Andy Green [Sat, 2 Feb 2013 15:02:56 +0000 (23:02 +0800)]
improve test server poll loop docs
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 2 Feb 2013 10:10:29 +0000 (18:10 +0800)]
add note about MIPS opewrt configure options
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 1 Feb 2013 03:17:02 +0000 (11:17 +0800)]
remove one more mention of broadcast callback
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 1 Feb 2013 02:50:15 +0000 (10:50 +0800)]
introduce library version plus git hash
This exposes the library version and git head hash it was built from
into LWS_LIBRARY_VERSION and LWS_BUILD_HASH.
These are combined into a version string that's both printed as a
notice log by the library and made available to the app using a new
api lws_get_library_version(). The version looks like
1.1 178d78c
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 1 Feb 2013 00:48:41 +0000 (08:48 +0800)]
remove stray reference to max broadcast size from readme.build
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 1 Feb 2013 00:42:15 +0000 (08:42 +0800)]
additional casts allow test server build as cpp
With these explicit casts that are not needed in C, it's possible to build
the test server using g++ like this, after building and installing the
library.
g++ -DINSTALL_DATADIR=\"/usr/share\" -ocpptest test.cpp -lwebsockets
Add a small documentation to README.coding
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 11:53:05 +0000 (19:53 +0800)]
update changelog tag chrome 26 firefox 18
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 11:50:57 +0000 (19:50 +0800)]
renovate test html
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 02:16:44 +0000 (10:16 +0800)]
add changelog v1.0 to v1.1
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 02:12:52 +0000 (10:12 +0800)]
bump version to 1.1 and soname to 2
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 02:34:52 +0000 (10:34 +0800)]
clean out remaining mentions of deprecated broadcast
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 02:05:43 +0000 (10:05 +0800)]
get error from getnameinfo if unable to improve hostname and use hostname
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 31 Jan 2013 01:57:05 +0000 (09:57 +0800)]
fixes for without server and without client
Signed-off-by: Andy Green <andy.green@linaro.org>
Jack Mitchell [Wed, 30 Jan 2013 10:53:36 +0000 (18:53 +0800)]
fix unused var if no enable openssl
Signed-off-by: Jack Mitchell <ml@communistcode.co.uk>
Andy Green [Wed, 30 Jan 2013 03:03:01 +0000 (11:03 +0800)]
introduce test echo
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 30 Jan 2013 04:28:34 +0000 (12:28 +0800)]
add info about why we close to more places
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 30 Jan 2013 04:27:27 +0000 (12:27 +0800)]
roubustness handle problems in read loop better
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 30 Jan 2013 04:26:14 +0000 (12:26 +0800)]
server allow NULL protocol map to protocol 0
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 30 Jan 2013 00:12:20 +0000 (08:12 +0800)]
change get_peer_addresses to use context wsi latency
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 29 Jan 2013 04:37:35 +0000 (12:37 +0800)]
instrument latency
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 29 Jan 2013 04:36:17 +0000 (12:36 +0800)]
introduce lws_latency
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 30 Jan 2013 00:31:43 +0000 (08:31 +0800)]
fix docs about protocol version supported
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 30 Jan 2013 00:02:26 +0000 (08:02 +0800)]
test server terminate cleanly on ctrl c
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 29 Jan 2013 09:57:39 +0000 (17:57 +0800)]
evict all broadcast support
Libwebsockets is fundamentally singlethreaded... the existence of the
fork and broadcast support, especially in the sample server is
giving the wrong idea about how to use it.
This replaces broadcast in the sample server with
libwebsocket_callback_on_writable_all_protocol(). The whole idea of
'broadcast' is removed.
All of the broadcast proxy stuff is removed: data must now be sent
from the callback only. Doing othherwise is not reliable since the
service loop may close the socket and free the wsi at any time,
invalidating a wsi pointer held by another thread (don't do that!)
Likewise the confirm_legit_wsi api added recently does not help the
other thread case, since if the wsi has been freed dereferencing the
wsi to study if it is legit or not will segfault in that case. So
this is removed too.
The overall effect is to push user code to only operate inside the
protocol callbacks or external poll loops, ie, single thread context.
Signed-off-by: Andy Green <andy.green@linaro.org>
Matthieu Riviere [Mon, 28 Jan 2013 22:57:21 +0000 (06:57 +0800)]
FreeBSD compatibility
Seems we just need to take care about __FreeBSD__ along with
__APPLE__ in a couple of places.
Signed-off-by: Matthieu Riviere <matthieu.riviere@leukensis.org>
Andy Green [Mon, 28 Jan 2013 22:52:00 +0000 (06:52 +0800)]
trac 6 expose libwebsockets read with note about not normally needed
Comes in handy if the original application poll loop is the boss,
in this case libwebsockets is optional and can't be the boss poll
loop
Requested-by: ajandhyala@wms.com
Signed-off-by: Andy Green <andy.green@linaro.org>
Edwin van den Oetelaar [Mon, 28 Jan 2013 13:53:53 +0000 (21:53 +0800)]
fix ssl bits outside of ifdef coverage
Signed-off-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com>
Andy Green [Mon, 28 Jan 2013 09:48:21 +0000 (17:48 +0800)]
force client ssl bio nonblocking
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 28 Jan 2013 09:45:34 +0000 (17:45 +0800)]
force ssl rw bios nonblocking
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 28 Jan 2013 09:20:41 +0000 (17:20 +0800)]
timeout coverage for ssl accept
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 28 Jan 2013 04:19:10 +0000 (12:19 +0800)]
break up ssl accept action
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 28 Jan 2013 03:23:52 +0000 (11:23 +0800)]
ensure accept is nonblocking
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 25 Jan 2013 09:34:15 +0000 (17:34 +0800)]
autocreate foreign broadcast sockets on broadcast
Also introduce libwebsockets_broadcast_foreign() separate from libwebsockets_broadcast()
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 23 Jan 2013 00:32:03 +0000 (08:32 +0800)]
trac 5 sa_restorer deprecated
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 22 Jan 2013 04:40:35 +0000 (12:40 +0800)]
trac 3 document write and context_user
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 22 Jan 2013 04:32:11 +0000 (12:32 +0800)]
trac 4 correct libebsocket_service_fd
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 05:20:33 +0000 (13:20 +0800)]
update numbers for minimal build footprint
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 05:06:38 +0000 (13:06 +0800)]
more LWS_NO_DAEMONIZE
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 04:58:04 +0000 (12:58 +0800)]
use correct LWS_NO_DAEMONIZE on test server
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 04:19:30 +0000 (12:19 +0800)]
bind gcc debug generation to_DEBUG
Either generate debug sections in output or optimize aggressively,
controlled by the existing --disable-debug
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 03:04:23 +0000 (11:04 +0800)]
unionize mutually exclusive wsi members
Large chunks of struct libwebsocket members actually have a mutually
exclusive lifecycle, eg, once the http headers are finished they sit
there unused until the instance is destroyed.
This makes a big improvement in memory efficiency by making four
categories of member: always needed, needed for header processing,
needed for http processing, and needed for ws processing. The last
three are mutually exclusive and bound into a union inside the wsi.
Care needs taking now at "union transitions", although we zeroed down
the struct at init, the other union siblings have been writing the
same memory by the time later member siblings start to use it. So
it must be cleared down appropriately when we cross from one
mutually-exclusive use to another.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 03:04:49 +0000 (11:04 +0800)]
key_b64 doesnt need to be in wsi
It's only used in some client code and never seen again
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 02:36:12 +0000 (10:36 +0800)]
avoid PATH_MAX in bss in daemonize
PATH_MAX is typically 4KB, let's malloc space for the
actual path instead and just have the pointer in bss
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 01:53:35 +0000 (09:53 +0800)]
remove all support for pre v13 protocols
Since v13 was defined as the released ietf version the older versions
are deprecated. This patch strips out everything to do with the older
versions and gets rid of the option to send stuff unmasked.
The in-tree md5 implementation is then also deleted as nothing needs
it any more, 1280 loc are shed in all
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Mon, 21 Jan 2013 01:09:52 +0000 (09:09 +0800)]
disable private broadcast sockets if enable no fork config option
The whole thing about count_protocols + 1 broadcast sockets and
associated dummy wsis is a workaround for getting a broadcast from
a different process context, if we are running with --enable-no-fork
then we don't need any of it in.
Signed-off-by: Andy Green <andy.green@linaro.org>
Edwin van den Oetelaar [Sun, 20 Jan 2013 12:51:14 +0000 (20:51 +0800)]
align test server extpoll with library dynamic approach
Signed-off-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com>
Andy Green [Sun, 20 Jan 2013 12:14:42 +0000 (20:14 +0800)]
document memory efficiency
This puts some numbers of library size with the various --without
and --disable options and about dynamic memory allocation
performance
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sun, 20 Jan 2013 12:21:54 +0000 (20:21 +0800)]
log major dynamic allocation info
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sun, 20 Jan 2013 10:26:20 +0000 (18:26 +0800)]
zlib not needed if no extensions
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sun, 20 Jan 2013 09:08:31 +0000 (17:08 +0800)]
introduce without extensions
The new --without-extensions config flag completely removes all code
and data related to extensions from the build throughout the library
when given.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sun, 20 Jan 2013 03:28:06 +0000 (11:28 +0800)]
refactor README
Signed-off-by: Andy Green <andy.green@linaro.org>
Edwin van den Oetelaar [Sat, 19 Jan 2013 12:01:01 +0000 (20:01 +0800)]
syslog requires format string
Signed-off-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com>
Andy Green [Sat, 19 Jan 2013 07:11:23 +0000 (15:11 +0800)]
revert zlib update 1.2.7
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 05:56:10 +0000 (13:56 +0800)]
make use of lock file
Signed-off-by: Andy Green <andy.green@linaro.org>
FNA [Sat, 19 Jan 2013 05:58:06 +0000 (13:58 +0800)]
windows compatibility changes for private libwebsockets
Signed-off-by: FNA <aqiruse@gmail.com>
Andy Green [Sat, 19 Jan 2013 03:11:42 +0000 (11:11 +0800)]
test server add daemonization flag
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 03:32:18 +0000 (11:32 +0800)]
test server use syslog logging
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 03:58:07 +0000 (11:58 +0800)]
allow_use_of_lwsl_logging in user code
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 03:12:16 +0000 (11:12 +0800)]
helper api: log through syslog
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 05:08:17 +0000 (13:08 +0800)]
logging select some lwsl_info usage to be lwsl_notice
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 04:18:07 +0000 (12:18 +0800)]
add lwsl_notice
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 03:17:56 +0000 (11:17 +0800)]
expose log level in emit
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 03:52:06 +0000 (11:52 +0800)]
change bitfield setting to avoid gcc warnings
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Sat, 19 Jan 2013 02:39:35 +0000 (10:39 +0800)]
improve frame_is_binary setting
Gregory Junker <ggjunker@gmail.com> noticed the binary flag is not
getting set right, or at all on client side. This should improve
matters.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 18 Jan 2013 14:00:22 +0000 (22:00 +0800)]
add lws_confirm_legit_wsi
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 18 Jan 2013 03:43:21 +0000 (11:43 +0800)]
refactor and introduce without server configure option
Move server-only stuff into their own files and make building
that depend on not having --without-server on the configure
Make fragments in other places conditional as well
Remove client-related members from struct libwebscket when
building LWS_NO_CLIENT
Apps:
normal: build test server, client, fraggle, ping
--without-client: build test server
--without-server: build test client, ping
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Fri, 18 Jan 2013 01:49:20 +0000 (09:49 +0800)]
introdice tracking if frame is binary
Signed-off-by: Andy Green <andy.green@linaro.org>
Edwin van den Oetelaar [Fri, 18 Jan 2013 01:20:54 +0000 (09:20 +0800)]
different compiler warning fixes
Solve some unchecked return codes in the new daemonization file
AG adapted a bit
Signed-off-by: Edwin van den Oetelaar <oetelaar.automatisering@gmail.com>
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 17:55:48 +0000 (01:55 +0800)]
add lexical parser for headers
Profiling what happens during the ab test, one of the hotspots
was strcasecmp in a loop looking for header name matches each time.
This patch introduces a lexical parser that creates a state machine
in 276 bytes that encodes all the known header names. The fsm is
walked bytewise as chaacters come in... most states do not need any
recursion to match or fail.
The state machine output is cut-and-pasted into parsers.c as an
unsigned char array.
The fsm generator is a bit rough and ready, included in the tree but
not built since normal mortals won't need to touch it.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 11:53:16 +0000 (19:53 +0800)]
deprecate x google mux
Unfortunately this code is beginning to rot due to lack of demand to
provide it and it being disabled by default.
If demand appears we can revert this and resume work on it, otherwise
let's bite the bullet for the moment.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 08:50:35 +0000 (16:50 +0800)]
solve flowcontrol problems
Problems with rx flow control implementation were the underlying cause
of the connection stalling issue that was covered up with the udelay()
patch that was removed recently.
This get rx flow control working properly and corrects problems with
fifo management in the test server mirror protocol code too.
The rxfow control api has been changed to just set a flag, so it's very cheap
to call from user code. After the callbacks that might use the rxflow control
api the flag is checked and any pending actions done.
rx flow control now stops any rx packet coming immediately, with compessed
connections "just what was left in the pipe" might be hundreds of KBytes. To
implement that the current packet being decoded is copied into a malloc'd buffer
by the rx processing code now.
When rxflow is allows to come again, the buffer is drained and freed before any
new packet content is accepted.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 06:46:43 +0000 (14:46 +0800)]
check errors on shutdown close
Also make sure CLOSE doesn't go through extension munging
Reduce wait for close ack to 1s
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 04:26:48 +0000 (12:26 +0800)]
replace hashtable polltable management
This rips out the connection hashtable implementation along with
MAX_CLIENTS and replaces it with a dynamically allocated fds array
and lookup table along the same lines as the new extpoll implementation
from Edwin van den Oetelaar.
It detects the max number of file descriptors possible at context init
time and allocates accordingly; this can be externally controlled by
ulimit and the server run as a specific user to facilitate targeting
specific ulimit rules at it.
Many operations that translated between socket descriptors and struct
websocket or pollfd objects have had iteration removed by this patch
and under load will be a lot faster.
Signed-off-by: Andy Green <andy.green@linaro.org>
Edwin van der Oetelaar [Thu, 17 Jan 2013 03:16:15 +0000 (11:16 +0800)]
use simple lookup table for extpoll
Hash stuff is overkill since Edwin found a max connection limit of 30000 on his
box anyway. Just use a simple preallocated lookup table and fds array.
AG Modified for style and removed debugging bits
Signed-off-by: Edwin van der Oetelaar <oetelaar.automatisering@gmail.com>
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 07:49:27 +0000 (15:49 +0800)]
include daemonization file whoops
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 07:02:02 +0000 (15:02 +0800)]
portability dont assume size of tv.t_usec
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Thu, 17 Jan 2013 02:15:52 +0000 (10:15 +0800)]
add disable debug to README configuration options list
Signed-off-by: Andy Green <andy.green@linaro.org>
Peter Young [Thu, 17 Jan 2013 02:10:10 +0000 (10:10 +0800)]
including assert h needed on osx
Signed-off-by: Peter Young <young40@qq.com>
Peter Young [Thu, 17 Jan 2013 02:08:16 +0000 (10:08 +0800)]
just use limits.h directly
Signed-off-by: Peter Young <young40@qq.com>
Andy Green [Thu, 17 Jan 2013 02:05:39 +0000 (10:05 +0800)]
introduce daemonize
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 16:50:48 +0000 (00:50 +0800)]
client allow remote server to accept with no protocol set
Needed if our clientside can connect to, eg, echo.websocket.org
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 09:46:00 +0000 (17:46 +0800)]
move array bounds gcc workaround outside function
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 06:35:12 +0000 (14:35 +0800)]
update ping test client and stop exposing payload to extensions
Ping and Pong payload in control messages need to be
above the fray of extension payload munging
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 06:35:27 +0000 (14:35 +0800)]
export lswl_hexdump
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 05:40:43 +0000 (13:40 +0800)]
roubustness only return 0 from fd service if handled
Otherwise our app outer loop will think the world is ending
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 03:47:40 +0000 (11:47 +0800)]
configure without client
This leverages the refactor patches to introduce the ability to
disable building any client side code in the library or the client
side test apps.
This will be a considerable size saving for embedded server-only
case.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 04:32:20 +0000 (12:32 +0800)]
test client remove usleep
Whatever caused the need for this has gone away
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 04:21:29 +0000 (12:21 +0800)]
refactor output.c
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 03:53:05 +0000 (11:53 +0800)]
refactor migrate client stuff to client.c
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 03:45:38 +0000 (11:45 +0800)]
add new context arg to libwebsockets_serve_http_file
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 03:43:53 +0000 (11:43 +0800)]
robustness server dont exit server on accept problems
Error message corrected to %s as well
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 02:21:34 +0000 (10:21 +0800)]
workaround for some gcc array bounds false positive
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 02:11:34 +0000 (10:11 +0800)]
add logo to test file
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 02:06:28 +0000 (10:06 +0800)]
update test server html serving callback to use aepd whitelist approach
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 02:00:39 +0000 (10:00 +0800)]
add libwebsockets.org logo to share
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Wed, 16 Jan 2013 00:37:48 +0000 (08:37 +0800)]
optimize http file sending
This adapts the approach from the single-packet-per-poll-loop improvement
to sending more packets while the socket can take them.
It still falls back to the multi-state scheme if the socket ever chokes,
which it certainly will on larger files, so it's safe while being highly
efficient at smaller file sizes.
Nor should it significantly add to latency for other sockets, it simply
stuffs the pipe asynchronously as much as the pipe can take.
We also increase the packet payoad size from 512 to 1400 a time.
This reduces the time taken in the 300 connection / 5000 transfers ab test
from >8s to ~3.4s, transferring the same amount of data.
$ ab -t 100 -n 5000 -c 300 'http://127.0.0.1:7681/'
This is ApacheBench, Version 2.3 <$Revision: 1373084 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Benchmarking 127.0.0.1 (be patient)
Completed 500 requests
Completed 1000 requests
Completed 1500 requests
Completed 2000 requests
Completed 2500 requests
Completed 3000 requests
Completed 3500 requests
Completed 4000 requests
Completed 4500 requests
Completed 5000 requests
Finished 5000 requests
Server Software: libwebsockets
Server Hostname: 127.0.0.1
Server Port: 7681
Document Path: /
Document Length: 8447 bytes
Concurrency Level: 300
Time taken for tests: 3.400 seconds
Complete requests: 5000
Failed requests: 0
Write errors: 0
Total transferred:
42680000 bytes
HTML transferred:
42235000 bytes
Requests per second: 1470.76 [#/sec] (mean)
Time per request: 203.976 [ms] (mean)
Time per request: 0.680 [ms] (mean, across all concurrent requests)
Transfer rate: 12260.17 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 7 24 15.6 20 125
Processing: 32 172 50.2 161 407
Waiting: 27 154 49.4 142 386
Total: 81 196 48.3 182 428
Percentage of the requests served within a certain time (ms)
50% 182
66% 185
75% 188
80% 194
90% 304
95% 316
98% 322
99% 328
100% 428 (longest request)
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 15 Jan 2013 23:59:47 +0000 (07:59 +0800)]
listen socket more frequent service
From an idea by Edwin van den Oetelaar <oetelaar.automatisering@gmail.com>
When testing libwebsockets with ab, Edwin found an unexpected bump in
the distribution of latencies, some connections were held back almost
the whole test duration.
http://ml.libwebsockets.org/pipermail/libwebsockets/2013-January/000006.html
Studying the problem revealed that when there are mass pending connections
amongst many active connections, we do not service the listen socket often
enough to clear the backlog, some seem to get stale violating FIFO ordering.
This patch introduces listen socket service "piggybacking", where every n
normal socket service actions we also check the listen socket and deal with
pending connections there.
Normally, it checks the listen socket gratuitously every 10 normal socket
services. However, if it finds something waiting, it forces a check on the
next normal socket service too by keeping stats on how often something was
waiting. If the probability of something waiting each time becomes high,
it will allow up to two waiting connections to be serviced for each normal
socket service.
In that way it has low burden in the normal case, but rapidly adapts by
detecting mass connection loads as found in ab.
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 15 Jan 2013 13:56:12 +0000 (21:56 +0800)]
add empty m4 dir as workaround for autoreconf issue
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 15 Jan 2013 12:52:29 +0000 (20:52 +0800)]
allow LWS_SOMAXCONN to be defined at configuretime
Default remains at SOMAXCONN, you can force it at configure time
along these lines
./configure CFLAGS="-DLWS_SOMAXCONN=16384"
Signed-off-by: Andy Green <andy.green@linaro.org>
Andy Green [Tue, 15 Jan 2013 11:44:33 +0000 (19:44 +0800)]
extpoll use hashtable for fd tracking
This implements a much faster, hashtable-based tracking scheme for
external poll fds.
Signed-off-by: Andy Green <andy.green@linaro.org>