Carnegie Mellon Exceptions and Processes Slides adapted from: Gregory Kesden and Markus Püschel of Carnegie Mellon University

Carnegie Mellon Control Flow inst 1 inst 2 inst 3 … inst n Processors do only one thing:  From startup to shutdown, a CPU simply reads and executes (interprets) a sequence of instructions, one at a time  This sequence is the CPU’s control flow (or flow of control) Physical control flow Time

Carnegie Mellon Altering the Control Flow Up to now: two mechanisms for changing control flow:  Jumps and branches  Call and return Both react to changes in program state Insufficient for a useful system: Difficult to react to changes in system state  data arrives from a disk or a network adapter  instruction divides by zero  user hits Ctrl-C at the keyboard  System timer expires System needs mechanisms for “exceptional control flow”

Carnegie Mellon Exceptional Control Flow Exists at all levels of a computer system Low level mechanisms  Exceptions  change in control flow in response to a system event (i.e., change in system state)  Combination of hardware and OS software Higher level mechanisms  Process context switch  Signals  Nonlocal jumps: setjmp()/longjmp()  Implemented by either:  OS software (context switch and signals)  C language runtime library (nonlocal jumps)

Carnegie Mellon Exceptions An exception is a transfer of control to the OS in response to some event (i.e., change in processor state) Examples: div by 0, arithmetic overflow, page fault, I/O request completes, Ctrl-C User ProcessOS exception exception processing by exception handler return to I_current return to I_next abort event I_current I_next

Carnegie Mellon n-1 Interrupt Vectors Each type of event has a unique exception number k k = index into exception table (a.k.a. interrupt vector) Handler k is called each time exception k occurs Exception Table code for exception handler 0 code for exception handler 1 code for exception handler 2 code for exception handler n-1... Exception numbers

Carnegie Mellon Asynchronous Exceptions (Interrupts) Caused by events external to the processor  Indicated by setting the processor’s interrupt pin  Handler returns to “next” instruction Examples:  I/O interrupts  hitting Ctrl-C at the keyboard  arrival of a packet from a network  arrival of data from a disk  Hard reset interrupt  hitting the reset button  Soft reset interrupt  hitting Ctrl-Alt-Delete on a PC

Carnegie Mellon Synchronous Exceptions Caused by events that occur as a result of executing an instruction:  Traps  Intentional  Examples: system calls, breakpoint traps, special instructions  Returns control to “next” instruction  Faults  Unintentional but possibly recoverable  Examples: page faults (recoverable), protection faults (unrecoverable), floating point exceptions  Either re-executes faulting (“current”) instruction or aborts  Aborts  unintentional and unrecoverable  Examples: parity error, machine check  Aborts current program

Carnegie Mellon Trap Example: Opening File User calls: open(filename, options) Function open executes system call instruction int OS must find or create file, get it ready for reading or writing Returns integer file descriptor 0804d070 : d082:cd 80 int $0x80 804d084:5b pop %ebx... User ProcessOS exception open file returns int pop

Carnegie Mellon Fault Example: Page Fault User writes to memory location That portion (page) of user’s memory is currently on disk Page handler must load page into physical memory Returns to faulting instruction Successful on second try int a[1000]; main () { a[500] = 13; } 80483b7:c d d movl $0xd,0x8049d10 User ProcessOS exception: page fault Create page and load into memory returns movl

Carnegie Mellon Fault Example: Invalid Memory Reference Page handler detects invalid address Sends SIGSEGV signal to user process User process exits with “segmentation fault” int a[1000]; main () { a[5000] = 13; } 80483b7:c e d movl $0xd,0x804e360 User ProcessOS exception: page fault detect invalid address movl signal process

Carnegie Mellon User mode vs. Kernel mode Privileged instructions

Carnegie Mellon Exception handlers Return address  Depends on the class of exception  Current instruction (page fault)  Next instruction Push some additional processor state  Necessary to restart the interrupted program when the handler returns If the control is being transferred from user to kernel  All these items are pushed on the kernel stack Exception handlers run in kernel mode

Carnegie Mellon Exception Table IA32 (Excerpt) Exception NumberDescriptionException Class 0Divide errorFault 13General protection faultFault 14Page faultFault 18Machine checkAbort OS-definedInterrupt or trap 128 (0x80)System callTrap OS-definedInterrupt or trap Check pp. 183:

Carnegie Mellon # write our string to stdout movl $len,%edx # third argument: message length movl $msg,%ecx # second argument: message to write movl $1,%ebx # first argument: file handle (stdout) movl $4,%eax # system call number (sys_write) int $0x80 # call kernel # and exit movl $0,%ebx # first argument: exit code movl $1,%eax # system call number (sys_exit) int $0x80 # call kernel.data # section declaration msg:.ascii "Hello, world!\n" # our dear string len =. - msg # length of our dear string int main() { write(1, “hello, world\n”, 13); exit(0); }

Carnegie Mellon Today Exceptional Control Flow Processes

Carnegie Mellon Processes Definition: A process is an instance of a running program.  One of the most profound ideas in computer science  Not the same as “program” or “processor” Process provides each program with two key abstractions:  Logical control flow  Each program seems to have exclusive use of the CPU  Private virtual address space  Each program seems to have exclusive use of main memory How are these Illusions maintained?  Process executions interleaved (multitasking)  Address spaces managed by virtual memory system

Carnegie Mellon What is a process? A process is the OS's abstraction for execution  A process represents a single running application on the system Process has three main components: 1) Address space  The memory that the process can access  Consists of various pieces: the program code, static variables, heap, stack, etc. 2) Processor state  The CPU registers associated with the running process  Includes general purpose registers, program counter, stack pointer, etc. 3) OS resources  Various OS state associated with the process  Examples: open files, network sockets, etc.

Carnegie Mellon Process Address Space  The range of virtual memory addresses that the process can access  Includes the code of the running program  The data of the running program (static variables and heap)‏  An execution stack  Local variables and saved registers for each procedure call Stack Heap Initialized vars (data segment) ‏ Code (text segment) ‏ Address space 0x xFFFFFFFF Stack pointer Program counter Uninitialized vars (BSS segment) ‏ (Reserved for OS) ‏

Carnegie Mellon Process Address Space Note!!! This is the process's own view of the address space --- physical memory may not be laid out this way at all.  In fact, on systems that support multiple running processes, it's pretty much guaranteed to look different. The virtual memory system provides this illusion to each process. Stack Heap Initialized vars (data segment) ‏ Code (text segment) ‏ Address space 0x xFFFFFFFF Stack pointer Program counter Uninitialized vars (BSS segment) ‏ (Reserved for OS) ‏

Carnegie Mellon Execution State of a Process Each process has an execution state  Indicates what the process is currently doing Running:  Process is currently using the CPU Ready:  Currently waiting to be assigned to a CPU  That is, the process could be running, but another process is using the CPU Waiting (or sleeping):  Process is waiting for an event  Such as completion of an I/O, a timer to go off, etc.  Why is this different than “ready” ? As the process executes, it moves between these states  What state is the process in most of the time?

Carnegie Mellon Process State Transitions  What causes schedule and unschedule transitions? New Terminated Ready Running Waiting create kill or exit I/O, page fault, etc. I/O done schedule unschedule

Carnegie Mellon Process Control Block OS maintains a Process Control Block (PCB) for each process The PCB is a big data structure with many fields:  Process ID  User ID  Execution state  ready, running, or waiting  Saved CPU state  CPU registers saved the last time the process was suspended.  OS resources  Open files, network sockets, etc.  Memory management info  Scheduling priority  Give some processes higher priority than others  Accounting information  Total CPU time, memory usage, etc.

Carnegie Mellon Context Switching Processes are managed by a shared chunk of OS code called the kernel  Important: the kernel is not a separate process, but rather runs as part of some user process Control flow passes from one process to another via a context switch Process AProcess B user code kernel code user code kernel code user code context switch Time

Carnegie Mellon Context Switching  The act of swapping a process state on or off the CPU is a context switch PC Registers PC Registers PID 1342 State: Running PC Registers PID 4277 State: Ready PC Registers PID 8109 State: Ready Save current CPU state Currently running process

Carnegie Mellon Context Switching  The act of swapping a process state on or off the CPU is a context switch PC Registers PC Registers PID 1342 State: Ready PC Registers PID 4277 State: Ready PC Registers PID 8109 State: Ready Suspend process

Carnegie Mellon Context Switching  The act of swapping a process state on or off the CPU is a context switch PC Registers PC Registers PID 1342 State: Ready PC Registers PID 4277 State: Running PC Registers PID 8109 State: Ready Restore CPU state of new process Pick next process PC Registers

Carnegie Mellon Context Switch Overhead Context switches are not cheap  Generally have a lot of CPU state to save and restore  Also must update various flags in the PCB  Picking the next process to run – scheduling – is also expensive Context switch overhead in Linux  About 5.4 usec on a 2.4 GHz Pentium 4  This is equivalent to about 13,200 CPU cycles!  Not quite that many instructions since CPI > 1

Carnegie Mellon Context Switching in Linux Process A time Process A is happily running along...

Carnegie Mellon Context Switching in Linux Process A time Timer interrupt handler 1) Timer interrupt fires 2) PC saved on stack User Kernel

Carnegie Mellon Context Switching in Linux Process A Timer interrupt handler time 1) Timer interrupt fires 2) PC saved on stack Scheduler 4) Call schedule() routine 3) Rest of CPU state saved in PCB User Kernel

Carnegie Mellon Context Switching in Linux Process A Timer interrupt handler time 1) Timer interrupt fires 2) PC saved on stack Scheduler 5) Decide next process to run 4) Call schedule() routine 3) Rest of CPU state saved in PCB Timer interrupt handler 6) Resume Process B (suspended within timer interrupt handler!)‏ User Kernel Process B 7) Return from interrupt handler – process CPU state restored

Carnegie Mellon State Queues The OS maintains a set of state queues for each process state  Separate queues for ready and waiting states  Generally separate queues for each kind of waiting process  e.g., One queue for processes waiting for disk I/O  Another queue for processes waiting for network I/O, etc. PC Registers PID 4277 State: Ready PC Registers PID 4110 State: Waiting PC Registers PID 4002 State: Waiting PC Registers PID 4923 State: Waiting PC Registers PID 4391 State: Ready Ready queue Disk I/O queue

Carnegie Mellon State Queue Transitions PCBs move between these queues as their state changes  When scheduling a process, pop the head off of the ready queue  When I/O has completed, move PCB from waiting queue to ready queue PC Registers PID 4277 State: Ready PC Registers PID 4110 State: Waiting PC Registers PID 4002 State: Waiting PC Registers PID 4391 State: Ready PC Registers PID 4923 State: Waiting Ready queue Disk I/O queue PC Registers PID 4923 State: Ready Disk I/O completes

Carnegie Mellon Process Creation One process can create, or fork, another process  The original process is the parent  The new process is the child  What creates the first process in the system, and when? Parent process defines resources and access rights of children  Just like real life...  e.g., child process inherits parent's user ID % pstree -p init(1)-+-apmd(687)‏ |-atd(847)‏ |-crond(793)‏ |-rxvt(2700)---bash(2702)---ooffice(2853)‏ `-rxvt(2752)---bash(2754)‏

Carnegie Mellon PC Registers PID 4110 State: Ready UNIX fork mechanism In UNIX, use the fork() system call to create a new process  This creates an exact duplicate of the parent process!!  Creates and initializes a new PCB  Creates a new address space  Copies entire contents of parent's address space into the child  Initializes CPU and OS resources to a copy of the parent's  Places new PCB on ready queue PC Registers PID 4109 State: Running PC Registers PID 4277 State: Ready PC Registers PID 4391 State: Ready Ready queue Process calls fork()‏ PC Registers PID 4110 State: Ready Copy state PC Registers PID 4110 State: Ready Add to end of ready queue

Carnegie Mellon fork : Creating New Processes int fork(void)  creates a new process (child process) that is identical to the calling process (parent process)  returns 0 to the child process  returns child’s pid to the parent process Fork is interesting (and often confusing) because it is called once but returns twice pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); }

Carnegie Mellon Understanding fork pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); } Process n pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); } Child Process m pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); } pid = m pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); } pid = 0 pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); } pid_t pid = fork(); if (pid == 0) { printf("hello from child\n"); } else { printf("hello from parent\n"); } hello from parenthello from child Which one is first?

Carnegie Mellon Fork Example #1 void fork1() { int x = 1; pid_t pid = fork(); if (pid == 0) { printf("Child has x = %d\n", ++x); } else { printf("Parent has x = %d\n", --x); } printf("Bye from process %d with x = %d\n", getpid(), x); } Parent and child both run same code  Distinguish parent from child by return value from fork Start with same state, but each has private copy  Including shared output file descriptor  Relative ordering of their print statements undefined

Carnegie Mellon Fork Example #2 void fork2() { printf("L0\n"); fork(); printf("L1\n"); fork(); printf("Bye\n"); } Both parent and child can continue forking L0 L1 Bye

Carnegie Mellon Fork Example #3 Both parent and child can continue forking void fork3() { printf("L0\n"); fork(); printf("L1\n"); fork(); printf("L2\n"); fork(); printf("Bye\n"); } L1L2 Bye L1L2 Bye L0

Carnegie Mellon Fork Example #4 Both parent and child can continue forking void fork4() { printf("L0\n"); if (fork() != 0) { printf("L1\n"); if (fork() != 0) { printf("L2\n"); fork(); } printf("Bye\n"); } L0 L1 Bye L2 Bye

Carnegie Mellon Fork Example #4 Both parent and child can continue forking void fork5() { printf("L0\n"); if (fork() == 0) { printf("L1\n"); if (fork() == 0) { printf("L2\n"); fork(); } printf("Bye\n"); } L0 Bye L1 Bye L2

Carnegie Mellon Why have fork() at all? Why make a copy of the parent process? Don't you usually want to start a new program instead? Where might “cloning” the parent be useful?

Carnegie Mellon Why have fork() at all? Why make a copy of the parent process? Don't you usually want to start a new program instead? Where might “cloning” the parent be useful?  Web server – make a copy for each incoming connection  Parallel processing – set up initial state, fork off multiple copies to do work UNIX philosophy: System calls should be minimal.  Don't overload system calls with extra functionality if it is not always needed.  Better to provide a flexible set of simple primitives and let programmers combine them in useful ways.

Carnegie Mellon Memory concerns So fork makes a copy of a process. What about memory usage? Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Parent Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #1 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #2 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #3 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #4 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #5 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #6 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #7 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #8 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #10 Stack Heap Initialized vars (data segment)‏ Code (text segment)‏ Uninitialized vars (BSS segment)‏ (Reserved for OS)‏ Child #9

Carnegie Mellon Problematic?

Carnegie Mellon Memory concerns OS aggressively tries to share memory between processes.  Especially processes that are fork()'d copies of each other Copies of a parent process do not actually get a private copy of the address space... ... Though that is the illusion that each process gets.  Instead, they share the same physical memory, until one of them makes a change. The virtual memory system is behind these shenanigans.  We will discuss this in much detail later in the course

Carnegie Mellon exit : Ending a process void exit(int status)  exits a process  Normally return with status 0  atexit() registers functions to be executed upon exit void cleanup(void) { printf("cleaning up\n"); } void fork6() { atexit(cleanup); fork(); exit(0); }

Carnegie Mellon Zombies Idea  When process terminates, still consumes system resources  Various tables maintained by OS  Called a “zombie”  Living corpse, half alive and half dead Reaping  Performed by parent on terminated child  Parent is given exit status information  Kernel discards process What if parent doesn’t reap?  If any parent terminates without reaping a child, then child will be reaped by init process  So, only need explicit reaping in long-running processes  e.g., shells and servers

Carnegie Mellon linux>./forks 7 & [1] 6639 Running Parent, PID = 6639 Terminating Child, PID = 6640 linux> ps PID TTY TIME CMD 6585 ttyp9 00:00:00 tcsh 6639 ttyp9 00:00:03 forks 6640 ttyp9 00:00:00 forks 6641 ttyp9 00:00:00 ps linux> kill 6639 [1] Terminated linux> ps PID TTY TIME CMD 6585 ttyp9 00:00:00 tcsh 6642 ttyp9 00:00:00 ps Zombie Example ps shows child process as “defunct” Killing parent allows child to be reaped by init void fork7() { if (fork() == 0) { /* Child */ printf("Terminating Child, PID = %d\n", getpid()); exit(0); } else { printf("Running Parent, PID = %d\n", getpid()); while (1) ; /* Infinite loop */ }

Carnegie Mellon linux>./forks 8 Terminating Parent, PID = 6675 Running Child, PID = 6676 linux> ps PID TTY TIME CMD 6585 ttyp9 00:00:00 tcsh 6676 ttyp9 00:00:06 forks 6677 ttyp9 00:00:00 ps linux> kill 6676 linux> ps PID TTY TIME CMD 6585 ttyp9 00:00:00 tcsh 6678 ttyp9 00:00:00 ps Nonterminating Child Example Child process still active even though parent has terminated Must kill explicitly, or else will keep running indefinitely void fork8() { if (fork() == 0) { /* Child */ printf("Running Child, PID = %d\n", getpid()); while (1) ; /* Infinite loop */ } else { printf("Terminating Parent, PID = %d\n", getpid()); exit(0); }

Carnegie Mellon wait : Synchronizing with Children int wait(int *child_status)  suspends current process until one of its children terminates  return value is the pid of the child process that terminated  if child_status != NULL, then the object it points to will be set to a status indicating why the child process terminated

Carnegie Mellon wait : Synchronizing with Children void fork9() { int child_status; if (fork() == 0) { printf("HC: hello from child\n"); } else { printf("HP: hello from parent\n"); wait(&child_status); printf("CT: child has terminated\n"); } printf("Bye\n"); exit(); } HP HCBye CTBye

Carnegie Mellon wait() Example If multiple children completed, will take in arbitrary order Can use macros WIFEXITED and WEXITSTATUS to get information about exit status void fork10() { pid_t pid[N]; int i; int child_status; for (i = 0; i < N; i++) if ((pid[i] = fork()) == 0) exit(100+i); /* Child */ for (i = 0; i < N; i++) { pid_t wpid = wait(&child_status); if (WIFEXITED(child_status)) printf("Child %d terminated with exit status %d\n", wpid, WEXITSTATUS(child_status)); else printf("Child %d terminate abnormally\n", wpid); }

Carnegie Mellon waitpid() : Waiting for a Specific Process waitpid(pid, &status, options)  suspends current process until specific process terminates  various options (that we won’t talk about) void fork11() { pid_t pid[N]; int i; int child_status; for (i = 0; i < N; i++) if ((pid[i] = fork()) == 0) exit(100+i); /* Child */ for (i = 0; i < N; i++) { pid_t wpid = waitpid(pid[i], &child_status, 0); if (WIFEXITED(child_status)) printf("Child %d terminated with exit status %d\n", wpid, WEXITSTATUS(child_status)); else printf("Child %d terminated abnormally\n", wpid); }

Carnegie Mellon fork() and execve() How do we start a new program, instead of just a copy of the old program?  Use the UNIX execve() system call  int execve(const char *filename, char *const argv [], char *const envp[]);  filename: name of executable file to run  argv: Command line arguments  envp: environment variable settings (e.g., $PATH, $HOME, etc.)‏

Carnegie Mellon fork() and execve() execve() does not fork a new process!  Rather, it replaces the address space and CPU state of the current process  Loads the new address space from the executable file and starts it from main()‏  So, to start a new program, use fork() followed by execve()‏

Carnegie Mellon execve : Loading and Running Programs int execve( char *filename, char *argv[], char *envp ) Loads and runs  Executable filename  With argument list argv  And environment variable list envp Does not return (unless error) Overwrites process, keeps pid Environment variables:  “name=value” strings Null-terminated environment variable strings unused Null-terminated commandline arg strings envp[n] = NULL envp[n-1] envp[0] … Linker vars argv[argc] = NULL argv[argc-1] argv[0] … envp argc argv Stack 0xbfffffff

Carnegie Mellon execve : Example envp[n] = NULL envp[n-1] envp[0] … argv[argc] = NULL argv[argc-1] argv[0] … “ls” “-lt” “/usr/include” “USER=droh” “PRINTER=iron” “PWD=/usr/droh”

Carnegie Mellon execl and exec Family int execl(char *path, char *arg0, char *arg1, …, 0) Loads and runs executable at path with args arg0, arg1, …  path is the complete path of an executable object file  By convention, arg0 is the name of the executable object file  “Real” arguments to the program start with arg1, etc.  List of args is terminated by a (char *)0 argument  Environment taken from char **environ, which points to an array of “name=value” strings:  USER=ganger  LOGNAME=ganger  HOME=/afs/cs.cmu.edu/user/ganger Returns -1 if error, otherwise doesn’t return! Family of functions includes execv, execve (base function), execvp, execl, execle, and execlp

Carnegie Mellon exec : Loading and Running Programs main() { if (fork() == 0) { execl("/usr/bin/cp", "cp", "foo", "bar", 0); } wait(NULL); printf("copy completed\n"); exit(); }

Carnegie Mellon Summary Exceptions  Events that require nonstandard control flow  Generated externally (interrupts) or internally (traps and faults) Processes  At any given time, system has multiple active processes  Only one can execute at a time, though  Each process appears to have total control of processor + private memory space

Carnegie Mellon Summary (cont.) Spawning processes  Call to fork  One call, two returns Process completion  Call exit  One call, no return Reaping and waiting for Processes  Call wait or waitpid Loading and running Programs  Call execl (or variant)  One call, (normally) no return