1 //
2 // Copyright (C) 2020 The Android Open Source Project
3 //
4 // Licensed under the Apache License, Version 2.0 (the "License");
5 // you may not use this file except in compliance with the License.
6 // You may obtain a copy of the License at
7 //
8 // http://www.apache.org/licenses/LICENSE-2.0
9 //
10 // Unless required by applicable law or agreed to in writing, software
11 // distributed under the License is distributed on an "AS IS" BASIS,
12 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 // See the License for the specific language governing permissions and
14 // limitations under the License.
15 //
16
17 #include <sys/types.h>
18 #include <unistd.h>
19
20 #include <optional>
21 #include <unordered_map>
22 #include <unordered_set>
23 #include <vector>
24
25 #include <android-base/file.h>
26 #include <android-base/logging.h>
27 #include <libsnapshot/cow_format.h>
28 #include <libsnapshot/cow_reader.h>
29 #include <storage_literals/storage_literals.h>
30 #include <zlib.h>
31
32 #include "cow_decompress.h"
33 #include "parser_v2.h"
34 #include "parser_v3.h"
35
36 namespace android {
37 namespace snapshot {
38
39 using namespace android::storage_literals;
40
ReadCowHeader(android::base::borrowed_fd fd,CowHeaderV3 * header)41 bool ReadCowHeader(android::base::borrowed_fd fd, CowHeaderV3* header) {
42 if (lseek(fd.get(), 0, SEEK_SET) < 0) {
43 PLOG(ERROR) << "lseek header failed";
44 return false;
45 }
46
47 memset(header, 0, sizeof(*header));
48
49 if (!android::base::ReadFully(fd, &header->prefix, sizeof(header->prefix))) {
50 return false;
51 }
52 if (header->prefix.magic != kCowMagicNumber) {
53 LOG(ERROR) << "Header Magic corrupted. Magic: " << header->prefix.magic
54 << "Expected: " << kCowMagicNumber;
55 return false;
56 }
57 if (header->prefix.header_size > sizeof(CowHeaderV3)) {
58 LOG(ERROR) << "Unknown CowHeader size (got " << header->prefix.header_size
59 << " bytes, expected at most " << sizeof(CowHeaderV3) << " bytes)";
60 return false;
61 }
62
63 if (lseek(fd.get(), 0, SEEK_SET) < 0) {
64 PLOG(ERROR) << "lseek header failed";
65 return false;
66 }
67 return android::base::ReadFully(fd, header, header->prefix.header_size);
68 }
69
CowReader(ReaderFlags reader_flag,bool is_merge)70 CowReader::CowReader(ReaderFlags reader_flag, bool is_merge)
71 : fd_(-1),
72 header_(),
73 fd_size_(0),
74 block_pos_index_(std::make_shared<std::vector<int>>()),
75 reader_flag_(reader_flag),
76 is_merge_(is_merge) {}
77
CloneCowReader()78 std::unique_ptr<CowReader> CowReader::CloneCowReader() {
79 auto cow = std::make_unique<CowReader>();
80 cow->owned_fd_.reset();
81 cow->header_ = header_;
82 cow->footer_ = footer_;
83 cow->fd_size_ = fd_size_;
84 cow->last_label_ = last_label_;
85 cow->ops_ = ops_;
86 cow->merge_op_start_ = merge_op_start_;
87 cow->num_total_data_ops_ = num_total_data_ops_;
88 cow->num_ordered_ops_to_merge_ = num_ordered_ops_to_merge_;
89 cow->xor_data_loc_ = xor_data_loc_;
90 cow->block_pos_index_ = block_pos_index_;
91 cow->is_merge_ = is_merge_;
92 return cow;
93 }
94
InitForMerge(android::base::unique_fd && fd)95 bool CowReader::InitForMerge(android::base::unique_fd&& fd) {
96 owned_fd_ = std::move(fd);
97 fd_ = owned_fd_.get();
98
99 auto pos = lseek(fd_.get(), 0, SEEK_END);
100 if (pos < 0) {
101 PLOG(ERROR) << "lseek end failed";
102 return false;
103 }
104 fd_size_ = pos;
105
106 if (lseek(fd_.get(), 0, SEEK_SET) < 0) {
107 PLOG(ERROR) << "lseek header failed";
108 return false;
109 }
110
111 CHECK_GE(header_.prefix.header_size, sizeof(CowHeader));
112 CHECK_LE(header_.prefix.header_size, sizeof(header_));
113
114 if (!android::base::ReadFully(fd_, &header_, header_.prefix.header_size)) {
115 PLOG(ERROR) << "read header failed";
116 return false;
117 }
118 return true;
119 }
120
Parse(android::base::unique_fd && fd,std::optional<uint64_t> label)121 bool CowReader::Parse(android::base::unique_fd&& fd, std::optional<uint64_t> label) {
122 owned_fd_ = std::move(fd);
123 return Parse(android::base::borrowed_fd{owned_fd_}, label);
124 }
125
Parse(android::base::borrowed_fd fd,std::optional<uint64_t> label)126 bool CowReader::Parse(android::base::borrowed_fd fd, std::optional<uint64_t> label) {
127 fd_ = fd;
128
129 if (!ReadCowHeader(fd, &header_)) {
130 return false;
131 }
132
133 std::unique_ptr<CowParserBase> parser;
134 switch (header_.prefix.major_version) {
135 case 1:
136 case 2:
137 parser = std::make_unique<CowParserV2>();
138 break;
139 case 3:
140 parser = std::make_unique<CowParserV3>();
141 break;
142 default:
143 LOG(ERROR) << "Unknown version: " << header_.prefix.major_version;
144 return false;
145 }
146 if (!parser->Parse(fd, header_, label)) {
147 return false;
148 }
149
150 TranslatedCowOps ops_info;
151 if (!parser->Translate(&ops_info)) {
152 return false;
153 }
154
155 header_ = ops_info.header;
156 ops_ = std::move(ops_info.ops);
157 footer_ = parser->footer();
158 fd_size_ = parser->fd_size();
159 last_label_ = parser->last_label();
160 xor_data_loc_ = parser->xor_data_loc();
161
162 // If we're resuming a write, we're not ready to merge
163 if (label.has_value()) return true;
164 return PrepMergeOps();
165 }
166
GetMaxCompressionSize()167 uint32_t CowReader::GetMaxCompressionSize() {
168 switch (header_.prefix.major_version) {
169 case 1:
170 case 2:
171 // Old versions supports only 4KB compression.
172 return header_.block_size;
173 ;
174 case 3:
175 return header_.max_compression_size;
176 default:
177 LOG(ERROR) << "Unknown version: " << header_.prefix.major_version;
178 return 0;
179 }
180 }
181
182 //
183 // This sets up the data needed for MergeOpIter. MergeOpIter presents
184 // data in the order we intend to merge in.
185 //
186 // We merge all order sensitive ops up front, and sort the rest to allow for
187 // batch merging. Order sensitive ops can either be presented in their proper
188 // order in the cow, or be ordered by sequence ops (kCowSequenceOp), in which
189 // case we want to merge those ops first, followed by any ops not specified by
190 // new_block value by the sequence op, in sorted order.
191 // We will re-arrange the vector in such a way that
192 // kernel can batch merge. Ex:
193 //
194 // Existing COW format; All the copy operations
195 // are at the beginning.
196 // =======================================
197 // Copy-op-1 - cow_op->new_block = 1
198 // Copy-op-2 - cow_op->new_block = 2
199 // Copy-op-3 - cow_op->new_block = 3
200 // Replace-op-4 - cow_op->new_block = 6
201 // Replace-op-5 - cow_op->new_block = 4
202 // Replace-op-6 - cow_op->new_block = 8
203 // Replace-op-7 - cow_op->new_block = 9
204 // Zero-op-8 - cow_op->new_block = 7
205 // Zero-op-9 - cow_op->new_block = 5
206 // =======================================
207 //
208 // First find the operation which isn't a copy-op
209 // and then sort all the operations in descending order
210 // with the key being cow_op->new_block (source block)
211 //
212 // The data-structure will look like:
213 //
214 // =======================================
215 // Copy-op-1 - cow_op->new_block = 1
216 // Copy-op-2 - cow_op->new_block = 2
217 // Copy-op-3 - cow_op->new_block = 3
218 // Replace-op-7 - cow_op->new_block = 9
219 // Replace-op-6 - cow_op->new_block = 8
220 // Zero-op-8 - cow_op->new_block = 7
221 // Replace-op-4 - cow_op->new_block = 6
222 // Zero-op-9 - cow_op->new_block = 5
223 // Replace-op-5 - cow_op->new_block = 4
224 // =======================================
225 //
226 // Daemon will read the above data-structure in reverse-order
227 // when reading metadata. Thus, kernel will get the metadata
228 // in the following order:
229 //
230 // ========================================
231 // Replace-op-5 - cow_op->new_block = 4
232 // Zero-op-9 - cow_op->new_block = 5
233 // Replace-op-4 - cow_op->new_block = 6
234 // Zero-op-8 - cow_op->new_block = 7
235 // Replace-op-6 - cow_op->new_block = 8
236 // Replace-op-7 - cow_op->new_block = 9
237 // Copy-op-3 - cow_op->new_block = 3
238 // Copy-op-2 - cow_op->new_block = 2
239 // Copy-op-1 - cow_op->new_block = 1
240 // ===========================================
241 //
242 // When merging begins, kernel will start from the last
243 // metadata which was read: In the above format, Copy-op-1
244 // will be the first merge operation.
245 //
246 // Now, batching of the merge operations happens only when
247 // 1: origin block numbers in the base device are contiguous
248 // (cow_op->new_block) and,
249 // 2: cow block numbers which are assigned by daemon in ReadMetadata()
250 // are contiguous. These are monotonically increasing numbers.
251 //
252 // When both (1) and (2) are true, kernel will batch merge the operations.
253 // In the above case, we have to ensure that the copy operations
254 // are merged first before replace operations are done. Hence,
255 // we will not change the order of copy operations. Since,
256 // cow_op->new_block numbers are contiguous, we will ensure that the
257 // cow block numbers assigned in ReadMetadata() for these respective copy
258 // operations are not contiguous forcing kernel to issue merge for each
259 // copy operations without batch merging.
260 //
261 // For all the other operations viz. Replace and Zero op, the cow block
262 // numbers assigned by daemon will be contiguous allowing kernel to batch
263 // merge.
264 //
265 // The final format after assiging COW block numbers by the daemon will
266 // look something like:
267 //
268 // =========================================================
269 // Replace-op-5 - cow_op->new_block = 4 cow-block-num = 2
270 // Zero-op-9 - cow_op->new_block = 5 cow-block-num = 3
271 // Replace-op-4 - cow_op->new_block = 6 cow-block-num = 4
272 // Zero-op-8 - cow_op->new_block = 7 cow-block-num = 5
273 // Replace-op-6 - cow_op->new_block = 8 cow-block-num = 6
274 // Replace-op-7 - cow_op->new_block = 9 cow-block-num = 7
275 // Copy-op-3 - cow_op->new_block = 3 cow-block-num = 9
276 // Copy-op-2 - cow_op->new_block = 2 cow-block-num = 11
277 // Copy-op-1 - cow_op->new_block = 1 cow-block-num = 13
278 // ==========================================================
279 //
280 // Merge sequence will look like:
281 //
282 // Merge-1 - Batch-merge { Copy-op-1, Copy-op-2, Copy-op-3 }
283 // Merge-2 - Batch-merge {Replace-op-7, Replace-op-6, Zero-op-8,
284 // Replace-op-4, Zero-op-9, Replace-op-5 }
285 //==============================================================
PrepMergeOps()286 bool CowReader::PrepMergeOps() {
287 std::vector<int> other_ops;
288 std::vector<uint32_t> merge_op_blocks;
289 std::unordered_map<uint32_t, int> block_map;
290
291 switch (header_.prefix.major_version) {
292 case 1:
293 case 2:
294 GetSequenceDataV2(&merge_op_blocks, &other_ops, &block_map);
295 break;
296 case 3:
297 GetSequenceData(&merge_op_blocks, &other_ops, &block_map);
298 break;
299 default:
300 break;
301 }
302
303 for (auto block : merge_op_blocks) {
304 if (block_map.count(block) == 0) {
305 LOG(ERROR) << "Invalid Sequence Ops. Could not find Cow Op for new block " << block;
306 return false;
307 }
308 }
309
310 if (merge_op_blocks.size() > header_.num_merge_ops) {
311 num_ordered_ops_to_merge_ = merge_op_blocks.size() - header_.num_merge_ops;
312 } else {
313 num_ordered_ops_to_merge_ = 0;
314 }
315
316 // Sort the vector in increasing order if merging in user-space as
317 // we can batch merge them when iterating from forward.
318 //
319 // dm-snapshot-merge requires decreasing order as we iterate the blocks
320 // in reverse order.
321 if (reader_flag_ == ReaderFlags::USERSPACE_MERGE) {
322 std::sort(other_ops.begin(), other_ops.end());
323 } else {
324 std::sort(other_ops.begin(), other_ops.end(), std::greater<int>());
325 }
326
327 merge_op_blocks.insert(merge_op_blocks.end(), other_ops.begin(), other_ops.end());
328
329 num_total_data_ops_ = merge_op_blocks.size();
330 if (header_.num_merge_ops > 0) {
331 merge_op_start_ = header_.num_merge_ops;
332 }
333
334 if (is_merge_) {
335 // Metadata ops are not required for merge. Thus, just re-arrange
336 // the ops vector as required for merge operations.
337 auto merge_ops_buffer = std::make_shared<std::vector<CowOperation>>();
338 merge_ops_buffer->reserve(num_total_data_ops_);
339 for (auto block : merge_op_blocks) {
340 merge_ops_buffer->emplace_back(ops_->data()[block_map.at(block)]);
341 }
342 ops_->clear();
343 ops_ = merge_ops_buffer;
344 ops_->shrink_to_fit();
345 } else {
346 for (auto block : merge_op_blocks) {
347 block_pos_index_->push_back(block_map.at(block));
348 }
349 }
350
351 block_map.clear();
352 merge_op_blocks.clear();
353
354 return true;
355 }
356
GetSequenceDataV2(std::vector<uint32_t> * merge_op_blocks,std::vector<int> * other_ops,std::unordered_map<uint32_t,int> * block_map)357 bool CowReader::GetSequenceDataV2(std::vector<uint32_t>* merge_op_blocks,
358 std::vector<int>* other_ops,
359 std::unordered_map<uint32_t, int>* block_map) {
360 auto seq_ops_set = std::unordered_set<uint32_t>();
361 size_t num_seqs = 0;
362 size_t read;
363 for (size_t i = 0; i < ops_->size(); i++) {
364 auto& current_op = ops_->data()[i];
365
366 if (current_op.type() == kCowSequenceOp) {
367 size_t seq_len = current_op.data_length / sizeof(uint32_t);
368
369 merge_op_blocks->resize(merge_op_blocks->size() + seq_len);
370 if (!GetRawBytes(¤t_op, &merge_op_blocks->data()[num_seqs],
371 current_op.data_length, &read)) {
372 PLOG(ERROR) << "Failed to read sequence op!";
373 return false;
374 }
375 for (size_t j = num_seqs; j < num_seqs + seq_len; j++) {
376 seq_ops_set.insert(merge_op_blocks->at(j));
377 }
378 num_seqs += seq_len;
379 }
380
381 if (IsMetadataOp(current_op)) {
382 continue;
383 }
384
385 // Sequence ops must be the first ops in the stream.
386 if (seq_ops_set.empty() && IsOrderedOp(current_op)) {
387 merge_op_blocks->emplace_back(current_op.new_block);
388 } else if (seq_ops_set.count(current_op.new_block) == 0) {
389 other_ops->push_back(current_op.new_block);
390 }
391 block_map->insert({current_op.new_block, i});
392 }
393 return false;
394 }
395
GetSequenceData(std::vector<uint32_t> * merge_op_blocks,std::vector<int> * other_ops,std::unordered_map<uint32_t,int> * block_map)396 bool CowReader::GetSequenceData(std::vector<uint32_t>* merge_op_blocks, std::vector<int>* other_ops,
397 std::unordered_map<uint32_t, int>* block_map) {
398 std::unordered_set<uint32_t> seq_ops_set;
399 // read sequence ops data
400 merge_op_blocks->resize(header_.sequence_data_count);
401 if (!android::base::ReadFullyAtOffset(
402 fd_, merge_op_blocks->data(),
403 header_.sequence_data_count * sizeof(merge_op_blocks->at(0)),
404 GetSequenceOffset(header_))) {
405 PLOG(ERROR) << "failed to read sequence buffer. seq_data_count: "
406 << header_.sequence_data_count << " at offset: " << GetSequenceOffset(header_);
407 return false;
408 }
409 seq_ops_set.reserve(merge_op_blocks->size());
410 for (auto& i : *merge_op_blocks) {
411 seq_ops_set.insert(i);
412 }
413 // read ordered op data
414 for (size_t i = 0; i < ops_->size(); i++) {
415 auto& current_op = ops_->data()[i];
416 // Sequence ops must be the first ops in the stream.
417 if (seq_ops_set.empty()) {
418 merge_op_blocks->emplace_back(current_op.new_block);
419 } else if (seq_ops_set.count(current_op.new_block) == 0) {
420 other_ops->push_back(current_op.new_block);
421 }
422 block_map->insert({current_op.new_block, i});
423 }
424 return true;
425 }
426
VerifyMergeOps()427 bool CowReader::VerifyMergeOps() {
428 auto itr = GetMergeOpIter(true);
429 std::unordered_map<uint64_t, const CowOperation*> overwritten_blocks;
430 bool non_ordered_op_found = false;
431
432 while (!itr->AtEnd()) {
433 const auto& op = itr->Get();
434 uint64_t offset;
435
436 // Op should not be a metadata
437 if (IsMetadataOp(*op)) {
438 LOG(ERROR) << "Metadata op: " << op << " found during merge sequence";
439 return false;
440 }
441
442 // Sequence ops should contain all the ordered ops followed
443 // by Replace and Zero ops. If we find the first op which
444 // is not ordered, that means all ordered ops processing
445 // has been completed.
446 if (!IsOrderedOp(*op)) {
447 non_ordered_op_found = true;
448 }
449
450 // Since, all ordered ops processing has been completed,
451 // check that the subsequent ops are not ordered.
452 if (non_ordered_op_found && IsOrderedOp(*op)) {
453 LOG(ERROR) << "Invalid sequence - non-ordered and ordered ops"
454 << " cannot be mixed during sequence generation";
455 return false;
456 }
457
458 if (!GetSourceOffset(op, &offset)) {
459 itr->Next();
460 continue;
461 }
462
463 uint64_t block = GetBlockFromOffset(header_, offset);
464 bool misaligned = (GetBlockRelativeOffset(header_, offset) != 0);
465
466 const CowOperation* overwrite = nullptr;
467 if (overwritten_blocks.count(block)) {
468 overwrite = overwritten_blocks[block];
469 LOG(ERROR) << "Invalid Sequence! Block needed for op:\n"
470 << op << "\noverwritten by previously merged op:\n"
471 << *overwrite;
472 }
473 if (misaligned && overwritten_blocks.count(block + 1)) {
474 overwrite = overwritten_blocks[block + 1];
475 LOG(ERROR) << "Invalid Sequence! Block needed for op:\n"
476 << op << "\noverwritten by previously merged op:\n"
477 << *overwrite;
478 }
479 if (overwrite != nullptr) return false;
480 overwritten_blocks[op->new_block] = op;
481 itr->Next();
482 }
483 return true;
484 }
485
GetFooter(CowFooter * footer)486 bool CowReader::GetFooter(CowFooter* footer) {
487 if (!footer_) return false;
488 *footer = footer_.value();
489 return true;
490 }
491
GetLastLabel(uint64_t * label)492 bool CowReader::GetLastLabel(uint64_t* label) {
493 if (!last_label_) return false;
494 *label = last_label_.value();
495 return true;
496 }
497
498 class CowOpIter final : public ICowOpIter {
499 public:
500 CowOpIter(std::shared_ptr<std::vector<CowOperation>>& ops, uint64_t start);
501
502 bool AtEnd() override;
503 const CowOperation* Get() override;
504 void Next() override;
505
506 void Prev() override;
507 bool AtBegin() override;
508
509 private:
510 std::shared_ptr<std::vector<CowOperation>> ops_;
511 std::vector<CowOperation>::iterator op_iter_;
512 };
513
CowOpIter(std::shared_ptr<std::vector<CowOperation>> & ops,uint64_t start)514 CowOpIter::CowOpIter(std::shared_ptr<std::vector<CowOperation>>& ops, uint64_t start) {
515 ops_ = ops;
516 op_iter_ = ops_->begin() + start;
517 }
518
AtBegin()519 bool CowOpIter::AtBegin() {
520 return op_iter_ == ops_->begin();
521 }
522
Prev()523 void CowOpIter::Prev() {
524 CHECK(!AtBegin());
525 op_iter_--;
526 }
527
AtEnd()528 bool CowOpIter::AtEnd() {
529 return op_iter_ == ops_->end();
530 }
531
Next()532 void CowOpIter::Next() {
533 CHECK(!AtEnd());
534 op_iter_++;
535 }
536
Get()537 const CowOperation* CowOpIter::Get() {
538 CHECK(!AtEnd());
539 return &(*op_iter_);
540 }
541
542 class CowRevMergeOpIter final : public ICowOpIter {
543 public:
544 explicit CowRevMergeOpIter(std::shared_ptr<std::vector<CowOperation>> ops,
545 std::shared_ptr<std::vector<int>> block_pos_index, uint64_t start);
546
547 bool AtEnd() override;
548 const CowOperation* Get() override;
549 void Next() override;
550
551 void Prev() override;
552 bool AtBegin() override;
553
554 private:
555 std::shared_ptr<std::vector<CowOperation>> ops_;
556 std::vector<int>::reverse_iterator block_riter_;
557 std::shared_ptr<std::vector<int>> cow_op_index_vec_;
558 uint64_t start_;
559 };
560
561 class CowMergeOpIter final : public ICowOpIter {
562 public:
563 explicit CowMergeOpIter(std::shared_ptr<std::vector<CowOperation>> ops,
564 std::shared_ptr<std::vector<int>> block_pos_index, uint64_t start);
565
566 bool AtEnd() override;
567 const CowOperation* Get() override;
568 void Next() override;
569
570 void Prev() override;
571 bool AtBegin() override;
572
573 private:
574 std::shared_ptr<std::vector<CowOperation>> ops_;
575 std::vector<int>::iterator block_iter_;
576 std::shared_ptr<std::vector<int>> cow_op_index_vec_;
577 uint64_t start_;
578 };
579
CowMergeOpIter(std::shared_ptr<std::vector<CowOperation>> ops,std::shared_ptr<std::vector<int>> block_pos_index,uint64_t start)580 CowMergeOpIter::CowMergeOpIter(std::shared_ptr<std::vector<CowOperation>> ops,
581 std::shared_ptr<std::vector<int>> block_pos_index, uint64_t start) {
582 ops_ = ops;
583 start_ = start;
584 cow_op_index_vec_ = block_pos_index;
585 block_iter_ = cow_op_index_vec_->begin() + start;
586 }
587
AtBegin()588 bool CowMergeOpIter::AtBegin() {
589 return block_iter_ == cow_op_index_vec_->begin();
590 }
591
Prev()592 void CowMergeOpIter::Prev() {
593 CHECK(!AtBegin());
594 block_iter_--;
595 }
596
AtEnd()597 bool CowMergeOpIter::AtEnd() {
598 return block_iter_ == cow_op_index_vec_->end();
599 }
600
Next()601 void CowMergeOpIter::Next() {
602 CHECK(!AtEnd());
603 block_iter_++;
604 }
605
Get()606 const CowOperation* CowMergeOpIter::Get() {
607 CHECK(!AtEnd());
608 return &ops_->data()[*block_iter_];
609 }
610
CowRevMergeOpIter(std::shared_ptr<std::vector<CowOperation>> ops,std::shared_ptr<std::vector<int>> block_pos_index,uint64_t start)611 CowRevMergeOpIter::CowRevMergeOpIter(std::shared_ptr<std::vector<CowOperation>> ops,
612 std::shared_ptr<std::vector<int>> block_pos_index,
613 uint64_t start) {
614 ops_ = ops;
615 start_ = start;
616 cow_op_index_vec_ = block_pos_index;
617 block_riter_ = cow_op_index_vec_->rbegin();
618 }
619
AtBegin()620 bool CowRevMergeOpIter::AtBegin() {
621 return block_riter_ == cow_op_index_vec_->rbegin();
622 }
623
Prev()624 void CowRevMergeOpIter::Prev() {
625 CHECK(!AtBegin());
626 block_riter_--;
627 }
628
AtEnd()629 bool CowRevMergeOpIter::AtEnd() {
630 return block_riter_ == cow_op_index_vec_->rend() - start_;
631 }
632
Next()633 void CowRevMergeOpIter::Next() {
634 CHECK(!AtEnd());
635 block_riter_++;
636 }
637
Get()638 const CowOperation* CowRevMergeOpIter::Get() {
639 CHECK(!AtEnd());
640 return &ops_->data()[*block_riter_];
641 }
642
GetOpIter(bool merge_progress)643 std::unique_ptr<ICowOpIter> CowReader::GetOpIter(bool merge_progress) {
644 return std::make_unique<CowOpIter>(ops_, merge_progress ? merge_op_start_ : 0);
645 }
646
GetRevMergeOpIter(bool ignore_progress)647 std::unique_ptr<ICowOpIter> CowReader::GetRevMergeOpIter(bool ignore_progress) {
648 return std::make_unique<CowRevMergeOpIter>(ops_, block_pos_index_,
649 ignore_progress ? 0 : merge_op_start_);
650 }
651
GetMergeOpIter(bool ignore_progress)652 std::unique_ptr<ICowOpIter> CowReader::GetMergeOpIter(bool ignore_progress) {
653 return std::make_unique<CowMergeOpIter>(ops_, block_pos_index_,
654 ignore_progress ? 0 : merge_op_start_);
655 }
656
GetRawBytes(const CowOperation * op,void * buffer,size_t len,size_t * read)657 bool CowReader::GetRawBytes(const CowOperation* op, void* buffer, size_t len, size_t* read) {
658 switch (op->type()) {
659 case kCowSequenceOp:
660 case kCowReplaceOp:
661 case kCowXorOp:
662 return GetRawBytes(op->source(), buffer, len, read);
663 default:
664 LOG(ERROR) << "Cannot get raw bytes of non-data op: " << *op;
665 return false;
666 }
667 }
668
GetRawBytes(uint64_t offset,void * buffer,size_t len,size_t * read)669 bool CowReader::GetRawBytes(uint64_t offset, void* buffer, size_t len, size_t* read) {
670 // Validate the offset, taking care to acknowledge possible overflow of offset+len.
671 if (offset < header_.prefix.header_size || offset >= fd_size_ || offset + len > fd_size_ ||
672 len >= fd_size_) {
673 LOG(ERROR) << "invalid data offset: " << offset << ", " << len << " bytes";
674 return false;
675 }
676 if (lseek(fd_.get(), offset, SEEK_SET) < 0) {
677 PLOG(ERROR) << "lseek to read raw bytes failed";
678 return false;
679 }
680 ssize_t rv = TEMP_FAILURE_RETRY(::read(fd_.get(), buffer, len));
681 if (rv < 0) {
682 PLOG(ERROR) << "read failed";
683 return false;
684 }
685 *read = rv;
686 return true;
687 }
688
689 class CowDataStream final : public IByteStream {
690 public:
CowDataStream(CowReader * reader,uint64_t offset,size_t data_length)691 CowDataStream(CowReader* reader, uint64_t offset, size_t data_length)
692 : reader_(reader), offset_(offset), data_length_(data_length) {
693 remaining_ = data_length_;
694 }
695
Read(void * buffer,size_t length)696 ssize_t Read(void* buffer, size_t length) override {
697 size_t to_read = std::min(length, remaining_);
698 if (!to_read) {
699 return 0;
700 }
701 size_t read;
702 if (!reader_->GetRawBytes(offset_, buffer, to_read, &read)) {
703 return -1;
704 }
705 offset_ += read;
706 remaining_ -= read;
707 return read;
708 }
709
Size() const710 size_t Size() const override { return data_length_; }
711
712 private:
713 CowReader* reader_;
714 uint64_t offset_;
715 size_t data_length_;
716 size_t remaining_;
717 };
718
GetCompressionType()719 uint8_t CowReader::GetCompressionType() {
720 return header_.compression_algorithm;
721 }
722
ReadData(const CowOperation * op,void * buffer,size_t buffer_size,size_t ignore_bytes)723 ssize_t CowReader::ReadData(const CowOperation* op, void* buffer, size_t buffer_size,
724 size_t ignore_bytes) {
725 std::unique_ptr<IDecompressor> decompressor;
726 const size_t op_buf_size = CowOpCompressionSize(op, header_.block_size);
727 if (!op_buf_size) {
728 LOG(ERROR) << "Compression size is zero. op: " << *op;
729 return -1;
730 }
731 switch (GetCompressionType()) {
732 case kCowCompressNone:
733 break;
734 case kCowCompressGz:
735 decompressor = IDecompressor::Gz();
736 break;
737 case kCowCompressBrotli:
738 decompressor = IDecompressor::Brotli();
739 break;
740 case kCowCompressZstd:
741 if (op_buf_size != op->data_length) {
742 decompressor = IDecompressor::Zstd();
743 }
744 break;
745 case kCowCompressLz4:
746 if (op_buf_size != op->data_length) {
747 decompressor = IDecompressor::Lz4();
748 }
749 break;
750 default:
751 LOG(ERROR) << "Unknown compression type: " << GetCompressionType();
752 return -1;
753 }
754
755 uint64_t offset;
756 if (op->type() == kCowXorOp) {
757 offset = xor_data_loc_->at(op->new_block);
758 } else {
759 offset = op->source();
760 }
761 if (!decompressor ||
762 ((op->data_length == op_buf_size) && (header_.prefix.major_version == 3))) {
763 CowDataStream stream(this, offset + ignore_bytes, op->data_length - ignore_bytes);
764 return stream.ReadFully(buffer, buffer_size);
765 }
766
767 CowDataStream stream(this, offset, op->data_length);
768 decompressor->set_stream(&stream);
769 return decompressor->Decompress(buffer, buffer_size, op_buf_size, ignore_bytes);
770 }
771
GetSourceOffset(const CowOperation * op,uint64_t * source_offset)772 bool CowReader::GetSourceOffset(const CowOperation* op, uint64_t* source_offset) {
773 switch (op->type()) {
774 case kCowCopyOp:
775 *source_offset = op->source() * header_.block_size;
776 return true;
777 case kCowXorOp:
778 *source_offset = op->source();
779 return true;
780 default:
781 return false;
782 }
783 }
784
785 } // namespace snapshot
786 } // namespace android
787