Date:	Fri, 21 Mar 2014 15:11:26 +0000
From:	Mark Rutland <mark.rutland@....com>
To:	Jean Pihet <jean.pihet@...aro.org>
Cc:	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"linaro-kernel@...ts.linaro.org" <linaro-kernel@...ts.linaro.org>,
	"linux-arm-kernel@...ts.infradead.org" 
	<linux-arm-kernel@...ts.infradead.org>,
	Arnaldo <acme@...stprotocols.net>,
	Ingo Molnar <mingo@...nel.org>, Jiri Olsa <jolsa@...hat.com>,
	Steve Capper <steve.capper@...aro.org>,
	"patches@...aro.org" <patches@...aro.org>,
	Corey Ashford <cjashfor@...ux.vnet.ibm.com>,
	Frederic Weisbecker <fweisbec@...il.com>,
	Namhyung Kim <namhyung@...nel.org>,
	Paul Mackerras <paulus@...ba.org>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	David Ahern <dsahern@...il.com>
Subject: Re: [PATCH 1/3] perf tests: Introduce perf_regs_load function on
 ARM64

Hi Jean,

On Wed, Mar 19, 2014 at 09:42:33AM +0000, Jean Pihet wrote:
> Introducing perf_regs_load function, which is going
> to be used for dwarf unwind test in following patches.
> 
> It takes single argument as a pointer to the regs dump
> buffer and populates it with current registers values, as
> expected by the perf built-in unwinding test.
> 
> Signed-off-by: Jean Pihet <jean.pihet@...aro.org>
> Cc: Steve Capper <steve.capper@...aro.org>
> Cc: Corey Ashford <cjashfor@...ux.vnet.ibm.com>
> Cc: Frederic Weisbecker <fweisbec@...il.com>
> Cc: Ingo Molnar <mingo@...nel.org>
> Cc: Namhyung Kim <namhyung@...nel.org>
> Cc: Paul Mackerras <paulus@...ba.org>
> Cc: Peter Zijlstra <a.p.zijlstra@...llo.nl>
> Cc: Arnaldo Carvalho de Melo <acme@...stprotocols.net>
> Cc: David Ahern <dsahern@...il.com>
> Cc: Jiri Olsa <jolsa@...hat.com>
> ---
>  tools/perf/arch/arm64/Makefile            |  1 +
>  tools/perf/arch/arm64/include/perf_regs.h |  2 ++
>  tools/perf/arch/arm64/tests/regs_load.S   | 39 +++++++++++++++++++++++++++++++
>  3 files changed, 42 insertions(+)
>  create mode 100644 tools/perf/arch/arm64/tests/regs_load.S
> 
> diff --git a/tools/perf/arch/arm64/Makefile b/tools/perf/arch/arm64/Makefile
> index 67e9b3d..9b8f87e 100644
> --- a/tools/perf/arch/arm64/Makefile
> +++ b/tools/perf/arch/arm64/Makefile
> @@ -4,4 +4,5 @@ LIB_OBJS += $(OUTPUT)arch/$(ARCH)/util/dwarf-regs.o
>  endif
>  ifndef NO_LIBUNWIND
>  LIB_OBJS += $(OUTPUT)arch/$(ARCH)/util/unwind-libunwind.o
> +LIB_OBJS += $(OUTPUT)arch/$(ARCH)/tests/regs_load.o
>  endif
> diff --git a/tools/perf/arch/arm64/include/perf_regs.h b/tools/perf/arch/arm64/include/perf_regs.h
> index 2359546..1e052f1 100644
> --- a/tools/perf/arch/arm64/include/perf_regs.h
> +++ b/tools/perf/arch/arm64/include/perf_regs.h
> @@ -9,6 +9,8 @@
>  #define PERF_REG_IP	PERF_REG_ARM64_PC
>  #define PERF_REG_SP	PERF_REG_ARM64_SP
>  
> +void perf_regs_load(u64 *regs);
> +
>  static inline const char *perf_reg_name(int id)
>  {
>  	switch (id) {
> diff --git a/tools/perf/arch/arm64/tests/regs_load.S b/tools/perf/arch/arm64/tests/regs_load.S
> new file mode 100644
> index 0000000..92ab968
> --- /dev/null
> +++ b/tools/perf/arch/arm64/tests/regs_load.S
> @@ -0,0 +1,39 @@
> +#include <linux/linkage.h>
> +
> +/*
> + * Implementation of void perf_regs_load(u64 *regs);
> + *
> + * This functions fills in the 'regs' buffer from the actual registers values,
> + * in the way the perf built-in unwinding test expects them:
> + * - the PC at the time at the call to this function. Since this function
> + *   is called using a bl instruction, the PC value is taken from LR,

Is it guaranteed that this function is always invoked with a branch with
link instruction, or is that just what current compiler versions are
doing? I couldn't see where we would get that guarantee from.

If it is called with a branch with link, then the LR value will be the
PC at call time + 4, rather than just the exact PC at call time. If not
then we don't have a guaranteed relationship between the PC at call time
and the current LR value.

If the only place that perf_regs_load is used is a single test which
doesn't care about the precise PC at the time of the call, then it's
probably OK to use the LR value, but we should be careful to document
what the faked-up PC actually is and how we expect it to be used.

> + * - the current SP (not touched by this function),
> + * - the current value of LR is merely retrieved and stored because the
> + *   value before the call to this function is unknown at this time; it will
> + *   be unwound from the dwarf information in unwind__get_entries.
> + */
> +
> +.text
> +.type perf_regs_load,%function
> +ENTRY(perf_regs_load)
> +	stp x0,  x1,  [x0], #16	// store x0..x29
> +	stp x2,  x3,  [x0], #16
> +	stp x4,  x5,  [x0], #16
> +	stp x6,  x7,  [x0], #16
> +	stp x8,  x9,  [x0], #16
> +	stp x10, x11, [x0], #16
> +	stp x12, x13, [x0], #16
> +	stp x14, x15, [x0], #16
> +	stp x16, x17, [x0], #16
> +	stp x18, x19, [x0], #16
> +	stp x20, x21, [x0], #16
> +	stp x22, x23, [x0], #16
> +	stp x24, x25, [x0], #16
> +	stp x26, x27, [x0], #16
> +	stp x28, x29, [x0], #16
> +	mov x1,  sp
> +	stp x30, x1,  [x0], #16	// store lr and sp
> +	str x30, [x0]		// store pc as lr in order to skip the call
> +				//  to this function

It might be better to word this a "store the lr in place of the pc". To
me at least the current wording implies the opposite of what the code
seems to be doing.

Cheers,
Mark.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists