Reimplement ggatec/ggated applications.

Change communication protocol to be much more resistant on network
problems and to allow for much better performance.

Better performance is achieved by creating two connections between
ggatec and ggated one for sending the data and one for receiving it.
Every connection is handled by separeted thread, so there is no more
synchronous data flow (send and wait for response), now one threads
sends all requests and another receives the data.

Use two threads in ggatec(8):
- sendtd, which takes I/O requests from the kernel and sends them to the
  ggated daemon on the other end;
- recvtd, which waits for ggated responses and forwards them to the kernel.

Use three threads in ggated(8):
- recvtd, which waits for I/O requests and puts them onto incoming queue;
- disktd, which takes requests from the incoming queue, does disk operations
  and puts finished requests onto outgoing queue;
- sendtd, which takes finished requests from the outgoing queue and sends
  responses back to ggatec.

Because there were major changes in communication protocol, there is no
backward compatibility, from now on, both client and server has to run
on 5.x or 6.x (or at least ggated should be from the same FreeBSD version
on which ggatec is running).

For Gbit networks some buffers need to be increased. I use those settings:
kern.ipc.maxsockbuf=16777216
net.inet.tcp.sendspace=8388608
net.inet.tcp.recvspace=8388608
and I use '-S 4194304 -R 4194304' options for both, ggatec and ggated.

Approved by:	re (scottl)

2 files changed, 136 insertions, 24 deletions

diff --git a/sbin/ggate/shared/ggate.c b/sbin/ggate/shared/ggate.c
index 08f13d7..c8428a6 100644
--- a/sbin/ggate/shared/ggate.c
+++ b/sbin/ggate/shared/ggate.c
@@ -38,6 +38,7 @@
 #include <sys/linker.h>
 #include <sys/module.h>
 #include <netinet/in.h>
+#include <netinet/tcp.h>
 #include <arpa/inet.h>
 #include <signal.h>
 #include <err.h>
@@ -206,17 +207,6 @@ g_gate_destroy(int unit, int force)
 	g_gate_ioctl(G_GATE_CMD_DESTROY, &ggio);
 }
 
-int
-g_gate_openflags(unsigned ggflags)
-{
-
-	if ((ggflags & G_GATE_FLAG_READONLY) != 0)
-		return (O_RDONLY);
-	else if ((ggflags & G_GATE_FLAG_WRITEONLY) != 0)
-		return (O_WRONLY);
-	return (O_RDWR);
-}
-
 void
 g_gate_load_module(void)
 {
@@ -232,6 +222,76 @@ g_gate_load_module(void)
 	}
 }
 
+ssize_t
+g_gate_send(int s, const void *buf, size_t len, int flags)
+{
+	ssize_t done = 0, done2;
+	const unsigned char *p = buf;
+
+	while (len > 0) {
+		done2 = send(s, p, len, flags);
+		if (done2 == 0)
+			break;
+		else if (done2 == -1) {
+			if (errno == EAGAIN) {
+				printf("%s: EAGAIN\n", __func__);
+				continue;
+			}
+			done = -1;
+			break;
+		}
+		done += done2;
+		p += done2;
+		len -= done2;
+	}
+	return (done);
+}
+
+ssize_t
+g_gate_recv(int s, void *buf, size_t len, int flags)
+{
+
+	return (recv(s, buf, len, flags));
+}
+
+int nagle = 1;
+unsigned rcvbuf = G_GATE_RCVBUF;
+unsigned sndbuf = G_GATE_SNDBUF;
+
+void
+g_gate_socket_settings(int sfd)
+{
+	struct timeval tv;
+	int bsize, on;
+
+	/* Socket settings. */
+	on = 1;
+	if (nagle) {
+		if (setsockopt(sfd, IPPROTO_TCP, TCP_NODELAY, &on, 
+		    sizeof(on)) == -1) {
+			g_gate_xlog("setsockopt() error: %s.", strerror(errno));
+		}
+	}
+	if (setsockopt(sfd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on)) == -1)
+		g_gate_xlog("setsockopt(SO_REUSEADDR): %s.", strerror(errno));
+	bsize = rcvbuf;
+	if (setsockopt(sfd, SOL_SOCKET, SO_RCVBUF, &bsize, sizeof(bsize)) == -1)
+		g_gate_xlog("setsockopt(SO_RCVBUF): %s.", strerror(errno));
+	bsize = sndbuf;
+	if (setsockopt(sfd, SOL_SOCKET, SO_SNDBUF, &bsize, sizeof(bsize)) == -1)
+		g_gate_xlog("setsockopt(SO_SNDBUF): %s.", strerror(errno));
+	tv.tv_sec = 1;
+	tv.tv_usec = 0;
+	if (setsockopt(sfd, SOL_SOCKET, SO_SNDTIMEO, &tv, sizeof(tv)) == -1) {
+		g_gate_log(LOG_ERR, "setsockopt(SO_SNDTIMEO) error: %s.",
+		    strerror(errno));
+	}
+	if (setsockopt(sfd, SOL_SOCKET, SO_RCVTIMEO, &tv, sizeof(tv)) == -1) {
+		g_gate_log(LOG_ERR, "setsockopt(SO_RCVTIMEO) error: %s.",
+		    strerror(errno));
+	}
+}
+
 #ifdef LIBGEOM
 static struct gclass *
 find_class(struct gmesh *mesh, const char *name)
diff --git a/sbin/ggate/shared/ggate.h b/sbin/ggate/shared/ggate.h
index 12dfe6d..acbdaaa 100644
--- a/sbin/ggate/shared/ggate.h
+++ b/sbin/ggate/shared/ggate.h
@@ -32,22 +32,49 @@
 #include <sys/endian.h>
 #include <stdarg.h>
 
-#define	G_GATE_BUFSIZE_START	65536
 #define	G_GATE_PORT		3080
 
 #define	G_GATE_RCVBUF		131072
 #define	G_GATE_SNDBUF		131072
 #define	G_GATE_QUEUE_SIZE	1024
-#define	G_GATE_TIMEOUT		30
+#define	G_GATE_TIMEOUT		0
+
+#define	GGATE_MAGIC		"GEOM_GATE       "
+#define	GGATE_VERSION		0
+
+#define	GGATE_FLAG_RDONLY	0x0001
+#define	GGATE_FLAG_WRONLY	0x0002
+/*
+ * If GGATE_FLAG_SEND not GGATE_FLAG_RECV flag is set, this is initial
+ * connection.
+ * If GGATE_FLAG_SEND flag is set - this is socket to send data.
+ * If GGATE_FLAG_RECV flag is set - this is socket to receive data.
+ */
+#define	GGATE_FLAG_SEND		0x0004
+#define	GGATE_FLAG_RECV		0x0008
+
+#define	GGATE_CMD_READ		0
+#define	GGATE_CMD_WRITE		1
 
 extern int g_gate_devfd;
 extern int g_gate_verbose;
 
+extern int nagle;
+extern unsigned rcvbuf, sndbuf;
+
+struct g_gate_version {
+	char		gv_magic[16];
+	uint16_t	gv_version;
+	uint16_t	gv_error;
+} __packed;
+
 /* Client's initial packet. */
 struct g_gate_cinit {
-	char	gc_path[PATH_MAX + 1];
-	uint8_t	gc_flags;
-};
+	char		gc_path[PATH_MAX + 1];
+	uint64_t	gc_flags;
+	uint16_t	gc_nconn;
+	uint32_t	gc_token;
+} __packed;
 
 /* Server's initial packet. */
 struct g_gate_sinit {
@@ -55,15 +82,16 @@ struct g_gate_sinit {
 	uint64_t	gs_mediasize;
 	uint32_t	gs_sectorsize;
 	uint16_t	gs_error;
-};
+} __packed;
 
 /* Control struct. */
 struct g_gate_hdr {
 	uint8_t		gh_cmd;		/* command */
 	uint64_t	gh_offset;	/* device offset */
 	uint32_t	gh_length;	/* size of block */
-	int16_t		gh_error;	/* error value (0 if ok) */
-};
+	uint64_t	gh_seq;		/* request number */
+	uint16_t	gh_error;	/* error value (0 if ok) */
+} __packed;
 
 void	g_gate_vlog(int priority, const char *message, va_list ap);
 void	g_gate_log(int priority, const char *message, ...);
@@ -75,8 +103,10 @@ void	g_gate_open_device(void);
 void	g_gate_close_device(void);
 void	g_gate_ioctl(unsigned long req, void *data);
 void	g_gate_destroy(int unit, int force);
-int	g_gate_openflags(unsigned ggflags);
 void	g_gate_load_module(void);
+ssize_t	g_gate_recv(int s, void *buf, size_t len, int flags);
+ssize_t	g_gate_send(int s, const void *buf, size_t len, int flags);
+void	g_gate_socket_settings(int sfd);
 #ifdef LIBGEOM
 void	g_gate_list(int unit, int verbose);
 #endif
@@ -89,17 +119,37 @@ in_addr_t g_gate_str2ip(const char *str);
  */
 
 static __inline void
-g_gate_swap2h_cinit(struct g_gate_cinit *cinit __unused)
+g_gate_swap2h_version(struct g_gate_version *ver)
+{
+
+	ver->gv_version = be16toh(ver->gv_version);
+	ver->gv_error = be16toh(ver->gv_error);
+}
+
+static __inline void
+g_gate_swap2n_version(struct g_gate_version *ver)
+{
+
+	ver->gv_version = htobe16(ver->gv_version);
+	ver->gv_error = htobe16(ver->gv_error);
+}
+
+static __inline void
+g_gate_swap2h_cinit(struct g_gate_cinit *cinit)
 {
 
-	/* Nothing here for now. */
+	cinit->gc_flags = be64toh(cinit->gc_flags);
+	cinit->gc_nconn = be16toh(cinit->gc_nconn);
+	cinit->gc_token = be32toh(cinit->gc_token);
 }
 
 static __inline void
-g_gate_swap2n_cinit(struct g_gate_cinit *cinit __unused)
+g_gate_swap2n_cinit(struct g_gate_cinit *cinit)
 {
 
-	/* Nothing here for now. */
+	cinit->gc_flags = htobe64(cinit->gc_flags);
+	cinit->gc_nconn = htobe16(cinit->gc_nconn);
+	cinit->gc_token = htobe32(cinit->gc_token);
 }
 
 static __inline void
@@ -129,6 +179,7 @@ g_gate_swap2h_hdr(struct g_gate_hdr *hdr)
 	/* Swap only used fields. */
 	hdr->gh_offset = be64toh(hdr->gh_offset);
 	hdr->gh_length = be32toh(hdr->gh_length);
+	hdr->gh_seq = be64toh(hdr->gh_seq);
 	hdr->gh_error = be16toh(hdr->gh_error);
 }
 
@@ -139,6 +190,7 @@ g_gate_swap2n_hdr(struct g_gate_hdr *hdr)
 	/* Swap only used fields. */
 	hdr->gh_offset = htobe64(hdr->gh_offset);
 	hdr->gh_length = htobe32(hdr->gh_length);
+	hdr->gh_seq = htobe64(hdr->gh_seq);
 	hdr->gh_error = htobe16(hdr->gh_error);
 }
 #endif	/* _GGATE_H_ */