revlog: automatically read from opened file handles
The revlog reading code commonly opens a new file handle for
reading on demand. There is support for passing a file handle
to revlog.revision(). But it is marked as an internal argument.
When revlogs are written, we write() data as it is available. But
we don't flush() data until all revisions are written.
Putting these two traits together, it is possible for an in-process
revlog reader during active writes to trigger the opening of a new
file handle on a file with unflushed writes. The reader won't have
access to all "available" revlog data (as it hasn't been flushed).
And with the introduction of the previous patch, this can lead to
the revlog raising an error due to a partial read.
I witnessed this behavior when applying changegroup data (via
`hg pull`) before issue6006 was fixed via different means. Having
this and the previous patch in play would have helped cause errors
earlier rather than manifesting as hash verification failures.
While this has been a long-standing issue, I believe the relatively
new delta computation code has tickled it into being more common.
This is because the new delta computation code will compute deltas
in more scenarios. This can lead to revlog reading. While the delta
computation code is probably supposed to reuse file handles, it
appears it isn't doing so in all circumstances.
But the issue runs deeper than that. Theoretically, any code can
access revision data during revlog writes. It appears we were just
getting lucky that it wasn't. (The "add revision callback" passed to
addgroup() provides an avenue to do this.)
If I changed the revlog's behavior to not cache the full revision
text or to clear caches after revision insertion during addgroup(),
I was able to produce crashes 100% of the time when writing changelog
revisions. This is because changelog's add revision callback attempts
to resolve the revision data to access the changed files list. And
without the revision's fulltext being cached, we performed a revlog
read, which required opening a new file handle. This attempted to read
unflushed data, leading to a partial read and a crash.
This commit teaches the revlog to store the file handles used for
writing multiple revisions during addgroup(). It also teaches the
code for resolving a file handle when reading to use these handles,
if available. This ensures that *any* reads (regardless of their
source) use the active writing file handles, if available. These
file handles have access to the unflushed data because they wrote it.
This allows reads to complete without issue.
Differential Revision: https://phab.mercurial-scm.org/D5267
# sshprotoext.py - Extension to test behavior of SSH 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 extension replaces the SSH server started via `hg serve --stdio`.
# The server behaves differently depending on environment variables.
from __future__ import absolute_import
from mercurial import (
error,
extensions,
registrar,
sshpeer,
wireprotoserver,
wireprotov1server,
)
configtable = {}
configitem = registrar.configitem(configtable)
configitem(b'sshpeer', b'mode', default=None)
configitem(b'sshpeer', b'handshake-mode', default=None)
class bannerserver(wireprotoserver.sshserver):
"""Server that sends a banner to stdout."""
def serve_forever(self):
for i in range(10):
self._fout.write(b'banner: line %d\n' % i)
super(bannerserver, self).serve_forever()
class prehelloserver(wireprotoserver.sshserver):
"""Tests behavior when connecting to <0.9.1 servers.
The ``hello`` wire protocol command was introduced in Mercurial
0.9.1. Modern clients send the ``hello`` command when connecting
to SSH servers. This mock server tests behavior of the handshake
when ``hello`` is not supported.
"""
def serve_forever(self):
l = self._fin.readline()
assert l == b'hello\n'
# Respond to unknown commands with an empty reply.
wireprotoserver._sshv1respondbytes(self._fout, b'')
l = self._fin.readline()
assert l == b'between\n'
proto = wireprotoserver.sshv1protocolhandler(self._ui, self._fin,
self._fout)
rsp = wireprotov1server.dispatch(self._repo, proto, b'between')
wireprotoserver._sshv1respondbytes(self._fout, rsp.data)
super(prehelloserver, self).serve_forever()
def performhandshake(orig, ui, stdin, stdout, stderr):
"""Wrapped version of sshpeer._performhandshake to send extra commands."""
mode = ui.config(b'sshpeer', b'handshake-mode')
if mode == b'pre-no-args':
ui.debug(b'sending no-args command\n')
stdin.write(b'no-args\n')
stdin.flush()
return orig(ui, stdin, stdout, stderr)
elif mode == b'pre-multiple-no-args':
ui.debug(b'sending unknown1 command\n')
stdin.write(b'unknown1\n')
ui.debug(b'sending unknown2 command\n')
stdin.write(b'unknown2\n')
ui.debug(b'sending unknown3 command\n')
stdin.write(b'unknown3\n')
stdin.flush()
return orig(ui, stdin, stdout, stderr)
else:
raise error.ProgrammingError(b'unknown HANDSHAKECOMMANDMODE: %s' %
mode)
def extsetup(ui):
# It's easier for tests to define the server behavior via environment
# variables than config options. This is because `hg serve --stdio`
# has to be invoked with a certain form for security reasons and
# `dummyssh` can't just add `--config` flags to the command line.
servermode = ui.environ.get(b'SSHSERVERMODE')
if servermode == b'banner':
wireprotoserver.sshserver = bannerserver
elif servermode == b'no-hello':
wireprotoserver.sshserver = prehelloserver
elif servermode:
raise error.ProgrammingError(b'unknown server mode: %s' % servermode)
peermode = ui.config(b'sshpeer', b'mode')
if peermode == b'extra-handshake-commands':
extensions.wrapfunction(sshpeer, '_performhandshake', performhandshake)
elif peermode:
raise error.ProgrammingError(b'unknown peer mode: %s' % peermode)