jcs.org
Serenity Operating System
1/*
2 * Copyright (c) 2020-2022, the SerenityOS developers.
3 * Copyright (c) 2021, Idan Horowitz <idan.horowitz@serenityos.org>
4 *
5 * SPDX-License-Identifier: BSD-2-Clause
6 */
7
8#include <LibCompress/Gzip.h>
9
10#include <AK/DeprecatedString.h>
11#include <AK/MemoryStream.h>
12#include <LibCore/DateTime.h>
13
14namespace Compress {
15
16bool GzipDecompressor::is_likely_compressed(ReadonlyBytes bytes)
17{
18 return bytes.size() >= 2 && bytes[0] == gzip_magic_1 && bytes[1] == gzip_magic_2;
19}
20
21bool BlockHeader::valid_magic_number() const
22{
23 return identification_1 == gzip_magic_1 && identification_2 == gzip_magic_2;
24}
25
26bool BlockHeader::supported_by_implementation() const
27{
28 if (compression_method != 0x08) {
29 // RFC 1952 does not define any compression methods other than deflate.
30 return false;
31 }
32
33 if (flags > Flags::MAX) {
34 // RFC 1952 does not define any more flags.
35 return false;
36 }
37
38 return true;
39}
40
41ErrorOr<NonnullOwnPtr<GzipDecompressor::Member>> GzipDecompressor::Member::construct(BlockHeader header, Stream& stream)
42{
43 auto deflate_stream = TRY(DeflateDecompressor::construct(MaybeOwned<Stream>(stream)));
44 return TRY(adopt_nonnull_own_or_enomem(new (nothrow) Member(header, move(deflate_stream))));
45}
46
47GzipDecompressor::Member::Member(BlockHeader header, NonnullOwnPtr<DeflateDecompressor> stream)
48 : m_header(header)
49 , m_stream(move(stream))
50{
51}
52
53GzipDecompressor::GzipDecompressor(NonnullOwnPtr<Stream> stream)
54 : m_input_stream(move(stream))
55{
56}
57
58GzipDecompressor::~GzipDecompressor()
59{
60 m_current_member.clear();
61}
62
63ErrorOr<Bytes> GzipDecompressor::read_some(Bytes bytes)
64{
65 size_t total_read = 0;
66 while (total_read < bytes.size()) {
67 if (is_eof())
68 break;
69
70 auto slice = bytes.slice(total_read);
71
72 if (m_current_member) {
73 auto current_slice = TRY(current_member().m_stream->read_some(slice));
74 current_member().m_checksum.update(current_slice);
75 current_member().m_nread += current_slice.size();
76
77 if (current_slice.size() < slice.size()) {
78 u32 crc32 = TRY(m_input_stream->read_value<LittleEndian<u32>>());
79 u32 input_size = TRY(m_input_stream->read_value<LittleEndian<u32>>());
80
81 if (crc32 != current_member().m_checksum.digest())
82 return Error::from_string_literal("Stored CRC32 does not match the calculated CRC32 of the current member");
83
84 if (input_size != current_member().m_nread)
85 return Error::from_string_literal("Input size does not match the number of read bytes");
86
87 m_current_member.clear();
88
89 total_read += current_slice.size();
90 continue;
91 }
92
93 total_read += current_slice.size();
94 continue;
95 } else {
96 auto current_partial_header_slice = Bytes { m_partial_header, sizeof(BlockHeader) }.slice(m_partial_header_offset);
97 auto current_partial_header_data = TRY(m_input_stream->read_some(current_partial_header_slice));
98 m_partial_header_offset += current_partial_header_data.size();
99
100 if (is_eof())
101 break;
102
103 if (m_partial_header_offset < sizeof(BlockHeader)) {
104 break; // partial header read
105 }
106 m_partial_header_offset = 0;
107
108 BlockHeader header = *(reinterpret_cast<BlockHeader*>(m_partial_header));
109
110 if (!header.valid_magic_number())
111 return Error::from_string_literal("Header does not have a valid magic number");
112
113 if (!header.supported_by_implementation())
114 return Error::from_string_literal("Header is not supported by implementation");
115
116 if (header.flags & Flags::FEXTRA) {
117 u16 subfield_id = TRY(m_input_stream->read_value<LittleEndian<u16>>());
118 u16 length = TRY(m_input_stream->read_value<LittleEndian<u16>>());
119 TRY(m_input_stream->discard(length));
120 (void)subfield_id;
121 }
122
123 auto discard_string = [&]() -> ErrorOr<void> {
124 char next_char;
125 do {
126 next_char = TRY(m_input_stream->read_value<char>());
127 } while (next_char);
128
129 return {};
130 };
131
132 if (header.flags & Flags::FNAME)
133 TRY(discard_string());
134
135 if (header.flags & Flags::FCOMMENT)
136 TRY(discard_string());
137
138 if (header.flags & Flags::FHCRC) {
139 u16 crc = TRY(m_input_stream->read_value<LittleEndian<u16>>());
140 // FIXME: we should probably verify this instead of just assuming it matches
141 (void)crc;
142 }
143
144 m_current_member = TRY(Member::construct(header, *m_input_stream));
145 continue;
146 }
147 }
148 return bytes.slice(0, total_read);
149}
150
151Optional<DeprecatedString> GzipDecompressor::describe_header(ReadonlyBytes bytes)
152{
153 if (bytes.size() < sizeof(BlockHeader))
154 return {};
155
156 auto& header = *(reinterpret_cast<BlockHeader const*>(bytes.data()));
157 if (!header.valid_magic_number() || !header.supported_by_implementation())
158 return {};
159
160 LittleEndian<u32> original_size = *reinterpret_cast<u32 const*>(bytes.offset(bytes.size() - sizeof(u32)));
161 return DeprecatedString::formatted("last modified: {}, original size {}", Core::DateTime::from_timestamp(header.modification_time).to_deprecated_string(), (u32)original_size);
162}
163
164ErrorOr<ByteBuffer> GzipDecompressor::decompress_all(ReadonlyBytes bytes)
165{
166 auto memory_stream = TRY(try_make<FixedMemoryStream>(bytes));
167 auto gzip_stream = make<GzipDecompressor>(move(memory_stream));
168 AllocatingMemoryStream output_stream;
169
170 auto buffer = TRY(ByteBuffer::create_uninitialized(4096));
171 while (!gzip_stream->is_eof()) {
172 auto const data = TRY(gzip_stream->read_some(buffer));
173 TRY(output_stream.write_until_depleted(data));
174 }
175
176 auto output_buffer = TRY(ByteBuffer::create_uninitialized(output_stream.used_buffer_size()));
177 TRY(output_stream.read_until_filled(output_buffer));
178 return output_buffer;
179}
180
181bool GzipDecompressor::is_eof() const { return m_input_stream->is_eof(); }
182
183ErrorOr<size_t> GzipDecompressor::write_some(ReadonlyBytes)
184{
185 return Error::from_errno(EBADF);
186}
187
188GzipCompressor::GzipCompressor(MaybeOwned<Stream> stream)
189 : m_output_stream(move(stream))
190{
191}
192
193ErrorOr<Bytes> GzipCompressor::read_some(Bytes)
194{
195 return Error::from_errno(EBADF);
196}
197
198ErrorOr<size_t> GzipCompressor::write_some(ReadonlyBytes bytes)
199{
200 BlockHeader header;
201 header.identification_1 = 0x1f;
202 header.identification_2 = 0x8b;
203 header.compression_method = 0x08;
204 header.flags = 0;
205 header.modification_time = 0;
206 header.extra_flags = 3; // DEFLATE sets 2 for maximum compression and 4 for minimum compression
207 header.operating_system = 3; // unix
208 TRY(m_output_stream->write_until_depleted({ &header, sizeof(header) }));
209 auto compressed_stream = TRY(DeflateCompressor::construct(MaybeOwned(*m_output_stream)));
210 TRY(compressed_stream->write_until_depleted(bytes));
211 TRY(compressed_stream->final_flush());
212 Crypto::Checksum::CRC32 crc32;
213 crc32.update(bytes);
214 LittleEndian<u32> digest = crc32.digest();
215 LittleEndian<u32> size = bytes.size();
216 TRY(m_output_stream->write_until_depleted(digest.bytes()));
217 TRY(m_output_stream->write_until_depleted(size.bytes()));
218 return bytes.size();
219}
220
221bool GzipCompressor::is_eof() const
222{
223 return true;
224}
225
226bool GzipCompressor::is_open() const
227{
228 return m_output_stream->is_open();
229}
230
231void GzipCompressor::close()
232{
233}
234
235ErrorOr<ByteBuffer> GzipCompressor::compress_all(ReadonlyBytes bytes)
236{
237 auto output_stream = TRY(try_make<AllocatingMemoryStream>());
238 GzipCompressor gzip_stream { MaybeOwned<Stream>(*output_stream) };
239
240 TRY(gzip_stream.write_until_depleted(bytes));
241
242 auto buffer = TRY(ByteBuffer::create_uninitialized(output_stream->used_buffer_size()));
243 TRY(output_stream->read_until_filled(buffer.bytes()));
244 return buffer;
245}
246
247}