asCa missing locking

[mdavidsaver]

I noticed that there is a distinct lack of any locking in asCa.c to guard access to the global 'pasbase' and all the tree accessible through it. A half hearted attempt is made with 'caInitializing'.

[Andrew Johnson (anj)]

I do see a lock being taken and released in asCaStart() and asCaStop() around the signal/wait for asCaTask() to do its thing. This isn't easy to follow and certainly not how I would have written it, but in conjunction with the RCU-like approach to modifying the volatile ASBASE *pasbase in asLibRoutines.c and the LOCK/UNLOCK pair in asInitialize() I suspect there is enough here to prevent races.

No objections to it being rewritten, but I don't see it as a priority.

[mdavidsaver (mdavidsaver)]

I was looking at asCa.c from the prospective of whether I would use it as a template for some similar code w/ PVA. Given that asLock isn't exposed outside of asLibRoutines.c, and the whole thing is a singleton anyway, my conclusion is that trying to use 'pasbase' from external code is a bad idea.

If I change anything it will be to hide 'pasbase' and friends behind EPICS_PRIVATE_API.