2 * SPDX-License-Identifier: BSD-2-Clause
4 * Copyright (c) 2013 David Chisnall
7 * This software was developed by SRI International and the University of
8 * Cambridge Computer Laboratory under DARPA/AFRL contract (FA8750-10-C-0237)
9 * ("CTSRD"), as part of the DARPA CRASH research programme.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
20 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 #ifndef _INPUT_BUFFER_HH_
34 #define _INPUT_BUFFER_HH_
39 #include <unordered_set>
46 typedef std::unique_ptr<expression> expression_ptr;
50 * Class encapsulating the input file. Can be used as a const char*, but has
51 * range checking. Attempting to access anything out of range will return a 0
52 * byte. The input buffer can be cheaply copied, without copying the
53 * underlying memory, however it is the user's responsibility to ensure that
54 * such copies do not persist beyond the lifetime of the underlying memory.
56 * This also contains methods for reporting errors and for consuming the token
61 friend class text_input_buffer;
64 * The buffer. This class doesn't own the buffer, but the
65 * mmap_input_buffer subclass does.
69 * The size of the buffer.
74 * The current place in the buffer where we are reading. This class
75 * keeps a separate size, pointer, and cursor so that we can move
76 * forwards and backwards and still have checks that we haven't fallen
81 * Private constructor. This is used to create input buffers that
82 * refer to the same memory, but have different cursors.
84 input_buffer(const char* b, int s, int c) : buffer(b), size(s),
88 * Returns the file name associated with this buffer.
90 virtual const std::string &filename() const
95 static std::unique_ptr<input_buffer> buffer_for_file(const std::string &path,
98 * Skips all characters in the input until the specified character is
103 * Parses up to a specified character and returns the intervening
104 * characters as a string.
106 std::string parse_to(char);
108 * Return whether all input has been consumed.
110 bool finished() { return cursor >= size; }
112 * Virtual destructor. Does nothing, but exists so that subclasses
113 * that own the memory can run cleanup code for deallocating it.
115 virtual ~input_buffer() {};
117 * Constructs an empty buffer.
119 input_buffer() : buffer(0), size(0), cursor(0) {}
121 * Constructs a new buffer with a specified memory region and size.
123 input_buffer(const char* b, int s) : buffer(b), size(s), cursor(0){}
125 * Returns a new input buffer referring into this input, clamped to the
126 * specified size. If the requested buffer would fall outside the
127 * range of this one, then it returns an empty buffer.
129 * The returned buffer shares the same underlying storage as the
130 * original. This is intended to be used for splitting up the various
131 * sections of a device tree blob. Requesting a size of 0 will give a
132 * buffer that extends to the end of the available memory.
134 input_buffer buffer_from_offset(int offset, int s=0);
136 * Dereferencing operator, allows the buffer to be treated as a char*
137 * and dereferenced to give a character. This returns a null byte if
138 * the cursor is out of range.
140 inline char operator*()
142 if (cursor >= size) { return '\0'; }
143 if (cursor < 0) { return '\0'; }
144 return buffer[cursor];
147 * Array subscripting operator, returns a character at the specified
148 * index offset from the current cursor. The offset may be negative,
149 * to reread characters that have already been read. If the current
150 * cursor plus offset is outside of the range, this returns a nul
153 inline char operator[](int offset)
155 if (cursor + offset >= size) { return '\0'; }
156 if (cursor + offset < 0) { return '\0'; }
157 return buffer[cursor + offset];
160 * Increments the cursor, iterating forward in the buffer.
162 inline input_buffer &operator++()
173 return buffer + size;
176 * Consumes a character. Moves the cursor one character forward if the
177 * next character matches the argument, returning true. If the current
178 * character does not match the argument, returns false.
180 inline bool consume(char c)
190 * Consumes a string. If the (null-terminated) string passed as the
191 * argument appears in the input, advances the cursor to the end and
192 * returns true. Returns false if the string does not appear at the
193 * current point in the input.
195 bool consume(const char *str);
197 * Reads an integer in base 8, 10, or 16. Returns true and advances
198 * the cursor to the end of the integer if the cursor points to an
199 * integer, returns false and does not move the cursor otherwise.
201 * The parsed value is returned via the argument.
203 bool consume_integer(unsigned long long &outInt);
205 * Reads an arithmetic expression (containing any of the normal C
206 * operators), evaluates it, and returns the result.
208 bool consume_integer_expression(unsigned long long &outInt);
210 * Consumes two hex digits and return the resulting byte via the first
211 * argument. If the next two characters are hex digits, returns true
212 * and advances the cursor. If not, then returns false and leaves the
215 bool consume_hex_byte(uint8_t &outByte);
217 * Template function that consumes a binary value in big-endian format
218 * from the input stream. Returns true and advances the cursor if
219 * there is a value of the correct size. This function assumes that
220 * all values must be natively aligned, and so advances the cursor to
221 * the correct alignment before reading.
224 bool consume_binary(T &out)
227 int type_size = sizeof(T);
228 if (cursor % type_size != 0)
230 align = type_size - (cursor % type_size);
232 if (size < cursor + align + type_size)
237 assert(cursor % type_size == 0);
239 for (int i=0 ; i<type_size ; ++i)
246 out |= (((T)buffer[cursor++]) & 0xff);
252 * Dumps the current cursor value and the unconsumed values in the
253 * input buffer to the standard error. This method is intended solely
260 * Explicit specialisation for reading a single byte.
263 inline bool input_buffer::consume_binary(uint8_t &out)
265 if (size < cursor + 1)
269 out = buffer[cursor++];
274 * An input buffer subclass used for parsing DTS files. This manages a stack
275 * of input buffers to handle /input/ operations.
277 class text_input_buffer
279 std::unordered_set<std::string> defines;
281 * The cursor is the input into the input stream where we are currently reading.
285 * The current stack of includes. The current input is always from the top
288 std::stack<std::shared_ptr<input_buffer>> input_stack;
292 const std::vector<std::string> include_paths;
294 * Reads forward past any spaces. The DTS format is not whitespace
295 * sensitive and so we want to scan past whitespace when reading it.
299 * Returns the character immediately after the current one.
301 * This method does not look between files.
305 * If a /include/ token is encountered, then look up the corresponding
306 * input file, push it onto the input stack, and continue.
308 void handle_include();
310 * The base directory for this file.
312 const std::string dir;
314 * The file where dependencies should be output.
319 * Construct a new text input buffer with the specified buffer as the start
320 * of parsing and the specified set of input paths for handling new
323 text_input_buffer(std::unique_ptr<input_buffer> &&b,
324 std::unordered_set<std::string> &&d,
325 std::vector<std::string> &&i,
326 const std::string directory,
328 : defines(d), include_paths(i), dir(directory), depfile(deps)
330 input_stack.push(std::move(b));
333 * Skips all characters in the input until the specified character is
338 * Parse an expression. If `stopAtParen` is set, then only parse a number
339 * or a parenthetical expression, otherwise assume that either is the
340 * left-hand side of a binary expression and try to parse the right-hand
343 expression_ptr parse_expression(bool stopAtParen=false);
345 * Parse a binary expression, having already parsed the right-hand side.
347 expression_ptr parse_binary_expression(expression_ptr lhs);
349 * Return whether all input has been consumed.
353 return input_stack.empty() ||
354 ((input_stack.size() == 1) && input_stack.top()->finished());
357 * Dereferencing operator. Returns the current character in the top input buffer.
359 inline char operator*()
361 if (input_stack.empty())
365 return *(*input_stack.top());
368 * Increments the cursor, iterating forward in the buffer.
370 inline text_input_buffer &operator++()
372 if (input_stack.empty())
377 auto &top = *input_stack.top();
386 * Consumes a character. Moves the cursor one character forward if the
387 * next character matches the argument, returning true. If the current
388 * character does not match the argument, returns false.
390 inline bool consume(char c)
400 * Consumes a string. If the (null-terminated) string passed as the
401 * argument appears in the input, advances the cursor to the end and
402 * returns true. Returns false if the string does not appear at the
403 * current point in the input.
405 * This method does not scan between files.
407 bool consume(const char *str)
409 if (input_stack.empty())
413 return input_stack.top()->consume(str);
416 * Reads an integer in base 8, 10, or 16. Returns true and advances
417 * the cursor to the end of the integer if the cursor points to an
418 * integer, returns false and does not move the cursor otherwise.
420 * The parsed value is returned via the argument.
422 * This method does not scan between files.
424 bool consume_integer(unsigned long long &outInt)
426 if (input_stack.empty())
430 return input_stack.top()->consume_integer(outInt);
433 * Reads an arithmetic expression (containing any of the normal C
434 * operators), evaluates it, and returns the result.
436 bool consume_integer_expression(unsigned long long &outInt);
438 * Consumes two hex digits and return the resulting byte via the first
439 * argument. If the next two characters are hex digits, returns true
440 * and advances the cursor. If not, then returns false and leaves the
443 * This method does not scan between files.
445 bool consume_hex_byte(uint8_t &outByte)
447 if (input_stack.empty())
451 return input_stack.top()->consume_hex_byte(outByte);
454 * Returns the longest string in the input buffer starting at the
455 * current cursor and composed entirely of characters that are valid in
458 std::string parse_node_name();
460 * Returns the longest string in the input buffer starting at the
461 * current cursor and composed entirely of characters that are valid in
464 std::string parse_property_name();
466 * Parses either a node or a property name. If is_property is true on
467 * entry, then only property names are parsed. If it is false, then it
468 * will be set, on return, to indicate whether the parsed name is only
469 * valid as a property.
471 std::string parse_node_or_property_name(bool &is_property);
473 * Parses up to a specified character and returns the intervening
474 * characters as a string.
476 std::string parse_to(char);
478 * Advances the cursor to the start of the next token, skipping
479 * comments and whitespace. If the cursor already points to the start
480 * of a token, then this function does nothing.
482 text_input_buffer &next_token();
484 * Location in the source file. This should never be interpreted by
485 * anything other than error reporting functions of this class. It will
486 * eventually become something more complex than an `int`.
488 class source_location
490 friend class text_input_buffer;
492 * The text buffer object that included `b`.
494 text_input_buffer &buffer;
496 * The underlying buffer that contains this location.
498 std::shared_ptr<input_buffer> b;
500 * The offset within the current buffer of the source location.
503 source_location(text_input_buffer &buf)
505 b(buf.input_stack.empty() ? nullptr : buf.input_stack.top()),
506 cursor(b ? b->cursor : 0) {}
509 * Report an error at this location.
511 void report_error(const char *msg)
515 buffer.parse_error(msg, *b, cursor);
519 buffer.parse_error(msg);
524 * Returns the current source location.
526 source_location location()
531 * Prints a message indicating the location of a parse error.
533 void parse_error(const char *msg);
535 * Reads the contents of a binary file into `b`. The file name is assumed
536 * to be relative to one of the include paths.
538 * Returns true if the file exists and can be read, false otherwise.
540 bool read_binary_file(const std::string &filename, byte_buffer &b);
543 * Prints a message indicating the location of a parse error, given a
544 * specified location. This is used when input has already moved beyond
545 * the location that caused the failure.
547 void parse_error(const char *msg, input_buffer &b, int loc);
552 #endif // !_INPUT_BUFFER_HH_