##// END OF EJS Templates
wireproto: add streams to frame-based protocol...
wireproto: add streams to frame-based protocol Previously, the frame-based protocol was just a series of frames, with each frame associated with a request ID. In order to scale the protocol, we'll want to enable the use of compression. While it is possible to enable compression at the socket/pipe level, this has its disadvantages. The big one is it undermines the point of frames being standalone, atomic units that can be read and written: if you add compression above the framing protocol, you are back to having a stream-based protocol as opposed to something frame-based. So in order to preserve frames, compression needs to occur at the frame payload level. Compressing each frame's payload individually will limit compression ratios because the window size of the compressor will be limited by the max frame size, which is 32-64kb as currently defined. It will also add CPU overhead, as it is more efficient for compressors to operate on fewer, larger blocks of data than more, smaller blocks. So compressing each frame independently is out. This means we need to compress each frame's payload as if it is part of a larger stream. The simplest approach is to have 1 stream per connection. This could certainly work. However, it has disadvantages (documented below). We could also have 1 stream per RPC/command invocation. (This is the model HTTP/2 goes with.) This also has disadvantages. The main disadvantage to one global stream is that it has the very real potential to create CPU bottlenecks doing compression. Networks are only getting faster and the performance of single CPU cores has been relatively flat. Newer compression formats like zstandard offer better CPU cycle efficiency than predecessors like zlib. But it still all too common to saturate your CPU with compression overhead long before you saturate the network pipe. The main disadvantage with streams per request is that you can't reap the benefits of the compression context for multiple requests. For example, if you send 1000 RPC requests (or HTTP/2 requests for that matter), the response to each would have its own compression context. The overall size of the raw responses would be larger because compression contexts wouldn't be able to reference data from another request or response. The approach for streams as implemented in this commit is to support N streams per connection and for streams to potentially span requests and responses. As explained by the added internals docs, this facilitates servers and clients delegating independent streams and compression to independent threads / CPU cores. This helps alleviate the CPU bottleneck of compression. This design also allows compression contexts to be reused across requests/responses. This can result in improved compression ratios and less overhead for compressors and decompressors having to build new contexts. Another feature that was defined was the ability for individual frames within a stream to declare whether that individual frame's payload uses the content encoding (read: compression) defined by the stream. The idea here is that some servers may serve data from a combination of caches and dynamic resolution. Data coming from caches may be pre-compressed. We want to facilitate servers being able to essentially stream bytes from caches to the wire with minimal overhead. Being able to mix and match with frames are compressed within a stream enables these types of advanced server functionality. This commit defines the new streams mechanism. Basic code for supporting streams in frames has been added. But that code is seriously lacking and doesn't fully conform to the defined protocol. For example, we don't close any streams. And support for content encoding within streams is not yet implemented. The change was rather invasive and I didn't think it would be reasonable to implement the entire feature in a single commit. For the record, I would have loved to reuse an existing multiplexing protocol to build the new wire protocol on top of. However, I couldn't find a protocol that offers the performance and scaling characteristics that I desired. Namely, it should support multiple compression contexts to facilitate scaling out to multiple CPU cores and compression contexts should be able to live longer than single RPC requests. HTTP/2 *almost* fits the bill. But the semantics of HTTP message exchange state that streams can only live for a single request-response. We /could/ tunnel on top of HTTP/2 streams and frames with HEADER and DATA frames. But there's no guarantee that HTTP/2 libraries and proxies would allow us to use HTTP/2 streams and frames without the HTTP message exchange semantics defined in RFC 7540 Section 8. Other RPC protocols like gRPC tunnel are built on top of HTTP/2 and thus preserve its semantics of stream per RPC invocation. Even QUIC does this. We could attempt to invent a higher-level stream that spans HTTP/2 streams. But this would be violating HTTP/2 because there is no guarantee that HTTP/2 streams are routed to the same server. The best we can do - which is what this protocol does - is shoehorn all request and response data into a single HTTP message and create streams within. At that point, we've defined a Content-Type in HTTP parlance. It just so happens our media type can also work as a standalone, stream-based protocol, without leaning on HTTP or similar protocol. Differential Revision: https://phab.mercurial-scm.org/D2907

File last commit:

r35600:72b91f90 default
r37304:9bfcbe4f default
Show More
store.py
579 lines | 17.3 KiB | text/x-python | PythonLexer
# store.py - repository store handling for Mercurial
#
# Copyright 2008 Matt Mackall <mpm@selenic.com>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
from __future__ import absolute_import
import errno
import hashlib
import os
import stat
from .i18n import _
from . import (
error,
node,
policy,
pycompat,
util,
vfs as vfsmod,
)
parsers = policy.importmod(r'parsers')
# This avoids a collision between a file named foo and a dir named
# foo.i or foo.d
def _encodedir(path):
'''
>>> _encodedir(b'data/foo.i')
'data/foo.i'
>>> _encodedir(b'data/foo.i/bla.i')
'data/foo.i.hg/bla.i'
>>> _encodedir(b'data/foo.i.hg/bla.i')
'data/foo.i.hg.hg/bla.i'
>>> _encodedir(b'data/foo.i\\ndata/foo.i/bla.i\\ndata/foo.i.hg/bla.i\\n')
'data/foo.i\\ndata/foo.i.hg/bla.i\\ndata/foo.i.hg.hg/bla.i\\n'
'''
return (path
.replace(".hg/", ".hg.hg/")
.replace(".i/", ".i.hg/")
.replace(".d/", ".d.hg/"))
encodedir = getattr(parsers, 'encodedir', _encodedir)
def decodedir(path):
'''
>>> decodedir(b'data/foo.i')
'data/foo.i'
>>> decodedir(b'data/foo.i.hg/bla.i')
'data/foo.i/bla.i'
>>> decodedir(b'data/foo.i.hg.hg/bla.i')
'data/foo.i.hg/bla.i'
'''
if ".hg/" not in path:
return path
return (path
.replace(".d.hg/", ".d/")
.replace(".i.hg/", ".i/")
.replace(".hg.hg/", ".hg/"))
def _reserved():
''' characters that are problematic for filesystems
* ascii escapes (0..31)
* ascii hi (126..255)
* windows specials
these characters will be escaped by encodefunctions
'''
winreserved = [ord(x) for x in u'\\:*?"<>|']
for x in range(32):
yield x
for x in range(126, 256):
yield x
for x in winreserved:
yield x
def _buildencodefun():
'''
>>> enc, dec = _buildencodefun()
>>> enc(b'nothing/special.txt')
'nothing/special.txt'
>>> dec(b'nothing/special.txt')
'nothing/special.txt'
>>> enc(b'HELLO')
'_h_e_l_l_o'
>>> dec(b'_h_e_l_l_o')
'HELLO'
>>> enc(b'hello:world?')
'hello~3aworld~3f'
>>> dec(b'hello~3aworld~3f')
'hello:world?'
>>> enc(b'the\\x07quick\\xADshot')
'the~07quick~adshot'
>>> dec(b'the~07quick~adshot')
'the\\x07quick\\xadshot'
'''
e = '_'
xchr = pycompat.bytechr
asciistr = list(map(xchr, range(127)))
capitals = list(range(ord("A"), ord("Z") + 1))
cmap = dict((x, x) for x in asciistr)
for x in _reserved():
cmap[xchr(x)] = "~%02x" % x
for x in capitals + [ord(e)]:
cmap[xchr(x)] = e + xchr(x).lower()
dmap = {}
for k, v in cmap.iteritems():
dmap[v] = k
def decode(s):
i = 0
while i < len(s):
for l in xrange(1, 4):
try:
yield dmap[s[i:i + l]]
i += l
break
except KeyError:
pass
else:
raise KeyError
return (lambda s: ''.join([cmap[s[c:c + 1]] for c in xrange(len(s))]),
lambda s: ''.join(list(decode(s))))
_encodefname, _decodefname = _buildencodefun()
def encodefilename(s):
'''
>>> encodefilename(b'foo.i/bar.d/bla.hg/hi:world?/HELLO')
'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o'
'''
return _encodefname(encodedir(s))
def decodefilename(s):
'''
>>> decodefilename(b'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o')
'foo.i/bar.d/bla.hg/hi:world?/HELLO'
'''
return decodedir(_decodefname(s))
def _buildlowerencodefun():
'''
>>> f = _buildlowerencodefun()
>>> f(b'nothing/special.txt')
'nothing/special.txt'
>>> f(b'HELLO')
'hello'
>>> f(b'hello:world?')
'hello~3aworld~3f'
>>> f(b'the\\x07quick\\xADshot')
'the~07quick~adshot'
'''
xchr = pycompat.bytechr
cmap = dict([(xchr(x), xchr(x)) for x in xrange(127)])
for x in _reserved():
cmap[xchr(x)] = "~%02x" % x
for x in range(ord("A"), ord("Z") + 1):
cmap[xchr(x)] = xchr(x).lower()
def lowerencode(s):
return "".join([cmap[c] for c in pycompat.iterbytestr(s)])
return lowerencode
lowerencode = getattr(parsers, 'lowerencode', None) or _buildlowerencodefun()
# Windows reserved names: con, prn, aux, nul, com1..com9, lpt1..lpt9
_winres3 = ('aux', 'con', 'prn', 'nul') # length 3
_winres4 = ('com', 'lpt') # length 4 (with trailing 1..9)
def _auxencode(path, dotencode):
'''
Encodes filenames containing names reserved by Windows or which end in
period or space. Does not touch other single reserved characters c.
Specifically, c in '\\:*?"<>|' or ord(c) <= 31 are *not* encoded here.
Additionally encodes space or period at the beginning, if dotencode is
True. Parameter path is assumed to be all lowercase.
A segment only needs encoding if a reserved name appears as a
basename (e.g. "aux", "aux.foo"). A directory or file named "foo.aux"
doesn't need encoding.
>>> s = b'.foo/aux.txt/txt.aux/con/prn/nul/foo.'
>>> _auxencode(s.split(b'/'), True)
['~2efoo', 'au~78.txt', 'txt.aux', 'co~6e', 'pr~6e', 'nu~6c', 'foo~2e']
>>> s = b'.com1com2/lpt9.lpt4.lpt1/conprn/com0/lpt0/foo.'
>>> _auxencode(s.split(b'/'), False)
['.com1com2', 'lp~749.lpt4.lpt1', 'conprn', 'com0', 'lpt0', 'foo~2e']
>>> _auxencode([b'foo. '], True)
['foo.~20']
>>> _auxencode([b' .foo'], True)
['~20.foo']
'''
for i, n in enumerate(path):
if not n:
continue
if dotencode and n[0] in '. ':
n = "~%02x" % ord(n[0:1]) + n[1:]
path[i] = n
else:
l = n.find('.')
if l == -1:
l = len(n)
if ((l == 3 and n[:3] in _winres3) or
(l == 4 and n[3:4] <= '9' and n[3:4] >= '1'
and n[:3] in _winres4)):
# encode third letter ('aux' -> 'au~78')
ec = "~%02x" % ord(n[2:3])
n = n[0:2] + ec + n[3:]
path[i] = n
if n[-1] in '. ':
# encode last period or space ('foo...' -> 'foo..~2e')
path[i] = n[:-1] + "~%02x" % ord(n[-1:])
return path
_maxstorepathlen = 120
_dirprefixlen = 8
_maxshortdirslen = 8 * (_dirprefixlen + 1) - 4
def _hashencode(path, dotencode):
digest = node.hex(hashlib.sha1(path).digest())
le = lowerencode(path[5:]).split('/') # skips prefix 'data/' or 'meta/'
parts = _auxencode(le, dotencode)
basename = parts[-1]
_root, ext = os.path.splitext(basename)
sdirs = []
sdirslen = 0
for p in parts[:-1]:
d = p[:_dirprefixlen]
if d[-1] in '. ':
# Windows can't access dirs ending in period or space
d = d[:-1] + '_'
if sdirslen == 0:
t = len(d)
else:
t = sdirslen + 1 + len(d)
if t > _maxshortdirslen:
break
sdirs.append(d)
sdirslen = t
dirs = '/'.join(sdirs)
if len(dirs) > 0:
dirs += '/'
res = 'dh/' + dirs + digest + ext
spaceleft = _maxstorepathlen - len(res)
if spaceleft > 0:
filler = basename[:spaceleft]
res = 'dh/' + dirs + filler + digest + ext
return res
def _hybridencode(path, dotencode):
'''encodes path with a length limit
Encodes all paths that begin with 'data/', according to the following.
Default encoding (reversible):
Encodes all uppercase letters 'X' as '_x'. All reserved or illegal
characters are encoded as '~xx', where xx is the two digit hex code
of the character (see encodefilename).
Relevant path components consisting of Windows reserved filenames are
masked by encoding the third character ('aux' -> 'au~78', see _auxencode).
Hashed encoding (not reversible):
If the default-encoded path is longer than _maxstorepathlen, a
non-reversible hybrid hashing of the path is done instead.
This encoding uses up to _dirprefixlen characters of all directory
levels of the lowerencoded path, but not more levels than can fit into
_maxshortdirslen.
Then follows the filler followed by the sha digest of the full path.
The filler is the beginning of the basename of the lowerencoded path
(the basename is everything after the last path separator). The filler
is as long as possible, filling in characters from the basename until
the encoded path has _maxstorepathlen characters (or all chars of the
basename have been taken).
The extension (e.g. '.i' or '.d') is preserved.
The string 'data/' at the beginning is replaced with 'dh/', if the hashed
encoding was used.
'''
path = encodedir(path)
ef = _encodefname(path).split('/')
res = '/'.join(_auxencode(ef, dotencode))
if len(res) > _maxstorepathlen:
res = _hashencode(path, dotencode)
return res
def _pathencode(path):
de = encodedir(path)
if len(path) > _maxstorepathlen:
return _hashencode(de, True)
ef = _encodefname(de).split('/')
res = '/'.join(_auxencode(ef, True))
if len(res) > _maxstorepathlen:
return _hashencode(de, True)
return res
_pathencode = getattr(parsers, 'pathencode', _pathencode)
def _plainhybridencode(f):
return _hybridencode(f, False)
def _calcmode(vfs):
try:
# files in .hg/ will be created using this mode
mode = vfs.stat().st_mode
# avoid some useless chmods
if (0o777 & ~util.umask) == (0o777 & mode):
mode = None
except OSError:
mode = None
return mode
_data = ('data meta 00manifest.d 00manifest.i 00changelog.d 00changelog.i'
' phaseroots obsstore')
class basicstore(object):
'''base class for local repository stores'''
def __init__(self, path, vfstype):
vfs = vfstype(path)
self.path = vfs.base
self.createmode = _calcmode(vfs)
vfs.createmode = self.createmode
self.rawvfs = vfs
self.vfs = vfsmod.filtervfs(vfs, encodedir)
self.opener = self.vfs
def join(self, f):
return self.path + '/' + encodedir(f)
def _walk(self, relpath, recurse):
'''yields (unencoded, encoded, size)'''
path = self.path
if relpath:
path += '/' + relpath
striplen = len(self.path) + 1
l = []
if self.rawvfs.isdir(path):
visit = [path]
readdir = self.rawvfs.readdir
while visit:
p = visit.pop()
for f, kind, st in readdir(p, stat=True):
fp = p + '/' + f
if kind == stat.S_IFREG and f[-2:] in ('.d', '.i'):
n = util.pconvert(fp[striplen:])
l.append((decodedir(n), n, st.st_size))
elif kind == stat.S_IFDIR and recurse:
visit.append(fp)
l.sort()
return l
def datafiles(self):
return self._walk('data', True) + self._walk('meta', True)
def topfiles(self):
# yield manifest before changelog
return reversed(self._walk('', False))
def walk(self):
'''yields (unencoded, encoded, size)'''
# yield data files first
for x in self.datafiles():
yield x
for x in self.topfiles():
yield x
def copylist(self):
return ['requires'] + _data.split()
def write(self, tr):
pass
def invalidatecaches(self):
pass
def markremoved(self, fn):
pass
def __contains__(self, path):
'''Checks if the store contains path'''
path = "/".join(("data", path))
# file?
if self.vfs.exists(path + ".i"):
return True
# dir?
if not path.endswith("/"):
path = path + "/"
return self.vfs.exists(path)
class encodedstore(basicstore):
def __init__(self, path, vfstype):
vfs = vfstype(path + '/store')
self.path = vfs.base
self.createmode = _calcmode(vfs)
vfs.createmode = self.createmode
self.rawvfs = vfs
self.vfs = vfsmod.filtervfs(vfs, encodefilename)
self.opener = self.vfs
def datafiles(self):
for a, b, size in super(encodedstore, self).datafiles():
try:
a = decodefilename(a)
except KeyError:
a = None
yield a, b, size
def join(self, f):
return self.path + '/' + encodefilename(f)
def copylist(self):
return (['requires', '00changelog.i'] +
['store/' + f for f in _data.split()])
class fncache(object):
# the filename used to be partially encoded
# hence the encodedir/decodedir dance
def __init__(self, vfs):
self.vfs = vfs
self.entries = None
self._dirty = False
def _load(self):
'''fill the entries from the fncache file'''
self._dirty = False
try:
fp = self.vfs('fncache', mode='rb')
except IOError:
# skip nonexistent file
self.entries = set()
return
self.entries = set(decodedir(fp.read()).splitlines())
if '' in self.entries:
fp.seek(0)
for n, line in enumerate(util.iterfile(fp)):
if not line.rstrip('\n'):
t = _('invalid entry in fncache, line %d') % (n + 1)
raise error.Abort(t)
fp.close()
def write(self, tr):
if self._dirty:
tr.addbackup('fncache')
fp = self.vfs('fncache', mode='wb', atomictemp=True)
if self.entries:
fp.write(encodedir('\n'.join(self.entries) + '\n'))
fp.close()
self._dirty = False
def add(self, fn):
if self.entries is None:
self._load()
if fn not in self.entries:
self._dirty = True
self.entries.add(fn)
def remove(self, fn):
if self.entries is None:
self._load()
try:
self.entries.remove(fn)
self._dirty = True
except KeyError:
pass
def __contains__(self, fn):
if self.entries is None:
self._load()
return fn in self.entries
def __iter__(self):
if self.entries is None:
self._load()
return iter(self.entries)
class _fncachevfs(vfsmod.abstractvfs, vfsmod.proxyvfs):
def __init__(self, vfs, fnc, encode):
vfsmod.proxyvfs.__init__(self, vfs)
self.fncache = fnc
self.encode = encode
def __call__(self, path, mode='r', *args, **kw):
if mode not in ('r', 'rb') and (path.startswith('data/') or
path.startswith('meta/')):
self.fncache.add(path)
return self.vfs(self.encode(path), mode, *args, **kw)
def join(self, path):
if path:
return self.vfs.join(self.encode(path))
else:
return self.vfs.join(path)
class fncachestore(basicstore):
def __init__(self, path, vfstype, dotencode):
if dotencode:
encode = _pathencode
else:
encode = _plainhybridencode
self.encode = encode
vfs = vfstype(path + '/store')
self.path = vfs.base
self.pathsep = self.path + '/'
self.createmode = _calcmode(vfs)
vfs.createmode = self.createmode
self.rawvfs = vfs
fnc = fncache(vfs)
self.fncache = fnc
self.vfs = _fncachevfs(vfs, fnc, encode)
self.opener = self.vfs
def join(self, f):
return self.pathsep + self.encode(f)
def getsize(self, path):
return self.rawvfs.stat(path).st_size
def datafiles(self):
for f in sorted(self.fncache):
ef = self.encode(f)
try:
yield f, ef, self.getsize(ef)
except OSError as err:
if err.errno != errno.ENOENT:
raise
def copylist(self):
d = ('data meta dh fncache phaseroots obsstore'
' 00manifest.d 00manifest.i 00changelog.d 00changelog.i')
return (['requires', '00changelog.i'] +
['store/' + f for f in d.split()])
def write(self, tr):
self.fncache.write(tr)
def invalidatecaches(self):
self.fncache.entries = None
def markremoved(self, fn):
self.fncache.remove(fn)
def _exists(self, f):
ef = self.encode(f)
try:
self.getsize(ef)
return True
except OSError as err:
if err.errno != errno.ENOENT:
raise
# nonexistent entry
return False
def __contains__(self, path):
'''Checks if the store contains path'''
path = "/".join(("data", path))
# check for files (exact match)
e = path + '.i'
if e in self.fncache and self._exists(e):
return True
# now check for directories (prefix match)
if not path.endswith('/'):
path += '/'
for e in self.fncache:
if e.startswith(path) and self._exists(e):
return True
return False
def store(requirements, path, vfstype):
if 'store' in requirements:
if 'fncache' in requirements:
return fncachestore(path, vfstype, 'dotencode' in requirements)
return encodedstore(path, vfstype)
return basicstore(path, vfstype)