Mercurial > public > mercurial-scm > hg-stable
view mercurial/wireprotoframing.py @ 37063:0a6c5cc09a88
wireproto: define human output side channel frame
Currently, the SSH protocol delivers output tailored for people over
the stderr file descriptor. The HTTP protocol doesn't have this
file descriptor (because it only has an input and output pipe). So
it encodes textual output intended for humans within the protocol
responses. So response types have a facility for capturing output
to be printed to users. Some don't. And sometimes the implementation
of how that output is conveyed is super hacky.
On top of that, bundle2 has an "output" part that is used to store
output that should be printed when this part is encountered.
bundle2 also has the concept of "interrupt" chunks, which can be
used to signal that the regular bundle2 stream is to be
preempted by an out-of-band part that should be processed immediately.
This "interrupt" part can be an "output" part and can be used to
print data on the receiver.
The status quo is inconsistent and insane. We can do better.
This commit introduces a dedicated frame type on the frame-based
protocol for denoting textual data that should be printed on the
receiver. This frame type effectively constitutes a side-channel
by which textual data can be printed on the receiver without
interfering with other in-progress transmissions, such as the
transmission of command responses.
But wait - there's more! Previous implementations that transferred
textual data basically instructed the client to "print these bytes."
This suffered from a few problems.
First, the text data that was transmitted and eventually printed
originated from a server with a specic i18n configuration. This
meant that clients would see text using whatever the i18n settings
were on the server. Someone in France could connect to a server in
Japan and see unlegible Japanese glyphs - or maybe even mojibake.
Second, the normalization of all text data originated on servers
resulted in the loss of the ability to apply formatting to that
data. Local Mercurial clients can apply specific formatting
settings to individual atoms of text. For example, a revision can
be colored differently from a commit message. With data over the
wire, the potential for this rich formatting was lost. The best you
could do (without parsing the text to be printed), was apply a
universal label to it and e.g. color it specially.
The new mechanism for instructing the peer to print data does
not have these limitations.
Frames instructing the peer to print text are composed of a
formatting string plus arguments. In other words, receivers can
plug the formatting string into the i18n database to see if a local
translation is available. In addition, each atom being instructed
to print has a series of "labels" associated with it. These labels
can be mapped to the Mercurial UI's labels so locally configured
coloring, styling, etc settings can be applied.
What this all means is that textual messages originating on servers
can be localized on the client and richly formatted, all while
respecting the client's settings. This is slightly more complicated
than "print these bytes." But it is vastly more user friendly.
FWIW, I'm not aware of other protocols that attempt to encode
i18n and textual styling in this manner. You could lobby the
claim that this feature is over-engineered. However, if I were to
sit in the shoes of a non-English speaker learning how to use
version control, I think I would *love* this feature because
it would enable me to see richly formatted text in my chosen
locale.
Anyway, we only implement support for encoding frames of this
type and basic tests for that encoding. We'll still need to
hook up the server and its ui instance to emit these frames.
I recognize this feature may be a bit more controversial than
other aspects of the wire protocol because it is a bit
"radical." So I'd figured I'd start small to test the waters and
see if others feel this feature is worthwhile.
Differential Revision: https://phab.mercurial-scm.org/D2872
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Wed, 14 Mar 2018 22:19:00 -0700 |
parents | c5e9c3b47366 |
children | 884a0c1604ad |
line wrap: on
line source
# wireprotoframing.py - unified framing protocol for wire protocol # # Copyright 2018 Gregory Szorc <gregory.szorc@gmail.com> # # This software may be used and distributed according to the terms of the # GNU General Public License version 2 or any later version. # This file contains functionality to support the unified frame-based wire # protocol. For details about the protocol, see # `hg help internals.wireprotocol`. from __future__ import absolute_import import struct from .i18n import _ from . import ( error, util, ) FRAME_HEADER_SIZE = 6 DEFAULT_MAX_FRAME_SIZE = 32768 FRAME_TYPE_COMMAND_NAME = 0x01 FRAME_TYPE_COMMAND_ARGUMENT = 0x02 FRAME_TYPE_COMMAND_DATA = 0x03 FRAME_TYPE_BYTES_RESPONSE = 0x04 FRAME_TYPE_ERROR_RESPONSE = 0x05 FRAME_TYPE_TEXT_OUTPUT = 0x06 FRAME_TYPES = { b'command-name': FRAME_TYPE_COMMAND_NAME, b'command-argument': FRAME_TYPE_COMMAND_ARGUMENT, b'command-data': FRAME_TYPE_COMMAND_DATA, b'bytes-response': FRAME_TYPE_BYTES_RESPONSE, b'error-response': FRAME_TYPE_ERROR_RESPONSE, b'text-output': FRAME_TYPE_TEXT_OUTPUT, } FLAG_COMMAND_NAME_EOS = 0x01 FLAG_COMMAND_NAME_HAVE_ARGS = 0x02 FLAG_COMMAND_NAME_HAVE_DATA = 0x04 FLAGS_COMMAND = { b'eos': FLAG_COMMAND_NAME_EOS, b'have-args': FLAG_COMMAND_NAME_HAVE_ARGS, b'have-data': FLAG_COMMAND_NAME_HAVE_DATA, } FLAG_COMMAND_ARGUMENT_CONTINUATION = 0x01 FLAG_COMMAND_ARGUMENT_EOA = 0x02 FLAGS_COMMAND_ARGUMENT = { b'continuation': FLAG_COMMAND_ARGUMENT_CONTINUATION, b'eoa': FLAG_COMMAND_ARGUMENT_EOA, } FLAG_COMMAND_DATA_CONTINUATION = 0x01 FLAG_COMMAND_DATA_EOS = 0x02 FLAGS_COMMAND_DATA = { b'continuation': FLAG_COMMAND_DATA_CONTINUATION, b'eos': FLAG_COMMAND_DATA_EOS, } FLAG_BYTES_RESPONSE_CONTINUATION = 0x01 FLAG_BYTES_RESPONSE_EOS = 0x02 FLAGS_BYTES_RESPONSE = { b'continuation': FLAG_BYTES_RESPONSE_CONTINUATION, b'eos': FLAG_BYTES_RESPONSE_EOS, } FLAG_ERROR_RESPONSE_PROTOCOL = 0x01 FLAG_ERROR_RESPONSE_APPLICATION = 0x02 FLAGS_ERROR_RESPONSE = { b'protocol': FLAG_ERROR_RESPONSE_PROTOCOL, b'application': FLAG_ERROR_RESPONSE_APPLICATION, } # Maps frame types to their available flags. FRAME_TYPE_FLAGS = { FRAME_TYPE_COMMAND_NAME: FLAGS_COMMAND, FRAME_TYPE_COMMAND_ARGUMENT: FLAGS_COMMAND_ARGUMENT, FRAME_TYPE_COMMAND_DATA: FLAGS_COMMAND_DATA, FRAME_TYPE_BYTES_RESPONSE: FLAGS_BYTES_RESPONSE, FRAME_TYPE_ERROR_RESPONSE: FLAGS_ERROR_RESPONSE, FRAME_TYPE_TEXT_OUTPUT: {}, } ARGUMENT_FRAME_HEADER = struct.Struct(r'<HH') def makeframe(requestid, frametype, frameflags, payload): """Assemble a frame into a byte array.""" # TODO assert size of payload. frame = bytearray(FRAME_HEADER_SIZE + len(payload)) # 24 bits length # 16 bits request id # 4 bits type # 4 bits flags l = struct.pack(r'<I', len(payload)) frame[0:3] = l[0:3] struct.pack_into(r'<H', frame, 3, requestid) frame[5] = (frametype << 4) | frameflags frame[6:] = payload return frame def makeframefromhumanstring(s): """Create a frame from a human readable string Strings have the form: <request-id> <type> <flags> <payload> This can be used by user-facing applications and tests for creating frames easily without having to type out a bunch of constants. Request ID is an integer. Frame type and flags can be specified by integer or named constant. Flags can be delimited by `|` to bitwise OR them together. """ requestid, frametype, frameflags, payload = s.split(b' ', 3) requestid = int(requestid) if frametype in FRAME_TYPES: frametype = FRAME_TYPES[frametype] else: frametype = int(frametype) finalflags = 0 validflags = FRAME_TYPE_FLAGS[frametype] for flag in frameflags.split(b'|'): if flag in validflags: finalflags |= validflags[flag] else: finalflags |= int(flag) payload = util.unescapestr(payload) return makeframe(requestid, frametype, finalflags, payload) def parseheader(data): """Parse a unified framing protocol frame header from a buffer. The header is expected to be in the buffer at offset 0 and the buffer is expected to be large enough to hold a full header. """ # 24 bits payload length (little endian) # 4 bits frame type # 4 bits frame flags # ... payload framelength = data[0] + 256 * data[1] + 16384 * data[2] requestid = struct.unpack_from(r'<H', data, 3)[0] typeflags = data[5] frametype = (typeflags & 0xf0) >> 4 frameflags = typeflags & 0x0f return requestid, frametype, frameflags, framelength def readframe(fh): """Read a unified framing protocol frame from a file object. Returns a 3-tuple of (type, flags, payload) for the decoded frame or None if no frame is available. May raise if a malformed frame is seen. """ header = bytearray(FRAME_HEADER_SIZE) readcount = fh.readinto(header) if readcount == 0: return None if readcount != FRAME_HEADER_SIZE: raise error.Abort(_('received incomplete frame: got %d bytes: %s') % (readcount, header)) requestid, frametype, frameflags, framelength = parseheader(header) payload = fh.read(framelength) if len(payload) != framelength: raise error.Abort(_('frame length error: expected %d; got %d') % (framelength, len(payload))) return requestid, frametype, frameflags, payload def createcommandframes(requestid, cmd, args, datafh=None): """Create frames necessary to transmit a request to run a command. This is a generator of bytearrays. Each item represents a frame ready to be sent over the wire to a peer. """ flags = 0 if args: flags |= FLAG_COMMAND_NAME_HAVE_ARGS if datafh: flags |= FLAG_COMMAND_NAME_HAVE_DATA if not flags: flags |= FLAG_COMMAND_NAME_EOS yield makeframe(requestid, FRAME_TYPE_COMMAND_NAME, flags, cmd) for i, k in enumerate(sorted(args)): v = args[k] last = i == len(args) - 1 # TODO handle splitting of argument values across frames. payload = bytearray(ARGUMENT_FRAME_HEADER.size + len(k) + len(v)) offset = 0 ARGUMENT_FRAME_HEADER.pack_into(payload, offset, len(k), len(v)) offset += ARGUMENT_FRAME_HEADER.size payload[offset:offset + len(k)] = k offset += len(k) payload[offset:offset + len(v)] = v flags = FLAG_COMMAND_ARGUMENT_EOA if last else 0 yield makeframe(requestid, FRAME_TYPE_COMMAND_ARGUMENT, flags, payload) if datafh: while True: data = datafh.read(DEFAULT_MAX_FRAME_SIZE) done = False if len(data) == DEFAULT_MAX_FRAME_SIZE: flags = FLAG_COMMAND_DATA_CONTINUATION else: flags = FLAG_COMMAND_DATA_EOS assert datafh.read(1) == b'' done = True yield makeframe(requestid, FRAME_TYPE_COMMAND_DATA, flags, data) if done: break def createbytesresponseframesfrombytes(requestid, data, maxframesize=DEFAULT_MAX_FRAME_SIZE): """Create a raw frame to send a bytes response from static bytes input. Returns a generator of bytearrays. """ # Simple case of a single frame. if len(data) <= maxframesize: yield makeframe(requestid, FRAME_TYPE_BYTES_RESPONSE, FLAG_BYTES_RESPONSE_EOS, data) return offset = 0 while True: chunk = data[offset:offset + maxframesize] offset += len(chunk) done = offset == len(data) if done: flags = FLAG_BYTES_RESPONSE_EOS else: flags = FLAG_BYTES_RESPONSE_CONTINUATION yield makeframe(requestid, FRAME_TYPE_BYTES_RESPONSE, flags, chunk) if done: break def createerrorframe(requestid, msg, protocol=False, application=False): # TODO properly handle frame size limits. assert len(msg) <= DEFAULT_MAX_FRAME_SIZE flags = 0 if protocol: flags |= FLAG_ERROR_RESPONSE_PROTOCOL if application: flags |= FLAG_ERROR_RESPONSE_APPLICATION yield makeframe(requestid, FRAME_TYPE_ERROR_RESPONSE, flags, msg) def createtextoutputframe(requestid, atoms): """Create a text output frame to render text to people. ``atoms`` is a 3-tuple of (formatting string, args, labels). The formatting string contains ``%s`` tokens to be replaced by the corresponding indexed entry in ``args``. ``labels`` is an iterable of formatters to be applied at rendering time. In terms of the ``ui`` class, each atom corresponds to a ``ui.write()``. """ bytesleft = DEFAULT_MAX_FRAME_SIZE atomchunks = [] for (formatting, args, labels) in atoms: if len(args) > 255: raise ValueError('cannot use more than 255 formatting arguments') if len(labels) > 255: raise ValueError('cannot use more than 255 labels') # TODO look for localstr, other types here? if not isinstance(formatting, bytes): raise ValueError('must use bytes formatting strings') for arg in args: if not isinstance(arg, bytes): raise ValueError('must use bytes for arguments') for label in labels: if not isinstance(label, bytes): raise ValueError('must use bytes for labels') # Formatting string must be UTF-8. formatting = formatting.decode(r'utf-8', r'replace').encode(r'utf-8') # Arguments must be UTF-8. args = [a.decode(r'utf-8', r'replace').encode(r'utf-8') for a in args] # Labels must be ASCII. labels = [l.decode(r'ascii', r'strict').encode(r'ascii') for l in labels] if len(formatting) > 65535: raise ValueError('formatting string cannot be longer than 64k') if any(len(a) > 65535 for a in args): raise ValueError('argument string cannot be longer than 64k') if any(len(l) > 255 for l in labels): raise ValueError('label string cannot be longer than 255 bytes') chunks = [ struct.pack(r'<H', len(formatting)), struct.pack(r'<BB', len(labels), len(args)), struct.pack(r'<' + r'B' * len(labels), *map(len, labels)), struct.pack(r'<' + r'H' * len(args), *map(len, args)), ] chunks.append(formatting) chunks.extend(labels) chunks.extend(args) atom = b''.join(chunks) atomchunks.append(atom) bytesleft -= len(atom) if bytesleft < 0: raise ValueError('cannot encode data in a single frame') yield makeframe(requestid, FRAME_TYPE_TEXT_OUTPUT, 0, b''.join(atomchunks)) class serverreactor(object): """Holds state of a server handling frame-based protocol requests. This class is the "brain" of the unified frame-based protocol server component. While the protocol is stateless from the perspective of requests/commands, something needs to track which frames have been received, what frames to expect, etc. This class is that thing. Instances are modeled as a state machine of sorts. Instances are also reactionary to external events. The point of this class is to encapsulate the state of the connection and the exchange of frames, not to perform work. Instead, callers tell this class when something occurs, like a frame arriving. If that activity is worthy of a follow-up action (say *run a command*), the return value of that handler will say so. I/O and CPU intensive operations are purposefully delegated outside of this class. Consumers are expected to tell instances when events occur. They do so by calling the various ``on*`` methods. These methods return a 2-tuple describing any follow-up action(s) to take. The first element is the name of an action to perform. The second is a data structure (usually a dict) specific to that action that contains more information. e.g. if the server wants to send frames back to the client, the data structure will contain a reference to those frames. Valid actions that consumers can be instructed to take are: sendframes Indicates that frames should be sent to the client. The ``framegen`` key contains a generator of frames that should be sent. The server assumes that all frames are sent to the client. error Indicates that an error occurred. Consumer should probably abort. runcommand Indicates that the consumer should run a wire protocol command. Details of the command to run are given in the data structure. wantframe Indicates that nothing of interest happened and the server is waiting on more frames from the client before anything interesting can be done. noop Indicates no additional action is required. Known Issues ------------ There are no limits to the number of partially received commands or their size. A malicious client could stream command request data and exhaust the server's memory. Partially received commands are not acted upon when end of input is reached. Should the server error if it receives a partial request? Should the client send a message to abort a partially transmitted request to facilitate graceful shutdown? Active requests that haven't been responded to aren't tracked. This means that if we receive a command and instruct its dispatch, another command with its request ID can come in over the wire and there will be a race between who responds to what. """ def __init__(self, deferoutput=False): """Construct a new server reactor. ``deferoutput`` can be used to indicate that no output frames should be instructed to be sent until input has been exhausted. In this mode, events that would normally generate output frames (such as a command response being ready) will instead defer instructing the consumer to send those frames. This is useful for half-duplex transports where the sender cannot receive until all data has been transmitted. """ self._deferoutput = deferoutput self._state = 'idle' self._bufferedframegens = [] # request id -> dict of commands that are actively being received. self._receivingcommands = {} def onframerecv(self, requestid, frametype, frameflags, payload): """Process a frame that has been received off the wire. Returns a dict with an ``action`` key that details what action, if any, the consumer should take next. """ handlers = { 'idle': self._onframeidle, 'command-receiving': self._onframecommandreceiving, 'errored': self._onframeerrored, } meth = handlers.get(self._state) if not meth: raise error.ProgrammingError('unhandled state: %s' % self._state) return meth(requestid, frametype, frameflags, payload) def onbytesresponseready(self, requestid, data): """Signal that a bytes response is ready to be sent to the client. The raw bytes response is passed as an argument. """ framegen = createbytesresponseframesfrombytes(requestid, data) if self._deferoutput: self._bufferedframegens.append(framegen) return 'noop', {} else: return 'sendframes', { 'framegen': framegen, } def oninputeof(self): """Signals that end of input has been received. No more frames will be received. All pending activity should be completed. """ # TODO should we do anything about in-flight commands? if not self._deferoutput or not self._bufferedframegens: return 'noop', {} # If we buffered all our responses, emit those. def makegen(): for gen in self._bufferedframegens: for frame in gen: yield frame return 'sendframes', { 'framegen': makegen(), } def onapplicationerror(self, requestid, msg): return 'sendframes', { 'framegen': createerrorframe(requestid, msg, application=True), } def _makeerrorresult(self, msg): return 'error', { 'message': msg, } def _makeruncommandresult(self, requestid): entry = self._receivingcommands[requestid] del self._receivingcommands[requestid] if self._receivingcommands: self._state = 'command-receiving' else: self._state = 'idle' return 'runcommand', { 'requestid': requestid, 'command': entry['command'], 'args': entry['args'], 'data': entry['data'].getvalue() if entry['data'] else None, } def _makewantframeresult(self): return 'wantframe', { 'state': self._state, } def _onframeidle(self, requestid, frametype, frameflags, payload): # The only frame type that should be received in this state is a # command request. if frametype != FRAME_TYPE_COMMAND_NAME: self._state = 'errored' return self._makeerrorresult( _('expected command frame; got %d') % frametype) if requestid in self._receivingcommands: self._state = 'errored' return self._makeerrorresult( _('request with ID %d already received') % requestid) expectingargs = bool(frameflags & FLAG_COMMAND_NAME_HAVE_ARGS) expectingdata = bool(frameflags & FLAG_COMMAND_NAME_HAVE_DATA) self._receivingcommands[requestid] = { 'command': payload, 'args': {}, 'data': None, 'expectingargs': expectingargs, 'expectingdata': expectingdata, } if frameflags & FLAG_COMMAND_NAME_EOS: return self._makeruncommandresult(requestid) if expectingargs or expectingdata: self._state = 'command-receiving' return self._makewantframeresult() else: self._state = 'errored' return self._makeerrorresult(_('missing frame flags on ' 'command frame')) def _onframecommandreceiving(self, requestid, frametype, frameflags, payload): # It could be a new command request. Process it as such. if frametype == FRAME_TYPE_COMMAND_NAME: return self._onframeidle(requestid, frametype, frameflags, payload) # All other frames should be related to a command that is currently # receiving. if requestid not in self._receivingcommands: self._state = 'errored' return self._makeerrorresult( _('received frame for request that is not receiving: %d') % requestid) entry = self._receivingcommands[requestid] if frametype == FRAME_TYPE_COMMAND_ARGUMENT: if not entry['expectingargs']: self._state = 'errored' return self._makeerrorresult(_( 'received command argument frame for request that is not ' 'expecting arguments: %d') % requestid) return self._handlecommandargsframe(requestid, entry, frametype, frameflags, payload) elif frametype == FRAME_TYPE_COMMAND_DATA: if not entry['expectingdata']: self._state = 'errored' return self._makeerrorresult(_( 'received command data frame for request that is not ' 'expecting data: %d') % requestid) if entry['data'] is None: entry['data'] = util.bytesio() return self._handlecommanddataframe(requestid, entry, frametype, frameflags, payload) def _handlecommandargsframe(self, requestid, entry, frametype, frameflags, payload): # The frame and state of command should have already been validated. assert frametype == FRAME_TYPE_COMMAND_ARGUMENT offset = 0 namesize, valuesize = ARGUMENT_FRAME_HEADER.unpack_from(payload) offset += ARGUMENT_FRAME_HEADER.size # The argument name MUST fit inside the frame. argname = bytes(payload[offset:offset + namesize]) offset += namesize if len(argname) != namesize: self._state = 'errored' return self._makeerrorresult(_('malformed argument frame: ' 'partial argument name')) argvalue = bytes(payload[offset:]) # Argument value spans multiple frames. Record our active state # and wait for the next frame. if frameflags & FLAG_COMMAND_ARGUMENT_CONTINUATION: raise error.ProgrammingError('not yet implemented') # Common case: the argument value is completely contained in this # frame. if len(argvalue) != valuesize: self._state = 'errored' return self._makeerrorresult(_('malformed argument frame: ' 'partial argument value')) entry['args'][argname] = argvalue if frameflags & FLAG_COMMAND_ARGUMENT_EOA: if entry['expectingdata']: # TODO signal request to run a command once we don't # buffer data frames. return self._makewantframeresult() else: return self._makeruncommandresult(requestid) else: return self._makewantframeresult() def _handlecommanddataframe(self, requestid, entry, frametype, frameflags, payload): assert frametype == FRAME_TYPE_COMMAND_DATA # TODO support streaming data instead of buffering it. entry['data'].write(payload) if frameflags & FLAG_COMMAND_DATA_CONTINUATION: return self._makewantframeresult() elif frameflags & FLAG_COMMAND_DATA_EOS: entry['data'].seek(0) return self._makeruncommandresult(requestid) else: self._state = 'errored' return self._makeerrorresult(_('command data frame without ' 'flags')) def _onframeerrored(self, requestid, frametype, frameflags, payload): return self._makeerrorresult(_('server already errored'))