1 /*
2  * Copyright (C) 2017 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 "otautil/rangeset.h"
18 
19 #include <limits.h>
20 #include <stddef.h>
21 
22 #include <algorithm>
23 #include <string>
24 #include <utility>
25 #include <vector>
26 
27 #include <android-base/logging.h>
28 #include <android-base/parseint.h>
29 #include <android-base/stringprintf.h>
30 #include <android-base/strings.h>
31 
RangeSet(std::vector<Range> && pairs)32 RangeSet::RangeSet(std::vector<Range>&& pairs) {
33   blocks_ = 0;
34   if (pairs.empty()) {
35     LOG(ERROR) << "Invalid number of tokens";
36     return;
37   }
38 
39   for (const auto& range : pairs) {
40     if (!PushBack(range)) {
41       Clear();
42       return;
43     }
44   }
45 }
46 
Parse(const std::string & range_text)47 RangeSet RangeSet::Parse(const std::string& range_text) {
48   std::vector<std::string> pieces = android::base::Split(range_text, ",");
49   if (pieces.size() < 3) {
50     LOG(ERROR) << "Invalid range text: " << range_text;
51     return {};
52   }
53 
54   size_t num;
55   if (!android::base::ParseUint(pieces[0], &num, static_cast<size_t>(INT_MAX))) {
56     LOG(ERROR) << "Failed to parse the number of tokens: " << range_text;
57     return {};
58   }
59   if (num == 0) {
60     LOG(ERROR) << "Invalid number of tokens: " << range_text;
61     return {};
62   }
63   if (num % 2 != 0) {
64     LOG(ERROR) << "Number of tokens must be even: " << range_text;
65     return {};
66   }
67   if (num != pieces.size() - 1) {
68     LOG(ERROR) << "Mismatching number of tokens: " << range_text;
69     return {};
70   }
71 
72   std::vector<Range> pairs;
73   for (size_t i = 0; i < num; i += 2) {
74     size_t first;
75     size_t second;
76     if (!android::base::ParseUint(pieces[i + 1], &first, static_cast<size_t>(INT_MAX)) ||
77         !android::base::ParseUint(pieces[i + 2], &second, static_cast<size_t>(INT_MAX))) {
78       return {};
79     }
80     pairs.emplace_back(first, second);
81   }
82   return RangeSet(std::move(pairs));
83 }
84 
PushBack(Range range)85 bool RangeSet::PushBack(Range range) {
86   if (range.first >= range.second) {
87     LOG(ERROR) << "Empty or negative range: " << range.first << ", " << range.second;
88     return false;
89   }
90   size_t sz = range.second - range.first;
91   if (blocks_ >= SIZE_MAX - sz) {
92     LOG(ERROR) << "RangeSet size overflow";
93     return false;
94   }
95 
96   ranges_.push_back(std::move(range));
97   blocks_ += sz;
98   return true;
99 }
100 
Clear()101 void RangeSet::Clear() {
102   ranges_.clear();
103   blocks_ = 0;
104 }
105 
Split(size_t groups) const106 std::vector<RangeSet> RangeSet::Split(size_t groups) const {
107   if (ranges_.empty() || groups == 0) return {};
108 
109   if (blocks_ < groups) {
110     groups = blocks_;
111   }
112 
113   // Evenly distribute blocks, with the first few groups possibly containing one more.
114   size_t mean = blocks_ / groups;
115   std::vector<size_t> blocks_per_group(groups, mean);
116   std::fill_n(blocks_per_group.begin(), blocks_ % groups, mean + 1);
117 
118   std::vector<RangeSet> result;
119 
120   // Forward iterate Ranges and fill up each group with the desired number of blocks.
121   auto it = ranges_.cbegin();
122   Range range = *it;
123   for (const auto& blocks : blocks_per_group) {
124     RangeSet buffer;
125     size_t needed = blocks;
126     while (needed > 0) {
127       size_t range_blocks = range.second - range.first;
128       if (range_blocks > needed) {
129         // Split the current range and don't advance the iterator.
130         buffer.PushBack({ range.first, range.first + needed });
131         range.first = range.first + needed;
132         break;
133       }
134       buffer.PushBack(range);
135       it++;
136       if (it != ranges_.cend()) {
137         range = *it;
138       }
139       needed -= range_blocks;
140     }
141     result.push_back(std::move(buffer));
142   }
143   return result;
144 }
145 
ToString() const146 std::string RangeSet::ToString() const {
147   if (ranges_.empty()) {
148     return "";
149   }
150   std::string result = std::to_string(ranges_.size() * 2);
151   for (const auto& [begin, end] : ranges_) {
152     result += android::base::StringPrintf(",%zu,%zu", begin, end);
153   }
154 
155   return result;
156 }
157 
158 // Get the block number for the i-th (starting from 0) block in the RangeSet.
GetBlockNumber(size_t idx) const159 size_t RangeSet::GetBlockNumber(size_t idx) const {
160   CHECK_LT(idx, blocks_) << "Out of bound index " << idx << " (total blocks: " << blocks_ << ")";
161 
162   for (const auto& [begin, end] : ranges_) {
163     if (idx < end - begin) {
164       return begin + idx;
165     }
166     idx -= (end - begin);
167   }
168 
169   CHECK(false) << "Failed to find block number for index " << idx;
170   return 0;  // Unreachable, but to make compiler happy.
171 }
172 
173 // RangeSet has half-closed half-open bounds. For example, "3,5" contains blocks 3 and 4. So "3,5"
174 // and "5,7" are not overlapped.
Overlaps(const RangeSet & other) const175 bool RangeSet::Overlaps(const RangeSet& other) const {
176   for (const auto& [begin, end] : ranges_) {
177     for (const auto& [other_begin, other_end] : other.ranges_) {
178       // [begin, end) vs [other_begin, other_end)
179       if (!(other_begin >= end || begin >= other_end)) {
180         return true;
181       }
182     }
183   }
184   return false;
185 }
186 
GetSubRanges(size_t start_index,size_t num_of_blocks) const187 std::optional<RangeSet> RangeSet::GetSubRanges(size_t start_index, size_t num_of_blocks) const {
188   size_t end_index = start_index + num_of_blocks;  // The index of final block to read plus one
189   if (start_index > end_index || end_index > blocks_) {
190     LOG(ERROR) << "Failed to get the sub ranges for start_index " << start_index
191                << " num_of_blocks " << num_of_blocks
192                << " total number of blocks the range contains is " << blocks_;
193     return std::nullopt;
194   }
195 
196   if (num_of_blocks == 0) {
197     LOG(WARNING) << "num_of_blocks is zero when calling GetSubRanges()";
198     return RangeSet();
199   }
200 
201   RangeSet result;
202   size_t current_index = 0;
203   for (const auto& [range_start, range_end] : ranges_) {
204     CHECK_LT(range_start, range_end);
205     size_t blocks_in_range = range_end - range_start;
206     // Linear search to skip the ranges until we reach start_block.
207     if (current_index + blocks_in_range <= start_index) {
208       current_index += blocks_in_range;
209       continue;
210     }
211 
212     size_t trimmed_range_start = range_start;
213     // We have found the first block range to read, trim the heading blocks.
214     if (current_index < start_index) {
215       trimmed_range_start += start_index - current_index;
216     }
217     // Trim the trailing blocks if the last range has more blocks than desired; also return the
218     // result.
219     if (current_index + blocks_in_range >= end_index) {
220       size_t trimmed_range_end = range_end - (current_index + blocks_in_range - end_index);
221       if (!result.PushBack({ trimmed_range_start, trimmed_range_end })) {
222         return std::nullopt;
223       }
224 
225       return result;
226     }
227 
228     if (!result.PushBack({ trimmed_range_start, range_end })) {
229       return std::nullopt;
230     }
231     current_index += blocks_in_range;
232   }
233 
234   LOG(ERROR) << "Failed to construct byte ranges to read, start_block: " << start_index
235              << ", num_of_blocks: " << num_of_blocks << " total number of blocks: " << blocks_;
236   return std::nullopt;
237 }
238 
239 // Ranges in the the set should be mutually exclusive; and they're sorted by the start block.
SortedRangeSet(std::vector<Range> && pairs)240 SortedRangeSet::SortedRangeSet(std::vector<Range>&& pairs) : RangeSet(std::move(pairs)) {
241   std::sort(ranges_.begin(), ranges_.end());
242 }
243 
Insert(const Range & to_insert)244 void SortedRangeSet::Insert(const Range& to_insert) {
245   SortedRangeSet rs({ to_insert });
246   Insert(rs);
247 }
248 
249 // Insert the input SortedRangeSet; keep the ranges sorted and merge the overlap ranges.
Insert(const SortedRangeSet & rs)250 void SortedRangeSet::Insert(const SortedRangeSet& rs) {
251   if (rs.size() == 0) {
252     return;
253   }
254   // Merge and sort the two RangeSets.
255   std::vector<Range> temp = std::move(ranges_);
256   std::copy(rs.begin(), rs.end(), std::back_inserter(temp));
257   std::sort(temp.begin(), temp.end());
258 
259   Clear();
260   // Trim overlaps and insert the result back to ranges_.
261   Range to_insert = temp.front();
262   for (auto it = temp.cbegin() + 1; it != temp.cend(); it++) {
263     if (it->first <= to_insert.second) {
264       to_insert.second = std::max(to_insert.second, it->second);
265     } else {
266       ranges_.push_back(to_insert);
267       blocks_ += (to_insert.second - to_insert.first);
268       to_insert = *it;
269     }
270   }
271   ranges_.push_back(to_insert);
272   blocks_ += (to_insert.second - to_insert.first);
273 }
274 
275 // Compute the block range the file occupies, and insert that range.
Insert(size_t start,size_t len)276 void SortedRangeSet::Insert(size_t start, size_t len) {
277   Range to_insert{ start / kBlockSize, (start + len - 1) / kBlockSize + 1 };
278   Insert(to_insert);
279 }
280 
Overlaps(size_t start,size_t len) const281 bool SortedRangeSet::Overlaps(size_t start, size_t len) const {
282   RangeSet rs({ { start / kBlockSize, (start + len - 1) / kBlockSize + 1 } });
283   return Overlaps(rs);
284 }
285 
286 // Given an offset of the file, checks if the corresponding block (by considering the file as
287 // 0-based continuous block ranges) is covered by the SortedRangeSet. If so, returns the offset
288 // within this SortedRangeSet.
289 //
290 // For example, the 4106-th byte of a file is from block 1, assuming a block size of 4096-byte.
291 // The mapped offset within a SortedRangeSet("1-9 15-19") is 10.
292 //
293 // An offset of 65546 falls into the 16-th block in a file. Block 16 is contained as the 10-th
294 // item in SortedRangeSet("1-9 15-19"). So its data can be found at offset 40970 (i.e. 4096 * 10
295 // + 10) in a range represented by this SortedRangeSet.
GetOffsetInRangeSet(size_t old_offset) const296 size_t SortedRangeSet::GetOffsetInRangeSet(size_t old_offset) const {
297   size_t old_block_start = old_offset / kBlockSize;
298   size_t new_block_start = 0;
299   for (const auto& [start, end] : ranges_) {
300     // Find the index of old_block_start.
301     if (old_block_start >= end) {
302       new_block_start += (end - start);
303     } else if (old_block_start >= start) {
304       new_block_start += (old_block_start - start);
305       return (new_block_start * kBlockSize + old_offset % kBlockSize);
306     } else {
307       CHECK(false) << "block_start " << old_block_start
308                    << " is missing between two ranges: " << ToString();
309       return 0;
310     }
311   }
312   CHECK(false) << "block_start " << old_block_start
313                << " exceeds the limit of current RangeSet: " << ToString();
314   return 0;
315 }
316