| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
| |
The use of task_abort to prematurely fail an aio at scheduling
time was a mistake, because it could have led to duplicate calls
to nng_aio_finish(). We do need to ensure that we leave an indicator
so that nni_aio_schedule can return the abort status to caller,
in the case that abort is called between the nni_aio_begin and
nni_aio_schedule calls.
|
| |
|
|
|
|
|
|
| |
Operations that might be performed during teardown, such as reaping,
waiting, closing, freeing, should only be done if the aio has properly
been initialized. This is important for certain simple cases where
inline aio objects are used, and initialization of an outer object can
fail before the enclosed aio is initialized.
|
| | |
|
| |
|
|
|
| |
Make sure *everything* is drained before proceeding all the way
to deallocation.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Probably other subsystems should get the same treatment. We need
to basically start the process of shutting down so that subsystems
know to cease operation before we rip memory out from underneath them.
This ensures that no new operations can be started as well, once we
have begun the process of teardown.
We also enhanced the completion of sleep to avoid some extra locking
contention, since the expiration *is* the completion.
Includes a test for this case.
|
| |
|
|
|
| |
We do not want to let operations restart if we're in the process of shutting
down. This ensures that they get a reasonable hard failure.
|
| |
|
|
|
|
|
|
|
| |
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.
|
| |
|
|
|
| |
If one tries to sleep indefinitely, a sign bug leads to constantly
waking calls, which causes an infinite cycle in the expire loop.
|
| | |
|
| |
|
|
| |
when an `aio` has no `a_cancel_fn` and the task is in `task_prep` abort it on `nni_aio_stop` call
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This further limits some of the thread counts, but principally it
offers a new runtime facility, nng_init_set_parameter(), which can
be used to set certain runtime parameters on the number of threads,
provided it is called before the rest of application start up.
This facility is quite intentionally "undocumented", at least for now,
as we want to limit our commitment to it. Still this should be helpful
for applications that need to reduce the number of threads that are
created.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Credit goes to Wu Xuan (@willwu1217) for diagnosing and proposing
a fix as part of #1695. This approach takes a revised approach
to avoid adding extra memory, and it also is slightly faster as we
do not need to update both pointers in the linked list, by reusing
the reap node.
As part of this a new internal API, nni_aio_completions, is introduced.
In all likelihood we will be able to use this to solve some similar
crashes in other areas of the code.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change makes expire threads tunable follows the same strategy as
taskq threads tunables.
Add NNG_NUM_EXPIRE_THREADS to override the default behavior (`n_cpu`
expire threads).
The NNG_MAX_EXPIRE_THREADS limit is always applied if > 0, even if you
specify the desired number of threads using NNG_NUM_EXPIRE_THREADS.
NNG_EXPIRE_THREADS is not used anymore. This was only referenced in the
code but never defined on CMake.
The logic to cap expire threads between 1 and 256 was removed. If users
set no limits, whatever value they choose will be used instead of being
silently overridden by us.
|
| | |
|
| |
|
|
| |
This reverts commit 8461c7207b440f5ba8c51b2236fcfa178f415a6f.
|
| |
|
| |
fixes #1543 by aborting tasks that may have been prepped, but not yet started.
|
| |
|
|
|
|
|
| |
This introduces a new API, nng_aio_busy(), that can be used
to query the status of the aio without blocking.
Some minor documentation fixes are included.
|
| |
|
|
|
|
| |
This takes one less parameter, and is simpler. It will let us
reclaim the aio_prov_extra data space as well, so that we can
use it for other purposes.
|
| | |
|
| |
|
|
|
|
| |
There were several problems with the array implementation, both
from performance and from correctness. This corrects those errors
(hopefully) and restores the expiration lists as linked lists.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
fixes #1048 nng_aio reuse error messages are unhelpful
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
fixes #1317 IPv6 listener get port is incorrect
fixes #1319 Want symbolic service names
This is phase 1 of reducing the memory foot-print of aios, and
also of pipes. This removes the largest consumer the socket
address information, from the aio, which was only used by a few
consumers.
|
| |
|
|
|
|
| |
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 #1219 nng_close occasionally hang on Windows
|
| | |
|
| |
|
|
|
| |
This only does it for rep, but it also has changes that should increase
the overall test coverage for the REP protocol
|
| | |
|
| |
|
|
| |
fixes #1097 aio prov_data not used at all
|
| | |
|
| |
|
|
|
|
|
|
|
| |
This is a major change, and includes changes to use a polymorphic
stream API for all transports. There have been related bugs fixed
along the way. Additionally the man pages have changed.
The old non-polymorphic APIs are removed now. This is a breaking
change, but the old APIs were never part of any released public API.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
| |
This changes the signature of the aio cancellation routines
to take the argument for cancellation directly, so we do not
need to lookup the argument using the nni_aio_get_prov_data.
We should probably consider eliminating nni_aio_get_prov_data,
and co, and changing the prov_extra to reflect prov_data. Later.
|
| | |
|
| | |
|
| |
|
|
|
| |
This changeset needs work. We are seeing errors described by
This reverts commit d7f7c896c0ede24249ef63b1e45b1878bf4bd473.
|
| |
|
|
|
|
|
|
|
|
| |
fixes #208 pipe start should occur before connect / accept
fixes #616 Race condition closing between header & body
This refactors the transports to handle their own connection
handshaking before passing the pipe to the socket. This
changes and simplifies the setup. This also fixes a rather
challenging race condition described by #616.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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.
|
