devbridge.c

著名操作系统Plan 9的第三版的部分核心源代码。现在很难找到了。Plan 9是bell实验室开发的Unix后继者。

#include "u.h"
#include "../port/lib.h"
#include "mem.h"
#include "dat.h"
#include "fns.h"
#include "../port/netif.h"
#include "../port/error.h"

typedef struct Bridge 	Bridge;
typedef struct Port 	Port;
typedef struct Centry	Centry;
typedef struct Iphdr	Iphdr;
typedef struct Tcphdr	Tcphdr;

enum
{
	Qtopdir=	1,		/* top level directory */

	Qbridgedir,			/* bridge* directory */
	Qbctl,
	Qstats,
	Qcache,
	Qlog,

	Qportdir,			/* directory for a protocol */
	Qpctl,
	Qlocal,
	Qstatus,

	MaxQ,

	Maxbridge=	4,
	Maxport=	64,		// power of 2
	CacheHash=	257,		// prime
	CacheLook=	5,		// how many cache entries to examine
	CacheSize=	(CacheHash+CacheLook-1),
	CacheTimeout=	5*60,		// timeout for cache entry in seconds

	TcpMssMax = 1300,			// max desirable Tcp MSS value
	TunnelMtu = 1400,
};

static Dirtab bridgedirtab[]={
	"ctl",		{Qbctl},	0,	0666,
	"stats",	{Qstats},	0,	0444,
	"cache",	{Qcache},	0,	0444,
	"log",		{Qlog},		0,	0666,
};

static Dirtab portdirtab[]={
	"ctl",		{Qpctl},	0,	0666,
	"local",	{Qlocal},	0,	0444,
	"status",	{Qstatus},	0,	0444,
};

enum {
	Logcache=	(1<<0),
	Logmcast=	(1<<1),
};

// types of interfaces
enum
{
	Tether,
	Ttun,
};

static Logflag logflags[] =
{
	{ "cache",	Logcache, },
	{ "multicast",	Logmcast, },
	{ nil,		0, },
};

static Dirtab	*dirtab[MaxQ];

#define TYPE(x) 	((x).path & 0xff)
#define PORT(x) 	(((x).path >> 8)&(Maxport-1))
#define QID(x, y) 	(((x)<<8) | (y))

struct Centry
{
	uchar	d[Eaddrlen];
	int	port;
	long	expire;		// entry expires this number of seconds after bootime
	long	src;
	long	dst;
};

struct Bridge
{
	QLock;
	int	nport;
	Port	*port[Maxport];
	Centry	cache[CacheSize];
	ulong	hit;
	ulong	miss;
	ulong	copy;
	long	delay0;		// constant microsecond delay per packet
	long	delayn;		// microsecond delay per byte
	int tcpmss;		// modify tcpmss value

	Log;
};

struct Port
{
	int	id;
	Bridge	*bridge;
	int	ref;
	int	closed;

	Chan	*data[2];	// channel to data

	int	mcast;		// send multi cast packets

	Proc	*readp;		// read proc
	
	// the following uniquely identifies the port
	int	type;
	char	name[NAMELEN];
	
	// owner hash - avoids bind/unbind races
	ulong	ownhash;

	// various stats
	int	in;		// number of packets read
	int	inmulti;	// multicast or broadcast
	int	inunknown;	// unknown address
	int	out;		// number of packets read
	int	outmulti;	// multicast or broadcast
	int	outunknown;	// unknown address
	int outfrag;	// fragmented the packet
	int	nentry;		// number of cache entries for this port
};

enum {
	IP_VER		= 0x40,		/* Using IP version 4 */
	IP_HLEN		= 0x05,		/* Header length in characters */
	IP_DF		= 0x4000,	/* Don't fragment */
	IP_MF		= 0x2000,	/* More fragments */
	IP_MAX		= (32*1024),	/* Maximum Internet packet size */
	IP_TCPPROTO = 6,
	EOLOPT		= 0,
	NOOPOPT		= 1,
	MSSOPT		= 2,
	MSS_LENGTH	= 4,		/* Mean segment size */
	SYN		= 0x02,		/* Pkt. is synchronise */
	IPHDR		= 20,		/* sizeof(Iphdr) */
};

struct Iphdr
{
	uchar	vihl;		/* Version and header length */
	uchar	tos;		/* Type of service */
	uchar	length[2];	/* packet length */
	uchar	id[2];		/* ip->identification */
	uchar	frag[2];	/* Fragment information */
	uchar	ttl;		/* Time to live */
	uchar	proto;		/* Protocol */
	uchar	cksum[2];	/* Header checksum */
	uchar	src[4];		/* IP source */
	uchar	dst[4];		/* IP destination */
};

struct Tcphdr
{
	uchar	sport[2];
	uchar	dport[2];
	uchar	seq[4];
	uchar	ack[4];
	uchar	flag[2];
	uchar	win[2];
	uchar	cksum[2];
	uchar	urg[2];
};

static Bridge bridgetab[Maxbridge];

static int m2p[] = {
	[OREAD]		4,
	[OWRITE]	2,
	[ORDWR]		6
};

static int	bridgegen(Chan *c, Dirtab*, int, int s, Dir *dp);
static void	portbind(Bridge *b, int argc, char *argv[]);
static void	portunbind(Bridge *b, int argc, char *argv[]);
static void	etherread(void *a);
static char	*cachedump(Bridge *b);
static void	portfree(Port *port);
static void	cacheflushport(Bridge *b, int port);
static void	etherwrite(Port *port, Block *bp);

extern ulong	parseip(uchar*, char*);
extern ushort	ipcsum(uchar *addr);

static void
bridgeinit(void)
{
	int i;
	Dirtab *dt;
	// setup dirtab with non directory entries
	for(i=0; i<nelem(bridgedirtab); i++) {
		dt = bridgedirtab + i;
		dirtab[TYPE(dt->qid)] = dt;
	}
	for(i=0; i<nelem(portdirtab); i++) {
		dt = portdirtab + i;
		dirtab[TYPE(dt->qid)] = dt;
	}
}

static Chan*
bridgeattach(char* spec)
{
	Chan *c;
	int dev;

	dev = atoi(spec);
	if(dev<0 || dev >= Maxbridge)
		error("bad specification");

	c = devattach('B', spec);
	c->qid = (Qid){QID(0, Qtopdir)|CHDIR, 0};
	c->dev = dev;

	return c;
}

static int
bridgewalk(Chan *c, char *name)
{
	return devwalk(c, name, 0, 0, bridgegen);
}

static void
bridgestat(Chan* c, char* db)
{
	devstat(c, db, nil, 0, bridgegen);
}

static Chan*
bridgeopen(Chan* c, int omode)
{
	int perm;
	Bridge *b;

	omode &= 3;
	perm = m2p[omode];
	USED(perm);

	b = bridgetab + c->dev;
	USED(b);

	switch(TYPE(c->qid)) {
	default:
		break;
	case Qlog:
		logopen(b);
		break;
	case Qcache:
		c->aux = cachedump(b);
		break;
	}
	c->mode = openmode(omode);
	c->flag |= COPEN;
	c->offset = 0;
	return c;
}

static void
bridgeclose(Chan* c)
{
	Bridge *b  = bridgetab + c->dev;

	switch(TYPE(c->qid)) {
	case Qcache:
		if(c->flag & COPEN)
			free(c->aux);
		break;
	case Qlog:
		if(c->flag & COPEN)
			logclose(b);
		break;
	}
}

static long
bridgeread(Chan *c, void *a, long n, vlong off)
{
	char buf[256];
	Bridge *b = bridgetab + c->dev;
	Port *port;
	int i, ingood, outgood;

	USED(off);
	switch(TYPE(c->qid)) {
	default:
		error(Eperm);
	case Qtopdir:
	case Qbridgedir:
	case Qportdir:
		return devdirread(c, a, n, 0, 0, bridgegen);
	case Qlog:
		return logread(b, a, off, n);
	case Qstatus:
		qlock(b);
		port = b->port[PORT(c->qid)];
		if(port == 0)
			strcpy(buf, "unbound\n");
		else {
			i = 0;
			switch(port->type) {
			default: panic("bridgeread: unknown port type: %d", port->type);
			case Tether:
				i += snprint(buf+i, sizeof(buf)-i, "ether %s: ", port->name);
				break;
			case Ttun:
				i += snprint(buf+i, sizeof(buf)-i, "tunnel %s: ", port->name);
				break;
			}
			ingood = port->in-port->inmulti-port->inunknown;
			outgood = port->out-port->outmulti-port->outunknown;
			i += snprint(buf+i, sizeof(buf)-i, "in=%d(%d:%d:%d) out=%d(%d:%d:%d:%d)\n",
				port->in, ingood, port->inmulti, port->inunknown,
				port->out, outgood, port->outmulti, port->outunknown, port->outfrag);
			USED(i);
		}
		n = readstr(off, a, n, buf);
		qunlock(b);
		return n;
	case Qbctl:
		snprint(buf, sizeof(buf), "%s tcpmss\ndelay %ld %ld\n", b->tcpmss ? "set" : "clear",
			b->delay0, b->delayn);
		n = readstr(off, a, n, buf);
		return n;
	case Qcache:
		n = readstr(off, a, n, c->aux);
		return n;
	case Qstats:
		snprint(buf, sizeof(buf), "hit=%uld miss=%uld copy=%uld\n",
			b->hit, b->miss, b->copy);
		n = readstr(off, a, n, buf);
		return n;
	}
}

static void
bridgeoption(Bridge *b, char *option, int value)
{
	if(strcmp(option, "tcpmss") == 0)
		b->tcpmss = value;
	else
		error("unknown bridge option");
}


static long
bridgewrite(Chan *c, void *a, long n, vlong off)
{
	Bridge *b = bridgetab + c->dev;
	Cmdbuf *cb;
	char *arg0;
	char *p;
	
	USED(off);
	switch(TYPE(c->qid)) {
	default:
		error(Eperm);
	case Qbctl:
		cb = parsecmd(a, n);
		qlock(b);
		if(waserror()) {
			qunlock(b);
			free(cb);
			nexterror();
		}
		if(cb->nf == 0)
			error("short write");
		arg0 = cb->f[0];
		if(strcmp(arg0, "bind") == 0) {
			portbind(b, cb->nf-1, cb->f+1);
		} else if(strcmp(arg0, "unbind") == 0) {
			portunbind(b, cb->nf-1, cb->f+1);
		} else if(strcmp(arg0, "cacheflush") == 0) {
			log(b, Logcache, "cache flush\n");
			memset(b->cache, 0, CacheSize*sizeof(Centry));
		} else if(strcmp(arg0, "set") == 0) {
			if(cb->nf != 2)
				error("usage: set option");
			bridgeoption(b, cb->f[1], 1);
		} else if(strcmp(arg0, "clear") == 0) {
			if(cb->nf != 2)
				error("usage: clear option");
			bridgeoption(b, cb->f[1], 0);
		} else if(strcmp(arg0, "delay") == 0) {
			if(cb->nf != 3)
				error("usage: delay delay0 delayn");
			b->delay0 = strtol(cb->f[1], nil, 10);
			b->delayn = strtol(cb->f[2], nil, 10);
		} else
			error("unknown control request");
		poperror();
		qunlock(b);
		free(cb);
		return n;
	case Qlog:
		cb = parsecmd(a, n);
		p = logctl(b, cb->nf, cb->f, logflags);
		free(cb);
		if(p != nil)
			error(p);
		return n;
	}
}

static int
bridgegen(Chan *c, Dirtab*, int, int s, Dir *dp)
{
	Bridge *b = bridgetab + c->dev;
	int type = TYPE(c->qid);
	char buf[32];
	Dirtab *dt;
	Qid qid;

	if(s  == DEVDOTDOT){
		switch(TYPE(c->qid)){
		case Qtopdir:
		case Qbridgedir:
			snprint(buf, sizeof(buf), "#B%ld", c->dev);
			devdir(c, (Qid){CHDIR|Qtopdir, 0}, buf, 0, eve, 0555, dp);
			break;
		case Qportdir:
			sprint(buf, "bridge%ld", c->dev);
			devdir(c, (Qid){CHDIR|Qbridgedir, 0}, buf, 0, eve, 0555, dp);
			break;
		default:
			panic("bridgewalk %lux", c->qid.path);
		}
		return 1;
	}

	switch(type) {
	default:
		// non directory entries end up here
		if(c->qid.path & CHDIR)
			panic("bridgegen: unexpected directory");	
		if(s != 0)
			return -1;
		dt = dirtab[TYPE(c->qid)];
		if(dt == nil)
			panic("bridgegen: unknown type: %d", TYPE(c->qid));
		devdir(c, c->qid, dt->name, dt->length, eve, dt->perm, dp);
		return 1;
	case Qtopdir:
		if(s != 0)
			return -1;
		sprint(buf, "bridge%ld", c->dev);
		devdir(c, (Qid){QID(0,Qbridgedir)|CHDIR,0}, buf, 0, eve, 0555, dp);
		return 1;
	case Qbridgedir:
		if(s<nelem(bridgedirtab)) {
			dt = bridgedirtab+s;
			devdir(c, dt->qid, dt->name, dt->length, eve, dt->perm, dp);
			return 1;
		}
		s -= nelem(bridgedirtab);
		if(s >= b->nport)
			return -1;
		qid = (Qid){QID(s,Qportdir)|CHDIR, 0};
		snprint(buf, sizeof(buf), "%d", s);
		devdir(c, qid, buf, 0, eve, 0555, dp);
		return 1;
	case Qportdir:
		if(s>=nelem(portdirtab))
			return -1;
		dt = portdirtab+s;
		qid = (Qid){QID(PORT(c->qid),TYPE(dt->qid)),0};
		devdir(c, qid, dt->name, dt->length, eve, dt->perm, dp);
		return 1;
	}
}

// also in netif.c
static int
parseaddr(uchar *to, char *from, int alen)
{
	char nip[4];
	char *p;
	int i;

	p = from;
	for(i = 0; i < alen; i++){
		if(*p == 0)
			return -1;
		nip[0] = *p++;
		if(*p == 0)
			return -1;
		nip[1] = *p++;
		nip[2] = 0;
		to[i] = strtoul(nip, 0, 16);
		if(*p == ':')
			p++;
	}
	return 0;
}

// assumes b is locked
static void
portbind(Bridge *b, int argc, char *argv[])
{
	Port *port;
	char path[8*NAMELEN];
	char buf[100];
	char *dev, *dev2=nil, *p;
	Chan *ctl;
	int type=0, i, n;
	char *usage = "usage: bind ether|tunnel name ownhash dev [dev2]";
	char name[NAMELEN];
	ulong ownhash;

	memset(name, 0, NAMELEN);
	if(argc < 4)
		error(usage);
	if(strcmp(argv[0], "ether") == 0) {
		if(argc != 4)
			error(usage);
		type = Tether;
		strncpy(name, argv[1], NAMELEN);
		name[NAMELEN-1] = 0;
//		parseaddr(addr, argv[1], Eaddrlen);
	} else if(strcmp(argv[0], "tunnel") == 0) {
		if(argc != 5)
			error(usage);
		type = Ttun;
		strncpy(name, argv[1], NAMELEN);
		name[NAMELEN-1] = 0;
//		parseip(addr, argv[1]);
		dev2 = argv[4];
	} else
		error(usage);
	ownhash = atoi(argv[2]);
	dev = argv[3];
	for(i=0; i<b->nport; i++) {
		port = b->port[i];
		if(port != nil)
		if(port->type == type)
		if(memcmp(port->name, name, NAMELEN) == 0)
			error("port in use");
	}
	for(i=0; i<Maxport; i++)
		if(b->port[i] == nil)
			break;
	if(i == Maxport)
		error("no more ports");
	port = smalloc(sizeof(Port));
	port->ref = 1;
	port->id = i;
	port->ownhash = ownhash;

	if(waserror()) {
		portfree(port);
		nexterror();
	}
	port->type = type;
	memmove(port->name, name, NAMELEN);
	switch(port->type) {
	default: panic("portbind: unknown port type: %d", type);
	case Tether:
		snprint(path, sizeof(path), "%s/clone", dev);
		ctl = namec(path, Aopen, ORDWR, 0);
		if(waserror()) {
			cclose(ctl);
			nexterror();
		}
		// check addr?

		// get directory name
		n = devtab[ctl->type]->read(ctl, buf, sizeof(buf), 0);
		buf[n] = 0;
		for(p = buf; *p == ' '; p++)
			;
		snprint(path, sizeof(path), "%s/%lud/data", dev, strtoul(p, 0, 0));

		// setup connection to be promiscuous
		snprint(buf, sizeof(buf), "connect -1");
		devtab[ctl->type]->write(ctl, buf, strlen(buf), 0);
		snprint(buf, sizeof(buf), "promiscuous");
		devtab[ctl->type]->write(ctl, buf, strlen(buf), 0);
		snprint(buf, sizeof(buf), "bridge");
		devtab[ctl->type]->write(ctl, buf, strlen(buf), 0);

		// open data port
		port->data[0] = namec(path, Aopen, ORDWR, 0);
		// dup it
		incref(port->data[0]);
		port->data[1] = port->data[0];

		poperror();
		cclose(ctl);		

		break;