KERNEL DESIGN ISSUES - Hard Real-Time Computing Systems: Predictable Scheduling Algorithms and Applications - page 389

10.6.2 CAB IMPLEMENTATION

The data structure used to implement a CAB is shown in Figure 10.17. A CAB con-
trol block must store the maximum number of buffers ( max buf ), their dimension
( dim buf ), a pointer to a list of free buffers ( free ), and a pointer to the most recent
buffer ( mrb ). Each buffer in the CAB can be implemented as a data structure with
three fields: a pointer ( next ) to maintain a list of free buffers, a counter ( use ) that
stores the current number of tasks accessing that buffer, and a memory area ( data ) for
storing the message.

The code of the four CAB primitives is shown below. Note that the main purpose of
the putmes primitive is to update the pointer to the most recent buffer (MRB). Before
doing that, however, it deallocates the old MRB if no tasks are accessing that buffer.
Similarly, the unget primitive decrements the number of tasks accessing that buffer
and deallocates the buffer only if no task is accessing it and it is not the MRB.

free

mrb

max_buf

dim_buf

next next

NIL

use use use use

reading task

most
recent

p

data empty empty

buffer

Figure 10.17 CAB data structure.