]>
git.proxmox.com Git - ovs.git/blob - ovn/lib/lex.c
2 * Copyright (c) 2015 Nicira, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at:
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
22 #include "dynamic-string.h"
26 /* Returns a string that represents 'format'. */
28 lex_format_to_string(enum lex_format format
)
33 case LEX_F_HEXADECIMAL
:
46 /* Initializes 'token'. */
48 lex_token_init(struct lex_token
*token
)
50 token
->type
= LEX_T_END
;
54 /* Frees memory owned by 'token'. */
56 lex_token_destroy(struct lex_token
*token
)
61 /* Exchanges 'a' and 'b'. */
63 lex_token_swap(struct lex_token
*a
, struct lex_token
*b
)
65 struct lex_token tmp
= *a
;
70 /* lex_token_format(). */
73 lex_token_n_zeros(enum lex_format format
)
76 case LEX_F_DECIMAL
: return offsetof(union mf_subvalue
, integer
);
77 case LEX_F_HEXADECIMAL
: return 0;
78 case LEX_F_IPV4
: return offsetof(union mf_subvalue
, ipv4
);
79 case LEX_F_IPV6
: return offsetof(union mf_subvalue
, ipv6
);
80 case LEX_F_ETHERNET
: return offsetof(union mf_subvalue
, mac
);
81 default: OVS_NOT_REACHED();
85 /* Returns the effective format for 'token', that is, the format in which it
86 * should actually be printed. This is ordinarily the same as 'token->format',
87 * but it's always possible that someone sets up a token with a format that
88 * won't work for a value, e.g. 'token->value' is wider than 32 bits but the
89 * format is LEX_F_IPV4. (The lexer itself won't do that; this is an attempt
90 * to avoid confusion in the future.) */
91 static enum lex_format
92 lex_token_get_format(const struct lex_token
*token
)
94 size_t n_zeros
= lex_token_n_zeros(token
->format
);
95 return (is_all_zeros(&token
->value
, n_zeros
)
96 && (token
->type
!= LEX_T_MASKED_INTEGER
97 || is_all_zeros(&token
->mask
, n_zeros
))
103 lex_token_format_value(const union mf_subvalue
*value
,
104 enum lex_format format
, struct ds
*s
)
108 ds_put_format(s
, "%"PRIu64
, ntohll(value
->integer
));
111 case LEX_F_HEXADECIMAL
:
112 mf_format_subvalue(value
, s
);
116 ds_put_format(s
, IP_FMT
, IP_ARGS(value
->ipv4
));
120 ipv6_format_addr(&value
->ipv6
, s
);
124 ds_put_format(s
, ETH_ADDR_FMT
, ETH_ADDR_ARGS(value
->mac
));
134 lex_token_format_masked_integer(const struct lex_token
*token
, struct ds
*s
)
136 enum lex_format format
= lex_token_get_format(token
);
138 lex_token_format_value(&token
->value
, format
, s
);
141 const union mf_subvalue
*mask
= &token
->mask
;
142 if (format
== LEX_F_IPV4
&& ip_is_cidr(mask
->ipv4
)) {
143 ds_put_format(s
, "%d", ip_count_cidr_bits(mask
->ipv4
));
144 } else if (token
->format
== LEX_F_IPV6
&& ipv6_is_cidr(&mask
->ipv6
)) {
145 ds_put_format(s
, "%d", ipv6_count_cidr_bits(&mask
->ipv6
));
147 lex_token_format_value(&token
->mask
, format
, s
);
151 /* Appends a string representation of 'token' to 's', in a format that can be
152 * losslessly parsed back by the lexer. (LEX_T_END and LEX_T_ERROR can't be
155 lex_token_format(const struct lex_token
*token
, struct ds
*s
)
157 switch (token
->type
) {
163 ds_put_cstr(s
, token
->s
);
167 ds_put_cstr(s
, "error(");
168 json_string_escape(token
->s
, s
);
173 json_string_escape(token
->s
, s
);
177 lex_token_format_value(&token
->value
, lex_token_get_format(token
), s
);
180 case LEX_T_MASKED_INTEGER
:
181 lex_token_format_masked_integer(token
, s
);
203 ds_put_cstr(s
, "==");
206 ds_put_cstr(s
, "!=");
212 ds_put_cstr(s
, "<=");
218 ds_put_cstr(s
, ">=");
224 ds_put_cstr(s
, "&&");
227 ds_put_cstr(s
, "||");
230 ds_put_cstr(s
, "..");
235 case LEX_T_SEMICOLON
:
242 ds_put_cstr(s
, "<->");
244 case LEX_T_DECREMENT
:
245 ds_put_cstr(s
, "--");
253 /* lex_token_parse(). */
255 static void OVS_PRINTF_FORMAT(2, 3)
256 lex_error(struct lex_token
*token
, const char *message
, ...)
258 ovs_assert(!token
->s
);
259 token
->type
= LEX_T_ERROR
;
262 va_start(args
, message
);
263 token
->s
= xvasprintf(message
, args
);
268 lex_parse_hex_integer(const char *start
, size_t len
, struct lex_token
*token
)
270 const char *in
= start
+ (len
- 1);
271 uint8_t *out
= token
->value
.u8
+ (sizeof token
->value
.u8
- 1);
273 for (int i
= 0; i
< len
; i
++) {
274 int hexit
= hexit_value(in
[-i
]);
276 lex_error(token
, "Invalid syntax in hexadecimal constant.");
279 if (hexit
&& i
/ 2 >= sizeof token
->value
.u8
) {
280 lex_error(token
, "Hexadecimal constant requires more than "
281 "%"PRIuSIZE
" bits.", 8 * sizeof token
->value
.u8
);
284 out
[-(i
/ 2)] |= i
% 2 ? hexit
<< 4 : hexit
;
286 token
->format
= LEX_F_HEXADECIMAL
;
290 lex_parse_integer__(const char *p
, struct lex_token
*token
)
292 lex_token_init(token
);
293 token
->type
= LEX_T_INTEGER
;
294 memset(&token
->value
, 0, sizeof token
->value
);
295 const char *start
= p
;
296 const char *end
= start
;
297 while (isalnum((unsigned char) *end
) || *end
== ':'
298 || (*end
== '.' && end
[1] != '.')) {
301 size_t len
= end
- start
;
307 lex_error(token
, "Integer constant expected.");
309 && ovs_scan(start
, ETH_ADDR_SCAN_FMT
"%n",
310 ETH_ADDR_SCAN_ARGS(mac
), &n
)
312 token
->value
.mac
= mac
;
313 token
->format
= LEX_F_ETHERNET
;
314 } else if (start
+ strspn(start
, "0123456789") == end
) {
315 if (p
[0] == '0' && len
> 1) {
316 lex_error(token
, "Decimal constants must not have leading zeros.");
318 unsigned long long int integer
;
322 integer
= strtoull(p
, &tail
, 10);
323 if (tail
!= end
|| errno
== ERANGE
) {
324 lex_error(token
, "Decimal constants must be less than 2**64.");
326 token
->value
.integer
= htonll(integer
);
327 token
->format
= LEX_F_DECIMAL
;
330 } else if (p
[0] == '0' && (p
[1] == 'x' || p
[1] == 'X')) {
332 lex_parse_hex_integer(start
+ 2, len
- 2, token
);
334 lex_error(token
, "Hex digits expected following 0%c.", p
[1]);
336 } else if (len
< INET6_ADDRSTRLEN
) {
337 char copy
[INET6_ADDRSTRLEN
];
338 memcpy(copy
, p
, len
);
342 struct in6_addr ipv6
;
343 if (inet_pton(AF_INET
, copy
, &ipv4
) == 1) {
344 token
->value
.ipv4
= ipv4
.s_addr
;
345 token
->format
= LEX_F_IPV4
;
346 } else if (inet_pton(AF_INET6
, copy
, &ipv6
) == 1) {
347 token
->value
.ipv6
= ipv6
;
348 token
->format
= LEX_F_IPV6
;
350 lex_error(token
, "Invalid numeric constant.");
353 lex_error(token
, "Invalid numeric constant.");
356 ovs_assert(token
->type
== LEX_T_INTEGER
|| token
->type
== LEX_T_ERROR
);
361 lex_parse_mask(const char *p
, struct lex_token
*token
)
363 struct lex_token mask
;
365 /* Parse just past the '/' as a second integer. Handle errors. */
366 p
= lex_parse_integer__(p
+ 1, &mask
);
367 if (mask
.type
== LEX_T_ERROR
) {
368 lex_token_swap(&mask
, token
);
369 lex_token_destroy(&mask
);
372 ovs_assert(mask
.type
== LEX_T_INTEGER
);
374 /* Now convert the value and mask into a masked integer token.
375 * We have a few special cases. */
376 token
->type
= LEX_T_MASKED_INTEGER
;
377 memset(&token
->mask
, 0, sizeof token
->mask
);
378 uint32_t prefix_bits
= ntohll(mask
.value
.integer
);
379 if (token
->format
== mask
.format
) {
380 /* Same format value and mask is always OK. */
381 token
->mask
= mask
.value
;
382 } else if (token
->format
== LEX_F_IPV4
383 && mask
.format
== LEX_F_DECIMAL
384 && prefix_bits
<= 32) {
385 /* IPv4 address with decimal mask is a CIDR prefix. */
386 token
->mask
.integer
= htonll(ntohl(be32_prefix_mask(prefix_bits
)));
387 } else if (token
->format
== LEX_F_IPV6
388 && mask
.format
== LEX_F_DECIMAL
389 && prefix_bits
<= 128) {
390 /* IPv6 address with decimal mask is a CIDR prefix. */
391 token
->mask
.ipv6
= ipv6_create_mask(prefix_bits
);
392 } else if (token
->format
== LEX_F_DECIMAL
393 && mask
.format
== LEX_F_HEXADECIMAL
394 && token
->value
.integer
== 0) {
395 /* Special case for e.g. 0/0x1234. */
396 token
->format
= LEX_F_HEXADECIMAL
;
397 token
->mask
= mask
.value
;
399 lex_error(token
, "Value and mask have incompatible formats.");
403 /* Check invariant that a 1-bit in the value corresponds to a 1-bit in the
405 for (int i
= 0; i
< ARRAY_SIZE(token
->mask
.be32
); i
++) {
406 ovs_be32 v
= token
->value
.be32
[i
];
407 ovs_be32 m
= token
->mask
.be32
[i
];
410 lex_error(token
, "Value contains unmasked 1-bits.");
416 lex_token_destroy(&mask
);
421 lex_parse_integer(const char *p
, struct lex_token
*token
)
423 p
= lex_parse_integer__(p
, token
);
424 if (token
->type
== LEX_T_INTEGER
&& *p
== '/') {
425 p
= lex_parse_mask(p
, token
);
431 lex_parse_string(const char *p
, struct lex_token
*token
)
433 const char *start
= ++p
;
437 lex_error(token
, "Input ends inside quoted string.");
441 token
->type
= (json_string_unescape(start
, p
- start
, &token
->s
)
442 ? LEX_T_STRING
: LEX_T_ERROR
);
460 lex_is_id1(unsigned char c
)
462 return ((c
>= 'a' && c
<= 'z') || (c
>= 'A' && c
<= 'Z')
463 || c
== '_' || c
== '.');
467 lex_is_idn(unsigned char c
)
469 return lex_is_id1(c
) || (c
>= '0' && c
<= '9');
473 lex_parse_id(const char *p
, struct lex_token
*token
)
475 const char *start
= p
;
479 } while (lex_is_idn(*p
));
481 token
->type
= LEX_T_ID
;
482 token
->s
= xmemdup0(start
, p
- start
);
486 /* Initializes 'token' and parses the first token from the beginning of
487 * null-terminated string 'p' into 'token'. Stores a pointer to the start of
488 * the token (after skipping white space and comments, if any) into '*startp'.
489 * Returns the character position at which to begin parsing the next token. */
491 lex_token_parse(struct lex_token
*token
, const char *p
, const char **startp
)
493 lex_token_init(token
);
499 token
->type
= LEX_T_END
;
502 case ' ': case '\t': case '\n': case '\r':
511 } while (*p
!= '\0' && *p
!= '\n');
513 } else if (*p
== '*') {
516 if (*p
== '*' && p
[1] == '/') {
519 } else if (*p
== '\0' || *p
== '\n') {
520 lex_error(token
, "`/*' without matching `*/'.");
529 "`/' is only valid as part of `//' or `/*'.");
534 token
->type
= LEX_T_LPAREN
;
539 token
->type
= LEX_T_RPAREN
;
544 token
->type
= LEX_T_LCURLY
;
549 token
->type
= LEX_T_RCURLY
;
554 token
->type
= LEX_T_LSQUARE
;
559 token
->type
= LEX_T_RSQUARE
;
566 token
->type
= LEX_T_EQ
;
569 token
->type
= LEX_T_EQUALS
;
576 token
->type
= LEX_T_NE
;
579 token
->type
= LEX_T_LOG_NOT
;
586 token
->type
= LEX_T_LOG_AND
;
589 lex_error(token
, "`&' is only valid as part of `&&'.");
596 token
->type
= LEX_T_LOG_OR
;
599 lex_error(token
, "`|' is only valid as part of `||'.");
606 token
->type
= LEX_T_LE
;
608 } else if (*p
== '-' && p
[1] == '>') {
609 token
->type
= LEX_T_EXCHANGE
;
612 token
->type
= LEX_T_LT
;
619 token
->type
= LEX_T_GE
;
622 token
->type
= LEX_T_GT
;
629 token
->type
= LEX_T_ELLIPSIS
;
632 lex_error(token
, "`.' is only valid as part of `..' or a number.");
638 token
->type
= LEX_T_COMMA
;
643 token
->type
= LEX_T_SEMICOLON
;
649 token
->type
= LEX_T_DECREMENT
;
652 lex_error(token
, "`-' is only valid as part of `--'.");
656 case '0': case '1': case '2': case '3': case '4':
657 case '5': case '6': case '7': case '8': case '9':
659 p
= lex_parse_integer(p
, token
);
663 p
= lex_parse_string(p
, token
);
666 case 'a': case 'b': case 'c': case 'd': case 'e': case 'f':
667 case 'A': case 'B': case 'C': case 'D': case 'E': case 'F':
668 /* We need to distinguish an Ethernet address or IPv6 address from an
669 * identifier. Fortunately, Ethernet addresses and IPv6 addresses that
670 * are ambiguous based on the first character, always start with hex
671 * digits followed by a colon, but identifiers never do. */
672 p
= (p
[strspn(p
, "0123456789abcdefABCDEF")] == ':'
673 ? lex_parse_integer(p
, token
)
674 : lex_parse_id(p
, token
));
678 if (lex_is_id1(*p
)) {
679 p
= lex_parse_id(p
, token
);
681 if (isprint((unsigned char) *p
)) {
682 lex_error(token
, "Invalid character `%c' in input.", *p
);
684 lex_error(token
, "Invalid byte 0x%d in input.", *p
);
694 /* Initializes 'lexer' for parsing 'input'.
696 * While the lexer is in use, 'input' must remain available, but the caller
697 * otherwise retains ownership of 'input'.
699 * The caller must call lexer_get() to obtain the first token. */
701 lexer_init(struct lexer
*lexer
, const char *input
)
703 lexer
->input
= input
;
705 lex_token_init(&lexer
->token
);
708 /* Frees storage associated with 'lexer'. */
710 lexer_destroy(struct lexer
*lexer
)
712 lex_token_destroy(&lexer
->token
);
715 /* Obtains the next token from 'lexer' into 'lexer->token', and returns the
716 * token's type. The caller may examine 'lexer->token' directly to obtain full
717 * information about the token. */
719 lexer_get(struct lexer
*lexer
)
721 lex_token_destroy(&lexer
->token
);
722 lexer
->input
= lex_token_parse(&lexer
->token
, lexer
->input
, &lexer
->start
);
723 return lexer
->token
.type
;
726 /* Returns the type of the next token that will be fetched by lexer_get(),
727 * without advancing 'lexer->token' to that token. */
729 lexer_lookahead(const struct lexer
*lexer
)
731 struct lex_token next
;
735 lex_token_parse(&next
, lexer
->input
, &start
);
737 lex_token_destroy(&next
);
741 /* If 'lexer''s current token has the given 'type', advances 'lexer' to the
742 * next token and returns true. Otherwise returns false. */
744 lexer_match(struct lexer
*lexer
, enum lex_type type
)
746 if (lexer
->token
.type
== type
) {
754 /* If 'lexer''s current token is the identifier given in 'id', advances 'lexer'
755 * to the next token and returns true. Otherwise returns false. */
757 lexer_match_id(struct lexer
*lexer
, const char *id
)
759 if (lexer
->token
.type
== LEX_T_ID
&& !strcmp(lexer
->token
.s
, id
)) {
768 lexer_is_int(const struct lexer
*lexer
)
770 return (lexer
->token
.type
== LEX_T_INTEGER
771 && lexer
->token
.format
== LEX_F_DECIMAL
772 && ntohll(lexer
->token
.value
.integer
) <= INT_MAX
);
776 lexer_get_int(struct lexer
*lexer
, int *value
)
778 if (lexer_is_int(lexer
)) {
779 *value
= ntohll(lexer
->token
.value
.integer
);