bundle2: increase payload part chunk size to 32kb
Bundle2 payload parts are framed chunks. Esentially, we obtain
data in equal size chunks of size `preferedchunksize` and emit those
to a generator. That generator is fed into a compressor (which can
be the no-op compressor, which just re-emits the generator). And
the output from the compressor likely goes to a file descriptor
or socket.
What this means is that small chunk sizes create more Python objects
and Python function calls than larger chunk sizes. And as we know,
Python object and function call overhead in performance sensitive
code matters (at least with CPython).
This commit increases the bundle2 part payload chunk size from 4k
to 32k. Practically speaking, this means that the chunks we feed
into a compressor (implemented in C code) or feed directly into a
file handle or socket write() are larger. It's possible the chunks
might be larger than what the receiver can handle in one logical
operation. But at that point, we're in C code, which is much more
efficient at dealing with splitting up the chunk and making multiple
function calls than Python is.
A downside to larger chunks is that the receiver has to wait for that
much data to arrive (either raw or from a decompressor) before it
can process the chunk. But 32kb still feels like a small buffer to
have to wait for. And in many cases, the client will convert from
8 read(4096) to 1 read(32768). That's happening in Python land. So
we cut down on the number of Python objects and function calls,
making the client faster as well. I don't think there are any
significant concerns to increasing the payload chunk size to 32kb.
The impact of this change on performance significant. Using `curl`
to obtain a stream clone bundle2 payload from a server on localhost
serving the mozilla-unified repository:
before: 20.78 user; 7.71 system; 80.5 MB/s
after: 13.90 user; 3.51 system; 132 MB/s
legacy: 9.72 user; 8.16 system; 132 MB/s
bundle2 stream clone generation is still more resource intensive than
legacy stream clone (that's likely because of the use of a
util.chunkbuffer). But the throughput is the same. We might
be in territory we're this is effectively a benchmark of the
networking stack or Python's syscall throughput.
From the client perspective, `hg clone -U --stream`:
before: 33.50 user; 7.95 system; 53.3 MB/s
after: 22.82 user; 7.33 system; 72.7 MB/s
legacy: 29.96 user; 7.94 system; 58.0 MB/s
And for `hg clone --stream` with a working directory update of
~230k files:
after: 119.55 user; 26.47 system; 0:57.08 wall
legacy: 126.98 user; 26.94 system; 1:05.56 wall
So, it appears that bundle2's stream clone is now definitively faster
than legacy stream clone!
Differential Revision: https://phab.mercurial-scm.org/D1932
revlog.parseindex must be able to parse the index file even if
an index entry is split between two 64k blocks. The ideal test
would be to create an index file with inline data where
64k < size < 64k + 64 (64k is the size of the read buffer, 64 is
the size of an index entry) and with an index entry starting right
before the 64k block boundary, and try to read it.
We approximate that by reducing the read buffer to 1 byte.
$ hg init a
$ cd a
$ echo abc > foo
$ hg add foo
$ hg commit -m 'add foo'
$ echo >> foo
$ hg commit -m 'change foo'
$ hg log -r 0:
changeset: 0:7c31755bf9b5
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: add foo
changeset: 1:26333235a41c
tag: tip
user: test
date: Thu Jan 01 00:00:00 1970 +0000
summary: change foo
$ cat >> test.py << EOF
> from mercurial import changelog, vfs
> from mercurial.node import *
>
> class singlebyteread(object):
> def __init__(self, real):
> self.real = real
>
> def read(self, size=-1):
> if size == 65536:
> size = 1
> return self.real.read(size)
>
> def __getattr__(self, key):
> return getattr(self.real, key)
>
> def opener(*args):
> o = vfs.vfs(*args)
> def wrapper(*a):
> f = o(*a)
> return singlebyteread(f)
> return wrapper
>
> cl = changelog.changelog(opener('.hg/store'))
> print len(cl), 'revisions:'
> for r in cl:
> print short(cl.node(r))
> EOF
$ $PYTHON test.py
2 revisions:
7c31755bf9b5
26333235a41c
$ cd ..
#if no-pure
Test SEGV caused by bad revision passed to reachableroots() (issue4775):
$ cd a
$ $PYTHON <<EOF
> from mercurial import changelog, vfs
> cl = changelog.changelog(vfs.vfs('.hg/store'))
> print 'good heads:'
> for head in [0, len(cl) - 1, -1]:
> print'%s: %r' % (head, cl.reachableroots(0, [head], [0]))
> print 'bad heads:'
> for head in [len(cl), 10000, -2, -10000, None]:
> print '%s:' % head,
> try:
> cl.reachableroots(0, [head], [0])
> print 'uncaught buffer overflow?'
> except (IndexError, TypeError) as inst:
> print inst
> print 'good roots:'
> for root in [0, len(cl) - 1, -1]:
> print '%s: %r' % (root, cl.reachableroots(root, [len(cl) - 1], [root]))
> print 'out-of-range roots are ignored:'
> for root in [len(cl), 10000, -2, -10000]:
> print '%s: %r' % (root, cl.reachableroots(root, [len(cl) - 1], [root]))
> print 'bad roots:'
> for root in [None]:
> print '%s:' % root,
> try:
> cl.reachableroots(root, [len(cl) - 1], [root])
> print 'uncaught error?'
> except TypeError as inst:
> print inst
> EOF
good heads:
0: [0]
1: [0]
-1: []
bad heads:
2: head out of range
10000: head out of range
-2: head out of range
-10000: head out of range
None: an integer is required
good roots:
0: [0]
1: [1]
-1: [-1]
out-of-range roots are ignored:
2: []
10000: []
-2: []
-10000: []
bad roots:
None: an integer is required
$ cd ..
Test corrupted p1/p2 fields that could cause SEGV at parsers.c:
$ mkdir invalidparent
$ cd invalidparent
$ hg clone --pull -q --config phases.publish=False ../a limit
$ hg clone --pull -q --config phases.publish=False ../a segv
$ rm -R limit/.hg/cache segv/.hg/cache
$ $PYTHON <<EOF
> data = open("limit/.hg/store/00changelog.i", "rb").read()
> for n, p in [('limit', '\0\0\0\x02'), ('segv', '\0\x01\0\0')]:
> # corrupt p1 at rev0 and p2 at rev1
> d = data[:24] + p + data[28:127 + 28] + p + data[127 + 32:]
> open(n + "/.hg/store/00changelog.i", "wb").write(d)
> EOF
$ hg debugindex -f1 limit/.hg/store/00changelog.i
rev flag offset length size base link p1 p2 nodeid
0 0000 0 63 62 0 0 2 -1 7c31755bf9b5
1 0000 63 66 65 1 1 0 2 26333235a41c
$ hg debugindex -f1 segv/.hg/store/00changelog.i
rev flag offset length size base link p1 p2 nodeid
0 0000 0 63 62 0 0 65536 -1 7c31755bf9b5
1 0000 63 66 65 1 1 0 65536 26333235a41c
$ cat <<EOF > test.py
> import sys
> from mercurial import changelog, vfs
> cl = changelog.changelog(vfs.vfs(sys.argv[1]))
> n0, n1 = cl.node(0), cl.node(1)
> ops = [
> ('reachableroots',
> lambda: cl.index.reachableroots2(0, [1], [0], False)),
> ('compute_phases_map_sets', lambda: cl.computephases([[0], []])),
> ('index_headrevs', lambda: cl.headrevs()),
> ('find_gca_candidates', lambda: cl.commonancestorsheads(n0, n1)),
> ('find_deepest', lambda: cl.ancestor(n0, n1)),
> ]
> for l, f in ops:
> print l + ':',
> try:
> f()
> print 'uncaught buffer overflow?'
> except ValueError, inst:
> print inst
> EOF
$ $PYTHON test.py limit/.hg/store
reachableroots: parent out of range
compute_phases_map_sets: parent out of range
index_headrevs: parent out of range
find_gca_candidates: parent out of range
find_deepest: parent out of range
$ $PYTHON test.py segv/.hg/store
reachableroots: parent out of range
compute_phases_map_sets: parent out of range
index_headrevs: parent out of range
find_gca_candidates: parent out of range
find_deepest: parent out of range
$ cd ..
#endif