1 /* vi: set sw=4 ts=4: */
3 * wget - retrieve a file using HTTP or FTP
5 * Chip Rosenthal Covad Communications <chip@laserlink.net>
6 * Licensed under GPLv2, see file LICENSE in this source tree.
8 * Copyright (C) 2010 Bradley M. Kuhn <bkuhn@ebb.org>
9 * Kuhn's copyrights are licensed GPLv2-or-later. File as a whole remains GPLv2.
16 //config: wget is a utility for non-interactive download of files from HTTP
17 //config: and FTP servers.
19 //config:config FEATURE_WGET_LONG_OPTIONS
20 //config: bool "Enable long options"
22 //config: depends on WGET && LONG_OPTS
24 //config:config FEATURE_WGET_STATUSBAR
25 //config: bool "Enable progress bar (+2k)"
27 //config: depends on WGET
29 //config:config FEATURE_WGET_AUTHENTICATION
30 //config: bool "Enable HTTP authentication"
32 //config: depends on WGET
34 //config: Support authenticated HTTP transfers.
36 //config:config FEATURE_WGET_TIMEOUT
37 //config: bool "Enable timeout option -T SEC"
39 //config: depends on WGET
41 //config: Supports network read and connect timeouts for wget,
42 //config: so that wget will give up and timeout, through the -T
43 //config: command line option.
45 //config: Currently only connect and network data read timeout are
46 //config: supported (i.e., timeout is not applied to the DNS query). When
47 //config: FEATURE_WGET_LONG_OPTIONS is also enabled, the --timeout option
48 //config: will work in addition to -T.
50 //config:config FEATURE_WGET_HTTPS
51 //config: bool "Support HTTPS using internal TLS code"
53 //config: depends on WGET
56 //config: wget will use internal TLS code to connect to https:// URLs.
58 //config: On NOMMU machines, ssl_helper applet should be available
59 //config: in the $PATH for this to work. Make sure to select that applet.
61 //config:config FEATURE_WGET_OPENSSL
62 //config: bool "Try to connect to HTTPS using openssl"
64 //config: depends on WGET
66 //config: Try to use openssl to handle HTTPS.
68 //config: OpenSSL has a simple SSL client for debug purposes.
69 //config: If you select this option, wget will effectively run:
70 //config: "openssl s_client -quiet -connect hostname:443
71 //config: -servername hostname 2>/dev/null" and pipe its data
72 //config: through it. -servername is not used if hostname is numeric.
73 //config: Note inconvenient API: host resolution is done twice,
74 //config: and there is no guarantee openssl's idea of IPv6 address
75 //config: format is the same as ours.
76 //config: Another problem is that s_client prints debug information
77 //config: to stderr, and it needs to be suppressed. This means
78 //config: all error messages get suppressed too.
79 //config: openssl is also a big binary, often dynamically linked
80 //config: against ~15 libraries.
82 //config: If openssl can't be executed, internal TLS code will be used
83 //config: (if you enabled it); if openssl can be executed but fails later,
84 //config: wget can't detect this, and download will fail.
86 //applet:IF_WGET(APPLET(wget, BB_DIR_USR_BIN, BB_SUID_DROP))
88 //kbuild:lib-$(CONFIG_WGET) += wget.o
90 //usage:#define wget_trivial_usage
91 //usage: IF_FEATURE_WGET_LONG_OPTIONS(
92 //usage: "[-c|--continue] [--spider] [-q|--quiet] [-O|--output-document FILE]\n"
93 //usage: " [--header 'header: value'] [-Y|--proxy on/off] [-P DIR]\n"
94 /* Since we ignore these opts, we don't show them in --help */
95 /* //usage: " [--no-check-certificate] [--no-cache] [--passive-ftp] [-t TRIES]" */
96 /* //usage: " [-nv] [-nc] [-nH] [-np]" */
97 //usage: " [-U|--user-agent AGENT]" IF_FEATURE_WGET_TIMEOUT(" [-T SEC]") " URL..."
99 //usage: IF_NOT_FEATURE_WGET_LONG_OPTIONS(
100 //usage: "[-cq] [-O FILE] [-Y on/off] [-P DIR] [-U AGENT]"
101 //usage: IF_FEATURE_WGET_TIMEOUT(" [-T SEC]") " URL..."
103 //usage:#define wget_full_usage "\n\n"
104 //usage: "Retrieve files via HTTP or FTP\n"
105 //usage: IF_FEATURE_WGET_LONG_OPTIONS(
106 //usage: "\n --spider Spider mode - only check file existence"
108 //usage: "\n -c Continue retrieval of aborted transfer"
109 //usage: "\n -q Quiet"
110 //usage: "\n -P DIR Save to DIR (default .)"
111 //usage: IF_FEATURE_WGET_TIMEOUT(
112 //usage: "\n -T SEC Network read timeout is SEC seconds"
114 //usage: "\n -O FILE Save to FILE ('-' for stdout)"
115 //usage: "\n -U STR Use STR for User-Agent header"
116 //usage: "\n -Y on/off Use proxy"
121 # define log_io(...) bb_error_msg(__VA_ARGS__)
122 # define SENDFMT(fp, fmt, ...) \
124 log_io("> " fmt, ##__VA_ARGS__); \
125 fprintf(fp, fmt, ##__VA_ARGS__); \
128 # define log_io(...) ((void)0)
129 # define SENDFMT(fp, fmt, ...) fprintf(fp, fmt, ##__VA_ARGS__)
133 #define SSL_SUPPORTED (ENABLE_FEATURE_WGET_OPENSSL || ENABLE_FEATURE_WGET_HTTPS)
139 const char *protocol;
143 static const char P_FTP[] ALIGN1 = "ftp";
144 static const char P_HTTP[] ALIGN1 = "http";
146 static const char P_HTTPS[] ALIGN1 = "https";
149 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
150 /* User-specified headers prevent using our corresponding built-in headers. */
153 HDR_USER_AGENT = (1<<1),
155 HDR_AUTH = (1<<3) * ENABLE_FEATURE_WGET_AUTHENTICATION,
156 HDR_PROXY_AUTH = (1<<4) * ENABLE_FEATURE_WGET_AUTHENTICATION,
158 static const char wget_user_headers[] ALIGN1 =
162 # if ENABLE_FEATURE_WGET_AUTHENTICATION
164 "Proxy-Authorization:\0"
167 # define USR_HEADER_HOST (G.user_headers & HDR_HOST)
168 # define USR_HEADER_USER_AGENT (G.user_headers & HDR_USER_AGENT)
169 # define USR_HEADER_RANGE (G.user_headers & HDR_RANGE)
170 # define USR_HEADER_AUTH (G.user_headers & HDR_AUTH)
171 # define USR_HEADER_PROXY_AUTH (G.user_headers & HDR_PROXY_AUTH)
172 #else /* No long options, no user-headers :( */
173 # define USR_HEADER_HOST 0
174 # define USR_HEADER_USER_AGENT 0
175 # define USR_HEADER_RANGE 0
176 # define USR_HEADER_AUTH 0
177 # define USR_HEADER_PROXY_AUTH 0
182 off_t content_len; /* Content-length of the file */
183 off_t beg_range; /* Range at which continue begins */
184 #if ENABLE_FEATURE_WGET_STATUSBAR
185 off_t transferred; /* Number of bytes transferred so far */
186 const char *curfile; /* Name of current file being transferred */
190 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
193 unsigned char user_headers; /* Headers mentioned by the user */
195 char *fname_out; /* where to direct output (-O) */
196 const char *proxy_flag; /* Use proxies if env vars are set */
197 const char *user_agent; /* "User-Agent" header field */
198 #if ENABLE_FEATURE_WGET_TIMEOUT
199 unsigned timeout_seconds;
200 bool die_if_timed_out;
204 smallint chunked; /* chunked transfer encoding */
205 smallint got_clen; /* got content-length: from server */
206 /* Local downloads do benefit from big buffer.
207 * With 512 byte buffer, it was measured to be
208 * an order of magnitude slower than with big one.
210 uint64_t just_to_align_next_member;
211 char wget_buf[CONFIG_FEATURE_COPYBUF_KB*1024];
213 #define G (*ptr_to_globals)
214 #define INIT_G() do { \
215 SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \
217 #define FINI_G() do { \
218 FREE_PTR_TO_GLOBALS(); \
222 /* Must match option string! */
224 WGET_OPT_CONTINUE = (1 << 0),
225 WGET_OPT_QUIET = (1 << 1),
226 WGET_OPT_OUTNAME = (1 << 2),
227 WGET_OPT_PREFIX = (1 << 3),
228 WGET_OPT_PROXY = (1 << 4),
229 WGET_OPT_USER_AGENT = (1 << 5),
230 WGET_OPT_NETWORK_READ_TIMEOUT = (1 << 6),
231 WGET_OPT_RETRIES = (1 << 7),
232 WGET_OPT_nsomething = (1 << 8),
233 WGET_OPT_HEADER = (1 << 9) * ENABLE_FEATURE_WGET_LONG_OPTIONS,
234 WGET_OPT_POST_DATA = (1 << 10) * ENABLE_FEATURE_WGET_LONG_OPTIONS,
235 WGET_OPT_SPIDER = (1 << 11) * ENABLE_FEATURE_WGET_LONG_OPTIONS,
243 #if ENABLE_FEATURE_WGET_STATUSBAR
244 static void progress_meter(int flag)
246 if (option_mask32 & WGET_OPT_QUIET)
249 if (flag == PROGRESS_START)
250 bb_progress_init(&G.pmt, G.curfile);
252 bb_progress_update(&G.pmt,
255 (G.chunked || !G.got_clen) ? 0 : G.beg_range + G.transferred + G.content_len
258 if (flag == PROGRESS_END) {
259 bb_progress_free(&G.pmt);
260 bb_putchar_stderr('\n');
265 static ALWAYS_INLINE void progress_meter(int flag UNUSED_PARAM) { }
269 /* IPv6 knows scoped address types i.e. link and site local addresses. Link
270 * local addresses can have a scope identifier to specify the
271 * interface/link an address is valid on (e.g. fe80::1%eth0). This scope
272 * identifier is only valid on a single node.
274 * RFC 4007 says that the scope identifier MUST NOT be sent across the wire,
275 * unless all nodes agree on the semantic. Apache e.g. regards zone identifiers
276 * in the Host header as invalid requests, see
277 * https://issues.apache.org/bugzilla/show_bug.cgi?id=35122
279 static void strip_ipv6_scope_id(char *host)
283 /* bbox wget actually handles IPv6 addresses without [], like
284 * wget "http://::1/xxx", but this is not standard.
285 * To save code, _here_ we do not support it. */
288 return; /* not IPv6 */
290 scope = strchr(host, '%');
294 /* Remove the IPv6 zone identifier from the host address */
295 cp = strchr(host, ']');
296 if (!cp || (cp[1] != ':' && cp[1] != '\0')) {
297 /* malformed address (not "[xx]:nn" or "[xx]") */
301 /* cp points to "]...", scope points to "%eth0]..." */
302 overlapping_strcpy(scope, cp);
305 #if ENABLE_FEATURE_WGET_AUTHENTICATION
306 /* Base64-encode character string. */
307 static char *base64enc(const char *str)
309 unsigned len = strlen(str);
310 if (len > sizeof(G.wget_buf)/4*3 - 10) /* paranoia */
311 len = sizeof(G.wget_buf)/4*3 - 10;
312 bb_uuencode(G.wget_buf, str, len, bb_uuenc_tbl_base64);
317 static char* sanitize_string(char *s)
319 unsigned char *p = (void *) s;
326 #if ENABLE_FEATURE_WGET_TIMEOUT
327 static void alarm_handler(int sig UNUSED_PARAM)
329 /* This is theoretically unsafe (uses stdio and malloc in signal handler) */
330 if (G.die_if_timed_out)
331 bb_error_msg_and_die("download timed out");
333 static void set_alarm(void)
335 if (G.timeout_seconds) {
336 alarm(G.timeout_seconds);
337 G.die_if_timed_out = 1;
340 # define clear_alarm() ((void)(G.die_if_timed_out = 0))
342 # define set_alarm() ((void)0)
343 # define clear_alarm() ((void)0)
346 #if ENABLE_FEATURE_WGET_OPENSSL
348 * is_ip_address() attempts to verify whether or not a string
349 * contains an IPv4 or IPv6 address (vs. an FQDN). The result
350 * of inet_pton() can be used to determine this.
352 * TODO add proper error checking when inet_pton() returns -1
353 * (some form of system error has occurred, and errno is set)
355 static int is_ip_address(const char *string)
357 struct sockaddr_in sa;
359 int result = inet_pton(AF_INET, string, &(sa.sin_addr));
360 # if ENABLE_FEATURE_IPV6
362 struct sockaddr_in6 sa6;
363 result = inet_pton(AF_INET6, string, &(sa6.sin6_addr));
366 return (result == 1);
370 static FILE *open_socket(len_and_sockaddr *lsa)
376 fd = xconnect_stream(lsa);
379 /* glibc 2.4 seems to try seeking on it - ??! */
380 /* hopefully it understands what ESPIPE means... */
381 fp = fdopen(fd, "r+");
383 bb_perror_msg_and_die(bb_msg_memory_exhausted);
388 /* Returns '\n' if it was seen, else '\0'. Trims at first '\r' or '\n' */
389 static char fgets_and_trim(FILE *fp)
395 if (fgets(G.wget_buf, sizeof(G.wget_buf) - 1, fp) == NULL)
396 bb_perror_msg_and_die("error getting response");
399 buf_ptr = strchrnul(G.wget_buf, '\n');
402 buf_ptr = strchrnul(G.wget_buf, '\r');
405 log_io("< %s", G.wget_buf);
410 static int ftpcmd(const char *s1, const char *s2, FILE *fp)
416 fprintf(fp, "%s%s\r\n", s1, s2);
418 log_io("> %s%s", s1, s2);
423 } while (!isdigit(G.wget_buf[0]) || G.wget_buf[3] != ' ');
425 G.wget_buf[3] = '\0';
426 result = xatoi_positive(G.wget_buf);
431 static void parse_url(const char *src_url, struct host_info *h)
436 h->allocated = url = xstrdup(src_url);
439 p = strstr(url, "://");
443 if (strcmp(url, P_FTP) == 0) {
444 h->port = bb_lookup_port(P_FTP, "tcp", 21);
447 if (strcmp(url, P_HTTPS) == 0) {
448 h->port = bb_lookup_port(P_HTTPS, "tcp", 443);
449 h->protocol = P_HTTPS;
452 if (strcmp(url, P_HTTP) == 0) {
454 h->port = bb_lookup_port(P_HTTP, "tcp", 80);
455 h->protocol = P_HTTP;
458 bb_error_msg_and_die("not an http or ftp url: %s", sanitize_string(url));
461 // GNU wget is user-friendly and falls back to http://
467 // "Real" wget 'http://busybox.net?var=a/b' sends this request:
468 // 'GET /?var=a/b HTTP/1.0'
469 // and saves 'index.html?var=a%2Fb' (we save 'b')
470 // wget 'http://busybox.net?login=john@doe':
471 // request: 'GET /?login=john@doe HTTP/1.0'
472 // saves: 'index.html?login=john@doe' (we save '?login=john@doe')
473 // wget 'http://busybox.net#test/test':
474 // request: 'GET / HTTP/1.0'
475 // saves: 'index.html' (we save 'test')
477 // We also don't add unique .N suffix if file exists...
478 sp = strchr(h->host, '/');
479 p = strchr(h->host, '?'); if (!sp || (p && sp > p)) sp = p;
480 p = strchr(h->host, '#'); if (!sp || (p && sp > p)) sp = p;
483 } else if (*sp == '/') {
486 } else { // '#' or '?'
487 // http://busybox.net?login=john@doe is a valid URL
488 // memmove converts to:
489 // http:/busybox.nett?login=john@doe...
490 memmove(h->host - 1, h->host, sp - h->host);
496 sp = strrchr(h->host, '@');
498 // URL-decode "user:password" string before base64-encoding:
499 // wget http://test:my%20pass@example.com should send
500 // Authorization: Basic dGVzdDpteSBwYXNz
501 // which decodes to "test:my pass".
502 // Standard wget and curl do this too.
505 h->user = xstrdup(percent_decode_in_place(h->host, /*strict:*/ 0));
508 /* else: h->user remains NULL, or as set by original request
509 * before redirect (if we are here after a redirect).
513 static char *gethdr(FILE *fp)
518 /* retrieve header line */
519 c = fgets_and_trim(fp);
521 /* end of the headers? */
522 if (G.wget_buf[0] == '\0')
525 /* convert the header name to lower case */
526 for (s = G.wget_buf; isalnum(*s) || *s == '-' || *s == '.' || *s == '_'; ++s) {
528 * No-op for 20-3f and 60-7f. "0-9a-z-." are in these ranges.
529 * 40-5f range ("@A-Z[\]^_") maps to 60-7f.
530 * "A-Z" maps to "a-z".
531 * "@[\]" can't occur in header names.
532 * "^_" maps to "~,DEL" (which is wrong).
533 * "^" was never seen yet, "_" was seen from web.archive.org
534 * (x-archive-orig-x_commoncrawl_Signature: HEXSTRING).
539 /* verify we are at the end of the header name */
541 bb_error_msg_and_die("bad header line: %s", sanitize_string(G.wget_buf));
543 /* locate the start of the header value */
545 hdrval = skip_whitespace(s);
548 /* Rats! The buffer isn't big enough to hold the entire header value */
549 while (c = getc(fp), c != EOF && c != '\n')
556 static void reset_beg_range_to_zero(void)
558 bb_error_msg("restart failed");
560 xlseek(G.output_fd, 0, SEEK_SET);
561 /* Done at the end instead: */
562 /* ftruncate(G.output_fd, 0); */
565 static FILE* prepare_ftp_session(FILE **dfpp, struct host_info *target, len_and_sockaddr *lsa)
572 target->user = xstrdup("anonymous:busybox@");
574 sfp = open_socket(lsa);
575 if (ftpcmd(NULL, NULL, sfp) != 220)
576 bb_error_msg_and_die("%s", sanitize_string(G.wget_buf + 4));
579 * Splitting username:password pair,
582 str = strchr(target->user, ':');
585 switch (ftpcmd("USER ", target->user, sfp)) {
589 if (ftpcmd("PASS ", str, sfp) == 230)
591 /* fall through (failed login) */
593 bb_error_msg_and_die("ftp login: %s", sanitize_string(G.wget_buf + 4));
596 ftpcmd("TYPE I", NULL, sfp);
601 if (ftpcmd("SIZE ", target->path, sfp) == 213) {
602 G.content_len = BB_STRTOOFF(G.wget_buf + 4, NULL, 10);
603 if (G.content_len < 0 || errno) {
604 bb_error_msg_and_die("SIZE value is garbage");
610 * Entering passive mode
612 if (ftpcmd("PASV", NULL, sfp) != 227) {
614 bb_error_msg_and_die("bad response to %s: %s", "PASV", sanitize_string(G.wget_buf));
616 // Response is "227 garbageN1,N2,N3,N4,P1,P2[)garbage]
617 // Server's IP is N1.N2.N3.N4 (we ignore it)
618 // Server's port for data connection is P1*256+P2
619 str = strrchr(G.wget_buf, ')');
620 if (str) str[0] = '\0';
621 str = strrchr(G.wget_buf, ',');
622 if (!str) goto pasv_error;
623 port = xatou_range(str+1, 0, 255);
625 str = strrchr(G.wget_buf, ',');
626 if (!str) goto pasv_error;
627 port += xatou_range(str+1, 0, 255) * 256;
628 set_nport(&lsa->u.sa, htons(port));
630 *dfpp = open_socket(lsa);
632 if (G.beg_range != 0) {
633 sprintf(G.wget_buf, "REST %"OFF_FMT"u", G.beg_range);
634 if (ftpcmd(G.wget_buf, NULL, sfp) == 350)
635 G.content_len -= G.beg_range;
637 reset_beg_range_to_zero();
640 if (ftpcmd("RETR ", target->path, sfp) > 150)
641 bb_error_msg_and_die("bad response to %s: %s", "RETR", sanitize_string(G.wget_buf));
646 #if ENABLE_FEATURE_WGET_OPENSSL
647 static int spawn_https_helper_openssl(const char *host, unsigned port)
649 char *allocated = NULL;
653 IF_FEATURE_WGET_HTTPS(volatile int child_failed = 0;)
655 if (socketpair(AF_UNIX, SOCK_STREAM, 0, sp) != 0)
656 /* Kernel can have AF_UNIX support disabled */
657 bb_perror_msg_and_die("socketpair");
659 if (!strchr(host, ':'))
660 host = allocated = xasprintf("%s:%u", host, port);
661 servername = xstrdup(host);
662 strrchr(servername, ':')[0] = '\0';
674 * openssl s_client -quiet -connect www.kernel.org:443 2>/dev/null
675 * It prints some debug stuff on stderr, don't know how to suppress it.
676 * Work around by dev-nulling stderr. We lose all error messages :(
679 xopen("/dev/null", O_RDWR);
680 memset(&argv, 0, sizeof(argv));
681 argv[0] = (char*)"openssl";
682 argv[1] = (char*)"s_client";
683 argv[2] = (char*)"-quiet";
684 argv[3] = (char*)"-connect";
685 argv[4] = (char*)host;
687 * Per RFC 6066 Section 3, the only permitted values in the
688 * TLS server_name (SNI) field are FQDNs (DNS hostnames).
689 * IPv4 and IPv6 addresses, port numbers are not allowed.
691 if (!is_ip_address(servername)) {
692 argv[5] = (char*)"-servername";
693 argv[6] = (char*)servername;
696 BB_EXECVP(argv[0], argv);
698 # if ENABLE_FEATURE_WGET_HTTPS
702 bb_perror_msg_and_die("can't execute '%s'", argv[0]);
711 # if ENABLE_FEATURE_WGET_HTTPS
721 #if ENABLE_FEATURE_WGET_HTTPS
722 static void spawn_ssl_client(const char *host, int network_fd)
726 char *servername, *p;
728 servername = xstrdup(host);
729 p = strrchr(servername, ':');
732 if (socketpair(AF_UNIX, SOCK_STREAM, 0, sp) != 0)
733 /* Kernel can have AF_UNIX support disabled */
734 bb_perror_msg_and_die("socketpair");
737 pid = BB_MMU ? xfork() : xvfork();
744 tls_state_t *tls = new_tls_state();
745 tls->ifd = tls->ofd = network_fd;
746 tls_handshake(tls, servername);
747 tls_run_copy_loop(tls);
751 xmove_fd(network_fd, 3);
752 argv[0] = (char*)"ssl_client";
753 argv[1] = (char*)"-s3";
754 //TODO: if (!is_ip_address(servername))...
755 argv[2] = (char*)"-n";
756 argv[3] = servername;
758 BB_EXECVP(argv[0], argv);
759 bb_perror_msg_and_die("can't execute '%s'", argv[0]);
767 xmove_fd(sp[0], network_fd);
771 static void NOINLINE retrieve_file_data(FILE *dfp)
773 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
774 # if ENABLE_FEATURE_WGET_TIMEOUT
775 unsigned second_cnt = G.timeout_seconds;
777 struct pollfd polldata;
779 polldata.fd = fileno(dfp);
780 polldata.events = POLLIN | POLLPRI;
782 progress_meter(PROGRESS_START);
787 /* Loops only if chunked */
790 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
791 /* Must use nonblocking I/O, otherwise fread will loop
792 * and *block* until it reads full buffer,
793 * which messes up progress bar and/or timeout logic.
794 * Because of nonblocking I/O, we need to dance
795 * very carefully around EAGAIN. See explanation at
798 ndelay_on(polldata.fd);
804 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
805 /* fread internally uses read loop, which in our case
806 * is usually exited when we get EAGAIN.
807 * In this case, libc sets error marker on the stream.
808 * Need to clear it before next fread to avoid possible
809 * rare false positive ferror below. Rare because usually
810 * fread gets more than zero bytes, and we don't fall
811 * into if (n <= 0) ...
816 rdsz = sizeof(G.wget_buf);
818 if (G.content_len < (off_t)sizeof(G.wget_buf)) {
819 if ((int)G.content_len <= 0)
821 rdsz = (unsigned)G.content_len;
824 n = fread(G.wget_buf, 1, rdsz, dfp);
827 xwrite(G.output_fd, G.wget_buf, n);
828 #if ENABLE_FEATURE_WGET_STATUSBAR
833 if (G.content_len == 0)
836 #if ENABLE_FEATURE_WGET_TIMEOUT
837 second_cnt = G.timeout_seconds;
844 * If error occurs, or EOF is reached, the return value
845 * is a short item count (or zero).
846 * fread does not distinguish between EOF and error.
848 if (errno != EAGAIN) {
850 progress_meter(PROGRESS_END);
851 bb_perror_msg_and_die(bb_msg_read_error);
853 break; /* EOF, not error */
856 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
857 /* It was EAGAIN. There is no data. Wait up to one second
858 * then abort if timed out, or update the bar and try reading again.
860 if (safe_poll(&polldata, 1, 1000) == 0) {
861 # if ENABLE_FEATURE_WGET_TIMEOUT
862 if (second_cnt != 0 && --second_cnt == 0) {
863 progress_meter(PROGRESS_END);
864 bb_error_msg_and_die("download timed out");
867 /* We used to loop back to poll here,
868 * but there is no great harm in letting fread
869 * to try reading anyway.
874 /* Need to do it _every_ second for "stalled" indicator
875 * to be shown properly.
877 progress_meter(PROGRESS_BUMP);
878 } /* while (reading data) */
880 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
882 ndelay_off(polldata.fd); /* else fgets can get very unhappy */
887 fgets_and_trim(dfp); /* Eat empty line */
890 G.content_len = STRTOOFF(G.wget_buf, NULL, 16);
891 /* FIXME: error check? */
892 if (G.content_len == 0)
893 break; /* all done! */
896 * Note that fgets may result in some data being buffered in dfp.
897 * We loop back to fread, which will retrieve this data.
898 * Also note that code has to be arranged so that fread
899 * is done _before_ one-second poll wait - poll doesn't know
900 * about stdio buffering and can result in spurious one second waits!
904 /* If -c failed, we restart from the beginning,
905 * but we do not truncate file then, we do it only now, at the end.
906 * This lets user to ^C if his 99% complete 10 GB file download
907 * failed to restart *without* losing the almost complete file.
910 off_t pos = lseek(G.output_fd, 0, SEEK_CUR);
911 if (pos != (off_t)-1)
912 ftruncate(G.output_fd, pos);
915 /* Draw full bar and free its resources */
916 G.chunked = 0; /* makes it show 100% even for chunked download */
917 G.got_clen = 1; /* makes it show 100% even for download of (formerly) unknown size */
918 progress_meter(PROGRESS_END);
921 static void download_one_url(const char *url)
923 bool use_proxy; /* Use proxies if env vars are set */
925 len_and_sockaddr *lsa;
926 FILE *sfp; /* socket to web/ftp server */
927 FILE *dfp; /* socket to ftp server (data) */
929 char *fname_out_alloc;
930 char *redirected_path = NULL;
931 struct host_info server;
932 struct host_info target;
934 server.allocated = NULL;
935 target.allocated = NULL;
939 parse_url(url, &target);
941 /* Use the proxy if necessary */
942 use_proxy = (strcmp(G.proxy_flag, "off") != 0);
944 proxy = getenv(target.protocol == P_FTP ? "ftp_proxy" : "http_proxy");
945 //FIXME: what if protocol is https? Ok to use http_proxy?
946 use_proxy = (proxy && proxy[0]);
948 parse_url(proxy, &server);
951 server.port = target.port;
952 if (ENABLE_FEATURE_IPV6) {
953 //free(server.allocated); - can't be non-NULL
954 server.host = server.allocated = xstrdup(target.host);
956 server.host = target.host;
960 if (ENABLE_FEATURE_IPV6)
961 strip_ipv6_scope_id(target.host);
963 /* If there was no -O FILE, guess output filename */
964 fname_out_alloc = NULL;
965 if (!(option_mask32 & WGET_OPT_OUTNAME)) {
966 G.fname_out = bb_get_last_path_component_nostrip(target.path);
967 /* handle "wget http://kernel.org//" */
968 if (G.fname_out[0] == '/' || !G.fname_out[0])
969 G.fname_out = (char*)"index.html";
970 /* -P DIR is considered only if there was no -O FILE */
972 G.fname_out = fname_out_alloc = concat_path_file(G.dir_prefix, G.fname_out);
974 /* redirects may free target.path later, need to make a copy */
975 G.fname_out = fname_out_alloc = xstrdup(G.fname_out);
978 #if ENABLE_FEATURE_WGET_STATUSBAR
979 G.curfile = bb_get_last_path_component_nostrip(G.fname_out);
982 /* Determine where to start transfer */
984 if (option_mask32 & WGET_OPT_CONTINUE) {
985 G.output_fd = open(G.fname_out, O_WRONLY);
986 if (G.output_fd >= 0) {
987 G.beg_range = xlseek(G.output_fd, 0, SEEK_END);
989 /* File doesn't exist. We do not create file here yet.
990 * We are not sure it exists on remote side */
995 lsa = xhost2sockaddr(server.host, server.port);
996 if (!(option_mask32 & WGET_OPT_QUIET)) {
997 char *s = xmalloc_sockaddr2dotted(&lsa->u.sa);
998 fprintf(stderr, "Connecting to %s (%s)\n", server.host, s);
1002 /*G.content_len = 0; - redundant, got_clen = 0 is enough */
1005 if (use_proxy || target.protocol != P_FTP) {
1012 /* Open socket to http(s) server */
1013 #if ENABLE_FEATURE_WGET_OPENSSL
1014 /* openssl (and maybe internal TLS) support is configured */
1015 if (target.protocol == P_HTTPS) {
1016 /* openssl-based helper
1017 * Inconvenient API since we can't give it an open fd
1019 int fd = spawn_https_helper_openssl(server.host, server.port);
1020 # if ENABLE_FEATURE_WGET_HTTPS
1021 if (fd < 0) { /* no openssl? try internal */
1022 sfp = open_socket(lsa);
1023 spawn_ssl_client(server.host, fileno(sfp));
1027 /* We don't check for exec("openssl") failure in this case */
1029 sfp = fdopen(fd, "r+");
1031 bb_perror_msg_and_die(bb_msg_memory_exhausted);
1034 sfp = open_socket(lsa);
1036 #elif ENABLE_FEATURE_WGET_HTTPS
1037 /* Only internal TLS support is configured */
1038 sfp = open_socket(lsa);
1039 if (target.protocol == P_HTTPS)
1040 spawn_ssl_client(server.host, fileno(sfp));
1042 /* ssl (https) support is not configured */
1043 sfp = open_socket(lsa);
1045 /* Send HTTP request */
1047 SENDFMT(sfp, "GET %s://%s/%s HTTP/1.1\r\n",
1048 target.protocol, target.host,
1051 SENDFMT(sfp, "%s /%s HTTP/1.1\r\n",
1052 (option_mask32 & WGET_OPT_POST_DATA) ? "POST" : "GET",
1055 if (!USR_HEADER_HOST)
1056 SENDFMT(sfp, "Host: %s\r\n", target.host);
1057 if (!USR_HEADER_USER_AGENT)
1058 SENDFMT(sfp, "User-Agent: %s\r\n", G.user_agent);
1060 /* Ask server to close the connection as soon as we are done
1061 * (IOW: we do not intend to send more requests)
1063 SENDFMT(sfp, "Connection: close\r\n");
1065 #if ENABLE_FEATURE_WGET_AUTHENTICATION
1066 if (target.user && !USR_HEADER_AUTH) {
1067 SENDFMT(sfp, "Proxy-Authorization: Basic %s\r\n"+6,
1068 base64enc(target.user));
1070 if (use_proxy && server.user && !USR_HEADER_PROXY_AUTH) {
1071 SENDFMT(sfp, "Proxy-Authorization: Basic %s\r\n",
1072 base64enc(server.user));
1076 if (G.beg_range != 0 && !USR_HEADER_RANGE)
1077 SENDFMT(sfp, "Range: bytes=%"OFF_FMT"u-\r\n", G.beg_range);
1079 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1080 if (G.extra_headers) {
1081 log_io(G.extra_headers);
1082 fputs(G.extra_headers, sfp);
1085 if (option_mask32 & WGET_OPT_POST_DATA) {
1087 "Content-Type: application/x-www-form-urlencoded\r\n"
1088 "Content-Length: %u\r\n"
1091 (int) strlen(G.post_data), G.post_data
1096 SENDFMT(sfp, "\r\n");
1101 /* Tried doing this unconditionally.
1102 * Cloudflare and nginx/1.11.5 are shocked to see SHUT_WR on non-HTTPS.
1105 if (target.protocol == P_HTTPS) {
1106 /* If we use SSL helper, keeping our end of the socket open for writing
1107 * makes our end (i.e. the same fd!) readable (EAGAIN instead of EOF)
1108 * even after child closes its copy of the fd.
1111 shutdown(fileno(sfp), SHUT_WR);
1116 * Retrieve HTTP response line and check for "200" status code.
1119 fgets_and_trim(sfp);
1122 str = skip_non_whitespace(str);
1123 str = skip_whitespace(str);
1124 // FIXME: no error check
1125 // xatou wouldn't work: "200 OK"
1130 while (gethdr(sfp) != NULL)
1131 /* eat all remaining headers */;
1134 /* Success responses */
1137 case 201: /* 201 Created */
1138 /* "The request has been fulfilled and resulted in a new resource being created" */
1139 /* Standard wget is reported to treat this as success */
1141 case 202: /* 202 Accepted */
1142 /* "The request has been accepted for processing, but the processing has not been completed" */
1143 /* Treat as success: fall through */
1144 case 203: /* 203 Non-Authoritative Information */
1145 /* "Use of this response code is not required and is only appropriate when the response would otherwise be 200 (OK)" */
1147 case 204: /* 204 No Content */
1149 Response 204 doesn't say "null file", it says "metadata
1150 has changed but data didn't":
1152 "10.2.5 204 No Content
1153 The server has fulfilled the request but does not need to return
1154 an entity-body, and might want to return updated metainformation.
1155 The response MAY include new or updated metainformation in the form
1156 of entity-headers, which if present SHOULD be associated with
1157 the requested variant.
1159 If the client is a user agent, it SHOULD NOT change its document
1160 view from that which caused the request to be sent. This response
1161 is primarily intended to allow input for actions to take place
1162 without causing a change to the user agent's active document view,
1163 although any new or updated metainformation SHOULD be applied
1164 to the document currently in the user agent's active view.
1166 The 204 response MUST NOT include a message-body, and thus
1167 is always terminated by the first empty line after the header fields."
1169 However, in real world it was observed that some web servers
1170 (e.g. Boa/0.94.14rc21) simply use code 204 when file size is zero.
1172 if (G.beg_range != 0) {
1173 /* "Range:..." was not honored by the server.
1174 * Restart download from the beginning.
1176 reset_beg_range_to_zero();
1179 /* 205 Reset Content ?? what to do on this ?? */
1181 case 300: /* redirection */
1187 case 206: /* Partial Content */
1188 if (G.beg_range != 0)
1189 /* "Range:..." worked. Good. */
1191 /* Partial Content even though we did not ask for it??? */
1194 bb_error_msg_and_die("server returned error: %s", sanitize_string(G.wget_buf));
1198 * Retrieve HTTP headers.
1200 while ((str = gethdr(sfp)) != NULL) {
1201 static const char keywords[] ALIGN1 =
1202 "content-length\0""transfer-encoding\0""location\0";
1204 KEY_content_length = 1, KEY_transfer_encoding, KEY_location
1208 /* gethdr converted "FOO:" string to lowercase */
1210 /* strip trailing whitespace */
1211 char *s = strchrnul(str, '\0') - 1;
1212 while (s >= str && (*s == ' ' || *s == '\t')) {
1216 key = index_in_strings(keywords, G.wget_buf) + 1;
1217 if (key == KEY_content_length) {
1218 G.content_len = BB_STRTOOFF(str, NULL, 10);
1219 if (G.content_len < 0 || errno) {
1220 bb_error_msg_and_die("content-length %s is garbage", sanitize_string(str));
1225 if (key == KEY_transfer_encoding) {
1226 if (strcmp(str_tolower(str), "chunked") != 0)
1227 bb_error_msg_and_die("transfer encoding '%s' is not supported", sanitize_string(str));
1230 if (key == KEY_location && status >= 300) {
1231 if (--redir_limit == 0)
1232 bb_error_msg_and_die("too many redirections");
1234 if (str[0] == '/') {
1235 free(redirected_path);
1236 target.path = redirected_path = xstrdup(str+1);
1237 /* lsa stays the same: it's on the same server */
1239 parse_url(str, &target);
1241 /* server.user remains untouched */
1242 free(server.allocated);
1243 server.allocated = NULL;
1244 server.host = target.host;
1245 /* strip_ipv6_scope_id(target.host); - no! */
1246 /* we assume remote never gives us IPv6 addr with scope id */
1247 server.port = target.port;
1250 } /* else: lsa stays the same: we use proxy */
1252 goto establish_session;
1255 // if (status >= 300)
1256 // bb_error_msg_and_die("bad redirection (no Location: header from server)");
1258 /* For HTTP, data is pumped over the same connection */
1264 sfp = prepare_ftp_session(&dfp, &target, lsa);
1269 if (!(option_mask32 & WGET_OPT_SPIDER)) {
1270 if (G.output_fd < 0)
1271 G.output_fd = xopen(G.fname_out, G.o_flags);
1272 retrieve_file_data(dfp);
1273 if (!(option_mask32 & WGET_OPT_OUTNAME)) {
1274 xclose(G.output_fd);
1280 /* It's ftp. Close data connection properly */
1282 if (ftpcmd(NULL, NULL, sfp) != 226)
1283 bb_error_msg_and_die("ftp error: %s", sanitize_string(G.wget_buf + 4));
1284 /* ftpcmd("QUIT", NULL, sfp); - why bother? */
1288 free(server.allocated);
1289 free(target.allocated);
1292 free(fname_out_alloc);
1293 free(redirected_path);
1296 int wget_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
1297 int wget_main(int argc UNUSED_PARAM, char **argv)
1299 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1300 static const char wget_longopts[] ALIGN1 =
1301 /* name, has_arg, val */
1302 "continue\0" No_argument "c"
1303 "quiet\0" No_argument "q"
1304 "output-document\0" Required_argument "O"
1305 "directory-prefix\0" Required_argument "P"
1306 "proxy\0" Required_argument "Y"
1307 "user-agent\0" Required_argument "U"
1308 IF_FEATURE_WGET_TIMEOUT(
1309 "timeout\0" Required_argument "T")
1311 IF_DESKTOP( "tries\0" Required_argument "t")
1312 "header\0" Required_argument "\xff"
1313 "post-data\0" Required_argument "\xfe"
1314 "spider\0" No_argument "\xfd"
1315 /* Ignored (we always use PASV): */
1316 IF_DESKTOP( "passive-ftp\0" No_argument "\xf0")
1317 /* Ignored (we don't do ssl) */
1318 IF_DESKTOP( "no-check-certificate\0" No_argument "\xf0")
1319 /* Ignored (we don't support caching) */
1320 IF_DESKTOP( "no-cache\0" No_argument "\xf0")
1321 IF_DESKTOP( "no-verbose\0" No_argument "\xf0")
1322 IF_DESKTOP( "no-clobber\0" No_argument "\xf0")
1323 IF_DESKTOP( "no-host-directories\0" No_argument "\xf0")
1324 IF_DESKTOP( "no-parent\0" No_argument "\xf0")
1328 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1329 llist_t *headers_llist = NULL;
1334 #if ENABLE_FEATURE_WGET_TIMEOUT
1335 G.timeout_seconds = 900;
1336 signal(SIGALRM, alarm_handler);
1338 G.proxy_flag = "on"; /* use proxies if env vars are set */
1339 G.user_agent = "Wget"; /* "User-Agent" header field */
1341 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1342 applet_long_options = wget_longopts;
1344 opt_complementary = "-1" /* at least one URL */
1345 IF_FEATURE_WGET_LONG_OPTIONS(":\xff::"); /* --header is a list */
1346 getopt32(argv, "cqO:P:Y:U:T:+"
1349 /* wget has exactly four -n<letter> opts, all of which we can ignore:
1350 * -nv --no-verbose: be moderately quiet (-q is full quiet)
1351 * -nc --no-clobber: abort if exists, neither download to FILE.n nor overwrite FILE
1352 * -nH --no-host-directories: wget -r http://host/ won't create host/
1354 * "n::" above says that we accept -n[ARG].
1355 * Specifying "n:" would be a bug: "-n ARG" would eat ARG!
1357 , &G.fname_out, &G.dir_prefix,
1358 &G.proxy_flag, &G.user_agent,
1359 IF_FEATURE_WGET_TIMEOUT(&G.timeout_seconds) IF_NOT_FEATURE_WGET_TIMEOUT(NULL),
1360 NULL, /* -t RETRIES */
1362 IF_FEATURE_WGET_LONG_OPTIONS(, &headers_llist)
1363 IF_FEATURE_WGET_LONG_OPTIONS(, &G.post_data)
1365 #if 0 /* option bits debug */
1366 if (option_mask32 & WGET_OPT_RETRIES) bb_error_msg("-t NUM");
1367 if (option_mask32 & WGET_OPT_nsomething) bb_error_msg("-nsomething");
1368 if (option_mask32 & WGET_OPT_HEADER) bb_error_msg("--header");
1369 if (option_mask32 & WGET_OPT_POST_DATA) bb_error_msg("--post-data");
1370 if (option_mask32 & WGET_OPT_SPIDER) bb_error_msg("--spider");
1375 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1376 if (headers_llist) {
1379 llist_t *ll = headers_llist;
1381 size += strlen(ll->data) + 2;
1384 G.extra_headers = hdr = xmalloc(size + 1);
1385 while (headers_llist) {
1389 size = sprintf(hdr, "%s\r\n",
1390 (char*)llist_pop(&headers_llist));
1391 /* a bit like index_in_substrings but don't match full key */
1393 words = wget_user_headers;
1395 if (strstr(hdr, words) == hdr) {
1396 G.user_headers |= bit;
1400 words += strlen(words) + 1;
1408 G.o_flags = O_WRONLY | O_CREAT | O_TRUNC | O_EXCL;
1409 if (G.fname_out) { /* -O FILE ? */
1410 if (LONE_DASH(G.fname_out)) { /* -O - ? */
1412 option_mask32 &= ~WGET_OPT_CONTINUE;
1414 /* compat with wget: -O FILE can overwrite */
1415 G.o_flags = O_WRONLY | O_CREAT | O_TRUNC;
1419 download_one_url(*argv++);
1421 if (G.output_fd >= 0)
1422 xclose(G.output_fd);
1424 #if ENABLE_FEATURE_CLEAN_UP && ENABLE_FEATURE_WGET_LONG_OPTIONS
1425 free(G.extra_headers);
1429 return EXIT_SUCCESS;