84c90e223f
This commit introduces new APIs for applications and signal handling, intended to replace isc_app for applications built on top of libisc. * isc_app will be replaced with isc_loopmgr, which handles the starting and stopping of applications. In isc_loopmgr, the main thread is not blocked, but is part of the working thread set. The loop manager will start a number of threads, each with a uv_loop event loop running. Setup and teardown functions can be assigned which will run when the loop starts and stops, and jobs can be scheduled to run in the meantime. When isc_loopmgr_shutdown() is run from any the loops, all loops will shut down and the application can terminate. * signal handling will now be handled with a separate isc_signal unit. isc_loopmgr only handles SIGTERM and SIGINT for application termination, but the application may install additional signal handlers, such as SIGHUP as a signal to reload configuration. * new job running primitives, isc_job and isc_async, have been added. Both units schedule callbacks (specifying a callback function and argument) on an event loop. The difference is that isc_job unit is unlocked and not thread-safe, so it can be used to efficiently run jobs in the same thread, while isc_async is thread-safe and uses locking, so it can be used to pass jobs from one thread to another. * isc_tid will be used to track the thread ID in isc_loop worker threads. * unit tests have been added for the new APIs.
4.4 KiB
Loop Manager
This document aims to describe the design of the basic event loop handling in the BIND 9.
Every application is expected to create and use a single isc_loopmgr_t
instance, but the isc_loopmgr API itself doesn't enforce this requirement.
Event Loops
The loop manager creates N event loops (of type isc_loop_t), where N is
specified by the caller when creating the loop manager. The number of event
loops is usually same as the number of logical CPUs. The minimum N is 1, and
maximum is limited only by the machine resources.
For each event loop, a thread is created by the loop manager. The
isc_loop_t object itself is built on top of uv_loop_t. uv_loop_t
is not thread-safe, and this is also true for isc_loop_t. If you need to
run an event on a different event loop, see below for the isc_async API.
The application can get a reference to the current event loop using
isc_loop_current(), to the main event loop (loop 0) that will always exist
using isc_loop_main() and to an arbitrary event loop using
isc_loop_get().
Application Start and Stop
Every application MUST add its initial events using isc_loopmgr_setup() to
be run on all initialized event loops or isc_loop_setup() to be run on a
selected event loop.
Applications MAY also add events to be run when the application is shut down by
calling isc_loopmgr_teardown() (or isc_loop_teardown() for a specific
event loop).
After the setup and teardown events have been configured, the application may
be started via isc_loopmgr_run(). isc_loopmgr_run() will block for the
caller while event loops are running. When the work is done,
isc_loopmgr_shutdown() must be run from within one of the event loops; this
will cause all loops to be shut down and isc_loopmgr_run() to return.
The most notable change from the isc_app API is the lack of a blocked
main thread. The loop manager starts the main event loop on the
main thread when the application is started.
This API now replaces the old isc_app API.
Signal Handling
The loop manager itself takes care of handling the SIGTERM and SIGINT
signals, but the application MAY add more handlers via isc_signal API. In
named, for example, SIGHUP is used to trigger an application reload.
Event scheduling
The application may add events to the event loop via isc_job_run() for jobs
on the same event loop, or via isc_async_run() for jobs to be passed to
other event loops. Both functions take the event loop, the callback and the
callback argument as parameters.
Generally isc_job_run() is more direct, as it schedules the event directly
on the event loop and doesn't use locking, and should be preferred unless you
need to run the event on a different thread.
isc_async_run() is the only new thread-safe function provided by the loop
manager, uses locked list to collect new jobs and uv_async() primitive to
enqueue the collected jobs onto the event loop.
Tasks
The isc_task API has been modified to run the tasks directly on the loop
manager. The new isc_job and isc_async APIs are preferred for simple
events; the isc_task API is provided for backward-compatibility purposes
and thus is also thread-safe because it uses locking and uv_async() to enqueue
events onto the event loop.
Timers
The isc_timer API is now built on top of the uv_timer_t object. It has
been changed to support only ticker and once timers, and now uses
isc_timer_start() and isc_timer_stop() instead of changing the timer
type to inactive. The isc_timer_t object is not thread-safe.
Network Manager
The network manager has been changed to use the loop manager event loops instead of managing its own event loops.
The new network manager calls are not thread-safe; all connect/read/write functions MUST be called from the thread that created the network manager socket.
The isc_nm_listen*() functions MUST be called from the main loop.