xref: /freebsd/usr.bin/dtc/input_buffer.hh (revision 287472b39c7985d968be84ea145c3e75a3e6b875)
1 /*-
2  * Copyright (c) 2013 David Chisnall
3  * All rights reserved.
4  *
5  * This software was developed by SRI International and the University of
6  * Cambridge Computer Laboratory under DARPA/AFRL contract (FA8750-10-C-0237)
7  * ("CTSRD"), as part of the DARPA CRASH research programme.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
19  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
22  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28  * SUCH DAMAGE.
29  *
30  * $FreeBSD$
31  */
32 
33 #ifndef _INPUT_BUFFER_HH_
34 #define _INPUT_BUFFER_HH_
35 #include "util.hh"
36 #include <assert.h>
37 
38 namespace dtc
39 {
40 
41 /**
42  * Class encapsulating the input file.  Can be used as a const char*, but has
43  * range checking.  Attempting to access anything out of range will return a 0
44  * byte.  The input buffer can be cheaply copied, without copying the
45  * underlying memory, however it is the user's responsibility to ensure that
46  * such copies do not persist beyond the lifetime of the underlying memory.
47  *
48  * This also contains methods for reporting errors and for consuming the token
49  * stream.
50  */
51 class input_buffer
52 {
53 	protected:
54 	/**
55 	 * The buffer.  This class doesn't own the buffer, but the
56 	 * mmap_input_buffer subclass does.
57 	 */
58 	const char* buffer;
59 	/**
60 	 * The size of the buffer.
61 	 */
62 	int size;
63 	private:
64 	/**
65 	 * The current place in the buffer where we are reading.  This class
66 	 * keeps a separate size, pointer, and cursor so that we can move
67 	 * forwards and backwards and still have checks that we haven't fallen
68 	 * off either end.
69 	 */
70 	int cursor;
71 	/**
72 	 * Private constructor.  This is used to create input buffers that
73 	 * refer to the same memory, but have different cursors.
74 	 */
75 	input_buffer(const char* b, int s, int c) : buffer(b), size(s),
76 		cursor(c) {}
77 	/**
78 	 * Reads forward past any spaces.  The DTS format is not whitespace
79 	 * sensitive and so we want to scan past whitespace when reading it.
80 	 */
81 	void skip_spaces();
82 	public:
83 	/**
84 	 * Virtual destructor.  Does nothing, but exists so that subclasses
85 	 * that own the memory can run cleanup code for deallocating it.
86 	 */
87 	virtual ~input_buffer() {};
88 	/**
89 	 * Constructs an empty buffer.
90 	 */
91 	input_buffer() : buffer(0), size(0), cursor(0) {}
92 	/**
93 	 * Constructs a new buffer with a specified memory region and size.
94 	 */
95 	input_buffer(const char* b, int s) : buffer(b), size(s), cursor(0){}
96 	/**
97 	 * Returns a new input buffer referring into this input, clamped to the
98 	 * specified size.  If the requested buffer would fall outside the
99 	 * range of this one, then it returns an empty buffer.
100 	 *
101 	 * The returned buffer shares the same underlying storage as the
102 	 * original.  This is intended to be used for splitting up the various
103 	 * sections of a device tree blob.  Requesting a size of 0 will give a
104 	 * buffer that extends to the end of the available memory.
105 	 */
106 	input_buffer buffer_from_offset(int offset, int s=0);
107 	/**
108 	 * Returns true if this buffer has no unconsumed space in it.
109 	 */
110 	inline bool empty()
111 	{
112 		return cursor >= size;
113 	}
114 	/**
115 	 * Dereferencing operator, allows the buffer to be treated as a char*
116 	 * and dereferenced to give a character.  This returns a null byte if
117 	 * the cursor is out of range.
118 	 */
119 	inline char operator*()
120 	{
121 		if (cursor >= size) { return '\0'; }
122 		if (cursor < 0) { return '\0'; }
123 		return buffer[cursor];
124 	}
125 	/**
126 	 * Array subscripting operator, returns a character at the specified
127 	 * index offset from the current cursor.  The offset may be negative,
128 	 * to reread characters that have already been read.  If the current
129 	 * cursor plus offset is outside of the range, this returns a nul
130 	 * byte.
131 	 */
132 	inline char operator[](int offset)
133 	{
134 		if (cursor + offset >= size) { return '\0'; }
135 		if (cursor + offset < 0) { return '\0'; }
136 		return buffer[cursor + offset];
137 	}
138 	/**
139 	 * Increments the cursor, iterating forward in the buffer.
140 	 */
141 	inline input_buffer &operator++()
142 	{
143 		cursor++;
144 		return *this;
145 	}
146 	/**
147 	 * Cast to char* operator.  Returns a pointer into the buffer that can
148 	 * be used for constructing strings.
149 	 */
150 	inline operator const char*()
151 	{
152 		if (cursor >= size) { return 0; }
153 		if (cursor < 0) { return 0; }
154 		return &buffer[cursor];
155 	}
156 	/**
157 	 * Consumes a character.  Moves the cursor one character forward if the
158 	 * next character matches the argument, returning true.  If the current
159 	 * character does not match the argument, returns false.
160 	 */
161 	inline bool consume(char c)
162 	{
163 		if ((*this)[0] == c)
164 		{
165 			++(*this);
166 			return true;
167 		}
168 		return false;
169 	}
170 	/**
171 	 * Consumes a string.  If the (null-terminated) string passed as the
172 	 * argument appears in the input, advances the cursor to the end and
173 	 * returns true.  Returns false if the string does not appear at the
174 	 * current point in the input.
175 	 */
176 	bool consume(const char *str);
177 	/**
178 	 * Reads an integer in base 8, 10, or 16.  Returns true and advances
179 	 * the cursor to the end of the integer if the cursor points to an
180 	 * integer, returns false and does not move the cursor otherwise.
181 	 *
182 	 * The parsed value is returned via the argument.
183 	 */
184 	bool consume_integer(long long &outInt);
185 	/**
186 	 * Template function that consumes a binary value in big-endian format
187 	 * from the input stream.  Returns true and advances the cursor if
188 	 * there is a value of the correct size.  This function assumes that
189 	 * all values must be natively aligned, and so advances the cursor to
190 	 * the correct alignment before reading.
191 	 */
192 	template<typename T>
193 	bool consume_binary(T &out)
194 	{
195 		int align = 0;
196 		int type_size = sizeof(T);
197 		if (cursor % type_size != 0)
198 		{
199 			align = type_size - (cursor % type_size);
200 		}
201 		if (size < cursor + align + type_size)
202 		{
203 			return false;
204 		}
205 		cursor += align;
206 		assert(cursor % type_size == 0);
207 		out = 0;
208 		for (int i=0 ; i<type_size ; ++i)
209 		{
210 			out <<= 8;
211 			out |= (((T)buffer[cursor++]) & 0xff);
212 		}
213 		return true;
214 	}
215 	/**
216 	 * Consumes two hex digits and return the resulting byte via the first
217 	 * argument.  If the next two characters are hex digits, returns true
218 	 * and advances the cursor.  If not, then returns false and leaves the
219 	 * cursor in place.
220 	 */
221 	bool consume_hex_byte(uint8_t &outByte);
222 	/**
223 	 * Advances the cursor to the start of the next token, skipping
224 	 * comments and whitespace.  If the cursor already points to the start
225 	 * of a token, then this function does nothing.
226 	 */
227 	input_buffer &next_token();
228 	/**
229 	 * Prints a message indicating the location of a parse error.
230 	 */
231 	void parse_error(const char *msg);
232 	/**
233 	 * Dumps the current cursor value and the unconsumed values in the
234 	 * input buffer to the standard error.  This method is intended solely
235 	 * for debugging.
236 	 */
237 	void dump();
238 };
239 /**
240  * Explicit specialisation for reading a single byte.
241  */
242 template<>
243 inline bool input_buffer::consume_binary(uint8_t &out)
244 {
245 	if (size < cursor + 1)
246 	{
247 		return false;
248 	}
249 	out = buffer[cursor++];
250 	return true;
251 }
252 
253 /**
254  * Subclass of input_buffer that mmap()s a file and owns the resulting memory.
255  * When this object is destroyed, the memory is unmapped.
256  */
257 struct mmap_input_buffer : public input_buffer
258 {
259 	/**
260 	 * Constructs a new buffer from the file passed in as a file
261 	 * descriptor.
262 	 */
263 	mmap_input_buffer(int fd);
264 	/**
265 	 * Unmaps the buffer, if one exists.
266 	 */
267 	virtual ~mmap_input_buffer();
268 };
269 /**
270  * Input buffer read from standard input.  This is used for reading device tree
271  * blobs and source from standard input.  It reads the entire input into
272  * malloc'd memory, so will be very slow for large inputs.  DTS and DTB files
273  * are very rarely more than 10KB though, so this is probably not a problem.
274  */
275 struct stream_input_buffer : public input_buffer
276 {
277 	/**
278 	 * The buffer that will store the data read from the standard input.
279 	 */
280 	std::vector<char> b;
281 	/**
282 	 * Constructs a new buffer from the standard input.
283 	 */
284 	stream_input_buffer();
285 };
286 
287 } // namespace dtc
288 
289 #endif // !_INPUT_BUFFER_HH_
290