epics/pvAccess
0
0
Fork 0
Code Issues Pull Requests Projects Wiki Activity
1,577 Commits 1 Branch 41 Tags
a413f8f9db23f3b7e1ad34ed76977912cd9de7d1
Commit Graph

1577 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Michael Davidsaver
a413f8f9db add dummy ChannelFind implementation. 
Since this class is not really used...
 2018-06-26 11:23:50 -07:00
Michael Davidsaver
656e2fcfd4 PVA server avoid possible data race 
The operation member variable is set by from the
returned pointer, and within the *Connect callback.
Make sure these down race.
 2018-06-26 11:23:50 -07:00
Michael Davidsaver
4baa72c91d avoid data race 
destroy() sets _serverSocketChannel=INVALID_SOCKET under the lock.
Other changes are made before the worker thread starts.
 2018-06-26 10:45:14 -07:00
Michael Davidsaver
1ba48868f6 PVA server avoid lock order violation 
There is a lock order violation when a server force disconnects it's clients while sending a final message to it (or concurrent close, or probably others). The entry points are channelStateChange() and operation send().

Fix is not to hold the channel mutex when destroy()ing operations.

epics::pvAccess::ServerChannelRPCRequesterImpl::destroy
  locks RPC requester mutex
epics::pvAccess::ServerChannel::destroy
  locks channel mutex
epics::pvAccess::ServerChannelRequesterImpl::channelStateChange


epics::pvAccess::ServerChannel::unregisterRequest
  locks channel mutex
epics::pvAccess::ServerChannelRPCRequesterImpl::destroy
  locks RPC requester mutex
epics::pvAccess::ServerChannelRPCRequesterImpl::send
 2018-06-25 19:52:10 -07:00
Michael Davidsaver
4b3237a20d BlockingUDPTransport::enqueueSendRequest cleanup? 
oh my was this bad code!  It isn't clear what
lock()/unlock() are guarding, but let them
continue as before, but don't swallow any exception,
or unlock twice.
 2018-06-22 13:28:49 -07:00
Michael Davidsaver
d424202323 client: exception propagation from callbacks 
Swallowing exceptions here hides some of our errors.
Particularly unrecoverable introspectionRegistry misses.
Instead, propagate up to Thread and close connection.
 2018-06-22 13:28:49 -07:00
Michael Davidsaver
1670ff8f7c SerializationHelper cleanup and add error checks 2018-06-22 13:28:49 -07:00
Michael Davidsaver
3abd77a42c fix IntrospectionRegistry error handling 
Calling code can't deal with "cache" miss,
which indicates either a protocol error (use of invalid ID),
or a logic error on our part (didn't save ID).
In either case, better to throw an exception than to crash.
 2018-06-22 13:28:49 -07:00
Michael Davidsaver
e08de8c2df detect Structure caching error 
The assert in the client side copy is hit during
some kind of failure to keep the Field cache
up to date.
 2018-06-22 13:28:49 -07:00
Michael Davidsaver
de6586b92f MonitorElement::Ref swap() and move, but no copy 2018-06-05 09:32:22 -07:00
Michael Davidsaver
578dbc9456 add another ChannelProviderRegistry::addSingleton() overload 
Construct a "factory" around a pre-allocated instance.
 2018-06-05 09:32:22 -07:00
Michael Davidsaver
8000ed0637 client.h default ctor for ClientProvider 
Allow an invalid ClientProvider to be created
for later assignment.
 2018-06-05 09:32:22 -07:00
Michael Davidsaver
c8292caf91 remote: Request larger stack for workers 
Overflows observed on RTEMS/ppc (mvme6100).
 2018-06-04 10:08:01 -07:00
Michael Davidsaver
794d72ba6b client.h MonitorSync missing init of member 2018-06-04 10:08:01 -07:00
Michael Davidsaver
44132be716 avoid MSVC min()/max() conflict 2018-05-28 10:19:48 -07:00
Michael Davidsaver
12e7b7864b fix s_channelNameToProvider hack 
Move from global to ServerContext, apply lock, and
properly test if no provider has claimed, eg an attempt
at unadvertised PV (still wouldn't work, but at least
wouldn't crash).
 2018-05-18 17:04:07 -07:00
Michael Davidsaver
7a523dd948 drop unused NOMINMAX 2018-05-18 15:20:26 -07:00
Michael Davidsaver
ba41fa6d62 notes 2018-05-18 15:17:27 -07:00
Michael Davidsaver
ce19199edd cleanup 2018-05-18 11:58:35 -07:00
Michael Davidsaver
6ccca1ce58 const-ify string constants 
yes really...
 2018-05-18 10:12:34 -07:00
Michael Davidsaver
6abfe9d196 server: cleanup handling of vector of providers 
avoid keeping extra refs.  situation is confusing
enough as it is.
 2018-05-18 08:53:52 -07:00
Michael Davidsaver
bd88e35448 client.h update doc 2018-05-16 09:14:49 -07:00
Michael Davidsaver
51fdd506d0 update release notes 6.1.0-pre1 2018-04-24 13:50:59 -07:00
Michael Davidsaver
a75720120c client: call ChannelSearchManager::cancel() 2018-04-24 12:33:54 -07:00
Michael Davidsaver
1adea89e77 pvtools: combine separator and fieldSeparator 
both always have the same value anyway...
 2018-04-24 08:33:25 -07:00
Michael Davidsaver
9c25057d80 pvtools: drop unnecessary accessors for global flags 
These don't provide any encapsulation or validation...
 2018-04-24 08:29:45 -07:00
Andrew C. Starritt
a2b60771c2 pvget -m show time/alarm 
MD: cleanups
 2018-04-24 07:36:06 -07:00
Michael Davidsaver
7cf78d5b9b client.h add ClientProvider::name() 2018-04-20 15:22:57 -07:00
Michael Davidsaver
3d85852c93 attempt to fix dllimport issue 2018-04-18 09:01:11 -07:00
Michael Davidsaver
61d43e498d client.h std::ostream printing 2018-04-10 16:19:03 -07:00
Michael Davidsaver
0d40464541 client.h: expose reset() 
can also be accomplished with assignment from empty.

op = pvac::Operation()
 2018-04-10 15:17:18 -07:00
Michael Davidsaver
0c6fe2c1c6 client.h: add validity test for all handles 2018-04-08 16:04:36 -07:00
Michael Davidsaver
ff5d51b221 server: shorter name for Timer thread 2018-04-04 20:55:57 -07:00
Michael Davidsaver
1e0523b7b9 client context. Prevent transport dtor under channel mutex. 
m_channelMutex is locked recursively, so places like
disconnect(bool,bool) which purport to unlock for the
Transport detory actually don't as they are called
with m_channelMutex already locked.

Make a rather ugly hack of saving a ref to m_channel
at the various entry point.
 2018-04-04 19:25:04 -07:00
Michael Davidsaver
c9c8fb369a client context explicitly close() timer queue 2018-04-04 18:53:22 -07:00
Michael Davidsaver
d4822f141c simplify InternalClientContextImpl::destroy() 
Collapse internalDestroy() into destroy()
 2018-04-04 18:50:58 -07:00
Michael Davidsaver
128d4b67f9 simplify InternalChannelImpl::destroy() 
Collapse channel destroy(bool) and destroyChannel() into destroy()
 2018-04-04 18:48:48 -07:00
Michael Davidsaver
e97354db61 client context collapse channel createChannel() into searchResponse() 2018-04-04 18:44:11 -07:00
Michael Davidsaver
bab32c332a codec: notes on thread self dtor 2018-04-04 18:40:46 -07:00
Michael Davidsaver
54bb30fd86 reftrack BlockingUDPTransport 2018-04-04 18:31:12 -07:00
Michael Davidsaver
a8e26123c7 minor doc 2018-04-04 18:31:12 -07:00
Michael Davidsaver
7f5eb1de4d BlockingClientTCPTransportCodec collapse closedNotifyClients() into internalClose() 2018-04-04 18:31:12 -07:00
Michael Davidsaver
0eabf10005 codec: drop unused _shutdownEvent 
using epicsThread::exitWait() instead
 2018-04-04 18:31:12 -07:00
Michael Davidsaver
192955e7b5 client context: don't call cancel() 
it doesn't do anything!
 2018-04-04 14:55:41 -07:00
Michael Davidsaver
83a941c070 channelSearchManager minor 2018-04-04 14:53:52 -07:00
Michael Davidsaver
18491f6eb3 const-ify AbstractClientResponseHandler::_context 2018-04-04 10:12:44 -07:00
Michael Davidsaver
9e090fa191 client context: avoid lock order violations 
avoid ordering violation of calling checkAndGetTransport(),
which locks channel mutex, while request mutex is locked.
 2018-04-04 10:12:44 -07:00
Michael Davidsaver
8068ab3454 eliminate unnecessary m_nullMonitorElement 2018-04-04 10:12:44 -07:00
Michael Davidsaver
40ecc93e9d client context: use mutex guards 
whether this locking is necessary is another question...
 2018-04-04 10:12:44 -07:00
Michael Davidsaver
f2164ddb5a fix typedef in security.h 
Can't refer to typedef in (private) remote.h
from (public) security.h
 2018-04-04 10:12:44 -07:00