1 /* $Id: http.c,v 1.29 2007-04-23 08:15:22 marc Exp $
2 Copyright (c) 2006-2007, Index Data.
4 This file is part of Pazpar2.
6 Pazpar2 is free software; you can redistribute it and/or modify it under
7 the terms of the GNU General Public License as published by the Free
8 Software Foundation; either version 2, or (at your option) any later
11 Pazpar2 is distributed in the hope that it will be useful, but WITHOUT ANY
12 WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
16 You should have received a copy of the GNU General Public License
17 along with Pazpar2; see the file LICENSE. If not, write to the
18 Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
23 #include <sys/socket.h>
24 #include <sys/types.h>
40 #include <netinet/in.h>
41 #include <arpa/inet.h>
44 #include <yaz/yaz-util.h>
45 #include <yaz/comstack.h>
53 #include "http_command.h"
55 static void proxy_io(IOCHAN i, int event);
56 static struct http_channel *http_create(const char *addr);
57 static void http_destroy(IOCHAN i);
59 // If this is set, we proxy normal HTTP requests
60 static struct sockaddr_in *proxy_addr = 0;
61 static char proxy_url[256] = "";
62 static char myurl[256] = "";
63 static struct http_buf *http_buf_freelist = 0;
64 static struct http_channel *http_channel_freelist = 0;
66 static struct http_buf *http_buf_create()
70 if (http_buf_freelist)
72 r = http_buf_freelist;
73 http_buf_freelist = http_buf_freelist->next;
76 r = xmalloc(sizeof(struct http_buf));
83 static void http_buf_destroy(struct http_buf *b)
85 b->next = http_buf_freelist;
86 http_buf_freelist = b;
89 static void http_buf_destroy_queue(struct http_buf *b)
101 // Calculate length of chain
102 static int http_buf_len(struct http_buf *b)
105 for (; b; b = b->next)
111 static struct http_buf *http_buf_bybuf(char *b, int len)
113 struct http_buf *res = 0;
114 struct http_buf **p = &res;
119 if (tocopy > HTTP_BUF_SIZE)
120 tocopy = HTTP_BUF_SIZE;
121 *p = http_buf_create();
122 memcpy((*p)->buf, b, tocopy);
131 // Add a (chain of) buffers to the end of an existing queue.
132 static void http_buf_enqueue(struct http_buf **queue, struct http_buf *b)
135 queue = &(*queue)->next;
139 static struct http_buf *http_buf_bywrbuf(WRBUF wrbuf)
141 // Heavens to Betsy (buf)!
142 return http_buf_bybuf(wrbuf_buf(wrbuf), wrbuf_len(wrbuf));
145 // Non-destructively collapse chain of buffers into a string (max *len)
147 static int http_buf_peek(struct http_buf *b, char *buf, int len)
150 while (b && rd < len)
152 int toread = len - rd;
155 memcpy(buf + rd, b->buf + b->offset, toread);
163 // Ddestructively munch up to len from head of queue.
164 static int http_buf_read(struct http_buf **b, char *buf, int len)
167 while ((*b) && rd < len)
169 int toread = len - rd;
170 if (toread > (*b)->len)
172 memcpy(buf + rd, (*b)->buf + (*b)->offset, toread);
174 if (toread < (*b)->len)
177 (*b)->offset += toread;
182 struct http_buf *n = (*b)->next;
183 http_buf_destroy(*b);
191 // Buffers may overlap.
192 static void urldecode(char *i, char *o)
204 sscanf(i, "%2hhx", o);
214 // Warning: Buffers may not overlap
215 void urlencode(const char *i, char *o)
219 if (strchr(" /:", *i))
221 sprintf(o, "%%%.2X", (int) *i);
231 void http_addheader(struct http_response *r, const char *name, const char *value)
233 struct http_channel *c = r->channel;
234 struct http_header *h = nmem_malloc(c->nmem, sizeof *h);
235 h->name = nmem_strdup(c->nmem, name);
236 h->value = nmem_strdup(c->nmem, value);
237 h->next = r->headers;
241 char *http_argbyname(struct http_request *r, char *name)
243 struct http_argument *p;
246 for (p = r->arguments; p; p = p->next)
247 if (!strcmp(p->name, name))
252 char *http_headerbyname(struct http_header *h, char *name)
254 for (; h; h = h->next)
255 if (!strcmp(h->name, name))
260 struct http_response *http_create_response(struct http_channel *c)
262 struct http_response *r = nmem_malloc(c->nmem, sizeof(*r));
263 strcpy(r->code, "200");
271 // Check if buf contains a package (minus payload)
272 static int package_check(const char *buf)
275 while (*buf) // Check if we have a sequence of lines terminated by an empty line
277 char *b = strstr(buf, "\r\n");
282 len += (b - buf) + 2;
290 // Check if we have a request. Return 0 or length
291 // (including trailing CRNL) FIXME: Does not deal gracefully with requests
292 // carrying payload but this is kind of OK since we will reject anything
293 // other than an empty GET
294 static int request_check(struct http_buf *queue)
298 http_buf_peek(queue, tmp, 4096);
299 return package_check(tmp);
302 struct http_response *http_parse_response_buf(struct http_channel *c, const char *buf, int len)
305 struct http_response *r = http_create_response(c);
307 struct http_header **hp = &r->headers;
311 memcpy(tmp, buf, len);
312 for (p = tmp; *p && *p != ' '; p++) // Skip HTTP version
316 for (p2 = p; *p2 && *p2 != ' ' && p2 - p < 3; p2++)
317 r->code[p2 - p] = *p2;
318 if (!(p = strstr(tmp, "\r\n")))
323 if (!(p2 = strstr(p, "\r\n")))
325 if (p == p2) // End of headers
329 struct http_header *h = *hp = nmem_malloc(c->nmem, sizeof(*h));
330 char *value = strchr(p, ':');
334 h->name = nmem_strdup(c->nmem, p);
335 while (isspace(*value))
337 if (value >= p2) // Empty header;
344 h->value = nmem_strdup(c->nmem, value);
353 struct http_request *http_parse_request(struct http_channel *c, struct http_buf **queue,
356 struct http_request *r = nmem_malloc(c->nmem, sizeof(*r));
363 if (http_buf_read(queue, buf, len) < len)
371 for (p = buf, p2 = r->method; *p && *p != ' ' && p - buf < 19; p++)
375 yaz_log(YLOG_WARN, "Unexpected HTTP method in request");
380 if (!(buf = strchr(buf, ' ')))
382 yaz_log(YLOG_WARN, "Syntax error in request (1)");
386 if (!(p = strchr(buf, ' ')))
388 yaz_log(YLOG_WARN, "Syntax error in request (2)");
392 if ((p2 = strchr(buf, '?'))) // Do we have arguments?
394 r->path = nmem_strdup(c->nmem, buf);
397 r->search = nmem_strdup(c->nmem, p2);
401 struct http_argument *a;
402 char *equal = strchr(p2, '=');
403 char *eoa = strchr(p2, '&');
406 yaz_log(YLOG_WARN, "Expected '=' in argument");
410 eoa = equal + strlen(equal); // last argument
413 a = nmem_malloc(c->nmem, sizeof(struct http_argument));
415 a->name = nmem_strdup(c->nmem, p2);
416 urldecode(equal, equal);
417 a->value = nmem_strdup(c->nmem, equal);
418 a->next = r->arguments;
425 if (strncmp(buf, "HTTP/", 5))
426 strcpy(r->http_version, "1.0");
430 if (!(p = strstr(buf, "\r\n")))
434 strcpy(r->http_version, buf);
437 strcpy(c->version, r->http_version);
442 if (!(p = strstr(buf, "\r\n")))
448 struct http_header *h = nmem_malloc(c->nmem, sizeof(*h));
449 if (!(p2 = strchr(buf, ':')))
452 h->name = nmem_strdup(c->nmem, buf);
455 if (p2 >= p) // Empty header?
461 h->value = nmem_strdup(c->nmem, p2);
462 h->next = r->headers;
471 static struct http_buf *http_serialize_response(struct http_channel *c,
472 struct http_response *r)
474 struct http_header *h;
476 wrbuf_rewind(c->wrbuf);
477 wrbuf_printf(c->wrbuf, "HTTP/1.1 %s %s\r\n", r->code, r->msg);
478 for (h = r->headers; h; h = h->next)
479 wrbuf_printf(c->wrbuf, "%s: %s\r\n", h->name, h->value);
482 wrbuf_printf(c->wrbuf, "Content-length: %d\r\n", r->payload ?
483 (int) strlen(r->payload) : 0);
484 wrbuf_printf(c->wrbuf, "Content-type: text/xml\r\n");
486 wrbuf_puts(c->wrbuf, "\r\n");
489 wrbuf_puts(c->wrbuf, r->payload);
491 return http_buf_bywrbuf(c->wrbuf);
494 // Serialize a HTTP request
495 static struct http_buf *http_serialize_request(struct http_request *r)
497 struct http_channel *c = r->channel;
498 struct http_header *h;
499 struct http_argument *a;
501 wrbuf_rewind(c->wrbuf);
502 wrbuf_printf(c->wrbuf, "%s %s", r->method, r->path);
506 wrbuf_putc(c->wrbuf, '?');
507 for (a = r->arguments; a; a = a->next) {
508 if (a != r->arguments)
509 wrbuf_putc(c->wrbuf, '&');
510 wrbuf_printf(c->wrbuf, "%s=%s", a->name, a->value);
514 wrbuf_printf(c->wrbuf, " HTTP/%s\r\n", r->http_version);
516 for (h = r->headers; h; h = h->next)
517 wrbuf_printf(c->wrbuf, "%s: %s\r\n", h->name, h->value);
519 wrbuf_puts(c->wrbuf, "\r\n");
521 return http_buf_bywrbuf(c->wrbuf);
525 static int http_weshouldproxy(struct http_request *rq)
527 if (proxy_addr && !strstr(rq->path, "search.pz2"))
533 struct http_header * http_header_append(struct http_channel *ch,
534 struct http_header * hp,
538 struct http_header *hpnew = 0;
543 while (hp && hp->next)
546 if(name && strlen(name)&& value && strlen(value)){
547 hpnew = nmem_malloc(ch->nmem, sizeof *hpnew);
548 hpnew->name = nmem_strdup(ch->nmem, name);
549 hpnew->value = nmem_strdup(ch->nmem, value);
563 static int http_proxy(struct http_request *rq)
565 struct http_channel *c = rq->channel;
566 struct http_proxy *p = c->proxy;
567 struct http_header *hp;
568 struct http_buf *requestbuf;
569 char server_via[128] = "";
570 char server_port[16] = "";
571 struct conf_server *ser = global_parameters.server;
573 if (!p) // This is a new connection. Create a proxy channel
580 if (!(pe = getprotobyname("tcp"))) {
583 if ((sock = socket(PF_INET, SOCK_STREAM, pe->p_proto)) < 0)
585 yaz_log(YLOG_WARN|YLOG_ERRNO, "socket");
588 if (setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (char*)
589 &one, sizeof(one)) < 0)
591 if ((flags = fcntl(sock, F_GETFL, 0)) < 0)
592 yaz_log(YLOG_FATAL|YLOG_ERRNO, "fcntl");
593 if (fcntl(sock, F_SETFL, flags | O_NONBLOCK) < 0)
594 yaz_log(YLOG_FATAL|YLOG_ERRNO, "fcntl2");
595 if (connect(sock, (struct sockaddr *) proxy_addr,
596 sizeof(*proxy_addr)) < 0)
597 if (errno != EINPROGRESS)
599 yaz_log(YLOG_WARN|YLOG_ERRNO, "Proxy connect");
603 p = xmalloc(sizeof(struct http_proxy));
606 p->first_response = 1;
608 // We will add EVENT_OUTPUT below
609 p->iochan = iochan_create(sock, proxy_io, EVENT_INPUT);
610 iochan_setdata(p->iochan, p);
611 pazpar2_add_channel(p->iochan);
614 // Do _not_ modify Host: header, just checking it's existence
615 for (hp = rq->headers; hp; hp = hp->next)
616 if (!strcmp(hp->name, "Host"))
620 yaz_log(YLOG_WARN, "Failed to find Host header in proxy");
624 // Add new header about paraz2 version, host, remote client address, etc.
627 hp = http_header_append(c, hp,
628 "X-Pazpar2-Version", PACKAGE_VERSION);
629 hp = http_header_append(c, hp,
630 "X-Pazpar2-Server-Host", ser->host);
631 sprintf(server_port, "%d", ser->port);
632 hp = http_header_append(c, hp,
633 "X-Pazpar2-Server-Port", server_port);
634 sprintf(server_via, "1.1 %s:%s (%s/%s)",
635 ser->host, server_port, PACKAGE_NAME, PACKAGE_VERSION);
636 hp = http_header_append(c, hp, "Via" , server_via);
637 hp = http_header_append(c, hp, "X-Forwarded-For", c->addr);
640 requestbuf = http_serialize_request(rq);
641 http_buf_enqueue(&p->oqueue, requestbuf);
642 iochan_setflag(p->iochan, EVENT_OUTPUT);
646 void http_send_response(struct http_channel *ch)
648 struct http_response *rs = ch->response;
652 hb = http_serialize_response(ch, rs);
655 yaz_log(YLOG_WARN, "Failed to serialize HTTP response");
656 http_destroy(ch->iochan);
660 http_buf_enqueue(&ch->oqueue, hb);
661 iochan_setflag(ch->iochan, EVENT_OUTPUT);
662 ch->state = Http_Idle;
666 static void http_io(IOCHAN i, int event)
668 struct http_channel *hc = iochan_getdata(i);
673 struct http_buf *htbuf;
676 htbuf = http_buf_create();
677 res = read(iochan_getfd(i), htbuf->buf, HTTP_BUF_SIZE -1);
678 if (res == -1 && errno == EAGAIN)
680 http_buf_destroy(htbuf);
685 http_buf_destroy(htbuf);
691 htbuf->buf[res] = '\0';
693 http_buf_enqueue(&hc->iqueue, htbuf);
696 if (hc->state == Http_Busy)
698 if ((reqlen = request_check(hc->iqueue)) <= 2)
701 nmem_reset(hc->nmem);
702 if (!(hc->request = http_parse_request(hc, &hc->iqueue, reqlen)))
704 yaz_log(YLOG_WARN, "Failed to parse request");
709 yaz_log(YLOG_LOG, "Request: %s %s%s%s", hc->request->method,
711 *hc->request->search ? "?" : "",
712 hc->request->search);
713 if (http_weshouldproxy(hc->request))
714 http_proxy(hc->request);
717 // Execute our business logic!
718 hc->state = Http_Busy;
723 yaz_log(YLOG_DEBUG, "We think we have more input to read. Forcing event");
724 iochan_setevent(i, EVENT_INPUT);
732 struct http_buf *wb = hc->oqueue;
733 res = write(iochan_getfd(hc->iochan), wb->buf + wb->offset, wb->len);
736 yaz_log(YLOG_WARN|YLOG_ERRNO, "write");
742 hc->oqueue = hc->oqueue->next;
743 http_buf_destroy(wb);
751 if (!strcmp(hc->version, "1.0"))
758 iochan_clearflag(i, EVENT_OUTPUT);
760 iochan_setevent(hc->iochan, EVENT_INPUT);
765 if (!hc->oqueue && hc->proxy && !hc->proxy->iochan)
766 http_destroy(i); // Server closed; we're done
769 yaz_log(YLOG_WARN, "Unexpected event on connection");
775 // If this hostname contains our proxy host as a prefix, replace with myurl
776 static char *sub_hostname(struct http_channel *c, char *buf)
779 if (strlen(buf) > 1023)
781 if (strncmp(buf, "http://", 7))
783 if (!strncmp(buf + 7, proxy_url, strlen(proxy_url)))
786 strcat(tmp, buf + strlen(proxy_url) + 7);
787 return nmem_strdup(c->nmem, tmp);
793 // Handles I/O on a client connection to a backend web server (proxy mode)
794 static void proxy_io(IOCHAN pi, int event)
796 struct http_proxy *pc = iochan_getdata(pi);
797 struct http_channel *hc = pc->channel;
802 struct http_buf *htbuf;
805 htbuf = http_buf_create();
806 res = read(iochan_getfd(pi), htbuf->buf, HTTP_BUF_SIZE -1);
807 if (res == 0 || (res < 0 && errno != EINPROGRESS))
811 yaz_log(YLOG_WARN, "Proxy read came up short");
812 // Close channel and alert client HTTP channel that we're gone
813 http_buf_destroy(htbuf);
814 close(iochan_getfd(pi));
820 http_destroy(hc->iochan);
826 htbuf->buf[res] = '\0';
830 if (pc->first_response) // Check if this is a redirect
833 if ((len = package_check(htbuf->buf)))
835 struct http_response *res = http_parse_response_buf(hc, htbuf->buf, len);
838 struct http_header *h;
839 for (h = res->headers; h; h = h->next)
840 if (!strcmp(h->name, "Location"))
842 // We found a location header. Rewrite it.
843 struct http_buf *buf;
844 h->value = sub_hostname(hc, h->value);
845 buf = http_serialize_response(hc, res);
846 yaz_log(YLOG_LOG, "Proxy rewrite");
847 http_buf_enqueue(&hc->oqueue, buf);
853 pc->first_response = 0;
856 // Write any remaining payload
857 if (htbuf->len - htbuf->offset > 0)
858 http_buf_enqueue(&hc->oqueue, htbuf);
860 iochan_setflag(hc->iochan, EVENT_OUTPUT);
863 if (!(htbuf = pc->oqueue))
865 iochan_clearflag(pi, EVENT_OUTPUT);
868 res = write(iochan_getfd(pi), htbuf->buf + htbuf->offset, htbuf->len);
871 yaz_log(YLOG_WARN|YLOG_ERRNO, "write");
872 http_destroy(hc->iochan);
875 if (res == htbuf->len)
877 struct http_buf *np = htbuf->next;
878 http_buf_destroy(htbuf);
884 htbuf->offset += res;
888 iochan_setflags(pi, EVENT_INPUT); // Turns off output flag
892 yaz_log(YLOG_WARN, "Unexpected event on connection");
893 http_destroy(hc->iochan);
898 static void http_destroy(IOCHAN i)
900 struct http_channel *s = iochan_getdata(i);
904 if (s->proxy->iochan)
906 close(iochan_getfd(s->proxy->iochan));
907 iochan_destroy(s->proxy->iochan);
909 http_buf_destroy_queue(s->proxy->oqueue);
912 s->next = http_channel_freelist;
913 http_channel_freelist = s;
914 close(iochan_getfd(i));
918 static struct http_channel *http_create(const char *addr)
920 struct http_channel *r = http_channel_freelist;
924 http_channel_freelist = r->next;
926 wrbuf_rewind(r->wrbuf);
930 r = xmalloc(sizeof(struct http_channel));
931 r->nmem = nmem_create();
932 r->wrbuf = wrbuf_alloc();
936 r->iqueue = r->oqueue = 0;
937 r->state = Http_Idle;
942 yaz_log(YLOG_WARN, "Invalid HTTP forward address");
945 strcpy(r->addr, addr);
950 /* Accept a new command connection */
951 static void http_accept(IOCHAN i, int event)
953 struct sockaddr_in addr;
954 int fd = iochan_getfd(i);
959 struct http_channel *ch;
962 if ((s = accept(fd, (struct sockaddr *) &addr, &len)) < 0)
964 yaz_log(YLOG_WARN|YLOG_ERRNO, "accept");
967 if ((flags = fcntl(s, F_GETFL, 0)) < 0)
968 yaz_log(YLOG_FATAL|YLOG_ERRNO, "fcntl");
969 if (fcntl(s, F_SETFL, flags | O_NONBLOCK) < 0)
970 yaz_log(YLOG_FATAL|YLOG_ERRNO, "fcntl2");
972 yaz_log(YLOG_DEBUG, "New command connection");
973 c = iochan_create(s, http_io, EVENT_INPUT | EVENT_EXCEPT);
975 ch = http_create(inet_ntoa(addr.sin_addr));
977 iochan_setdata(c, ch);
979 pazpar2_add_channel(c);
982 /* Create a http-channel listener, syntax [host:]port */
983 void http_init(const char *addr)
988 struct sockaddr_in myaddr;
993 yaz_log(YLOG_LOG, "HTTP listener %s", addr);
995 memset(&myaddr, 0, sizeof myaddr);
996 myaddr.sin_family = AF_INET;
997 pp = strchr(addr, ':');
1000 int len = pp - addr;
1004 strncpy(hostname, addr, len);
1005 hostname[len] = '\0';
1006 if (!(he = gethostbyname(hostname)))
1007 yaz_log(YLOG_FATAL, "Unable to resolve '%s'", hostname);
1009 memcpy(&myaddr.sin_addr.s_addr, he->h_addr_list[0], he->h_length);
1010 port = atoi(pp + 1);
1012 yaz_log(YLOG_LOG, "HTTP address %s:%d",
1013 "" == he->h_addr_list[0] ? he->h_addr_list[0] : "127.0.0.1" ,
1020 myaddr.sin_addr.s_addr = INADDR_ANY;
1023 myaddr.sin_port = htons(port);
1025 if (!(p = getprotobyname("tcp"))) {
1028 if ((l = socket(PF_INET, SOCK_STREAM, p->p_proto)) < 0)
1029 yaz_log(YLOG_FATAL|YLOG_ERRNO, "socket");
1030 if (setsockopt(l, SOL_SOCKET, SO_REUSEADDR, (char*)
1031 &one, sizeof(one)) < 0)
1034 if (bind(l, (struct sockaddr *) &myaddr, sizeof myaddr) < 0)
1035 yaz_log(YLOG_FATAL|YLOG_ERRNO, "bind");
1036 if (listen(l, SOMAXCONN) < 0)
1037 yaz_log(YLOG_FATAL|YLOG_ERRNO, "listen");
1039 c = iochan_create(l, http_accept, EVENT_INPUT | EVENT_EXCEPT);
1040 pazpar2_add_channel(c);
1043 void http_set_proxyaddr(char *host, char *base_url)
1049 strcpy(myurl, base_url);
1050 strcpy(proxy_url, host);
1051 p = strchr(host, ':');
1052 yaz_log(YLOG_DEBUG, "Proxying for %s", host);
1053 yaz_log(YLOG_LOG, "HTTP backend %s", proxy_url);
1060 if (!(he = gethostbyname(host)))
1062 fprintf(stderr, "Failed to lookup '%s'\n", host);
1065 proxy_addr = xmalloc(sizeof(struct sockaddr_in));
1066 proxy_addr->sin_family = he->h_addrtype;
1067 memcpy(&proxy_addr->sin_addr.s_addr, he->h_addr_list[0], he->h_length);
1068 proxy_addr->sin_port = htons(port);
1074 * indent-tabs-mode: nil
1076 * vim: shiftwidth=4 tabstop=8 expandtab