[PATCH] perf, x86: Widen Haswell OFFCORE mask

[PATCH] perf, x86: Widen Haswell OFFCORE mask

[Andi Kleen]

From: Andi Kleen <ak@linux.intel.com>

Haswell supports more bits in the offcore_rsp_* MSRs than Sandy
Bridge. Previously the Haswell code was using the Sandy Bridge
extra register definitions, which prevented users from setting
all of these bits. This in term did not allow to set some valid
SNOOP_* bits, among others.

I allowed all bits the CPU does not #GP on. This is ok because
it's protected by a model check.

Add a new extra_regs table for Haswell and use. Except for the
widened mask it is identical to Sandy Bridge.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
---
 arch/x86/kernel/cpu/perf_event_intel.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index adb02aa..5d41e2b 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -189,6 +189,14 @@ static struct extra_reg intel_snbep_extra_regs[] __read_mostly = {
 	EVENT_EXTRA_END
 };
 
+static struct extra_reg intel_hsw_extra_regs[] __read_mostly = {
+	/* must define OFFCORE_RSP_X first, see intel_fixup_er() */
+	INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0x3fffff8fffull, RSP_0),
+	INTEL_UEVENT_EXTRA_REG(0x01bb, MSR_OFFCORE_RSP_1, 0x3fffff8fffull, RSP_1),
+	INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(0x01cd),
+	EVENT_EXTRA_END
+};
+
 EVENT_ATTR_STR(mem-loads,	mem_ld_nhm,	"event=0x0b,umask=0x10,ldlat=3");
 EVENT_ATTR_STR(mem-loads,	mem_ld_snb,	"event=0xcd,umask=0x1,ldlat=3");
 EVENT_ATTR_STR(mem-stores,	mem_st_snb,	"event=0xcd,umask=0x2");
@@ -2504,7 +2512,7 @@ __init int intel_pmu_init(void)
 
 		x86_pmu.event_constraints = intel_hsw_event_constraints;
 		x86_pmu.pebs_constraints = intel_hsw_pebs_event_constraints;
-		x86_pmu.extra_regs = intel_snb_extra_regs;
+		x86_pmu.extra_regs = intel_hsw_extra_regs;
 		x86_pmu.pebs_aliases = intel_pebs_aliases_snb;
 		/* all extra regs are per-cpu when HT is on */
 		x86_pmu.er_flags |= ERF_HAS_RSP_1;
-- 
1.9.3

Re: [PATCH] perf, x86: Widen Haswell OFFCORE mask

[Stephane Eranian]

On Sat, Jul 12, 2014 at 2:06 AM, Andi Kleen <andi@firstfloor.org> wrote:
> From: Andi Kleen <ak@linux.intel.com>
>
> Haswell supports more bits in the offcore_rsp_* MSRs than Sandy
> Bridge. Previously the Haswell code was using the Sandy Bridge
> extra register definitions, which prevented users from setting
> all of these bits. This in term did not allow to set some valid
> SNOOP_* bits, among others.
>
> I allowed all bits the CPU does not #GP on. This is ok because
> it's protected by a model check.
>
> Add a new extra_regs table for Haswell and use. Except for the
> widened mask it is identical to Sandy Bridge.
>
You're adding an extra_regs[] array which is identical to the one
used by SNB-EP. So why not use that one instead of defining yet
another extra_regs array?


> Signed-off-by: Andi Kleen <ak@linux.intel.com>
> ---
>  arch/x86/kernel/cpu/perf_event_intel.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
> index adb02aa..5d41e2b 100644
> --- a/arch/x86/kernel/cpu/perf_event_intel.c
> +++ b/arch/x86/kernel/cpu/perf_event_intel.c
> @@ -189,6 +189,14 @@ static struct extra_reg intel_snbep_extra_regs[] __read_mostly = {
>         EVENT_EXTRA_END
>  };
>
> +static struct extra_reg intel_hsw_extra_regs[] __read_mostly = {
> +       /* must define OFFCORE_RSP_X first, see intel_fixup_er() */
> +       INTEL_UEVENT_EXTRA_REG(0x01b7, MSR_OFFCORE_RSP_0, 0x3fffff8fffull, RSP_0),
> +       INTEL_UEVENT_EXTRA_REG(0x01bb, MSR_OFFCORE_RSP_1, 0x3fffff8fffull, RSP_1),
> +       INTEL_UEVENT_PEBS_LDLAT_EXTRA_REG(0x01cd),
> +       EVENT_EXTRA_END
> +};
> +
>  EVENT_ATTR_STR(mem-loads,      mem_ld_nhm,     "event=0x0b,umask=0x10,ldlat=3");
>  EVENT_ATTR_STR(mem-loads,      mem_ld_snb,     "event=0xcd,umask=0x1,ldlat=3");
>  EVENT_ATTR_STR(mem-stores,     mem_st_snb,     "event=0xcd,umask=0x2");
> @@ -2504,7 +2512,7 @@ __init int intel_pmu_init(void)
>
>                 x86_pmu.event_constraints = intel_hsw_event_constraints;
>                 x86_pmu.pebs_constraints = intel_hsw_pebs_event_constraints;
> -               x86_pmu.extra_regs = intel_snb_extra_regs;
> +               x86_pmu.extra_regs = intel_hsw_extra_regs;
>                 x86_pmu.pebs_aliases = intel_pebs_aliases_snb;
>                 /* all extra regs are per-cpu when HT is on */
>                 x86_pmu.er_flags |= ERF_HAS_RSP_1;
> --
> 1.9.3
>