Re: [PATCH v4 1/2] function_graph: Support recording and printing the return value of function

["Russell King (Oracle)" <linux@armlinux.org.uk>]

On Wed, Mar 15, 2023 at 06:39:10AM -0700, Donglin Peng wrote:
> diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
> index e24a9820e12f..ad03fc868f34 100644
> --- a/arch/arm/Kconfig
> +++ b/arch/arm/Kconfig
> @@ -99,6 +99,7 @@ config ARM
>  	select HAVE_FTRACE_MCOUNT_RECORD if !XIP_KERNEL
>  	select HAVE_FUNCTION_ERROR_INJECTION
>  	select HAVE_FUNCTION_GRAPH_TRACER
> +	select HAVE_FUNCTION_GRAPH_RETVAL if HAVE_FUNCTION_GRAPH_TRACER
>  	select HAVE_FUNCTION_TRACER if !XIP_KERNEL
>  	select HAVE_GCC_PLUGINS
>  	select HAVE_HW_BREAKPOINT if PERF_EVENTS && (CPU_V6 || CPU_V6K || CPU_V7)
> diff --git a/arch/arm/kernel/entry-ftrace.S b/arch/arm/kernel/entry-ftrace.S
> index 3e7bcaca5e07..0151d2ce9958 100644
> --- a/arch/arm/kernel/entry-ftrace.S
> +++ b/arch/arm/kernel/entry-ftrace.S
> @@ -258,6 +258,10 @@ ENDPROC(ftrace_graph_regs_caller)
>  #ifdef CONFIG_FUNCTION_GRAPH_TRACER
>  ENTRY(return_to_handler)
>  	stmdb	sp!, {r0-r3}
> +#ifdef CONFIG_FUNCTION_GRAPH_RETVAL
> +	/* Pass the function return value to ftrace_return_to_handler */
> +	mov	r1, r0

In a similar vein to Peter's comment, do we care about 64-bit return
values here, because the above only covers 32-bit values.

If we do care about 64-bit values, then we get into EABI/OABI
stickyness, because on EABI the 64-bit value would have to be passed
in r2,r3, and OABI would need r1,r2.

it would be better to have the 64-bit argument as the first argument
to ftrace_return_to_handler() which would eliminate that variability,
but I don't know what effect that would have for other architectures.

Things get more icky if we want 128-bit values. For EABI, we've
conveniently just stacked that. For OABI, that would need to be in
r1-r3 and the final high bits on the stack.

With a 128-bit argument as the first, that would be r0-r3 with the
existing stack pointer argument stored... on the stack.

So, really it depends what size of return value we want to report.
Also, please bear in mind that where a function returns a 32-bit
value, that will be in r0, and r1 will be whatever happened to be
in it at function exit - there's no defined value for r1.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!