Tim Duesterhus | dbd25c3 | 2021-04-15 21:45:55 +0200 | [diff] [blame] | 1 | /* |
| 2 | * HTTP request URI normalization. |
| 3 | * |
| 4 | * Copyright 2021 Tim Duesterhus <tim@bastelstu.be> |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU General Public License |
| 8 | * as published by the Free Software Foundation; either version |
| 9 | * 2 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | */ |
| 12 | |
Tim Duesterhus | d371e99 | 2021-04-15 21:45:58 +0200 | [diff] [blame] | 13 | #include <import/ist.h> |
| 14 | |
Tim Duesterhus | dbd25c3 | 2021-04-15 21:45:55 +0200 | [diff] [blame] | 15 | #include <haproxy/api.h> |
Tim Duesterhus | d7b89be | 2021-04-15 21:46:01 +0200 | [diff] [blame] | 16 | #include <haproxy/buf.h> |
| 17 | #include <haproxy/chunk.h> |
Tim Duesterhus | a407193 | 2021-04-15 21:46:02 +0200 | [diff] [blame] | 18 | #include <haproxy/tools.h> |
Tim Duesterhus | dbd25c3 | 2021-04-15 21:45:55 +0200 | [diff] [blame] | 19 | #include <haproxy/uri_normalizer.h> |
| 20 | |
Tim Duesterhus | a407193 | 2021-04-15 21:46:02 +0200 | [diff] [blame] | 21 | /* Uppercases letters used in percent encoding. |
| 22 | * |
| 23 | * If `strict` is set to 0 then percent characters that are not followed by a |
| 24 | * hexadecimal digit are returned as-is without modifying the following letters. |
| 25 | * If `strict` is set to 1 then `URI_NORMALIZER_ERR_INVALID_INPUT` is returned |
| 26 | * for invalid sequences. |
| 27 | */ |
| 28 | enum uri_normalizer_err uri_normalizer_percent_upper(const struct ist input, int strict, struct ist *dst) |
| 29 | { |
| 30 | enum uri_normalizer_err err; |
| 31 | |
| 32 | const size_t size = istclear(dst); |
| 33 | struct ist output = *dst; |
| 34 | |
| 35 | struct ist scanner = input; |
| 36 | |
| 37 | /* The output will have the same length. */ |
| 38 | if (size < istlen(input)) { |
| 39 | err = URI_NORMALIZER_ERR_ALLOC; |
| 40 | goto fail; |
| 41 | } |
| 42 | |
| 43 | while (istlen(scanner)) { |
| 44 | const char current = istshift(&scanner); |
| 45 | |
| 46 | if (current == '%') { |
| 47 | if (istlen(scanner) >= 2) { |
| 48 | if (ishex(istptr(scanner)[0]) && ishex(istptr(scanner)[1])) { |
| 49 | output = __istappend(output, current); |
| 50 | output = __istappend(output, toupper(istshift(&scanner))); |
| 51 | output = __istappend(output, toupper(istshift(&scanner))); |
| 52 | continue; |
| 53 | } |
| 54 | } |
| 55 | |
| 56 | if (strict) { |
| 57 | err = URI_NORMALIZER_ERR_INVALID_INPUT; |
| 58 | goto fail; |
| 59 | } |
| 60 | else { |
| 61 | output = __istappend(output, current); |
| 62 | } |
| 63 | } |
| 64 | else { |
| 65 | output = __istappend(output, current); |
| 66 | } |
| 67 | } |
| 68 | |
| 69 | *dst = output; |
| 70 | |
| 71 | return URI_NORMALIZER_ERR_NONE; |
| 72 | |
| 73 | fail: |
| 74 | |
| 75 | return err; |
| 76 | } |
| 77 | |
Tim Duesterhus | 560e1a6 | 2021-04-15 21:46:00 +0200 | [diff] [blame] | 78 | /* Merges `/../` with preceding path segments. |
| 79 | * |
| 80 | * If `full` is set to `0` then `/../` will be printed at the start of the resulting |
| 81 | * path if the number of `/../` exceeds the number of other segments. If `full` is |
| 82 | * set to `1` these will not be printed. |
| 83 | */ |
| 84 | enum uri_normalizer_err uri_normalizer_path_dotdot(const struct ist path, int full, struct ist *dst) |
Tim Duesterhus | 9982fc2 | 2021-04-15 21:45:59 +0200 | [diff] [blame] | 85 | { |
| 86 | enum uri_normalizer_err err; |
| 87 | |
| 88 | const size_t size = istclear(dst); |
| 89 | char * const tail = istptr(*dst) + size; |
| 90 | char *head = tail; |
| 91 | |
| 92 | ssize_t offset = istlen(path) - 1; |
| 93 | |
| 94 | int up = 0; |
| 95 | |
| 96 | /* The path will either be shortened or have the same length. */ |
| 97 | if (size < istlen(path)) { |
| 98 | err = URI_NORMALIZER_ERR_ALLOC; |
| 99 | goto fail; |
| 100 | } |
| 101 | |
| 102 | /* Handle `/..` at the end of the path without a trailing slash. */ |
| 103 | if (offset >= 2 && istmatch(istadv(path, offset - 2), ist("/.."))) { |
| 104 | up++; |
| 105 | offset -= 2; |
| 106 | } |
| 107 | |
| 108 | while (offset >= 0) { |
| 109 | if (offset >= 3 && istmatch(istadv(path, offset - 3), ist("/../"))) { |
| 110 | up++; |
| 111 | offset -= 3; |
| 112 | continue; |
| 113 | } |
| 114 | |
| 115 | if (up > 0) { |
| 116 | /* Skip the slash. */ |
| 117 | offset--; |
| 118 | |
| 119 | /* First check whether we already reached the start of the path, |
| 120 | * before popping the current `/../`. |
| 121 | */ |
| 122 | if (offset >= 0) { |
| 123 | up--; |
| 124 | |
| 125 | /* Skip the current path segment. */ |
| 126 | while (offset >= 0 && istptr(path)[offset] != '/') |
| 127 | offset--; |
| 128 | } |
| 129 | } |
| 130 | else { |
| 131 | /* Prepend the slash. */ |
| 132 | *(--head) = istptr(path)[offset]; |
| 133 | offset--; |
| 134 | |
| 135 | /* Prepend the current path segment. */ |
| 136 | while (offset >= 0 && istptr(path)[offset] != '/') { |
| 137 | *(--head) = istptr(path)[offset]; |
| 138 | offset--; |
| 139 | } |
| 140 | } |
| 141 | } |
| 142 | |
| 143 | if (up > 0) { |
| 144 | /* Prepend a trailing slash. */ |
| 145 | *(--head) = '/'; |
| 146 | |
Tim Duesterhus | 560e1a6 | 2021-04-15 21:46:00 +0200 | [diff] [blame] | 147 | if (!full) { |
| 148 | /* Prepend unconsumed `/..`. */ |
| 149 | do { |
| 150 | *(--head) = '.'; |
| 151 | *(--head) = '.'; |
| 152 | *(--head) = '/'; |
| 153 | up--; |
| 154 | } while (up > 0); |
| 155 | } |
Tim Duesterhus | 9982fc2 | 2021-04-15 21:45:59 +0200 | [diff] [blame] | 156 | } |
| 157 | |
| 158 | *dst = ist2(head, tail - head); |
| 159 | |
| 160 | return URI_NORMALIZER_ERR_NONE; |
| 161 | |
| 162 | fail: |
| 163 | |
| 164 | return err; |
| 165 | } |
| 166 | |
Tim Duesterhus | d371e99 | 2021-04-15 21:45:58 +0200 | [diff] [blame] | 167 | /* Merges adjacent slashes in the given path. */ |
| 168 | enum uri_normalizer_err uri_normalizer_path_merge_slashes(const struct ist path, struct ist *dst) |
| 169 | { |
| 170 | enum uri_normalizer_err err; |
| 171 | |
| 172 | const size_t size = istclear(dst); |
| 173 | struct ist newpath = *dst; |
| 174 | |
| 175 | struct ist scanner = path; |
| 176 | |
| 177 | /* The path will either be shortened or have the same length. */ |
| 178 | if (size < istlen(path)) { |
| 179 | err = URI_NORMALIZER_ERR_ALLOC; |
| 180 | goto fail; |
| 181 | } |
| 182 | |
| 183 | while (istlen(scanner) > 0) { |
| 184 | const char current = istshift(&scanner); |
| 185 | |
| 186 | if (current == '/') { |
| 187 | while (istlen(scanner) > 0 && *istptr(scanner) == '/') |
| 188 | scanner = istnext(scanner); |
| 189 | } |
| 190 | |
| 191 | newpath = __istappend(newpath, current); |
| 192 | } |
| 193 | |
| 194 | *dst = newpath; |
| 195 | |
| 196 | return URI_NORMALIZER_ERR_NONE; |
| 197 | |
| 198 | fail: |
| 199 | |
| 200 | return err; |
| 201 | } |
| 202 | |
Tim Duesterhus | d7b89be | 2021-04-15 21:46:01 +0200 | [diff] [blame] | 203 | /* Compares two query parameters by name. Query parameters are ordered |
| 204 | * as with memcmp. Shorter parameter names are ordered lower. Identical |
| 205 | * parameter names are compared by their pointer to maintain a stable |
| 206 | * sort. |
| 207 | */ |
| 208 | static int query_param_cmp(const void *a, const void *b) |
| 209 | { |
| 210 | const struct ist param_a = *(struct ist*)a; |
| 211 | const struct ist param_b = *(struct ist*)b; |
| 212 | const struct ist param_a_name = iststop(param_a, '='); |
| 213 | const struct ist param_b_name = iststop(param_b, '='); |
| 214 | |
| 215 | int cmp = istdiff(param_a_name, param_b_name); |
| 216 | |
| 217 | if (cmp != 0) |
| 218 | return cmp; |
| 219 | |
| 220 | /* The contents are identical: Compare the pointer. */ |
| 221 | if (istptr(param_a) < istptr(param_b)) |
| 222 | return -1; |
| 223 | |
| 224 | if (istptr(param_a) > istptr(param_b)) |
| 225 | return 1; |
| 226 | |
| 227 | return 0; |
| 228 | } |
| 229 | |
| 230 | /* Sorts the parameters within the given query string. */ |
| 231 | enum uri_normalizer_err uri_normalizer_query_sort(const struct ist query, const char delim, struct ist *dst) |
| 232 | { |
| 233 | enum uri_normalizer_err err; |
| 234 | |
| 235 | const size_t size = istclear(dst); |
| 236 | struct ist newquery = *dst; |
| 237 | |
| 238 | struct ist scanner = query; |
| 239 | |
| 240 | const struct buffer *trash = get_trash_chunk(); |
| 241 | struct ist *params = (struct ist *)b_orig(trash); |
| 242 | const size_t max_param = b_size(trash) / sizeof(*params); |
| 243 | size_t param_count = 0; |
| 244 | |
| 245 | size_t i; |
| 246 | |
| 247 | /* The query will have the same length. */ |
| 248 | if (size < istlen(query)) { |
| 249 | err = URI_NORMALIZER_ERR_ALLOC; |
| 250 | goto fail; |
| 251 | } |
| 252 | |
| 253 | /* Handle the leading '?'. */ |
| 254 | newquery = __istappend(newquery, istshift(&scanner)); |
| 255 | |
| 256 | while (istlen(scanner) > 0) { |
| 257 | const struct ist param = istsplit(&scanner, delim); |
| 258 | |
| 259 | if (param_count + 1 > max_param) { |
| 260 | err = URI_NORMALIZER_ERR_ALLOC; |
| 261 | goto fail; |
| 262 | } |
| 263 | |
| 264 | params[param_count] = param; |
| 265 | param_count++; |
| 266 | } |
| 267 | |
| 268 | qsort(params, param_count, sizeof(*params), query_param_cmp); |
| 269 | |
| 270 | for (i = 0; i < param_count; i++) { |
| 271 | if (i > 0) |
| 272 | newquery = __istappend(newquery, '&'); |
| 273 | |
| 274 | if (istcat(&newquery, params[i], size) < 0) { |
| 275 | /* This is impossible, because we checked the size of the destination buffer. */ |
| 276 | my_unreachable(); |
| 277 | err = URI_NORMALIZER_ERR_INTERNAL_ERROR; |
| 278 | goto fail; |
| 279 | } |
| 280 | } |
| 281 | |
| 282 | *dst = newquery; |
| 283 | |
| 284 | return URI_NORMALIZER_ERR_NONE; |
| 285 | |
| 286 | fail: |
| 287 | |
| 288 | return err; |
| 289 | } |
Tim Duesterhus | d371e99 | 2021-04-15 21:45:58 +0200 | [diff] [blame] | 290 | |
Tim Duesterhus | dbd25c3 | 2021-04-15 21:45:55 +0200 | [diff] [blame] | 291 | /* |
| 292 | * Local variables: |
| 293 | * c-indent-level: 8 |
| 294 | * c-basic-offset: 8 |
| 295 | * End: |
| 296 | */ |