| Commit message (Collapse) | Author | Age |
|---|
| |
|
|
|
|
|
|
|
|
|
|
| |
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.
|
| |
|
|
|
|
|
| |
In some places, we use ifdef, and others if.
This normalizes for always using ifdef, so we can compile when
this macro is not defined.
|
| |
|
|
|
|
|
| |
In some places, we use ifdef, and others if.
This normalizes for always using ifdef, so we can compile when
this macro is not defined.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
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.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
We try to move the msgq close up earler. While here we can stop
dropping and reacquiring the lock -- this is likely left over
and may lead to races.
|
| |
|
|
| |
This reverts commit 8461c7207b440f5ba8c51b2236fcfa178f415a6f.
|
| | |
|
| |
|
| |
fixes #1543 by aborting tasks that may have been prepped, but not yet started.
|
| |
|
| |
Fix for UBSAN error. As `id_reg_map` is initialised as NULL and passing NULL to `memcpy()` is undefined. Should make no difference to compiled code. Purely to appease the automated checks I have to deal with on my side. Thanks!
|
| | |
|
| |
|
| |
The socket member `st_name` is not available when NNG_ENABLE_STATS is OFF.
|
| |
|
|
|
|
|
| |
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.
|
| |
|
|
|
| |
Also this fixes problems with uninitialized socket names, and the
socket name stat not being adjusted correctly when set via API.
|
| |
|
|
|
|
| |
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.
|
| | |
|
| | |
|
| |
|
|
|
|
| |
This function is like nng_device(), but runs asynchronously.
Also, this fixes #1503 nng_device causes nng_close to blocking
|
| |
|
|
|
| |
This makes these functions entirely bullet proof, and eliminates
yet more error handling cases.
|
| |
|
|
|
| |
This allows us to make nni_lmq_init() non-failing. (Although
the buffer size requested at initialization might not be granted.)
|
| | |
|
| |
|
|
|
| |
This eliminates more failure paths, and brings us still closer
to eliminating the possibility of failure during socket init.
|
| | |
|
| |
|
|
|
| |
Note that one of these warning is a real bug that would prevent
TLS from functioning properly on Windows.
|
| |
|
|
|
|
|
|
| |
This should help greatly with performance on older systems such
as CentOS 7 and GCC 4.8. Though, such folks really should update
to newer compilers. Folks running version of GCC earlier than 4.7
will still pay a rather significant performance penalty, as they
still implement atomics with a global mutex.
|
| |
|
|
| |
This also arranges to clean up the maps at nng_fini time.
|
| | |
|
| |
|
|
|
|
| |
This eliminates some run-time initialization, moving it to compile time.
Additional follow up work will expand on this to simplify initialization
and reduce the need for certain locks.
|
| |
|
|
|
| |
This is initially used for TLS to make loading the engine pointer
faster, eliminating a much more expensive lock operation.
|
| | |
|
| |
|
|
|
|
| |
(Note that this includes some incorrect formatting due to an apparent
bug in clang-format 13. Hopefully they'll fix it later, but for now
I'm preserving the bad whitespace.)
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
This eliminates several mutex operations done each time a pipe
is created or destroyed. For large scale systems this should
reduce overall pressure on the memory subsystem, and scale better
as many threads are coming and going.
This also reduces the overall size of nni_pipe -- on Linux by
36 bytes typically.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
Also, remove an extraneous initialization call.
|
| |
|
|
|
|
| |
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.
|
| | |
|
| |
|
|
|
| |
Nothing is using this, but it adds complexity and also requires
additional lock activity each time a socket is opened.
|
| | |
|
| | |
|
