4 * Copyright IBM, Corp. 2009
9 * This work is licensed under the terms of the GNU LGPL, version 2.1 or later.
10 * See the COPYING.LIB file in the top-level directory.
14 #include "qemu/osdep.h"
15 #include "json-parser-int.h"
17 #define MAX_TOKEN_SIZE (64ULL << 20)
20 * From RFC 8259 "The JavaScript Object Notation (JSON) Data
21 * Interchange Format", with [comments in brackets]:
23 * The set of tokens includes six structural characters, strings,
24 * numbers, and three literal names.
26 * These are the six structural characters:
28 * begin-array = ws %x5B ws ; [ left square bracket
29 * begin-object = ws %x7B ws ; { left curly bracket
30 * end-array = ws %x5D ws ; ] right square bracket
31 * end-object = ws %x7D ws ; } right curly bracket
32 * name-separator = ws %x3A ws ; : colon
33 * value-separator = ws %x2C ws ; , comma
35 * Insignificant whitespace is allowed before or after any of the six
36 * structural characters.
37 * [This lexer accepts it before or after any token, which is actually
38 * the same, as the grammar always has structural characters between
43 * %x09 / ; Horizontal tab
44 * %x0A / ; Line feed or New line
45 * %x0D ) ; Carriage return
47 * [...] three literal names:
49 * [This lexer accepts [a-z]+, and leaves rejecting unknown literal
50 * names to the parser.]
54 * number = [ minus ] int [ frac ] [ exp ]
55 * decimal-point = %x2E ; .
56 * digit1-9 = %x31-39 ; 1-9
57 * e = %x65 / %x45 ; e E
58 * exp = e [ minus / plus ] 1*DIGIT
59 * frac = decimal-point 1*DIGIT
60 * int = zero / ( digit1-9 *DIGIT )
66 * string = quotation-mark *char quotation-mark
70 * %x22 / ; " quotation mark U+0022
71 * %x5C / ; \ reverse solidus U+005C
72 * %x2F / ; / solidus U+002F
73 * %x62 / ; b backspace U+0008
74 * %x66 / ; f form feed U+000C
75 * %x6E / ; n line feed U+000A
76 * %x72 / ; r carriage return U+000D
77 * %x74 / ; t tab U+0009
78 * %x75 4HEXDIG ) ; uXXXX U+XXXX
80 * quotation-mark = %x22 ; "
81 * unescaped = %x20-21 / %x23-5B / %x5D-10FFFF
82 * [This lexer accepts any non-control character after escape, and
83 * leaves rejecting invalid ones to the parser.]
86 * Extensions over RFC 8259:
87 * - Extra escape sequence in strings:
88 * 0x27 (apostrophe) is recognized after escape, too
89 * - Single-quoted strings:
90 * Like double-quoted strings, except they're delimited by %x27
91 * (apostrophe) instead of %x22 (quotation mark), and can't contain
92 * unescaped apostrophe, but can contain unescaped quotation mark.
93 * - Interpolation, if enabled:
94 * The lexer accepts %[A-Za-z0-9]*, and leaves rejecting invalid
98 * - Input must be encoded in modified UTF-8.
99 * - Decoding and validating is left to the parser.
102 enum json_lexer_state {
103 IN_ERROR = 0, /* must really be 0, see json_lexer[] */
120 IN_START_INTERP, /* must be IN_START + 1 */
123 QEMU_BUILD_BUG_ON((int)JSON_MIN <= (int)IN_START_INTERP);
124 QEMU_BUILD_BUG_ON(JSON_MAX >= 0x80);
125 QEMU_BUILD_BUG_ON(IN_START_INTERP != IN_START + 1);
127 #define LOOKAHEAD 0x80
128 #define TERMINAL(state) [0 ... 0xFF] = ((state) | LOOKAHEAD)
130 static const uint8_t json_lexer[][256] = {
131 /* Relies on default initialization to IN_ERROR! */
133 /* double quote string */
134 [IN_DQ_STRING_ESCAPE] = {
135 [0x20 ... 0xFD] = IN_DQ_STRING,
138 [0x20 ... 0xFD] = IN_DQ_STRING,
139 ['\\'] = IN_DQ_STRING_ESCAPE,
143 /* single quote string */
144 [IN_SQ_STRING_ESCAPE] = {
145 [0x20 ... 0xFD] = IN_SQ_STRING,
148 [0x20 ... 0xFD] = IN_SQ_STRING,
149 ['\\'] = IN_SQ_STRING_ESCAPE,
150 ['\''] = JSON_STRING,
155 TERMINAL(JSON_INTEGER),
156 ['0' ... '9'] = IN_ERROR,
162 TERMINAL(JSON_FLOAT),
163 ['0' ... '9'] = IN_EXP_DIGITS,
167 ['0' ... '9'] = IN_EXP_DIGITS,
173 ['0' ... '9'] = IN_EXP_DIGITS,
176 [IN_MANTISSA_DIGITS] = {
177 TERMINAL(JSON_FLOAT),
178 ['0' ... '9'] = IN_MANTISSA_DIGITS,
184 ['0' ... '9'] = IN_MANTISSA_DIGITS,
189 TERMINAL(JSON_INTEGER),
190 ['0' ... '9'] = IN_DIGITS,
198 ['1' ... '9'] = IN_DIGITS,
203 TERMINAL(JSON_KEYWORD),
204 ['a' ... 'z'] = IN_KEYWORD,
210 [' '] = IN_WHITESPACE,
211 ['\t'] = IN_WHITESPACE,
212 ['\r'] = IN_WHITESPACE,
213 ['\n'] = IN_WHITESPACE,
218 TERMINAL(JSON_INTERP),
219 ['A' ... 'Z'] = IN_INTERP,
220 ['a' ... 'z'] = IN_INTERP,
221 ['0' ... '9'] = IN_INTERP,
226 * - IN_START recognizes JSON tokens with our string extensions
227 * - IN_START_INTERP additionally recognizes interpolation.
229 [IN_START ... IN_START_INTERP] = {
230 ['"'] = IN_DQ_STRING,
231 ['\''] = IN_SQ_STRING,
233 ['1' ... '9'] = IN_DIGITS,
237 ['['] = JSON_LSQUARE,
238 [']'] = JSON_RSQUARE,
241 ['a' ... 'z'] = IN_KEYWORD,
242 [' '] = IN_WHITESPACE,
243 ['\t'] = IN_WHITESPACE,
244 ['\r'] = IN_WHITESPACE,
245 ['\n'] = IN_WHITESPACE,
247 [IN_START_INTERP]['%'] = IN_INTERP,
250 static inline uint8_t next_state(JSONLexer *lexer, char ch, bool flush,
255 assert(lexer->state <= ARRAY_SIZE(json_lexer));
256 next = json_lexer[lexer->state][(uint8_t)ch];
257 *char_consumed = !flush && !(next & LOOKAHEAD);
258 return next & ~LOOKAHEAD;
261 void json_lexer_init(JSONLexer *lexer, bool enable_interpolation)
263 lexer->start_state = lexer->state = enable_interpolation
264 ? IN_START_INTERP : IN_START;
265 lexer->token = g_string_sized_new(3);
266 lexer->x = lexer->y = 0;
269 static void json_lexer_feed_char(JSONLexer *lexer, char ch, bool flush)
272 bool char_consumed = false;
280 while (flush ? lexer->state != lexer->start_state : !char_consumed) {
281 new_state = next_state(lexer, ch, flush, &char_consumed);
284 g_string_append_c(lexer->token, ch);
299 json_message_process_token(lexer, lexer->token, new_state,
303 g_string_truncate(lexer->token, 0);
304 new_state = lexer->start_state;
307 /* XXX: To avoid having previous bad input leaving the parser in an
308 * unresponsive state where we consume unpredictable amounts of
309 * subsequent "good" input, percolate this error state up to the
310 * parser by emitting a JSON_ERROR token, then reset lexer state.
312 * Also note that this handling is required for reliable channel
313 * negotiation between QMP and the guest agent, since chr(0xFF)
314 * is placed at the beginning of certain events to ensure proper
315 * delivery when the channel is in an unknown state. chr(0xFF) is
316 * never a valid ASCII/UTF-8 sequence, so this should reliably
317 * induce an error/flush state.
319 json_message_process_token(lexer, lexer->token, JSON_ERROR,
321 g_string_truncate(lexer->token, 0);
322 lexer->state = lexer->start_state;
327 lexer->state = new_state;
330 /* Do not let a single token grow to an arbitrarily large size,
331 * this is a security consideration.
333 if (lexer->token->len > MAX_TOKEN_SIZE) {
334 json_message_process_token(lexer, lexer->token, lexer->state,
336 g_string_truncate(lexer->token, 0);
337 lexer->state = lexer->start_state;
341 void json_lexer_feed(JSONLexer *lexer, const char *buffer, size_t size)
345 for (i = 0; i < size; i++) {
346 json_lexer_feed_char(lexer, buffer[i], false);
350 void json_lexer_flush(JSONLexer *lexer)
352 json_lexer_feed_char(lexer, 0, true);
353 assert(lexer->state == lexer->start_state);
354 json_message_process_token(lexer, lexer->token, JSON_END_OF_INPUT,
358 void json_lexer_destroy(JSONLexer *lexer)
360 g_string_free(lexer->token, true);