Merge branch 'jk/prune-with-corrupt-refs' into maint
[alt-git.git] / connect.c
blobce0e1214234244236a407e9f3004e619d26c54a8
1 #include "git-compat-util.h"
2 #include "cache.h"
3 #include "pkt-line.h"
4 #include "quote.h"
5 #include "refs.h"
6 #include "run-command.h"
7 #include "remote.h"
8 #include "connect.h"
9 #include "url.h"
10 #include "string-list.h"
11 #include "sha1-array.h"
13 static char *server_capabilities;
14 static const char *parse_feature_value(const char *, const char *, int *);
16 static int check_ref(const char *name, unsigned int flags)
18 if (!flags)
19 return 1;
21 if (!skip_prefix(name, "refs/", &name))
22 return 0;
24 /* REF_NORMAL means that we don't want the magic fake tag refs */
25 if ((flags & REF_NORMAL) && check_refname_format(name, 0))
26 return 0;
28 /* REF_HEADS means that we want regular branch heads */
29 if ((flags & REF_HEADS) && starts_with(name, "heads/"))
30 return 1;
32 /* REF_TAGS means that we want tags */
33 if ((flags & REF_TAGS) && starts_with(name, "tags/"))
34 return 1;
36 /* All type bits clear means that we are ok with anything */
37 return !(flags & ~REF_NORMAL);
40 int check_ref_type(const struct ref *ref, int flags)
42 return check_ref(ref->name, flags);
45 static void die_initial_contact(int got_at_least_one_head)
47 if (got_at_least_one_head)
48 die("The remote end hung up upon initial contact");
49 else
50 die("Could not read from remote repository.\n\n"
51 "Please make sure you have the correct access rights\n"
52 "and the repository exists.");
55 static void parse_one_symref_info(struct string_list *symref, const char *val, int len)
57 char *sym, *target;
58 struct string_list_item *item;
60 if (!len)
61 return; /* just "symref" */
62 /* e.g. "symref=HEAD:refs/heads/master" */
63 sym = xmemdupz(val, len);
64 target = strchr(sym, ':');
65 if (!target)
66 /* just "symref=something" */
67 goto reject;
68 *(target++) = '\0';
69 if (check_refname_format(sym, REFNAME_ALLOW_ONELEVEL) ||
70 check_refname_format(target, REFNAME_ALLOW_ONELEVEL))
71 /* "symref=bogus:pair */
72 goto reject;
73 item = string_list_append(symref, sym);
74 item->util = target;
75 return;
76 reject:
77 free(sym);
78 return;
81 static void annotate_refs_with_symref_info(struct ref *ref)
83 struct string_list symref = STRING_LIST_INIT_DUP;
84 const char *feature_list = server_capabilities;
86 while (feature_list) {
87 int len;
88 const char *val;
90 val = parse_feature_value(feature_list, "symref", &len);
91 if (!val)
92 break;
93 parse_one_symref_info(&symref, val, len);
94 feature_list = val + 1;
96 string_list_sort(&symref);
98 for (; ref; ref = ref->next) {
99 struct string_list_item *item;
100 item = string_list_lookup(&symref, ref->name);
101 if (!item)
102 continue;
103 ref->symref = xstrdup((char *)item->util);
105 string_list_clear(&symref, 0);
109 * Read all the refs from the other end
111 struct ref **get_remote_heads(int in, char *src_buf, size_t src_len,
112 struct ref **list, unsigned int flags,
113 struct sha1_array *extra_have,
114 struct sha1_array *shallow_points)
116 struct ref **orig_list = list;
117 int got_at_least_one_head = 0;
119 *list = NULL;
120 for (;;) {
121 struct ref *ref;
122 unsigned char old_sha1[20];
123 char *name;
124 int len, name_len;
125 char *buffer = packet_buffer;
126 const char *arg;
128 len = packet_read(in, &src_buf, &src_len,
129 packet_buffer, sizeof(packet_buffer),
130 PACKET_READ_GENTLE_ON_EOF |
131 PACKET_READ_CHOMP_NEWLINE);
132 if (len < 0)
133 die_initial_contact(got_at_least_one_head);
135 if (!len)
136 break;
138 if (len > 4 && skip_prefix(buffer, "ERR ", &arg))
139 die("remote error: %s", arg);
141 if (len == 48 && skip_prefix(buffer, "shallow ", &arg)) {
142 if (get_sha1_hex(arg, old_sha1))
143 die("protocol error: expected shallow sha-1, got '%s'", arg);
144 if (!shallow_points)
145 die("repository on the other end cannot be shallow");
146 sha1_array_append(shallow_points, old_sha1);
147 continue;
150 if (len < 42 || get_sha1_hex(buffer, old_sha1) || buffer[40] != ' ')
151 die("protocol error: expected sha/ref, got '%s'", buffer);
152 name = buffer + 41;
154 name_len = strlen(name);
155 if (len != name_len + 41) {
156 free(server_capabilities);
157 server_capabilities = xstrdup(name + name_len + 1);
160 if (extra_have && !strcmp(name, ".have")) {
161 sha1_array_append(extra_have, old_sha1);
162 continue;
165 if (!check_ref(name, flags))
166 continue;
167 ref = alloc_ref(buffer + 41);
168 hashcpy(ref->old_sha1, old_sha1);
169 *list = ref;
170 list = &ref->next;
171 got_at_least_one_head = 1;
174 annotate_refs_with_symref_info(*orig_list);
176 return list;
179 static const char *parse_feature_value(const char *feature_list, const char *feature, int *lenp)
181 int len;
183 if (!feature_list)
184 return NULL;
186 len = strlen(feature);
187 while (*feature_list) {
188 const char *found = strstr(feature_list, feature);
189 if (!found)
190 return NULL;
191 if (feature_list == found || isspace(found[-1])) {
192 const char *value = found + len;
193 /* feature with no value (e.g., "thin-pack") */
194 if (!*value || isspace(*value)) {
195 if (lenp)
196 *lenp = 0;
197 return value;
199 /* feature with a value (e.g., "agent=git/1.2.3") */
200 else if (*value == '=') {
201 value++;
202 if (lenp)
203 *lenp = strcspn(value, " \t\n");
204 return value;
207 * otherwise we matched a substring of another feature;
208 * keep looking
211 feature_list = found + 1;
213 return NULL;
216 int parse_feature_request(const char *feature_list, const char *feature)
218 return !!parse_feature_value(feature_list, feature, NULL);
221 const char *server_feature_value(const char *feature, int *len)
223 return parse_feature_value(server_capabilities, feature, len);
226 int server_supports(const char *feature)
228 return !!server_feature_value(feature, NULL);
231 enum protocol {
232 PROTO_LOCAL = 1,
233 PROTO_FILE,
234 PROTO_SSH,
235 PROTO_GIT
238 int url_is_local_not_ssh(const char *url)
240 const char *colon = strchr(url, ':');
241 const char *slash = strchr(url, '/');
242 return !colon || (slash && slash < colon) ||
243 has_dos_drive_prefix(url);
246 static const char *prot_name(enum protocol protocol)
248 switch (protocol) {
249 case PROTO_LOCAL:
250 case PROTO_FILE:
251 return "file";
252 case PROTO_SSH:
253 return "ssh";
254 case PROTO_GIT:
255 return "git";
256 default:
257 return "unkown protocol";
261 static enum protocol get_protocol(const char *name)
263 if (!strcmp(name, "ssh"))
264 return PROTO_SSH;
265 if (!strcmp(name, "git"))
266 return PROTO_GIT;
267 if (!strcmp(name, "git+ssh"))
268 return PROTO_SSH;
269 if (!strcmp(name, "ssh+git"))
270 return PROTO_SSH;
271 if (!strcmp(name, "file"))
272 return PROTO_FILE;
273 die("I don't handle protocol '%s'", name);
276 static char *host_end(char **hoststart, int removebrackets)
278 char *host = *hoststart;
279 char *end;
280 char *start = strstr(host, "@[");
281 if (start)
282 start++; /* Jump over '@' */
283 else
284 start = host;
285 if (start[0] == '[') {
286 end = strchr(start + 1, ']');
287 if (end) {
288 if (removebrackets) {
289 *end = 0;
290 memmove(start, start + 1, end - start);
291 end++;
293 } else
294 end = host;
295 } else
296 end = host;
297 return end;
300 #define STR_(s) # s
301 #define STR(s) STR_(s)
303 static void get_host_and_port(char **host, const char **port)
305 char *colon, *end;
306 end = host_end(host, 1);
307 colon = strchr(end, ':');
308 if (colon) {
309 long portnr = strtol(colon + 1, &end, 10);
310 if (end != colon + 1 && *end == '\0' && 0 <= portnr && portnr < 65536) {
311 *colon = 0;
312 *port = colon + 1;
317 static void enable_keepalive(int sockfd)
319 int ka = 1;
321 if (setsockopt(sockfd, SOL_SOCKET, SO_KEEPALIVE, &ka, sizeof(ka)) < 0)
322 fprintf(stderr, "unable to set SO_KEEPALIVE on socket: %s\n",
323 strerror(errno));
326 #ifndef NO_IPV6
328 static const char *ai_name(const struct addrinfo *ai)
330 static char addr[NI_MAXHOST];
331 if (getnameinfo(ai->ai_addr, ai->ai_addrlen, addr, sizeof(addr), NULL, 0,
332 NI_NUMERICHOST) != 0)
333 strcpy(addr, "(unknown)");
335 return addr;
339 * Returns a connected socket() fd, or else die()s.
341 static int git_tcp_connect_sock(char *host, int flags)
343 struct strbuf error_message = STRBUF_INIT;
344 int sockfd = -1;
345 const char *port = STR(DEFAULT_GIT_PORT);
346 struct addrinfo hints, *ai0, *ai;
347 int gai;
348 int cnt = 0;
350 get_host_and_port(&host, &port);
351 if (!*port)
352 port = "<none>";
354 memset(&hints, 0, sizeof(hints));
355 hints.ai_socktype = SOCK_STREAM;
356 hints.ai_protocol = IPPROTO_TCP;
358 if (flags & CONNECT_VERBOSE)
359 fprintf(stderr, "Looking up %s ... ", host);
361 gai = getaddrinfo(host, port, &hints, &ai);
362 if (gai)
363 die("Unable to look up %s (port %s) (%s)", host, port, gai_strerror(gai));
365 if (flags & CONNECT_VERBOSE)
366 fprintf(stderr, "done.\nConnecting to %s (port %s) ... ", host, port);
368 for (ai0 = ai; ai; ai = ai->ai_next, cnt++) {
369 sockfd = socket(ai->ai_family,
370 ai->ai_socktype, ai->ai_protocol);
371 if ((sockfd < 0) ||
372 (connect(sockfd, ai->ai_addr, ai->ai_addrlen) < 0)) {
373 strbuf_addf(&error_message, "%s[%d: %s]: errno=%s\n",
374 host, cnt, ai_name(ai), strerror(errno));
375 if (0 <= sockfd)
376 close(sockfd);
377 sockfd = -1;
378 continue;
380 if (flags & CONNECT_VERBOSE)
381 fprintf(stderr, "%s ", ai_name(ai));
382 break;
385 freeaddrinfo(ai0);
387 if (sockfd < 0)
388 die("unable to connect to %s:\n%s", host, error_message.buf);
390 enable_keepalive(sockfd);
392 if (flags & CONNECT_VERBOSE)
393 fprintf(stderr, "done.\n");
395 strbuf_release(&error_message);
397 return sockfd;
400 #else /* NO_IPV6 */
403 * Returns a connected socket() fd, or else die()s.
405 static int git_tcp_connect_sock(char *host, int flags)
407 struct strbuf error_message = STRBUF_INIT;
408 int sockfd = -1;
409 const char *port = STR(DEFAULT_GIT_PORT);
410 char *ep;
411 struct hostent *he;
412 struct sockaddr_in sa;
413 char **ap;
414 unsigned int nport;
415 int cnt;
417 get_host_and_port(&host, &port);
419 if (flags & CONNECT_VERBOSE)
420 fprintf(stderr, "Looking up %s ... ", host);
422 he = gethostbyname(host);
423 if (!he)
424 die("Unable to look up %s (%s)", host, hstrerror(h_errno));
425 nport = strtoul(port, &ep, 10);
426 if ( ep == port || *ep ) {
427 /* Not numeric */
428 struct servent *se = getservbyname(port,"tcp");
429 if ( !se )
430 die("Unknown port %s", port);
431 nport = se->s_port;
434 if (flags & CONNECT_VERBOSE)
435 fprintf(stderr, "done.\nConnecting to %s (port %s) ... ", host, port);
437 for (cnt = 0, ap = he->h_addr_list; *ap; ap++, cnt++) {
438 memset(&sa, 0, sizeof sa);
439 sa.sin_family = he->h_addrtype;
440 sa.sin_port = htons(nport);
441 memcpy(&sa.sin_addr, *ap, he->h_length);
443 sockfd = socket(he->h_addrtype, SOCK_STREAM, 0);
444 if ((sockfd < 0) ||
445 connect(sockfd, (struct sockaddr *)&sa, sizeof sa) < 0) {
446 strbuf_addf(&error_message, "%s[%d: %s]: errno=%s\n",
447 host,
448 cnt,
449 inet_ntoa(*(struct in_addr *)&sa.sin_addr),
450 strerror(errno));
451 if (0 <= sockfd)
452 close(sockfd);
453 sockfd = -1;
454 continue;
456 if (flags & CONNECT_VERBOSE)
457 fprintf(stderr, "%s ",
458 inet_ntoa(*(struct in_addr *)&sa.sin_addr));
459 break;
462 if (sockfd < 0)
463 die("unable to connect to %s:\n%s", host, error_message.buf);
465 enable_keepalive(sockfd);
467 if (flags & CONNECT_VERBOSE)
468 fprintf(stderr, "done.\n");
470 return sockfd;
473 #endif /* NO_IPV6 */
476 static void git_tcp_connect(int fd[2], char *host, int flags)
478 int sockfd = git_tcp_connect_sock(host, flags);
480 fd[0] = sockfd;
481 fd[1] = dup(sockfd);
485 static char *git_proxy_command;
487 static int git_proxy_command_options(const char *var, const char *value,
488 void *cb)
490 if (!strcmp(var, "core.gitproxy")) {
491 const char *for_pos;
492 int matchlen = -1;
493 int hostlen;
494 const char *rhost_name = cb;
495 int rhost_len = strlen(rhost_name);
497 if (git_proxy_command)
498 return 0;
499 if (!value)
500 return config_error_nonbool(var);
501 /* [core]
502 * ;# matches www.kernel.org as well
503 * gitproxy = netcatter-1 for kernel.org
504 * gitproxy = netcatter-2 for sample.xz
505 * gitproxy = netcatter-default
507 for_pos = strstr(value, " for ");
508 if (!for_pos)
509 /* matches everybody */
510 matchlen = strlen(value);
511 else {
512 hostlen = strlen(for_pos + 5);
513 if (rhost_len < hostlen)
514 matchlen = -1;
515 else if (!strncmp(for_pos + 5,
516 rhost_name + rhost_len - hostlen,
517 hostlen) &&
518 ((rhost_len == hostlen) ||
519 rhost_name[rhost_len - hostlen -1] == '.'))
520 matchlen = for_pos - value;
521 else
522 matchlen = -1;
524 if (0 <= matchlen) {
525 /* core.gitproxy = none for kernel.org */
526 if (matchlen == 4 &&
527 !memcmp(value, "none", 4))
528 matchlen = 0;
529 git_proxy_command = xmemdupz(value, matchlen);
531 return 0;
534 return git_default_config(var, value, cb);
537 static int git_use_proxy(const char *host)
539 git_proxy_command = getenv("GIT_PROXY_COMMAND");
540 git_config(git_proxy_command_options, (void*)host);
541 return (git_proxy_command && *git_proxy_command);
544 static struct child_process *git_proxy_connect(int fd[2], char *host)
546 const char *port = STR(DEFAULT_GIT_PORT);
547 struct child_process *proxy;
549 get_host_and_port(&host, &port);
551 proxy = xmalloc(sizeof(*proxy));
552 child_process_init(proxy);
553 argv_array_push(&proxy->args, git_proxy_command);
554 argv_array_push(&proxy->args, host);
555 argv_array_push(&proxy->args, port);
556 proxy->in = -1;
557 proxy->out = -1;
558 if (start_command(proxy))
559 die("cannot start proxy %s", git_proxy_command);
560 fd[0] = proxy->out; /* read from proxy stdout */
561 fd[1] = proxy->in; /* write to proxy stdin */
562 return proxy;
565 static char *get_port(char *host)
567 char *end;
568 char *p = strchr(host, ':');
570 if (p) {
571 long port = strtol(p + 1, &end, 10);
572 if (end != p + 1 && *end == '\0' && 0 <= port && port < 65536) {
573 *p = '\0';
574 return p+1;
578 return NULL;
582 * Extract protocol and relevant parts from the specified connection URL.
583 * The caller must free() the returned strings.
585 static enum protocol parse_connect_url(const char *url_orig, char **ret_host,
586 char **ret_path)
588 char *url;
589 char *host, *path;
590 char *end;
591 int separator = '/';
592 enum protocol protocol = PROTO_LOCAL;
594 if (is_url(url_orig))
595 url = url_decode(url_orig);
596 else
597 url = xstrdup(url_orig);
599 host = strstr(url, "://");
600 if (host) {
601 *host = '\0';
602 protocol = get_protocol(url);
603 host += 3;
604 } else {
605 host = url;
606 if (!url_is_local_not_ssh(url)) {
607 protocol = PROTO_SSH;
608 separator = ':';
613 * Don't do destructive transforms as protocol code does
614 * '[]' unwrapping in get_host_and_port()
616 end = host_end(&host, 0);
618 if (protocol == PROTO_LOCAL)
619 path = end;
620 else if (protocol == PROTO_FILE && has_dos_drive_prefix(end))
621 path = end; /* "file://$(pwd)" may be "file://C:/projects/repo" */
622 else
623 path = strchr(end, separator);
625 if (!path || !*path)
626 die("No path specified. See 'man git-pull' for valid url syntax");
629 * null-terminate hostname and point path to ~ for URL's like this:
630 * ssh://host.xz/~user/repo
633 end = path; /* Need to \0 terminate host here */
634 if (separator == ':')
635 path++; /* path starts after ':' */
636 if (protocol == PROTO_GIT || protocol == PROTO_SSH) {
637 if (path[1] == '~')
638 path++;
641 path = xstrdup(path);
642 *end = '\0';
644 *ret_host = xstrdup(host);
645 *ret_path = path;
646 free(url);
647 return protocol;
650 static struct child_process no_fork = CHILD_PROCESS_INIT;
653 * This returns a dummy child_process if the transport protocol does not
654 * need fork(2), or a struct child_process object if it does. Once done,
655 * finish the connection with finish_connect() with the value returned from
656 * this function (it is safe to call finish_connect() with NULL to support
657 * the former case).
659 * If it returns, the connect is successful; it just dies on errors (this
660 * will hopefully be changed in a libification effort, to return NULL when
661 * the connection failed).
663 struct child_process *git_connect(int fd[2], const char *url,
664 const char *prog, int flags)
666 char *hostandport, *path;
667 struct child_process *conn = &no_fork;
668 enum protocol protocol;
669 struct strbuf cmd = STRBUF_INIT;
671 /* Without this we cannot rely on waitpid() to tell
672 * what happened to our children.
674 signal(SIGCHLD, SIG_DFL);
676 protocol = parse_connect_url(url, &hostandport, &path);
677 if ((flags & CONNECT_DIAG_URL) && (protocol != PROTO_SSH)) {
678 printf("Diag: url=%s\n", url ? url : "NULL");
679 printf("Diag: protocol=%s\n", prot_name(protocol));
680 printf("Diag: hostandport=%s\n", hostandport ? hostandport : "NULL");
681 printf("Diag: path=%s\n", path ? path : "NULL");
682 conn = NULL;
683 } else if (protocol == PROTO_GIT) {
685 * Set up virtual host information based on where we will
686 * connect, unless the user has overridden us in
687 * the environment.
689 char *target_host = getenv("GIT_OVERRIDE_VIRTUAL_HOST");
690 if (target_host)
691 target_host = xstrdup(target_host);
692 else
693 target_host = xstrdup(hostandport);
695 /* These underlying connection commands die() if they
696 * cannot connect.
698 if (git_use_proxy(hostandport))
699 conn = git_proxy_connect(fd, hostandport);
700 else
701 git_tcp_connect(fd, hostandport, flags);
703 * Separate original protocol components prog and path
704 * from extended host header with a NUL byte.
706 * Note: Do not add any other headers here! Doing so
707 * will cause older git-daemon servers to crash.
709 packet_write(fd[1],
710 "%s %s%chost=%s%c",
711 prog, path, 0,
712 target_host, 0);
713 free(target_host);
714 } else {
715 conn = xmalloc(sizeof(*conn));
716 child_process_init(conn);
718 strbuf_addstr(&cmd, prog);
719 strbuf_addch(&cmd, ' ');
720 sq_quote_buf(&cmd, path);
722 conn->in = conn->out = -1;
723 if (protocol == PROTO_SSH) {
724 const char *ssh;
725 int putty;
726 char *ssh_host = hostandport;
727 const char *port = NULL;
728 get_host_and_port(&ssh_host, &port);
730 if (!port)
731 port = get_port(ssh_host);
733 if (flags & CONNECT_DIAG_URL) {
734 printf("Diag: url=%s\n", url ? url : "NULL");
735 printf("Diag: protocol=%s\n", prot_name(protocol));
736 printf("Diag: userandhost=%s\n", ssh_host ? ssh_host : "NULL");
737 printf("Diag: port=%s\n", port ? port : "NONE");
738 printf("Diag: path=%s\n", path ? path : "NULL");
740 free(hostandport);
741 free(path);
742 return NULL;
743 } else {
744 ssh = getenv("GIT_SSH_COMMAND");
745 if (ssh) {
746 conn->use_shell = 1;
747 putty = 0;
748 } else {
749 ssh = getenv("GIT_SSH");
750 if (!ssh)
751 ssh = "ssh";
752 putty = !!strcasestr(ssh, "plink");
755 argv_array_push(&conn->args, ssh);
756 if (putty && !strcasestr(ssh, "tortoiseplink"))
757 argv_array_push(&conn->args, "-batch");
758 if (port) {
759 /* P is for PuTTY, p is for OpenSSH */
760 argv_array_push(&conn->args, putty ? "-P" : "-p");
761 argv_array_push(&conn->args, port);
763 argv_array_push(&conn->args, ssh_host);
765 } else {
766 /* remove repo-local variables from the environment */
767 conn->env = local_repo_env;
768 conn->use_shell = 1;
770 argv_array_push(&conn->args, cmd.buf);
772 if (start_command(conn))
773 die("unable to fork");
775 fd[0] = conn->out; /* read from child's stdout */
776 fd[1] = conn->in; /* write to child's stdin */
777 strbuf_release(&cmd);
779 free(hostandport);
780 free(path);
781 return conn;
784 int git_connection_is_socket(struct child_process *conn)
786 return conn == &no_fork;
789 int finish_connect(struct child_process *conn)
791 int code;
792 if (!conn || git_connection_is_socket(conn))
793 return 0;
795 code = finish_command(conn);
796 free(conn);
797 return code;