[PATCH 0/3] perf: make Arm CCI driver modular

[PATCH 0/3] perf: make Arm CCI driver modular

[Robin Murphy]

Now that it has been surgically removed from the MCPM port-control code,
we can let the CCI PMU driver be modular. Probing the PMU in the first
place still depends on the bus driver stub being built-in, but it's a
small price to pay compared to the major upheaval of completely reworking
the DT-handling code across the two drivers.

Robin.


Robin Murphy (3):
  perf/arm-cc*: Fix MODULE_LICENSE() tags
  perf/arm-cci: Remove pointless PMU disabling
  perf/arm-cci: Allow building as a module

 drivers/perf/Kconfig   | 31 +++++++++++++++++--------------
 drivers/perf/arm-cci.c | 32 ++++++++++++++++++++------------
 drivers/perf/arm-ccn.c |  2 +-
 3 files changed, 38 insertions(+), 27 deletions(-)

-- 
2.17.0.dirty

[PATCH 1/3] perf/arm-cc*: Fix MODULE_LICENSE() tags

[Robin Murphy]

The CCI/CCN drivers are licensed under GPLv2, but the MODULE_LICENSE()
tags are using the bare "GPL" string implying GPLv2 or later. Fix them
to match their actual file license.

Acked-by: Pawel Moll <pawel.moll@arm.com>
Acked-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---
 drivers/perf/arm-cci.c | 2 +-
 drivers/perf/arm-ccn.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/perf/arm-cci.c b/drivers/perf/arm-cci.c
index 383b2d3dcbc6..53c774b20563 100644
--- a/drivers/perf/arm-cci.c
+++ b/drivers/perf/arm-cci.c
@@ -1718,5 +1718,5 @@ static struct platform_driver cci_pmu_driver = {
 };
 
 builtin_platform_driver(cci_pmu_driver);
-MODULE_LICENSE("GPL");
+MODULE_LICENSE("GPL v2");
 MODULE_DESCRIPTION("ARM CCI PMU support");
diff --git a/drivers/perf/arm-ccn.c b/drivers/perf/arm-ccn.c
index 65b7e4042ece..917b47e776df 100644
--- a/drivers/perf/arm-ccn.c
+++ b/drivers/perf/arm-ccn.c
@@ -1594,4 +1594,4 @@ module_init(arm_ccn_init);
 module_exit(arm_ccn_exit);
 
 MODULE_AUTHOR("Pawel Moll <pawel.moll@arm.com>");
-MODULE_LICENSE("GPL");
+MODULE_LICENSE("GPL v2");
-- 
2.17.0.dirty

[PATCH 2/3] perf/arm-cci: Remove pointless PMU disabling

[Robin Murphy]

The CCI PMU driver bears some legacy remnants of the arm_pmu framework
from when it was split in c6f85cb4305b ("bus: cci: move away from
arm_pmu framework"). In particular this perf_pmu_{dis,en}able() dance
around pmu->add which was fixed for arm_pmu in a9e469d1c89b
("drivers/perf: arm_pmu: remove pointless PMU disabling").

For the exact same reasons (i.e. perf core already does this around the
call anyway), give cci_pmu_add() the exact same change, which also
prevents having to export those core functions to build it as a module.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---
 drivers/perf/arm-cci.c | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)

diff --git a/drivers/perf/arm-cci.c b/drivers/perf/arm-cci.c
index 53c774b20563..7029d8fe8f44 100644
--- a/drivers/perf/arm-cci.c
+++ b/drivers/perf/arm-cci.c
@@ -1184,16 +1184,11 @@ static int cci_pmu_add(struct perf_event *event, int flags)
 	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
 	struct hw_perf_event *hwc = &event->hw;
 	int idx;
-	int err = 0;
-
-	perf_pmu_disable(event->pmu);
 
 	/* If we don't have a space for the counter then finish early. */
 	idx = pmu_get_event_idx(hw_events, event);
-	if (idx < 0) {
-		err = idx;
-		goto out;
-	}
+	if (idx < 0)
+		return idx;
 
 	event->hw.idx = idx;
 	hw_events->events[idx] = event;
@@ -1205,9 +1200,7 @@ static int cci_pmu_add(struct perf_event *event, int flags)
 	/* Propagate our changes to the userspace mapping. */
 	perf_event_update_userpage(event);
 
-out:
-	perf_pmu_enable(event->pmu);
-	return err;
+	return 0;
 }
 
 static void cci_pmu_del(struct perf_event *event, int flags)
-- 
2.17.0.dirty

[PATCH 3/3] perf/arm-cci: Allow building as a module

[Robin Murphy]

Fill in the few extra bits and annotations needed to make the driver
work properly as a module, and jiggle the Kconfig to expose the
driver-level ARM_CCI_PMU option.

Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---

Preserving the current level of configurability does end up allowing a
rather pointless module configuration which supports no hardware variants,
but I couldn't find a way to enforce "at least one sub-option selected"
logic without Kbuild claiming a recursive dependency.

 drivers/perf/Kconfig   | 31 +++++++++++++++++--------------
 drivers/perf/arm-cci.c | 17 ++++++++++++++++-
 2 files changed, 33 insertions(+), 15 deletions(-)

diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig
index 28bb5a029558..269331bfb746 100644
--- a/drivers/perf/Kconfig
+++ b/drivers/perf/Kconfig
@@ -6,30 +6,33 @@ menu "Performance monitor support"
 	depends on PERF_EVENTS
 
 config ARM_CCI_PMU
-	bool
+	tristate "ARM CCI PMU driver"
 	select ARM_CCI
+	help
+	  Support for PMU events monitoring on the ARM CCI (Cache Coherent
+	  Interconnect) family of products.
+
+	  If compiled as a module, it will be called arm-cci.
 
 config ARM_CCI400_PMU
-	bool "ARM CCI400 PMU support"
+	bool "support CCI-400"
+	default y
 	depends on (ARM && CPU_V7) || ARM64
+	depends on ARM_CCI_PMU
 	select ARM_CCI400_COMMON
-	select ARM_CCI_PMU
 	help
-	  Support for PMU events monitoring on the ARM CCI-400 (cache coherent
-	  interconnect). CCI-400 supports counting events related to the
-	  connected slave/master interfaces.
+	  CCI-400 provides 4 independent event counters counting events related
+	  to the connected slave/master interfaces, plus a cycle counter.
 
 config ARM_CCI5xx_PMU
-	bool "ARM CCI-500/CCI-550 PMU support"
+	bool "support CCI-500/CCI-550"
+	default y
 	depends on (ARM && CPU_V7) || ARM64
-	select ARM_CCI_PMU
+	depends on ARM_CCI_PMU
 	help
-	  Support for PMU events monitoring on the ARM CCI-500/CCI-550 cache
-	  coherent interconnects. Both of them provide 8 independent event counters,
-	  which can count events pertaining to the slave/master interfaces as well
-	  as the internal events to the CCI.
-
-	  If unsure, say Y
+	  CCI-500/CCI-550 both provide 8 independent event counters, which can
+	  count events pertaining to the slave/master interfaces as well as the
+	  internal events to the CCI.
 
 config ARM_CCN
 	tristate "ARM CCN driver support"
diff --git a/drivers/perf/arm-cci.c b/drivers/perf/arm-cci.c
index 7029d8fe8f44..09938dd8eb6f 100644
--- a/drivers/perf/arm-cci.c
+++ b/drivers/perf/arm-cci.c
@@ -1416,6 +1416,7 @@ static int cci_pmu_init(struct cci_pmu *cci_pmu, struct platform_device *pdev)
 	pmu_format_attr_group.attrs = model->format_attrs;
 
 	cci_pmu->pmu = (struct pmu) {
+		.module		= THIS_MODULE,
 		.name		= cci_pmu->model->name,
 		.task_ctx_nr	= perf_invalid_context,
 		.pmu_enable	= cci_pmu_enable,
@@ -1581,6 +1582,7 @@ static const struct of_device_id arm_cci_pmu_matches[] = {
 #endif
 	{},
 };
+MODULE_DEVICE_TABLE(of, arm_cci_pmu_matches);
 
 static bool is_duplicate_irq(int irq, int *irqs, int nr_irqs)
 {
@@ -1702,14 +1704,27 @@ static int cci_pmu_probe(struct platform_device *pdev)
 	return 0;
 }
 
+static int cci_pmu_remove(struct platform_device *pdev)
+{
+	if (!g_cci_pmu)
+		return 0;
+
+	cpuhp_remove_state(CPUHP_AP_PERF_ARM_CCI_ONLINE);
+	perf_pmu_unregister(&g_cci_pmu->pmu);
+	g_cci_pmu = NULL;
+
+	return 0;
+}
+
 static struct platform_driver cci_pmu_driver = {
 	.driver = {
 		   .name = DRIVER_NAME,
 		   .of_match_table = arm_cci_pmu_matches,
 		  },
 	.probe = cci_pmu_probe,
+	.remove = cci_pmu_remove,
 };
 
-builtin_platform_driver(cci_pmu_driver);
+module_platform_driver(cci_pmu_driver);
 MODULE_LICENSE("GPL v2");
 MODULE_DESCRIPTION("ARM CCI PMU support");
-- 
2.17.0.dirty

Re: [PATCH 2/3] perf/arm-cci: Remove pointless PMU disabling

[Mark Rutland]

On Fri, May 11, 2018 at 03:29:13PM +0100, Robin Murphy wrote:
> The CCI PMU driver bears some legacy remnants of the arm_pmu framework
> from when it was split in c6f85cb4305b ("bus: cci: move away from
> arm_pmu framework"). In particular this perf_pmu_{dis,en}able() dance
> around pmu->add which was fixed for arm_pmu in a9e469d1c89b
> ("drivers/perf: arm_pmu: remove pointless PMU disabling").
> 
> For the exact same reasons (i.e. perf core already does this around the
> call anyway), give cci_pmu_add() the exact same change, which also
> prevents having to export those core functions to build it as a module.
> 
> Signed-off-by: Robin Murphy <robin.murphy@arm.com>

Acked-by: Mark Rutland <mark.rutland@arm.com>

Mark.

> ---
>  drivers/perf/arm-cci.c | 13 +++----------
>  1 file changed, 3 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/perf/arm-cci.c b/drivers/perf/arm-cci.c
> index 53c774b20563..7029d8fe8f44 100644
> --- a/drivers/perf/arm-cci.c
> +++ b/drivers/perf/arm-cci.c
> @@ -1184,16 +1184,11 @@ static int cci_pmu_add(struct perf_event *event, int flags)
>  	struct cci_pmu_hw_events *hw_events = &cci_pmu->hw_events;
>  	struct hw_perf_event *hwc = &event->hw;
>  	int idx;
> -	int err = 0;
> -
> -	perf_pmu_disable(event->pmu);
>  
>  	/* If we don't have a space for the counter then finish early. */
>  	idx = pmu_get_event_idx(hw_events, event);
> -	if (idx < 0) {
> -		err = idx;
> -		goto out;
> -	}
> +	if (idx < 0)
> +		return idx;
>  
>  	event->hw.idx = idx;
>  	hw_events->events[idx] = event;
> @@ -1205,9 +1200,7 @@ static int cci_pmu_add(struct perf_event *event, int flags)
>  	/* Propagate our changes to the userspace mapping. */
>  	perf_event_update_userpage(event);
>  
> -out:
> -	perf_pmu_enable(event->pmu);
> -	return err;
> +	return 0;
>  }
>  
>  static void cci_pmu_del(struct perf_event *event, int flags)
> -- 
> 2.17.0.dirty
>