CS 452/652 Spring 2025 - Lecture 4

ARM Processor, Context Switch

May 15, 2025 prev next

AEM - AArch64 Exception model

Context Switch

• context = registers + stack
  • registers: fast operands for processor operations
  • stack: dynamic storage for automatic variables
• stack switch
  • save registers (to somewhere) without changing them
  • stack switch: save/restore SP register
  • restore registers
• mode switch (privilege level) - processor exception
  • some state automatically saved
  • usually implies stack switch for safety & security
• system call & interrupt: stack switch & mode switch
• aisde: task vs. coroutine differ in scheduling

ARMv8 Basics

• RISC - 64-bit memory, but 32-bit instructions: opcode + operands => lots of addressing modes
  • immediate constants with shifting; pc-relative, etc., restricted
  • move, arithmetic, branch, load/store (index and/or increment)

ARMv8 Processor State

• execution state: 32 vs 64 bit mode (AArch32 vs. AArch64)
• exception level: 0...3 (cf. Intel/AMD "protection ring")
  0: user program
  1: kernel
  2: hypervisor (boot)
  3: secure monitor → example: trust/DRM
• security state: normal (non-secure) vs. secure world
• register file: X0 ... X30=LR, X31=XZR, PC, SP (EL 0-3)
  • program counter, stack pointer (banked: SP_ELx)
• pstate: current processor status (implicit)
  • condition codes (N,Z,C,V), interrupt flags, execution state, exception level
  • Negative, Zero, Carry and oVerflow -> conditional branching
• named system registers; access with MRS/read, MSR/save
  • such as: ELR, ESR, SPSR, VBAR
• SP selector (SPSel): use current (banked) SP_ELx or SP_EL0
  • using SP_EL0 is unsafe

Exception Vector

• see AEM Table 4
• vectors: VBAR_ELn exceptions to EL n
  • 4 groups (from where): Current EL SP0/x, Lower EL 64/32
  • 4 vectors (32 instructions): Synchronous, IRQ, FIQ, Error
    → total of 2KB = 4 * 4 * 128 bytes
• recommendation: set up dummy handlers for everything!

System Call

• synchronous exception - this is how a task asks the kernel for something
• dedicated instruction: svc N
• ESR_EL1 holds exception code and N
• ELR_EL1 holds return address (next PC after svc)
• SPSR_EL1 hold pstate before exception
• processor in EL1 using SP_EL1 (assuming SPSel=1)
• execution continues (PC) at hard-coded handler (exception vector)
• what needs to happen next?
  • save general-purpose registers
  • save ELR_EL1, SPSR_EL1, SP_EL0
    • we might not return to same task
  • restore kernel state
  • access system call arguments?
• resume user task:
  • save kernel state
  • restore ELR_EL1, SPSR_EL1, SP_EL0
  • return from exception: eret
    • restores PC from ELR, pstate from SPSR
    • returns to EL0 (then using SP_EL0)
• initialize user task: set up stack and "fake" context and resume