| Commit message (Collapse) | Author | Age |
|---|
| | |
|
| |
|
|
| |
This is done for kqueue and poll. Others coming soon.
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
This change moves the posix pollers to inline the PFD and makes
the callbacks constant, so that we can dispense with tests, failures,
and locks. It is anticipated that this will reduce lock based
pressure on the bus and increase performance modestly.
|
| |
|
|
|
|
|
|
|
|
|
| |
We preallocate the arrays used for pollfds, based on what the
system can tolerate (tunable with NNG_MAX_OPEN), and we change
the code for inserting and removing pollfds from the list so
that it can run without acquiring the locks during the main loop,
only when adding or removing files.
The poll() implementation is very nearly lock free in the hot
code path, and soon will be.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The poller selection in the previous poller changes for select were
not quite functional. Also, while testing poll() based poller, there
were problems where it simply did not work correctly, so this addresses
those, and it seems to work now.
The pfd structures are exposed as we intend to allow inlining them
to eliminate the separate allocation and potential for failure during
initialization. We also want to have plans afoot to eliminate a
lot of the extra locking done done on each I/O iteration, and this
is setting the foundation for that.
|
| |
|
|
|
|
|
|
|
| |
Applications must now call nng_init(), but they can supply
a set of parameters optionally. The code is now safe for
multiple libraries to do this concurrently, meaning nng_fini
no longer can race against another instance starting up.
The nni_init checks on all public APIs are removed now.
|
| |
|
|
|
|
| |
This also exposes an nng_thread_set_name() function for
applications to use. All NNG thread names start with "nng:".
Note that support is highly dependent on the operating system.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
fixes #179 DNS resolution should be done at connect time
fixes #586 Windows IO completion port work could be better
fixes #339 Windows iocp could use synchronous completions
fixes #280 TCP abstraction improvements
This is a rather monstrous set of changes, which refactors TCP, and
the underlying Windows I/O completion path logic, in order to obtain
a cleaner, simpler API, with support for asynchronous DNS lookups performed
on connect rather than initialization time, the ability to have multiple
connects or accepts pending, as well as fewer extraneous function calls.
The Windows code also benefits from greatly reduced context switching,
fewer lock operations performed, and a reduced number of system calls
on the hot code path. (We use automatic event resetting instead of manual.)
Some dead code was removed as well, and a few potential edge case leaks
on failure paths (in the websocket code) were plugged.
Note that all TCP based transports benefit from this work. The IPC code
on Windows still uses the legacy IOCP for now, as does the UDP code (used
for ZeroTier.) We will be converting those soon too.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
fixes #326 consider nni_taskq_exec_synch()
fixes #410 kqueue implementation could be smarter
fixes #411 epoll_implementation could be smarter
fixes #426 synchronous completion can lead to panic
fixes #421 pipe close race condition/duplicate destroy
This is a major refactoring of two significant parts of the code base,
which are closely interrelated.
First the aio and taskq framework have undergone a number of simplifications,
and improvements. We have ditched a few parts of the internal API (for
example tasks no longer support cancellation) that weren't terribly useful
but added a lot of complexity, and we've made aio_schedule something that
now checks for cancellation or other "premature" completions. The
aio framework now uses the tasks more tightly, so that aio wait can
devolve into just nni_task_wait(). We did have to add a "task_prep()"
step to prevent race conditions.
Second, the entire POSIX poller framework has been simplified, and made
more robust, and more scalable. There were some fairly inherent race
conditions around the shutdown/close code, where we *thought* we were
synchronizing against the other thread, but weren't doing so adequately.
With a cleaner design, we've been able to tighten up the implementation
to remove these race conditions, while substantially reducing the chance
for lock contention, thereby improving scalability. The illumos poller
also got a performance boost by polling for multiple events.
In highly "busy" systems, we expect to see vast reductions in lock
contention, and therefore greater scalability, in addition to overall
improved reliability.
One area where we currently can do better is that there is still only
a single poller thread run. Scaling this out is a task that has to be done
differently for each poller, and carefuly to ensure that close conditions
are safe on all pollers, and that no chance for deadlock/livelock waiting
for pfd finalizers can occur.
|
| |
|
|
|
|
|
|
| |
This replaces the epoll support with proper illumos/SunOS port
events. The port event support is structured so that it actually
is superior to epoll and kqueue, because it avoids a single master
lock on the poller. In the future we will explore this for macOS
and Linux pollers.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
It was possible for pollq arm to be called on a node that was removed
in some circumstances -- particularly and ep that was closed in the
callback.
While here, lets use normal booleans for closed state, and only call
the arm function (which is not free -- typicall it involves a mutex
and may even involve a system call) if we are going to arm some events.
We also initialize these things properly, and clean up a stale comment.
This work is done to faciliate the kqueue work by @liamstask.
|
| |
|
|
|
|
|
|
|
|
| |
This change is being made to facilitate the work done for the
kqueue port. We have created two new functions, nni_posix_pollq_init
and nni_posix_pollq_fini, which can be used when creating or destroying
the pollq nodes. Then nodes are *added* and *removed* from the pollq
structure with nni_posix_pollq_add and nni_posix_pollq_remove. The
add function in particular MUST NEVER be called unless the node has
a valid file descriptor.
|
| |
|
|
|
|
|
| |
We enabled verbose compiler warnings, and found a lot of issues.
Some of these were even real bugs. As a bonus, we actually save
some initialization steps in the compat layer, and avoid passing
some variables we don't need.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
We introduced richer, deeper tests for UDP functionality.
These tests uncovered a number of issues which this commit fixes.
The Windows IOCP code needs to support multiple aios on a single
nni_win_event. A redesign of the IOCP handling addresses that.
The POSIX UDP code also needed fixes; foremost among them is the
fact that the UDP file descriptor is not placed into non-blocking
mode, leading to potential hangs.
A number of race conditions and bugs along the implementation of
the above items were uncovered and fixed. To the best of our knowledge
the current code is bug-free.
|
| |
|
|
|
|
|
|
|
|
| |
We only compile files that are appropriate for the platform. (We
still have guards in place, to allow for a future single .C file
to be built from all the sources.) We also remove the subsystem defines;
if a new platform needs to deviate from POSIX in ways beyond what we
intended here, then that platform should just copy those parts into
a new platform directory, rather than cross including portions from
POSIX.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If the underlying platform fails (FreeBSD is the only one I'm aware
of that does this!), we use a global lock or condition variable instead.
This means that our lock initializers never ever fail.
Probably we could eliminate most of this for Linux and Darwin, since
on those platforms, mutex and condvar initialization reasonably never
fails. Initial benchmarks show little difference either way -- so we
can revisit (optimize) later.
This removes a lot of otherwise untested code in error cases and so forth,
improving coverage and resilience in the face of allocation failures.
Platforms other than POSIX should follow a similar pattern if they need
this. (VxWorks, I'm thinking of you.) Most sane platforms won't have
an issue here, since normally these initializations do not need to allocate
memory. (Reportedly, even FreeBSD has plans to "fix" this in libthr2.)
While here, some bugs were fixed in initialization & teardown.
The fallback code is properly tested with dedicated test cases.
|
| |
|
|
|
| |
This creates a use-after-free bug if nni_fini() is run, then new
sockets are created.
|
| | |
|
| |
|
|
|
|
| |
Apparently there are circumstances when a pipedesc may get orphaned form the
pollq. This triggers an assertion failure when it occurs. I am still
trying to understand how this can occur. Stay tuned.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The connect & accept logic for IPC is now fully asynchronous.
This will serve as a straight-forward template for TCP. Note that
the upper logic still uses a thread to run this "synchronously", but
that will be able to be removed once the last transport (TCP) is made
fully async.
The unified ipcsock is also now separated, and we anticipate being
able to remove the posix_sock.c logic shortly. Separating out the
endpoint logic from the pipe logic helps makes things clearer, and
may faciliate a day where endpoints have multiple addresses (for
example with a connect() endpoint that uses a round-robin DNS list
and tries to run the entire list in parallel, stopping with the first
connection made.)
The platform header got a little cleanup while we were here.
|
| |
|
