System calls are special function calls to access kernel functionality in a controlled and managed way.

They allow the processor to enter kernel mode and exit back into user mode without the user needing to explicitly implement that functionality. They also prevent access of secure locations

Fetch-Execute Cycle

Fetch
Set PC (??? before or after ???)
Execute

x86 - Privileged-only Instructions

cli - clear interrupt (disable interrupts)
sti - set interrupt (enable interrupts)

Only executes in kernel mode.

Divide the memory into a kernel-only region, and a shared region.

Steps to make a System Call

Push arguments to the stack
Set up relevant registers
Execute syscall call
Kernel Trap!!!
- We are now in Kernel Space (the OS)
Dispatcher checks trapframe
Do whatever needs to be done for the syscall
Return to caller
Increment SP

User -> Privileged Mode change

Processor mode switch
Stack pointer switch
Program counter switch
Registers used to pass data to and from kernel.
Memory used to pass data to and from kernel.

MIPS R3000

Has a co-processor (CP0) that manages the state of the main processor.

c0_cause register - cause of the most recent exception
c0_status register - current status of the CPU
c0_epc register - address of the instruction that triggered an exception

Accessing CP0 is done through the two kernel mode instructions mtc0 (move to c0), and mfc0 (move from c0).

Important Bits

c0_status

c0_cause

Important Exception Codes
- 0 - Interrupt
- 8 - Syscall
General Exception Vector is at 0x8000 0800

Returning from an Exception

Set the PC to the value of the EPC (using the jr instruction)
Return from the exception (using the rfe instruction)

OS/161 System Calls

Arguments are passed and returned via normal C function calling convention.

When a syscall returns successfully:
- Register a3 = 0 on success
Register a3 is non-zero if a failure has occurred
- v0 contains the error number
- v0 is then stored in errno
- v0 set to -1

// TODO: Slides

Contents