Admission policy for protected objects

On multiprocessor systems, a spin lock is typically used to gain access to a protected object in order to execute a protected action. Most multiprocessor locking algorithms prescribe that if there is more than one request competing for the same resource, the requests are served in FIFO order. This bounds the time it takes for a lower priority task to gain access to a protected object. With Ravenscar all tasks are statically allocated to processors, and if each protected object used by tasks on different processors is given a high ceiling priority then the blocking time for each task can be computed.

Previously the language did not presume any ordering or queuing for tasks competing to start a protected action, Admission policy defined for acquiring a protected object resource (AI12-0276) adds:

to specify that FIFO_Spinning is used (see RM D.4.1). A task will inherit the ceiling priority of the protected object. Other, implementation-defined, Admission_Policies may also be specified.

6.6 Admission policy for protected objects