CS 454

1
CS 454

• Process State Transitions and System Calls

2

• Why use getpid() instead of currpid?
• Portability of Code It is not always possible
  to access global variables, such as currpid, from
  any process in some operating systems.
• What is wrong with kill()?
• It does not follow the C calling convention. The
  return address is not right above the arguments
  in the stack the process ID is. The arguments
  are therefore not at the correct offset with
  respect to the base pointer. It will try to take
  the PID as argument 0.

3
Process State Transitions
4
Process State Transitions

• Suspend (suspend.c)
• check pid
• check for current or ready
• if ready then
• dequeue the process from the ready queue
• put it in suspended state

5
Process State Transitions

• Suspend (suspend.c)
• if current then
• put it in suspended state
• call resched(), which calls ctxsw()
• NOTE
• system calls suspend, resume, kill, resched
• utility functions getprio, getpid, chprio

6
System Calls

• The precautions, that system calls take to verify
  that an operation is illegal, make the general
  purpose subroutines that can be invoked by any
  process at any time.
• System calls stand between a naive user program
  and the rest of the operating system, which must
  protect the internal system from illegal use.

7
System Calls

• System calls define the exterior of the OS by
  providing an interface through which the user
  accesses all system services.
• Difference between library and system calls
• When a program calls a subroutine, the code is
  always part of the final object program, even if
  it was linked in from a library.

8
System Calls

• Difference between library and system calls
• With a system call, the major part of the code
  excecuted is actually part of the kernel itself
  and not the calling program.
• In other word, the calling program is making
  direct use of the facilities provided by the
  kernel.

9
Chapter 6 Process Coordination

• Look at Semaphores and their functions
• wait(sem)
• signal(sem)
• screate(count)
• newsem()
• sdelete(sem)
• sreset(sem, count)

10
When are semaphores used?

• For the collaboration and coordination of
  processes
• When processes used shared resources
• i.e. the producer and consumer type of process
• Protects the shared resource

11
What other ways can we protect shared resources?

• Disable Interrupts
• Not Good
• Miss interrupts from I/O devices or other things
  that dont effect the current process
• Performance degradation
• Not all processes may need to access the shared
  resource

12
XINU Semaphores

• In XINU, each semaphore has 4 fields
• struct sentry state / used or free
  / semcount sqhead sqtail

13
Semaphore Invariant

• A non-negative semaphore count means that the
  queue is empty. A semaphore count of negative n
  means that the queue contains n waiting processes

14
How many semaphores are needed?

• For a producer consumer problem with one buffer
• Similar to ex5.c page 13-14 of text
• 2 for producer/consumer
• What are they initialized to
• if both 0 neither can run, both wait
• if both 1 numbers may be produced twice

15

• What if change which is initialized to 1 and
  which is initialized to 0?
• if consumed 0 produced 1 then ex5.c will
  print 0 - 1999
• if consumed 1 produced 0 then ex5.c will
  print 1 - 2000
• if you just want mutual exclusion to the shared
  resource, just need 1 semaphore (see ex6.c on
  pages 14-15 of the text)