diff options
author | Matthew Olsson <matthewcolsson@gmail.com> | 2021-05-24 13:57:16 -0700 |
---|---|---|
committer | Ali Mohammad Pur <Ali.mpfard@gmail.com> | 2021-06-12 22:45:01 +0430 |
commit | 1ef5071d1b46caf76a1156a5bbef82e253df6474 (patch) | |
tree | 26e3e34c19a35ba3f5315bc94c4441a741603f45 /Userland/Libraries/LibPDF | |
parent | d654fe0e41071bd769b920c08b30f302951eb638 (diff) | |
download | serenity-1ef5071d1b46caf76a1156a5bbef82e253df6474.zip |
LibPDF: Harden the document/parser against errors
Diffstat (limited to 'Userland/Libraries/LibPDF')
-rw-r--r-- | Userland/Libraries/LibPDF/Document.cpp | 67 | ||||
-rw-r--r-- | Userland/Libraries/LibPDF/Document.h | 10 | ||||
-rw-r--r-- | Userland/Libraries/LibPDF/Parser.cpp | 203 | ||||
-rw-r--r-- | Userland/Libraries/LibPDF/Parser.h | 31 | ||||
-rw-r--r-- | Userland/Libraries/LibPDF/Value.h | 10 |
5 files changed, 214 insertions, 107 deletions
diff --git a/Userland/Libraries/LibPDF/Document.cpp b/Userland/Libraries/LibPDF/Document.cpp index 1241a68865..deeb3ca46c 100644 --- a/Userland/Libraries/LibPDF/Document.cpp +++ b/Userland/Libraries/LibPDF/Document.cpp @@ -34,20 +34,32 @@ String OutlineItem::to_string(int indent) const return builder.to_string(); } -Document::Document(const ReadonlyBytes& bytes) - : m_parser(Parser({}, bytes)) +RefPtr<Document> Document::create(const ReadonlyBytes& bytes) { - m_parser.set_document(this); + auto parser = adopt_ref(*new Parser({}, bytes)); + auto document = adopt_ref(*new Document(parser)); - VERIFY(m_parser.perform_validation()); - auto [xref_table, trailer] = m_parser.parse_last_xref_table_and_trailer(); + VERIFY(parser->perform_validation()); + auto xref_table_and_trailer_opt = parser->parse_last_xref_table_and_trailer(); + if (!xref_table_and_trailer_opt.has_value()) + return {}; - m_xref_table = xref_table; - m_trailer = trailer; + auto [xref_table, trailer] = xref_table_and_trailer_opt.value(); - m_catalog = m_trailer->get_dict(this, CommonNames::Root); - build_page_tree(); - build_outline(); + document->m_xref_table = xref_table; + document->m_trailer = trailer; + + document->m_catalog = document->m_trailer->get_dict(document, CommonNames::Root); + document->build_page_tree(); + document->build_outline(); + + return document; +} + +Document::Document(const NonnullRefPtr<Parser>& parser) + : m_parser(parser) +{ + m_parser->set_document(this); } Value Document::get_or_load_value(u32 index) @@ -58,7 +70,7 @@ Value Document::get_or_load_value(u32 index) VERIFY(m_xref_table.has_object(index)); auto byte_offset = m_xref_table.byte_offset_for_object(index); - auto indirect_value = m_parser.parse_indirect_value_at_offset(byte_offset); + auto indirect_value = m_parser->parse_indirect_value_at_offset(byte_offset); VERIFY(indirect_value->index() == index); value = indirect_value->value(); m_values.set(index, value); @@ -145,14 +157,19 @@ Value Document::resolve(const Value& value) return obj; } -void Document::build_page_tree() +bool Document::build_page_tree() { + if (!m_catalog->contains(CommonNames::Pages)) + return false; auto page_tree = m_catalog->get_dict(this, CommonNames::Pages); - add_page_tree_node_to_page_tree(page_tree); + return add_page_tree_node_to_page_tree(page_tree); } -void Document::add_page_tree_node_to_page_tree(NonnullRefPtr<DictObject> page_tree) +bool Document::add_page_tree_node_to_page_tree(NonnullRefPtr<DictObject> page_tree) { + if (!page_tree->contains(CommonNames::Kids) || !page_tree->contains(CommonNames::Count)) + return false; + auto kids_array = page_tree->get_array(this, CommonNames::Kids); auto page_count = page_tree->get(CommonNames::Count).value().as_int(); @@ -163,20 +180,24 @@ void Document::add_page_tree_node_to_page_tree(NonnullRefPtr<DictObject> page_tr for (auto& value : *kids_array) { auto reference_index = value.as_ref_index(); auto byte_offset = m_xref_table.byte_offset_for_object(reference_index); - auto maybe_page_tree_node = m_parser.conditionally_parse_page_tree_node_at_offset(byte_offset); + bool ok; + auto maybe_page_tree_node = m_parser->conditionally_parse_page_tree_node_at_offset(byte_offset, ok); + if (!ok) + return false; if (maybe_page_tree_node) { - add_page_tree_node_to_page_tree(maybe_page_tree_node.release_nonnull()); + if (!add_page_tree_node_to_page_tree(maybe_page_tree_node.release_nonnull())) + return false; } else { m_page_object_indices.append(reference_index); } } - - return; + } else { + // We know all of the kids are leaf nodes + for (auto& value : *kids_array) + m_page_object_indices.append(value.as_ref_index()); } - // We know all of the kids are leaf nodes - for (auto& value : *kids_array) - m_page_object_indices.append(value.as_ref_index()); + return true; } void Document::build_outline() @@ -187,8 +208,8 @@ void Document::build_outline() auto outline_dict = m_catalog->get_dict(this, CommonNames::Outlines); if (!outline_dict->contains(CommonNames::First)) return; - - VERIFY(outline_dict->contains(CommonNames::Last)); + if (!outline_dict->contains(CommonNames::Last)) + return; auto first_ref = outline_dict->get_value(CommonNames::First); auto last_ref = outline_dict->get_value(CommonNames::Last); diff --git a/Userland/Libraries/LibPDF/Document.h b/Userland/Libraries/LibPDF/Document.h index a52c1eda47..655b0450eb 100644 --- a/Userland/Libraries/LibPDF/Document.h +++ b/Userland/Libraries/LibPDF/Document.h @@ -73,7 +73,7 @@ struct OutlineDict final : public RefCounted<OutlineDict> { class Document final : public RefCounted<Document> { public: - explicit Document(const ReadonlyBytes& bytes); + static RefPtr<Document> create(const ReadonlyBytes& bytes); ALWAYS_INLINE const XRefTable& xref_table() const { return m_xref_table; } ALWAYS_INLINE const DictObject& trailer() const { return *m_trailer; } @@ -123,20 +123,22 @@ public: } private: + explicit Document(const NonnullRefPtr<Parser>& parser); + // FIXME: Currently, to improve performance, we don't load any pages at Document // construction, rather we just load the page structure and populate // m_page_object_indices. However, we can be even lazier and defer page tree node // parsing, as good PDF writers will layout the page tree in a balanced tree to // improve lookup time. This would reduce the initial overhead by not loading // every page tree node of, say, a 1000+ page PDF file. - void build_page_tree(); - void add_page_tree_node_to_page_tree(NonnullRefPtr<DictObject> page_tree); + bool build_page_tree(); + bool add_page_tree_node_to_page_tree(NonnullRefPtr<DictObject> page_tree); void build_outline(); NonnullRefPtr<OutlineItem> build_outline_item(NonnullRefPtr<DictObject> outline_item_dict); NonnullRefPtrVector<OutlineItem> build_outline_item_chain(const Value& first_ref, const Value& last_ref); - Parser m_parser; + NonnullRefPtr<Parser> m_parser; XRefTable m_xref_table; RefPtr<DictObject> m_trailer; RefPtr<DictObject> m_catalog; diff --git a/Userland/Libraries/LibPDF/Parser.cpp b/Userland/Libraries/LibPDF/Parser.cpp index 7ec53eb15c..f868629525 100644 --- a/Userland/Libraries/LibPDF/Parser.cpp +++ b/Userland/Libraries/LibPDF/Parser.cpp @@ -24,8 +24,8 @@ static NonnullRefPtr<T> make_object(Args... args) requires(IsBaseOf<Object, T>) Vector<Command> Parser::parse_graphics_commands(const ReadonlyBytes& bytes) { - Parser parser(bytes); - return parser.parse_graphics_commands(); + auto parser = adopt_ref(*new Parser(bytes)); + return parser->parse_graphics_commands(); } Parser::Parser(Badge<Document>, const ReadonlyBytes& bytes) @@ -43,26 +43,34 @@ bool Parser::perform_validation() return !sloppy_is_linearized() && parse_header(); } -Parser::XRefTableAndTrailer Parser::parse_last_xref_table_and_trailer() +Optional<Parser::XRefTableAndTrailer> Parser::parse_last_xref_table_and_trailer() { m_reader.move_to(m_reader.bytes().size() - 1); - VERIFY(navigate_to_before_eof_marker()); - navigate_to_after_startxref(); - VERIFY(!m_reader.done()); + if (!navigate_to_before_eof_marker()) + return {}; + if (!navigate_to_after_startxref()) + return {}; + if (m_reader.done()) + return {}; m_reader.set_reading_forwards(); auto xref_offset_value = parse_number(); - VERIFY(xref_offset_value.is_int()); + if (!xref_offset_value.is_int()) + return {}; auto xref_offset = xref_offset_value.as_int(); m_reader.move_to(xref_offset); auto xref_table = parse_xref_table(); + if (!xref_table.has_value()) + return {}; auto trailer = parse_file_trailer(); + if (!trailer) + return {}; - return { xref_table, trailer }; + return XRefTableAndTrailer { xref_table.value(), trailer.release_nonnull() }; } -NonnullRefPtr<IndirectValue> Parser::parse_indirect_value_at_offset(size_t offset) +RefPtr<IndirectValue> Parser::parse_indirect_value_at_offset(size_t offset) { m_reader.set_reading_forwards(); m_reader.move_to(offset); @@ -103,11 +111,13 @@ bool Parser::parse_header() return true; } -XRefTable Parser::parse_xref_table() +Optional<XRefTable> Parser::parse_xref_table() { - VERIFY(m_reader.matches("xref")); + if (!m_reader.matches("xref")) + return {}; m_reader.move_by(4); - consume_eol(); + if (!consume_eol()) + return {}; XRefTable table; @@ -125,23 +135,28 @@ XRefTable Parser::parse_xref_table() for (int i = 0; i < object_count; i++) { auto offset_string = String(m_reader.bytes().slice(m_reader.offset(), 10)); m_reader.move_by(10); - consume(' '); + if (!consume(' ')) + return {}; auto generation_string = String(m_reader.bytes().slice(m_reader.offset(), 5)); m_reader.move_by(5); - consume(' '); + if (!consume(' ')) + return {}; auto letter = m_reader.read(); - VERIFY(letter == 'n' || letter == 'f'); + if (letter != 'n' && letter != 'f') + return {}; // The line ending sequence can be one of the following: // SP CR, SP LF, or CR LF if (m_reader.matches(' ')) { consume(); auto ch = consume(); - VERIFY(ch == '\r' || ch == '\n'); + if (ch != '\r' && ch != '\n') + return {}; } else { - VERIFY(m_reader.matches("\r\n")); + if (!m_reader.matches("\r\n")) + return {}; m_reader.move_by(2); } @@ -157,23 +172,27 @@ XRefTable Parser::parse_xref_table() return table; } -NonnullRefPtr<DictObject> Parser::parse_file_trailer() +RefPtr<DictObject> Parser::parse_file_trailer() { - VERIFY(m_reader.matches("trailer")); + if (!m_reader.matches("trailer")) + return {}; m_reader.move_by(7); consume_whitespace(); auto dict = parse_dict(); + if (!dict) + return {}; - VERIFY(m_reader.matches("startxref")); + if (!m_reader.matches("startxref")) + return {}; m_reader.move_by(9); consume_whitespace(); m_reader.move_until([&](auto) { return matches_eol(); }); - consume_eol(); - VERIFY(m_reader.matches("%%EOF")); + VERIFY(consume_eol()); + if (!m_reader.matches("%%EOF")) + return {}; m_reader.move_by(5); consume_whitespace(); - VERIFY(m_reader.done()); return dict; } @@ -291,8 +310,10 @@ Value Parser::parse_value() if (m_reader.matches("<<")) { auto dict = parse_dict(); + if (!dict) + return {}; if (m_reader.matches("stream\n")) - return parse_stream(dict); + return parse_stream(dict.release_nonnull()); return dict; } @@ -335,23 +356,28 @@ Value Parser::parse_possible_indirect_value_or_ref() return first_number; } -NonnullRefPtr<IndirectValue> Parser::parse_indirect_value(int index, int generation) +RefPtr<IndirectValue> Parser::parse_indirect_value(int index, int generation) { - VERIFY(m_reader.matches("obj")); + if (!m_reader.matches("obj")) + return {}; m_reader.move_by(3); if (matches_eol()) consume_eol(); auto value = parse_value(); - VERIFY(m_reader.matches("endobj")); + if (!m_reader.matches("endobj")) + return {}; return make_object<IndirectValue>(index, generation, value); } -NonnullRefPtr<IndirectValue> Parser::parse_indirect_value() +RefPtr<IndirectValue> Parser::parse_indirect_value() { auto first_number = parse_number(); + if (!first_number.is_int()) + return {}; auto second_number = parse_number(); - VERIFY(first_number.is_int() && second_number.is_int()); + if (!second_number.is_int()) + return {}; return parse_indirect_value(first_number.as_int(), second_number.as_int()); } @@ -387,9 +413,10 @@ Value Parser::parse_number() return Value(static_cast<int>(f)); } -NonnullRefPtr<NameObject> Parser::parse_name() +RefPtr<NameObject> Parser::parse_name() { - consume('/'); + if (!consume('/')) + return {}; StringBuilder builder; while (true) { @@ -400,7 +427,8 @@ NonnullRefPtr<NameObject> Parser::parse_name() int hex_value = 0; for (int i = 0; i < 2; i++) { auto ch = consume(); - VERIFY(isxdigit(ch)); + if (!isxdigit(ch)) + return {}; hex_value *= 16; if (ch <= '9') { hex_value += ch - '0'; @@ -420,7 +448,7 @@ NonnullRefPtr<NameObject> Parser::parse_name() return make_object<NameObject>(builder.to_string()); } -NonnullRefPtr<StringObject> Parser::parse_string() +RefPtr<StringObject> Parser::parse_string() { ScopeGuard guard([&] { consume_whitespace(); }); @@ -435,6 +463,9 @@ NonnullRefPtr<StringObject> Parser::parse_string() is_binary_string = true; } + if (string.is_null()) + return {}; + if (string.bytes().starts_with(Array<u8, 2> { 0xfe, 0xff })) { // The string is encoded in UTF16-BE string = TextCodec::decoder_for("utf-16be")->to_utf8(string.substring(2)); @@ -449,7 +480,8 @@ NonnullRefPtr<StringObject> Parser::parse_string() String Parser::parse_literal_string() { - consume('('); + if (!consume('(')) + return {}; StringBuilder builder; auto opened_parens = 0; @@ -470,7 +502,9 @@ String Parser::parse_literal_string() continue; } - VERIFY(!m_reader.done()); + if (m_reader.done()) + return {}; + auto ch = consume(); switch (ch) { case 'n': @@ -520,13 +554,16 @@ String Parser::parse_literal_string() } } - VERIFY(opened_parens == 0); + if (opened_parens != 0) + return {}; + return builder.to_string(); } String Parser::parse_hex_string() { - consume('<'); + if (!consume('<')) + return {}; StringBuilder builder; while (true) { @@ -546,7 +583,9 @@ String Parser::parse_hex_string() builder.append(static_cast<char>(hex_value)); return builder.to_string(); } - VERIFY(isxdigit(ch)); + + if (!isxdigit(ch)) + return {}; hex_value *= 16; if (ch <= '9') { @@ -561,25 +600,31 @@ String Parser::parse_hex_string() } } -NonnullRefPtr<ArrayObject> Parser::parse_array() +RefPtr<ArrayObject> Parser::parse_array() { - consume('['); + if (!consume('[')) + return {}; consume_whitespace(); Vector<Value> values; - while (!m_reader.matches(']')) - values.append(parse_value()); + while (!m_reader.matches(']')) { + auto value = parse_value(); + if (!value) + return {}; + values.append(value); + } - consume(']'); + if (!consume(']')) + return {}; consume_whitespace(); return make_object<ArrayObject>(values); } -NonnullRefPtr<DictObject> Parser::parse_dict() +RefPtr<DictObject> Parser::parse_dict() { - consume('<'); - consume('<'); + if (!consume('<') || !consume('<')) + return {}; consume_whitespace(); HashMap<FlyString, Value> map; @@ -587,28 +632,38 @@ NonnullRefPtr<DictObject> Parser::parse_dict() if (m_reader.matches(">>")) break; auto name = parse_name(); + if (!name) + return {}; auto value = parse_value(); + if (!value) + return {}; map.set(name->name(), value); } - consume('>'); - consume('>'); + if (!consume('>') || !consume('>')) + return {}; consume_whitespace(); return make_object<DictObject>(map); } -RefPtr<DictObject> Parser::conditionally_parse_page_tree_node_at_offset(size_t offset) +RefPtr<DictObject> Parser::conditionally_parse_page_tree_node_at_offset(size_t offset, bool& ok) { + ok = true; + m_reader.move_to(offset); parse_number(); parse_number(); - VERIFY(m_reader.matches("obj")); + if (!m_reader.matches("obj")) { + ok = false; + return {}; + } + m_reader.move_by(3); consume_whitespace(); - consume('<'); - consume('<'); + if (!consume('<') || !consume('<')) + return {}; consume_whitespace(); HashMap<FlyString, Value> map; @@ -616,12 +671,21 @@ RefPtr<DictObject> Parser::conditionally_parse_page_tree_node_at_offset(size_t o if (m_reader.matches(">>")) break; auto name = parse_name(); + if (!name) { + ok = false; + return {}; + } + auto name_string = name->name(); if (!name_string.is_one_of(CommonNames::Type, CommonNames::Parent, CommonNames::Kids, CommonNames::Count)) { // This is a page, not a page tree node return {}; } auto value = parse_value(); + if (!value) { + ok = false; + return {}; + } if (name_string == CommonNames::Type) { if (!value.is_object()) return {}; @@ -635,18 +699,20 @@ RefPtr<DictObject> Parser::conditionally_parse_page_tree_node_at_offset(size_t o map.set(name->name(), value); } - consume('>'); - consume('>'); + if (!consume('>') || !consume('>')) + return {}; consume_whitespace(); return make_object<DictObject>(map); } -NonnullRefPtr<StreamObject> Parser::parse_stream(NonnullRefPtr<DictObject> dict) +RefPtr<StreamObject> Parser::parse_stream(NonnullRefPtr<DictObject> dict) { - VERIFY(m_reader.matches("stream")); + if (!m_reader.matches("stream")) + return {}; m_reader.move_by(6); - consume_eol(); + if (!consume_eol()) + return {}; ReadonlyBytes bytes; @@ -681,8 +747,8 @@ NonnullRefPtr<StreamObject> Parser::parse_stream(NonnullRefPtr<DictObject> dict) if (dict->contains(CommonNames::Filter)) { auto filter_type = dict->get_name(m_document, CommonNames::Filter)->name(); auto maybe_bytes = Filter::decode(bytes, filter_type); - // FIXME: Handle error condition - VERIFY(maybe_bytes.has_value()); + if (!maybe_bytes.has_value()) + return {}; return make_object<EncodedStreamObject>(dict, move(maybe_bytes.value())); } @@ -752,14 +818,15 @@ bool Parser::matches_regular_character() const return !matches_delimiter() && !matches_whitespace(); } -void Parser::consume_eol() +bool Parser::consume_eol() { if (m_reader.matches("\r\n")) { consume(2); - } else { - auto consumed = consume(); - VERIFY(consumed == 0xd || consumed == 0xa); + return true; } + + auto consumed = consume(); + return consumed == 0xd || consumed == 0xa; } bool Parser::consume_whitespace() @@ -777,9 +844,15 @@ char Parser::consume() return m_reader.read(); } -void Parser::consume(char ch) +void Parser::consume(int amount) +{ + for (size_t i = 0; i < static_cast<size_t>(amount); i++) + consume(); +} + +bool Parser::consume(char ch) { - VERIFY(consume() == ch); + return consume() == ch; } } diff --git a/Userland/Libraries/LibPDF/Parser.h b/Userland/Libraries/LibPDF/Parser.h index 216c9e6da8..6774acb93b 100644 --- a/Userland/Libraries/LibPDF/Parser.h +++ b/Userland/Libraries/LibPDF/Parser.h @@ -16,7 +16,7 @@ namespace PDF { class Document; -class Parser { +class Parser final : public RefCounted<Parser> { public: static Vector<Command> parse_graphics_commands(const ReadonlyBytes&); @@ -30,18 +30,18 @@ public: XRefTable xref_table; NonnullRefPtr<DictObject> trailer; }; - XRefTableAndTrailer parse_last_xref_table_and_trailer(); + Optional<XRefTableAndTrailer> parse_last_xref_table_and_trailer(); - NonnullRefPtr<IndirectValue> parse_indirect_value_at_offset(size_t offset); + RefPtr<IndirectValue> parse_indirect_value_at_offset(size_t offset); - RefPtr<DictObject> conditionally_parse_page_tree_node_at_offset(size_t offset); + RefPtr<DictObject> conditionally_parse_page_tree_node_at_offset(size_t offset, bool& ok); private: explicit Parser(const ReadonlyBytes&); bool parse_header(); - XRefTable parse_xref_table(); - NonnullRefPtr<DictObject> parse_file_trailer(); + Optional<XRefTable> parse_xref_table(); + RefPtr<DictObject> parse_file_trailer(); bool navigate_to_before_eof_marker(); bool navigate_to_after_startxref(); @@ -58,16 +58,16 @@ private: Value parse_value(); Value parse_possible_indirect_value_or_ref(); - NonnullRefPtr<IndirectValue> parse_indirect_value(int index, int generation); - NonnullRefPtr<IndirectValue> parse_indirect_value(); + RefPtr<IndirectValue> parse_indirect_value(int index, int generation); + RefPtr<IndirectValue> parse_indirect_value(); Value parse_number(); - NonnullRefPtr<NameObject> parse_name(); - NonnullRefPtr<StringObject> parse_string(); + RefPtr<NameObject> parse_name(); + RefPtr<StringObject> parse_string(); String parse_literal_string(); String parse_hex_string(); - NonnullRefPtr<ArrayObject> parse_array(); - NonnullRefPtr<DictObject> parse_dict(); - NonnullRefPtr<StreamObject> parse_stream(NonnullRefPtr<DictObject> dict); + RefPtr<ArrayObject> parse_array(); + RefPtr<DictObject> parse_dict(); + RefPtr<StreamObject> parse_stream(NonnullRefPtr<DictObject> dict); Vector<Command> parse_graphics_commands(); @@ -77,10 +77,11 @@ private: bool matches_delimiter() const; bool matches_regular_character() const; - void consume_eol(); + bool consume_eol(); bool consume_whitespace(); char consume(); - void consume(char); + void consume(int amount); + bool consume(char); Reader m_reader; RefPtr<Document> m_document; diff --git a/Userland/Libraries/LibPDF/Value.h b/Userland/Libraries/LibPDF/Value.h index f6b3b61899..432be4750e 100644 --- a/Userland/Libraries/LibPDF/Value.h +++ b/Userland/Libraries/LibPDF/Value.h @@ -62,6 +62,16 @@ public: } template<IsObject T> + Value(RefPtr<T> obj) + : m_type(obj ? Type::Object : Type::Empty) + { + if (obj) { + obj->ref(); + m_as_object = obj; + } + } + + template<IsObject T> Value(NonnullRefPtr<T> obj) : m_type(Type::Object) { |