[PATCH v2] riscv: Improve exception and system call latency

[Jisheng Zhang]

On Thu, Jun 06, 2024 at 11:13:35PM -0700, Cyril Bur wrote:
> From: Anton Blanchard <antonb at tenstorrent.com>
> 
> Many CPUs implement return address branch prediction as a stack. The
> RISCV architecture refers to this as a return address stack (RAS). If
> this gets corrupted then the CPU will mispredict at least one but
> potentally many function returns.
> 
> There are two issues with the current RISCV exception code:
> 
> - We are using the alternate link stack (x5/t0) for the indirect branch
>   which makes the hardware think this is a function return. This will
>   corrupt the RAS.
> 
> - We modify the return address of handle_exception to point to
>   ret_from_exception. This will also corrupt the RAS.
> 
> Testing the null system call latency before and after the patch:
> 
> Visionfive2 (StarFive JH7110 / U74)
> baseline: 189.87 ns
> patched:  176.76 ns
> 
> Lichee pi 4a (T-Head TH1520 / C910)
> baseline: 666.58 ns
> patched:  636.90 ns
> 
> Just over 7% on the U74 and just over 4% on the C910.
> 
> Signed-off-by: Anton Blanchard <antonb at tenstorrent.com>
> Signed-off-by: Cyril Bur <cyrilbur at tenstorrent.com>

Tested-by: Jisheng Zhang <jszhang at kernel.org>
Reviewed-by: Jisheng Zhang <jszhang at kernel.org>
> ---
> v2:
>    Simplify jalr ra,t1 to jalr t1
>    Drop extra .globl from entry.S and use ra == handle_exception
> ---
>  arch/riscv/kernel/entry.S      | 17 ++++++++++-------
>  arch/riscv/kernel/stacktrace.c |  4 ++--
>  2 files changed, 12 insertions(+), 9 deletions(-)
> 
> diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> index 68a24cf9481a..c933460ed3e9 100644
> --- a/arch/riscv/kernel/entry.S
> +++ b/arch/riscv/kernel/entry.S
> @@ -88,7 +88,6 @@ SYM_CODE_START(handle_exception)
>  	call riscv_v_context_nesting_start
>  #endif
>  	move a0, sp /* pt_regs */
> -	la ra, ret_from_exception
>  
>  	/*
>  	 * MSB of cause differentiates between
> @@ -97,7 +96,8 @@ SYM_CODE_START(handle_exception)
>  	bge s4, zero, 1f
>  
>  	/* Handle interrupts */
> -	tail do_irq
> +	call do_irq
> +	j ret_from_exception
>  1:
>  	/* Handle other exceptions */
>  	slli t0, s4, RISCV_LGPTR
> @@ -105,11 +105,14 @@ SYM_CODE_START(handle_exception)
>  	la t2, excp_vect_table_end
>  	add t0, t1, t0
>  	/* Check if exception code lies within bounds */
> -	bgeu t0, t2, 1f
> -	REG_L t0, 0(t0)
> -	jr t0
> -1:
> -	tail do_trap_unknown
> +	bgeu t0, t2, 3f
> +	REG_L t1, 0(t0)
> +2:	jalr t1
> +	j ret_from_exception
> +3:
> +
> +	la t1, do_trap_unknown
> +	j 2b
>  SYM_CODE_END(handle_exception)
>  ASM_NOKPROBE(handle_exception)
>  
> diff --git a/arch/riscv/kernel/stacktrace.c b/arch/riscv/kernel/stacktrace.c
> index 528ec7cc9a62..5eb3d135b717 100644
> --- a/arch/riscv/kernel/stacktrace.c
> +++ b/arch/riscv/kernel/stacktrace.c
> @@ -16,7 +16,7 @@
>  
>  #ifdef CONFIG_FRAME_POINTER
>  
> -extern asmlinkage void ret_from_exception(void);
> +extern asmlinkage void handle_exception(void);
>  
>  static inline int fp_is_valid(unsigned long fp, unsigned long sp)
>  {
> @@ -70,7 +70,7 @@ void notrace walk_stackframe(struct task_struct *task, struct pt_regs *regs,
>  			fp = frame->fp;
>  			pc = ftrace_graph_ret_addr(current, NULL, frame->ra,
>  						   &frame->ra);
> -			if (pc == (unsigned long)ret_from_exception) {
> +			if (pc == (unsigned long)handle_exception) {
>  				if (unlikely(!__kernel_text_address(pc) || !fn(arg, pc)))
>  					break;
>  
> -- 
> 2.25.1
> 
> 
> _______________________________________________
> linux-riscv mailing list
> linux-riscv at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv