On Sun, Sep 15, 2024 at 06:11:14PM +0900, Masami Hiramatsu (Google) wrote:
> From: Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>
>
> Add ftrace_fill_perf_regs() which should be compatible with the
> perf_fetch_caller_regs(). In other words, the pt_regs returned from the
> ftrace_fill_perf_regs() must satisfy 'user_mode(regs) == false' and can be
> used for stack tracing.
>
> Signed-off-by: Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>
...
> arch/s390/include/asm/ftrace.h | 5 +++++
...
> diff --git a/arch/s390/include/asm/ftrace.h b/arch/s390/include/asm/ftrace.h
> index 9cdd48a46bf7..0d9f6df21f81 100644
> --- a/arch/s390/include/asm/ftrace.h
> +++ b/arch/s390/include/asm/ftrace.h
> @@ -84,6 +84,11 @@ ftrace_regs_get_frame_pointer(struct ftrace_regs *fregs)
> return sp[0]; /* return backchain */
> }
>
> +#define arch_ftrace_fill_perf_regs(fregs, _regs) do { \
> + (_regs)->psw.addr = (fregs)->regs.psw.addr; \
> + (_regs)->gprs[15] = (fregs)->regs.gprs[15]; \
> + } while (0)
After reading your commit description and looking at the code I think the
s390 implementation of perf_arch_fetch_caller_regs() is at least
suboptimal: it is not possible to tell if an address belongs to user or
kernel space by looking only at the address (psw.addr); this also requires
to look at psw.mask. It _looks_ like all current callers of
perf_arch_fetch_caller_regs() initialize the passed in pt_regs to zero,
which of course also sets psw.mask to zero, and therefore
user_mode(regs) == false is satisfied.
However I'd rather make that explicit and don't want to rely on
callers. Therefore the above arch_ftrace_fill_perf_regs() should
set the mask to zero
(_regs)->psw.mask = 0;
I will change perf_arch_fetch_caller_regs() accordingly.
