diff options
author | Ulion <ulion2002@gmail.com> | 2013-01-21 23:20:09 +0100 |
---|---|---|
committer | Daniel Stenberg <daniel@haxx.se> | 2013-01-22 15:43:29 +0100 |
commit | 2698520aef593cbd746a64f79021a4c8d7c83d65 (patch) | |
tree | 69cccacf56e002c7413d8ab7356a92ecfc94212c /src | |
parent | fa176376c8af90fcce632db63ac691a471577e8e (diff) | |
download | curl-2698520aef593cbd746a64f79021a4c8d7c83d65.tar.gz |
formpost: support quotes, commas and semicolon in file names
- document the double-quote and backslash need be escaped if quoting.
- libcurl formdata escape double-quote in filename by backslash.
- curl formparse can parse filename both contains '"' and ',' or ';'.
- curl now can uploading file with ',' or ';' in filename.
Bug: http://curl.haxx.se/bug/view.cgi?id=1171
Diffstat (limited to 'src')
-rw-r--r-- | src/tool_formparse.c | 234 |
1 files changed, 144 insertions, 90 deletions
diff --git a/src/tool_formparse.c b/src/tool_formparse.c index 12b1a9d46..fe357f504 100644 --- a/src/tool_formparse.c +++ b/src/tool_formparse.c @@ -5,7 +5,7 @@ * | (__| |_| | _ <| |___ * \___|\___/|_| \_\_____| * - * Copyright (C) 1998 - 2012, Daniel Stenberg, <daniel@haxx.se>, et al. + * Copyright (C) 1998 - 2013, 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 @@ -34,13 +34,73 @@ #include "memdebug.h" /* keep this as LAST include */ + +/* + * helper function to get a word from form param + * after call get_parm_word, str either point to string end + * or point to any of end chars. + */ +static char *get_param_word(char **str, char **end_pos) +{ + char *ptr = *str; + char *word_begin = NULL; + char *ptr2; + char *escape = NULL; + const char *end_chars = ";,"; + + /* the first non-space char is here */ + word_begin = ptr; + if(*ptr == '"') { + ++ptr; + while(*ptr) { + if(*ptr == '\\') { + if(ptr[1] == '\\' || ptr[1] == '"') { + /* remember the first escape position */ + if(!escape) + escape = ptr; + /* skip escape of back-slash or double-quote */ + ptr += 2; + continue; + } + } + if(*ptr == '"') { + *end_pos = ptr; + if(escape) { + /* has escape, we restore the unescaped string here */ + ptr = ptr2 = escape; + do { + if(*ptr == '\\' && (ptr[1] == '\\' || ptr[1] == '"')) + ++ptr; + *ptr2++ = *ptr++; + } + while(ptr < *end_pos); + *end_pos = ptr2; + } + while(*ptr && NULL==strchr(end_chars, *ptr)) + ++ptr; + *str = ptr; + return word_begin+1; + } + ++ptr; + } + /* end quote is missing, treat it as non-quoted. */ + ptr = word_begin; + } + + while(*ptr && NULL==strchr(end_chars, *ptr)) + ++ptr; + *str = *end_pos = ptr; + return word_begin; +} + /*************************************************************************** * * formparse() * * Reads a 'name=value' parameter and builds the appropriate linked list. * - * Specify files to upload with 'name=@filename'. Supports specified + * Specify files to upload with 'name=@filename', or 'name=@"filename"' + * in case the filename contain ',' or ';'. Supports specified * given Content-Type of the files. Such as ';type=<content-type>'. * * If literal_value is set, any initial '@' or '<' in the value string @@ -51,6 +111,10 @@ * * 'name=@filename,filename2,filename3' * + * or use double-quotes quote the filename: + * + * 'name=@"filename","filename2","filename3"' + * * If you want content-types specified for each too, write them like: * * 'name=@filename;type=image/gif,filename2,filename3' @@ -64,7 +128,12 @@ * To upload a file, but to fake the file name that will be included in the * formpost, do like this: * - * 'name=@filename;filename=/dev/null' + * 'name=@filename;filename=/dev/null' or quote the faked filename like: + * 'name=@filename;filename="play, play, and play.txt"' + * + * If filename/path contains ',' or ';', it must be quoted by double-quotes, + * else curl will fail to figure out the correct filename. if the filename + * tobe quoted contains '"' or '\', '"' and '\' must be escaped by backslash. * * This function uses curl_formadd to fulfill it's job. Is heavily based on * the old curl_formparse code. @@ -86,7 +155,6 @@ int formparse(struct Configurable *config, char *contp; const char *type = NULL; char *sep; - char *sep2; if((1 == sscanf(input, "%255[^=]=", name)) && ((contp = strchr(input, '=')) != NULL)) { @@ -107,118 +175,104 @@ int formparse(struct Configurable *config, struct multi_files *multi_start = NULL; struct multi_files *multi_current = NULL; - contp++; + char *ptr = contp; + char *end = ptr + strlen(ptr); do { /* since this was a file, it may have a content-type specifier at the end too, or a filename. Or both. */ - char *ptr; char *filename = NULL; - - sep = strchr(contp, ';'); - sep2 = strchr(contp, ','); - - /* pick the closest */ - if(sep2 && (sep2 < sep)) { - sep = sep2; - - /* no type was specified! */ - } + char *word_end; + bool semicolon; type = NULL; - if(sep) { - bool semicolon = (';' == *sep) ? TRUE : FALSE; - - *sep = '\0'; /* terminate file name at separator */ - - ptr = sep+1; /* point to the text following the separator */ - - while(semicolon && ptr && (','!= *ptr)) { - - /* pass all white spaces */ - while(ISSPACE(*ptr)) - ptr++; - - if(checkprefix("type=", ptr)) { - /* set type pointer */ - type = &ptr[5]; - - /* verify that this is a fine type specifier */ - if(2 != sscanf(type, "%127[^/]/%127[^;,\n]", - type_major, type_minor)) { - warnf(config, "Illegally formatted content-type field!\n"); - Curl_safefree(contents); - FreeMultiInfo(&multi_start, &multi_current); - return 2; /* illegal content-type syntax! */ - } - - /* now point beyond the content-type specifier */ - sep = (char *)type + strlen(type_major)+strlen(type_minor)+1; - - /* there's a semicolon following - we check if it is a filename - specified and if not we simply assume that it is text that - the user wants included in the type and include that too up - to the next zero or semicolon. */ - if(*sep==';') { - if(!checkprefix(";filename=", sep)) { - sep2 = strchr(sep+1, ';'); - if(sep2) - sep = sep2; - else - sep = sep + strlen(sep); /* point to end of string */ - } - } - else - semicolon = FALSE; - - if(*sep) { - *sep = '\0'; /* zero terminate type string */ - - ptr = sep+1; - } - else - ptr = NULL; /* end */ + ++ptr; + contp = get_param_word(&ptr, &word_end); + semicolon = (';' == *ptr) ? TRUE : FALSE; + *word_end = '\0'; /* terminate the contp */ + + /* have other content, continue parse */ + while(semicolon) { + /* have type or filename field */ + ++ptr; + while(*ptr && (ISSPACE(*ptr))) + ++ptr; + + if(checkprefix("type=", ptr)) { + /* set type pointer */ + type = &ptr[5]; + + /* verify that this is a fine type specifier */ + if(2 != sscanf(type, "%127[^/]/%127[^;,\n]", + type_major, type_minor)) { + warnf(config, "Illegally formatted content-type field!\n"); + Curl_safefree(contents); + FreeMultiInfo(&multi_start, &multi_current); + return 2; /* illegal content-type syntax! */ } - else if(checkprefix("filename=", ptr)) { - filename = &ptr[9]; - ptr = strchr(filename, ';'); - if(!ptr) { - ptr = strchr(filename, ','); - } - if(ptr) { - *ptr = '\0'; /* zero terminate */ - ptr++; + + /* now point beyond the content-type specifier */ + sep = (char *)type + strlen(type_major)+strlen(type_minor)+1; + + /* there's a semicolon following - we check if it is a filename + specified and if not we simply assume that it is text that + the user wants included in the type and include that too up + to the next sep. */ + ptr = sep; + if(*sep==';') { + if(!checkprefix(";filename=", sep)) { + ptr = sep + 1; + (void)get_param_word(&ptr, &sep); + semicolon = (';' == *ptr) ? TRUE : FALSE; } } else - /* confusion, bail out of loop */ - break; - } + semicolon = FALSE; - sep = ptr; + if(*sep) + *sep = '\0'; /* zero terminate type string */ + } + else if(checkprefix("filename=", ptr)) { + ptr += 9; + filename = get_param_word(&ptr, &word_end); + semicolon = (';' == *ptr) ? TRUE : FALSE; + *word_end = '\0'; + } + else { + /* unknown prefix, skip to next block */ + char *unknown = NULL; + unknown = get_param_word(&ptr, &word_end); + semicolon = (';' == *ptr) ? TRUE : FALSE; + if(*unknown) { + *word_end = '\0'; + warnf(config, "skip unknown form field: %s\n", unknown); + } + } } + /* now ptr point to comma or string end */ + /* if type == NULL curl_formadd takes care of the problem */ - if(!AddMultiFiles(contp, type, filename, &multi_start, + if(*contp && !AddMultiFiles(contp, type, filename, &multi_start, &multi_current)) { warnf(config, "Error building form post!\n"); Curl_safefree(contents); FreeMultiInfo(&multi_start, &multi_current); return 3; } - contp = sep; /* move the contents pointer to after the separator */ - } while(sep && *sep); /* loop if there's another file name */ + /* *ptr could be '\0', so we just check with the string end */ + } while(ptr < end); /* loop if there's another file name */ /* now we add the multiple files section */ if(multi_start) { struct curl_forms *forms = NULL; - struct multi_files *ptr = multi_start; + struct multi_files *start = multi_start; unsigned int i, count = 0; - while(ptr) { - ptr = ptr->next; + while(start) { + start = start->next; ++count; } forms = malloc((count+1)*sizeof(struct curl_forms)); @@ -228,9 +282,9 @@ int formparse(struct Configurable *config, FreeMultiInfo(&multi_start, &multi_current); return 4; } - for(i = 0, ptr = multi_start; i < count; ++i, ptr = ptr->next) { - forms[i].option = ptr->form.option; - forms[i].value = ptr->form.value; + for(i = 0, start = multi_start; i < count; ++i, start = start->next) { + forms[i].option = start->form.option; + forms[i].value = start->form.value; } forms[count].option = CURLFORM_END; FreeMultiInfo(&multi_start, &multi_current); |