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_STATUSBAR
20 //config: bool "Enable a nifty process meter (+2k)"
22 //config: depends on WGET
24 //config: Enable the transfer progress bar for wget transfers.
26 //config:config FEATURE_WGET_AUTHENTICATION
27 //config: bool "Enable HTTP authentication"
29 //config: depends on WGET
31 //config: Support authenticated HTTP transfers.
33 //config:config FEATURE_WGET_LONG_OPTIONS
34 //config: bool "Enable long options"
36 //config: depends on WGET && LONG_OPTS
38 //config: Support long options for the wget applet.
40 //config:config FEATURE_WGET_TIMEOUT
41 //config: bool "Enable timeout option -T SEC"
43 //config: depends on WGET
45 //config: Supports network read and connect timeouts for wget,
46 //config: so that wget will give up and timeout, through the -T
47 //config: command line option.
49 //config: Currently only connect and network data read timeout are
50 //config: supported (i.e., timeout is not applied to the DNS query). When
51 //config: FEATURE_WGET_LONG_OPTIONS is also enabled, the --timeout option
52 //config: will work in addition to -T.
54 //config:config FEATURE_WGET_OPENSSL
55 //config: bool "Try to connect to HTTPS using openssl"
57 //config: depends on WGET
59 //config: Choose how wget establishes SSL connection for https:// URLs.
61 //config: Busybox itself contains no SSL code. wget will spawn
62 //config: a helper program to talk over HTTPS.
64 //config: OpenSSL has a simple SSL client for debug purposes.
65 //config: If you select "openssl" helper, wget will effectively run:
66 //config: "openssl s_client -quiet -connect hostname:443
67 //config: -servername hostname 2>/dev/null" and pipe its data
68 //config: through it. -servername is not used if hostname is numeric.
69 //config: Note inconvenient API: host resolution is done twice,
70 //config: and there is no guarantee openssl's idea of IPv6 address
71 //config: format is the same as ours.
72 //config: Another problem is that s_client prints debug information
73 //config: to stderr, and it needs to be suppressed. This means
74 //config: all error messages get suppressed too.
75 //config: openssl is also a big binary, often dynamically linked
76 //config: against ~15 libraries.
78 //config:config FEATURE_WGET_SSL_HELPER
79 //config: bool "Try to connect to HTTPS using ssl_helper"
81 //config: depends on WGET
83 //config: Choose how wget establishes SSL connection for https:// URLs.
85 //config: Busybox itself contains no SSL code. wget will spawn
86 //config: a helper program to talk over HTTPS.
88 //config: ssl_helper is a tool which can be built statically
89 //config: from busybox sources against a small embedded SSL library.
90 //config: Please see networking/ssl_helper/README.
91 //config: It does not require double host resolution and emits
92 //config: error messages to stderr.
94 //config: Precompiled static binary may be available at
95 //config: http://busybox.net/downloads/binaries/
97 //applet:IF_WGET(APPLET(wget, BB_DIR_USR_BIN, BB_SUID_DROP))
99 //kbuild:lib-$(CONFIG_WGET) += wget.o
101 //usage:#define wget_trivial_usage
102 //usage: IF_FEATURE_WGET_LONG_OPTIONS(
103 //usage: "[-c|--continue] [--spider] [-q|--quiet] [-O|--output-document FILE]\n"
104 //usage: " [--header 'header: value'] [-Y|--proxy on/off] [-P DIR]\n"
105 /* Since we ignore these opts, we don't show them in --help */
106 /* //usage: " [--no-check-certificate] [--no-cache] [--passive-ftp] [-t TRIES]" */
107 /* //usage: " [-nv] [-nc] [-nH] [-np]" */
108 //usage: " [-U|--user-agent AGENT]" IF_FEATURE_WGET_TIMEOUT(" [-T SEC]") " URL..."
110 //usage: IF_NOT_FEATURE_WGET_LONG_OPTIONS(
111 //usage: "[-cq] [-O FILE] [-Y on/off] [-P DIR] [-U AGENT]"
112 //usage: IF_FEATURE_WGET_TIMEOUT(" [-T SEC]") " URL..."
114 //usage:#define wget_full_usage "\n\n"
115 //usage: "Retrieve files via HTTP or FTP\n"
116 //usage: IF_FEATURE_WGET_LONG_OPTIONS(
117 //usage: "\n --spider Spider mode - only check file existence"
119 //usage: "\n -c Continue retrieval of aborted transfer"
120 //usage: "\n -q Quiet"
121 //usage: "\n -P DIR Save to DIR (default .)"
122 //usage: IF_FEATURE_WGET_TIMEOUT(
123 //usage: "\n -T SEC Network read timeout is SEC seconds"
125 //usage: "\n -O FILE Save to FILE ('-' for stdout)"
126 //usage: "\n -U STR Use STR for User-Agent header"
127 //usage: "\n -Y on/off Use proxy"
132 # define log_io(...) bb_error_msg(__VA_ARGS__)
133 # define SENDFMT(fp, fmt, ...) \
135 log_io("> " fmt, ##__VA_ARGS__); \
136 fprintf(fp, fmt, ##__VA_ARGS__); \
139 # define log_io(...) ((void)0)
140 # define SENDFMT(fp, fmt, ...) fprintf(fp, fmt, ##__VA_ARGS__)
144 #define SSL_SUPPORTED (ENABLE_FEATURE_WGET_OPENSSL || ENABLE_FEATURE_WGET_SSL_HELPER)
150 const char *protocol;
154 static const char P_FTP[] ALIGN1 = "ftp";
155 static const char P_HTTP[] ALIGN1 = "http";
157 static const char P_HTTPS[] ALIGN1 = "https";
160 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
161 /* User-specified headers prevent using our corresponding built-in headers. */
164 HDR_USER_AGENT = (1<<1),
166 HDR_AUTH = (1<<3) * ENABLE_FEATURE_WGET_AUTHENTICATION,
167 HDR_PROXY_AUTH = (1<<4) * ENABLE_FEATURE_WGET_AUTHENTICATION,
169 static const char wget_user_headers[] ALIGN1 =
173 # if ENABLE_FEATURE_WGET_AUTHENTICATION
175 "Proxy-Authorization:\0"
178 # define USR_HEADER_HOST (G.user_headers & HDR_HOST)
179 # define USR_HEADER_USER_AGENT (G.user_headers & HDR_USER_AGENT)
180 # define USR_HEADER_RANGE (G.user_headers & HDR_RANGE)
181 # define USR_HEADER_AUTH (G.user_headers & HDR_AUTH)
182 # define USR_HEADER_PROXY_AUTH (G.user_headers & HDR_PROXY_AUTH)
183 #else /* No long options, no user-headers :( */
184 # define USR_HEADER_HOST 0
185 # define USR_HEADER_USER_AGENT 0
186 # define USR_HEADER_RANGE 0
187 # define USR_HEADER_AUTH 0
188 # define USR_HEADER_PROXY_AUTH 0
193 off_t content_len; /* Content-length of the file */
194 off_t beg_range; /* Range at which continue begins */
195 #if ENABLE_FEATURE_WGET_STATUSBAR
196 off_t transferred; /* Number of bytes transferred so far */
197 const char *curfile; /* Name of current file being transferred */
201 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
204 unsigned char user_headers; /* Headers mentioned by the user */
206 char *fname_out; /* where to direct output (-O) */
207 const char *proxy_flag; /* Use proxies if env vars are set */
208 const char *user_agent; /* "User-Agent" header field */
209 #if ENABLE_FEATURE_WGET_TIMEOUT
210 unsigned timeout_seconds;
211 bool die_if_timed_out;
215 smallint chunked; /* chunked transfer encoding */
216 smallint got_clen; /* got content-length: from server */
217 /* Local downloads do benefit from big buffer.
218 * With 512 byte buffer, it was measured to be
219 * an order of magnitude slower than with big one.
221 uint64_t just_to_align_next_member;
222 char wget_buf[CONFIG_FEATURE_COPYBUF_KB*1024];
224 #define G (*ptr_to_globals)
225 #define INIT_G() do { \
226 SET_PTR_TO_GLOBALS(xzalloc(sizeof(G))); \
228 #define FINI_G() do { \
229 FREE_PTR_TO_GLOBALS(); \
233 /* Must match option string! */
235 WGET_OPT_CONTINUE = (1 << 0),
236 WGET_OPT_QUIET = (1 << 1),
237 WGET_OPT_OUTNAME = (1 << 2),
238 WGET_OPT_PREFIX = (1 << 3),
239 WGET_OPT_PROXY = (1 << 4),
240 WGET_OPT_USER_AGENT = (1 << 5),
241 WGET_OPT_NETWORK_READ_TIMEOUT = (1 << 6),
242 WGET_OPT_RETRIES = (1 << 7),
243 WGET_OPT_nsomething = (1 << 8),
244 WGET_OPT_HEADER = (1 << 9) * ENABLE_FEATURE_WGET_LONG_OPTIONS,
245 WGET_OPT_POST_DATA = (1 << 10) * ENABLE_FEATURE_WGET_LONG_OPTIONS,
246 WGET_OPT_SPIDER = (1 << 11) * ENABLE_FEATURE_WGET_LONG_OPTIONS,
254 #if ENABLE_FEATURE_WGET_STATUSBAR
255 static void progress_meter(int flag)
257 if (option_mask32 & WGET_OPT_QUIET)
260 if (flag == PROGRESS_START)
261 bb_progress_init(&G.pmt, G.curfile);
263 bb_progress_update(&G.pmt,
266 (G.chunked || !G.got_clen) ? 0 : G.beg_range + G.transferred + G.content_len
269 if (flag == PROGRESS_END) {
270 bb_progress_free(&G.pmt);
271 bb_putchar_stderr('\n');
276 static ALWAYS_INLINE void progress_meter(int flag UNUSED_PARAM) { }
280 /* IPv6 knows scoped address types i.e. link and site local addresses. Link
281 * local addresses can have a scope identifier to specify the
282 * interface/link an address is valid on (e.g. fe80::1%eth0). This scope
283 * identifier is only valid on a single node.
285 * RFC 4007 says that the scope identifier MUST NOT be sent across the wire,
286 * unless all nodes agree on the semantic. Apache e.g. regards zone identifiers
287 * in the Host header as invalid requests, see
288 * https://issues.apache.org/bugzilla/show_bug.cgi?id=35122
290 static void strip_ipv6_scope_id(char *host)
294 /* bbox wget actually handles IPv6 addresses without [], like
295 * wget "http://::1/xxx", but this is not standard.
296 * To save code, _here_ we do not support it. */
299 return; /* not IPv6 */
301 scope = strchr(host, '%');
305 /* Remove the IPv6 zone identifier from the host address */
306 cp = strchr(host, ']');
307 if (!cp || (cp[1] != ':' && cp[1] != '\0')) {
308 /* malformed address (not "[xx]:nn" or "[xx]") */
312 /* cp points to "]...", scope points to "%eth0]..." */
313 overlapping_strcpy(scope, cp);
316 #if ENABLE_FEATURE_WGET_AUTHENTICATION
317 /* Base64-encode character string. */
318 static char *base64enc(const char *str)
320 unsigned len = strlen(str);
321 if (len > sizeof(G.wget_buf)/4*3 - 10) /* paranoia */
322 len = sizeof(G.wget_buf)/4*3 - 10;
323 bb_uuencode(G.wget_buf, str, len, bb_uuenc_tbl_base64);
328 static char* sanitize_string(char *s)
330 unsigned char *p = (void *) s;
337 #if ENABLE_FEATURE_WGET_TIMEOUT
338 static void alarm_handler(int sig UNUSED_PARAM)
340 /* This is theoretically unsafe (uses stdio and malloc in signal handler) */
341 if (G.die_if_timed_out)
342 bb_error_msg_and_die("download timed out");
344 static void set_alarm(void)
346 if (G.timeout_seconds) {
347 alarm(G.timeout_seconds);
348 G.die_if_timed_out = 1;
351 # define clear_alarm() ((void)(G.die_if_timed_out = 0))
353 # define set_alarm() ((void)0)
354 # define clear_alarm() ((void)0)
357 #if ENABLE_FEATURE_WGET_OPENSSL
359 * is_ip_address() attempts to verify whether or not a string
360 * contains an IPv4 or IPv6 address (vs. an FQDN). The result
361 * of inet_pton() can be used to determine this.
363 * TODO add proper error checking when inet_pton() returns -1
364 * (some form of system error has occurred, and errno is set)
366 static int is_ip_address(const char *string)
368 struct sockaddr_in sa;
370 int result = inet_pton(AF_INET, string, &(sa.sin_addr));
371 # if ENABLE_FEATURE_IPV6
373 struct sockaddr_in6 sa6;
374 result = inet_pton(AF_INET6, string, &(sa6.sin6_addr));
377 return (result == 1);
381 static FILE *open_socket(len_and_sockaddr *lsa)
387 fd = xconnect_stream(lsa);
390 /* glibc 2.4 seems to try seeking on it - ??! */
391 /* hopefully it understands what ESPIPE means... */
392 fp = fdopen(fd, "r+");
394 bb_perror_msg_and_die(bb_msg_memory_exhausted);
399 /* Returns '\n' if it was seen, else '\0'. Trims at first '\r' or '\n' */
400 static char fgets_and_trim(FILE *fp)
406 if (fgets(G.wget_buf, sizeof(G.wget_buf) - 1, fp) == NULL)
407 bb_perror_msg_and_die("error getting response");
410 buf_ptr = strchrnul(G.wget_buf, '\n');
413 buf_ptr = strchrnul(G.wget_buf, '\r');
416 log_io("< %s", G.wget_buf);
421 static int ftpcmd(const char *s1, const char *s2, FILE *fp)
427 fprintf(fp, "%s%s\r\n", s1, s2);
429 log_io("> %s%s", s1, s2);
434 } while (!isdigit(G.wget_buf[0]) || G.wget_buf[3] != ' ');
436 G.wget_buf[3] = '\0';
437 result = xatoi_positive(G.wget_buf);
442 static void parse_url(const char *src_url, struct host_info *h)
447 h->allocated = url = xstrdup(src_url);
450 p = strstr(url, "://");
454 if (strcmp(url, P_FTP) == 0) {
455 h->port = bb_lookup_port(P_FTP, "tcp", 21);
458 if (strcmp(url, P_HTTPS) == 0) {
459 h->port = bb_lookup_port(P_HTTPS, "tcp", 443);
460 h->protocol = P_HTTPS;
463 if (strcmp(url, P_HTTP) == 0) {
465 h->port = bb_lookup_port(P_HTTP, "tcp", 80);
466 h->protocol = P_HTTP;
469 bb_error_msg_and_die("not an http or ftp url: %s", sanitize_string(url));
472 // GNU wget is user-friendly and falls back to http://
478 // "Real" wget 'http://busybox.net?var=a/b' sends this request:
479 // 'GET /?var=a/b HTTP 1.0'
480 // and saves 'index.html?var=a%2Fb' (we save 'b')
481 // wget 'http://busybox.net?login=john@doe':
482 // request: 'GET /?login=john@doe HTTP/1.0'
483 // saves: 'index.html?login=john@doe' (we save '?login=john@doe')
484 // wget 'http://busybox.net#test/test':
485 // request: 'GET / HTTP/1.0'
486 // saves: 'index.html' (we save 'test')
488 // We also don't add unique .N suffix if file exists...
489 sp = strchr(h->host, '/');
490 p = strchr(h->host, '?'); if (!sp || (p && sp > p)) sp = p;
491 p = strchr(h->host, '#'); if (!sp || (p && sp > p)) sp = p;
494 } else if (*sp == '/') {
497 } else { // '#' or '?'
498 // http://busybox.net?login=john@doe is a valid URL
499 // memmove converts to:
500 // http:/busybox.nett?login=john@doe...
501 memmove(h->host - 1, h->host, sp - h->host);
507 sp = strrchr(h->host, '@');
509 // URL-decode "user:password" string before base64-encoding:
510 // wget http://test:my%20pass@example.com should send
511 // Authorization: Basic dGVzdDpteSBwYXNz
512 // which decodes to "test:my pass".
513 // Standard wget and curl do this too.
516 h->user = xstrdup(percent_decode_in_place(h->host, /*strict:*/ 0));
519 /* else: h->user remains NULL, or as set by original request
520 * before redirect (if we are here after a redirect).
524 static char *gethdr(FILE *fp)
529 /* retrieve header line */
530 c = fgets_and_trim(fp);
532 /* end of the headers? */
533 if (G.wget_buf[0] == '\0')
536 /* convert the header name to lower case */
537 for (s = G.wget_buf; isalnum(*s) || *s == '-' || *s == '.' || *s == '_'; ++s) {
539 * No-op for 20-3f and 60-7f. "0-9a-z-." are in these ranges.
540 * 40-5f range ("@A-Z[\]^_") maps to 60-7f.
541 * "A-Z" maps to "a-z".
542 * "@[\]" can't occur in header names.
543 * "^_" maps to "~,DEL" (which is wrong).
544 * "^" was never seen yet, "_" was seen from web.archive.org
545 * (x-archive-orig-x_commoncrawl_Signature: HEXSTRING).
550 /* verify we are at the end of the header name */
552 bb_error_msg_and_die("bad header line: %s", sanitize_string(G.wget_buf));
554 /* locate the start of the header value */
556 hdrval = skip_whitespace(s);
559 /* Rats! The buffer isn't big enough to hold the entire header value */
560 while (c = getc(fp), c != EOF && c != '\n')
567 static void reset_beg_range_to_zero(void)
569 bb_error_msg("restart failed");
571 xlseek(G.output_fd, 0, SEEK_SET);
572 /* Done at the end instead: */
573 /* ftruncate(G.output_fd, 0); */
576 static FILE* prepare_ftp_session(FILE **dfpp, struct host_info *target, len_and_sockaddr *lsa)
583 target->user = xstrdup("anonymous:busybox@");
585 sfp = open_socket(lsa);
586 if (ftpcmd(NULL, NULL, sfp) != 220)
587 bb_error_msg_and_die("%s", sanitize_string(G.wget_buf + 4));
590 * Splitting username:password pair,
593 str = strchr(target->user, ':');
596 switch (ftpcmd("USER ", target->user, sfp)) {
600 if (ftpcmd("PASS ", str, sfp) == 230)
602 /* fall through (failed login) */
604 bb_error_msg_and_die("ftp login: %s", sanitize_string(G.wget_buf + 4));
607 ftpcmd("TYPE I", NULL, sfp);
612 if (ftpcmd("SIZE ", target->path, sfp) == 213) {
613 G.content_len = BB_STRTOOFF(G.wget_buf + 4, NULL, 10);
614 if (G.content_len < 0 || errno) {
615 bb_error_msg_and_die("SIZE value is garbage");
621 * Entering passive mode
623 if (ftpcmd("PASV", NULL, sfp) != 227) {
625 bb_error_msg_and_die("bad response to %s: %s", "PASV", sanitize_string(G.wget_buf));
627 // Response is "227 garbageN1,N2,N3,N4,P1,P2[)garbage]
628 // Server's IP is N1.N2.N3.N4 (we ignore it)
629 // Server's port for data connection is P1*256+P2
630 str = strrchr(G.wget_buf, ')');
631 if (str) str[0] = '\0';
632 str = strrchr(G.wget_buf, ',');
633 if (!str) goto pasv_error;
634 port = xatou_range(str+1, 0, 255);
636 str = strrchr(G.wget_buf, ',');
637 if (!str) goto pasv_error;
638 port += xatou_range(str+1, 0, 255) * 256;
639 set_nport(&lsa->u.sa, htons(port));
641 *dfpp = open_socket(lsa);
643 if (G.beg_range != 0) {
644 sprintf(G.wget_buf, "REST %"OFF_FMT"u", G.beg_range);
645 if (ftpcmd(G.wget_buf, NULL, sfp) == 350)
646 G.content_len -= G.beg_range;
648 reset_beg_range_to_zero();
651 if (ftpcmd("RETR ", target->path, sfp) > 150)
652 bb_error_msg_and_die("bad response to %s: %s", "RETR", sanitize_string(G.wget_buf));
657 #if ENABLE_FEATURE_WGET_OPENSSL
658 static int spawn_https_helper_openssl(const char *host, unsigned port)
660 char *allocated = NULL;
664 IF_FEATURE_WGET_SSL_HELPER(volatile int child_failed = 0;)
666 if (socketpair(AF_UNIX, SOCK_STREAM, 0, sp) != 0)
667 /* Kernel can have AF_UNIX support disabled */
668 bb_perror_msg_and_die("socketpair");
670 if (!strchr(host, ':'))
671 host = allocated = xasprintf("%s:%u", host, port);
672 servername = xstrdup(host);
673 strrchr(servername, ':')[0] = '\0';
685 * openssl s_client -quiet -connect www.kernel.org:443 2>/dev/null
686 * It prints some debug stuff on stderr, don't know how to suppress it.
687 * Work around by dev-nulling stderr. We lose all error messages :(
690 xopen("/dev/null", O_RDWR);
691 memset(&argv, 0, sizeof(argv));
692 argv[0] = (char*)"openssl";
693 argv[1] = (char*)"s_client";
694 argv[2] = (char*)"-quiet";
695 argv[3] = (char*)"-connect";
696 argv[4] = (char*)host;
698 * Per RFC 6066 Section 3, the only permitted values in the
699 * TLS server_name (SNI) field are FQDNs (DNS hostnames).
700 * IPv4 and IPv6 addresses, port numbers are not allowed.
702 if (!is_ip_address(servername)) {
703 argv[5] = (char*)"-servername";
704 argv[6] = (char*)servername;
707 BB_EXECVP(argv[0], argv);
709 # if ENABLE_FEATURE_WGET_SSL_HELPER
713 bb_perror_msg_and_die("can't execute '%s'", argv[0]);
722 # if ENABLE_FEATURE_WGET_SSL_HELPER
732 /* See networking/ssl_helper/README how to build one */
733 #if ENABLE_FEATURE_WGET_SSL_HELPER
734 static void spawn_https_helper_small(int network_fd)
739 if (socketpair(AF_UNIX, SOCK_STREAM, 0, sp) != 0)
740 /* Kernel can have AF_UNIX support disabled */
741 bb_perror_msg_and_die("socketpair");
743 pid = BB_MMU ? xfork() : xvfork();
751 xmove_fd(network_fd, 3);
753 * A simple ssl/tls helper
755 argv[0] = (char*)"ssl_helper";
756 argv[1] = (char*)"-d3";
758 BB_EXECVP(argv[0], argv);
759 bb_perror_msg_and_die("can't execute '%s'", argv[0]);
765 xmove_fd(sp[0], network_fd);
769 static void NOINLINE retrieve_file_data(FILE *dfp)
771 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
772 # if ENABLE_FEATURE_WGET_TIMEOUT
773 unsigned second_cnt = G.timeout_seconds;
775 struct pollfd polldata;
777 polldata.fd = fileno(dfp);
778 polldata.events = POLLIN | POLLPRI;
780 progress_meter(PROGRESS_START);
785 /* Loops only if chunked */
788 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
789 /* Must use nonblocking I/O, otherwise fread will loop
790 * and *block* until it reads full buffer,
791 * which messes up progress bar and/or timeout logic.
792 * Because of nonblocking I/O, we need to dance
793 * very carefully around EAGAIN. See explanation at
796 ndelay_on(polldata.fd);
802 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
803 /* fread internally uses read loop, which in our case
804 * is usually exited when we get EAGAIN.
805 * In this case, libc sets error marker on the stream.
806 * Need to clear it before next fread to avoid possible
807 * rare false positive ferror below. Rare because usually
808 * fread gets more than zero bytes, and we don't fall
809 * into if (n <= 0) ...
814 rdsz = sizeof(G.wget_buf);
816 if (G.content_len < (off_t)sizeof(G.wget_buf)) {
817 if ((int)G.content_len <= 0)
819 rdsz = (unsigned)G.content_len;
822 n = fread(G.wget_buf, 1, rdsz, dfp);
825 xwrite(G.output_fd, G.wget_buf, n);
826 #if ENABLE_FEATURE_WGET_STATUSBAR
831 if (G.content_len == 0)
834 #if ENABLE_FEATURE_WGET_TIMEOUT
835 second_cnt = G.timeout_seconds;
842 * If error occurs, or EOF is reached, the return value
843 * is a short item count (or zero).
844 * fread does not distinguish between EOF and error.
846 if (errno != EAGAIN) {
848 progress_meter(PROGRESS_END);
849 bb_perror_msg_and_die(bb_msg_read_error);
851 break; /* EOF, not error */
854 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
855 /* It was EAGAIN. There is no data. Wait up to one second
856 * then abort if timed out, or update the bar and try reading again.
858 if (safe_poll(&polldata, 1, 1000) == 0) {
859 # if ENABLE_FEATURE_WGET_TIMEOUT
860 if (second_cnt != 0 && --second_cnt == 0) {
861 progress_meter(PROGRESS_END);
862 bb_error_msg_and_die("download timed out");
865 /* We used to loop back to poll here,
866 * but there is no great harm in letting fread
867 * to try reading anyway.
872 /* Need to do it _every_ second for "stalled" indicator
873 * to be shown properly.
875 progress_meter(PROGRESS_BUMP);
876 } /* while (reading data) */
878 #if ENABLE_FEATURE_WGET_STATUSBAR || ENABLE_FEATURE_WGET_TIMEOUT
880 ndelay_off(polldata.fd); /* else fgets can get very unhappy */
885 fgets_and_trim(dfp); /* Eat empty line */
888 G.content_len = STRTOOFF(G.wget_buf, NULL, 16);
889 /* FIXME: error check? */
890 if (G.content_len == 0)
891 break; /* all done! */
894 * Note that fgets may result in some data being buffered in dfp.
895 * We loop back to fread, which will retrieve this data.
896 * Also note that code has to be arranged so that fread
897 * is done _before_ one-second poll wait - poll doesn't know
898 * about stdio buffering and can result in spurious one second waits!
902 /* If -c failed, we restart from the beginning,
903 * but we do not truncate file then, we do it only now, at the end.
904 * This lets user to ^C if his 99% complete 10 GB file download
905 * failed to restart *without* losing the almost complete file.
908 off_t pos = lseek(G.output_fd, 0, SEEK_CUR);
909 if (pos != (off_t)-1)
910 ftruncate(G.output_fd, pos);
913 /* Draw full bar and free its resources */
914 G.chunked = 0; /* makes it show 100% even for chunked download */
915 G.got_clen = 1; /* makes it show 100% even for download of (formerly) unknown size */
916 progress_meter(PROGRESS_END);
919 static void download_one_url(const char *url)
921 bool use_proxy; /* Use proxies if env vars are set */
923 len_and_sockaddr *lsa;
924 FILE *sfp; /* socket to web/ftp server */
925 FILE *dfp; /* socket to ftp server (data) */
927 char *fname_out_alloc;
928 char *redirected_path = NULL;
929 struct host_info server;
930 struct host_info target;
932 server.allocated = NULL;
933 target.allocated = NULL;
937 parse_url(url, &target);
939 /* Use the proxy if necessary */
940 use_proxy = (strcmp(G.proxy_flag, "off") != 0);
942 proxy = getenv(target.protocol == P_FTP ? "ftp_proxy" : "http_proxy");
943 //FIXME: what if protocol is https? Ok to use http_proxy?
944 use_proxy = (proxy && proxy[0]);
946 parse_url(proxy, &server);
949 server.port = target.port;
950 if (ENABLE_FEATURE_IPV6) {
951 //free(server.allocated); - can't be non-NULL
952 server.host = server.allocated = xstrdup(target.host);
954 server.host = target.host;
958 if (ENABLE_FEATURE_IPV6)
959 strip_ipv6_scope_id(target.host);
961 /* If there was no -O FILE, guess output filename */
962 fname_out_alloc = NULL;
963 if (!(option_mask32 & WGET_OPT_OUTNAME)) {
964 G.fname_out = bb_get_last_path_component_nostrip(target.path);
965 /* handle "wget http://kernel.org//" */
966 if (G.fname_out[0] == '/' || !G.fname_out[0])
967 G.fname_out = (char*)"index.html";
968 /* -P DIR is considered only if there was no -O FILE */
970 G.fname_out = fname_out_alloc = concat_path_file(G.dir_prefix, G.fname_out);
972 /* redirects may free target.path later, need to make a copy */
973 G.fname_out = fname_out_alloc = xstrdup(G.fname_out);
976 #if ENABLE_FEATURE_WGET_STATUSBAR
977 G.curfile = bb_get_last_path_component_nostrip(G.fname_out);
980 /* Determine where to start transfer */
982 if (option_mask32 & WGET_OPT_CONTINUE) {
983 G.output_fd = open(G.fname_out, O_WRONLY);
984 if (G.output_fd >= 0) {
985 G.beg_range = xlseek(G.output_fd, 0, SEEK_END);
987 /* File doesn't exist. We do not create file here yet.
988 * We are not sure it exists on remote side */
993 lsa = xhost2sockaddr(server.host, server.port);
994 if (!(option_mask32 & WGET_OPT_QUIET)) {
995 char *s = xmalloc_sockaddr2dotted(&lsa->u.sa);
996 fprintf(stderr, "Connecting to %s (%s)\n", server.host, s);
1000 /*G.content_len = 0; - redundant, got_clen = 0 is enough */
1003 if (use_proxy || target.protocol != P_FTP) {
1010 /* Open socket to http(s) server */
1011 #if ENABLE_FEATURE_WGET_OPENSSL
1012 /* openssl (and maybe ssl_helper) support is configured */
1013 if (target.protocol == P_HTTPS) {
1014 /* openssl-based helper
1015 * Inconvenient API since we can't give it an open fd
1017 int fd = spawn_https_helper_openssl(server.host, server.port);
1018 # if ENABLE_FEATURE_WGET_SSL_HELPER
1019 if (fd < 0) { /* no openssl? try ssl_helper */
1020 sfp = open_socket(lsa);
1021 spawn_https_helper_small(fileno(sfp));
1025 /* We don't check for exec("openssl") failure in this case */
1027 sfp = fdopen(fd, "r+");
1029 bb_perror_msg_and_die(bb_msg_memory_exhausted);
1032 sfp = open_socket(lsa);
1034 #elif ENABLE_FEATURE_WGET_SSL_HELPER
1035 /* Only ssl_helper support is configured */
1036 sfp = open_socket(lsa);
1037 if (target.protocol == P_HTTPS)
1038 spawn_https_helper_small(fileno(sfp));
1040 /* ssl (https) support is not configured */
1041 sfp = open_socket(lsa);
1043 /* Send HTTP request */
1045 SENDFMT(sfp, "GET %s://%s/%s HTTP/1.1\r\n",
1046 target.protocol, target.host,
1049 SENDFMT(sfp, "%s /%s HTTP/1.1\r\n",
1050 (option_mask32 & WGET_OPT_POST_DATA) ? "POST" : "GET",
1053 if (!USR_HEADER_HOST)
1054 SENDFMT(sfp, "Host: %s\r\n", target.host);
1055 if (!USR_HEADER_USER_AGENT)
1056 SENDFMT(sfp, "User-Agent: %s\r\n", G.user_agent);
1058 /* Ask server to close the connection as soon as we are done
1059 * (IOW: we do not intend to send more requests)
1061 SENDFMT(sfp, "Connection: close\r\n");
1063 #if ENABLE_FEATURE_WGET_AUTHENTICATION
1064 if (target.user && !USR_HEADER_AUTH) {
1065 SENDFMT(sfp, "Proxy-Authorization: Basic %s\r\n"+6,
1066 base64enc(target.user));
1068 if (use_proxy && server.user && !USR_HEADER_PROXY_AUTH) {
1069 SENDFMT(sfp, "Proxy-Authorization: Basic %s\r\n",
1070 base64enc(server.user));
1074 if (G.beg_range != 0 && !USR_HEADER_RANGE)
1075 SENDFMT(sfp, "Range: bytes=%"OFF_FMT"u-\r\n", G.beg_range);
1077 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1078 if (G.extra_headers) {
1079 log_io(G.extra_headers);
1080 fputs(G.extra_headers, sfp);
1083 if (option_mask32 & WGET_OPT_POST_DATA) {
1085 "Content-Type: application/x-www-form-urlencoded\r\n"
1086 "Content-Length: %u\r\n"
1089 (int) strlen(G.post_data), G.post_data
1094 SENDFMT(sfp, "\r\n");
1099 /* Tried doing this unconditionally.
1100 * Cloudflare and nginx/1.11.5 are shocked to see SHUT_WR on non-HTTPS.
1103 if (target.protocol == P_HTTPS) {
1104 /* If we use SSL helper, keeping our end of the socket open for writing
1105 * makes our end (i.e. the same fd!) readable (EAGAIN instead of EOF)
1106 * even after child closes its copy of the fd.
1109 shutdown(fileno(sfp), SHUT_WR);
1114 * Retrieve HTTP response line and check for "200" status code.
1117 fgets_and_trim(sfp);
1120 str = skip_non_whitespace(str);
1121 str = skip_whitespace(str);
1122 // FIXME: no error check
1123 // xatou wouldn't work: "200 OK"
1128 while (gethdr(sfp) != NULL)
1129 /* eat all remaining headers */;
1132 /* Success responses */
1135 case 201: /* 201 Created */
1136 /* "The request has been fulfilled and resulted in a new resource being created" */
1137 /* Standard wget is reported to treat this as success */
1139 case 202: /* 202 Accepted */
1140 /* "The request has been accepted for processing, but the processing has not been completed" */
1141 /* Treat as success: fall through */
1142 case 203: /* 203 Non-Authoritative Information */
1143 /* "Use of this response code is not required and is only appropriate when the response would otherwise be 200 (OK)" */
1145 case 204: /* 204 No Content */
1147 Response 204 doesn't say "null file", it says "metadata
1148 has changed but data didn't":
1150 "10.2.5 204 No Content
1151 The server has fulfilled the request but does not need to return
1152 an entity-body, and might want to return updated metainformation.
1153 The response MAY include new or updated metainformation in the form
1154 of entity-headers, which if present SHOULD be associated with
1155 the requested variant.
1157 If the client is a user agent, it SHOULD NOT change its document
1158 view from that which caused the request to be sent. This response
1159 is primarily intended to allow input for actions to take place
1160 without causing a change to the user agent's active document view,
1161 although any new or updated metainformation SHOULD be applied
1162 to the document currently in the user agent's active view.
1164 The 204 response MUST NOT include a message-body, and thus
1165 is always terminated by the first empty line after the header fields."
1167 However, in real world it was observed that some web servers
1168 (e.g. Boa/0.94.14rc21) simply use code 204 when file size is zero.
1170 if (G.beg_range != 0) {
1171 /* "Range:..." was not honored by the server.
1172 * Restart download from the beginning.
1174 reset_beg_range_to_zero();
1177 /* 205 Reset Content ?? what to do on this ?? */
1179 case 300: /* redirection */
1185 case 206: /* Partial Content */
1186 if (G.beg_range != 0)
1187 /* "Range:..." worked. Good. */
1189 /* Partial Content even though we did not ask for it??? */
1192 bb_error_msg_and_die("server returned error: %s", sanitize_string(G.wget_buf));
1196 * Retrieve HTTP headers.
1198 while ((str = gethdr(sfp)) != NULL) {
1199 static const char keywords[] ALIGN1 =
1200 "content-length\0""transfer-encoding\0""location\0";
1202 KEY_content_length = 1, KEY_transfer_encoding, KEY_location
1206 /* gethdr converted "FOO:" string to lowercase */
1208 /* strip trailing whitespace */
1209 char *s = strchrnul(str, '\0') - 1;
1210 while (s >= str && (*s == ' ' || *s == '\t')) {
1214 key = index_in_strings(keywords, G.wget_buf) + 1;
1215 if (key == KEY_content_length) {
1216 G.content_len = BB_STRTOOFF(str, NULL, 10);
1217 if (G.content_len < 0 || errno) {
1218 bb_error_msg_and_die("content-length %s is garbage", sanitize_string(str));
1223 if (key == KEY_transfer_encoding) {
1224 if (strcmp(str_tolower(str), "chunked") != 0)
1225 bb_error_msg_and_die("transfer encoding '%s' is not supported", sanitize_string(str));
1228 if (key == KEY_location && status >= 300) {
1229 if (--redir_limit == 0)
1230 bb_error_msg_and_die("too many redirections");
1232 if (str[0] == '/') {
1233 free(redirected_path);
1234 target.path = redirected_path = xstrdup(str+1);
1235 /* lsa stays the same: it's on the same server */
1237 parse_url(str, &target);
1239 /* server.user remains untouched */
1240 free(server.allocated);
1241 server.allocated = NULL;
1242 server.host = target.host;
1243 /* strip_ipv6_scope_id(target.host); - no! */
1244 /* we assume remote never gives us IPv6 addr with scope id */
1245 server.port = target.port;
1248 } /* else: lsa stays the same: we use proxy */
1250 goto establish_session;
1253 // if (status >= 300)
1254 // bb_error_msg_and_die("bad redirection (no Location: header from server)");
1256 /* For HTTP, data is pumped over the same connection */
1262 sfp = prepare_ftp_session(&dfp, &target, lsa);
1267 if (!(option_mask32 & WGET_OPT_SPIDER)) {
1268 if (G.output_fd < 0)
1269 G.output_fd = xopen(G.fname_out, G.o_flags);
1270 retrieve_file_data(dfp);
1271 if (!(option_mask32 & WGET_OPT_OUTNAME)) {
1272 xclose(G.output_fd);
1278 /* It's ftp. Close data connection properly */
1280 if (ftpcmd(NULL, NULL, sfp) != 226)
1281 bb_error_msg_and_die("ftp error: %s", sanitize_string(G.wget_buf + 4));
1282 /* ftpcmd("QUIT", NULL, sfp); - why bother? */
1286 free(server.allocated);
1287 free(target.allocated);
1290 free(fname_out_alloc);
1291 free(redirected_path);
1294 int wget_main(int argc, char **argv) MAIN_EXTERNALLY_VISIBLE;
1295 int wget_main(int argc UNUSED_PARAM, char **argv)
1297 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1298 static const char wget_longopts[] ALIGN1 =
1299 /* name, has_arg, val */
1300 "continue\0" No_argument "c"
1301 "quiet\0" No_argument "q"
1302 "output-document\0" Required_argument "O"
1303 "directory-prefix\0" Required_argument "P"
1304 "proxy\0" Required_argument "Y"
1305 "user-agent\0" Required_argument "U"
1306 IF_FEATURE_WGET_TIMEOUT(
1307 "timeout\0" Required_argument "T")
1309 IF_DESKTOP( "tries\0" Required_argument "t")
1310 "header\0" Required_argument "\xff"
1311 "post-data\0" Required_argument "\xfe"
1312 "spider\0" No_argument "\xfd"
1313 /* Ignored (we always use PASV): */
1314 IF_DESKTOP( "passive-ftp\0" No_argument "\xf0")
1315 /* Ignored (we don't do ssl) */
1316 IF_DESKTOP( "no-check-certificate\0" No_argument "\xf0")
1317 /* Ignored (we don't support caching) */
1318 IF_DESKTOP( "no-cache\0" No_argument "\xf0")
1319 IF_DESKTOP( "no-verbose\0" No_argument "\xf0")
1320 IF_DESKTOP( "no-clobber\0" No_argument "\xf0")
1321 IF_DESKTOP( "no-host-directories\0" No_argument "\xf0")
1322 IF_DESKTOP( "no-parent\0" No_argument "\xf0")
1326 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1327 llist_t *headers_llist = NULL;
1332 #if ENABLE_FEATURE_WGET_TIMEOUT
1333 G.timeout_seconds = 900;
1334 signal(SIGALRM, alarm_handler);
1336 G.proxy_flag = "on"; /* use proxies if env vars are set */
1337 G.user_agent = "Wget"; /* "User-Agent" header field */
1339 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1340 applet_long_options = wget_longopts;
1342 opt_complementary = "-1" /* at least one URL */
1343 IF_FEATURE_WGET_LONG_OPTIONS(":\xff::"); /* --header is a list */
1344 getopt32(argv, "cqO:P:Y:U:T:+"
1347 /* wget has exactly four -n<letter> opts, all of which we can ignore:
1348 * -nv --no-verbose: be moderately quiet (-q is full quiet)
1349 * -nc --no-clobber: abort if exists, neither download to FILE.n nor overwrite FILE
1350 * -nH --no-host-directories: wget -r http://host/ won't create host/
1352 * "n::" above says that we accept -n[ARG].
1353 * Specifying "n:" would be a bug: "-n ARG" would eat ARG!
1355 , &G.fname_out, &G.dir_prefix,
1356 &G.proxy_flag, &G.user_agent,
1357 IF_FEATURE_WGET_TIMEOUT(&G.timeout_seconds) IF_NOT_FEATURE_WGET_TIMEOUT(NULL),
1358 NULL, /* -t RETRIES */
1360 IF_FEATURE_WGET_LONG_OPTIONS(, &headers_llist)
1361 IF_FEATURE_WGET_LONG_OPTIONS(, &G.post_data)
1363 #if 0 /* option bits debug */
1364 if (option_mask32 & WGET_OPT_RETRIES) bb_error_msg("-t NUM");
1365 if (option_mask32 & WGET_OPT_nsomething) bb_error_msg("-nsomething");
1366 if (option_mask32 & WGET_OPT_HEADER) bb_error_msg("--header");
1367 if (option_mask32 & WGET_OPT_POST_DATA) bb_error_msg("--post-data");
1368 if (option_mask32 & WGET_OPT_SPIDER) bb_error_msg("--spider");
1373 #if ENABLE_FEATURE_WGET_LONG_OPTIONS
1374 if (headers_llist) {
1377 llist_t *ll = headers_llist;
1379 size += strlen(ll->data) + 2;
1382 G.extra_headers = hdr = xmalloc(size + 1);
1383 while (headers_llist) {
1387 size = sprintf(hdr, "%s\r\n",
1388 (char*)llist_pop(&headers_llist));
1389 /* a bit like index_in_substrings but don't match full key */
1391 words = wget_user_headers;
1393 if (strstr(hdr, words) == hdr) {
1394 G.user_headers |= bit;
1398 words += strlen(words) + 1;
1406 G.o_flags = O_WRONLY | O_CREAT | O_TRUNC | O_EXCL;
1407 if (G.fname_out) { /* -O FILE ? */
1408 if (LONE_DASH(G.fname_out)) { /* -O - ? */
1410 option_mask32 &= ~WGET_OPT_CONTINUE;
1412 /* compat with wget: -O FILE can overwrite */
1413 G.o_flags = O_WRONLY | O_CREAT | O_TRUNC;
1417 download_one_url(*argv++);
1419 if (G.output_fd >= 0)
1420 xclose(G.output_fd);
1422 #if ENABLE_FEATURE_CLEAN_UP && ENABLE_FEATURE_WGET_LONG_OPTIONS
1423 free(G.extra_headers);
1427 return EXIT_SUCCESS;