blob: 384e59fed9051192069bf4d355f6f478bebde936 [file] [log] [blame]
/*
* Copyright (C) 2019 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specic language governing permissions and
* limitations under the License.
*/
#include "include/libfuse_jni/RedactionInfo.h"
#include <android-base/logging.h>
using std::unique_ptr;
using std::vector;
namespace mediaprovider {
namespace fuse {
/**
* Merges any overlapping ranges into 1 range.
*
* Given ranges should be sorted, and they remain sorted.
*/
static void mergeOverlappingRedactionRanges(vector<RedactionRange>& ranges) {
if (ranges.size() == 0) return;
int newRangesSize = ranges.size();
for (int i = 0; i < ranges.size() - 1; ++i) {
if (ranges[i].second >= ranges[i + 1].first) {
ranges[i + 1].first = ranges[i].first;
ranges[i + 1].second = std::max(ranges[i].second, ranges[i + 1].second);
// Invalidate the redundant range
ranges[i].first = LONG_MAX;
ranges[i].second = LONG_MAX;
newRangesSize--;
}
}
if (newRangesSize < ranges.size()) {
// Move invalid ranges to end of array
std::sort(ranges.begin(), ranges.end());
ranges.resize(newRangesSize);
}
}
/**
* Removes any range with zero size.
*
* If ranges are modified, it will be guaranteed to be sorted.
*/
static void removeZeroSizeRedactionRanges(vector<RedactionRange>& ranges) {
int newRangesSize = ranges.size();
for (int i = 0; i < ranges.size(); ++i) {
if (ranges[i].first == ranges[i].second) {
// This redaction range is of length zero, hence we don't have anything
// to redact in this range, so remove it from the redaction_ranges_.
ranges[i].first = LONG_MAX;
ranges[i].second = LONG_MAX;
newRangesSize--;
}
}
if (newRangesSize < ranges.size()) {
// Move invalid ranges to end of array
std::sort(ranges.begin(), ranges.end());
ranges.resize(newRangesSize);
}
}
/**
* Determine whether the read request overlaps with the redaction ranges
* defined by the given RedactionInfo.
*
* This function assumes redaction_ranges_ within RedactionInfo is sorted.
*/
bool RedactionInfo::hasOverlapWithReadRequest(size_t size, off64_t off) const {
if (!isRedactionNeeded() || off >= redaction_ranges_.back().second ||
off + size <= redaction_ranges_.front().first) {
return false;
}
return true;
}
/**
* Sets the redaction ranges in RedactionInfo, sort the ranges and merge
* overlapping ranges.
*/
void RedactionInfo::processRedactionRanges(int redaction_ranges_num,
const off64_t* redaction_ranges) {
redaction_ranges_.resize(redaction_ranges_num);
for (int i = 0; i < redaction_ranges_num; ++i) {
redaction_ranges_[i].first = static_cast<off64_t>(redaction_ranges[2 * i]);
redaction_ranges_[i].second = static_cast<off64_t>(redaction_ranges[2 * i + 1]);
}
std::sort(redaction_ranges_.begin(), redaction_ranges_.end());
removeZeroSizeRedactionRanges(redaction_ranges_);
mergeOverlappingRedactionRanges(redaction_ranges_);
}
int RedactionInfo::size() const {
return redaction_ranges_.size();
}
bool RedactionInfo::isRedactionNeeded() const {
return size() > 0;
}
RedactionInfo::RedactionInfo(int redaction_ranges_num, const off64_t* redaction_ranges) {
if (redaction_ranges == 0) return;
processRedactionRanges(redaction_ranges_num, redaction_ranges);
}
unique_ptr<vector<RedactionRange>> RedactionInfo::getOverlappingRedactionRanges(size_t size,
off64_t off) const {
if (hasOverlapWithReadRequest(size, off)) {
const off64_t start = off;
const off64_t end = static_cast<off64_t>(off + size);
auto first_redaction = redaction_ranges_.end();
auto last_redaction = redaction_ranges_.begin();
for (auto iter = redaction_ranges_.begin(); iter != redaction_ranges_.end(); ++iter) {
if (iter->second > start && iter->first < end) {
if (iter < first_redaction) first_redaction = iter;
if (iter > last_redaction) last_redaction = iter;
}
if (iter->first >= end) {
break;
}
}
if (first_redaction != redaction_ranges_.end()) {
CHECK(first_redaction <= last_redaction);
return std::make_unique<vector<RedactionRange>>(first_redaction, last_redaction + 1);
}
}
return std::make_unique<vector<RedactionRange>>();
}
void RedactionInfo::getReadRanges(off64_t off, size_t size, std::vector<ReadRange>* out) const {
const auto rr = getOverlappingRedactionRanges(size, off);
const size_t num_ranges = rr->size();
if (num_ranges == 0) {
return;
}
const off64_t read_start = off;
const off64_t read_end = static_cast<off64_t>(read_start + size);
// The algorithm for computing redaction ranges is very simple.
// Given a set of overlapping redaction ranges [s1, e1) [s2, e2) .. [sN, eN) for a read
// [s, e)
//
// We can construct a series of indices that we know will be the starts of every read range
// that we intend to return. Then, it's relatively simple to compute the lengths of the ranges.
// Also note that the read ranges we return always alternate in whether they're redacting or
// not. i.e, we will never return two consecutive redacting ranges or non redacting ranges.
std::vector<off64_t> sorted_indices;
// Compute the list of indices -- this list will always contain { e1, s2, e2... sN }
// In addition, it may contain s or both (s and s1), depending on the start index.
// In addition, it may contain e or both (e and eN), depending on the end index.
//
// For a concrete example, consider ranges [10, 20) and [30, 40)
// For a read [0, 60) : sorted_indices will be { 0, 10, 20, 30, 40, 60 } is_first = false
// For a read [15, 60) : sorted_indices will be { 15, 20, 30, 40, 60 } is_first = true
// For a read [0, 35) : sorted_indices will be { 0, 10, 20, 30, 35 } is_first = false
// For a read [15, 35) : sorted_indices will be { 15, 20, 30, 35 } is_first = true
for (int i = 0; i < num_ranges; ++i) {
sorted_indices.push_back(rr->at(i).first);
sorted_indices.push_back(rr->at(i).second);
}
// Find the right position for read_start in sorted_indices
// Either insert at the beginning or replace s1 with read_start
bool is_first_range_redaction = true;
if (read_start < rr->at(0).first) {
is_first_range_redaction = false;
sorted_indices.insert(sorted_indices.begin(), read_start);
} else {
sorted_indices.front() = read_start;
}
// Find the right position for read_end in sorted_indices
// Either insert at the end or replace eN with read_end
if (read_end > rr->at(num_ranges - 1).second) {
sorted_indices.push_back(read_end);
} else {
sorted_indices.back() = read_end;
}
bool is_redaction = is_first_range_redaction;
for (int i = 0; i < (sorted_indices.size() - 1); ++i) {
const off64_t read_size = sorted_indices[i + 1] - sorted_indices[i];
CHECK(read_size > 0);
out->push_back(ReadRange(sorted_indices[i], read_size, is_redaction));
is_redaction = !is_redaction;
}
}
} // namespace fuse
} // namespace mediaprovider