48366de1d1
An lws context usually contains a processwide fd -> wsi lookup table. This allows any possible fd returned by a *nix type OS to be immediately converted to a wsi just by indexing an array of struct lws * the size of the highest possible fd, as found by ulimit -n or similar. This works modestly for Linux type systems where the default ulimit -n for a process is 1024, it means a 4KB or 8KB lookup table for 32-bit or 64-bit systems. However in the case your lws usage is much simpler, like one outgoing client connection and no serving, this represents increasing waste. It's made much worse if the system has a much larger default ulimit -n, eg 1M, the table is occupying 4MB or 8MB, of which you will only use one. Even so, because lws can't be sure the OS won't return a socket fd at any number up to (ulimit -n - 1), it has to allocate the whole lookup table at the moment. This patch looks to see if the context creation info is setting info->fd_limit_per_thread... if it leaves it at the default 0, then everything is as it was before this patch. However if finds that (info->fd_limit_per_thread * actual_number_of_service_threads) where the default number of service threads is 1, is less than the fd limit set by ulimit -n, lws switches to a slower lookup table scheme, which only allocates the requested number of slots. Lookups happen then by iterating the table and comparing rather than indexing the array directly, which is obviously somewhat of a performance hit. However in the case where you know lws will only have a very few wsi maximum, this method can very usefully trade off speed to be able to avoid the allocation sized by ulimit -n. minimal examples for client that can make use of this are also modified by this patch to use the smaller context allocations. |
||
|---|---|---|
| .. | ||
| api-tests | ||
| client-server | ||
| crypto | ||
| dbus-client | ||
| dbus-server | ||
| http-client | ||
| http-server | ||
| raw | ||
| ws-client | ||
| ws-server | ||
| README.md | ||
| selftests-library.sh | ||
| selftests.sh | ||
| name | demonstrates |
|---|---|
| client-server | Minimal examples providing client and server connections simultaneously |
| crypto | Minimal examples related to using lws crypto apis |
| dbus-server | Minimal examples showing how to integrate DBUS into lws event loop |
| http-client | Minimal examples providing an http client |
| http-server | Minimal examples providing an http server |
| raw | Minimal examples related to adopting raw file or socket descriptors into the event loop |
| ws-client | Minimal examples providing a ws client |
| ws-server | Minimal examples providing a ws server (and an http server) |
FAQ
Getting started
Build and install lws itself first (note that after installing lws on *nix, you need to run ldconfig one time so the OS can learn about the new library. Lws installs in /usr/local by default, Debian / Ubuntu ldconfig knows to look there already, but Fedora / CentOS need you to add the line /usr/local/lib to /etc/ld.so.conf and run ldconfig)
Then start with the simplest:
http-server/minimal-http-server
Why are most of the sources split into a main C file file and a protocol file?
Lws supports three ways to implement the protocol callback code:
-
you can just add it all in the same source file
-
you can separate it as these examples do, and #include it into the main sources
-
you can build it as a standalone plugin that is discovered and loaded at runtime.
The way these examples are structured, you can easily also build the protocol callback as a plugin just with a different CMakeLists.txt... see https://github.com/warmcat/libwebsockets/tree/master/plugin-standalone for an example.
Why would we want the protocol as a plugin?
You will notice a lot of the main C code is the same boilerplate repeated for each example. The actual interesting part is in the protocol callback only.
Lws provides (-DLWS_WITH_LWSWS=1) a generic lightweight server app called 'lwsws' that can be configured by JSON. Combined with your protocol as a plugin, it means you don't actually have to make a special server "app" part, you can just use lwsws and pass per-vhost configuration from JSON into your protocol. (Of course in some cases you have an existing app you are bolting lws on to, then you don't care about this for that particular case).
Because lwsws has no dependency on whatever your plugin does, it can mix and match different protocols randomly without needing any code changes. It reduces the size of the task to just writing the code you care about in your protocol handler, and nothing else to write or maintain.
Lwsws supports advanced features like reload, where it starts a new server instance with changed config or different plugins, while keeping the old instance around until the last connection to it closes.
I get why there is a pss, but why is there a vhd?
The pss is instantiated per-connection. But there are almost always other variables that have a lifetime longer than a single connection.
You could make these variables "filescope" one-time globals, but that means your protocol cannot instantiate multiple times.
Lws supports vhosts (virtual hosts), for example both https://warmcat.com and https://libwebsockets are running on the same lwsws instance on the same server and same IP... each of these is a separate vhost.
Your protocol may be enabled on multiple vhosts, each of these vhosts provides a different vhd specific to the protocol instance on that vhost. For example many of the samples keep a linked-list head to a list of live pss in the vhd... that means it's cleanly a list of pss opened on that vhost. If another vhost has the protocol enabled, connections to that will point to a different vhd, and the linked-list head on that vhd will only list connections to his vhost.
The example "ws-server/minimal-ws-server-threads" demonstrates how to deliver external configuration data to a specific vhost + protocol combination using code. In lwsws, this is simply a matter of setting the desired JSON config.