net: allow multiple icmp sockets; autocalculate checksum; support IP_RECVTTL?

apps/ping.c

@@ -78,6 +78,9 @@ int main(int argc, char * argv[]) {
 		return 1;
 	}
 
+	int yes = 1;
+	setsockopt(sock, IPPROTO_IP, IP_RECVTTL, &yes, sizeof(int));
+
 	signal(SIGINT, sig_break_loop);
 
 	struct sockaddr_in dest;
@@ -118,14 +121,40 @@ int main(int argc, char * argv[]) {
 
 		if (ret > 0) {
 			char data[4096];
+			char control[4096];
 			struct sockaddr_in source;
 			socklen_t source_size = sizeof(struct sockaddr_in);
-			ssize_t len = recvfrom(sock, data, 4096, 0, (struct sockaddr*)&source, &source_size);
+			struct iovec _iovec = {
+				data, 4096
+			};
+			struct msghdr msg = {
+				&source,
+				source_size,
+				&_iovec,
+				1,
+				control,
+				4096,
+				0
+			};
+			ssize_t len = recvmsg(sock, &msg, 0);
 			unsigned long rcvd_at = clocktime();
 			if (len > 0) {
 				/* Is it actually a PING response ? */
 
 				struct ICMP_Header * icmp = (void*)data;
+				unsigned char ttl = 0;
+
+				if (msg.msg_controllen) {
+					char * control_msg = control;
+					while (control_msg - control + sizeof(struct cmsghdr) <= msg.msg_controllen) {
+						struct cmsghdr * cmsg = (void*)control_msg;
+						if (cmsg->cmsg_level == IPPROTO_IP && (cmsg->cmsg_type == IP_RECVTTL || cmsg->cmsg_type == IP_TTL)) {
+							memcpy(&ttl, CMSG_DATA(cmsg), 1);
+							break;
+						}
+						control_msg += cmsg->cmsg_len;
+					}
+				}
 
 				if (icmp->type == 0) {
 					/* How much data, minus the header? */
@@ -133,7 +162,7 @@ int main(int argc, char * argv[]) {
 					char * from = inet_ntoa(source.sin_addr);
 					int time_taken = (rcvd_at - sent_at);
 					printf("%zd bytes from %s: icmp_seq=%d ttl=%d time=%d",
-						len, from, ntohs(icmp->sequence_number), (unsigned char)source.sin_zero[0], /* we hide the ttl in here */
+						len, from, ntohs(icmp->sequence_number), (unsigned char)ttl,
 						time_taken / 1000);
 					if (time_taken < 1000) {
 						printf(".%03d", time_taken % 1000);

base/usr/include/kernel/net/ipv4.h

@@ -19,7 +19,7 @@ struct icmp_header {
 	uint8_t type;
 	uint8_t code;
 	uint16_t csum;
-	uint16_t rest_of_header;
+	uint16_t identifier;
 	uint8_t data[];
 } __attribute__((packed)) __attribute__((aligned(2)));
 

base/usr/include/netinet/in.h

@@ -21,4 +21,7 @@ struct sockaddr_in {
 in_addr_t inet_addr(const char *cp);
 char *inet_ntoa(struct in_addr in);
 
+#define IP_TTL 2
+#define IP_RECVTTL 12
+
 _End_C_Header

base/usr/include/sys/socket.h

@@ -21,7 +21,7 @@ _Begin_C_Header
 #define IPPROTO_TCP 6
 #define IPPROTO_UDP 17
 
-#define SOL_SOCKET 0
+#define SOL_SOCKET 1
 
 #define SO_KEEPALIVE 1
 #define SO_REUSEADDR 2
@@ -65,6 +65,15 @@ struct sockaddr_storage {
 	char _ss_pad[128];
 };
 
+struct cmsghdr {
+	socklen_t cmsg_len;
+	int cmsg_level;
+	int cmsg_type;
+	unsigned char cmsg_data[];
+};
+
+#define CMSG_DATA(cmsg) (&((struct cmsghdr*)(cmsg))->cmsg_data)
+
 #ifndef _KERNEL_
 extern ssize_t recv(int sockfd, void *buf, size_t len, int flags);
 extern ssize_t recvfrom(int sockfd, void *buf, size_t len, int flags, struct sockaddr *src_addr, socklen_t *addrlen);

kernel/net/ipv4.c

@@ -117,6 +117,16 @@ uint16_t calculate_tcp_checksum(struct tcp_check_header * p, struct tcp_header *
 	return ~(sum & 0xFFFF) & 0xFFFF;
 }
 
+static hashmap_t * udp_sockets = NULL;
+static hashmap_t * tcp_sockets = NULL;
+static hashmap_t * icmp_sockets = NULL;
+
+void ipv4_install(void) {
+	udp_sockets = hashmap_create_int(10);
+	tcp_sockets = hashmap_create_int(10);
+	icmp_sockets = hashmap_create_int(10);
+}
+
 int net_ipv4_send(struct ipv4_packet * response, fs_node_t * nic) {
 	/* TODO: This should be routing, with a _hint_ about the interface, not the actual nic to send from! */
 	struct EthernetDevice * enic = nic->device;
@@ -152,8 +162,6 @@ int net_ipv4_send(struct ipv4_packet * response, fs_node_t * nic) {
 	return 0;
 }
 
-static sock_t * icmp_handler = NULL;
-
 static void icmp_handle(struct ipv4_packet * packet, const char * src, const char * dest, fs_node_t * nic) {
 	struct icmp_header * header = (void*)&packet->payload;
 
@@ -187,10 +195,10 @@ static void icmp_handle(struct ipv4_packet * packet, const char * src, const cha
 		net_ipv4_send(response,nic);
 		free(response);
 	} else if (header->type == 0 && header->code == 0) {
-		printf("net: ping reply\n");
 		/* Did we have a client waiting for this? */
-		if (icmp_handler) {
+		sock_t * handler = hashmap_get(icmp_sockets, (void*)(uintptr_t)ntohs(header->identifier));
-			net_sock_add(icmp_handler, packet, ntohs(packet->length));
+		if (handler) {
+			net_sock_add(handler, packet, ntohs(packet->length));
 		}
 	} else {
 		printf("net: ipv4: %s: %s -> %s ICMP %d (code = %d)\n", nic->name, src, dest, header->type, header->code);
@@ -198,7 +206,7 @@ static void icmp_handle(struct ipv4_packet * packet, const char * src, const cha
 }
 
 static void sock_icmp_close(sock_t * sock) {
-	icmp_handler = NULL;
+	hashmap_remove(icmp_sockets, (void*)(uintptr_t)sock->priv32[0]);
 }
 
 static long sock_icmp_recv(sock_t * sock, struct msghdr * msg, int flags) {
@@ -225,23 +233,35 @@ static long sock_icmp_recv(sock_t * sock, struct msghdr * msg, int flags) {
 			((struct sockaddr_in*)msg->msg_name)->sin_family = AF_INET;
 			((struct sockaddr_in*)msg->msg_name)->sin_port = 0;
 			((struct sockaddr_in*)msg->msg_name)->sin_addr.s_addr = src->source;
-			((struct sockaddr_in*)msg->msg_name)->sin_zero[0] = src->ttl;
 		}
 	}
 
+	/* TODO should check flag IP_RECVTTL ? Also this should be common to all IPPROTO_IP sockets */
+	if (sock->priv32[2] && msg->msg_controllen > sizeof(struct cmsghdr) + 1) {
+		struct cmsghdr * out = msg->msg_control;
+		out->cmsg_len = sizeof(struct cmsghdr) + 1;
+		out->cmsg_level = IPPROTO_IP;
+		out->cmsg_type = IP_RECVTTL;
+		out->cmsg_data[0] = src->ttl;
+		msg->msg_controllen = sizeof(struct cmsghdr) + 1;
+	} else {
+		msg->msg_controllen = 0;
+	}
+
 	memcpy(msg->msg_iov[0].iov_base, src->payload, packet_size);
 	free(packet);
 	return packet_size;
 }
 
 static long sock_icmp_send(sock_t * sock, const struct msghdr *msg, int flags) {
-	if (msg->msg_iovlen > 1) {
+	if (msg->msg_iovlen > 1) return -ENOTSUP;
-		return -ENOTSUP;
-	}
 	if (msg->msg_iovlen == 0) return 0;
-	if (msg->msg_namelen != sizeof(struct sockaddr_in)) {
+	if (msg->msg_namelen != sizeof(struct sockaddr_in)) return -EINVAL;
-		return -EINVAL;
+	if (msg->msg_iov[0].iov_len < sizeof(struct icmp_header)) return -EINVAL;
-	}
+
+	struct icmp_header * icmp = msg->msg_iov[0].iov_base;
+	if (icmp->type != 8 || icmp->code != 0) return -EINVAL;
+	if (icmp->identifier != 0) return -EINVAL;
 
 	struct sockaddr_in * name = msg->msg_name;
 	fs_node_t * nic = net_if_route(name->sin_addr.s_addr);
@@ -262,6 +282,11 @@ static long sock_icmp_send(sock_t * sock, const struct msghdr *msg, int flags) {
 	response->checksum = htons(calculate_ipv4_checksum(response));
 
 	memcpy(response->payload, msg->msg_iov[0].iov_base, msg->msg_iov[0].iov_len);
+	struct icmp_header * micmp = (struct icmp_header*)response->payload;
+	micmp->identifier = htons(sock->priv32[0]);
+	micmp->csum = 0;
+	micmp->csum = htons(icmp_checksum(response));
+
 	net_ipv4_send(response,nic);
 	free(response);
 
@@ -269,24 +294,16 @@ static long sock_icmp_send(sock_t * sock, const struct msghdr *msg, int flags) {
 }
 
 static int icmp_socket(void) {
-	printf("icmp socket...\n");
+	if (hashmap_has(icmp_sockets, (void*)(uintptr_t)this_core->current_process->id)) return -EINVAL;
-	if (icmp_handler) return -EINVAL;
 	sock_t * sock = net_sock_create();
 	sock->sock_recv = sock_icmp_recv;
 	sock->sock_send = sock_icmp_send;
 	sock->sock_close = sock_icmp_close;
-	icmp_handler = sock;
+	sock->priv32[0] = this_core->current_process->id;
+	hashmap_set(icmp_sockets, (void*)(uintptr_t)sock->priv32[0], sock);
 	return process_append_fd((process_t *)this_core->current_process, (fs_node_t *)sock);
 }
 
-static hashmap_t * udp_sockets = NULL;
-static hashmap_t * tcp_sockets = NULL;
-
-void ipv4_install(void) {
-	udp_sockets = hashmap_create_int(10);
-	tcp_sockets = hashmap_create_int(10);
-}
-
 #define TCP_FLAGS_FIN (1 << 0)
 #define TCP_FLAGS_SYN (1 << 1)
 #define TCP_FLAGS_RST (1 << 2)
@@ -1034,3 +1051,15 @@ long net_ipv4_socket(int type, int protocol) {
 			return -EINVAL;
 	}
 }
+
+long net_so_ipv4_socket(struct SockData * sock, int optname, const void *optval, socklen_t optlen) {
+	switch (optname) {
+		case IP_RECVTTL:
+			if (optlen != sizeof(int)) return -EINVAL;
+			/* TODO ugh bad */
+			sock->priv32[2] = *(int*)optval;
+			return 0;
+		default:
+			return -ENOPROTOOPT;
+	}
+}

kernel/net/socket.c

@@ -215,6 +215,8 @@ long net_so_socket(struct SockData * sock, int optname, const void *optval, sock
 	}
 }
 
+extern long net_so_ipv4_socket(struct SockData * sock, int optname, const void *optval, socklen_t optlen);
+
 static inline int is_socket(int sockfd) {
 	if (!FD_CHECK(sockfd)) return -EBADF;
 	fs_node_t * node = FD_ENTRY(sockfd);
@@ -241,6 +243,8 @@ long net_setsockopt(int sockfd, int level, int optname, const void *optval, sock
 	switch (level) {
 		case SOL_SOCKET:
 			return net_so_socket(node,optname,optval,optlen);
+		case IPPROTO_IP:
+			return net_so_ipv4_socket(node,optname,optval,optlen);
 		default:
 			return -ENOPROTOOPT;
 	}