Class: Unicorn::HttpParser
Relationships & Source Files | |
Inherits: | Object |
Defined in: | lib/unicorn/http_request.rb, ext/unicorn_http/unicorn_http.rl |
Constant Summary
-
CHUNK_MAX =
The maximum size a single chunk when using chunked transfer encoding. This is only a theoretical maximum used to detect errors in clients, it is highly unlikely to encounter clients that send more than several kilobytes at once.
OFFT2NUM(UH_OFF_T_MAX)
-
DEFAULTS =
Internal use only
default parameters we merge into the request env for Rack handlers
{ "rack.errors" => $stderr, "rack.multiprocess" => true, "rack.multithread" => false, "rack.run_once" => false, "rack.version" => [1, 2], "rack.hijack?" => true, "SCRIPT_NAME" => "", # this is not in the Rack spec, but some apps may rely on it "SERVER_SOFTWARE" => "Unicorn #{Unicorn::Const::UNICORN_VERSION}" }
-
EMPTY_ARRAY =
Internal use only
# File 'lib/unicorn/http_request.rb', line 29[].freeze
-
HTTP_RESPONSE_START =
Internal use only
# File 'lib/unicorn/http_request.rb', line 28[ 'HTTP'.freeze, '/1.1 '.freeze ]
-
LENGTH_MAX =
The maximum size of the body as specified by Content-Length. This is only a theoretical maximum, the actual limit is subject to the limits of the file system used for
Dir.tmpdir
.OFFT2NUM(UH_OFF_T_MAX)
-
NULL_IO =
Internal use only
# File 'lib/unicorn/http_request.rb', line 25StringIO.new("")
-
TCPI =
Internal use only
# File 'lib/unicorn/http_request.rb', line 109Raindrops::TCP_Info.allocate
Class Attribute Summary
-
.max_header_len=(len)
writeonly
this is only intended for use with Rainbows!
- .check_client_connection rw Internal use only
- .check_client_connection=(bool) rw Internal use only
- .input_class rw Internal use only
- .input_class=(klass) rw Internal use only
Class Method Summary
-
.new ⇒ parser
constructor
Creates a new parser.
-
.is_chunked?(v) ⇒ Boolean
Internal use only
called by ext/unicorn_http/unicorn_http.rl via rb_funcall.
Instance Attribute Summary
-
#body_eof? ⇒ Boolean
readonly
Detects if we’re done filtering the body or not.
-
#headers? ⇒ Boolean
readonly
This should be used to detect if a request has headers (and if the response will have headers as well).
-
#keepalive? ⇒ Boolean
readonly
This should be used to detect if a request can really handle keepalives and pipelining.
-
#next? ⇒ Boolean
readonly
Exactly like #keepalive?, except it will reset the internal parser state on next parse if it returns true.
-
#response_start_sent
rw
ignored by Ruby anyways.
- #response_start_sent=(boolean) rw
- #chunkable_response? ⇒ Boolean readonly Internal use only
- #hijacked? ⇒ Boolean readonly Internal use only
Instance Method Summary
-
#add_parse(buffer) ⇒ env?
adds the contents of
buffer
to the internal buffer and attempts to continue parsing. - #buf
-
#clear ⇒ parser
Resets the parser to it’s initial state so that you can reuse it rather than making new ones.
-
#content_length ⇒ nil, Integer
Returns the number of bytes left to run through #filter_body.
- #env
-
#filter_body(dst, src) ⇒ nil/src
Takes a String of
src
, will modify data if dechunking is done. - #headers(env, buf) (also: #trailers) readonly
- #hijacked!
-
#parse ⇒ env?
Takes a Hash and a String of data, parses the String of data filling in the Hash returning the Hash if parsing is finished, nil otherwise When returning the env Hash, it may modify data to point to where body processing should begin.
-
#trailers(env, buf)
Alias for #headers.
-
#call
Internal use only
for rack.hijack, we respond to this method so no extra allocation of a proc object.
-
#check_client_connection(socket, ai)
Internal use only
Ruby 2.2+ can show struct tcp_info as a string
Socket::Option#inspect
. -
#closed_state?(state) ⇒ Boolean
Internal use only
raindrops before 0.18 only supported TCP_INFO under Linux.
- #closed_state_str?(state) ⇒ Boolean Internal use only
-
#read_headers(socket, ai)
Internal use only
Does the majority of the IO processing.
- #write_http_header(socket) Internal use only
Constructor Details
.new ⇒ parser
Creates a new parser.
# File 'ext/unicorn_http/unicorn_http.rl', line 638
static VALUE HttpParser_init(VALUE self) { struct http_parser *hp = data_get(self); http_parser_init(hp); hp->buf = rb_str_new(NULL, 0); hp->env = rb_hash_new(); return self; }
Class Attribute Details
.check_client_connection (rw)
# File 'lib/unicorn/http_request.rb', line 42
def self.check_client_connection @@check_client_connection end
.check_client_connection=(bool) (rw)
# File 'lib/unicorn/http_request.rb', line 46
def self.check_client_connection=(bool) @@check_client_connection = bool end
.input_class (rw)
# File 'lib/unicorn/http_request.rb', line 34
def self.input_class @@input_class end
.input_class=(klass) (rw)
# File 'lib/unicorn/http_request.rb', line 38
def self.input_class=(klass) @@input_class = klass end
.max_header_len=(len) (writeonly)
this is only intended for use with Rainbows!
# File 'ext/unicorn_http/unicorn_http.rl', line 43
static VALUE set_maxhdrlen(VALUE self, VALUE len) { return UINT2NUM(MAX_HEADER_LEN = NUM2UINT(len)); }
Class Method Details
.is_chunked?(v) ⇒ Boolean
called by ext/unicorn_http/unicorn_http.rl via rb_funcall
# File 'lib/unicorn/http_request.rb', line 193
def self.is_chunked?(v) # :nodoc: vals = v.split(/[ \t]*,[ \t]*/).map!(&:downcase) if vals.pop == 'chunked'.freeze return true unless vals.include?('chunked'.freeze) raise Unicorn::HttpParserError, 'double chunked', [] end return false unless vals.include?('chunked'.freeze) raise Unicorn::HttpParserError, 'chunked not last', [] end
Instance Attribute Details
#body_eof? ⇒ Boolean
(readonly)
Detects if we’re done filtering the body or not. This can be used to detect when to stop calling #filter_body.
# File 'ext/unicorn_http/unicorn_http.rl', line 799
static VALUE HttpParser_body_eof(VALUE self) { struct http_parser *hp = data_get(self); if (HP_FL_TEST(hp, CHUNKED)) return chunked_eof(hp) ? Qtrue : Qfalse; return hp->len.content == 0 ? Qtrue : Qfalse; }
#chunkable_response? ⇒ Boolean
(readonly)
# File 'ext/unicorn_http/unicorn_http.rl', line 828
static VALUE chunkable_response_p(VALUE self) { const struct http_parser *hp = data_get(self); return HP_FL_ALL(hp, RES_CHUNKABLE) ? Qtrue : Qfalse; }
#headers? ⇒ Boolean
(readonly)
This should be used to detect if a request has headers (and if the response will have headers as well). HTTP/0.9 requests should return false, all subsequent HTTP versions will return true
# File 'ext/unicorn_http/unicorn_http.rl', line 861
static VALUE HttpParser_has_headers(VALUE self) { struct http_parser *hp = data_get(self); return HP_FL_TEST(hp, HASHEADER) ? Qtrue : Qfalse; }
#hijacked? ⇒ Boolean
(readonly)
# File 'lib/unicorn/http_request.rb', line 104
def hijacked? env.include?('rack.hijack_io'.freeze) end
#keepalive? ⇒ Boolean
(readonly)
This should be used to detect if a request can really handle keepalives and pipelining. Currently, the rules are:
-
MUST be a GET or HEAD request
-
MUST be HTTP/1.1
or
HTTP/1.0 with “Connection: keep-alive” -
MUST NOT have “Connection: close” set
# File 'ext/unicorn_http/unicorn_http.rl', line 820
static VALUE HttpParser_keepalive(VALUE self) { struct http_parser *hp = data_get(self); return HP_FL_ALL(hp, KEEPALIVE) ? Qtrue : Qfalse; }
#next? ⇒ Boolean
(readonly)
Exactly like #keepalive?, except it will reset the internal parser state on next parse if it returns true.
# File 'ext/unicorn_http/unicorn_http.rl', line 842
static VALUE HttpParser_next(VALUE self) { struct http_parser *hp = data_get(self); if (HP_FL_ALL(hp, KEEPALIVE)) { HP_FL_SET(hp, TO_CLEAR); return Qtrue; } return Qfalse; }
#response_start_sent (rw)
ignored by Ruby anyways
# File 'ext/unicorn_http/unicorn_http.rl', line 977
static VALUE HttpParser_rssget(VALUE self) { struct http_parser *hp = data_get(self); return HP_FL_TEST(hp, RESSTART) ? Qtrue : Qfalse; }
#response_start_sent=(boolean) (rw)
[ GitHub ]# File 'ext/unicorn_http/unicorn_http.rl', line 965
static VALUE HttpParser_rssset(VALUE self, VALUE boolean) { struct http_parser *hp = data_get(self); if (RTEST(boolean)) HP_FL_SET(hp, RESSTART); else HP_FL_UNSET(hp, RESSTART); return boolean; /* ignored by Ruby anyways */ }
Instance Method Details
#add_parse(buffer) ⇒ env?
adds the contents of buffer
to the internal buffer and attempts to continue parsing. Returns the #env Hash on success or nil if more data is needed.
Raises HttpParserError if there are parsing errors.
# File 'ext/unicorn_http/unicorn_http.rl', line 756
static VALUE HttpParser_add_parse(VALUE self, VALUE buffer) { struct http_parser *hp = data_get(self); Check_Type(buffer, T_STRING); rb_str_buf_append(hp->buf, buffer); return HttpParser_parse(self); }
#buf
[ GitHub ]# File 'ext/unicorn_http/unicorn_http.rl', line 868
static VALUE HttpParser_buf(VALUE self) { return data_get(self)->buf; }
#call
for rack.hijack, we respond to this method so no extra allocation of a proc object
#check_client_connection(socket, ai)
Ruby 2.2+ can show struct tcp_info as a string Socket::Option#inspect
. Not that efficient, but probably still better than doing unnecessary work after a client gives up.
See additional method definition at line 111.
# File 'lib/unicorn/http_request.rb', line 156
def check_client_connection(socket, ai) # :nodoc: if ai.ip? # Raindrops::TCP_Info#get!, #state (reads struct tcp_info#tcpi_state) raise Errno::EPIPE, "client closed connection".freeze, EMPTY_ARRAY if closed_state?(TCPI.get!(socket).state) else write_http_header(socket) end end
#clear ⇒ parser
Resets the parser to it’s initial state so that you can reuse it rather than making new ones.
# File 'ext/unicorn_http/unicorn_http.rl', line 656
static VALUE HttpParser_clear(VALUE self) { struct http_parser *hp = data_get(self); /* we can't safely reuse .buf and .env if hijacked */ if (HP_FL_TEST(hp, HIJACK)) return HttpParser_init(self); http_parser_init(hp); rb_hash_clear(hp->env); return self; }
#closed_state?(state) ⇒ Boolean
raindrops before 0.18 only supported TCP_INFO under Linux
# File 'lib/unicorn/http_request.rb', line 140
def closed_state?(state) # :nodoc: case state when 1 # ESTABLISHED false when 8, 6, 7, 9, 11 # CLOSE_WAIT, TIME_WAIT, CLOSE, LAST_ACK, CLOSING true else false end end
#closed_state_str?(state) ⇒ Boolean
# File 'lib/unicorn/http_request.rb', line 172
def closed_state_str?(state) case state when 'ESTABLISHED' false # not a typo, ruby maps TCP_CLOSE (no 'D') to state=CLOSED (w/ 'D') when 'CLOSE_WAIT', 'TIME_WAIT', 'CLOSED', 'LAST_ACK', 'CLOSING' true else false end end
#content_length ⇒ nil
, Integer
Returns the number of bytes left to run through #filter_body. This will initially be the value of the “Content-Length” HTTP header after header parsing is complete and will decrease in value as #filter_body is called for each chunk. This should return zero for requests with no body.
This will return nil on “Transfer-Encoding: chunked” requests.
# File 'ext/unicorn_http/unicorn_http.rl', line 698
static VALUE HttpParser_content_length(VALUE self) { struct http_parser *hp = data_get(self); return HP_FL_TEST(hp, CHUNKED) ? Qnil : OFFT2NUM(hp->len.content); }
#env
[ GitHub ]# File 'ext/unicorn_http/unicorn_http.rl', line 873
static VALUE HttpParser_env(VALUE self) { return data_get(self)->env; }
#filter_body(dst, src) ⇒ nil
/src
Takes a String of src
, will modify data if dechunking is done. Returns nil
if there is more data left to process. Returns src
if body processing is complete. When returning src
, it may modify src
so the start of the string points to where the body ended so that trailer processing can begin.
Raises HttpParserError if there are dechunking errors. Basically this is a glorified memcpy(3) that copies src
into #buf while filtering it through the dechunker.
# File 'ext/unicorn_http/unicorn_http.rl', line 901
static VALUE HttpParser_filter_body(VALUE self, VALUE dst, VALUE src) { struct http_parser *hp = data_get(self); char *srcptr; long srclen; srcptr = RSTRING_PTR(src); srclen = RSTRING_LEN(src); StringValue(dst); if (HP_FL_TEST(hp, CHUNKED)) { if (!chunked_eof(hp)) { rb_str_modify(dst); rb_str_resize(dst, srclen); /* we can never copy more than srclen bytes */ hp->s.dest_offset = 0; hp->cont = dst; hp->buf = src; http_parser_execute(hp, srcptr, srclen); if (hp->cs == http_parser_error) parser_raise(eHttpParserError, "Invalid HTTP format, parsing fails."); assert(hp->s.dest_offset <= hp->offset && "destination buffer overflow"); advance_str(src, hp->offset); rb_str_set_len(dst, hp->s.dest_offset); if (RSTRING_LEN(dst) == 0 && chunked_eof(hp)) { assert(hp->len.chunk == 0 && "chunk at EOF but more to parse"); } else { src = Qnil; } } } else { /* no need to enter the Ragel machine for unchunked transfers */ assert(hp->len.content >= 0 && "negative Content-Length"); if (hp->len.content > 0) { long nr = MIN(srclen, hp->len.content); rb_str_modify(dst); rb_str_resize(dst, nr); /* * using rb_str_replace() to avoid memcpy() doesn't help in * most cases because a GC-aware programmer will pass an explicit * buffer to env["rack.input"].read and reuse the buffer in a loop. * This causes copy-on-write behavior to be triggered anyways * when the src buffer is modified (when reading off the socket). */ hp->buf = src; memcpy(RSTRING_PTR(dst), srcptr, nr); hp->len.content -= nr; if (hp->len.content == 0) { HP_FL_SET(hp, REQEOF); hp->cs = http_parser_first_final; } advance_str(src, nr); src = Qnil; } } hp->offset = 0; /* for trailer parsing */ return src; }
#headers(env, buf) (readonly) Also known as: #trailers
[ GitHub ]# File 'ext/unicorn_http/unicorn_http.rl', line 777
static VALUE HttpParser_headers(VALUE self, VALUE env, VALUE buf) { struct http_parser *hp = data_get(self); hp->env = env; hp->buf = buf; return HttpParser_parse(self); }
#hijacked!
[ GitHub ]# File 'ext/unicorn_http/unicorn_http.rl', line 878
static VALUE HttpParser_hijacked_bang(VALUE self) { struct http_parser *hp = data_get(self); HP_FL_SET(hp, HIJACK); return self; }
#parse ⇒ env?
Takes a Hash and a String of data, parses the String of data filling in the Hash returning the Hash if parsing is finished, nil otherwise When returning the env Hash, it may modify data to point to where body processing should begin.
Raises HttpParserError if there are parsing errors.
# File 'ext/unicorn_http/unicorn_http.rl', line 717
static VALUE HttpParser_parse(VALUE self) { struct http_parser *hp = data_get(self); VALUE data = hp->buf; if (HP_FL_TEST(hp, TO_CLEAR)) HttpParser_clear(self); http_parser_execute(hp, RSTRING_PTR(data), RSTRING_LEN(data)); if (hp->offset > MAX_HEADER_LEN) parser_raise(e413, "HTTP header is too large"); if (hp->cs == http_parser_first_final || hp->cs == http_parser_en_ChunkedBody) { advance_str(data, hp->offset + 1); hp->offset = 0; if (HP_FL_TEST(hp, INTRAILER)) HP_FL_SET(hp, REQEOF); return hp->env; } if (hp->cs == http_parser_error) parser_raise(eHttpParserError, "Invalid HTTP format, parsing fails."); return Qnil; }
#read_headers(socket, ai)
Does the majority of the IO processing. It has been written in Ruby using about 8 different IO processing strategies.
It is currently carefully constructed to make sure that it gets the best possible performance for the common case: GET requests that are fully complete after a single read(2)
Anyone who thinks they can make it faster is more than welcome to take a crack at it.
returns an environment hash suitable for Rack if successful This does minimal exception trapping and it is up to the caller to handle any socket errors (e.g. user aborted upload).
# File 'lib/unicorn/http_request.rb', line 65
def read_headers(socket, ai) e = env # From https://www.ietf.org/rfc/rfc3875: # "Script authors should be aware that the REMOTE_ADDR and # REMOTE_HOST meta-variables (see sections 4.1.8 and 4.1.9) # may not identify the ultimate source of the request. They # identify the client for the immediate request to the server; # that client may be a proxy, gateway, or other intermediary # acting on behalf of the actual source client." e['REMOTE_ADDR'] = ai.unix? ? '127.0.0.1' : ai.ip_address # short circuit the common case with small GET requests first socket.readpartial(16384, buf) if parse.nil? # Parser is not done, queue up more data to read and continue parsing # an Exception thrown from the parser will throw us out of the loop false until add_parse(socket.readpartial(16384)) end check_client_connection(socket, ai) if @@check_client_connection e['rack.input'] = 0 == content_length ? NULL_IO : @@input_class.new(socket, self) # for Rack hijacking in Rack 1.5 and later e['unicorn.socket'] = socket e['rack.hijack'] = self e.merge!(DEFAULTS) end
#trailers(env, buf)
#write_http_header(socket)
# File 'lib/unicorn/http_request.rb', line 185
def write_http_header(socket) # :nodoc: if headers? self.response_start_sent = true HTTP_RESPONSE_START.each { |c| socket.write(c) } end end