1 | // Copyright (C) 2020 The Qt Company Ltd. |
---|---|
2 | // SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only |
3 | |
4 | #include "qdecompresshelper_p.h" |
5 | |
6 | #include <QtCore/qiodevice.h> |
7 | #include <QtCore/qcoreapplication.h> |
8 | |
9 | #include <limits> |
10 | #include <zlib.h> |
11 | |
12 | #if QT_CONFIG(brotli) |
13 | # include <brotli/decode.h> |
14 | #endif |
15 | |
16 | #if QT_CONFIG(zstd) |
17 | # include <zstd.h> |
18 | #endif |
19 | |
20 | #include <array> |
21 | |
22 | QT_BEGIN_NAMESPACE |
23 | namespace { |
24 | struct ContentEncodingMapping |
25 | { |
26 | QByteArrayView name; |
27 | QDecompressHelper::ContentEncoding encoding; |
28 | }; |
29 | |
30 | constexpr ContentEncodingMapping contentEncodingMapping[] { |
31 | #if QT_CONFIG(zstd) |
32 | { .name: "zstd", .encoding: QDecompressHelper::Zstandard }, |
33 | #endif |
34 | #if QT_CONFIG(brotli) |
35 | { .name: "br", .encoding: QDecompressHelper::Brotli }, |
36 | #endif |
37 | { .name: "gzip", .encoding: QDecompressHelper::GZip }, |
38 | { .name: "deflate", .encoding: QDecompressHelper::Deflate }, |
39 | }; |
40 | |
41 | QDecompressHelper::ContentEncoding encodingFromByteArray(QByteArrayView ce) noexcept |
42 | { |
43 | for (const auto &mapping : contentEncodingMapping) { |
44 | if (ce.compare(a: mapping.name, cs: Qt::CaseInsensitive) == 0) |
45 | return mapping.encoding; |
46 | } |
47 | return QDecompressHelper::None; |
48 | } |
49 | |
50 | z_stream *toZlibPointer(void *ptr) |
51 | { |
52 | return static_cast<z_stream_s *>(ptr); |
53 | } |
54 | |
55 | #if QT_CONFIG(brotli) |
56 | BrotliDecoderState *toBrotliPointer(void *ptr) |
57 | { |
58 | return static_cast<BrotliDecoderState *>(ptr); |
59 | } |
60 | #endif |
61 | |
62 | #if QT_CONFIG(zstd) |
63 | ZSTD_DStream *toZstandardPointer(void *ptr) |
64 | { |
65 | return static_cast<ZSTD_DStream *>(ptr); |
66 | } |
67 | #endif |
68 | } |
69 | |
70 | bool QDecompressHelper::isSupportedEncoding(QByteArrayView encoding) |
71 | { |
72 | return encodingFromByteArray(ce: encoding) != QDecompressHelper::None; |
73 | } |
74 | |
75 | QByteArrayList QDecompressHelper::acceptedEncoding() |
76 | { |
77 | QByteArrayList list; |
78 | list.reserve(asize: std::size(contentEncodingMapping)); |
79 | for (const auto &mapping : contentEncodingMapping) { |
80 | list << mapping.name.toByteArray(); |
81 | } |
82 | return list; |
83 | } |
84 | |
85 | QDecompressHelper::~QDecompressHelper() |
86 | { |
87 | clear(); |
88 | } |
89 | |
90 | bool QDecompressHelper::setEncoding(QByteArrayView encoding) |
91 | { |
92 | Q_ASSERT(contentEncoding == QDecompressHelper::None); |
93 | if (contentEncoding != QDecompressHelper::None) { |
94 | qWarning(msg: "Encoding is already set."); |
95 | // This isn't an error, so it doesn't set errorStr, it's just wrong usage. |
96 | return false; |
97 | } |
98 | ContentEncoding ce = encodingFromByteArray(ce: encoding); |
99 | if (ce == None) { |
100 | errorStr = QCoreApplication::translate(context: "QHttp", key: "Unsupported content encoding: %1") |
101 | .arg(a: QLatin1String(encoding)); |
102 | return false; |
103 | } |
104 | errorStr = QString(); // clear error |
105 | return setEncoding(ce); |
106 | } |
107 | |
108 | bool QDecompressHelper::setEncoding(ContentEncoding ce) |
109 | { |
110 | Q_ASSERT(contentEncoding == None); |
111 | contentEncoding = ce; |
112 | switch (contentEncoding) { |
113 | case None: |
114 | Q_UNREACHABLE(); |
115 | break; |
116 | case Deflate: |
117 | case GZip: { |
118 | z_stream *inflateStream = new z_stream; |
119 | memset(s: inflateStream, c: 0, n: sizeof(z_stream)); |
120 | // "windowBits can also be greater than 15 for optional gzip decoding. |
121 | // Add 32 to windowBits to enable zlib and gzip decoding with automatic header detection" |
122 | // http://www.zlib.net/manual.html |
123 | if (inflateInit2(inflateStream, MAX_WBITS + 32) != Z_OK) { |
124 | delete inflateStream; |
125 | inflateStream = nullptr; |
126 | } |
127 | decoderPointer = inflateStream; |
128 | break; |
129 | } |
130 | case Brotli: |
131 | #if QT_CONFIG(brotli) |
132 | decoderPointer = BrotliDecoderCreateInstance(alloc_func: nullptr, free_func: nullptr, opaque: nullptr); |
133 | #else |
134 | Q_UNREACHABLE(); |
135 | #endif |
136 | break; |
137 | case Zstandard: |
138 | #if QT_CONFIG(zstd) |
139 | decoderPointer = ZSTD_createDStream(); |
140 | #else |
141 | Q_UNREACHABLE(); |
142 | #endif |
143 | break; |
144 | } |
145 | if (!decoderPointer) { |
146 | errorStr = QCoreApplication::translate(context: "QHttp", |
147 | key: "Failed to initialize the compression decoder."); |
148 | contentEncoding = QDecompressHelper::None; |
149 | return false; |
150 | } |
151 | return true; |
152 | } |
153 | |
154 | /*! |
155 | \internal |
156 | |
157 | Returns true if the QDecompressHelper is measuring the |
158 | size of the decompressed data. |
159 | |
160 | \sa setCountingBytesEnabled, uncompressedSize |
161 | */ |
162 | bool QDecompressHelper::isCountingBytes() const |
163 | { |
164 | return countDecompressed; |
165 | } |
166 | |
167 | /*! |
168 | \internal |
169 | |
170 | Enable or disable counting the decompressed size of the data |
171 | based on \a shouldCount. Enabling this means the data will be |
172 | decompressed twice (once for counting and once when data is |
173 | being read). |
174 | |
175 | \note Can only be called before contentEncoding is set and data |
176 | is fed to the object. |
177 | |
178 | \sa isCountingBytes, uncompressedSize |
179 | */ |
180 | void QDecompressHelper::setCountingBytesEnabled(bool shouldCount) |
181 | { |
182 | // These are a best-effort check to ensure that no data has already been processed before this |
183 | // gets enabled |
184 | Q_ASSERT(compressedDataBuffer.byteAmount() == 0); |
185 | Q_ASSERT(contentEncoding == None); |
186 | countDecompressed = shouldCount; |
187 | } |
188 | |
189 | /*! |
190 | \internal |
191 | |
192 | Returns the amount of uncompressed bytes left. |
193 | |
194 | \note Since this is only based on the data received |
195 | so far the final size could be larger. |
196 | |
197 | \note It is only valid to call this if isCountingBytes() |
198 | returns true |
199 | |
200 | \sa isCountingBytes, setCountBytes |
201 | */ |
202 | qint64 QDecompressHelper::uncompressedSize() const |
203 | { |
204 | Q_ASSERT(countDecompressed); |
205 | // Use the 'totalUncompressedBytes' from the countHelper if it exceeds the amount of bytes |
206 | // that we know about. |
207 | auto totalUncompressed = |
208 | countHelper && countHelper->totalUncompressedBytes > totalUncompressedBytes |
209 | ? countHelper->totalUncompressedBytes |
210 | : totalUncompressedBytes; |
211 | return totalUncompressed - totalBytesRead; |
212 | } |
213 | |
214 | /*! |
215 | \internal |
216 | \overload |
217 | */ |
218 | void QDecompressHelper::feed(const QByteArray &data) |
219 | { |
220 | return feed(data: QByteArray(data)); |
221 | } |
222 | |
223 | /*! |
224 | \internal |
225 | Give \a data to the QDecompressHelper which will be stored until |
226 | a read is attempted. |
227 | |
228 | If \c isCountingBytes() is true then it will decompress immediately |
229 | before discarding the data, but will count the uncompressed byte |
230 | size. |
231 | */ |
232 | void QDecompressHelper::feed(QByteArray &&data) |
233 | { |
234 | Q_ASSERT(contentEncoding != None); |
235 | totalCompressedBytes += data.size(); |
236 | compressedDataBuffer.append(bd: std::move(data)); |
237 | if (!countInternal(data: compressedDataBuffer[compressedDataBuffer.bufferCount() - 1])) { |
238 | if (errorStr.isEmpty() && countHelper) |
239 | errorStr = countHelper->errorString(); |
240 | clear(); // If our counting brother failed then so will we :| |
241 | } |
242 | } |
243 | |
244 | /*! |
245 | \internal |
246 | \overload |
247 | */ |
248 | void QDecompressHelper::feed(const QByteDataBuffer &buffer) |
249 | { |
250 | Q_ASSERT(contentEncoding != None); |
251 | totalCompressedBytes += buffer.byteAmount(); |
252 | compressedDataBuffer.append(other: buffer); |
253 | if (!countInternal(buffer)) { |
254 | if (errorStr.isEmpty() && countHelper) |
255 | errorStr = countHelper->errorString(); |
256 | clear(); // If our counting brother failed then so will we :| |
257 | } |
258 | } |
259 | |
260 | /*! |
261 | \internal |
262 | \overload |
263 | */ |
264 | void QDecompressHelper::feed(QByteDataBuffer &&buffer) |
265 | { |
266 | Q_ASSERT(contentEncoding != None); |
267 | totalCompressedBytes += buffer.byteAmount(); |
268 | const QByteDataBuffer copy(buffer); |
269 | compressedDataBuffer.append(other: std::move(buffer)); |
270 | if (!countInternal(buffer: copy)) { |
271 | if (errorStr.isEmpty() && countHelper) |
272 | errorStr = countHelper->errorString(); |
273 | clear(); // If our counting brother failed then so will we :| |
274 | } |
275 | } |
276 | |
277 | /*! |
278 | \internal |
279 | Decompress the data internally and immediately discard the |
280 | uncompressed data, but count how many bytes were decoded. |
281 | This lets us know the final size, unfortunately at the cost of |
282 | increased computation. |
283 | |
284 | To save on some of the computation we will store the data until |
285 | we reach \c MaxDecompressedDataBufferSize stored. In this case the |
286 | "penalty" is completely removed from users who read the data on |
287 | readyRead rather than waiting for it all to be received. And |
288 | any file smaller than \c MaxDecompressedDataBufferSize will |
289 | avoid this issue as well. |
290 | */ |
291 | bool QDecompressHelper::countInternal() |
292 | { |
293 | Q_ASSERT(countDecompressed); |
294 | while (hasDataInternal() |
295 | && decompressedDataBuffer.byteAmount() < MaxDecompressedDataBufferSize) { |
296 | const qsizetype toRead = 256 * 1024; |
297 | QByteArray buffer(toRead, Qt::Uninitialized); |
298 | qsizetype bytesRead = readInternal(data: buffer.data(), maxSize: buffer.size()); |
299 | if (bytesRead == -1) |
300 | return false; |
301 | buffer.truncate(pos: bytesRead); |
302 | decompressedDataBuffer.append(bd: std::move(buffer)); |
303 | } |
304 | if (!hasDataInternal()) |
305 | return true; // handled all the data so far, just return |
306 | |
307 | while (countHelper->hasData()) { |
308 | std::array<char, 1024> temp; |
309 | qsizetype bytesRead = countHelper->read(data: temp.data(), maxSize: temp.size()); |
310 | if (bytesRead == -1) |
311 | return false; |
312 | } |
313 | return true; |
314 | } |
315 | |
316 | /*! |
317 | \internal |
318 | \overload |
319 | */ |
320 | bool QDecompressHelper::countInternal(const QByteArray &data) |
321 | { |
322 | if (countDecompressed) { |
323 | if (!countHelper) { |
324 | countHelper = std::make_unique<QDecompressHelper>(); |
325 | countHelper->setDecompressedSafetyCheckThreshold(archiveBombCheckThreshold); |
326 | countHelper->setEncoding(contentEncoding); |
327 | } |
328 | countHelper->feed(data); |
329 | return countInternal(); |
330 | } |
331 | return true; |
332 | } |
333 | |
334 | /*! |
335 | \internal |
336 | \overload |
337 | */ |
338 | bool QDecompressHelper::countInternal(const QByteDataBuffer &buffer) |
339 | { |
340 | if (countDecompressed) { |
341 | if (!countHelper) { |
342 | countHelper = std::make_unique<QDecompressHelper>(); |
343 | countHelper->setDecompressedSafetyCheckThreshold(archiveBombCheckThreshold); |
344 | countHelper->setEncoding(contentEncoding); |
345 | } |
346 | countHelper->feed(buffer); |
347 | return countInternal(); |
348 | } |
349 | return true; |
350 | } |
351 | |
352 | qsizetype QDecompressHelper::read(char *data, qsizetype maxSize) |
353 | { |
354 | if (maxSize <= 0) |
355 | return 0; |
356 | |
357 | if (!isValid()) |
358 | return -1; |
359 | |
360 | if (!hasData()) |
361 | return 0; |
362 | |
363 | qsizetype cachedRead = 0; |
364 | if (!decompressedDataBuffer.isEmpty()) { |
365 | cachedRead = decompressedDataBuffer.read(dst: data, amount: maxSize); |
366 | data += cachedRead; |
367 | maxSize -= cachedRead; |
368 | } |
369 | |
370 | qsizetype bytesRead = readInternal(data, maxSize); |
371 | if (bytesRead == -1) |
372 | return -1; |
373 | totalBytesRead += bytesRead + cachedRead; |
374 | return bytesRead + cachedRead; |
375 | } |
376 | |
377 | /*! |
378 | \internal |
379 | Like read() but without attempting to read the |
380 | cached/already-decompressed data. |
381 | */ |
382 | qsizetype QDecompressHelper::readInternal(char *data, qsizetype maxSize) |
383 | { |
384 | Q_ASSERT(isValid()); |
385 | |
386 | if (maxSize <= 0) |
387 | return 0; |
388 | |
389 | if (!hasDataInternal()) |
390 | return 0; |
391 | |
392 | qsizetype bytesRead = -1; |
393 | switch (contentEncoding) { |
394 | case None: |
395 | Q_UNREACHABLE(); |
396 | break; |
397 | case Deflate: |
398 | case GZip: |
399 | bytesRead = readZLib(data, maxSize); |
400 | break; |
401 | case Brotli: |
402 | bytesRead = readBrotli(data, maxSize); |
403 | break; |
404 | case Zstandard: |
405 | bytesRead = readZstandard(data, maxSize); |
406 | break; |
407 | } |
408 | if (bytesRead == -1) |
409 | clear(); |
410 | |
411 | totalUncompressedBytes += bytesRead; |
412 | if (isPotentialArchiveBomb()) { |
413 | errorStr = QCoreApplication::translate( |
414 | context: "QHttp", |
415 | key: "The decompressed output exceeds the limits specified by " |
416 | "QNetworkRequest::decompressedSafetyCheckThreshold()"); |
417 | return -1; |
418 | } |
419 | |
420 | return bytesRead; |
421 | } |
422 | |
423 | /*! |
424 | \internal |
425 | Set the \a threshold required before the archive bomb detection kicks in. |
426 | By default this is 10MB. Setting it to -1 is treated as disabling the |
427 | feature. |
428 | */ |
429 | void QDecompressHelper::setDecompressedSafetyCheckThreshold(qint64 threshold) |
430 | { |
431 | if (threshold == -1) |
432 | threshold = std::numeric_limits<qint64>::max(); |
433 | archiveBombCheckThreshold = threshold; |
434 | } |
435 | |
436 | bool QDecompressHelper::isPotentialArchiveBomb() const |
437 | { |
438 | if (totalCompressedBytes == 0) |
439 | return false; |
440 | |
441 | if (totalUncompressedBytes <= archiveBombCheckThreshold) |
442 | return false; |
443 | |
444 | // Some protection against malicious or corrupted compressed files that expand far more than |
445 | // is reasonable. |
446 | double ratio = double(totalUncompressedBytes) / double(totalCompressedBytes); |
447 | switch (contentEncoding) { |
448 | case None: |
449 | Q_UNREACHABLE(); |
450 | break; |
451 | case Deflate: |
452 | case GZip: |
453 | // This value is mentioned in docs for |
454 | // QNetworkRequest::setMinimumArchiveBombSize, keep synchronized |
455 | if (ratio > 40) { |
456 | return true; |
457 | } |
458 | break; |
459 | case Brotli: |
460 | case Zstandard: |
461 | // This value is mentioned in docs for |
462 | // QNetworkRequest::setMinimumArchiveBombSize, keep synchronized |
463 | if (ratio > 100) { |
464 | return true; |
465 | } |
466 | break; |
467 | } |
468 | return false; |
469 | } |
470 | |
471 | /*! |
472 | \internal |
473 | Returns true if there are encoded bytes left or there is some |
474 | indication that the decoder still has data left internally. |
475 | |
476 | \note Even if this returns true the next call to read() might |
477 | read 0 bytes. This most likely means the decompression is done. |
478 | */ |
479 | bool QDecompressHelper::hasData() const |
480 | { |
481 | return hasDataInternal() || !decompressedDataBuffer.isEmpty(); |
482 | } |
483 | |
484 | /*! |
485 | \internal |
486 | Like hasData() but internally the buffer of decompressed data is |
487 | not interesting. |
488 | */ |
489 | bool QDecompressHelper::hasDataInternal() const |
490 | { |
491 | return encodedBytesAvailable() || decoderHasData; |
492 | } |
493 | |
494 | qint64 QDecompressHelper::encodedBytesAvailable() const |
495 | { |
496 | return compressedDataBuffer.byteAmount(); |
497 | } |
498 | |
499 | /*! |
500 | \internal |
501 | Returns whether or not the object is valid. |
502 | If it becomes invalid after an operation has been performed |
503 | then an error has occurred. |
504 | \sa errorString() |
505 | */ |
506 | bool QDecompressHelper::isValid() const |
507 | { |
508 | return contentEncoding != None; |
509 | } |
510 | |
511 | /*! |
512 | \internal |
513 | Returns a string describing the error that occurred or an empty |
514 | string if no error occurred. |
515 | \sa isValid() |
516 | */ |
517 | QString QDecompressHelper::errorString() const |
518 | { |
519 | return errorStr; |
520 | } |
521 | |
522 | void QDecompressHelper::clear() |
523 | { |
524 | switch (contentEncoding) { |
525 | case None: |
526 | break; |
527 | case Deflate: |
528 | case GZip: { |
529 | z_stream *inflateStream = toZlibPointer(ptr: decoderPointer); |
530 | if (inflateStream) |
531 | inflateEnd(strm: inflateStream); |
532 | delete inflateStream; |
533 | break; |
534 | } |
535 | case Brotli: { |
536 | #if QT_CONFIG(brotli) |
537 | BrotliDecoderState *brotliDecoderState = toBrotliPointer(ptr: decoderPointer); |
538 | if (brotliDecoderState) |
539 | BrotliDecoderDestroyInstance(state: brotliDecoderState); |
540 | #endif |
541 | break; |
542 | } |
543 | case Zstandard: { |
544 | #if QT_CONFIG(zstd) |
545 | ZSTD_DStream *zstdStream = toZstandardPointer(ptr: decoderPointer); |
546 | if (zstdStream) |
547 | ZSTD_freeDStream(zds: zstdStream); |
548 | #endif |
549 | break; |
550 | } |
551 | } |
552 | decoderPointer = nullptr; |
553 | contentEncoding = None; |
554 | |
555 | compressedDataBuffer.clear(); |
556 | decompressedDataBuffer.clear(); |
557 | decoderHasData = false; |
558 | |
559 | countDecompressed = false; |
560 | countHelper.reset(); |
561 | totalBytesRead = 0; |
562 | totalUncompressedBytes = 0; |
563 | totalCompressedBytes = 0; |
564 | } |
565 | |
566 | qsizetype QDecompressHelper::readZLib(char *data, const qsizetype maxSize) |
567 | { |
568 | bool triedRawDeflate = false; |
569 | |
570 | z_stream *inflateStream = toZlibPointer(ptr: decoderPointer); |
571 | static const size_t zlibMaxSize = |
572 | size_t(std::numeric_limits<decltype(inflateStream->avail_in)>::max()); |
573 | |
574 | QByteArrayView input = compressedDataBuffer.readPointer(); |
575 | if (size_t(input.size()) > zlibMaxSize) |
576 | input = input.sliced(pos: zlibMaxSize); |
577 | |
578 | inflateStream->avail_in = input.size(); |
579 | inflateStream->next_in = reinterpret_cast<Bytef *>(const_cast<char *>(input.data())); |
580 | |
581 | bool bigMaxSize = (zlibMaxSize < size_t(maxSize)); |
582 | qsizetype adjustedAvailableOut = bigMaxSize ? qsizetype(zlibMaxSize) : maxSize; |
583 | inflateStream->avail_out = adjustedAvailableOut; |
584 | inflateStream->next_out = reinterpret_cast<Bytef *>(data); |
585 | |
586 | qsizetype bytesDecoded = 0; |
587 | do { |
588 | auto previous_avail_out = inflateStream->avail_out; |
589 | int ret = inflate(strm: inflateStream, Z_NO_FLUSH); |
590 | // All negative return codes are errors, in the context of HTTP compression, Z_NEED_DICT is |
591 | // also an error. |
592 | // in the case where we get Z_DATA_ERROR this could be because we received raw deflate |
593 | // compressed data. |
594 | if (ret == Z_DATA_ERROR && !triedRawDeflate) { |
595 | inflateEnd(strm: inflateStream); |
596 | triedRawDeflate = true; |
597 | inflateStream->zalloc = Z_NULL; |
598 | inflateStream->zfree = Z_NULL; |
599 | inflateStream->opaque = Z_NULL; |
600 | inflateStream->avail_in = 0; |
601 | inflateStream->next_in = Z_NULL; |
602 | int ret = inflateInit2(inflateStream, -MAX_WBITS); |
603 | if (ret != Z_OK) { |
604 | return -1; |
605 | } else { |
606 | inflateStream->avail_in = input.size(); |
607 | inflateStream->next_in = |
608 | reinterpret_cast<Bytef *>(const_cast<char *>(input.data())); |
609 | continue; |
610 | } |
611 | } else if (ret < 0 || ret == Z_NEED_DICT) { |
612 | return -1; |
613 | } |
614 | bytesDecoded += qsizetype(previous_avail_out - inflateStream->avail_out); |
615 | if (ret == Z_STREAM_END) { |
616 | |
617 | // If there's more data after the stream then this is probably composed of multiple |
618 | // streams. |
619 | if (inflateStream->avail_in != 0) { |
620 | inflateEnd(strm: inflateStream); |
621 | Bytef *next_in = inflateStream->next_in; |
622 | uInt avail_in = inflateStream->avail_in; |
623 | inflateStream->zalloc = Z_NULL; |
624 | inflateStream->zfree = Z_NULL; |
625 | inflateStream->opaque = Z_NULL; |
626 | if (inflateInit2(inflateStream, MAX_WBITS + 32) != Z_OK) { |
627 | delete inflateStream; |
628 | decoderPointer = nullptr; |
629 | // Failed to reinitialize, so we'll just return what we have |
630 | compressedDataBuffer.advanceReadPointer(distance: input.size() - avail_in); |
631 | return bytesDecoded; |
632 | } else { |
633 | inflateStream->next_in = next_in; |
634 | inflateStream->avail_in = avail_in; |
635 | // Keep going to handle the other cases below |
636 | } |
637 | } else { |
638 | // No extra data, stream is at the end. We're done. |
639 | compressedDataBuffer.advanceReadPointer(distance: input.size()); |
640 | return bytesDecoded; |
641 | } |
642 | } |
643 | |
644 | if (bigMaxSize && inflateStream->avail_out == 0) { |
645 | // Need to adjust the next_out and avail_out parameters since we reached the end |
646 | // of the current range |
647 | bigMaxSize = (zlibMaxSize < size_t(maxSize - bytesDecoded)); |
648 | inflateStream->avail_out = bigMaxSize ? qsizetype(zlibMaxSize) : maxSize - bytesDecoded; |
649 | inflateStream->next_out = reinterpret_cast<Bytef *>(data + bytesDecoded); |
650 | } |
651 | |
652 | if (inflateStream->avail_in == 0 && inflateStream->avail_out > 0) { |
653 | // Grab the next input! |
654 | compressedDataBuffer.advanceReadPointer(distance: input.size()); |
655 | input = compressedDataBuffer.readPointer(); |
656 | if (size_t(input.size()) > zlibMaxSize) |
657 | input = input.sliced(pos: zlibMaxSize); |
658 | inflateStream->avail_in = input.size(); |
659 | inflateStream->next_in = |
660 | reinterpret_cast<Bytef *>(const_cast<char *>(input.data())); |
661 | } |
662 | } while (inflateStream->avail_out > 0 && inflateStream->avail_in > 0); |
663 | |
664 | compressedDataBuffer.advanceReadPointer(distance: input.size() - inflateStream->avail_in); |
665 | |
666 | return bytesDecoded; |
667 | } |
668 | |
669 | qsizetype QDecompressHelper::readBrotli(char *data, const qsizetype maxSize) |
670 | { |
671 | #if !QT_CONFIG(brotli) |
672 | Q_UNUSED(data); |
673 | Q_UNUSED(maxSize); |
674 | Q_UNREACHABLE(); |
675 | #else |
676 | qint64 bytesDecoded = 0; |
677 | |
678 | BrotliDecoderState *brotliDecoderState = toBrotliPointer(ptr: decoderPointer); |
679 | |
680 | while (decoderHasData && bytesDecoded < maxSize) { |
681 | Q_ASSERT(brotliUnconsumedDataPtr || BrotliDecoderHasMoreOutput(brotliDecoderState)); |
682 | if (brotliUnconsumedDataPtr) { |
683 | Q_ASSERT(brotliUnconsumedAmount); |
684 | size_t toRead = std::min(a: size_t(maxSize - bytesDecoded), b: brotliUnconsumedAmount); |
685 | memcpy(dest: data + bytesDecoded, src: brotliUnconsumedDataPtr, n: toRead); |
686 | bytesDecoded += toRead; |
687 | brotliUnconsumedAmount -= toRead; |
688 | brotliUnconsumedDataPtr += toRead; |
689 | if (brotliUnconsumedAmount == 0) { |
690 | brotliUnconsumedDataPtr = nullptr; |
691 | decoderHasData = false; |
692 | } |
693 | } |
694 | if (BrotliDecoderHasMoreOutput(state: brotliDecoderState) == BROTLI_TRUE) { |
695 | brotliUnconsumedDataPtr = |
696 | BrotliDecoderTakeOutput(state: brotliDecoderState, size: &brotliUnconsumedAmount); |
697 | decoderHasData = true; |
698 | } |
699 | } |
700 | if (bytesDecoded == maxSize) |
701 | return bytesDecoded; |
702 | Q_ASSERT(bytesDecoded < maxSize); |
703 | |
704 | QByteArrayView input = compressedDataBuffer.readPointer(); |
705 | const uint8_t *encodedPtr = reinterpret_cast<const uint8_t *>(input.data()); |
706 | size_t encodedBytesRemaining = input.size(); |
707 | |
708 | uint8_t *decodedPtr = reinterpret_cast<uint8_t *>(data + bytesDecoded); |
709 | size_t unusedDecodedSize = size_t(maxSize - bytesDecoded); |
710 | while (unusedDecodedSize > 0) { |
711 | auto previousUnusedDecodedSize = unusedDecodedSize; |
712 | BrotliDecoderResult result = BrotliDecoderDecompressStream( |
713 | state: brotliDecoderState, available_in: &encodedBytesRemaining, next_in: &encodedPtr, available_out: &unusedDecodedSize, |
714 | next_out: &decodedPtr, total_out: nullptr); |
715 | bytesDecoded += previousUnusedDecodedSize - unusedDecodedSize; |
716 | |
717 | switch (result) { |
718 | case BROTLI_DECODER_RESULT_ERROR: |
719 | errorStr = QLatin1String("Brotli error: %1") |
720 | .arg(args: QString::fromUtf8(utf8: BrotliDecoderErrorString( |
721 | c: BrotliDecoderGetErrorCode(state: brotliDecoderState)))); |
722 | return -1; |
723 | case BROTLI_DECODER_RESULT_SUCCESS: |
724 | BrotliDecoderDestroyInstance(state: brotliDecoderState); |
725 | decoderPointer = nullptr; |
726 | compressedDataBuffer.clear(); |
727 | return bytesDecoded; |
728 | case BROTLI_DECODER_RESULT_NEEDS_MORE_INPUT: |
729 | compressedDataBuffer.advanceReadPointer(distance: input.size()); |
730 | input = compressedDataBuffer.readPointer(); |
731 | if (!input.isEmpty()) { |
732 | encodedPtr = reinterpret_cast<const uint8_t *>(input.constData()); |
733 | encodedBytesRemaining = input.size(); |
734 | break; |
735 | } |
736 | return bytesDecoded; |
737 | case BROTLI_DECODER_RESULT_NEEDS_MORE_OUTPUT: |
738 | // Some data is leftover inside the brotli decoder, remember for next time |
739 | decoderHasData = BrotliDecoderHasMoreOutput(state: brotliDecoderState); |
740 | Q_ASSERT(unusedDecodedSize == 0); |
741 | break; |
742 | } |
743 | } |
744 | compressedDataBuffer.advanceReadPointer(distance: input.size() - encodedBytesRemaining); |
745 | return bytesDecoded; |
746 | #endif |
747 | } |
748 | |
749 | qsizetype QDecompressHelper::readZstandard(char *data, const qsizetype maxSize) |
750 | { |
751 | #if !QT_CONFIG(zstd) |
752 | Q_UNUSED(data); |
753 | Q_UNUSED(maxSize); |
754 | Q_UNREACHABLE(); |
755 | #else |
756 | ZSTD_DStream *zstdStream = toZstandardPointer(ptr: decoderPointer); |
757 | |
758 | QByteArrayView input = compressedDataBuffer.readPointer(); |
759 | ZSTD_inBuffer inBuf { .src: input.data(), .size: size_t(input.size()), .pos: 0 }; |
760 | |
761 | ZSTD_outBuffer outBuf { .dst: data, .size: size_t(maxSize), .pos: 0 }; |
762 | |
763 | qsizetype bytesDecoded = 0; |
764 | while (outBuf.pos < outBuf.size && (inBuf.pos < inBuf.size || decoderHasData)) { |
765 | size_t retValue = ZSTD_decompressStream(zds: zstdStream, output: &outBuf, input: &inBuf); |
766 | if (ZSTD_isError(code: retValue)) { |
767 | errorStr = QLatin1String("ZStandard error: %1") |
768 | .arg(args: QString::fromUtf8(utf8: ZSTD_getErrorName(code: retValue))); |
769 | return -1; |
770 | } else { |
771 | decoderHasData = false; |
772 | bytesDecoded = outBuf.pos; |
773 | // if pos == size then there may be data left over in internal buffers |
774 | if (outBuf.pos == outBuf.size) { |
775 | decoderHasData = true; |
776 | } else if (inBuf.pos == inBuf.size) { |
777 | compressedDataBuffer.advanceReadPointer(distance: input.size()); |
778 | input = compressedDataBuffer.readPointer(); |
779 | inBuf = { .src: input.constData(), .size: size_t(input.size()), .pos: 0 }; |
780 | } |
781 | } |
782 | } |
783 | compressedDataBuffer.advanceReadPointer(distance: inBuf.pos); |
784 | return bytesDecoded; |
785 | #endif |
786 | } |
787 | |
788 | QT_END_NAMESPACE |
789 |
Definitions
- ContentEncodingMapping
- contentEncodingMapping
- encodingFromByteArray
- toZlibPointer
- toBrotliPointer
- toZstandardPointer
- isSupportedEncoding
- acceptedEncoding
- ~QDecompressHelper
- setEncoding
- setEncoding
- isCountingBytes
- setCountingBytesEnabled
- uncompressedSize
- feed
- feed
- feed
- feed
- countInternal
- countInternal
- countInternal
- read
- readInternal
- setDecompressedSafetyCheckThreshold
- isPotentialArchiveBomb
- hasData
- hasDataInternal
- encodedBytesAvailable
- isValid
- errorString
- clear
- readZLib
- readBrotli
Learn Advanced QML with KDAB
Find out more