,
 	return ztls_setsockopt_ctx(obj, level, optname, optval, optlen);
 }
 
-static int tls_sock_close_vmeth(void *obj)
+static int tls_sock_close2_vmeth(void *obj, int sock)
 {
-	return ztls_close_ctx(obj);
+	return ztls_close_ctx(obj, sock);
 }
 
 static int tls_sock_getpeername_vmeth(void *obj, struct sockaddr *addr,
@@ -3851,7 +3855,7 @@ static const struct socket_op_vtable tls_sock_fd_op_vtable = {
 	.fd_vtable = {
 		.read = tls_sock_read_vmeth,
 		.write = tls_sock_write_vmeth,
-		.close = tls_sock_close_vmeth,
+		.close2 = tls_sock_close2_vmeth,
 		.ioctl = tls_sock_ioctl_vmeth,
 	},
 	.shutdown = tls_sock_shutdown_vmeth,