Firebird internals : Asyncronous System Trap
What is Asyncronous System Trap ? : This is similar to Unixes signal’s or OS interrupts.
Technically this is callbacks from Lock Manager into the Engine called when
somebody acquires lock at level not compatible with current level. In this case
all processes who owns the lock receives notification and should take appropriate
action to release lock (immediately or deferred-ASAP).
The main trick with AST’s handlers is that AST is send asynchronously and
handler must take care to work correctly. Since v2.5 all ATS handlers runs in
separate thread. Before v2.5 it works using signals on Unixes and threads on
PS the issue with old glibc and FB 2.5 is that since v2.5 we abandoned usage of old
System V semaphores (do you remember “semaphores are exhausted” error ? now
you can forget it completely) and start to use pthread mutexes. And old glibc have a
bugs in pthread mutex implementation.