2 * Copyright (C) 1996-2021 The Squid Software Foundation and contributors
4 * Squid software is distributed under GPLv2+ license and includes
5 * contributions from numerous individuals and organizations.
6 * Please see the COPYING and CONTRIBUTORS files for details.
9 #ifndef _SQUID_SRC_HTTP_ONE_PARSER_H
10 #define _SQUID_SRC_HTTP_ONE_PARSER_H
12 #include "anyp/ProtocolVersion.h"
13 #include "http/one/forward.h"
14 #include "http/StatusCode.h"
15 #include "parser/forward.h"
16 #include "sbuf/SBuf.h"
23 HTTP_PARSE_NONE
, ///< initialized, but nothing usefully parsed yet
24 HTTP_PARSE_FIRST
, ///< HTTP/1 message first-line
25 HTTP_PARSE_CHUNK_SZ
, ///< HTTP/1.1 chunked encoding chunk-size
26 HTTP_PARSE_CHUNK_EXT
, ///< HTTP/1.1 chunked encoding chunk-ext
27 HTTP_PARSE_CHUNK
, ///< HTTP/1.1 chunked encoding chunk-data
28 HTTP_PARSE_MIME
, ///< HTTP/1 mime-header block
29 HTTP_PARSE_DONE
///< parsed a message header, or reached a terminal syntax error
32 /** HTTP/1.x protocol parser
34 * Works on a raw character I/O buffer and tokenizes the content into
35 * the major CRLF delimited segments of an HTTP/1 procotol message:
37 * \li first-line (request-line / simple-request / status-line)
38 * \li mime-header 0*( header-name ':' SP field-value CRLF)
40 class Parser
: public RefCountable
43 typedef SBuf::size_type size_type
;
44 typedef ::Parser::Tokenizer Tokenizer
;
47 Parser(const Parser
&) = default;
48 Parser
&operator =(const Parser
&) = default;
49 Parser(Parser
&&) = default;
50 Parser
&operator =(Parser
&&) = default;
53 /// Set this parser back to a default state.
54 /// Will DROP any reference to a buffer (does not free).
55 virtual void clear() = 0;
57 /// attempt to parse a message from the buffer
58 /// \retval true if a full message was found and parsed
59 /// \retval false if incomplete, invalid or no message was found
60 virtual bool parse(const SBuf
&aBuf
) = 0;
62 /** Whether the parser is waiting on more data to complete parsing a message.
63 * Use to distinguish between incomplete data and error results
64 * when parse() returns false.
66 bool needsMoreData() const {return parsingStage_
!=HTTP_PARSE_DONE
;}
68 /// size in bytes of the first line including CRLF terminator
69 virtual size_type
firstLineSize() const = 0;
71 /// size in bytes of the message headers including CRLF terminator(s)
72 /// but excluding first-line bytes
73 size_type
headerBlockSize() const {return mimeHeaderBlock_
.length();}
75 /// size in bytes of HTTP message block, includes first-line and mime headers
76 /// excludes any body/entity/payload bytes
77 /// excludes any garbage prefix before the first-line
78 size_type
messageHeaderSize() const {return firstLineSize() + headerBlockSize();}
80 /// buffer containing HTTP mime headers, excluding message first-line.
81 SBuf
mimeHeader() const {return mimeHeaderBlock_
;}
83 /// the protocol label for this message
84 const AnyP::ProtocolVersion
& messageProtocol() const {return msgProtocol_
;}
87 * Scan the mime header block (badly) for a Host header.
89 * BUG: omits lines when searching for headers with obs-fold or multiple entries.
91 * BUG: limits output to just 1KB when Squid accepts up to 64KB line length.
93 * \return A pointer to a field-value of the first matching field-name, or NULL.
95 char *getHostHeaderField();
97 /// the remaining unprocessed section of buffer
98 const SBuf
&remaining() const {return buf_
;}
101 * HTTP status code resulting from the parse process.
102 * to be used on the invalid message handling.
104 * Http::scNone indicates incomplete parse,
105 * Http::scOkay indicates no error,
106 * other codes represent a parse error.
108 Http::StatusCode parseStatusCode
= Http::scNone
;
110 /// Whitespace between regular protocol elements.
111 /// Seen in RFCs as OWS, RWS, BWS, SP/HTAB but may be "relaxed" by us.
112 /// See also: DelimiterCharacters().
113 static const CharacterSet
&WhitespaceCharacters();
115 /// Whitespace between protocol elements in restricted contexts like
116 /// request line, status line, asctime-date, and credentials
117 /// Seen in RFCs as SP but may be "relaxed" by us.
118 /// See also: WhitespaceCharacters().
119 /// XXX: Misnamed and overused.
120 static const CharacterSet
&DelimiterCharacters();
124 * detect and skip the CRLF or (if tolerant) LF line terminator
125 * consume from the tokenizer.
127 * \throws exception on bad or InsuffientInput.
128 * \retval true only if line terminator found.
129 * \retval false incomplete or missing line terminator, need more data.
131 void skipLineTerminator(Tokenizer
&) const;
134 * Scan to find the mime headers block for current message.
136 * \retval true If mime block (or a blocks non-existence) has been
137 * identified accurately within limit characters.
138 * mimeHeaderBlock_ has been updated and buf_ consumed.
140 * \retval false An error occurred, or no mime terminator found within limit.
142 bool grabMimeBlock(const char *which
, const size_t limit
);
144 /// RFC 7230 section 2.6 - 7 magic octets
145 static const SBuf Http1magic
;
147 /// bytes remaining to be parsed
150 /// what stage the parser is currently up to
151 ParseState parsingStage_
= HTTP_PARSE_NONE
;
153 /// what protocol label has been found in the first line (if any)
154 AnyP::ProtocolVersion msgProtocol_
;
156 /// buffer holding the mime headers (if any)
157 SBuf mimeHeaderBlock_
;
159 /// Whether the invalid HTTP as HTTP/0.9 hack expects a mime header block
160 bool hackExpectsMime_
= false;
163 void cleanMimePrefix();
167 /// skips and, if needed, warns about RFC 7230 BWS ("bad" whitespace)
168 /// \throws InsufficientInput when the end of BWS cannot be confirmed
169 void ParseBws(Parser::Tokenizer
&);
171 /// the right debugs() level for logging HTTP violation messages
177 #endif /* _SQUID_SRC_HTTP_ONE_PARSER_H */