freebsd-nq/lib/libfetch/fetch.c
Stefan Eßer 73b3e4df87 The fix for schemeless and hostless URLs (rev. 1.27) broke the schemeless
proxy specification, which seems to be valid according to the man page.

Change the logic to consider "hostname:port" a hostname and port instead
of a file URL.

Approved by:	des
2001-07-28 21:28:14 +00:00

429 lines
9.9 KiB
C

/*-
* Copyright (c) 1998 Dag-Erling Coïdan Smørgrav
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer
* in this position and unchanged.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* $FreeBSD$
*/
#include <sys/param.h>
#include <sys/errno.h>
#include <ctype.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "fetch.h"
#include "common.h"
auth_t fetchAuthMethod;
int fetchLastErrCode;
char fetchLastErrString[MAXERRSTRING];
int fetchTimeout;
int fetchRestartCalls = 1;
/*** Local data **************************************************************/
/*
* Error messages for parser errors
*/
#define URL_MALFORMED 1
#define URL_BAD_SCHEME 2
#define URL_BAD_PORT 3
static struct fetcherr _url_errlist[] = {
{ URL_MALFORMED, FETCH_URL, "Malformed URL" },
{ URL_BAD_SCHEME, FETCH_URL, "Invalid URL scheme" },
{ URL_BAD_PORT, FETCH_URL, "Invalid server port" },
{ -1, FETCH_UNKNOWN, "Unknown parser error" }
};
/*** Public API **************************************************************/
/*
* Select the appropriate protocol for the URL scheme, and return a
* read-only stream connected to the document referenced by the URL.
* Also fill out the struct url_stat.
*/
FILE *
fetchXGet(struct url *URL, struct url_stat *us, const char *flags)
{
int direct;
direct = CHECK_FLAG('d');
if (strcasecmp(URL->scheme, SCHEME_FILE) == 0)
return fetchXGetFile(URL, us, flags);
else if (strcasecmp(URL->scheme, SCHEME_HTTP) == 0)
return fetchXGetHTTP(URL, us, flags);
else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
return fetchXGetFTP(URL, us, flags);
} else {
_url_seterr(URL_BAD_SCHEME);
return NULL;
}
}
/*
* Select the appropriate protocol for the URL scheme, and return a
* read-only stream connected to the document referenced by the URL.
*/
FILE *
fetchGet(struct url *URL, const char *flags)
{
return fetchXGet(URL, NULL, flags);
}
/*
* Select the appropriate protocol for the URL scheme, and return a
* write-only stream connected to the document referenced by the URL.
*/
FILE *
fetchPut(struct url *URL, const char *flags)
{
int direct;
direct = CHECK_FLAG('d');
if (strcasecmp(URL->scheme, SCHEME_FILE) == 0)
return fetchPutFile(URL, flags);
else if (strcasecmp(URL->scheme, SCHEME_HTTP) == 0)
return fetchPutHTTP(URL, flags);
else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
return fetchPutFTP(URL, flags);
} else {
_url_seterr(URL_BAD_SCHEME);
return NULL;
}
}
/*
* Select the appropriate protocol for the URL scheme, and return the
* size of the document referenced by the URL if it exists.
*/
int
fetchStat(struct url *URL, struct url_stat *us, const char *flags)
{
int direct;
direct = CHECK_FLAG('d');
if (strcasecmp(URL->scheme, SCHEME_FILE) == 0)
return fetchStatFile(URL, us, flags);
else if (strcasecmp(URL->scheme, SCHEME_HTTP) == 0)
return fetchStatHTTP(URL, us, flags);
else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
return fetchStatFTP(URL, us, flags);
} else {
_url_seterr(URL_BAD_SCHEME);
return -1;
}
}
/*
* Select the appropriate protocol for the URL scheme, and return a
* list of files in the directory pointed to by the URL.
*/
struct url_ent *
fetchList(struct url *URL, const char *flags)
{
int direct;
direct = CHECK_FLAG('d');
if (strcasecmp(URL->scheme, SCHEME_FILE) == 0)
return fetchListFile(URL, flags);
else if (strcasecmp(URL->scheme, SCHEME_HTTP) == 0)
return fetchListHTTP(URL, flags);
else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) {
return fetchListFTP(URL, flags);
} else {
_url_seterr(URL_BAD_SCHEME);
return NULL;
}
}
/*
* Attempt to parse the given URL; if successful, call fetchXGet().
*/
FILE *
fetchXGetURL(const char *URL, struct url_stat *us, const char *flags)
{
struct url *u;
FILE *f;
if ((u = fetchParseURL(URL)) == NULL)
return NULL;
f = fetchXGet(u, us, flags);
fetchFreeURL(u);
return f;
}
/*
* Attempt to parse the given URL; if successful, call fetchGet().
*/
FILE *
fetchGetURL(const char *URL, const char *flags)
{
return fetchXGetURL(URL, NULL, flags);
}
/*
* Attempt to parse the given URL; if successful, call fetchPut().
*/
FILE *
fetchPutURL(const char *URL, const char *flags)
{
struct url *u;
FILE *f;
if ((u = fetchParseURL(URL)) == NULL)
return NULL;
f = fetchPut(u, flags);
fetchFreeURL(u);
return f;
}
/*
* Attempt to parse the given URL; if successful, call fetchStat().
*/
int
fetchStatURL(const char *URL, struct url_stat *us, const char *flags)
{
struct url *u;
int s;
if ((u = fetchParseURL(URL)) == NULL)
return -1;
s = fetchStat(u, us, flags);
fetchFreeURL(u);
return s;
}
/*
* Attempt to parse the given URL; if successful, call fetchList().
*/
struct url_ent *
fetchListURL(const char *URL, const char *flags)
{
struct url *u;
struct url_ent *ue;
if ((u = fetchParseURL(URL)) == NULL)
return NULL;
ue = fetchList(u, flags);
fetchFreeURL(u);
return ue;
}
/*
* Make a URL
*/
struct url *
fetchMakeURL(const char *scheme, const char *host, int port, const char *doc,
const char *user, const char *pwd)
{
struct url *u;
if (!scheme || (!host && !doc)) {
_url_seterr(URL_MALFORMED);
return NULL;
}
if (port < 0 || port > 65535) {
_url_seterr(URL_BAD_PORT);
return NULL;
}
/* allocate struct url */
if ((u = calloc(1, sizeof *u)) == NULL) {
_fetch_syserr();
return NULL;
}
if ((u->doc = strdup(doc ? doc : "/")) == NULL) {
_fetch_syserr();
free(u);
return NULL;
}
#define seturl(x) snprintf(u->x, sizeof u->x, "%s", x)
seturl(scheme);
seturl(host);
seturl(user);
seturl(pwd);
#undef seturl
u->port = port;
return u;
}
/*
* Split an URL into components. URL syntax is:
* [method:/][/[user[:pwd]@]host[:port]/][document]
* This almost, but not quite, RFC1738 URL syntax.
*/
struct url *
fetchParseURL(const char *URL)
{
char *doc;
const char *p, *q;
struct url *u;
int i;
/* allocate struct url */
if ((u = calloc(1, sizeof *u)) == NULL) {
_fetch_syserr();
return NULL;
}
/* scheme name */
if ((p = strstr(URL, ":/"))) {
snprintf(u->scheme, URL_SCHEMELEN+1, "%.*s", (int)(p - URL), URL);
URL = ++p;
/*
* Only one slash: no host, leave slash as part of document
* Two slashes: host follows, strip slashes
*/
if (URL[1] == '/')
URL = (p += 2);
} else {
p = URL;
}
if (!*URL || *URL == '/' || *URL == '.' ||
(u->scheme[0] == '\0' &&
strchr(URL, '/') == NULL && strchr(URL, ':') == NULL))
goto nohost;
p = strpbrk(URL, "/@");
if (p && *p == '@') {
/* username */
for (q = URL, i = 0; (*q != ':') && (*q != '@'); q++)
if (i < URL_USERLEN)
u->user[i++] = *q;
/* password */
if (*q == ':')
for (q++, i = 0; (*q != ':') && (*q != '@'); q++)
if (i < URL_PWDLEN)
u->pwd[i++] = *q;
p++;
} else {
p = URL;
}
/* hostname */
#ifdef INET6
if (*p == '[' && (q = strchr(p + 1, ']')) != NULL &&
(*++q == '\0' || *q == '/' || *q == ':')) {
if ((i = q - p - 2) > MAXHOSTNAMELEN)
i = MAXHOSTNAMELEN;
strncpy(u->host, ++p, i);
p = q;
} else
#endif
for (i = 0; *p && (*p != '/') && (*p != ':'); p++)
if (i < MAXHOSTNAMELEN)
u->host[i++] = *p;
/* port */
if (*p == ':') {
for (q = ++p; *q && (*q != '/'); q++)
if (isdigit(*q))
u->port = u->port * 10 + (*q - '0');
else {
/* invalid port */
_url_seterr(URL_BAD_PORT);
goto ouch;
}
while (*p && (*p != '/'))
p++;
}
nohost:
/* document */
if (!*p)
p = "/";
if (strcasecmp(u->scheme, SCHEME_HTTP) == 0 ||
strcasecmp(u->scheme, SCHEME_HTTPS) == 0) {
const char hexnums[] = "0123456789abcdef";
/* percent-escape whitespace. */
if ((doc = malloc(strlen(p) * 3 + 1)) == NULL) {
_fetch_syserr();
goto ouch;
}
u->doc = doc;
while (*p != '\0') {
if (!isspace(*p)) {
*doc++ = *p++;
} else {
*doc++ = '%';
*doc++ = hexnums[((unsigned int)*p) >> 4];
*doc++ = hexnums[((unsigned int)*p) & 0xf];
p++;
}
}
*doc = '\0';
} else if ((u->doc = strdup(p)) == NULL) {
_fetch_syserr();
goto ouch;
}
DEBUG(fprintf(stderr,
"scheme: [\033[1m%s\033[m]\n"
"user: [\033[1m%s\033[m]\n"
"password: [\033[1m%s\033[m]\n"
"host: [\033[1m%s\033[m]\n"
"port: [\033[1m%d\033[m]\n"
"document: [\033[1m%s\033[m]\n",
u->scheme, u->user, u->pwd,
u->host, u->port, u->doc));
return u;
ouch:
free(u);
return NULL;
}
/*
* Free a URL
*/
void
fetchFreeURL(struct url *u)
{
free(u->doc);
free(u);
}