Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Mon, 27 Oct 2008 10:44:24 -0500

On Monday 27 October 2008 09:48:00 Dirk Zimoch wrote:
>
> It might be possible to make this thread-safe by disabling interrupts
> before calling dbPvdAdd().

That won't work on any workstation OS, where epicsInterruptLock() just takes a 
global mutex so it won't stop another thread from calling dvPvdFind while the 
dbPvdAdd() is in the middle of initializing an unused bucket or adding a new 
node (unfortunately both actions currently insert their new object before 
initializing it completely).

On vxWorks and RTEMS that might be safe as long as the record has been 
completely initialized before the lock is released, but I'm not sure you 
could safely insert the new record into a lockset since you have to take a 
lock to do that, and the OS allows itself to re-enable interrupts if your 
code blocks or releases a lock.  Read the vxWorks manual entry for intLock() 
for details.

- Andrew
-- 
Talk is cheap. Show me the code. -- Linus Torvalds

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System