blob: 2b5178f8c072324996feac25b33e5992c70859d8 [file] [log] [blame]
Willy Tarreaue11f7272017-05-30 17:49:36 +02001/*
2 * include/common/ist.h
3 * Very simple indirect string manipulation functions.
4 *
5 * Copyright (C) 2014-2017 Willy Tarreau - w@1wt.eu
6 *
7 * Permission is hereby granted, free of charge, to any person obtaining
8 * a copy of this software and associated documentation files (the
9 * "Software"), to deal in the Software without restriction, including
10 * without limitation the rights to use, copy, modify, merge, publish,
11 * distribute, sublicense, and/or sell copies of the Software, and to
12 * permit persons to whom the Software is furnished to do so, subject to
13 * the following conditions:
14 *
15 * The above copyright notice and this permission notice shall be
16 * included in all copies or substantial portions of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
19 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
20 * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
21 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
22 * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
23 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
24 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
25 * OTHER DEALINGS IN THE SOFTWARE.
26 */
27
28#ifndef _COMMON_IST_H
29#define _COMMON_IST_H
30
Christopher Faulet20761452018-06-06 16:33:53 +020031#include <ctype.h>
Willy Tarreaue11f7272017-05-30 17:49:36 +020032#include <string.h>
Willy Tarreaua7280a12018-11-26 19:41:40 +010033#include <unistd.h>
Willy Tarreaue11f7272017-05-30 17:49:36 +020034
Tim Duesterhus35005d02020-03-05 17:56:32 +010035#ifndef IST_FREESTANDING
36#include <stdlib.h>
37#endif
38
Willy Tarreaue11f7272017-05-30 17:49:36 +020039#include <common/config.h>
40
Willy Tarreaud6735d62018-12-07 09:40:01 +010041/* ASCII to lower case conversion table */
Willy Tarreau0f35c592019-05-15 16:07:36 +020042#define _IST_LC ((const unsigned char[256]){ \
43 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
44 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, \
45 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, \
46 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, \
47 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, \
48 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, \
49 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, \
50 0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, \
51 0x40, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, \
52 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, \
53 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, \
54 0x78, 0x79, 0x7a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f, \
55 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, \
56 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, \
57 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, \
58 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, \
59 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, \
60 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, \
61 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, \
62 0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, \
63 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, \
64 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, \
65 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, \
66 0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf, \
67 0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7, \
68 0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf, \
69 0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, \
70 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf, \
71 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, \
72 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, \
73 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, \
74 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, \
75})
Willy Tarreaud6735d62018-12-07 09:40:01 +010076
77/* ASCII to upper case conversion table */
Willy Tarreau0f35c592019-05-15 16:07:36 +020078#define _IST_UC ((const unsigned char[256]){ \
79 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, \
80 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, \
81 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, \
82 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, \
83 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, \
84 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, \
85 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, \
86 0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, \
87 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, \
88 0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, \
89 0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, \
90 0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f, \
91 0x60, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, \
92 0x48, 0x49, 0x4a, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, \
93 0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, \
94 0x58, 0x59, 0x5a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f, \
95 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, \
96 0x88, 0x89, 0x8a, 0x8b, 0x8c, 0x8d, 0x8e, 0x8f, \
97 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, \
98 0x98, 0x99, 0x9a, 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, \
99 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, \
100 0xa8, 0xa9, 0xaa, 0xab, 0xac, 0xad, 0xae, 0xaf, \
101 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, \
102 0xb8, 0xb9, 0xba, 0xbb, 0xbc, 0xbd, 0xbe, 0xbf, \
103 0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7, \
104 0xc8, 0xc9, 0xca, 0xcb, 0xcc, 0xcd, 0xce, 0xcf, \
105 0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, \
106 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf, \
107 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, \
108 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, \
109 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, \
110 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, \
111})
112
113#ifdef USE_OBSOLETE_LINKER
114/* some old linkers and some non-ELF platforms have issues with the weak
115 * attribute so we turn these arrays to literals there.
116 */
117#define ist_lc _IST_LC
118#define ist_uc _IST_UC
119#else
120const unsigned char ist_lc[256] __attribute__((weak)) = _IST_LC;
121const unsigned char ist_uc[256] __attribute__((weak)) = _IST_UC;
122#endif
Willy Tarreaud6735d62018-12-07 09:40:01 +0100123
Willy Tarreaue11f7272017-05-30 17:49:36 +0200124/* This string definition will most often be used to represent a read-only
125 * string returned from a function, based on the starting point and its length
126 * in bytes. No storage is provided, only a pointer and a length. The types
127 * here are important as we only want to have 2 native machine words there so
128 * that on modern architectures the compiler is capable of efficiently
129 * returning a register pair without having to allocate stack room from the
130 * caller. This is done with -freg-struct which is often enabled by default.
131 */
132struct ist {
133 char *ptr;
134 size_t len;
135};
136
Willy Tarreau2ba67272017-09-21 15:24:10 +0200137/* makes a constant ist from a constant string, for use in array declarations */
138#define IST(str) { .ptr = str "", .len = (sizeof str "") - 1 }
139
Tim Duesterhus241e29e2020-03-05 17:56:30 +0100140/* IST_NULL is equivalent to an `ist` with `.ptr = NULL` and `.len = 0` */
141#define IST_NULL ((const struct ist){ .ptr = 0, .len = 0 })
142
Willy Tarreaue11f7272017-05-30 17:49:36 +0200143/* makes an ist from a regular zero terminated string. Null has length 0.
144 * Constants are detected and replaced with constant initializers. Other values
145 * are measured by hand without strlen() as it's much cheaper and inlinable on
146 * small strings. The construct is complex because we must never call
147 * __builtin_strlen() with an expression otherwise it involves a real
148 * measurement.
149 */
150#if __GNUC__ >= 4
151// gcc >= 4 detects constant propagation of str through __x and resolves the
152// length of constant strings easily.
153#define ist(str) ({ \
154 char *__x = (void *)(str); \
155 (struct ist){ \
156 .ptr = __x, \
157 .len = __builtin_constant_p(str) ? \
158 ((void *)str == (void *)0) ? 0 : \
159 __builtin_strlen(__x) : \
160 ({ \
161 size_t __l = 0; \
162 if (__x) for (__l--; __x[++__l]; ) ; \
163 __l; \
164 }) \
165 }; \
166})
167#else
168// gcc < 4 can't do this, and the side effect is a warning each time a NULL is
169// passed to ist() due to the check on __builtin_strlen(). It doesn't have the
170// ability to know that this code is never called.
171#define ist(str) ({ \
172 char *__x = (void *)(str); \
173 (struct ist){ \
174 .ptr = __x, \
175 .len = __builtin_constant_p(str) ? \
176 ((void *)str == (void *)0) ? 0 : \
177 __builtin_strlen(str) : \
178 ({ \
179 size_t __l = 0; \
180 if (__x) for (__l--; __x[++__l]; ) ; \
181 __l; \
182 }) \
183 }; \
184})
185#endif
186
187/* makes an ist struct from a string and a length */
188static inline struct ist ist2(const void *ptr, size_t len)
189{
190 return (struct ist){ .ptr = (char *)ptr, .len = len };
191}
192
Tim Duesterhuse296d3e2020-03-05 17:56:31 +0100193/* returns the result of `ist.ptr != NULL` */
194static inline int isttest(const struct ist ist)
195{
196 return ist.ptr != NULL;
197}
198
Willy Tarreaue67c4e52017-10-19 06:28:23 +0200199/* This function MODIFIES the string to add a zero AFTER the end, and returns
200 * the start pointer. The purpose is to use it on strings extracted by parsers
201 * from larger strings cut with delimiters that are not important and can be
202 * destroyed. It allows any such string to be used with regular string
203 * functions. It's also convenient to use with printf() to show data extracted
204 * from writable areas. The caller is obviously responsible for ensuring that
205 * the string is valid and that the first byte past the end is writable. If
206 * these conditions cannot be satisfied, use istpad() below instead.
207 */
208static inline char *ist0(struct ist ist)
209{
210 ist.ptr[ist.len] = 0;
211 return ist.ptr;
212}
213
Christopher Faulet04179752020-04-21 10:46:43 +0200214/* returns the pointer of the string */
215static inline char *istptr(const struct ist ist)
216{
217 return ist.ptr;
218}
219
Willy Tarreaue11f7272017-05-30 17:49:36 +0200220/* returns the length of the string */
221static inline size_t istlen(const struct ist ist)
222{
223 return ist.len;
224}
225
226/* skips to next character in the string, always stops at the end */
227static inline struct ist istnext(const struct ist ist)
228{
229 struct ist ret = ist;
230
231 if (ret.len) {
232 ret.len--;
233 ret.ptr++;
234 }
235 return ret;
236}
237
238/* copies the contents from string <ist> to buffer <buf> and adds a trailing
239 * zero. The caller must ensure <buf> is large enough.
240 */
241static inline struct ist istpad(void *buf, const struct ist ist)
242{
243 struct ist ret = { .ptr = buf, .len = ist.len };
244
245 for (ret.len = 0; ret.len < ist.len; ret.len++)
246 ret.ptr[ret.len] = ist.ptr[ret.len];
247
248 ret.ptr[ret.len] = 0;
249 return ret;
250}
251
252/* trims string <ist> to no more than <size> characters. The string is
253 * returned.
254 */
255static inline struct ist isttrim(const struct ist ist, size_t size)
256{
257 struct ist ret = ist;
258
259 if (ret.len > size)
260 ret.len = size;
261 return ret;
262}
263
264/* trims string <ist> to no more than <size>-1 characters and ensures that a
265 * zero is placed after <ist.len> (possibly reduced by one) and before <size>,
266 * unless <size> is already zero. The string is returned. This is mostly aimed
267 * at building printable strings that need to be zero-terminated.
268 */
269static inline struct ist istzero(const struct ist ist, size_t size)
270{
271 struct ist ret = ist;
272
273 if (!size)
274 ret.len = 0;
275 else {
276 if (ret.len > size - 1)
277 ret.len = size - 1;
278 ret.ptr[ret.len] = 0;
279 }
280 return ret;
281}
282
283/* returns the ordinal difference between two strings :
284 * < 0 if ist1 < ist2
285 * = 0 if ist1 == ist2
286 * > 0 if ist1 > ist2
287 */
288static inline int istdiff(const struct ist ist1, const struct ist ist2)
289{
290 struct ist l = ist1;
291 struct ist r = ist2;
292
293 do {
294 if (!l.len--)
295 return -r.len;
296 if (!r.len--)
297 return 1;
298 } while (*l.ptr++ == *r.ptr++);
299
300 return *(unsigned char *)(l.ptr - 1) - *(unsigned char *)(r.ptr - 1);
301}
302
303/* returns non-zero if <ist1> starts like <ist2> (empty strings do match) */
304static inline int istmatch(const struct ist ist1, const struct ist ist2)
305{
306 struct ist l = ist1;
307 struct ist r = ist2;
308
309 if (l.len < r.len)
310 return 0;
311
312 while (r.len--) {
313 if (*l.ptr++ != *r.ptr++)
314 return 0;
315 }
316 return 1;
317}
318
319/* returns non-zero if <ist1> starts like <ist2> on the first <count>
320 * characters (empty strings do match).
321 */
322static inline int istnmatch(const struct ist ist1, const struct ist ist2, size_t count)
323{
324 struct ist l = ist1;
325 struct ist r = ist2;
326
327 if (l.len > count)
328 l.len = count;
329 if (r.len > count)
330 r.len = count;
331 return istmatch(l, r);
332}
333
334/* returns non-zero if <ist1> equals <ist2> (empty strings are equal) */
335static inline int isteq(const struct ist ist1, const struct ist ist2)
336{
337 struct ist l = ist1;
338 struct ist r = ist2;
339
340 if (l.len != r.len)
341 return 0;
342
343 while (l.len--) {
344 if (*l.ptr++ != *r.ptr++)
345 return 0;
346 }
347 return 1;
348}
349
Christopher Faulet20761452018-06-06 16:33:53 +0200350/* returns non-zero if <ist1> equals <ist2>, ignoring the case (empty strings are equal) */
351static inline int isteqi(const struct ist ist1, const struct ist ist2)
352{
353 struct ist l = ist1;
354 struct ist r = ist2;
355
356 if (l.len != r.len)
357 return 0;
358
359 while (l.len--) {
Willy Tarreaud6735d62018-12-07 09:40:01 +0100360 if (*l.ptr != *r.ptr &&
361 ist_lc[(unsigned char)*l.ptr] != ist_lc[(unsigned char)*r.ptr])
Christopher Faulet20761452018-06-06 16:33:53 +0200362 return 0;
Willy Tarreaud6735d62018-12-07 09:40:01 +0100363
Christopher Faulet20761452018-06-06 16:33:53 +0200364 l.ptr++;
365 r.ptr++;
366 }
367 return 1;
368}
369
Willy Tarreaue11f7272017-05-30 17:49:36 +0200370/* returns non-zero if <ist1> equals <ist2> on the first <count> characters
371 * (empty strings are equal).
372 */
373static inline int istneq(const struct ist ist1, const struct ist ist2, size_t count)
374{
375 struct ist l = ist1;
376 struct ist r = ist2;
377
378 if (l.len > count)
379 l.len = count;
380 if (r.len > count)
381 r.len = count;
382 return isteq(l, r);
383}
384
385/* copies <src> over <dst> for a maximum of <count> bytes. Returns the number
386 * of characters copied (src.len), or -1 if it does not fit. In all cases, the
387 * contents are copied prior to reporting an error, so that the destination
388 * at least contains a valid but truncated string.
389 */
390static inline ssize_t istcpy(struct ist *dst, const struct ist src, size_t count)
391{
392 dst->len = 0;
393
394 if (count > src.len)
395 count = src.len;
396
397 while (dst->len < count) {
398 dst->ptr[dst->len] = src.ptr[dst->len];
399 dst->len++;
400 }
401
402 if (dst->len == src.len)
403 return src.len;
404
405 return -1;
406}
407
408/* copies <src> over <dst> for a maximum of <count> bytes. Returns the number
409 * of characters copied, or -1 if it does not fit. A (possibly truncated) valid
410 * copy of <src> is always left into <dst>, and a trailing \0 is appended as
411 * long as <count> is not null, even if that results in reducing the string by
412 * one character.
413 */
414static inline ssize_t istscpy(struct ist *dst, const struct ist src, size_t count)
415{
416 dst->len = 0;
417
418 if (!count)
419 goto fail;
420
421 if (count > src.len)
422 count = src.len + 1;
423
424 while (dst->len < count - 1) {
425 dst->ptr[dst->len] = src.ptr[dst->len];
426 dst->len++;
427 }
428
429 dst->ptr[dst->len] = 0;
430 if (dst->len == src.len)
431 return src.len;
432 fail:
433 return -1;
434}
435
436/* appends <src> after <dst> for a maximum of <count> total bytes in <dst> after
437 * the copy. <dst> is assumed to be <count> or less before the call. The new
438 * string's length is returned, or -1 if a truncation happened. In all cases,
439 * the contents are copied prior to reporting an error, so that the destination
440 * at least contains a valid but truncated string.
441 */
442static inline ssize_t istcat(struct ist *dst, const struct ist src, size_t count)
443{
444 const char *s = src.ptr;
445
446 while (dst->len < count && s != src.ptr + src.len)
447 dst->ptr[dst->len++] = *s++;
448
449 if (s == src.ptr + src.len)
450 return dst->len;
451
452 return -1;
453}
454
455/* appends <src> after <dst> for a maximum of <count> total bytes in <dst> after
456 * the copy. <dst> is assumed to be <count> or less before the call. The new
457 * string's length is returned, or -1 if a truncation happened. In all cases,
458 * the contents are copied prior to reporting an error, so that the destination
459 * at least contains a valid but truncated string.
460 */
461static inline ssize_t istscat(struct ist *dst, const struct ist src, size_t count)
462{
463 const char *s = src.ptr;
464
465 if (!count)
466 goto fail;
467
468 while (dst->len < count - 1 && s != src.ptr + src.len) {
469 dst->ptr[dst->len++] = *s++;
470 }
471
472 dst->ptr[dst->len] = 0;
473 if (s == src.ptr + src.len)
474 return dst->len;
475 fail:
476 return -1;
477}
478
Willy Tarreau3f2d6962018-12-07 08:35:07 +0100479/* copies the entire <src> over <dst>, which must be allocated large enough to
480 * hold the whole contents. No trailing zero is appended, this is mainly used
481 * for protocol processing where the frame length has already been checked. An
482 * ist made of the output and its length are returned. The destination is not
483 * touched if src.len is null.
484 */
485static inline struct ist ist2bin(char *dst, const struct ist src)
486{
487 size_t ofs = 0;
488
489 /* discourage the compiler from trying to optimize for large strings,
490 * but tell it that most of our strings are not empty.
491 */
492 if (__builtin_expect(ofs < src.len, 1)) {
493 do {
494 dst[ofs] = src.ptr[ofs];
495 ofs++;
496 } while (__builtin_expect(ofs < src.len, 0));
497 }
498 return ist2(dst, ofs);
499}
500
501/* copies the entire <src> over <dst>, which must be allocated large enough to
502 * hold the whole contents as well as a trailing zero which is always appended.
503 * This is mainly used for protocol conversions where the frame length has
504 * already been checked. An ist made of the output and its length (not counting
505 * the trailing zero) are returned.
506 */
507static inline struct ist ist2str(char *dst, const struct ist src, size_t count)
508{
509 size_t ofs = 0;
510
511 /* discourage the compiler from trying to optimize for large strings,
512 * but tell it that most of our strings are not empty.
513 */
514 if (__builtin_expect(ofs < src.len, 1)) {
515 do {
516 dst[ofs] = src.ptr[ofs];
517 ofs++;
518 } while (__builtin_expect(ofs < src.len, 0));
519 }
520 dst[ofs] = 0;
521 return ist2(dst, ofs);
522}
523
524/* makes a lower case copy of the entire <src> into <dst>, which must have been
525 * allocated large enough to hold the whole contents. No trailing zero is
526 * appended, this is mainly used for protocol processing where the frame length
527 * has already been checked. An ist made of the output and its length are
528 * returned. The destination is not touched if src.len is null.
529 */
530static inline struct ist ist2bin_lc(char *dst, const struct ist src)
531{
532 size_t ofs = 0;
533
534 /* discourage the compiler from trying to optimize for large strings,
535 * but tell it that most of our strings are not empty.
536 */
537 if (__builtin_expect(ofs < src.len, 1)) {
538 do {
Willy Tarreaud6735d62018-12-07 09:40:01 +0100539 dst[ofs] = ist_lc[(unsigned char)src.ptr[ofs]];
Willy Tarreau3f2d6962018-12-07 08:35:07 +0100540 ofs++;
541 } while (__builtin_expect(ofs < src.len, 0));
542 }
543 return ist2(dst, ofs);
544}
545
546/* makes a lower case copy of the entire <src> into <dst>, which must have been
547 * allocated large enough to hold the whole contents as well as a trailing zero
548 * which is always appended. This is mainly used for protocol conversions where
549 * the frame length has already been checked. An ist made of the output and its
550 * length (not counting the trailing zero) are returned.
551 */
552static inline struct ist ist2str_lc(char *dst, const struct ist src, size_t count)
553{
554 size_t ofs = 0;
555
556 /* discourage the compiler from trying to optimize for large strings,
557 * but tell it that most of our strings are not empty.
558 */
559 if (__builtin_expect(ofs < src.len, 1)) {
560 do {
Willy Tarreaud6735d62018-12-07 09:40:01 +0100561 dst[ofs] = ist_lc[(unsigned char)src.ptr[ofs]];
Willy Tarreau3f2d6962018-12-07 08:35:07 +0100562 ofs++;
563 } while (__builtin_expect(ofs < src.len, 0));
564 }
565 dst[ofs] = 0;
566 return ist2(dst, ofs);
567}
568
569/* makes an upper case copy of the entire <src> into <dst>, which must have
570 * been allocated large enough to hold the whole contents. No trailing zero is
571 * appended, this is mainly used for protocol processing where the frame length
572 * has already been checked. An ist made of the output and its length are
573 * returned. The destination is not touched if src.len is null.
574 */
575static inline struct ist ist2bin_uc(char *dst, const struct ist src)
576{
577 size_t ofs = 0;
578
579 /* discourage the compiler from trying to optimize for large strings,
580 * but tell it that most of our strings are not empty.
581 */
582 if (__builtin_expect(ofs < src.len, 1)) {
583 do {
Willy Tarreaud6735d62018-12-07 09:40:01 +0100584 dst[ofs] = ist_uc[(unsigned char)src.ptr[ofs]];
Willy Tarreau3f2d6962018-12-07 08:35:07 +0100585 ofs++;
586 } while (__builtin_expect(ofs < src.len, 0));
587 }
588 return ist2(dst, ofs);
589}
590
591/* makes an upper case copy of the entire <src> into <dst>, which must have been
592 * allocated large enough to hold the whole contents as well as a trailing zero
593 * which is always appended. This is mainly used for protocol conversions where
594 * the frame length has already been checked. An ist made of the output and its
595 * length (not counting the trailing zero) are returned.
596 */
597static inline struct ist ist2str_uc(char *dst, const struct ist src, size_t count)
598{
599 size_t ofs = 0;
600
601 /* discourage the compiler from trying to optimize for large strings,
602 * but tell it that most of our strings are not empty.
603 */
604 if (__builtin_expect(ofs < src.len, 1)) {
605 do {
Willy Tarreaud6735d62018-12-07 09:40:01 +0100606 dst[ofs] = ist_uc[(unsigned char)src.ptr[ofs]];
Willy Tarreau3f2d6962018-12-07 08:35:07 +0100607 ofs++;
608 } while (__builtin_expect(ofs < src.len, 0));
609 }
610 dst[ofs] = 0;
611 return ist2(dst, ofs);
612}
613
Willy Tarreaue11f7272017-05-30 17:49:36 +0200614/* looks for first occurrence of character <chr> in string <ist>. Returns the
615 * pointer if found, or NULL if not found.
616 */
617static inline char *istchr(const struct ist ist, char chr)
618{
619 char *s = ist.ptr;
620
621 do {
622 if (s >= ist.ptr + ist.len)
623 return NULL;
624 } while (*s++ != chr);
625 return s - 1;
626}
627
Willy Tarreau8f3ce062019-11-22 15:58:53 +0100628/* Returns a pointer to the first control character found in <ist>, or NULL if
629 * none is present. A control character is defined as a byte whose value is
630 * between 0x00 and 0x1F included. The function is optimized for strings having
631 * no CTL chars by processing up to sizeof(long) bytes at once on architectures
632 * supporting efficient unaligned accesses. Despite this it is not very fast
633 * (~0.43 byte/cycle) and should mostly be used on low match probability when
634 * it can save a call to a much slower function.
635 */
636static inline const char *ist_find_ctl(const struct ist ist)
637{
638 const union { unsigned long v; } __attribute__((packed)) *u;
639 const char *curr = (void *)ist.ptr - sizeof(long);
640 const char *last = curr + ist.len;
641 unsigned long l1, l2;
642
643 do {
644 curr += sizeof(long);
645 if (curr > last)
646 break;
647 u = (void *)curr;
648 /* subtract 0x202020...20 to the value to generate a carry in
649 * the lower byte if the byte contains a lower value. If we
650 * generate a bit 7 that was not there, it means the byte was
651 * within 0x00..0x1F.
652 */
653 l2 = u->v;
654 l1 = ~l2 & ((~0UL / 255) * 0x80); /* 0x808080...80 */
655 l2 -= (~0UL / 255) * 0x20; /* 0x202020...20 */
656 } while ((l1 & l2) == 0);
657
658 last += sizeof(long);
659 if (__builtin_expect(curr < last, 0)) {
660 do {
661 if ((uint8_t)*curr < 0x20)
662 return curr;
663 curr++;
664 } while (curr < last);
665 }
666 return NULL;
667}
668
Willy Tarreaue11f7272017-05-30 17:49:36 +0200669/* looks for first occurrence of character <chr> in string <ist> and returns
670 * the tail of the string starting with this character, or (ist.end,0) if not
671 * found.
672 */
673static inline struct ist istfind(const struct ist ist, char chr)
674{
675 struct ist ret = ist;
676
677 while (ret.len--) {
678 if (*ret.ptr++ == chr)
679 return ist2(ret.ptr - 1, ret.len + 1);
680 }
681 return ist2(ret.ptr, 0);
682}
683
684/* looks for first occurrence of character different from <chr> in string <ist>
685 * and returns the tail of the string starting at this character, or (ist_end,0)
686 * if not found.
687 */
688static inline struct ist istskip(const struct ist ist, char chr)
689{
690 struct ist ret = ist;
691
692 while (ret.len--) {
693 if (*ret.ptr++ != chr)
694 return ist2(ret.ptr - 1, ret.len + 1);
695 }
696 return ist2(ret.ptr, 0);
697}
698
699/* looks for first occurrence of string <pat> in string <ist> and returns the
700 * tail of the string starting at this position, or (NULL,0) if not found. The
701 * empty pattern is found everywhere.
702 */
703static inline struct ist istist(const struct ist ist, const struct ist pat)
704{
705 struct ist ret = ist;
706 size_t pos;
707
708 if (!pat.len)
709 return ret;
710
711 while (1) {
712 loop:
713 ret = istfind(ret, *pat.ptr);
714 if (ret.len < pat.len)
715 break;
716
717 /* ret.len >= 1, pat.len >= 1 and *ret.ptr == *pat.ptr */
718
719 ret = istnext(ret);
720 for (pos = 0; pos < pat.len - 1; ) {
721 ++pos;
722 if (ret.ptr[pos - 1] != pat.ptr[pos])
723 goto loop;
724 }
725 return ist2(ret.ptr - 1, ret.len + 1);
726 }
Tim Duesterhus241e29e2020-03-05 17:56:30 +0100727 return IST_NULL;
Willy Tarreaue11f7272017-05-30 17:49:36 +0200728}
729
Jerome Magnin9dde0b22020-02-21 10:33:12 +0100730/*
Ilya Shipitsin77e3b4a2020-03-10 12:06:11 +0500731 * looks for the first occurrence of <chr> in string <ist> and returns a shorter
Jerome Magnin9dde0b22020-02-21 10:33:12 +0100732 * ist if char is found.
733 */
734static inline struct ist iststop(const struct ist ist, char chr)
735{
736 size_t len = 0;
737
738 while (len++ < ist.len && ist.ptr[len - 1] != chr)
739 ;
740 return ist2(ist.ptr, len - 1);
741}
Tim Duesterhus35005d02020-03-05 17:56:32 +0100742
743#ifndef IST_FREESTANDING
744/* This function allocates <size> bytes and returns an `ist` pointing to
745 * the allocated area with size `0`.
746 *
747 * If this function fails to allocate memory the return value is equivalent
748 * to IST_NULL.
749 */
750static inline struct ist istalloc(const size_t size)
751{
752 return ist2(malloc(size), 0);
753}
754
755/* This function performs the equivalent of free() on the given <ist>.
756 *
757 * After this function returns the value of the given <ist> will be
758 * modified to be equivalent to IST_NULL.
759 */
760static inline void istfree(struct ist *ist)
761{
762 free(ist->ptr);
763 *ist = IST_NULL;
764}
Tim Duesterhus9576ab72020-03-05 17:56:34 +0100765
766/* This function performs the equivalent of strdup() on the given <src>.
767 *
768 * If this function fails to allocate memory the return value is equivalent
769 * to IST_NULL.
770 */
771static inline struct ist istdup(const struct ist src)
772{
773 const size_t src_size = src.len;
774
775 /* Allocate at least 1 byte to allow duplicating an empty string with
776 * malloc implementations that return NULL for a 0-size allocation.
777 */
778 struct ist dst = istalloc(MAX(src_size, 1));
779
780 if (isttest(dst)) {
781 istcpy(&dst, src, src_size);
782 }
783
784 return dst;
785}
Tim Duesterhus35005d02020-03-05 17:56:32 +0100786#endif
787
Willy Tarreaue11f7272017-05-30 17:49:36 +0200788#endif