1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
4  * All rights reserved.
5  *
6  * This source code is licensed under the BSD-style license found in the
7  * LICENSE file in the root directory of https://github.com/facebook/zstd.
8  * An additional grant of patent rights can be found in the PATENTS file in the
9  * same directory.
10  *
11  * This program is free software; you can redistribute it and/or modify it under
12  * the terms of the GNU General Public License version 2 as published by the
13  * Free Software Foundation. This program is dual-licensed; you may select
14  * either version 2 of the GNU General Public License ("GPL") or BSD license
15  * ("BSD").
16  */
17 
18 #ifndef LINUX_ZSTD_H
19 #define LINUX_ZSTD_H
20 
21 /**
22  * This is a kernel-style API that wraps the upstream zstd API, which cannot be
23  * used directly because the symbols aren't exported. It exposes the minimal
24  * functionality which is currently required by users of zstd in the kernel.
25  * Expose extra functions from lib/zstd/zstd.h as needed.
26  */
27 
28 /* ======   Dependency   ====== */
29 #include <linux/types.h>
30 
31 /* ======   Helper Functions   ====== */
32 /**
33  * zstd_compress_bound() - maximum compressed size in worst case scenario
34  * @src_size: The size of the data to compress.
35  *
36  * Return:    The maximum compressed size in the worst case scenario.
37  */
38 size_t zstd_compress_bound(size_t src_size);
39 
40 /**
41  * zstd_is_error() - tells if a size_t function result is an error code
42  * @code:  The function result to check for error.
43  *
44  * Return: Non-zero iff the code is an error.
45  */
46 unsigned int zstd_is_error(size_t code);
47 
48 /**
49  * zstd_get_error_code() - translates an error function result to an error code
50  * @code:  The function result for which zstd_is_error(code) is true.
51  *
52  * Return: A unique error code for this error.
53  */
54 int zstd_get_error_code(size_t code);
55 
56 /**
57  * zstd_get_error_name() - translates an error function result to a string
58  * @code:  The function result for which zstd_is_error(code) is true.
59  *
60  * Return: An error string corresponding to the error code.
61  */
62 const char *zstd_get_error_name(size_t code);
63 
64 /* ======   Parameter Selection   ====== */
65 
66 /**
67  * enum zstd_strategy - zstd compression search strategy
68  *
69  * From faster to stronger.
70  */
71 enum zstd_strategy {
72 	zstd_fast = 1,
73 	zstd_dfast = 2,
74 	zstd_greedy = 3,
75 	zstd_lazy = 4,
76 	zstd_lazy2 = 5,
77 	zstd_btlazy2 = 6,
78 	zstd_btopt = 7,
79 	zstd_btultra = 8,
80 	zstd_btultra2 = 9
81 };
82 
83 /**
84  * struct zstd_compression_parameters - zstd compression parameters
85  * @window_log:    Log of the largest match distance. Larger means more
86  *                 compression, and more memory needed during decompression.
87  * @chain_log:     Fully searched segment. Larger means more compression,
88  *                 slower, and more memory (useless for fast).
89  * @hash_log:      Dispatch table. Larger means more compression,
90  *                 slower, and more memory.
91  * @search_log:    Number of searches. Larger means more compression and slower.
92  * @search_length: Match length searched. Larger means faster decompression,
93  *                 sometimes less compression.
94  * @target_length: Acceptable match size for optimal parser (only). Larger means
95  *                 more compression, and slower.
96  * @strategy:      The zstd compression strategy.
97  */
98 struct zstd_compression_parameters {
99 	unsigned int window_log;
100 	unsigned int chain_log;
101 	unsigned int hash_log;
102 	unsigned int search_log;
103 	unsigned int search_length;
104 	unsigned int target_length;
105 	enum zstd_strategy strategy;
106 };
107 
108 /**
109  * struct zstd_frame_parameters - zstd frame parameters
110  * @content_size_flag: Controls whether content size will be present in the
111  *                     frame header (when known).
112  * @checksum_flag:     Controls whether a 32-bit checksum is generated at the
113  *                     end of the frame for error detection.
114  * @no_dict_id_flag:   Controls whether dictID will be saved into the frame
115  *                     header when using dictionary compression.
116  *
117  * The default value is all fields set to 0.
118  */
119 struct zstd_frame_parameters {
120 	unsigned int content_size_flag;
121 	unsigned int checksum_flag;
122 	unsigned int no_dict_id_flag;
123 };
124 
125 /**
126  * struct zstd_parameters - zstd parameters
127  * @cparams: The compression parameters.
128  * @fparams: The frame parameters.
129  */
130 struct zstd_parameters {
131 	struct zstd_compression_parameters cparams;
132 	struct zstd_frame_parameters fparams;
133 };
134 
135 /**
136  * zstd_get_params() - returns zstd_parameters for selected level
137  * @level:              The compression level
138  * @estimated_src_size: The estimated source size to compress or 0
139  *                      if unknown.
140  *
141  * Return:              The selected zstd_parameters.
142  */
143 struct zstd_parameters zstd_get_params(int level,
144 	unsigned long long estimated_src_size);
145 
146 /* ======   Single-pass Compression   ====== */
147 
148 typedef struct ZSTD_CCtx_s zstd_cctx;
149 
150 /**
151  * zstd_cctx_workspace_bound() - max memory needed to initialize a zstd_cctx
152  * @parameters: The compression parameters to be used.
153  *
154  * If multiple compression parameters might be used, the caller must call
155  * zstd_cctx_workspace_bound() for each set of parameters and use the maximum
156  * size.
157  *
158  * Return:      A lower bound on the size of the workspace that is passed to
159  *              zstd_init_cctx().
160  */
161 size_t zstd_cctx_workspace_bound(
162 	const struct zstd_compression_parameters *parameters);
163 
164 /**
165  * zstd_init_cctx() - initialize a zstd compression context
166  * @workspace:      The workspace to emplace the context into. It must outlive
167  *                  the returned context.
168  * @workspace_size: The size of workspace. Use zstd_cctx_workspace_bound() to
169  *                  determine how large the workspace must be.
170  *
171  * Return:          A zstd compression context or NULL on error.
172  */
173 zstd_cctx *zstd_init_cctx(void *workspace, size_t workspace_size);
174 
175 /**
176  * zstd_compress_cctx() - compress src into dst with the initialized parameters
177  * @cctx:         The context. Must have been initialized with zstd_init_cctx().
178  * @dst:          The buffer to compress src into.
179  * @dst_capacity: The size of the destination buffer. May be any size, but
180  *                ZSTD_compressBound(srcSize) is guaranteed to be large enough.
181  * @src:          The data to compress.
182  * @src_size:     The size of the data to compress.
183  * @parameters:   The compression parameters to be used.
184  *
185  * Return:        The compressed size or an error, which can be checked using
186  *                zstd_is_error().
187  */
188 size_t zstd_compress_cctx(zstd_cctx *cctx, void *dst, size_t dst_capacity,
189 	const void *src, size_t src_size, const struct zstd_parameters *parameters);
190 
191 /* ======   Single-pass Decompression   ====== */
192 
193 typedef struct ZSTD_DCtx_s zstd_dctx;
194 
195 /**
196  * zstd_dctx_workspace_bound() - max memory needed to initialize a zstd_dctx
197  *
198  * Return: A lower bound on the size of the workspace that is passed to
199  *         zstd_init_dctx().
200  */
201 size_t zstd_dctx_workspace_bound(void);
202 
203 /**
204  * zstd_init_dctx() - initialize a zstd decompression context
205  * @workspace:      The workspace to emplace the context into. It must outlive
206  *                  the returned context.
207  * @workspace_size: The size of workspace. Use zstd_dctx_workspace_bound() to
208  *                  determine how large the workspace must be.
209  *
210  * Return:          A zstd decompression context or NULL on error.
211  */
212 zstd_dctx *zstd_init_dctx(void *workspace, size_t workspace_size);
213 
214 /**
215  * zstd_decompress_dctx() - decompress zstd compressed src into dst
216  * @dctx:         The decompression context.
217  * @dst:          The buffer to decompress src into.
218  * @dst_capacity: The size of the destination buffer. Must be at least as large
219  *                as the decompressed size. If the caller cannot upper bound the
220  *                decompressed size, then it's better to use the streaming API.
221  * @src:          The zstd compressed data to decompress. Multiple concatenated
222  *                frames and skippable frames are allowed.
223  * @src_size:     The exact size of the data to decompress.
224  *
225  * Return:        The decompressed size or an error, which can be checked using
226  *                zstd_is_error().
227  */
228 size_t zstd_decompress_dctx(zstd_dctx *dctx, void *dst, size_t dst_capacity,
229 	const void *src, size_t src_size);
230 
231 /* ======   Streaming Buffers   ====== */
232 
233 /**
234  * struct zstd_in_buffer - input buffer for streaming
235  * @src:  Start of the input buffer.
236  * @size: Size of the input buffer.
237  * @pos:  Position where reading stopped. Will be updated.
238  *        Necessarily 0 <= pos <= size.
239  */
240 struct zstd_in_buffer {
241 	const void *src;
242 	size_t size;
243 	size_t pos;
244 };
245 
246 /**
247  * struct zstd_out_buffer - output buffer for streaming
248  * @dst:  Start of the output buffer.
249  * @size: Size of the output buffer.
250  * @pos:  Position where writing stopped. Will be updated.
251  *        Necessarily 0 <= pos <= size.
252  */
253 struct zstd_out_buffer {
254 	void *dst;
255 	size_t size;
256 	size_t pos;
257 };
258 
259 /* ======   Streaming Compression   ====== */
260 
261 typedef struct ZSTD_CCtx_s zstd_cstream;
262 
263 /**
264  * zstd_cstream_workspace_bound() - memory needed to initialize a zstd_cstream
265  * @cparams: The compression parameters to be used for compression.
266  *
267  * Return:   A lower bound on the size of the workspace that is passed to
268  *           zstd_init_cstream().
269  */
270 size_t zstd_cstream_workspace_bound(
271 	const struct zstd_compression_parameters *cparams);
272 
273 /**
274  * zstd_init_cstream() - initialize a zstd streaming compression context
275  * @parameters        The zstd parameters to use for compression.
276  * @pledged_src_size: If params.fParams.contentSizeFlag == 1 then the caller
277  *                    must pass the source size (zero means empty source).
278  *                    Otherwise, the caller may optionally pass the source
279  *                    size, or zero if unknown.
280  * @workspace:        The workspace to emplace the context into. It must outlive
281  *                    the returned context.
282  * @workspace_size:   The size of workspace.
283  *                    Use zstd_cstream_workspace_bound(params->cparams) to
284  *                    determine how large the workspace must be.
285  *
286  * Return:            The zstd streaming compression context or NULL on error.
287  */
288 zstd_cstream *zstd_init_cstream(const struct zstd_parameters *parameters,
289 	unsigned long long pledged_src_size, void *workspace, size_t workspace_size);
290 
291 /**
292  * zstd_reset_cstream() - reset the context using parameters from creation
293  * @cstream:          The zstd streaming compression context to reset.
294  * @pledged_src_size: Optionally the source size, or zero if unknown.
295  *
296  * Resets the context using the parameters from creation. Skips dictionary
297  * loading, since it can be reused. If `pledged_src_size` is non-zero the frame
298  * content size is always written into the frame header.
299  *
300  * Return:            Zero or an error, which can be checked using
301  *                    zstd_is_error().
302  */
303 size_t zstd_reset_cstream(zstd_cstream *cstream,
304 	unsigned long long pledged_src_size);
305 
306 /**
307  * zstd_compress_stream() - streaming compress some of input into output
308  * @cstream: The zstd streaming compression context.
309  * @output:  Destination buffer. `output->pos` is updated to indicate how much
310  *           compressed data was written.
311  * @input:   Source buffer. `input->pos` is updated to indicate how much data
312  *           was read. Note that it may not consume the entire input, in which
313  *           case `input->pos < input->size`, and it's up to the caller to
314  *           present remaining data again.
315  *
316  * The `input` and `output` buffers may be any size. Guaranteed to make some
317  * forward progress if `input` and `output` are not empty.
318  *
319  * Return:   A hint for the number of bytes to use as the input for the next
320  *           function call or an error, which can be checked using
321  *           zstd_is_error().
322  */
323 size_t zstd_compress_stream(zstd_cstream *cstream,
324 	struct zstd_out_buffer *output, struct zstd_in_buffer *input);
325 
326 /**
327  * zstd_flush_stream() - flush internal buffers into output
328  * @cstream: The zstd streaming compression context.
329  * @output:  Destination buffer. `output->pos` is updated to indicate how much
330  *           compressed data was written.
331  *
332  * zstd_flush_stream() must be called until it returns 0, meaning all the data
333  * has been flushed. Since zstd_flush_stream() causes a block to be ended,
334  * calling it too often will degrade the compression ratio.
335  *
336  * Return:   The number of bytes still present within internal buffers or an
337  *           error, which can be checked using zstd_is_error().
338  */
339 size_t zstd_flush_stream(zstd_cstream *cstream, struct zstd_out_buffer *output);
340 
341 /**
342  * zstd_end_stream() - flush internal buffers into output and end the frame
343  * @cstream: The zstd streaming compression context.
344  * @output:  Destination buffer. `output->pos` is updated to indicate how much
345  *           compressed data was written.
346  *
347  * zstd_end_stream() must be called until it returns 0, meaning all the data has
348  * been flushed and the frame epilogue has been written.
349  *
350  * Return:   The number of bytes still present within internal buffers or an
351  *           error, which can be checked using zstd_is_error().
352  */
353 size_t zstd_end_stream(zstd_cstream *cstream, struct zstd_out_buffer *output);
354 
355 /* ======   Streaming Decompression   ====== */
356 
357 typedef struct ZSTD_DCtx_s zstd_dstream;
358 
359 /**
360  * zstd_dstream_workspace_bound() - memory needed to initialize a zstd_dstream
361  * @max_window_size: The maximum window size allowed for compressed frames.
362  *
363  * Return:           A lower bound on the size of the workspace that is passed
364  *                   to zstd_init_dstream().
365  */
366 size_t zstd_dstream_workspace_bound(size_t max_window_size);
367 
368 /**
369  * zstd_init_dstream() - initialize a zstd streaming decompression context
370  * @max_window_size: The maximum window size allowed for compressed frames.
371  * @workspace:       The workspace to emplace the context into. It must outlive
372  *                   the returned context.
373  * @workspaceSize:   The size of workspace.
374  *                   Use zstd_dstream_workspace_bound(max_window_size) to
375  *                   determine how large the workspace must be.
376  *
377  * Return:           The zstd streaming decompression context.
378  */
379 zstd_dstream *zstd_init_dstream(size_t max_window_size, void *workspace,
380 	size_t workspace_size);
381 
382 /**
383  * zstd_reset_dstream() - reset the context using parameters from creation
384  * @dstream: The zstd streaming decompression context to reset.
385  *
386  * Resets the context using the parameters from creation. Skips dictionary
387  * loading, since it can be reused.
388  *
389  * Return:   Zero or an error, which can be checked using zstd_is_error().
390  */
391 size_t zstd_reset_dstream(zstd_dstream *dstream);
392 
393 /**
394  * zstd_decompress_stream() - streaming decompress some of input into output
395  * @dstream: The zstd streaming decompression context.
396  * @output:  Destination buffer. `output.pos` is updated to indicate how much
397  *           decompressed data was written.
398  * @input:   Source buffer. `input.pos` is updated to indicate how much data was
399  *           read. Note that it may not consume the entire input, in which case
400  *           `input.pos < input.size`, and it's up to the caller to present
401  *           remaining data again.
402  *
403  * The `input` and `output` buffers may be any size. Guaranteed to make some
404  * forward progress if `input` and `output` are not empty.
405  * zstd_decompress_stream() will not consume the last byte of the frame until
406  * the entire frame is flushed.
407  *
408  * Return:   Returns 0 iff a frame is completely decoded and fully flushed.
409  *           Otherwise returns a hint for the number of bytes to use as the
410  *           input for the next function call or an error, which can be checked
411  *           using zstd_is_error(). The size hint will never load more than the
412  *           frame.
413  */
414 size_t zstd_decompress_stream(zstd_dstream *dstream,
415 	struct zstd_out_buffer *output, struct zstd_in_buffer *input);
416 
417 /* ======   Frame Inspection Functions ====== */
418 
419 /**
420  * zstd_find_frame_compressed_size() - returns the size of a compressed frame
421  * @src:      Source buffer. It should point to the start of a zstd encoded
422  *            frame or a skippable frame.
423  * @src_size: The size of the source buffer. It must be at least as large as the
424  *            size of the frame.
425  *
426  * Return:    The compressed size of the frame pointed to by `src` or an error,
427  *            which can be check with zstd_is_error().
428  *            Suitable to pass to ZSTD_decompress() or similar functions.
429  */
430 size_t zstd_find_frame_compressed_size(const void *src, size_t src_size);
431 
432 /**
433  * struct zstd_frame_params - zstd frame parameters stored in the frame header
434  * @frame_content_size: The frame content size, or 0 if not present.
435  * @window_size:        The window size, or 0 if the frame is a skippable frame.
436  * @dict_id:            The dictionary id, or 0 if not present.
437  * @checksum_flag:      Whether a checksum was used.
438  */
439 struct zstd_frame_params {
440 	unsigned long long frame_content_size;
441 	unsigned int window_size;
442 	unsigned int dict_id;
443 	unsigned int checksum_flag;
444 };
445 
446 /**
447  * zstd_get_frame_params() - extracts parameters from a zstd or skippable frame
448  * @params:   On success the frame parameters are written here.
449  * @src:      The source buffer. It must point to a zstd or skippable frame.
450  * @src_size: The size of the source buffer.
451  *
452  * Return:    0 on success. If more data is required it returns how many bytes
453  *            must be provided to make forward progress. Otherwise it returns
454  *            an error, which can be checked using zstd_is_error().
455  */
456 size_t zstd_get_frame_params(struct zstd_frame_params *params, const void *src,
457 	size_t src_size);
458 
459 #endif  /* LINUX_ZSTD_H */
460