HTTP server protocols have had for a while LWS_CALLBACK_HTTP_DROP/BIND_PROTOCOL
callbacks that mark when a wsi is attched to a protocol and detached.
It turns out this is generally useful for everything to know when a wsi is
joining a protocol and definitively completely finished with a protocol.
Particularly with client wsi where you provided the userdata externally, this
makes a clear point to free() it on the protocol binding being dropped.
This patch adds protocol bind / unbind callbacks to the role definition and
lets them operate on all roles. For the various roles
HTTP server: LWS_CALLBACK_HTTP_BIND/DROP_PROTOCOL as before
HTTP client: LWS_CALLBACK_CLIENT_HTTP_BIND/DROP_PROTOCOL
ws server: LWS_CALLBACK_WS_SERVER_BIND/DROP_PROTOCOL
ws client: LWS_CALLBACK_WS_CLIENT_BIND/DROP_PROTOCOL
raw file: LWS_CALLBACK_RAW_FILE_BIND/DROP_PROTOCOL
raw skt: LWS_CALLBACK_RAW_SKT_BIND/DROP_PROTOCOL
1) This moves the service tid detection stuff from context to pt.
2) If LWS_MAX_SMP > 1, a default pthread tid detection callback is provided
on the dummy callback. Callback handlers that call through to the dummy
handler will inherit this. It provides an int truncation of the pthread
tid.
3) If there has been any service calls on the service threads, the pts now
know the low sizeof(int) bytes of their tid. When you ask for a client
connection to be created, it looks through the pts to see if the calling
thread is a pt service thread. If so, the new client is set to use the
same pt as the caller.
This adds a plugin that interfaces to libjsongit2
https://warmcat.com/git/libjsongit2
to provide a per-vhost service for presenting bare git repos in a
web interface.
This changes the vhost destroy flow to only hand off the listen
socket if another vhost sharing it, and mark the vhost as
being_destroyed.
Each tsi calls lws_check_deferred_free() once a second, if it sees
any vhost being_destroyed there, it closes all wsi on its tsi on
the same vhost, one time.
As the wsi on the vhost complete close (ie, after libuv async close
if on libuv event loop), they decrement a reference count for all
wsi open on the vhost. The tsi who closes the last one then
completes the destroy flow for the vhost itself... it's random
which tsi completes the vhost destroy but since there are no
wsi left on the vhost, and it holds the context lock, nothing
can conflict.
The advantage of this is that owning tsi do the close for wsi
that are bound to the vhost under destruction, at a time when
they are guaranteed to be idle for service, and they do it with
both vhost and context locks owned, so no other service thread
can conflict for stuff protected by those either.
For the situation the user code may have allocations attached to
the vhost, this adds args to lws_vhost_destroy() to allow destroying
the user allocations just before the vhost is freed.
- split raw role into separate skt and file
- remove all special knowledge from the adoption
apis and migrate to core
- remove all special knowledge from client_connect
stuff, and have it discovered by iterating the
role callbacks to let those choose how to bind;
migrate to core
- retire the old deprecated client apis pre-
client_connect_info
Several new ops are planned for tls... so better to bite the bullet and
clean it out to the same level as roles + event-libs first.
Also adds a new travis target "mbedtls" and all the tests except
autobahn against mbedtls build.
This replaces the old test-app for echo with separate client and server
minimal versions.
The autobahn test script is made more autonomous and tests both
client and server.
1) Remove the whole ah rxbuf and put things on to the wsi buflist
This eliminates the whole detachability thing based on ah rxbuf
state... ah can always be detached.
2) Remove h2 scratch and put it on the wsi buflist
3) Remove preamble_rx and use the wsi buflist
This was used in the case adopted sockets had already been read.
Basically there are now only three forced service scenarios
- something in buflist (and not in state LRS_DEFERRING_ACTION)
- tls layer has buffered rx
- extension has buffered rx
This is a net removal of around 400 lines of special-casing.
This converts several of the selftests to return a status in their exit code
about whether they 'worked'.
A small bash script framework is added, with a selftest.sh in the mininmal
example dirs that support it, and a ./minimal-examples/selftests.sh script
that can be run from the build dir with no args that discovers and runs all
the selftest.sh scripts underneath.
That is also integrated into travis and the enabled tests must pass now for
travis to pass. Travis does not have a modern libuv so it can't run a
couple of tests which are nulled out if it sees it's running in travis env.
For h1 / ws, a combination of removing POLLIN wait and
stashing any unused rx lets us immediately respond to
rx flow control requests in a simple and effective way,
because the tcp socket is the stream.
But for muxed protocols like h2, that technique cannot
be used because we cannot silence the whole bundle of
streams because one can't handle any more rx dynamically.
There are control frames and content for other streams
serialized inbetween the flow controlled stream content.
We have no choice but to read to so we can see the other
things. Therefore for muxed protocols like h2 and spdy,
rx flow control boils down to tx credit manipulation
on individual streams to staunch the flow at the peer.
However that requires a round trip to take effect, any
transmitted packets that were in flight before the tx credit
reduction arrives at the remote peer are still going to come
and have to be dealt with by adding them to the stash.
This patch introduces lws_buflist scatter-gather type
buffer management for rxflow handling, so we can append
buffer segments in a linked-list to handle whatever rx
is unavoidably in flight on a stream that is trying to
assert rx flow control.
Since new roles may be incompatible with http, add support for
alpn names at the role struct, automatic generation of the
default list of alpn names that servers advertise, and the
ability to override the used alpn names per-vhost and per-
client connection.
This not only lets you modulate visibility or use of h2,
but also enables vhosts that only offer non-http roles,
as well as restricting http role vhosts to only alpn
identifiers related to http roles.
This adds h2 http support for the client api.
The public client api requires no changes, it will detect by
ALPN if the server can handle http/2, if so, it will use it.
Multiple client connections using the lws api will be mapped on
to the same single http/2 + tls socket using http/2 streams
that are serviced simultaneously where possible.
This completely removes the loop self-running stuff.
Static allocations (uv_idle, timers etc) are referenced-counted in the context
same as the wsi are. When lws wants to close, he first closes all his wsi, then
when that is completed in the uv close callbacks, he closes all of his static
uv handles. When that is also completed in the uv callbacks, he stops the loop
so the lws context can destroy and exit.
Any direct libuv allocations in protocol handlers must participate in the
reference counting. Two new apis are provided
- lws_libuv_static_refcount_add(handle, context) to mark the handle with
a pointer to the context and increment the global uv object counter
- lws_libuv_static_refcount_del() which should be used as the close callback
for your own libuv objects declared in the protocol scope.
