Processes in Minix (2.5)

Processes: internal structure

• Layer 1: catches interrupts, traps, provides interprocess communications, scheduling, saving and restoring registers, providing process abstraction
• Layer 2: tasks = device drivers, one per device type and the "system task", do not have full privileges but more than user tasks
• Layer 3: system services useful to user tasks: fork, exec, brk, read, mount, chdir
• run all the time, started by the system
• processes have user-level privileges
• Layer 4: user processes, including command shells, editors, compilers, user-written (a.out type) programs

Processes: management

• process spawn other process, starting with "init", using the fork/exec cycle
• startup (floppy): HW reads first sector of first track and executes bootstrap program
• startup (disk): master boot record at (0,0) with disk partition information (up to 4 physical partitions), searching for active partition, then go to active partition and read (0,0) there for bootstrap
• bootstrap looks for the following parts:
• kernel
• memory manager
• file system
• init
• kernal starts tasks, then when they all block, starts init
• init:
• reads /etc/ttytab - lists all potential terminal devices, can be used as login terminals if there is an entry in getty field - if so, init forks a child process for each.
• child executes /usr/bin/getty prints login prompt, then runs /usr/bin/login
• success means that /usr/bin/login executes shell specified in /etc/passwd file, typically, /bin/sh or /usr/bin/ash
• shells wait for commands and fork new processes as required
• process cycle: parent does a "fork" (make a copy); one child continues as before, other sets up an environment then "exec" to swap current program for the desired program, runs the new program and eventually, "exit" (by its own hand, stopped by another process, or gets a kill signal)
• process exec and exit involve: memory manager, file system and kernel tables

Processes: interprocess communication

• C library procedures
• send (destination, message)
• receive (source, message)
• send_rec (sourceDestination, message) - blocks
• may communicate in same layer or directly below - enforced by system

Processes: scheduling

• three queues: layers 2, 3 and 4
• tasks, memory manager and file system are never preempted, may be interrupted
• user tasks run as round-robin, 100ms time slice

Implementation of Processes (2.6)

ProcImp: organization of source code (2.6.1)

ProcImp: common header files (ends in .h) (2.6.2)

• include: ansi, limits, errno, unistd, string, signal, fcntl, stdlib, termios, a.out
• include/sys: types, ioctl, sigcontext, ptrace, stat, dir, wait
• differences between 16 and 32 bit versions, primarily size of various types - see fig 2-29, pg 106

ProcImp: MINIX header files (2.6.3)

• include/minix: config, const, type, syslib, callnr, com, boot, keymap, partition
• include/ibm: partition
• Defines 6 message types in MINIX - Fig 2-30, pg 110

ProcImp: data files and header files (2.6.4)

• parts (starting page numbers)
• kernel - 570
• const - 570
• type - 573
• proto - 575
• glo - 579, global variables
• proc - 580
• protect - 582
• sconst - 584
• assert - 585
• table.c - 586, contains all kernel data structures
• mpx.s - 588, 16-bit version
• mpx386.s - 589, 32-bit version
• start.c - 599
• main.c - 601
• proc.c - 604
• exception.c - 613
• i8259.c - 614, interrupt controller
• protect.c - 616
• klib.s, klib386.s - 621
• misc.c - 632
• driver, driver.c - 635
• drvlib.c - 642
• memory.c - 645, memory for drivers
• wini.c, at_wini.c - 650, winchester drives
• clock.c - 666
• tty, tty.c - 675
• keyboard.c - 699
• console.c - 708, IBM video output
• dmp.c - 724, debug dump
• system.c - 726, system task messages

ProcImp: bootstrapping (2.6.5)

• See fig 2-31, pg 120, for a map of disk drives
• installboot - writes boot sector or boot block
• boot - secondary MINIX loader, can run as monitor program, a minimal operatins system that can be used to change some system boot parameters
• can be modified to boot other operating systems or pass parameters to minix
• can use minix file system to find files
• default: looks for /minix, or /minix/*, but can be configured to look for any file
• when done, control is passed to the kernel
• possible to boot over network - requires smart ROM code

ProcImp: system initialization (2.6.6)

• 16 or 32 bit modes available
• Start at line 6051, pg 591,
• set up stack frame for C code, tables for memory segments, processor registers
• call cstart at line 6109 to 6524, pg 599, Global and Interrupt Descriptor Tables
• end with jump to main (6131 to 6721, pg 601), kernel entry point
• configure interrupt handlers
• sets just messages to handle spurious interrupts
• init memory map of system
• set up process table with initial processes
• special processes: IDLE, HARDWARE
• transfer control to "restart" and exit
• restart, line 6322, pg 596, starts processing the entries in the process table

ProcImp: interrupt handling (2.6.7)

• Hardware diagram: fig 2-33, pg 128
• 15 distinct IRQ lines, some tied to specific hardware
• uses separate stack, stores current process info on that stack
• interrupts can be interrupted
• 32-bint Intel chip handles much of interrupt mechanism
• 15 macros to handle interrupts (lines 6164 to 6251)
• start by saving other registers
• shut down interrupter hardware
• enable other sources
• jump using jump table to handle device
• handle the interrupt, may involve scheduling processes
• finally return, disable all interrupts, reset the interrupting device
• re-enable all interrupts, then return from interrupt
• hardware can tell if returning to user mode or to another interrupt
• compare to system call - fig 2-34, pg 134
• also compare exception handling
• magic number at start of file for boot
• reserve stack space for kernel

ProcImp: interprocess communication (2.6.8)

• based on rendezvous principle:
• proc1 does a send to proc2
• lowest kernel layer checks if proc2 waiting on message from proc1 or from ANY
• if yes, message delivered to receivers buffer and both procs marked as ready to run
• else, proc1 blocks, put on queue of proc's waiting to send to proc2
• when proc2 does receive:
• if message in queue, copy to proc2 buffer and mark both as ready to run
• else block
• code in proc.h (pg 580) and proc.c (line 6900-7419, pg 604-612)
• responds to hardware and software interrupts requiring message transfers
• software only from layer 2 processes - system calls
• Summary by functions:
• handle hardware interrupt
• handle software interrupt
• check to see source and destination are valid processes
• send message as required
• receive message as required
• SIGINT, SIGQUIT and SIGALARM are sent to the memory manager if required
• details on pgs 138-140

ProcImp: scheduling (2.6.9)

• Three queues, acting as priority levels:
• TASK_Q - highest priority
• SERVER_Q - middle priority
• USER_Q - lowest priority
• linked lists with pointers to head and tail
• code in proc.c, lines 7179-7205, pg 608

ProcImp: hardware-dependent kernel support (2.6.10)

• exception.c, line 7512, pg 613, errors from user processes, see code for list
• i8259.c, line 7600-7692, pg 614, init Intel 8259 interrupt controller chips, masks used to turn on and off various interrupts
• protect.c, line 7700-7997, pg 616, handles protected modes of Intel processors:
• Global Descriptor Table, accessed through special CPU register
• Local Descriptor Table, accessed through GDT - one for each process
• Interrupt Descriptor Table, accessed through special CPU register
• Segment descriptor, see fig 2-37, pg 144
• Task State Segment - includes registers needed to restart a task
• convert variable location using segment descriptors
• used when new processes start
• back door to allow direct read/write of I/O ports through /dev/mem and /dev/kmem (pg 145)

ProcImp: utilities and the kernel library (2.6.11)

• klib.s (lines 8000-8008, pg 621) - selects machine specific version of klib
• klib386.s (lines 8100-8777, pg 621-632) - set of basic utilities to support the O/S
• monitor, check memory, copy message, exit, read byte/word, write byte/word, read/write port, enable/disable interrupts (lock/unlock, irq's also), copy memory, load segment, reset system, copy to video memory, call level0 function
• misc.c (lines 8800-8958, pg 632-635) - init memory, parse boot params, bad assertion, bad compare (last two if DEBUG)

I/O in Minix (3.4)

I/O: interrupt handlers (3.4.1)

• Actually, see interrupt handlers in 2.6.7.
  

I/O: device drivers (3.4.2)

I/O: device independent (3.4.3)

I/O: user-level (3.4.4)

I/O: deadlock handling (3.4.5)

Block Devices (3.5)

Block devices: overview of drivers (3.5.1)

Block devices: common drivers (3.5.2)

Block devices: driver library (3.5.3)

RAM Disks (3.6)

RAM Disk: hardware and software (3.6.1)

RAM Disk: overview of driver (3.6.2)

RAM Disk: implementation of driver (3.6.3)

Disks (3.7)

Disks: hardware (3.7.1)

Disks: software (3.7.2)

Disks: overview of drivers (3.7.3)

Disks: implementation (3.7.4)

Disks: floppies (3.7.5)

Clocks (3.8)

Clock: hardware (3.8.1)

Clock: software (3.8.2)

Clock: overview of driver (3.8.3)

Clock: implementation of driver (3.8.4)

Terminals (3.9)

Term: hardware (3.9.1)

Term: software (3.9.2)

Term: overview of driver (3.9.3)

Term: device independent driver (3.9.4)

Term: keyboard driver (3.9.5)

Term: display driver (3.9.6)

Memory Management (4.7)

Memory: layout (4.7.1)

Memory: handling (4.7.2)

Memory: data structures and algorithms (4.7.3)

Memory: fork, exit, wait system calls (4.7.4)

Memory: exec system call (4.7.5)

Memory: brk system call (4.7.6)

Memory: signal handling (4.7.7)

Memory: other system calls (4.7.8)

Implementation of Memory Management (4.8)

MMimp: header files and data structures (4.8.1)

MMimp: main program (4.8.2)

MMimp: fork, exit wait (4.8.3)

MMimp: exec (4.8.4)

MMimp: brk (4.8.5)

MMimp: signal handling (4.8.6)

MMimp: other system calls (4.8.7)

MMimp: utilities (4.8.8)

Minix File System (5.6)

Files: messages (5.6.1)

Files: layout (5.6.2)

Files: bit maps (5.6.3)

Files: i-nodes (5.6.4)

Files: block cache (5.6.5)

Files: directories and paths(5.6.6)

Files: file descriptors (5.6.7)

Files: locking (5.6.8)

Files: pipes and special files (5.6.9)

Files: example: read system call (5.6.10)

Implementation of the File System (5.7)

Files: header files and global data structures (5.7.1)

Files: table management (5.7.2)

Files: main program (5.7.3)

Files: operations on individual files (5.7.4)

Files: directories and paths (5.7.5)

Files: other system calls (5.7.6)

Files: I/O device interface (5.7.7)

Files: general utilities (5.7.8)