[PATCH V4 2/2] x86/perf/intel/uncore: Make Intel rapl modular

From: kan . liang
Date: Sun Mar 20 2016 - 11:59:37 EST


From: Kan Liang <kan.liang@xxxxxxxxx>

This patch makes Intel rapl pmu driver modular.
By default, the rapl driver will be built in the kernel. If it is
configured as module, the supported cpu model can be auto loaded.
This patch also cleans up the code of rapl_pmu_init.

Based-on-a-patch-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Signed-off-by: Kan Liang <kan.liang@xxxxxxxxx>
Reviewed-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
---

Changes since V3:
- The driver needs to migrate its stuff away before the core
code gets notified about the cpu going down

Changes since V2:
- Add const __initconst for all struct intel_rapl_init_fun

Changes since V1:
- should use _begin _done, not _done _done.

Changes since Thomas Gleixner's version
- Move codes to new location
- The module can be autoprobe/load.
- Clean up the code of rapl_pmu_init.

arch/x86/Kconfig.perf | 8 +++
arch/x86/events/intel/Makefile | 3 +-
arch/x86/events/intel/rapl.c | 121 ++++++++++++++++++++++++++++-------------
3 files changed, 94 insertions(+), 38 deletions(-)

diff --git a/arch/x86/Kconfig.perf b/arch/x86/Kconfig.perf
index 90b7f58..b239ad5 100644
--- a/arch/x86/Kconfig.perf
+++ b/arch/x86/Kconfig.perf
@@ -8,4 +8,12 @@ config PERF_EVENTS_INTEL_UNCORE
Include support for Intel uncore performance events. These are
available on NehalemEX and more modern processors.

+config PERF_EVENTS_INTEL_RAPL
+ tristate "Intel rapl performance events"
+ depends on PERF_EVENTS && CPU_SUP_INTEL && PCI
+ default y
+ ---help---
+ Include support for Intel rapl performance events for power
+ monitoring on modern processors.
+
endmenu
diff --git a/arch/x86/events/intel/Makefile b/arch/x86/events/intel/Makefile
index a6c7448..27adbba 100644
--- a/arch/x86/events/intel/Makefile
+++ b/arch/x86/events/intel/Makefile
@@ -1,6 +1,7 @@
obj-$(CONFIG_CPU_SUP_INTEL) += core.o bts.o cqm.o
obj-$(CONFIG_CPU_SUP_INTEL) += cstate.o ds.o knc.o
obj-$(CONFIG_CPU_SUP_INTEL) += lbr.o p4.o p6.o pt.o
-obj-$(CONFIG_CPU_SUP_INTEL) += rapl.o
+obj-$(CONFIG_PERF_EVENTS_INTEL_RAPL) += intel-rapl.o
+intel-rapl-objs := rapl.o
obj-$(CONFIG_PERF_EVENTS_INTEL_UNCORE) += intel-uncore.o
intel-uncore-objs := uncore.o uncore_nhmex.o uncore_snb.o uncore_snbep.o
diff --git a/arch/x86/events/intel/rapl.c b/arch/x86/events/intel/rapl.c
index b834a3f..5fb71b0 100644
--- a/arch/x86/events/intel/rapl.c
+++ b/arch/x86/events/intel/rapl.c
@@ -53,6 +53,8 @@
#include <asm/cpu_device_id.h>
#include "../perf_event.h"

+MODULE_LICENSE("GPL");
+
/*
* RAPL energy status counters
*/
@@ -592,6 +594,15 @@ static int rapl_cpu_notifier(struct notifier_block *self,
return NOTIFY_OK;
}

+static struct notifier_block rapl_cpu_nb = {
+ .notifier_call = rapl_cpu_notifier,
+ /*
+ * to migrate rapl events, our notifier should be executed
+ * before perf core's notifier.
+ */
+ .priority = CPU_PRI_PERF + 1,
+};
+
static int rapl_check_hw_unit(bool apply_quirk)
{
u64 msr_rapl_power_unit_bits;
@@ -660,7 +671,7 @@ static int __init rapl_prepare_cpus(void)
return 0;
}

-static void __init cleanup_rapl_pmus(void)
+static void cleanup_rapl_pmus(void)
{
int i;

@@ -691,49 +702,75 @@ static int __init init_rapl_pmus(void)
return 0;
}

+#define X86_RAPL_MODEL_MATCH(model, init) \
+ { X86_VENDOR_INTEL, 6, model, X86_FEATURE_ANY, (unsigned long)&init }
+
+struct intel_rapl_init_fun {
+ bool apply_quirk;
+ int cntr_mask;
+ struct attribute **attrs;
+};
+
+static const struct intel_rapl_init_fun snb_rapl_init __initconst = {
+ .apply_quirk = false,
+ .cntr_mask = RAPL_IDX_CLN,
+ .attrs = rapl_events_cln_attr,
+};
+
+static const struct intel_rapl_init_fun hsx_rapl_init __initconst = {
+ .apply_quirk = true,
+ .cntr_mask = RAPL_IDX_SRV,
+ .attrs = rapl_events_srv_attr,
+};
+
+static const struct intel_rapl_init_fun hsw_rapl_init __initconst = {
+ .apply_quirk = false,
+ .cntr_mask = RAPL_IDX_HSW,
+ .attrs = rapl_events_hsw_attr,
+};
+
+static const struct intel_rapl_init_fun snbep_rapl_init __initconst = {
+ .apply_quirk = false,
+ .cntr_mask = RAPL_IDX_SRV,
+ .attrs = rapl_events_srv_attr,
+};
+
+static const struct intel_rapl_init_fun knl_rapl_init __initconst = {
+ .apply_quirk = true,
+ .cntr_mask = RAPL_IDX_KNL,
+ .attrs = rapl_events_knl_attr,
+};
+
static const struct x86_cpu_id rapl_cpu_match[] __initconst = {
- [0] = { .vendor = X86_VENDOR_INTEL, .family = 6 },
- [1] = {},
+ X86_RAPL_MODEL_MATCH(42, snb_rapl_init), /* Sandy Bridge */
+ X86_RAPL_MODEL_MATCH(58, snb_rapl_init), /* Ivy Bridge */
+ X86_RAPL_MODEL_MATCH(63, hsx_rapl_init), /* Haswell-Server */
+ X86_RAPL_MODEL_MATCH(60, hsw_rapl_init), /* Haswell */
+ X86_RAPL_MODEL_MATCH(69, hsw_rapl_init), /* Haswell-Celeron */
+ X86_RAPL_MODEL_MATCH(61, hsw_rapl_init), /* Broadwell */
+ X86_RAPL_MODEL_MATCH(45, snbep_rapl_init), /* Sandy Bridge-EP */
+ X86_RAPL_MODEL_MATCH(62, snbep_rapl_init), /* IvyTown */
+ X86_RAPL_MODEL_MATCH(87, knl_rapl_init), /* Knights Landing */
+ {},
};

+MODULE_DEVICE_TABLE(x86cpu, rapl_cpu_match);
+
static int __init rapl_pmu_init(void)
{
- bool apply_quirk = false;
+ const struct x86_cpu_id *id;
+ struct intel_rapl_init_fun *rapl_init;
+ bool apply_quirk;
int ret;

- if (!x86_match_cpu(rapl_cpu_match))
+ id = x86_match_cpu(rapl_cpu_match);
+ if (!id)
return -ENODEV;

- switch (boot_cpu_data.x86_model) {
- case 42: /* Sandy Bridge */
- case 58: /* Ivy Bridge */
- rapl_cntr_mask = RAPL_IDX_CLN;
- rapl_pmu_events_group.attrs = rapl_events_cln_attr;
- break;
- case 63: /* Haswell-Server */
- apply_quirk = true;
- rapl_cntr_mask = RAPL_IDX_SRV;
- rapl_pmu_events_group.attrs = rapl_events_srv_attr;
- break;
- case 60: /* Haswell */
- case 69: /* Haswell-Celeron */
- case 61: /* Broadwell */
- rapl_cntr_mask = RAPL_IDX_HSW;
- rapl_pmu_events_group.attrs = rapl_events_hsw_attr;
- break;
- case 45: /* Sandy Bridge-EP */
- case 62: /* IvyTown */
- rapl_cntr_mask = RAPL_IDX_SRV;
- rapl_pmu_events_group.attrs = rapl_events_srv_attr;
- break;
- case 87: /* Knights Landing */
- apply_quirk = true;
- rapl_cntr_mask = RAPL_IDX_KNL;
- rapl_pmu_events_group.attrs = rapl_events_knl_attr;
- break;
- default:
- return -ENODEV;
- }
+ rapl_init = (struct intel_rapl_init_fun *)id->driver_data;
+ apply_quirk = rapl_init->apply_quirk;
+ rapl_cntr_mask = rapl_init->cntr_mask;
+ rapl_pmu_events_group.attrs = rapl_init->attrs;

ret = rapl_check_hw_unit(apply_quirk);
if (ret)
@@ -753,7 +790,7 @@ static int __init rapl_pmu_init(void)
if (ret)
goto out;

- __perf_cpu_notifier(rapl_cpu_notifier);
+ __register_cpu_notifier(&rapl_cpu_nb);
cpu_notifier_register_done();
rapl_advertise();
return 0;
@@ -764,4 +801,14 @@ out:
cpu_notifier_register_done();
return ret;
}
-device_initcall(rapl_pmu_init);
+module_init(rapl_pmu_init);
+
+static void __exit intel_rapl_exit(void)
+{
+ cpu_notifier_register_begin();
+ __unregister_cpu_notifier(&rapl_cpu_nb);
+ perf_pmu_unregister(&rapl_pmus->pmu);
+ cleanup_rapl_pmus();
+ cpu_notifier_register_done();
+}
+module_exit(intel_rapl_exit);
--
2.5.0