Transcript Operating Systems

Operating Systems
CMPSC 473
Processes (3)
September 17 2008 - Lecture 9
Instructor: Bhuvan Urgaonkar
Announcements
• Suggested Reading for this lecture: Sections
3.1-3.3
• Quiz 1: Will have several questions that came
up during previous lectures
•
Overview of ProcessHow a process is born
related Topics
– Parent/child relationship
– fork, clone, …
• How it leads its life
– Loaded: Later in the course
– Executed
• CPU scheduling
• Context switching
• Where a process “lives”: Address space
– OS maintains some info. for each process: PCB
– Process = Address Space + PCB
• How processes request services from the OS
– System calls
• How processes communicate
• Some variants of processes: LWPs and threads
• How processes die
•
Overview of ProcessHow a process is born
related Topics
– Parent/child relationship
– fork, clone, …
• How it leads its life
– Loaded: Later in the course
– Executed
• CPU scheduling
• Context switching
more today
• Where a process “lives”: Address space
– OS maintains some info. for each process: PCB
– Process = Address Space + PCB
• How processes request services from the OS
– System calls
• How processes communicate
• Some variants of processes: LWPs and threads
• How processes die
The Process/Kernel
Model
Process 1
Process 1
Process 2
Process 2
USER MODE
KERNEL MODE
Sys call
handler
Scheduler
Interrupt
handler
Time
• Transitions between User and Kernel modes: An example
Re-entrant Kernels
Process 1
Process 1
Process 2
USER MODE
KERNEL MODE
Excp
Time
Intr
Intr
Intr
•
Note: Not showing scheduler invocations
•
Re-entrant kernel: Several processes may be in Kernel Mode at the same time
–
•
A re-entrant kernel is able to suspend the current running process even if it is in the Kernel Mode
Note: Traps are a type of exceptions. We will encounter more types later.
Re-entrant Kernels
Process 1
Process 1
Process 2
USER MODE
KERNEL MODE
Excp
Intr
Intr
Kernel control paths
Time
Intr
•
Note: Not showing scheduler invocations
•
Re-entrant kernel: Several processes may be in Kernel Mode at the same time
–
A re-entrant kernel is able to suspend the current running process even if it is in the Kernel Mode
Re-entrant Kernels
Process 1
Process 1
Process 2
USER MODE
KERNEL MODE
Excp
Intr
Intr
Kernel control paths
Time
Intr
•
Note: Not showing scheduler invocations
•
Re-entrant kernel: Several processes may be in Kernel Mode at the same time
–
•
A re-entrant kernel is able to suspend the current running process even if it is in the Kernel Mode
A kernel control path denotes the sequence of instructions executed by the
kernel to handle a system call, an exception, or an interrupt
Re-entrant Kernels
Process 1
Process 1
Process 2
USER MODE
KERNEL MODE
Excp
Intr
Intr
Kernel control paths
Time
Intr
•
Note: Not showing scheduler invocations
• Why re-entrancy?
Re-entrant Kernels
Process 1
Process 1
Process 2
USER MODE
KERNEL MODE
Excp
Intr
Intr
Kernel control paths
Time
Intr
•
Note: Not showing scheduler invocations
•
Why re-entrancy?
–
–
Improves throughput of devices controllers that raise interrupts
Allows priorities among interrupts
Realizing
re-entrancy: Take 1
• Write kernel functions that only modify local
variables and do not alter global variables
– Re-entrant functions
• Pros and Cons?
Realizing
re-entrancy: Take 1
• Take 1: Write kernel functions that only
modify local variables and do not alter global
variables
– Re-entrant functions
• Pros/Cons
– Simplifies/complicates kernel programming (?)
Realizing re-entrancy:
Take 2
• Kernel Mode Stacks
• We know: a process running in User Model refers to
its private stack
• To allow re-entrancy, there is a Kernel Mode Stack
for each process
• Each kernel control path uses its own private kernel
stack
• Kept in part of RAM reserved for the kernel
Kernel Mode Stack
Stack
esp
PCB
(task_struct)
• KM stack and PCB
need to be able to find
each other
• KM stack must have
access to a pointer to
the PCB
Kernel Mode Stack
Stack
esp
thread_info
structure
task
curent
PCB
(task_struct)
• KM stack and PCB need
to be able to find each
other
• KM stack must have
access to a pointer to the
PCB
– Linux: thread_info
• PCB must have access to
KM stack
Kernel Mode Stack
Stack
esp
thread_info
structure
PCB
(task_struct)
task
curent
thread_info
• KM stack and PCB need
to be able to find each
other
• KM stack must have
access to a pointer to the
PCB
– Linux: thread_info
• PCB must have access to
KM stack
Kernel Mode Stack
•
Stack
esp
thread_info
structure
PCB
(task_struct)
task
curent
thread_info
Since KM stacks make little use of the
stack, only a few thousand bytes suffice
– An example of “Design for the most
common case”, we’ll see more
– Linux: 8KB , thread_info 52 bytes
Kernel Mode Stack
•
Stack
esp
thread_info
structure
PCB
(task_struct)
task
curent
thread_info
union thread_union {
struct thread_info
thread_info;
unsigned long stack[2048];
};
Since KM stacks make little use of the
stack, only a few thousand bytes suffice
– An example of “Design for the most
common case”, we’ll see more
– Linux: 8KB
•
Why combine KM stack and
thread_info into a union?
Kernel Mode Stack
•
Stack
esp
thread_info
structure
PCB
(task_struct)
task
curent
thread_info
union thread_union {
struct thread_info
thread_info;
unsigned long stack[2048];
};
Since KM stacks make little use of the stack,
only a few thousand bytes suffice
– An example of “Design for the most
common case”, we’ll see more
– Linux: KM Stack 8KB, thread_info 52 bytes
•
Why combine KM stack and thread_info into
a union?
– You might think spatial locality
– The kernel can easily obtain the address of
the thread_info structure of the process
currently running on the CPU from the value
of the esp register
– task field is at offset 0
– Other benefits apply to multi-processors:
makes it easy to efficiently find the current
process on each processor
• Earlier approach: Have an array of
current pointers
Enumerating # Possible
CPU Multiplexing
Between Processes
• Consider two processes P1 and P2
– P1: n instructions
– P2: m instructions
– No jump instrictions => No loops
• How many unique executions are possible?