Knox Shell - Operating Systems and Networks (IIIT Hyderabad)

Functionality

Prompt with username, hostname and directory
Basic commands : cd, pwd, echo, ls, repeat, pinfo
Ability to run external commands as foreground and background

Installation

Run make to build the executable
Run make run or ./vsh to start the shell

File structure

Functionality	Files
Prompt	prompt.c prompt.h
cd, pwd, echo, repeat, jobs, fg, bg, sig, replay	builtins.c builtins.h
ls	ls.c ls.h
Baywatch	baywatch.c baywatch.h
Input parser	parse.c parse.h
External commands	execute.c execute.h
Background jobs	jobs.c jobs.h
Data structure for commands	args.c args.h
Error handling, etc	utils.c utils.h
History	history.c history.h
Up arrow key	rawio.c rawio.h
IO Redirection	execute.c execute.h
Signal Handling	Handled when process is set as fg in execute.c

Specification 1: Syscalls

`strace`

The proc.h file was modified to add a new field, int strace_m, to the proc structure. This field stores the mask that indicates which syscalls are to be traced; it is initialised to 0 by default in freeproc().

Now, the sys_trace() syscall, when called, sets the strace_m field of the calling process p to its first argument, the mask.

All syscalls have been modified to check the corresponding bit of the strace_m field of the calling process and print accordingly, just before they return. For example, in the case of sbrk(), we have added

if (p->strace_m & (1 << SYS_sbrk))
  printf("%d: syscall sbrk (%d) -> %d\n", p->pid, n, addr);

`sigalarm`

sigalarm() proceeds by first saving the frequency n and the handler periodic_fn_ptr, and then saving the trapframe of the calling process. It then acquires the lock of the process, and if its runtime is a multiple of n, it changes the instruction pointer of the trapframe to the pointer of the handler.

sigreturn(), correspondingly, restores the saved trapframe and increments the instruction pointer so it can continue where it left off.

Specification 2: Scheduling Algorithms

First Come First Served

If the FCFS macro is defined, the scheduler first makes a pass through the proc list to determine the runnable process with the earliest creation time. If no process is runnable, here to_be_run remains at its initial value 0 and the function skips all the following steps (effectively, it continues to search proc[] for runnable processes).

Once the process is identified, the scheduler acquires its lock and checks if it is still runnable. If it is, it runs it until it stops. At this point, it resets the proc field of the CPU and goes back to searching proc[] for runnable processes.

Additionally, we ensure that timer-interrupts (in trap.c) are checked for only when FCFS is not the scheduler. This is to avoid preemption.

Lottery-Based Scheduler

This scheduler also works in a straightforward way. First, it passes over proc[] to find the total number of tickets assigned to all runnable processes, and then finds a random ticket in the range [0, total_tickets).

Then, it finds which process this ticket belongs to in another pass over the runnable processes in proc[]. This is where to_be_run is assigned. The last part of the function, as in the other algorithms, acquires the lock, checks the state, and runs it. Since clock interrupts are checked for in this algorithm, it is preemptive.

Additionally, the sys_settickets() syscall has been implemented, to set the number of tickets of the calling process. The allocproc() function has been modified to initialise the tickets field of the proc structure to 1, and the fork() function to initialise the tickets field of np (the child process) from that of p (the parent process).

Priority Based Scheduling

This scheduler works on the basis of assigning priorities to each process. It runs over proc[] and selects the process with the highest priority.

The get_priority function helps assigning a priority to a process based on the static priority, dynamic priority and niceness score as mentioned in the specifications. Ties are broken by seeing the number of times a process is scheduled, and the start time of a process.

The set_priority syscall is also implemented which also resets niceness score to 5, it returns the old static priority of a process . There is also a setpriority user program which uses the syscall to change the priority. It can be used by calling setpriority <priority> <pid>.

victorknox/Knox-Shell