webcookies.c

这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易于我们学习和理解

			fprint(2, "bad wday separator (%s)\n", os);
		return -1;
	}

	/* 25- */
	if(!isdigit(s[0]) || !isdigit(s[1]) || (s[2]!='-' && s[2]!=' ')){
		if(debug)
			fprint(2, "bad day of month (%s)\n", os);
		return -1;
	}
	tm.mday = strtol(s, 0, 10);
	s += 3;

	/* Jan- */
	for(i=0; i<nelem(mon); i++)
		if(cistrncmp(s, mon[i], 3) == 0){
			tm.mon = i;
			s += 3;
			break;
		}
	if(i==nelem(mon)){
		if(debug)
			fprint(2, "bad month (%s)\n", os);
		return -1;
	}
	if(s[0] != '-' && s[0] != ' '){
		if(debug)
			fprint(2, "bad month separator (%s)\n", os);
		return -1;
	}
	s++;

	/* 2002 */
	if(!isdigit(s[0]) || !isdigit(s[1])){
		if(debug)
			fprint(2, "bad year (%s)\n", os);
		return -1;
	}
	tm.year = strtol(s, 0, 10);
	s += 2;
	if(isdigit(s[0]) && isdigit(s[1]))
		s += 2;
	else{
		if(tm.year <= 68)
			tm.year += 2000;
		else
			tm.year += 1900;
	}
	if(tm.mday==0 || tm.mday > mday[isleap(tm.year)][tm.mon]){
		if(debug)
			fprint(2, "invalid day of month (%s)\n", os);
		return -1;
	}
	tm.year -= 1900;
	if(*s++ != ' '){
		if(debug)
			fprint(2, "bad year separator (%s)\n", os);
		return -1;
	}

	if(!isdigit(s[0]) || !isdigit(s[1]) || s[2]!=':'
	|| !isdigit(s[3]) || !isdigit(s[4]) || s[5]!=':'
	|| !isdigit(s[6]) || !isdigit(s[7]) || s[8]!=' '){
		if(debug)
			fprint(2, "bad time (%s)\n", os);
		return -1;
	}

	tm.hour = atoi(s);
	tm.min = atoi(s+3);
	tm.sec = atoi(s+6);
	if(tm.hour >= 24 || tm.min >= 60 || tm.sec >= 60){
		if(debug)
			fprint(2, "invalid time (%s)\n", os);
		return -1;
	}
	s += 9;

	if(cistrcmp(s, "GMT") != 0){
		if(debug)
			fprint(2, "time zone not GMT (%s)\n", os);
		return -1;
	}
	strcpy(tm.zone, "GMT");
	tm.yday = 0;
	return tm2sec(&tm);
}

/*
 * skip linear whitespace.  we're a bit more lenient than RFC2616 2.2.
 */
char*
skipspace(char *s)
{
	while(*s=='\r' || *s=='\n' || *s==' ' || *s=='\t')
		s++;
	return s;
}

/*
 * Try to identify old netscape headers.
 * The old headers:
 *	- didn't allow spaces around the '='
 *	- used an 'Expires' attribute
 *	- had no 'Version' attribute
 *	- had no quotes
 *	- allowed whitespace in values
 *	- apparently separated attr/value pairs with ';' exclusively
 */
int
isnetscape(char *hdr)
{
	char *s;

	for(s=hdr; (s=strchr(s, '=')) != nil; s++){
		if(isspace(s[1]) || (s > hdr && isspace(s[-1])))
			return 0;
		if(s[1]=='"')
			return 0;
	}
	if(cistrstr(hdr, "version="))
		return 0;
	return 1;
}

/*
 * Parse HTTP response headers, adding cookies to jar.
 * Overwrites the headers.  May overwrite path.
 */
char* parsecookie(Cookie*, char*, char**, int, char*, char*);
int
parsehttp(Jar *jar, char *hdr, char *dom, char *path)
{
	static char setcookie[] = "Set-Cookie:";
	char *e, *p, *nextp;
	Cookie c;
	int isns, n;

	isns = isnetscape(hdr);
	n = 0;
	for(p=hdr; p; p=nextp){
		p = skipspace(p);
		if(*p == '\0')
			break;
		nextp = strchr(p, '\n');
		if(nextp != nil)
			*nextp++ = '\0';
		if(debug)
			fprint(2, "?%s\n", p);
		if(cistrncmp(p, setcookie, strlen(setcookie)) != 0)
			continue;
		if(debug)
			fprint(2, "%s\n", p);
		p = skipspace(p+strlen(setcookie));
		for(; *p; p=skipspace(p)){
			if((e = parsecookie(&c, p, &p, isns, dom, path)) != nil){
				if(debug)
					fprint(2, "parse cookie: %s\n", e);
				break;
			}
			if((e = isbadcookie(&c, dom, path)) != nil){
				if(debug)
					fprint(2, "reject cookie; %s\n", e);
				continue;
			}
			addcookie(jar, &c);
			n++;
		}
	}
	return n;
}

static char*
skipquoted(char *s)
{
	/*
	 * Sec 2.2 of RFC2616 defines a "quoted-string" as:
	 *
	 *  quoted-string  = ( <"> *(qdtext | quoted-pair ) <"> )
	 *  qdtext         = <any TEXT except <">>
	 *  quoted-pair    = "\" CHAR
	 *
	 * TEXT is any octet except CTLs, but including LWS;
	 * LWS is [CR LF] 1*(SP | HT);
	 * CHARs are ASCII octets 0-127;  (NOTE: we reject 0's)
	 * CTLs are octets 0-31 and 127;
	 */
	if(*s != '"')
		return s;

	for(s++; 32 <= *s && *s < 127 && *s != '"'; s++)
		if(*s == '\\' && *(s+1) != '\0')
			s++;
	return s;
}

static char*
skiptoken(char *s)
{
	/*
	 * Sec 2.2 of RFC2616 defines a "token" as
 	 *  1*<any CHAR except CTLs or separators>;
	 * CHARs are ASCII octets 0-127;
	 * CTLs are octets 0-31 and 127;
	 * separators are "()<>@,;:\/[]?={}", double-quote, SP (32), and HT (9)
	 */
	while(32 <= *s && *s < 127 && strchr("()<>@,;:[]?={}\" \t\\", *s)==nil)
		s++;

	return s;
}

static char*
skipvalue(char *s, int isns)
{
	char *t;

	/*
	 * An RFC2109 value is an HTTP token or an HTTP quoted string.
	 * Netscape servers ignore the spec and rely on semicolons, apparently.
	 */
	if(isns){
		if((t = strchr(s, ';')) == nil)
			t = s+strlen(s);
		return t;
	}
	if(*s == '"')
		return skipquoted(s);
	return skiptoken(s);
}

/*
 * RMID=80b186bb64c03c65fab767f8; expires=Monday, 10-Feb-2003 04:44:39 GMT; 
 *	path=/; domain=.nytimes.com
 */
char*
parsecookie(Cookie *c, char *p, char **e, int isns, char *dom, char *path)
{
	int i, done;
	char *t, *u, *attr, *val;

	memset(c, 0, sizeof *c);
	c->expire = ~0;

	/* NAME=VALUE */
	t = skiptoken(p);
	c->name = p;
	p = skipspace(t);
	if(*p != '='){
	Badname:
		return "malformed cookie: no NAME=VALUE";
	}
	*t = '\0';
	p = skipspace(p+1);
	t = skipvalue(p, isns);
	if(*t)
		*t++ = '\0';
	c->value = p;
	p = skipspace(t);
	if(c->name[0]=='\0' || c->value[0]=='\0')
		goto Badname;

	done = 0;
	for(; *p && !done; p=skipspace(p)){
		attr = p;
		t = skiptoken(p);
		u = skipspace(t);
		switch(*u){
		case '\0':
			*t = '\0';
			p = val = u;
			break;
		case ';':
			*t = '\0';
			val = "";
			p = u+1;
			break;
		case '=':
			*t = '\0';
			val = skipspace(u+1);
			p = skipvalue(val, isns);
			if(*p==',')
				done = 1;
			if(*p)
				*p++ = '\0';
			break;
		case ',':
			if(!isns){
				val = "";
				p = u;
				*p++ = '\0';
				done = 1;
				break;
			}
		default:
			if(debug)
				fprint(2, "syntax: %s\n", p);
			return "syntax error";
		}
		for(i=0; i<nelem(stab); i++)
			if(stab[i].ishttp && cistrcmp(stab[i].s, attr)==0)
				*(char**)((char*)c+stab[i].offset) = val;
		if(cistrcmp(attr, "expires") == 0){
			if(!isns)
				return "non-netscape cookie has Expires tag";
			if(!val[0])
				return "bad expires tag";
			c->expire = strtotime(val);
			if(c->expire == ~0)
				return "cannot parse netscape expires tag";
		}
		if(cistrcmp(attr, "max-age") == 0)
			c->expire = time(0)+atoi(val);
		if(cistrcmp(attr, "secure") == 0)
			c->secure = 1;
	}

	if(c->dom)
		c->explicitdom = 1;
	else
		c->dom = dom;
	if(c->path)
		c->explicitpath = 1;
	else{
		c->path = path;
		if((t = strchr(c->path, '?')) != 0)
			*t = '\0';
		if((t = strrchr(c->path, '/')) != 0)
			*t = '\0';
	}
	c->netscapestyle = isns;
	*e = p;

	return nil;
}

Jar *jar;

enum
{
	Xhttp = 1,
	Xcookies,

	NeedUrl = 0,
	HaveUrl,
};

typedef struct Aux Aux;
struct Aux
{
	int state;
	char *dom;
	char *path;
	char *inhttp;
	char *outhttp;
	char *ctext;
	int rdoff;
};
enum
{
	AuxBuf = 4096,
	MaxCtext = 16*1024*1024,
};

void
fsopen(Req *r)
{
	char *s, *es;
	int i, sz;
	Aux *a;

	switch((uintptr)r->fid->file->aux){
	case Xhttp:
		syncjar(jar);
		a = emalloc9p(sizeof(Aux));
		r->fid->aux = a;
		a->inhttp = emalloc9p(AuxBuf);
		a->outhttp = emalloc9p(AuxBuf);
		break;

	case Xcookies:
		syncjar(jar);
		a = emalloc9p(sizeof(Aux));
		r->fid->aux = a;
		if(r->ifcall.mode&OTRUNC){
			a->ctext = emalloc9p(1);
			a->ctext[0] = '\0';
		}else{
			sz = 256*jar->nc+1024;	/* BUG should do better */
			a->ctext = emalloc9p(sz);
			a->ctext[0] = '\0';
			s = a->ctext;
			es = s+sz;
			for(i=0; i<jar->nc; i++)
				s = seprint(s, es, "%K\n", &jar->c[i]);
		}
		break;
	}
	respond(r, nil);
}

void
fsread(Req *r)
{
	Aux *a;

	a = r->fid->aux;
	switch((uintptr)r->fid->file->aux){
	case Xhttp:
		if(a->state == NeedUrl){
			respond(r, "must write url before read");
			return;
		}
		r->ifcall.offset = a->rdoff;
		readstr(r, a->outhttp);
		a->rdoff += r->ofcall.count;
		respond(r, nil);
		return;

	case Xcookies:
		readstr(r, a->ctext);
		respond(r, nil);
		return;

	default:
		respond(r, "bug in webcookies");
		return;
	}
}

void
fswrite(Req *r)
{
	Aux *a;
	int i, sz, hlen, issecure;
	char buf[1024], *p;
	Jar *j;

	a = r->fid->aux;
	switch((uintptr)r->fid->file->aux){
	case Xhttp:
		if(a->state == NeedUrl){
			if(r->ifcall.count >= sizeof buf){
				respond(r, "url too long");
				return;
			}
			memmove(buf, r->ifcall.data, r->ifcall.count);
			buf[r->ifcall.count] = '\0';
			issecure = 0;
			if(cistrncmp(buf, "http://", 7) == 0)
				hlen = 7;
			else if(cistrncmp(buf, "https://", 8) == 0){
				hlen = 8;
				issecure = 1;
			}else{
				respond(r, "url must begin http:// or https://");
				return;
			}
			if(buf[hlen]=='/'){
				respond(r, "url without host name");
				return;
			}
			p = strchr(buf+hlen, '/');
			if(p == nil)
				a->path = estrdup9p("/");
			else{
				a->path = estrdup9p(p);
				*p = '\0';
			}
			a->dom = estrdup9p(buf+hlen);
			a->state = HaveUrl;
			j = cookiesearch(jar, a->dom, a->path, issecure);
			if(debug){
				fprint(2, "search %s %s got %p\n", a->dom, a->path, j);
				if(j){
					fprint(2, "%d cookies\n", j->nc);
					for(i=0; i<j->nc; i++)
						fprint(2, "%K\n", &j->c[i]);
				}
			}
			snprint(a->outhttp, AuxBuf, "%J", j);
			if(j)
				closejar(j);
		}else{
			if(strlen(a->inhttp)+r->ifcall.count >= AuxBuf){
				respond(r, "http headers too large");
				return;
			}
			memmove(a->inhttp+strlen(a->inhttp), r->ifcall.data, r->ifcall.count);
		}
		r->ofcall.count = r->ifcall.count;
		respond(r, nil);
		return;

	case Xcookies:
		sz = r->ifcall.count+r->ifcall.offset;
		if(sz > strlen(a->ctext)){
			if(sz >= MaxCtext){
				respond(r, "cookie file too large");
				return;
			}
			a->ctext = erealloc9p(a->ctext, sz+1);
			a->ctext[sz] = '\0';
		}
		memmove(a->ctext+r->ifcall.offset, r->ifcall.data, r->ifcall.count);
		r->ofcall.count = r->ifcall.count;
		respond(r, nil);
		return;

	default:
		respond(r, "bug in webcookies");
		return;
	}
}

void
fsdestroyfid(Fid *fid)
{
	char *p, *nextp;
	Aux *a;
	int i;

	a = fid->aux;
	if(a == nil)
		return;
	switch((uintptr)fid->file->aux){
	case Xhttp:
		parsehttp(jar, a->inhttp, a->dom, a->path);
		break;
	case Xcookies:
		for(i=0; i<jar->nc; i++)
			jar->c[i].mark = 1;
		for(p=a->ctext; *p; p=nextp){
			if((nextp = strchr(p, '\n')) != nil)
				*nextp++ = '\0';
			else
				nextp = "";
			addtojar(jar, p, 0);
		}
		for(i=0; i<jar->nc; i++)
			if(jar->c[i].mark)
				delcookie(jar, &jar->c[i]);
		break;
	}
	syncjar(jar);
	free(a->dom);
	free(a->path);
	free(a->inhttp);
	free(a->outhttp);
	free(a->ctext);
	free(a);
}

void
fsend(Srv*)
{
	closejar(jar);
}

Srv fs = 
{
.open=		fsopen,
.read=		fsread,
.write=		fswrite,
.destroyfid=	fsdestroyfid,
.end=		fsend,
};

void
usage(void)
{
	fprint(2, "usage: webcookies [-f file] [-m mtpt] [-s service]\n");
	exits("usage");
}
	
void
main(int argc, char **argv)
{
	char *file, *mtpt, *home, *srv;

	file = nil;
	srv = nil;
	mtpt = "/mnt/webcookies";
	ARGBEGIN{
	case 'D':
		chatty9p++;
		break;
	case 'd':
		debug = 1;
		break;
	case 'f':
		file = EARGF(usage());
		break;
	case 's':
		srv = EARGF(usage());
		break;
	case 'm':
		mtpt = EARGF(usage());
		break;
	default:
		usage();
	}ARGEND

	if(argc != 0)
		usage();

	quotefmtinstall();
	fmtinstall('J', jarfmt);
	fmtinstall('K', cookiefmt);

	if(file == nil){
		home = getenv("home");
		if(home == nil)
			sysfatal("no cookie file specified and no $home");
		file = emalloc9p(strlen(home)+30);
		strcpy(file, home);
		strcat(file, "/lib/webcookies");
	}
	if(access(file, AEXIST) < 0)
		close(create(file, OWRITE, 0666));

	jar = readjar(file);
	if(jar == nil)
		sysfatal("readjar: %r");

	fs.tree = alloctree("cookie", "cookie", DMDIR|0555, nil);
	closefile(createfile(fs.tree->root, "http", "cookie", 0666, (void*)Xhttp));
	closefile(createfile(fs.tree->root, "cookies", "cookie", 0666, (void*)Xcookies));

	postmountsrv(&fs, srv, mtpt, MREPL);
	exits(nil);
}