diff options
author | Brad King <brad.king@kitware.com> | 2018-01-24 19:17:42 (GMT) |
---|---|---|
committer | Brad King <brad.king@kitware.com> | 2018-01-24 19:18:16 (GMT) |
commit | cd8e31a1bf7429514078c2923a1a9580113f9d4f (patch) | |
tree | 6a0b8a01e3e0e52bb4b3abb66ce9e992614af6a7 /Utilities/cmcurl/lib/url.c | |
parent | e9c8ea75575afdb4e87b262641ee4071ef42b4c6 (diff) | |
parent | af9e654045f11028e50dac4781e297834129a749 (diff) | |
download | CMake-cd8e31a1bf7429514078c2923a1a9580113f9d4f.zip CMake-cd8e31a1bf7429514078c2923a1a9580113f9d4f.tar.gz CMake-cd8e31a1bf7429514078c2923a1a9580113f9d4f.tar.bz2 |
Merge branch 'upstream-curl' into update-curl
* upstream-curl:
curl 2018-01-23 (d6c21c8e)
Diffstat (limited to 'Utilities/cmcurl/lib/url.c')
-rw-r--r-- | Utilities/cmcurl/lib/url.c | 2965 |
1 files changed, 303 insertions, 2662 deletions
diff --git a/Utilities/cmcurl/lib/url.c b/Utilities/cmcurl/lib/url.c index 584635b..74813e8 100644 --- a/Utilities/cmcurl/lib/url.c +++ b/Utilities/cmcurl/lib/url.c @@ -5,7 +5,7 @@ * | (__| |_| | _ <| |___ * \___|\___/|_| \_\_____| * - * Copyright (C) 1998 - 2017, Daniel Stenberg, <daniel@haxx.se>, et al. + * Copyright (C) 1998 - 2018, Daniel Stenberg, <daniel@haxx.se>, et al. * * This software is licensed as described in the file COPYING, which * you should have received as part of this distribution. The terms @@ -55,9 +55,7 @@ #error "We can't compile without socket() support!" #endif -#ifdef HAVE_LIMITS_H #include <limits.h> -#endif #ifdef USE_LIBIDN2 #include <idn2.h> @@ -120,15 +118,13 @@ bool curl_win32_idn_to_ascii(const char *in, char **out); #include "pipeline.h" #include "dotdot.h" #include "strdup.h" +#include "setopt.h" + /* The last 3 #include files should be in this order */ #include "curl_printf.h" #include "curl_memory.h" #include "memdebug.h" -/* Local static prototypes */ -static struct connectdata * -find_oldest_idle_connection_in_bundle(struct Curl_easy *data, - struct connectbundle *bundle); static void conn_free(struct connectdata *conn); static void free_fixed_hostname(struct hostname *host); static void signalPipeClose(struct curl_llist *pipeline, bool pipe_broke); @@ -136,15 +132,8 @@ static CURLcode parse_url_login(struct Curl_easy *data, struct connectdata *conn, char **userptr, char **passwdptr, char **optionsptr); -static CURLcode parse_login_details(const char *login, const size_t len, - char **userptr, char **passwdptr, - char **optionsptr); static unsigned int get_protocol_family(unsigned int protocol); -#define READBUFFER_SIZE CURL_MAX_WRITE_SIZE -#define READBUFFER_MAX CURL_MAX_READ_SIZE -#define READBUFFER_MIN 1024 - /* Some parts of the code (e.g. chunked encoding) assume this buffer has at * more than just a few bytes to play with. Don't let it become too small or * bad things will happen. @@ -201,8 +190,11 @@ static const struct Curl_handler * const protocols[] = { &Curl_handler_tftp, #endif -#ifdef USE_LIBSSH2 +#if defined(USE_LIBSSH2) || defined(USE_LIBSSH) &Curl_handler_scp, +#endif + +#if defined(USE_LIBSSH2) || defined(USE_LIBSSH) &Curl_handler_sftp, #endif @@ -299,98 +291,8 @@ void Curl_freeset(struct Curl_easy *data) data->change.url_alloc = FALSE; } data->change.url = NULL; -} - -static CURLcode setstropt(char **charp, const char *s) -{ - /* Release the previous storage at `charp' and replace by a dynamic storage - copy of `s'. Return CURLE_OK or CURLE_OUT_OF_MEMORY. */ - - Curl_safefree(*charp); - - if(s) { - char *str = strdup(s); - - if(!str) - return CURLE_OUT_OF_MEMORY; - - *charp = str; - } - - return CURLE_OK; -} - -static CURLcode setstropt_userpwd(char *option, char **userp, char **passwdp) -{ - CURLcode result = CURLE_OK; - char *user = NULL; - char *passwd = NULL; - - /* Parse the login details if specified. It not then we treat NULL as a hint - to clear the existing data */ - if(option) { - result = parse_login_details(option, strlen(option), - (userp ? &user : NULL), - (passwdp ? &passwd : NULL), - NULL); - } - - if(!result) { - /* Store the username part of option if required */ - if(userp) { - if(!user && option && option[0] == ':') { - /* Allocate an empty string instead of returning NULL as user name */ - user = strdup(""); - if(!user) - result = CURLE_OUT_OF_MEMORY; - } - - Curl_safefree(*userp); - *userp = user; - } - - /* Store the password part of option if required */ - if(passwdp) { - Curl_safefree(*passwdp); - *passwdp = passwd; - } - } - - return result; -} - -CURLcode Curl_dupset(struct Curl_easy *dst, struct Curl_easy *src) -{ - CURLcode result = CURLE_OK; - enum dupstring i; - - /* Copy src->set into dst->set first, then deal with the strings - afterwards */ - dst->set = src->set; - - /* clear all string pointers first */ - memset(dst->set.str, 0, STRING_LAST * sizeof(char *)); - /* duplicate all strings */ - for(i = (enum dupstring)0; i< STRING_LASTZEROTERMINATED; i++) { - result = setstropt(&dst->set.str[i], src->set.str[i]); - if(result) - return result; - } - - /* duplicate memory areas pointed to */ - i = STRING_COPYPOSTFIELDS; - if(src->set.postfieldsize && src->set.str[i]) { - /* postfieldsize is curl_off_t, Curl_memdup() takes a size_t ... */ - dst->set.str[i] = Curl_memdup(src->set.str[i], - curlx_sotouz(src->set.postfieldsize)); - if(!dst->set.str[i]) - return CURLE_OUT_OF_MEMORY; - /* point to the new copy */ - dst->set.postfields = dst->set.str[i]; - } - - return CURLE_OK; + Curl_mime_cleanpart(&data->set.mimepost); } /* @@ -480,8 +382,6 @@ CURLcode Curl_close(struct Curl_easy *data) Curl_http2_cleanup_dependencies(data); Curl_convert_close(data); - Curl_mime_cleanpart(&data->set.mimepost); - /* No longer a dirty share, if it exists */ if(data->share) { Curl_share_lock(data, CURL_LOCK_DATA_SHARE, CURL_LOCK_ACCESS_SINGLE); @@ -489,12 +389,8 @@ CURLcode Curl_close(struct Curl_easy *data) Curl_share_unlock(data, CURL_LOCK_DATA_SHARE); } - if(data->set.wildcardmatch) { - /* destruct wildcard structures if it is needed */ - struct WildcardData *wc = &data->wildcard; - Curl_wildcard_dtor(wc); - } - + /* destruct wildcard structures if it is needed */ + Curl_wildcard_dtor(&data->wildcard); Curl_freeset(data); free(data); return CURLE_OK; @@ -504,8 +400,9 @@ CURLcode Curl_close(struct Curl_easy *data) * Initialize the UserDefined fields within a Curl_easy. * This may be safely called on a new or existing Curl_easy. */ -CURLcode Curl_init_userdefined(struct UserDefined *set) +CURLcode Curl_init_userdefined(struct Curl_easy *data) { + struct UserDefined *set = &data->set; CURLcode result = CURLE_OK; set->out = stdout; /* default output to stdout */ @@ -555,6 +452,8 @@ CURLcode Curl_init_userdefined(struct UserDefined *set) /* make libcurl quiet by default: */ set->hide_progress = TRUE; /* CURLOPT_NOPROGRESS changes these */ + Curl_mime_initpart(&set->mimepost, data); + /* * libcurl 7.10 introduced SSL verification *by default*! This needs to be * switched off unless wanted. @@ -591,25 +490,25 @@ CURLcode Curl_init_userdefined(struct UserDefined *set) /* This is our preferred CA cert bundle/path since install time */ #if defined(CURL_CA_BUNDLE) - result = setstropt(&set->str[STRING_SSL_CAFILE_ORIG], CURL_CA_BUNDLE); + result = Curl_setstropt(&set->str[STRING_SSL_CAFILE_ORIG], CURL_CA_BUNDLE); if(result) return result; - result = setstropt(&set->str[STRING_SSL_CAFILE_PROXY], CURL_CA_BUNDLE); + result = Curl_setstropt(&set->str[STRING_SSL_CAFILE_PROXY], CURL_CA_BUNDLE); if(result) return result; #endif #if defined(CURL_CA_PATH) - result = setstropt(&set->str[STRING_SSL_CAPATH_ORIG], CURL_CA_PATH); + result = Curl_setstropt(&set->str[STRING_SSL_CAPATH_ORIG], CURL_CA_PATH); if(result) return result; - result = setstropt(&set->str[STRING_SSL_CAPATH_PROXY], CURL_CA_PATH); + result = Curl_setstropt(&set->str[STRING_SSL_CAPATH_PROXY], CURL_CA_PATH); if(result) return result; #endif - set->wildcardmatch = FALSE; + set->wildcard_enabled = FALSE; set->chunk_bgn = ZERO_NULL; set->chunk_end = ZERO_NULL; @@ -670,32 +569,29 @@ CURLcode Curl_open(struct Curl_easy **curl) DEBUGF(fprintf(stderr, "Error: malloc of buffer failed\n")); result = CURLE_OUT_OF_MEMORY; } - - Curl_mime_initpart(&data->set.mimepost, data); - - data->state.headerbuff = malloc(HEADERSIZE); - if(!data->state.headerbuff) { - DEBUGF(fprintf(stderr, "Error: malloc of headerbuff failed\n")); - result = CURLE_OUT_OF_MEMORY; - } else { - result = Curl_init_userdefined(&data->set); - - data->state.headersize = HEADERSIZE; - - Curl_convert_init(data); + data->state.headerbuff = malloc(HEADERSIZE); + if(!data->state.headerbuff) { + DEBUGF(fprintf(stderr, "Error: malloc of headerbuff failed\n")); + result = CURLE_OUT_OF_MEMORY; + } + else { + result = Curl_init_userdefined(data); - Curl_initinfo(data); + data->state.headersize = HEADERSIZE; + Curl_convert_init(data); + Curl_initinfo(data); - /* most recent connection is not yet defined */ - data->state.lastconnect = NULL; + /* most recent connection is not yet defined */ + data->state.lastconnect = NULL; - data->progress.flags |= PGRS_HIDE; - data->state.current_speed = -1; /* init to negative == impossible */ - data->set.fnmatch = ZERO_NULL; - data->set.maxconnects = DEFAULT_CONNCACHE_SIZE; /* for easy handles */ + data->progress.flags |= PGRS_HIDE; + data->state.current_speed = -1; /* init to negative == impossible */ + data->set.fnmatch = ZERO_NULL; + data->set.maxconnects = DEFAULT_CONNCACHE_SIZE; /* for easy handles */ - Curl_http2_init_state(&data->state); + Curl_http2_init_state(&data->state); + } } if(result) { @@ -712,2269 +608,6 @@ CURLcode Curl_open(struct Curl_easy **curl) return result; } -#define C_SSLVERSION_VALUE(x) (x & 0xffff) -#define C_SSLVERSION_MAX_VALUE(x) (x & 0xffff0000) - -CURLcode Curl_setopt(struct Curl_easy *data, CURLoption option, - va_list param) -{ - char *argptr; - CURLcode result = CURLE_OK; - long arg; -#ifndef CURL_DISABLE_HTTP - curl_off_t bigsize; -#endif - - switch(option) { - case CURLOPT_DNS_CACHE_TIMEOUT: - data->set.dns_cache_timeout = va_arg(param, long); - break; - case CURLOPT_DNS_USE_GLOBAL_CACHE: - /* remember we want this enabled */ - arg = va_arg(param, long); - data->set.global_dns_cache = (0 != arg) ? TRUE : FALSE; - break; - case CURLOPT_SSL_CIPHER_LIST: - /* set a list of cipher we want to use in the SSL connection */ - result = setstropt(&data->set.str[STRING_SSL_CIPHER_LIST_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_SSL_CIPHER_LIST: - /* set a list of cipher we want to use in the SSL connection for proxy */ - result = setstropt(&data->set.str[STRING_SSL_CIPHER_LIST_PROXY], - va_arg(param, char *)); - break; - - case CURLOPT_RANDOM_FILE: - /* - * This is the path name to a file that contains random data to seed - * the random SSL stuff with. The file is only used for reading. - */ - result = setstropt(&data->set.str[STRING_SSL_RANDOM_FILE], - va_arg(param, char *)); - break; - case CURLOPT_EGDSOCKET: - /* - * The Entropy Gathering Daemon socket pathname - */ - result = setstropt(&data->set.str[STRING_SSL_EGDSOCKET], - va_arg(param, char *)); - break; - case CURLOPT_MAXCONNECTS: - /* - * Set the absolute number of maximum simultaneous alive connection that - * libcurl is allowed to have. - */ - data->set.maxconnects = va_arg(param, long); - break; - case CURLOPT_FORBID_REUSE: - /* - * When this transfer is done, it must not be left to be reused by a - * subsequent transfer but shall be closed immediately. - */ - data->set.reuse_forbid = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_FRESH_CONNECT: - /* - * This transfer shall not use a previously cached connection but - * should be made with a fresh new connect! - */ - data->set.reuse_fresh = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_VERBOSE: - /* - * Verbose means infof() calls that give a lot of information about - * the connection and transfer procedures as well as internal choices. - */ - data->set.verbose = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_HEADER: - /* - * Set to include the header in the general data output stream. - */ - data->set.include_header = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_NOPROGRESS: - /* - * Shut off the internal supported progress meter - */ - data->set.hide_progress = (0 != va_arg(param, long)) ? TRUE : FALSE; - if(data->set.hide_progress) - data->progress.flags |= PGRS_HIDE; - else - data->progress.flags &= ~PGRS_HIDE; - break; - case CURLOPT_NOBODY: - /* - * Do not include the body part in the output data stream. - */ - data->set.opt_no_body = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_FAILONERROR: - /* - * Don't output the >=400 error code HTML-page, but instead only - * return error. - */ - data->set.http_fail_on_error = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_KEEP_SENDING_ON_ERROR: - data->set.http_keep_sending_on_error = (0 != va_arg(param, long)) ? - TRUE : FALSE; - break; - case CURLOPT_UPLOAD: - case CURLOPT_PUT: - /* - * We want to sent data to the remote host. If this is HTTP, that equals - * using the PUT request. - */ - data->set.upload = (0 != va_arg(param, long)) ? TRUE : FALSE; - if(data->set.upload) { - /* If this is HTTP, PUT is what's needed to "upload" */ - data->set.httpreq = HTTPREQ_PUT; - data->set.opt_no_body = FALSE; /* this is implied */ - } - else - /* In HTTP, the opposite of upload is GET (unless NOBODY is true as - then this can be changed to HEAD later on) */ - data->set.httpreq = HTTPREQ_GET; - break; - case CURLOPT_REQUEST_TARGET: - result = setstropt(&data->set.str[STRING_TARGET], - va_arg(param, char *)); - break; - case CURLOPT_FILETIME: - /* - * Try to get the file time of the remote document. The time will - * later (possibly) become available using curl_easy_getinfo(). - */ - data->set.get_filetime = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_FTP_CREATE_MISSING_DIRS: - /* - * An FTP option that modifies an upload to create missing directories on - * the server. - */ - switch(va_arg(param, long)) { - case 0: - data->set.ftp_create_missing_dirs = 0; - break; - case 1: - data->set.ftp_create_missing_dirs = 1; - break; - case 2: - data->set.ftp_create_missing_dirs = 2; - break; - default: - /* reserve other values for future use */ - result = CURLE_UNKNOWN_OPTION; - break; - } - break; - case CURLOPT_SERVER_RESPONSE_TIMEOUT: - /* - * Option that specifies how quickly an server response must be obtained - * before it is considered failure. For pingpong protocols. - */ - data->set.server_response_timeout = va_arg(param, long) * 1000; - break; - case CURLOPT_TFTP_NO_OPTIONS: - /* - * Option that prevents libcurl from sending TFTP option requests to the - * server. - */ - data->set.tftp_no_options = va_arg(param, long) != 0; - break; - case CURLOPT_TFTP_BLKSIZE: - /* - * TFTP option that specifies the block size to use for data transmission. - */ - data->set.tftp_blksize = va_arg(param, long); - break; - case CURLOPT_DIRLISTONLY: - /* - * An option that changes the command to one that asks for a list - * only, no file info details. - */ - data->set.ftp_list_only = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_APPEND: - /* - * We want to upload and append to an existing file. - */ - data->set.ftp_append = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_FTP_FILEMETHOD: - /* - * How do access files over FTP. - */ - data->set.ftp_filemethod = (curl_ftpfile)va_arg(param, long); - break; - case CURLOPT_NETRC: - /* - * Parse the $HOME/.netrc file - */ - data->set.use_netrc = (enum CURL_NETRC_OPTION)va_arg(param, long); - break; - case CURLOPT_NETRC_FILE: - /* - * Use this file instead of the $HOME/.netrc file - */ - result = setstropt(&data->set.str[STRING_NETRC_FILE], - va_arg(param, char *)); - break; - case CURLOPT_TRANSFERTEXT: - /* - * This option was previously named 'FTPASCII'. Renamed to work with - * more protocols than merely FTP. - * - * Transfer using ASCII (instead of BINARY). - */ - data->set.prefer_ascii = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_TIMECONDITION: - /* - * Set HTTP time condition. This must be one of the defines in the - * curl/curl.h header file. - */ - data->set.timecondition = (curl_TimeCond)va_arg(param, long); - break; - case CURLOPT_TIMEVALUE: - /* - * This is the value to compare with the remote document with the - * method set with CURLOPT_TIMECONDITION - */ - data->set.timevalue = (time_t)va_arg(param, long); - break; - case CURLOPT_SSLVERSION: - /* - * Set explicit SSL version to try to connect with, as some SSL - * implementations are lame. - */ -#ifdef USE_SSL - arg = va_arg(param, long); - data->set.ssl.primary.version = C_SSLVERSION_VALUE(arg); - data->set.ssl.primary.version_max = C_SSLVERSION_MAX_VALUE(arg); -#else - result = CURLE_UNKNOWN_OPTION; -#endif - break; - case CURLOPT_PROXY_SSLVERSION: - /* - * Set explicit SSL version to try to connect with for proxy, as some SSL - * implementations are lame. - */ -#ifdef USE_SSL - arg = va_arg(param, long); - data->set.proxy_ssl.primary.version = C_SSLVERSION_VALUE(arg); - data->set.proxy_ssl.primary.version_max = C_SSLVERSION_MAX_VALUE(arg); -#else - result = CURLE_UNKNOWN_OPTION; -#endif - break; - -#ifndef CURL_DISABLE_HTTP - case CURLOPT_AUTOREFERER: - /* - * Switch on automatic referer that gets set if curl follows locations. - */ - data->set.http_auto_referer = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_ACCEPT_ENCODING: - /* - * String to use at the value of Accept-Encoding header. - * - * If the encoding is set to "" we use an Accept-Encoding header that - * encompasses all the encodings we support. - * If the encoding is set to NULL we don't send an Accept-Encoding header - * and ignore an received Content-Encoding header. - * - */ - argptr = va_arg(param, char *); - result = setstropt(&data->set.str[STRING_ENCODING], - (argptr && !*argptr)? - ALL_CONTENT_ENCODINGS: argptr); - break; - - case CURLOPT_TRANSFER_ENCODING: - data->set.http_transfer_encoding = (0 != va_arg(param, long)) ? - TRUE : FALSE; - break; - - case CURLOPT_FOLLOWLOCATION: - /* - * Follow Location: header hints on a HTTP-server. - */ - data->set.http_follow_location = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_UNRESTRICTED_AUTH: - /* - * Send authentication (user+password) when following locations, even when - * hostname changed. - */ - data->set.http_disable_hostname_check_before_authentication = - (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_MAXREDIRS: - /* - * The maximum amount of hops you allow curl to follow Location: - * headers. This should mostly be used to detect never-ending loops. - */ - data->set.maxredirs = va_arg(param, long); - break; - - case CURLOPT_POSTREDIR: - { - /* - * Set the behaviour of POST when redirecting - * CURL_REDIR_GET_ALL - POST is changed to GET after 301 and 302 - * CURL_REDIR_POST_301 - POST is kept as POST after 301 - * CURL_REDIR_POST_302 - POST is kept as POST after 302 - * CURL_REDIR_POST_303 - POST is kept as POST after 303 - * CURL_REDIR_POST_ALL - POST is kept as POST after 301, 302 and 303 - * other - POST is kept as POST after 301 and 302 - */ - arg = va_arg(param, long); - data->set.keep_post = arg & CURL_REDIR_POST_ALL; - } - break; - - case CURLOPT_POST: - /* Does this option serve a purpose anymore? Yes it does, when - CURLOPT_POSTFIELDS isn't used and the POST data is read off the - callback! */ - if(va_arg(param, long)) { - data->set.httpreq = HTTPREQ_POST; - data->set.opt_no_body = FALSE; /* this is implied */ - } - else - data->set.httpreq = HTTPREQ_GET; - break; - - case CURLOPT_COPYPOSTFIELDS: - /* - * A string with POST data. Makes curl HTTP POST. Even if it is NULL. - * If needed, CURLOPT_POSTFIELDSIZE must have been set prior to - * CURLOPT_COPYPOSTFIELDS and not altered later. - */ - argptr = va_arg(param, char *); - - if(!argptr || data->set.postfieldsize == -1) - result = setstropt(&data->set.str[STRING_COPYPOSTFIELDS], argptr); - else { - /* - * Check that requested length does not overflow the size_t type. - */ - - if((data->set.postfieldsize < 0) || - ((sizeof(curl_off_t) != sizeof(size_t)) && - (data->set.postfieldsize > (curl_off_t)((size_t)-1)))) - result = CURLE_OUT_OF_MEMORY; - else { - char *p; - - (void) setstropt(&data->set.str[STRING_COPYPOSTFIELDS], NULL); - - /* Allocate even when size == 0. This satisfies the need of possible - later address compare to detect the COPYPOSTFIELDS mode, and - to mark that postfields is used rather than read function or - form data. - */ - p = malloc((size_t)(data->set.postfieldsize? - data->set.postfieldsize:1)); - - if(!p) - result = CURLE_OUT_OF_MEMORY; - else { - if(data->set.postfieldsize) - memcpy(p, argptr, (size_t)data->set.postfieldsize); - - data->set.str[STRING_COPYPOSTFIELDS] = p; - } - } - } - - data->set.postfields = data->set.str[STRING_COPYPOSTFIELDS]; - data->set.httpreq = HTTPREQ_POST; - break; - - case CURLOPT_POSTFIELDS: - /* - * Like above, but use static data instead of copying it. - */ - data->set.postfields = va_arg(param, void *); - /* Release old copied data. */ - (void) setstropt(&data->set.str[STRING_COPYPOSTFIELDS], NULL); - data->set.httpreq = HTTPREQ_POST; - break; - - case CURLOPT_POSTFIELDSIZE: - /* - * The size of the POSTFIELD data to prevent libcurl to do strlen() to - * figure it out. Enables binary posts. - */ - bigsize = va_arg(param, long); - - if(data->set.postfieldsize < bigsize && - data->set.postfields == data->set.str[STRING_COPYPOSTFIELDS]) { - /* Previous CURLOPT_COPYPOSTFIELDS is no longer valid. */ - (void) setstropt(&data->set.str[STRING_COPYPOSTFIELDS], NULL); - data->set.postfields = NULL; - } - - data->set.postfieldsize = bigsize; - break; - - case CURLOPT_POSTFIELDSIZE_LARGE: - /* - * The size of the POSTFIELD data to prevent libcurl to do strlen() to - * figure it out. Enables binary posts. - */ - bigsize = va_arg(param, curl_off_t); - - if(data->set.postfieldsize < bigsize && - data->set.postfields == data->set.str[STRING_COPYPOSTFIELDS]) { - /* Previous CURLOPT_COPYPOSTFIELDS is no longer valid. */ - (void) setstropt(&data->set.str[STRING_COPYPOSTFIELDS], NULL); - data->set.postfields = NULL; - } - - data->set.postfieldsize = bigsize; - break; - - case CURLOPT_HTTPPOST: - /* - * Set to make us do HTTP POST - */ - data->set.httppost = va_arg(param, struct curl_httppost *); - data->set.httpreq = HTTPREQ_POST_FORM; - data->set.opt_no_body = FALSE; /* this is implied */ - break; -#endif /* CURL_DISABLE_HTTP */ - - case CURLOPT_MIMEPOST: - /* - * Set to make us do MIME/form POST - */ - result = curl_mime_subparts(&data->set.mimepost, - va_arg(param, curl_mime *)); - if(!result) { - data->set.mimepost.freefunc = NULL; /* Avoid free upon easy cleanup. */ - data->set.httpreq = HTTPREQ_POST_MIME; - data->set.opt_no_body = FALSE; /* this is implied */ - } - break; - - case CURLOPT_REFERER: - /* - * String to set in the HTTP Referer: field. - */ - if(data->change.referer_alloc) { - Curl_safefree(data->change.referer); - data->change.referer_alloc = FALSE; - } - result = setstropt(&data->set.str[STRING_SET_REFERER], - va_arg(param, char *)); - data->change.referer = data->set.str[STRING_SET_REFERER]; - break; - - case CURLOPT_USERAGENT: - /* - * String to use in the HTTP User-Agent field - */ - result = setstropt(&data->set.str[STRING_USERAGENT], - va_arg(param, char *)); - break; - - case CURLOPT_HTTPHEADER: - /* - * Set a list with HTTP headers to use (or replace internals with) - */ - data->set.headers = va_arg(param, struct curl_slist *); - break; - -#ifndef CURL_DISABLE_HTTP - case CURLOPT_PROXYHEADER: - /* - * Set a list with proxy headers to use (or replace internals with) - * - * Since CURLOPT_HTTPHEADER was the only way to set HTTP headers for a - * long time we remain doing it this way until CURLOPT_PROXYHEADER is - * used. As soon as this option has been used, if set to anything but - * NULL, custom headers for proxies are only picked from this list. - * - * Set this option to NULL to restore the previous behavior. - */ - data->set.proxyheaders = va_arg(param, struct curl_slist *); - break; - - case CURLOPT_HEADEROPT: - /* - * Set header option. - */ - arg = va_arg(param, long); - data->set.sep_headers = (arg & CURLHEADER_SEPARATE)? TRUE: FALSE; - break; - - case CURLOPT_HTTP200ALIASES: - /* - * Set a list of aliases for HTTP 200 in response header - */ - data->set.http200aliases = va_arg(param, struct curl_slist *); - break; - -#if !defined(CURL_DISABLE_COOKIES) - case CURLOPT_COOKIE: - /* - * Cookie string to send to the remote server in the request. - */ - result = setstropt(&data->set.str[STRING_COOKIE], - va_arg(param, char *)); - break; - - case CURLOPT_COOKIEFILE: - /* - * Set cookie file to read and parse. Can be used multiple times. - */ - argptr = (char *)va_arg(param, void *); - if(argptr) { - struct curl_slist *cl; - /* append the cookie file name to the list of file names, and deal with - them later */ - cl = curl_slist_append(data->change.cookielist, argptr); - if(!cl) { - curl_slist_free_all(data->change.cookielist); - data->change.cookielist = NULL; - return CURLE_OUT_OF_MEMORY; - } - data->change.cookielist = cl; /* store the list for later use */ - } - break; - - case CURLOPT_COOKIEJAR: - /* - * Set cookie file name to dump all cookies to when we're done. - */ - { - struct CookieInfo *newcookies; - result = setstropt(&data->set.str[STRING_COOKIEJAR], - va_arg(param, char *)); - - /* - * Activate the cookie parser. This may or may not already - * have been made. - */ - newcookies = Curl_cookie_init(data, NULL, data->cookies, - data->set.cookiesession); - if(!newcookies) - result = CURLE_OUT_OF_MEMORY; - data->cookies = newcookies; - } - break; - - case CURLOPT_COOKIESESSION: - /* - * Set this option to TRUE to start a new "cookie session". It will - * prevent the forthcoming read-cookies-from-file actions to accept - * cookies that are marked as being session cookies, as they belong to a - * previous session. - * - * In the original Netscape cookie spec, "session cookies" are cookies - * with no expire date set. RFC2109 describes the same action if no - * 'Max-Age' is set and RFC2965 includes the RFC2109 description and adds - * a 'Discard' action that can enforce the discard even for cookies that - * have a Max-Age. - * - * We run mostly with the original cookie spec, as hardly anyone implements - * anything else. - */ - data->set.cookiesession = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_COOKIELIST: - argptr = va_arg(param, char *); - - if(argptr == NULL) - break; - - if(strcasecompare(argptr, "ALL")) { - /* clear all cookies */ - Curl_share_lock(data, CURL_LOCK_DATA_COOKIE, CURL_LOCK_ACCESS_SINGLE); - Curl_cookie_clearall(data->cookies); - Curl_share_unlock(data, CURL_LOCK_DATA_COOKIE); - } - else if(strcasecompare(argptr, "SESS")) { - /* clear session cookies */ - Curl_share_lock(data, CURL_LOCK_DATA_COOKIE, CURL_LOCK_ACCESS_SINGLE); - Curl_cookie_clearsess(data->cookies); - Curl_share_unlock(data, CURL_LOCK_DATA_COOKIE); - } - else if(strcasecompare(argptr, "FLUSH")) { - /* flush cookies to file, takes care of the locking */ - Curl_flush_cookies(data, 0); - } - else if(strcasecompare(argptr, "RELOAD")) { - /* reload cookies from file */ - Curl_cookie_loadfiles(data); - break; - } - else { - if(!data->cookies) - /* if cookie engine was not running, activate it */ - data->cookies = Curl_cookie_init(data, NULL, NULL, TRUE); - - argptr = strdup(argptr); - if(!argptr || !data->cookies) { - result = CURLE_OUT_OF_MEMORY; - free(argptr); - } - else { - Curl_share_lock(data, CURL_LOCK_DATA_COOKIE, CURL_LOCK_ACCESS_SINGLE); - - if(checkprefix("Set-Cookie:", argptr)) - /* HTTP Header format line */ - Curl_cookie_add(data, data->cookies, TRUE, argptr + 11, NULL, NULL); - - else - /* Netscape format line */ - Curl_cookie_add(data, data->cookies, FALSE, argptr, NULL, NULL); - - Curl_share_unlock(data, CURL_LOCK_DATA_COOKIE); - free(argptr); - } - } - - break; -#endif /* !CURL_DISABLE_COOKIES */ - - case CURLOPT_HTTPGET: - /* - * Set to force us do HTTP GET - */ - if(va_arg(param, long)) { - data->set.httpreq = HTTPREQ_GET; - data->set.upload = FALSE; /* switch off upload */ - data->set.opt_no_body = FALSE; /* this is implied */ - } - break; - - case CURLOPT_HTTP_VERSION: - /* - * This sets a requested HTTP version to be used. The value is one of - * the listed enums in curl/curl.h. - */ - arg = va_arg(param, long); -#ifndef USE_NGHTTP2 - if(arg >= CURL_HTTP_VERSION_2) - return CURLE_UNSUPPORTED_PROTOCOL; -#endif - data->set.httpversion = arg; - break; - - case CURLOPT_EXPECT_100_TIMEOUT_MS: - /* - * Time to wait for a response to a HTTP request containing an - * Expect: 100-continue header before sending the data anyway. - */ - data->set.expect_100_timeout = va_arg(param, long); - break; - -#endif /* CURL_DISABLE_HTTP */ - - case CURLOPT_HTTPAUTH: - /* - * Set HTTP Authentication type BITMASK. - */ - { - int bitcheck; - bool authbits; - unsigned long auth = va_arg(param, unsigned long); - - if(auth == CURLAUTH_NONE) { - data->set.httpauth = auth; - break; - } - - /* the DIGEST_IE bit is only used to set a special marker, for all the - rest we need to handle it as normal DIGEST */ - data->state.authhost.iestyle = (auth & CURLAUTH_DIGEST_IE) ? TRUE : FALSE; - - if(auth & CURLAUTH_DIGEST_IE) { - auth |= CURLAUTH_DIGEST; /* set standard digest bit */ - auth &= ~CURLAUTH_DIGEST_IE; /* unset ie digest bit */ - } - - /* switch off bits we can't support */ -#ifndef USE_NTLM - auth &= ~CURLAUTH_NTLM; /* no NTLM support */ - auth &= ~CURLAUTH_NTLM_WB; /* no NTLM_WB support */ -#elif !defined(NTLM_WB_ENABLED) - auth &= ~CURLAUTH_NTLM_WB; /* no NTLM_WB support */ -#endif -#ifndef USE_SPNEGO - auth &= ~CURLAUTH_NEGOTIATE; /* no Negotiate (SPNEGO) auth without - GSS-API or SSPI */ -#endif - - /* check if any auth bit lower than CURLAUTH_ONLY is still set */ - bitcheck = 0; - authbits = FALSE; - while(bitcheck < 31) { - if(auth & (1UL << bitcheck++)) { - authbits = TRUE; - break; - } - } - if(!authbits) - return CURLE_NOT_BUILT_IN; /* no supported types left! */ - - data->set.httpauth = auth; - } - break; - - case CURLOPT_CUSTOMREQUEST: - /* - * Set a custom string to use as request - */ - result = setstropt(&data->set.str[STRING_CUSTOMREQUEST], - va_arg(param, char *)); - - /* we don't set - data->set.httpreq = HTTPREQ_CUSTOM; - here, we continue as if we were using the already set type - and this just changes the actual request keyword */ - break; - -#ifndef CURL_DISABLE_PROXY - case CURLOPT_HTTPPROXYTUNNEL: - /* - * Tunnel operations through the proxy instead of normal proxy use - */ - data->set.tunnel_thru_httpproxy = (0 != va_arg(param, long)) ? - TRUE : FALSE; - break; - - case CURLOPT_PROXYPORT: - /* - * Explicitly set HTTP proxy port number. - */ - data->set.proxyport = va_arg(param, long); - break; - - case CURLOPT_PROXYAUTH: - /* - * Set HTTP Authentication type BITMASK. - */ - { - int bitcheck; - bool authbits; - unsigned long auth = va_arg(param, unsigned long); - - if(auth == CURLAUTH_NONE) { - data->set.proxyauth = auth; - break; - } - - /* the DIGEST_IE bit is only used to set a special marker, for all the - rest we need to handle it as normal DIGEST */ - data->state.authproxy.iestyle = (auth & CURLAUTH_DIGEST_IE) ? TRUE : FALSE; - - if(auth & CURLAUTH_DIGEST_IE) { - auth |= CURLAUTH_DIGEST; /* set standard digest bit */ - auth &= ~CURLAUTH_DIGEST_IE; /* unset ie digest bit */ - } - /* switch off bits we can't support */ -#ifndef USE_NTLM - auth &= ~CURLAUTH_NTLM; /* no NTLM support */ - auth &= ~CURLAUTH_NTLM_WB; /* no NTLM_WB support */ -#elif !defined(NTLM_WB_ENABLED) - auth &= ~CURLAUTH_NTLM_WB; /* no NTLM_WB support */ -#endif -#ifndef USE_SPNEGO - auth &= ~CURLAUTH_NEGOTIATE; /* no Negotiate (SPNEGO) auth without - GSS-API or SSPI */ -#endif - - /* check if any auth bit lower than CURLAUTH_ONLY is still set */ - bitcheck = 0; - authbits = FALSE; - while(bitcheck < 31) { - if(auth & (1UL << bitcheck++)) { - authbits = TRUE; - break; - } - } - if(!authbits) - return CURLE_NOT_BUILT_IN; /* no supported types left! */ - - data->set.proxyauth = auth; - } - break; - - case CURLOPT_PROXY: - /* - * Set proxy server:port to use as proxy. - * - * If the proxy is set to "" (and CURLOPT_SOCKS_PROXY is set to "" or NULL) - * we explicitly say that we don't want to use a proxy - * (even though there might be environment variables saying so). - * - * Setting it to NULL, means no proxy but allows the environment variables - * to decide for us (if CURLOPT_SOCKS_PROXY setting it to NULL). - */ - result = setstropt(&data->set.str[STRING_PROXY], - va_arg(param, char *)); - break; - - case CURLOPT_PRE_PROXY: - /* - * Set proxy server:port to use as SOCKS proxy. - * - * If the proxy is set to "" or NULL we explicitly say that we don't want - * to use the socks proxy. - */ - result = setstropt(&data->set.str[STRING_PRE_PROXY], - va_arg(param, char *)); - break; - - case CURLOPT_PROXYTYPE: - /* - * Set proxy type. HTTP/HTTP_1_0/SOCKS4/SOCKS4a/SOCKS5/SOCKS5_HOSTNAME - */ - data->set.proxytype = (curl_proxytype)va_arg(param, long); - break; - - case CURLOPT_PROXY_TRANSFER_MODE: - /* - * set transfer mode (;type=<a|i>) when doing FTP via an HTTP proxy - */ - switch(va_arg(param, long)) { - case 0: - data->set.proxy_transfer_mode = FALSE; - break; - case 1: - data->set.proxy_transfer_mode = TRUE; - break; - default: - /* reserve other values for future use */ - result = CURLE_UNKNOWN_OPTION; - break; - } - break; -#endif /* CURL_DISABLE_PROXY */ - - case CURLOPT_SOCKS5_AUTH: - data->set.socks5auth = va_arg(param, unsigned long); - if(data->set.socks5auth & ~(CURLAUTH_BASIC | CURLAUTH_GSSAPI)) - result = CURLE_NOT_BUILT_IN; - break; -#if defined(HAVE_GSSAPI) || defined(USE_WINDOWS_SSPI) - case CURLOPT_SOCKS5_GSSAPI_NEC: - /* - * Set flag for NEC SOCK5 support - */ - data->set.socks5_gssapi_nec = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_SOCKS5_GSSAPI_SERVICE: - case CURLOPT_PROXY_SERVICE_NAME: - /* - * Set proxy authentication service name for Kerberos 5 and SPNEGO - */ - result = setstropt(&data->set.str[STRING_PROXY_SERVICE_NAME], - va_arg(param, char *)); - break; -#endif - -#if !defined(CURL_DISABLE_CRYPTO_AUTH) || defined(USE_KERBEROS5) || \ - defined(USE_SPNEGO) - case CURLOPT_SERVICE_NAME: - /* - * Set authentication service name for DIGEST-MD5, Kerberos 5 and SPNEGO - */ - result = setstropt(&data->set.str[STRING_SERVICE_NAME], - va_arg(param, char *)); - break; - -#endif - - case CURLOPT_HEADERDATA: - /* - * Custom pointer to pass the header write callback function - */ - data->set.writeheader = (void *)va_arg(param, void *); - break; - case CURLOPT_ERRORBUFFER: - /* - * Error buffer provided by the caller to get the human readable - * error string in. - */ - data->set.errorbuffer = va_arg(param, char *); - break; - case CURLOPT_WRITEDATA: - /* - * FILE pointer to write to. Or possibly - * used as argument to the write callback. - */ - data->set.out = va_arg(param, void *); - break; - case CURLOPT_FTPPORT: - /* - * Use FTP PORT, this also specifies which IP address to use - */ - result = setstropt(&data->set.str[STRING_FTPPORT], - va_arg(param, char *)); - data->set.ftp_use_port = (data->set.str[STRING_FTPPORT]) ? TRUE : FALSE; - break; - - case CURLOPT_FTP_USE_EPRT: - data->set.ftp_use_eprt = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_FTP_USE_EPSV: - data->set.ftp_use_epsv = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_FTP_USE_PRET: - data->set.ftp_use_pret = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_FTP_SSL_CCC: - data->set.ftp_ccc = (curl_ftpccc)va_arg(param, long); - break; - - case CURLOPT_FTP_SKIP_PASV_IP: - /* - * Enable or disable FTP_SKIP_PASV_IP, which will disable/enable the - * bypass of the IP address in PASV responses. - */ - data->set.ftp_skip_ip = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_READDATA: - /* - * FILE pointer to read the file to be uploaded from. Or possibly - * used as argument to the read callback. - */ - data->set.in_set = va_arg(param, void *); - break; - case CURLOPT_INFILESIZE: - /* - * If known, this should inform curl about the file size of the - * to-be-uploaded file. - */ - data->set.filesize = va_arg(param, long); - break; - case CURLOPT_INFILESIZE_LARGE: - /* - * If known, this should inform curl about the file size of the - * to-be-uploaded file. - */ - data->set.filesize = va_arg(param, curl_off_t); - break; - case CURLOPT_LOW_SPEED_LIMIT: - /* - * The low speed limit that if transfers are below this for - * CURLOPT_LOW_SPEED_TIME, the transfer is aborted. - */ - data->set.low_speed_limit = va_arg(param, long); - break; - case CURLOPT_MAX_SEND_SPEED_LARGE: - /* - * When transfer uploads are faster then CURLOPT_MAX_SEND_SPEED_LARGE - * bytes per second the transfer is throttled.. - */ - data->set.max_send_speed = va_arg(param, curl_off_t); - break; - case CURLOPT_MAX_RECV_SPEED_LARGE: - /* - * When receiving data faster than CURLOPT_MAX_RECV_SPEED_LARGE bytes per - * second the transfer is throttled.. - */ - data->set.max_recv_speed = va_arg(param, curl_off_t); - break; - case CURLOPT_LOW_SPEED_TIME: - /* - * The low speed time that if transfers are below the set - * CURLOPT_LOW_SPEED_LIMIT during this time, the transfer is aborted. - */ - data->set.low_speed_time = va_arg(param, long); - break; - case CURLOPT_URL: - /* - * The URL to fetch. - */ - if(data->change.url_alloc) { - /* the already set URL is allocated, free it first! */ - Curl_safefree(data->change.url); - data->change.url_alloc = FALSE; - } - result = setstropt(&data->set.str[STRING_SET_URL], - va_arg(param, char *)); - data->change.url = data->set.str[STRING_SET_URL]; - break; - case CURLOPT_PORT: - /* - * The port number to use when getting the URL - */ - data->set.use_port = va_arg(param, long); - break; - case CURLOPT_TIMEOUT: - /* - * The maximum time you allow curl to use for a single transfer - * operation. - */ - data->set.timeout = va_arg(param, long) * 1000L; - break; - - case CURLOPT_TIMEOUT_MS: - data->set.timeout = va_arg(param, long); - break; - - case CURLOPT_CONNECTTIMEOUT: - /* - * The maximum time you allow curl to use to connect. - */ - data->set.connecttimeout = va_arg(param, long) * 1000L; - break; - - case CURLOPT_CONNECTTIMEOUT_MS: - data->set.connecttimeout = va_arg(param, long); - break; - - case CURLOPT_ACCEPTTIMEOUT_MS: - /* - * The maximum time you allow curl to wait for server connect - */ - data->set.accepttimeout = va_arg(param, long); - break; - - case CURLOPT_USERPWD: - /* - * user:password to use in the operation - */ - result = setstropt_userpwd(va_arg(param, char *), - &data->set.str[STRING_USERNAME], - &data->set.str[STRING_PASSWORD]); - break; - - case CURLOPT_USERNAME: - /* - * authentication user name to use in the operation - */ - result = setstropt(&data->set.str[STRING_USERNAME], - va_arg(param, char *)); - break; - - case CURLOPT_PASSWORD: - /* - * authentication password to use in the operation - */ - result = setstropt(&data->set.str[STRING_PASSWORD], - va_arg(param, char *)); - break; - - case CURLOPT_LOGIN_OPTIONS: - /* - * authentication options to use in the operation - */ - result = setstropt(&data->set.str[STRING_OPTIONS], - va_arg(param, char *)); - break; - - case CURLOPT_XOAUTH2_BEARER: - /* - * OAuth 2.0 bearer token to use in the operation - */ - result = setstropt(&data->set.str[STRING_BEARER], - va_arg(param, char *)); - break; - - case CURLOPT_POSTQUOTE: - /* - * List of RAW FTP commands to use after a transfer - */ - data->set.postquote = va_arg(param, struct curl_slist *); - break; - case CURLOPT_PREQUOTE: - /* - * List of RAW FTP commands to use prior to RETR (Wesley Laxton) - */ - data->set.prequote = va_arg(param, struct curl_slist *); - break; - case CURLOPT_QUOTE: - /* - * List of RAW FTP commands to use before a transfer - */ - data->set.quote = va_arg(param, struct curl_slist *); - break; - case CURLOPT_RESOLVE: - /* - * List of NAME:[address] names to populate the DNS cache with - * Prefix the NAME with dash (-) to _remove_ the name from the cache. - * - * Names added with this API will remain in the cache until explicitly - * removed or the handle is cleaned up. - * - * This API can remove any name from the DNS cache, but only entries - * that aren't actually in use right now will be pruned immediately. - */ - data->set.resolve = va_arg(param, struct curl_slist *); - data->change.resolve = data->set.resolve; - break; - case CURLOPT_PROGRESSFUNCTION: - /* - * Progress callback function - */ - data->set.fprogress = va_arg(param, curl_progress_callback); - if(data->set.fprogress) - data->progress.callback = TRUE; /* no longer internal */ - else - data->progress.callback = FALSE; /* NULL enforces internal */ - break; - - case CURLOPT_XFERINFOFUNCTION: - /* - * Transfer info callback function - */ - data->set.fxferinfo = va_arg(param, curl_xferinfo_callback); - if(data->set.fxferinfo) - data->progress.callback = TRUE; /* no longer internal */ - else - data->progress.callback = FALSE; /* NULL enforces internal */ - - break; - - case CURLOPT_PROGRESSDATA: - /* - * Custom client data to pass to the progress callback - */ - data->set.progress_client = va_arg(param, void *); - break; - -#ifndef CURL_DISABLE_PROXY - case CURLOPT_PROXYUSERPWD: - /* - * user:password needed to use the proxy - */ - result = setstropt_userpwd(va_arg(param, char *), - &data->set.str[STRING_PROXYUSERNAME], - &data->set.str[STRING_PROXYPASSWORD]); - break; - case CURLOPT_PROXYUSERNAME: - /* - * authentication user name to use in the operation - */ - result = setstropt(&data->set.str[STRING_PROXYUSERNAME], - va_arg(param, char *)); - break; - case CURLOPT_PROXYPASSWORD: - /* - * authentication password to use in the operation - */ - result = setstropt(&data->set.str[STRING_PROXYPASSWORD], - va_arg(param, char *)); - break; - case CURLOPT_NOPROXY: - /* - * proxy exception list - */ - result = setstropt(&data->set.str[STRING_NOPROXY], - va_arg(param, char *)); - break; -#endif - - case CURLOPT_RANGE: - /* - * What range of the file you want to transfer - */ - result = setstropt(&data->set.str[STRING_SET_RANGE], - va_arg(param, char *)); - break; - case CURLOPT_RESUME_FROM: - /* - * Resume transfer at the given file position - */ - data->set.set_resume_from = va_arg(param, long); - break; - case CURLOPT_RESUME_FROM_LARGE: - /* - * Resume transfer at the given file position - */ - data->set.set_resume_from = va_arg(param, curl_off_t); - break; - case CURLOPT_DEBUGFUNCTION: - /* - * stderr write callback. - */ - data->set.fdebug = va_arg(param, curl_debug_callback); - /* - * if the callback provided is NULL, it'll use the default callback - */ - break; - case CURLOPT_DEBUGDATA: - /* - * Set to a void * that should receive all error writes. This - * defaults to CURLOPT_STDERR for normal operations. - */ - data->set.debugdata = va_arg(param, void *); - break; - case CURLOPT_STDERR: - /* - * Set to a FILE * that should receive all error writes. This - * defaults to stderr for normal operations. - */ - data->set.err = va_arg(param, FILE *); - if(!data->set.err) - data->set.err = stderr; - break; - case CURLOPT_HEADERFUNCTION: - /* - * Set header write callback - */ - data->set.fwrite_header = va_arg(param, curl_write_callback); - break; - case CURLOPT_WRITEFUNCTION: - /* - * Set data write callback - */ - data->set.fwrite_func = va_arg(param, curl_write_callback); - if(!data->set.fwrite_func) { - data->set.is_fwrite_set = 0; - /* When set to NULL, reset to our internal default function */ - data->set.fwrite_func = (curl_write_callback)fwrite; - } - else - data->set.is_fwrite_set = 1; - break; - case CURLOPT_READFUNCTION: - /* - * Read data callback - */ - data->set.fread_func_set = va_arg(param, curl_read_callback); - if(!data->set.fread_func_set) { - data->set.is_fread_set = 0; - /* When set to NULL, reset to our internal default function */ - data->set.fread_func_set = (curl_read_callback)fread; - } - else - data->set.is_fread_set = 1; - break; - case CURLOPT_SEEKFUNCTION: - /* - * Seek callback. Might be NULL. - */ - data->set.seek_func = va_arg(param, curl_seek_callback); - break; - case CURLOPT_SEEKDATA: - /* - * Seek control callback. Might be NULL. - */ - data->set.seek_client = va_arg(param, void *); - break; - case CURLOPT_CONV_FROM_NETWORK_FUNCTION: - /* - * "Convert from network encoding" callback - */ - data->set.convfromnetwork = va_arg(param, curl_conv_callback); - break; - case CURLOPT_CONV_TO_NETWORK_FUNCTION: - /* - * "Convert to network encoding" callback - */ - data->set.convtonetwork = va_arg(param, curl_conv_callback); - break; - case CURLOPT_CONV_FROM_UTF8_FUNCTION: - /* - * "Convert from UTF-8 encoding" callback - */ - data->set.convfromutf8 = va_arg(param, curl_conv_callback); - break; - case CURLOPT_IOCTLFUNCTION: - /* - * I/O control callback. Might be NULL. - */ - data->set.ioctl_func = va_arg(param, curl_ioctl_callback); - break; - case CURLOPT_IOCTLDATA: - /* - * I/O control data pointer. Might be NULL. - */ - data->set.ioctl_client = va_arg(param, void *); - break; - case CURLOPT_SSLCERT: - /* - * String that holds file name of the SSL certificate to use - */ - result = setstropt(&data->set.str[STRING_CERT_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_SSLCERT: - /* - * String that holds file name of the SSL certificate to use for proxy - */ - result = setstropt(&data->set.str[STRING_CERT_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_SSLCERTTYPE: - /* - * String that holds file type of the SSL certificate to use - */ - result = setstropt(&data->set.str[STRING_CERT_TYPE_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_SSLCERTTYPE: - /* - * String that holds file type of the SSL certificate to use for proxy - */ - result = setstropt(&data->set.str[STRING_CERT_TYPE_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_SSLKEY: - /* - * String that holds file name of the SSL key to use - */ - result = setstropt(&data->set.str[STRING_KEY_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_SSLKEY: - /* - * String that holds file name of the SSL key to use for proxy - */ - result = setstropt(&data->set.str[STRING_KEY_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_SSLKEYTYPE: - /* - * String that holds file type of the SSL key to use - */ - result = setstropt(&data->set.str[STRING_KEY_TYPE_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_SSLKEYTYPE: - /* - * String that holds file type of the SSL key to use for proxy - */ - result = setstropt(&data->set.str[STRING_KEY_TYPE_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_KEYPASSWD: - /* - * String that holds the SSL or SSH private key password. - */ - result = setstropt(&data->set.str[STRING_KEY_PASSWD_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_KEYPASSWD: - /* - * String that holds the SSL private key password for proxy. - */ - result = setstropt(&data->set.str[STRING_KEY_PASSWD_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_SSLENGINE: - /* - * String that holds the SSL crypto engine. - */ - argptr = va_arg(param, char *); - if(argptr && argptr[0]) - result = Curl_ssl_set_engine(data, argptr); - break; - - case CURLOPT_SSLENGINE_DEFAULT: - /* - * flag to set engine as default. - */ - result = Curl_ssl_set_engine_default(data); - break; - case CURLOPT_CRLF: - /* - * Kludgy option to enable CRLF conversions. Subject for removal. - */ - data->set.crlf = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_INTERFACE: - /* - * Set what interface or address/hostname to bind the socket to when - * performing an operation and thus what from-IP your connection will use. - */ - result = setstropt(&data->set.str[STRING_DEVICE], - va_arg(param, char *)); - break; - case CURLOPT_LOCALPORT: - /* - * Set what local port to bind the socket to when performing an operation. - */ - arg = va_arg(param, long); - if((arg < 0) || (arg > 65535)) - return CURLE_BAD_FUNCTION_ARGUMENT; - data->set.localport = curlx_sltous(arg); - break; - case CURLOPT_LOCALPORTRANGE: - /* - * Set number of local ports to try, starting with CURLOPT_LOCALPORT. - */ - arg = va_arg(param, long); - if((arg < 0) || (arg > 65535)) - return CURLE_BAD_FUNCTION_ARGUMENT; - data->set.localportrange = curlx_sltosi(arg); - break; - case CURLOPT_KRBLEVEL: - /* - * A string that defines the kerberos security level. - */ - result = setstropt(&data->set.str[STRING_KRB_LEVEL], - va_arg(param, char *)); - data->set.krb = (data->set.str[STRING_KRB_LEVEL]) ? TRUE : FALSE; - break; - case CURLOPT_GSSAPI_DELEGATION: - /* - * GSS-API credential delegation - */ - data->set.gssapi_delegation = va_arg(param, long); - break; - case CURLOPT_SSL_VERIFYPEER: - /* - * Enable peer SSL verifying. - */ - data->set.ssl.primary.verifypeer = (0 != va_arg(param, long)) ? - TRUE : FALSE; - break; - case CURLOPT_PROXY_SSL_VERIFYPEER: - /* - * Enable peer SSL verifying for proxy. - */ - data->set.proxy_ssl.primary.verifypeer = - (0 != va_arg(param, long))?TRUE:FALSE; - break; - case CURLOPT_SSL_VERIFYHOST: - /* - * Enable verification of the host name in the peer certificate - */ - arg = va_arg(param, long); - - /* Obviously people are not reading documentation and too many thought - this argument took a boolean when it wasn't and misused it. We thus ban - 1 as a sensible input and we warn about its use. Then we only have the - 2 action internally stored as TRUE. */ - - if(1 == arg) { - failf(data, "CURLOPT_SSL_VERIFYHOST no longer supports 1 as value!"); - return CURLE_BAD_FUNCTION_ARGUMENT; - } - - data->set.ssl.primary.verifyhost = (0 != arg) ? TRUE : FALSE; - break; - case CURLOPT_PROXY_SSL_VERIFYHOST: - /* - * Enable verification of the host name in the peer certificate for proxy - */ - arg = va_arg(param, long); - - /* Obviously people are not reading documentation and too many thought - this argument took a boolean when it wasn't and misused it. We thus ban - 1 as a sensible input and we warn about its use. Then we only have the - 2 action internally stored as TRUE. */ - - if(1 == arg) { - failf(data, "CURLOPT_SSL_VERIFYHOST no longer supports 1 as value!"); - return CURLE_BAD_FUNCTION_ARGUMENT; - } - - data->set.proxy_ssl.primary.verifyhost = (0 != arg)?TRUE:FALSE; - break; - case CURLOPT_SSL_VERIFYSTATUS: - /* - * Enable certificate status verifying. - */ - if(!Curl_ssl_cert_status_request()) { - result = CURLE_NOT_BUILT_IN; - break; - } - - data->set.ssl.primary.verifystatus = (0 != va_arg(param, long)) ? - TRUE : FALSE; - break; - case CURLOPT_SSL_CTX_FUNCTION: - /* - * Set a SSL_CTX callback - */ -#ifdef USE_SSL - if(Curl_ssl->have_ssl_ctx) - data->set.ssl.fsslctx = va_arg(param, curl_ssl_ctx_callback); - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_SSL_CTX_DATA: - /* - * Set a SSL_CTX callback parameter pointer - */ -#ifdef USE_SSL - if(Curl_ssl->have_ssl_ctx) - data->set.ssl.fsslctxp = va_arg(param, void *); - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_SSL_FALSESTART: - /* - * Enable TLS false start. - */ - if(!Curl_ssl_false_start()) { - result = CURLE_NOT_BUILT_IN; - break; - } - - data->set.ssl.falsestart = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_CERTINFO: -#ifdef USE_SSL - if(Curl_ssl->have_certinfo) - data->set.ssl.certinfo = (0 != va_arg(param, long)) ? TRUE : FALSE; - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_PINNEDPUBLICKEY: - /* - * Set pinned public key for SSL connection. - * Specify file name of the public key in DER format. - */ -#ifdef USE_SSL - if(Curl_ssl->have_pinnedpubkey) - result = setstropt(&data->set.str[STRING_SSL_PINNEDPUBLICKEY_ORIG], - va_arg(param, char *)); - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_PROXY_PINNEDPUBLICKEY: - /* - * Set pinned public key for SSL connection. - * Specify file name of the public key in DER format. - */ -#ifdef USE_SSL - if(Curl_ssl->have_pinnedpubkey) - result = setstropt(&data->set.str[STRING_SSL_PINNEDPUBLICKEY_PROXY], - va_arg(param, char *)); - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_CAINFO: - /* - * Set CA info for SSL connection. Specify file name of the CA certificate - */ - result = setstropt(&data->set.str[STRING_SSL_CAFILE_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_CAINFO: - /* - * Set CA info SSL connection for proxy. Specify file name of the - * CA certificate - */ - result = setstropt(&data->set.str[STRING_SSL_CAFILE_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_CAPATH: - /* - * Set CA path info for SSL connection. Specify directory name of the CA - * certificates which have been prepared using openssl c_rehash utility. - */ -#ifdef USE_SSL - if(Curl_ssl->have_ca_path) - /* This does not work on windows. */ - result = setstropt(&data->set.str[STRING_SSL_CAPATH_ORIG], - va_arg(param, char *)); - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_PROXY_CAPATH: - /* - * Set CA path info for SSL connection proxy. Specify directory name of the - * CA certificates which have been prepared using openssl c_rehash utility. - */ -#ifdef USE_SSL - if(Curl_ssl->have_ca_path) - /* This does not work on windows. */ - result = setstropt(&data->set.str[STRING_SSL_CAPATH_PROXY], - va_arg(param, char *)); - else -#endif - result = CURLE_NOT_BUILT_IN; - break; - case CURLOPT_CRLFILE: - /* - * Set CRL file info for SSL connection. Specify file name of the CRL - * to check certificates revocation - */ - result = setstropt(&data->set.str[STRING_SSL_CRLFILE_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_PROXY_CRLFILE: - /* - * Set CRL file info for SSL connection for proxy. Specify file name of the - * CRL to check certificates revocation - */ - result = setstropt(&data->set.str[STRING_SSL_CRLFILE_PROXY], - va_arg(param, char *)); - break; - case CURLOPT_ISSUERCERT: - /* - * Set Issuer certificate file - * to check certificates issuer - */ - result = setstropt(&data->set.str[STRING_SSL_ISSUERCERT_ORIG], - va_arg(param, char *)); - break; - case CURLOPT_TELNETOPTIONS: - /* - * Set a linked list of telnet options - */ - data->set.telnet_options = va_arg(param, struct curl_slist *); - break; - - case CURLOPT_BUFFERSIZE: - /* - * The application kindly asks for a differently sized receive buffer. - * If it seems reasonable, we'll use it. - */ - arg = va_arg(param, long); - - if(arg > READBUFFER_MAX) - arg = READBUFFER_MAX; - else if(arg < 1) - arg = READBUFFER_SIZE; - else if(arg < READBUFFER_MIN) - arg = READBUFFER_MIN; - - /* Resize if new size */ - if(arg != data->set.buffer_size) { - char *newbuff = realloc(data->state.buffer, arg + 1); - if(!newbuff) { - DEBUGF(fprintf(stderr, "Error: realloc of buffer failed\n")); - result = CURLE_OUT_OF_MEMORY; - } - else - data->state.buffer = newbuff; - } - data->set.buffer_size = arg; - - break; - - case CURLOPT_NOSIGNAL: - /* - * The application asks not to set any signal() or alarm() handlers, - * even when using a timeout. - */ - data->set.no_signal = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_SHARE: - { - struct Curl_share *set; - set = va_arg(param, struct Curl_share *); - - /* disconnect from old share, if any */ - if(data->share) { - Curl_share_lock(data, CURL_LOCK_DATA_SHARE, CURL_LOCK_ACCESS_SINGLE); - - if(data->dns.hostcachetype == HCACHE_SHARED) { - data->dns.hostcache = NULL; - data->dns.hostcachetype = HCACHE_NONE; - } - -#if !defined(CURL_DISABLE_HTTP) && !defined(CURL_DISABLE_COOKIES) - if(data->share->cookies == data->cookies) - data->cookies = NULL; -#endif - - if(data->share->sslsession == data->state.session) - data->state.session = NULL; - - data->share->dirty--; - - Curl_share_unlock(data, CURL_LOCK_DATA_SHARE); - data->share = NULL; - } - - /* use new share if it set */ - data->share = set; - if(data->share) { - - Curl_share_lock(data, CURL_LOCK_DATA_SHARE, CURL_LOCK_ACCESS_SINGLE); - - data->share->dirty++; - - if(data->share->specifier & (1<< CURL_LOCK_DATA_DNS)) { - /* use shared host cache */ - data->dns.hostcache = &data->share->hostcache; - data->dns.hostcachetype = HCACHE_SHARED; - } -#if !defined(CURL_DISABLE_HTTP) && !defined(CURL_DISABLE_COOKIES) - if(data->share->cookies) { - /* use shared cookie list, first free own one if any */ - Curl_cookie_cleanup(data->cookies); - /* enable cookies since we now use a share that uses cookies! */ - data->cookies = data->share->cookies; - } -#endif /* CURL_DISABLE_HTTP */ - if(data->share->sslsession) { - data->set.general_ssl.max_ssl_sessions = data->share->max_ssl_sessions; - data->state.session = data->share->sslsession; - } - Curl_share_unlock(data, CURL_LOCK_DATA_SHARE); - - } - /* check for host cache not needed, - * it will be done by curl_easy_perform */ - } - break; - - case CURLOPT_PRIVATE: - /* - * Set private data pointer. - */ - data->set.private_data = va_arg(param, void *); - break; - - case CURLOPT_MAXFILESIZE: - /* - * Set the maximum size of a file to download. - */ - data->set.max_filesize = va_arg(param, long); - break; - -#ifdef USE_SSL - case CURLOPT_USE_SSL: - /* - * Make transfers attempt to use SSL/TLS. - */ - data->set.use_ssl = (curl_usessl)va_arg(param, long); - break; - - case CURLOPT_SSL_OPTIONS: - arg = va_arg(param, long); - data->set.ssl.enable_beast = arg&CURLSSLOPT_ALLOW_BEAST?TRUE:FALSE; - data->set.ssl.no_revoke = !!(arg & CURLSSLOPT_NO_REVOKE); - break; - - case CURLOPT_PROXY_SSL_OPTIONS: - arg = va_arg(param, long); - data->set.proxy_ssl.enable_beast = arg&CURLSSLOPT_ALLOW_BEAST?TRUE:FALSE; - data->set.proxy_ssl.no_revoke = !!(arg & CURLSSLOPT_NO_REVOKE); - break; - -#endif - case CURLOPT_FTPSSLAUTH: - /* - * Set a specific auth for FTP-SSL transfers. - */ - data->set.ftpsslauth = (curl_ftpauth)va_arg(param, long); - break; - - case CURLOPT_IPRESOLVE: - data->set.ipver = va_arg(param, long); - break; - - case CURLOPT_MAXFILESIZE_LARGE: - /* - * Set the maximum size of a file to download. - */ - data->set.max_filesize = va_arg(param, curl_off_t); - break; - - case CURLOPT_TCP_NODELAY: - /* - * Enable or disable TCP_NODELAY, which will disable/enable the Nagle - * algorithm - */ - data->set.tcp_nodelay = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_FTP_ACCOUNT: - result = setstropt(&data->set.str[STRING_FTP_ACCOUNT], - va_arg(param, char *)); - break; - - case CURLOPT_IGNORE_CONTENT_LENGTH: - data->set.ignorecl = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_CONNECT_ONLY: - /* - * No data transfer, set up connection and let application use the socket - */ - data->set.connect_only = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_FTP_ALTERNATIVE_TO_USER: - result = setstropt(&data->set.str[STRING_FTP_ALTERNATIVE_TO_USER], - va_arg(param, char *)); - break; - - case CURLOPT_SOCKOPTFUNCTION: - /* - * socket callback function: called after socket() but before connect() - */ - data->set.fsockopt = va_arg(param, curl_sockopt_callback); - break; - - case CURLOPT_SOCKOPTDATA: - /* - * socket callback data pointer. Might be NULL. - */ - data->set.sockopt_client = va_arg(param, void *); - break; - - case CURLOPT_OPENSOCKETFUNCTION: - /* - * open/create socket callback function: called instead of socket(), - * before connect() - */ - data->set.fopensocket = va_arg(param, curl_opensocket_callback); - break; - - case CURLOPT_OPENSOCKETDATA: - /* - * socket callback data pointer. Might be NULL. - */ - data->set.opensocket_client = va_arg(param, void *); - break; - - case CURLOPT_CLOSESOCKETFUNCTION: - /* - * close socket callback function: called instead of close() - * when shutting down a connection - */ - data->set.fclosesocket = va_arg(param, curl_closesocket_callback); - break; - - case CURLOPT_CLOSESOCKETDATA: - /* - * socket callback data pointer. Might be NULL. - */ - data->set.closesocket_client = va_arg(param, void *); - break; - - case CURLOPT_SSL_SESSIONID_CACHE: - data->set.ssl.primary.sessionid = (0 != va_arg(param, long)) ? - TRUE : FALSE; - data->set.proxy_ssl.primary.sessionid = data->set.ssl.primary.sessionid; - break; - -#ifdef USE_LIBSSH2 - /* we only include SSH options if explicitly built to support SSH */ - case CURLOPT_SSH_AUTH_TYPES: - data->set.ssh_auth_types = va_arg(param, long); - break; - - case CURLOPT_SSH_PUBLIC_KEYFILE: - /* - * Use this file instead of the $HOME/.ssh/id_dsa.pub file - */ - result = setstropt(&data->set.str[STRING_SSH_PUBLIC_KEY], - va_arg(param, char *)); - break; - - case CURLOPT_SSH_PRIVATE_KEYFILE: - /* - * Use this file instead of the $HOME/.ssh/id_dsa file - */ - result = setstropt(&data->set.str[STRING_SSH_PRIVATE_KEY], - va_arg(param, char *)); - break; - case CURLOPT_SSH_HOST_PUBLIC_KEY_MD5: - /* - * Option to allow for the MD5 of the host public key to be checked - * for validation purposes. - */ - result = setstropt(&data->set.str[STRING_SSH_HOST_PUBLIC_KEY_MD5], - va_arg(param, char *)); - break; -#ifdef HAVE_LIBSSH2_KNOWNHOST_API - case CURLOPT_SSH_KNOWNHOSTS: - /* - * Store the file name to read known hosts from. - */ - result = setstropt(&data->set.str[STRING_SSH_KNOWNHOSTS], - va_arg(param, char *)); - break; - - case CURLOPT_SSH_KEYFUNCTION: - /* setting to NULL is fine since the ssh.c functions themselves will - then rever to use the internal default */ - data->set.ssh_keyfunc = va_arg(param, curl_sshkeycallback); - break; - - case CURLOPT_SSH_KEYDATA: - /* - * Custom client data to pass to the SSH keyfunc callback - */ - data->set.ssh_keyfunc_userp = va_arg(param, void *); - break; -#endif /* HAVE_LIBSSH2_KNOWNHOST_API */ - -#endif /* USE_LIBSSH2 */ - - case CURLOPT_HTTP_TRANSFER_DECODING: - /* - * disable libcurl transfer encoding is used - */ - data->set.http_te_skip = (0 == va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_HTTP_CONTENT_DECODING: - /* - * raw data passed to the application when content encoding is used - */ - data->set.http_ce_skip = (0 == va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_NEW_FILE_PERMS: - /* - * Uses these permissions instead of 0644 - */ - data->set.new_file_perms = va_arg(param, long); - break; - - case CURLOPT_NEW_DIRECTORY_PERMS: - /* - * Uses these permissions instead of 0755 - */ - data->set.new_directory_perms = va_arg(param, long); - break; - - case CURLOPT_ADDRESS_SCOPE: - /* - * We always get longs when passed plain numericals, but for this value we - * know that an unsigned int will always hold the value so we blindly - * typecast to this type - */ - arg = va_arg(param, long); - if((arg < 0) || (arg > 0xf)) - return CURLE_BAD_FUNCTION_ARGUMENT; - data->set.scope_id = curlx_sltoui(arg); - break; - - case CURLOPT_PROTOCOLS: - /* set the bitmask for the protocols that are allowed to be used for the - transfer, which thus helps the app which takes URLs from users or other - external inputs and want to restrict what protocol(s) to deal - with. Defaults to CURLPROTO_ALL. */ - data->set.allowed_protocols = va_arg(param, long); - break; - - case CURLOPT_REDIR_PROTOCOLS: - /* set the bitmask for the protocols that libcurl is allowed to follow to, - as a subset of the CURLOPT_PROTOCOLS ones. That means the protocol needs - to be set in both bitmasks to be allowed to get redirected to. Defaults - to all protocols except FILE and SCP. */ - data->set.redir_protocols = va_arg(param, long); - break; - - case CURLOPT_DEFAULT_PROTOCOL: - /* Set the protocol to use when the URL doesn't include any protocol */ - result = setstropt(&data->set.str[STRING_DEFAULT_PROTOCOL], - va_arg(param, char *)); - break; - - case CURLOPT_MAIL_FROM: - /* Set the SMTP mail originator */ - result = setstropt(&data->set.str[STRING_MAIL_FROM], - va_arg(param, char *)); - break; - - case CURLOPT_MAIL_AUTH: - /* Set the SMTP auth originator */ - result = setstropt(&data->set.str[STRING_MAIL_AUTH], - va_arg(param, char *)); - break; - - case CURLOPT_MAIL_RCPT: - /* Set the list of mail recipients */ - data->set.mail_rcpt = va_arg(param, struct curl_slist *); - break; - - case CURLOPT_SASL_IR: - /* Enable/disable SASL initial response */ - data->set.sasl_ir = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - - case CURLOPT_RTSP_REQUEST: - { - /* - * Set the RTSP request method (OPTIONS, SETUP, PLAY, etc...) - * Would this be better if the RTSPREQ_* were just moved into here? - */ - long curl_rtspreq = va_arg(param, long); - Curl_RtspReq rtspreq = RTSPREQ_NONE; - switch(curl_rtspreq) { - case CURL_RTSPREQ_OPTIONS: - rtspreq = RTSPREQ_OPTIONS; - break; - - case CURL_RTSPREQ_DESCRIBE: - rtspreq = RTSPREQ_DESCRIBE; - break; - - case CURL_RTSPREQ_ANNOUNCE: - rtspreq = RTSPREQ_ANNOUNCE; - break; - - case CURL_RTSPREQ_SETUP: - rtspreq = RTSPREQ_SETUP; - break; - - case CURL_RTSPREQ_PLAY: - rtspreq = RTSPREQ_PLAY; - break; - - case CURL_RTSPREQ_PAUSE: - rtspreq = RTSPREQ_PAUSE; - break; - - case CURL_RTSPREQ_TEARDOWN: - rtspreq = RTSPREQ_TEARDOWN; - break; - - case CURL_RTSPREQ_GET_PARAMETER: - rtspreq = RTSPREQ_GET_PARAMETER; - break; - - case CURL_RTSPREQ_SET_PARAMETER: - rtspreq = RTSPREQ_SET_PARAMETER; - break; - - case CURL_RTSPREQ_RECORD: - rtspreq = RTSPREQ_RECORD; - break; - - case CURL_RTSPREQ_RECEIVE: - rtspreq = RTSPREQ_RECEIVE; - break; - default: - rtspreq = RTSPREQ_NONE; - } - - data->set.rtspreq = rtspreq; - break; - } - - - case CURLOPT_RTSP_SESSION_ID: - /* - * Set the RTSP Session ID manually. Useful if the application is - * resuming a previously established RTSP session - */ - result = setstropt(&data->set.str[STRING_RTSP_SESSION_ID], - va_arg(param, char *)); - break; - - case CURLOPT_RTSP_STREAM_URI: - /* - * Set the Stream URI for the RTSP request. Unless the request is - * for generic server options, the application will need to set this. - */ - result = setstropt(&data->set.str[STRING_RTSP_STREAM_URI], - va_arg(param, char *)); - break; - - case CURLOPT_RTSP_TRANSPORT: - /* - * The content of the Transport: header for the RTSP request - */ - result = setstropt(&data->set.str[STRING_RTSP_TRANSPORT], - va_arg(param, char *)); - break; - - case CURLOPT_RTSP_CLIENT_CSEQ: - /* - * Set the CSEQ number to issue for the next RTSP request. Useful if the - * application is resuming a previously broken connection. The CSEQ - * will increment from this new number henceforth. - */ - data->state.rtsp_next_client_CSeq = va_arg(param, long); - break; - - case CURLOPT_RTSP_SERVER_CSEQ: - /* Same as the above, but for server-initiated requests */ - data->state.rtsp_next_client_CSeq = va_arg(param, long); - break; - - case CURLOPT_INTERLEAVEDATA: - data->set.rtp_out = va_arg(param, void *); - break; - case CURLOPT_INTERLEAVEFUNCTION: - /* Set the user defined RTP write function */ - data->set.fwrite_rtp = va_arg(param, curl_write_callback); - break; - - case CURLOPT_WILDCARDMATCH: - data->set.wildcardmatch = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_CHUNK_BGN_FUNCTION: - data->set.chunk_bgn = va_arg(param, curl_chunk_bgn_callback); - break; - case CURLOPT_CHUNK_END_FUNCTION: - data->set.chunk_end = va_arg(param, curl_chunk_end_callback); - break; - case CURLOPT_FNMATCH_FUNCTION: - data->set.fnmatch = va_arg(param, curl_fnmatch_callback); - break; - case CURLOPT_CHUNK_DATA: - data->wildcard.customptr = va_arg(param, void *); - break; - case CURLOPT_FNMATCH_DATA: - data->set.fnmatch_data = va_arg(param, void *); - break; -#ifdef USE_TLS_SRP - case CURLOPT_TLSAUTH_USERNAME: - result = setstropt(&data->set.str[STRING_TLSAUTH_USERNAME_ORIG], - va_arg(param, char *)); - if(data->set.str[STRING_TLSAUTH_USERNAME_ORIG] && !data->set.ssl.authtype) - data->set.ssl.authtype = CURL_TLSAUTH_SRP; /* default to SRP */ - break; - case CURLOPT_PROXY_TLSAUTH_USERNAME: - result = setstropt(&data->set.str[STRING_TLSAUTH_USERNAME_PROXY], - va_arg(param, char *)); - if(data->set.str[STRING_TLSAUTH_USERNAME_PROXY] && - !data->set.proxy_ssl.authtype) - data->set.proxy_ssl.authtype = CURL_TLSAUTH_SRP; /* default to SRP */ - break; - case CURLOPT_TLSAUTH_PASSWORD: - result = setstropt(&data->set.str[STRING_TLSAUTH_PASSWORD_ORIG], - va_arg(param, char *)); - if(data->set.str[STRING_TLSAUTH_USERNAME_ORIG] && !data->set.ssl.authtype) - data->set.ssl.authtype = CURL_TLSAUTH_SRP; /* default to SRP */ - break; - case CURLOPT_PROXY_TLSAUTH_PASSWORD: - result = setstropt(&data->set.str[STRING_TLSAUTH_PASSWORD_PROXY], - va_arg(param, char *)); - if(data->set.str[STRING_TLSAUTH_USERNAME_PROXY] && - !data->set.proxy_ssl.authtype) - data->set.proxy_ssl.authtype = CURL_TLSAUTH_SRP; /* default to SRP */ - break; - case CURLOPT_TLSAUTH_TYPE: - argptr = va_arg(param, char *); - if(!argptr || - strncasecompare(argptr, "SRP", strlen("SRP"))) - data->set.ssl.authtype = CURL_TLSAUTH_SRP; - else - data->set.ssl.authtype = CURL_TLSAUTH_NONE; - break; - case CURLOPT_PROXY_TLSAUTH_TYPE: - argptr = va_arg(param, char *); - if(!argptr || - strncasecompare(argptr, "SRP", strlen("SRP"))) - data->set.proxy_ssl.authtype = CURL_TLSAUTH_SRP; - else - data->set.proxy_ssl.authtype = CURL_TLSAUTH_NONE; - break; -#endif - case CURLOPT_DNS_SERVERS: - result = Curl_set_dns_servers(data, va_arg(param, char *)); - break; - case CURLOPT_DNS_INTERFACE: - result = Curl_set_dns_interface(data, va_arg(param, char *)); - break; - case CURLOPT_DNS_LOCAL_IP4: - result = Curl_set_dns_local_ip4(data, va_arg(param, char *)); - break; - case CURLOPT_DNS_LOCAL_IP6: - result = Curl_set_dns_local_ip6(data, va_arg(param, char *)); - break; - - case CURLOPT_TCP_KEEPALIVE: - data->set.tcp_keepalive = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_TCP_KEEPIDLE: - data->set.tcp_keepidle = va_arg(param, long); - break; - case CURLOPT_TCP_KEEPINTVL: - data->set.tcp_keepintvl = va_arg(param, long); - break; - case CURLOPT_TCP_FASTOPEN: -#if defined(CONNECT_DATA_IDEMPOTENT) || defined(MSG_FASTOPEN) - data->set.tcp_fastopen = (0 != va_arg(param, long))?TRUE:FALSE; -#else - result = CURLE_NOT_BUILT_IN; -#endif - break; - case CURLOPT_SSL_ENABLE_NPN: - data->set.ssl_enable_npn = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_SSL_ENABLE_ALPN: - data->set.ssl_enable_alpn = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - -#ifdef USE_UNIX_SOCKETS - case CURLOPT_UNIX_SOCKET_PATH: - data->set.abstract_unix_socket = FALSE; - result = setstropt(&data->set.str[STRING_UNIX_SOCKET_PATH], - va_arg(param, char *)); - break; - case CURLOPT_ABSTRACT_UNIX_SOCKET: - data->set.abstract_unix_socket = TRUE; - result = setstropt(&data->set.str[STRING_UNIX_SOCKET_PATH], - va_arg(param, char *)); - break; -#endif - - case CURLOPT_PATH_AS_IS: - data->set.path_as_is = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_PIPEWAIT: - data->set.pipewait = (0 != va_arg(param, long)) ? TRUE : FALSE; - break; - case CURLOPT_STREAM_WEIGHT: -#ifndef USE_NGHTTP2 - return CURLE_NOT_BUILT_IN; -#else - arg = va_arg(param, long); - if((arg >= 1) && (arg <= 256)) - data->set.stream_weight = (int)arg; - break; -#endif - case CURLOPT_STREAM_DEPENDS: - case CURLOPT_STREAM_DEPENDS_E: - { -#ifndef USE_NGHTTP2 - return CURLE_NOT_BUILT_IN; -#else - struct Curl_easy *dep = va_arg(param, struct Curl_easy *); - if(!dep || GOOD_EASY_HANDLE(dep)) { - if(data->set.stream_depends_on) { - Curl_http2_remove_child(data->set.stream_depends_on, data); - } - Curl_http2_add_child(dep, data, (option == CURLOPT_STREAM_DEPENDS_E)); - } - break; -#endif - } - case CURLOPT_CONNECT_TO: - data->set.connect_to = va_arg(param, struct curl_slist *); - break; - case CURLOPT_SUPPRESS_CONNECT_HEADERS: - data->set.suppress_connect_headers = (0 != va_arg(param, long))?TRUE:FALSE; - break; - case CURLOPT_SSH_COMPRESSION: - data->set.ssh_compression = (0 != va_arg(param, long))?TRUE:FALSE; - break; - default: - /* unknown tag and its companion, just ignore: */ - result = CURLE_UNKNOWN_OPTION; - break; - } - - return result; -} - #ifdef USE_RECV_BEFORE_SEND_WORKAROUND static void conn_reset_postponed_data(struct connectdata *conn, int num) { @@ -3081,6 +714,9 @@ static void conn_free(struct connectdata *conn) Curl_safefree(conn->unix_domain_socket); #endif +#ifdef USE_SSL + Curl_safefree(conn->ssl_extra); +#endif free(conn); /* free all the connection oriented data */ } @@ -3135,7 +771,7 @@ CURLcode Curl_disconnect(struct connectdata *conn, bool dead_connection) /* unlink ourselves! */ infof(data, "Closing connection %ld\n", conn->connection_id); - Curl_conncache_remove_conn(data->state.conn_cache, conn); + Curl_conncache_remove_conn(conn, TRUE); free_fixed_hostname(&conn->host); free_fixed_hostname(&conn->conn_to_host); @@ -3289,58 +925,6 @@ static void signalPipeClose(struct curl_llist *pipeline, bool pipe_broke) } } -/* - * This function finds the connection in the connection - * cache that has been unused for the longest time. - * - * Returns the pointer to the oldest idle connection, or NULL if none was - * found. - */ -struct connectdata * -Curl_oldest_idle_connection(struct Curl_easy *data) -{ - struct conncache *bc = data->state.conn_cache; - struct curl_hash_iterator iter; - struct curl_llist_element *curr; - struct curl_hash_element *he; - time_t highscore =- 1; - time_t score; - struct curltime now; - struct connectdata *conn_candidate = NULL; - struct connectbundle *bundle; - - now = Curl_tvnow(); - - Curl_hash_start_iterate(&bc->hash, &iter); - - he = Curl_hash_next_element(&iter); - while(he) { - struct connectdata *conn; - - bundle = he->ptr; - - curr = bundle->conn_list.head; - while(curr) { - conn = curr->ptr; - - if(!conn->inuse) { - /* Set higher score for the age passed since the connection was used */ - score = Curl_tvdiff(now, conn->now); - - if(score > highscore) { - highscore = score; - conn_candidate = conn; - } - } - curr = curr->next; - } - - he = Curl_hash_next_element(&iter); - } - - return conn_candidate; -} - static bool proxy_info_matches(const struct proxy_info* data, const struct proxy_info* needle) @@ -3353,56 +937,17 @@ proxy_info_matches(const struct proxy_info* data, return FALSE; } - /* - * This function finds the connection in the connection - * bundle that has been unused for the longest time. + * This function checks if the given connection is dead and extracts it from + * the connection cache if so. * - * Returns the pointer to the oldest idle connection, or NULL if none was - * found. - */ -static struct connectdata * -find_oldest_idle_connection_in_bundle(struct Curl_easy *data, - struct connectbundle *bundle) -{ - struct curl_llist_element *curr; - time_t highscore = -1; - time_t score; - struct curltime now; - struct connectdata *conn_candidate = NULL; - struct connectdata *conn; - - (void)data; - - now = Curl_tvnow(); - - curr = bundle->conn_list.head; - while(curr) { - conn = curr->ptr; - - if(!conn->inuse) { - /* Set higher score for the age passed since the connection was used */ - score = Curl_tvdiff(now, conn->now); - - if(score > highscore) { - highscore = score; - conn_candidate = conn; - } - } - curr = curr->next; - } - - return conn_candidate; -} - -/* - * This function checks if given connection is dead and disconnects if so. - * (That also removes it from the connection cache.) + * When this is called as a Curl_conncache_foreach() callback, the connection + * cache lock is held! * - * Returns TRUE if the connection actually was dead and disconnected. + * Returns TRUE if the connection was dead and extracted. */ -static bool disconnect_if_dead(struct connectdata *conn, - struct Curl_easy *data) +static bool extract_if_dead(struct connectdata *conn, + struct Curl_easy *data) { size_t pipeLen = conn->send_pipe.size + conn->recv_pipe.size; if(!pipeLen && !conn->inuse) { @@ -3427,25 +972,30 @@ static bool disconnect_if_dead(struct connectdata *conn, if(dead) { conn->data = data; infof(data, "Connection %ld seems to be dead!\n", conn->connection_id); - - /* disconnect resources */ - Curl_disconnect(conn, /* dead_connection */TRUE); + Curl_conncache_remove_conn(conn, FALSE); return TRUE; } } return FALSE; } +struct prunedead { + struct Curl_easy *data; + struct connectdata *extracted; +}; + /* - * Wrapper to use disconnect_if_dead() function in Curl_conncache_foreach() + * Wrapper to use extract_if_dead() function in Curl_conncache_foreach() * - * Returns always 0. */ -static int call_disconnect_if_dead(struct connectdata *conn, - void *param) +static int call_extract_if_dead(struct connectdata *conn, void *param) { - struct Curl_easy* data = (struct Curl_easy*)param; - disconnect_if_dead(conn, data); + struct prunedead *p = (struct prunedead *)param; + if(extract_if_dead(conn, p->data)) { + /* stop the iteration here, pass back the connection that was extracted */ + p->extracted = conn; + return 1; + } return 0; /* continue iteration */ } @@ -3456,12 +1006,18 @@ static int call_disconnect_if_dead(struct connectdata *conn, */ static void prune_dead_connections(struct Curl_easy *data) { - struct curltime now = Curl_tvnow(); - time_t elapsed = Curl_tvdiff(now, data->state.conn_cache->last_cleanup); + struct curltime now = Curl_now(); + time_t elapsed = Curl_timediff(now, data->state.conn_cache->last_cleanup); if(elapsed >= 1000L) { - Curl_conncache_foreach(data->state.conn_cache, data, - call_disconnect_if_dead); + struct prunedead prune; + prune.data = data; + prune.extracted = NULL; + while(Curl_conncache_foreach(data, data->state.conn_cache, &prune, + call_extract_if_dead)) { + /* disconnect it */ + (void)Curl_disconnect(prune.extracted, /* dead_connection */TRUE); + } data->state.conn_cache->last_cleanup = now; } } @@ -3516,8 +1072,8 @@ ConnectionExists(struct Curl_easy *data, Curl_pipeline_site_blacklisted(data, needle)) canpipe &= ~ CURLPIPE_HTTP1; - /* Look up the bundle with all the connections to this - particular host */ + /* Look up the bundle with all the connections to this particular host. + Locks the connection cache, beware of early returns! */ bundle = Curl_conncache_find_bundle(needle, data->state.conn_cache); if(bundle) { /* Max pipe length is zero (unlimited) for multiplexed connections */ @@ -3540,6 +1096,7 @@ ConnectionExists(struct Curl_easy *data, if((bundle->multiuse == BUNDLE_UNKNOWN) && data->set.pipewait) { infof(data, "Server doesn't support multi-use yet, wait\n"); *waitpipe = TRUE; + Curl_conncache_unlock(needle); return FALSE; /* no re-use */ } @@ -3571,8 +1128,11 @@ ConnectionExists(struct Curl_easy *data, check = curr->ptr; curr = curr->next; - if(disconnect_if_dead(check, data)) + if(extract_if_dead(check, data)) { + /* disconnect it */ + (void)Curl_disconnect(check, /* dead_connection */TRUE); continue; + } pipeLen = check->send_pipe.size + check->recv_pipe.size; @@ -3703,6 +1263,11 @@ ConnectionExists(struct Curl_easy *data, already in use so we skip it */ continue; + if((check->inuse) && (check->data->multi != needle->data->multi)) + /* this could be subject for pipeline/multiplex use, but only + if they belong to the same multi handle */ + continue; + if(needle->localdev || needle->localport) { /* If we are bound to a specific local end (IP+port), we must not re-use a random other one, although if we didn't ask for a @@ -3889,9 +1454,13 @@ ConnectionExists(struct Curl_easy *data, } if(chosen) { + /* mark it as used before releasing the lock */ + chosen->inuse = TRUE; + Curl_conncache_unlock(needle); *usethis = chosen; return TRUE; /* yes, we found one to use! */ } + Curl_conncache_unlock(needle); if(foundPendingCandidate && data->set.pipewait) { infof(data, @@ -4104,7 +1673,7 @@ static bool is_ASCII_name(const char *hostname) /* * Perform any necessary IDN conversion of hostname */ -static void fix_hostname(struct connectdata *conn, struct hostname *host) +static CURLcode fix_hostname(struct connectdata *conn, struct hostname *host) { size_t len; struct Curl_easy *data = conn->data; @@ -4144,9 +1713,11 @@ static void fix_hostname(struct connectdata *conn, struct hostname *host) /* change the name pointer to point to the encoded hostname */ host->name = host->encalloc; } - else - infof(data, "Failed to convert %s to ACE; %s\n", host->name, + else { + failf(data, "Failed to convert %s to ACE; %s\n", host->name, idn2_strerror(rc)); + return CURLE_URL_MALFORMAT; + } } #elif defined(USE_WIN32_IDN) char *ace_hostname = NULL; @@ -4156,12 +1727,24 @@ static void fix_hostname(struct connectdata *conn, struct hostname *host) /* change the name pointer to point to the encoded hostname */ host->name = host->encalloc; } - else - infof(data, "Failed to convert %s to ACE;\n", host->name); + else { + failf(data, "Failed to convert %s to ACE;\n", host->name); + return CURLE_URL_MALFORMAT; + } #else infof(data, "IDN support not present, can't parse Unicode domains\n"); #endif } + { + char *hostp; + for(hostp = host->name; *hostp; hostp++) { + if(*hostp <= 32) { + failf(data, "Host name '%s' contains bad letter", host->name); + return CURLE_URL_MALFORMAT; + } + } + } + return CURLE_OK; } /* @@ -4196,15 +1779,30 @@ static void llist_dtor(void *user, void *element) */ static struct connectdata *allocate_conn(struct Curl_easy *data) { -#ifdef USE_SSL -#define SSL_EXTRA + 4 * Curl_ssl->sizeof_ssl_backend_data - sizeof(long long) -#else -#define SSL_EXTRA 0 -#endif - struct connectdata *conn = calloc(1, sizeof(struct connectdata) + SSL_EXTRA); + struct connectdata *conn = calloc(1, sizeof(struct connectdata)); if(!conn) return NULL; +#ifdef USE_SSL + /* The SSL backend-specific data (ssl_backend_data) objects are allocated as + a separate array to ensure suitable alignment. + Note that these backend pointers can be swapped by vtls (eg ssl backend + data becomes proxy backend data). */ + { + size_t sslsize = Curl_ssl->sizeof_ssl_backend_data; + char *ssl = calloc(4, sslsize); + if(!ssl) { + free(conn); + return NULL; + } + conn->ssl_extra = ssl; + conn->ssl[0].backend = (void *)ssl; + conn->ssl[1].backend = (void *)(ssl + sslsize); + conn->proxy_ssl[0].backend = (void *)(ssl + 2 * sslsize); + conn->proxy_ssl[1].backend = (void *)(ssl + 3 * sslsize); + } +#endif + conn->handler = &Curl_handler_dummy; /* Be sure we have a handler defined already from start to avoid NULL situations and checks */ @@ -4229,7 +1827,7 @@ static struct connectdata *allocate_conn(struct Curl_easy *data) connclose(conn, "Default to force-close"); /* Store creation time to help future close decision making */ - conn->created = Curl_tvnow(); + conn->created = Curl_now(); conn->data = data; /* Setup the association between this connection and the Curl_easy */ @@ -4284,23 +1882,6 @@ static struct connectdata *allocate_conn(struct Curl_easy *data) conn->ip_version = data->set.ipver; -#ifdef USE_SSL - /* - * To save on malloc()s, the SSL backend-specific data has been allocated - * at the end of the connectdata struct. - */ - { - char *p = (char *)&conn->align_data__do_not_use; - conn->ssl[0].backend = (struct ssl_backend_data *)p; - conn->ssl[1].backend = - (struct ssl_backend_data *)(p + Curl_ssl->sizeof_ssl_backend_data); - conn->proxy_ssl[0].backend = - (struct ssl_backend_data *)(p + Curl_ssl->sizeof_ssl_backend_data * 2); - conn->proxy_ssl[1].backend = - (struct ssl_backend_data *)(p + Curl_ssl->sizeof_ssl_backend_data * 3); - } -#endif - #if !defined(CURL_DISABLE_HTTP) && defined(USE_NTLM) && \ defined(NTLM_WB_ENABLED) conn->ntlm_auth_hlpr_socket = CURL_SOCKET_BAD; @@ -4347,6 +1928,9 @@ static struct connectdata *allocate_conn(struct Curl_easy *data) free(conn->master_buffer); free(conn->localdev); +#ifdef USE_SSL + free(conn->ssl_extra); +#endif free(conn); return NULL; } @@ -4442,6 +2026,14 @@ static CURLcode parseurlandfillconn(struct Curl_easy *data, ('A' <= str[0] && str[0] <= 'Z')) && \ (str[1] == ':')) + /* MSDOS/Windows style drive prefix, optionally with + * a '|' instead of ':', followed by a slash or NUL */ +#define STARTS_WITH_URL_DRIVE_PREFIX(str) \ + ((('a' <= (str)[0] && (str)[0] <= 'z') || \ + ('A' <= (str)[0] && (str)[0] <= 'Z')) && \ + ((str)[1] == ':' || (str)[1] == '|') && \ + ((str)[2] == '/' || (str)[2] == '\\' || (str)[2] == 0)) + /* Don't mistake a drive letter for a scheme if the default protocol is file. curld --proto-default file c:/foo/bar.txt */ if(STARTS_WITH_DRIVE_PREFIX(data->change.url) && @@ -4474,63 +2066,90 @@ static CURLcode parseurlandfillconn(struct Curl_easy *data, return CURLE_URL_MALFORMAT; } - if(url_has_scheme && path[0] == '/' && path[1] == '/') { - /* Allow omitted hostname (e.g. file:/<path>). This is not strictly - * speaking a valid file: URL by RFC 1738, but treating file:/<path> as - * file://localhost/<path> is similar to how other schemes treat missing - * hostnames. See RFC 1808. */ - - /* This cannot be done with strcpy() in a portable manner, since the - memory areas overlap! */ - memmove(path, path + 2, strlen(path + 2) + 1); + if(url_has_scheme && path[0] == '/' && path[1] == '/' && + path[2] == '/' && path[3] == '/') { + /* This appears to be a UNC string (usually indicating a SMB share). + * We don't do SMB in file: URLs. (TODO?) + */ + failf(data, "SMB shares are not supported in file: URLs."); + return CURLE_URL_MALFORMAT; } - /* - * we deal with file://<host>/<path> differently since it supports no - * hostname other than "localhost" and "127.0.0.1", which is unique among - * the URL protocols specified in RFC 1738 + /* Extra handling URLs with an authority component (i.e. that start with + * "file://") + * + * We allow omitted hostname (e.g. file:/<path>) -- valid according to + * RFC 8089, but not the (current) WHAT-WG URL spec. */ - if(path[0] != '/' && !STARTS_WITH_DRIVE_PREFIX(path)) { - /* the URL includes a host name, it must match "localhost" or - "127.0.0.1" to be valid */ - char *ptr; - if(!checkprefix("localhost/", path) && - !checkprefix("127.0.0.1/", path)) { - failf(data, "Invalid file://hostname/, " - "expected localhost or 127.0.0.1 or none"); - return CURLE_URL_MALFORMAT; - } - ptr = &path[9]; /* now points to the slash after the host */ - - /* there was a host name and slash present - - RFC1738 (section 3.1, page 5) says: - - The rest of the locator consists of data specific to the scheme, - and is known as the "url-path". It supplies the details of how the - specified resource can be accessed. Note that the "/" between the - host (or port) and the url-path is NOT part of the url-path. + if(url_has_scheme && path[0] == '/' && path[1] == '/') { + /* swallow the two slashes */ + char *ptr = &path[2]; - As most agents use file://localhost/foo to get '/foo' although the - slash preceding foo is a separator and not a slash for the path, - a URL as file://localhost//foo must be valid as well, to refer to - the same file with an absolute path. - */ + /* + * According to RFC 8089, a file: URL can be reliably dereferenced if: + * + * o it has no/blank hostname, or + * + * o the hostname matches "localhost" (case-insensitively), or + * + * o the hostname is a FQDN that resolves to this machine. + * + * For brevity, we only consider URLs with empty, "localhost", or + * "127.0.0.1" hostnames as local. + * + * Additionally, there is an exception for URLs with a Windows drive + * letter in the authority (which was accidentally omitted from RFC 8089 + * Appendix E, but believe me, it was meant to be there. --MK) + */ + if(ptr[0] != '/' && !STARTS_WITH_URL_DRIVE_PREFIX(ptr)) { + /* the URL includes a host name, it must match "localhost" or + "127.0.0.1" to be valid */ + if(!checkprefix("localhost/", ptr) && + !checkprefix("127.0.0.1/", ptr)) { + failf(data, "Invalid file://hostname/, " + "expected localhost or 127.0.0.1 or none"); + return CURLE_URL_MALFORMAT; + } + ptr += 9; /* now points to the slash after the host */ + } - if('/' == ptr[1]) - /* if there was two slashes, we skip the first one as that is then - used truly as a separator */ + /* + * RFC 8089, Appendix D, Section D.1, says: + * + * > In a POSIX file system, the root of the file system is represented + * > as a directory with a zero-length name, usually written as "/"; the + * > presence of this root in a file URI can be taken as given by the + * > initial slash in the "path-absolute" rule. + * + * i.e. the first slash is part of the path. + * + * However in RFC 1738 the "/" between the host (or port) and the + * URL-path was NOT part of the URL-path. Any agent that followed the + * older spec strictly, and wanted to refer to a file with an absolute + * path, would have included a second slash. So if there are two + * slashes, swallow one. + */ + if('/' == ptr[1]) /* note: the only way ptr[0]!='/' is if ptr[1]==':' */ ptr++; - /* This cannot be made with strcpy, as the memory chunks overlap! */ + /* This cannot be done with strcpy, as the memory chunks overlap! */ memmove(path, ptr, strlen(ptr) + 1); } #if !defined(MSDOS) && !defined(WIN32) && !defined(__CYGWIN__) - if(STARTS_WITH_DRIVE_PREFIX(path)) { + /* Don't allow Windows drive letters when not in Windows. + * This catches both "file:/c:" and "file:c:" */ + if(('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) || + STARTS_WITH_URL_DRIVE_PREFIX(path)) { failf(data, "File drive letters are only accepted in MSDOS/Windows."); return CURLE_URL_MALFORMAT; } +#else + /* If the path starts with a slash and a drive letter, ditch the slash */ + if('/' == path[0] && STARTS_WITH_URL_DRIVE_PREFIX(&path[1])) { + /* This cannot be done with strcpy, as the memory chunks overlap! */ + memmove(path, &path[1], strlen(&path[1]) + 1); + } #endif protop = "file"; /* protocol string */ @@ -5142,8 +2761,8 @@ static CURLcode parse_proxy(struct Curl_easy *data, atsign = strchr(proxyptr, '@'); if(atsign) { CURLcode result = - parse_login_details(proxyptr, atsign - proxyptr, - &proxyuser, &proxypasswd, NULL); + Curl_parse_login_details(proxyptr, atsign - proxyptr, + &proxyuser, &proxypasswd, NULL); if(result) return result; proxyptr = atsign + 1; @@ -5534,10 +3153,11 @@ static CURLcode parse_url_login(struct Curl_easy *data, /* We could use the login information in the URL so extract it. Only parse options if the handler says we should. */ - result = parse_login_details(login, ptr - login - 1, - &userp, &passwdp, - (conn->handler->flags & PROTOPT_URLOPTIONS)? - &optionsp:NULL); + result = + Curl_parse_login_details(login, ptr - login - 1, + &userp, &passwdp, + (conn->handler->flags & PROTOPT_URLOPTIONS)? + &optionsp:NULL); if(result) goto out; @@ -5593,7 +3213,7 @@ static CURLcode parse_url_login(struct Curl_easy *data, } /* - * parse_login_details() + * Curl_parse_login_details() * * This is used to parse a login string for user name, password and options in * the following formats: @@ -5621,9 +3241,9 @@ static CURLcode parse_url_login(struct Curl_easy *data, * * Returns CURLE_OK on success. */ -static CURLcode parse_login_details(const char *login, const size_t len, - char **userp, char **passwdp, - char **optionsp) +CURLcode Curl_parse_login_details(const char *login, const size_t len, + char **userp, char **passwdp, + char **optionsp) { CURLcode result = CURLE_OK; char *ubuf = NULL; @@ -5752,7 +3372,13 @@ static CURLcode parse_remote_port(struct Curl_easy *data, portptr = strchr(conn->host.name, ']'); if(portptr) { *portptr++ = '\0'; /* zero terminate, killing the bracket */ - if(':' != *portptr) + if(*portptr) { + if (*portptr != ':') { + failf(data, "IPv6 closing bracket followed by '%c'", *portptr); + return CURLE_URL_MALFORMAT; + } + } + else portptr = NULL; /* no port number available */ } } @@ -5986,6 +3612,7 @@ static CURLcode parse_connect_to_host_port(struct Curl_easy *data, /* detect and extract RFC6874-style IPv6-addresses */ if(*hostptr == '[') { +#ifdef ENABLE_IPV6 char *ptr = ++hostptr; /* advance beyond the initial bracket */ while(*ptr && (ISXDIGIT(*ptr) || (*ptr == ':') || (*ptr == '.'))) ptr++; @@ -6009,6 +3636,11 @@ static CURLcode parse_connect_to_host_port(struct Curl_easy *data, * hostptr first, but I can't see anything wrong with that as no host * name nor a numeric can legally start with a bracket. */ +#else + failf(data, "Use of IPv6 in *_CONNECT_TO without IPv6 support built-in!"); + free(host_dup); + return CURLE_NOT_BUILT_IN; +#endif } /* Get port number off server.com:1080 */ @@ -6171,7 +3803,7 @@ static CURLcode resolve_server(struct Curl_easy *data, bool *async) { CURLcode result = CURLE_OK; - time_t timeout_ms = Curl_timeleft(data, NULL, TRUE); + timediff_t timeout_ms = Curl_timeleft(data, NULL, TRUE); /************************************************************* * Resolve the name of the server or proxy @@ -6588,13 +4220,24 @@ static CURLcode create_conn(struct Curl_easy *data, /************************************************************* * IDN-fix the hostnames *************************************************************/ - fix_hostname(conn, &conn->host); - if(conn->bits.conn_to_host) - fix_hostname(conn, &conn->conn_to_host); - if(conn->bits.httpproxy) - fix_hostname(conn, &conn->http_proxy.host); - if(conn->bits.socksproxy) - fix_hostname(conn, &conn->socks_proxy.host); + result = fix_hostname(conn, &conn->host); + if(result) + goto out; + if(conn->bits.conn_to_host) { + result = fix_hostname(conn, &conn->conn_to_host); + if(result) + goto out; + } + if(conn->bits.httpproxy) { + result = fix_hostname(conn, &conn->http_proxy.host); + if(result) + goto out; + } + if(conn->bits.socksproxy) { + result = fix_hostname(conn, &conn->socks_proxy.host); + if(result) + goto out; + } /************************************************************* * Check whether the host and the "connect to host" are equal. @@ -6751,20 +4394,21 @@ static CURLcode create_conn(struct Curl_easy *data, else reuse = ConnectionExists(data, conn, &conn_temp, &force_reuse, &waitpipe); - /* If we found a reusable connection, we may still want to - open a new connection if we are pipelining. */ + /* If we found a reusable connection that is now marked as in use, we may + still want to open a new connection if we are pipelining. */ if(reuse && !force_reuse && IsPipeliningPossible(data, conn_temp)) { size_t pipelen = conn_temp->send_pipe.size + conn_temp->recv_pipe.size; if(pipelen > 0) { infof(data, "Found connection %ld, with requests in the pipe (%zu)\n", conn_temp->connection_id, pipelen); - if(conn_temp->bundle->num_connections < max_host_connections && - data->state.conn_cache->num_connections < max_total_connections) { + if(Curl_conncache_bundle_size(conn_temp) < max_host_connections && + Curl_conncache_size(data) < max_total_connections) { /* We want a new connection anyway */ reuse = FALSE; infof(data, "We can reuse, but we want a new connection anyway\n"); + Curl_conncache_return_conn(conn_temp); } } } @@ -6776,9 +4420,10 @@ static CURLcode create_conn(struct Curl_easy *data, * just allocated before we can move along and use the previously * existing one. */ - conn_temp->inuse = TRUE; /* mark this as being in use so that no other - handle in a multi stack may nick it */ reuse_conn(conn, conn_temp); +#ifdef USE_SSL + free(conn->ssl_extra); +#endif free(conn); /* we don't need this anymore */ conn = conn_temp; *in_connect = conn; @@ -6794,7 +4439,6 @@ static CURLcode create_conn(struct Curl_easy *data, /* We have decided that we want a new connection. However, we may not be able to do that if we have reached the limit of how many connections we are allowed to open. */ - struct connectbundle *bundle = NULL; if(conn->handler->flags & PROTOPT_ALPN_NPN) { /* The protocol wants it, so set the bits if enabled in the easy handle @@ -6809,35 +4453,42 @@ static CURLcode create_conn(struct Curl_easy *data, /* There is a connection that *might* become usable for pipelining "soon", and we wait for that */ connections_available = FALSE; - else - bundle = Curl_conncache_find_bundle(conn, data->state.conn_cache); - - if(max_host_connections > 0 && bundle && - (bundle->num_connections >= max_host_connections)) { - struct connectdata *conn_candidate; - - /* The bundle is full. Let's see if we can kill a connection. */ - conn_candidate = find_oldest_idle_connection_in_bundle(data, bundle); - - if(conn_candidate) { - /* Set the connection's owner correctly, then kill it */ - conn_candidate->data = data; - (void)Curl_disconnect(conn_candidate, /* dead_connection */ FALSE); - } - else { - infof(data, "No more connections allowed to host: %d\n", - max_host_connections); - connections_available = FALSE; + else { + /* this gets a lock on the conncache */ + struct connectbundle *bundle = + Curl_conncache_find_bundle(conn, data->state.conn_cache); + + if(max_host_connections > 0 && bundle && + (bundle->num_connections >= max_host_connections)) { + struct connectdata *conn_candidate; + + /* The bundle is full. Extract the oldest connection. */ + conn_candidate = Curl_conncache_extract_bundle(data, bundle); + Curl_conncache_unlock(conn); + + if(conn_candidate) { + /* Set the connection's owner correctly, then kill it */ + conn_candidate->data = data; + (void)Curl_disconnect(conn_candidate, /* dead_connection */ FALSE); + } + else { + infof(data, "No more connections allowed to host: %d\n", + max_host_connections); + connections_available = FALSE; + } } + else + Curl_conncache_unlock(conn); + } if(connections_available && (max_total_connections > 0) && - (data->state.conn_cache->num_connections >= max_total_connections)) { + (Curl_conncache_size(data) >= max_total_connections)) { struct connectdata *conn_candidate; /* The cache is full. Let's see if we can kill a connection. */ - conn_candidate = Curl_oldest_idle_connection(data); + conn_candidate = Curl_conncache_extract_oldest(data); if(conn_candidate) { /* Set the connection's owner correctly, then kill it */ @@ -6860,6 +4511,9 @@ static CURLcode create_conn(struct Curl_easy *data, goto out; } else { + /* Mark the connection as used, before we add it */ + conn->inuse = TRUE; + /* * This is a brand new connection, so let's store it in the connection * cache of ours! @@ -6887,9 +4541,6 @@ static CURLcode create_conn(struct Curl_easy *data, #endif } - /* Mark the connection as used */ - conn->inuse = TRUE; - /* Setup and init stuff before DO starts, in preparing for the transfer. */ Curl_init_do(data, conn); @@ -6972,7 +4623,7 @@ CURLcode Curl_setup_conn(struct connectdata *conn, /* set start time here for timeout purposes in the connect procedure, it is later set again for the progress meter purpose */ - conn->now = Curl_tvnow(); + conn->now = Curl_now(); if(CURL_SOCKET_BAD == conn->sock[FIRSTSOCKET]) { conn->bits.tcpconnect[FIRSTSOCKET] = FALSE; @@ -6989,22 +4640,8 @@ CURLcode Curl_setup_conn(struct connectdata *conn, Curl_verboseconnect(conn); } - conn->now = Curl_tvnow(); /* time this *after* the connect is done, we - set this here perhaps a second time */ - -#ifdef __EMX__ - /* - * This check is quite a hack. We're calling _fsetmode to fix the problem - * with fwrite converting newline characters (you get mangled text files, - * and corrupted binary files when you download to stdout and redirect it to - * a file). - */ - - if((data->set.out)->_handle == NULL) { - _fsetmode(stdout, "b"); - } -#endif - + conn->now = Curl_now(); /* time this *after* the connect is done, we set + this here perhaps a second time */ return result; } @@ -7062,13 +4699,17 @@ CURLcode Curl_init_do(struct Curl_easy *data, struct connectdata *conn) { struct SingleRequest *k = &data->req; - if(conn) - conn->bits.do_more = FALSE; /* by default there's no curl_do_more() to - * use */ + conn->bits.do_more = FALSE; /* by default there's no curl_do_more() to + use */ data->state.done = FALSE; /* *_done() is not called yet */ data->state.expect100header = FALSE; + /* if the protocol used doesn't support wildcards, switch it off */ + if(data->state.wildcardmatch && + !(conn->handler->flags & PROTOPT_WILDCARD)) + data->state.wildcardmatch = FALSE; + if(data->set.opt_no_body) /* in HTTP lingo, no body means using the HEAD request... */ data->set.httpreq = HTTPREQ_HEAD; @@ -7080,7 +4721,7 @@ CURLcode Curl_init_do(struct Curl_easy *data, struct connectdata *conn) HTTP. */ data->set.httpreq = HTTPREQ_GET; - k->start = Curl_tvnow(); /* start time */ + k->start = Curl_now(); /* start time */ k->now = k->start; /* current time is now */ k->header = TRUE; /* assume header */ |