/* vim:set ts=2 sw=2 sts=2 et: */ /** * \author Marcus Holland-Moritz (github@mhxnet.de) * \copyright Copyright (c) Marcus Holland-Moritz * * This file is part of dwarfs. * * dwarfs is free software: you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation, either version 3 of the License, or * (at your option) any later version. * * dwarfs is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with dwarfs. If not, see . */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "mmap_mock.h" #include "test_helpers.h" #include "test_logger.h" using namespace dwarfs; namespace { class mock_cached_block : public reader::internal::cached_block { public: mock_cached_block() = default; mock_cached_block(std::span span) : span_{span} {} size_t range_end() const override { return span_ ? span_->size() : 0; } const uint8_t* data() const override { return span_ ? span_->data() : nullptr; } void decompress_until(size_t) override {} size_t uncompressed_size() const override { return 0; } void touch() override {} bool last_used_before(std::chrono::steady_clock::time_point) const override { return false; } bool any_pages_swapped_out(std::vector&) const override { return false; } private: std::optional> span_; }; } // namespace TEST(block_range, uncompressed) { std::vector data(100); std::iota(data.begin(), data.end(), 0); { reader::block_range range{data.data(), 0, data.size()}; EXPECT_EQ(range.data(), data.data()); EXPECT_EQ(range.size(), 100); EXPECT_TRUE(std::equal(range.begin(), range.end(), data.begin())); } { reader::block_range range{data.data(), 10, 20}; EXPECT_EQ(range.size(), 20); EXPECT_TRUE(std::equal(range.begin(), range.end(), data.begin() + 10)); } EXPECT_THAT([] { reader::block_range range(nullptr, 0, 0); }, ::testing::ThrowsMessage( ::testing::HasSubstr("block_range: block data is null"))); } TEST(block_range, compressed) { std::vector data(100); std::iota(data.begin(), data.end(), 0); { auto block = std::make_shared(data); reader::block_range range{block, 0, data.size()}; EXPECT_EQ(range.data(), data.data()); EXPECT_EQ(range.size(), 100); EXPECT_TRUE(std::equal(range.begin(), range.end(), data.begin())); } { auto block = std::make_shared(data); reader::block_range range{block, 10, 20}; EXPECT_EQ(range.size(), 20); EXPECT_TRUE(std::equal(range.begin(), range.end(), data.begin() + 10)); } EXPECT_THAT( [] { auto block = std::make_shared(); reader::block_range range(block, 0, 0); }, ::testing::ThrowsMessage( ::testing::HasSubstr("block_range: block data is null"))); EXPECT_THAT( [&] { auto block = std::make_shared(data); reader::block_range range(block, 100, 1); }, ::testing::ThrowsMessage( ::testing::HasSubstr("block_range: size out of range (101 > 100)"))); } class options_test : public ::testing::TestWithParam { DWARFS_SLOW_FIXTURE }; TEST_P(options_test, cache_stress) { static constexpr size_t num_threads{8}; static constexpr size_t num_read_reqs{1024}; auto const& cache_opts = GetParam(); auto os = std::make_shared(); { static constexpr size_t const num_files{256}; static constexpr size_t const avg_size{5000}; static constexpr size_t const max_size{16 * avg_size}; std::mt19937_64 rng{42}; std::exponential_distribution<> size_dist{1.0 / avg_size}; os->add("", {1, 040755, 1, 0, 0, 10, 42, 0, 0, 0}); for (size_t x = 0; x < num_files; ++x) { auto size = std::min(max_size, static_cast(size_dist(rng))); os->add_file(std::to_string(x), test::create_random_string(size, 32, 127, rng)); } } std::shared_ptr mm; { auto fa = std::make_shared(); test::test_iolayer iol{os, fa}; #if defined(DWARFS_HAVE_LIBBROTLI) std::string compression{"brotli:quality=0"}; #elif defined(DWARFS_HAVE_LIBLZMA) std::string compression{"lzma:level=0"}; #else std::string compression{"zstd:level=5"}; #endif std::vector args{"mkdwarfs", "-i", "/", "-o", "-", "-l3", "-S16", "-C", compression}; EXPECT_EQ(0, tool::main_adapter(tool::mkdwarfs_main)(args, iol.get())); mm = std::make_shared(iol.out()); } test::test_logger lgr(logger::TRACE); reader::filesystem_options opts{ .block_cache = cache_opts, }; reader::filesystem_v2 fs(lgr, *os, mm, opts); EXPECT_NO_THROW(fs.set_cache_tidy_config( {.strategy = reader::cache_tidy_strategy::NONE})); EXPECT_THAT( [&] { fs.set_cache_tidy_config({ .strategy = reader::cache_tidy_strategy::BLOCK_SWAPPED_OUT, .interval = std::chrono::seconds::zero(), }); }, ::testing::ThrowsMessage( ::testing::HasSubstr("tidy interval is zero"))); fs.set_cache_tidy_config({ .strategy = reader::cache_tidy_strategy::BLOCK_SWAPPED_OUT, }); fs.set_num_workers(cache_opts.num_workers); fs.set_cache_tidy_config({ .strategy = reader::cache_tidy_strategy::EXPIRY_TIME, .interval = std::chrono::milliseconds(1), .expiry_time = std::chrono::milliseconds(2), }); std::vector inodes; fs.walk([&](auto e) { inodes.push_back(e.inode()); }); std::uniform_int_distribution inode_dist(0, inodes.size() - 1); struct read_request { reader::inode_view inode; size_t offset; size_t size; }; std::vector> data(num_threads); std::mt19937_64 rng{42}; for (auto& reqs : data) { while (reqs.size() < num_read_reqs) { auto iv = inodes[inode_dist(rng)]; auto stat = fs.getattr(iv); if (stat.is_regular_file()) { auto offset = rng() % stat.size(); auto size = rng() % (stat.size() - offset); reqs.push_back({iv, offset, size}); while (reqs.size() < num_read_reqs && offset + size < static_cast(stat.size() / 2)) { offset += rng() % (stat.size() - (offset + size)); size = rng() % (stat.size() - offset); reqs.push_back({iv, offset, size}); } } } } std::vector threads; std::vector success(num_threads); for (auto const& [i, reqs] : ranges::views::enumerate(data)) { auto& succ = success[i]; // TODO: preqs is a workaround for older Clang versions threads.emplace_back([&, preqs = &reqs] { for (auto const& req : *preqs) { auto fh = fs.open(req.inode); std::error_code ec; auto ranges = fs.readv(fh, req.size, req.offset, ec); if (ec) { std::cerr << "read failed: " << ec.message() << std::endl; std::terminate(); } try { for (auto& b : ranges) { b.get(); } } catch (std::exception const& e) { std::cerr << "read failed: " << e.what() << std::endl; std::terminate(); } ++succ; } }); } for (auto& t : threads) { t.join(); } for (auto const& [i, reqs] : ranges::views::enumerate(success)) { EXPECT_EQ(reqs, num_read_reqs) << i; } } namespace { using reader::block_cache_options; constexpr std::array const cache_options{ block_cache_options{.max_bytes = 0, .num_workers = 0}, block_cache_options{.max_bytes = 256 * 1024, .num_workers = 0}, block_cache_options{.max_bytes = 256 * 1024, .num_workers = 1}, block_cache_options{.max_bytes = 256 * 1024, .num_workers = 3}, block_cache_options{.max_bytes = 256 * 1024, .num_workers = 7}, block_cache_options{.max_bytes = 1024 * 1024, .num_workers = 5}, block_cache_options{ .max_bytes = 1024 * 1024, .num_workers = 5, .decompress_ratio = 0.1}, block_cache_options{ .max_bytes = 1024 * 1024, .num_workers = 5, .decompress_ratio = 0.5}, block_cache_options{ .max_bytes = 1024 * 1024, .num_workers = 5, .decompress_ratio = 0.9}, block_cache_options{.max_bytes = 512 * 1024, .num_workers = 4, .mm_release = false, .disable_block_integrity_check = true}, }; } // namespace INSTANTIATE_TEST_SUITE_P(block_cache, options_test, ::testing::ValuesIn(cache_options));