[PATCH 1/2] ARM: perf: fix prototype of release_pmu

[PATCH 1/2] ARM: perf: fix prototype of release_pmu

[Mark Rutland]

Commit f12482c9 ("ARM: 6974/1: pmu: refactor reservation") changed the
prototype of release_pmu, but missed the stub for when
CONFIG_CPU_HAS_PMU is not selected by the platform.

This patch changes the prototype of the stub, preventing possible build
failures when CONFIG_CPU_HAS_PMU is not selected.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 arch/arm/include/asm/pmu.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/arm/include/asm/pmu.h b/arch/arm/include/asm/pmu.h
index 67c70a3..8ae32ba 100644
--- a/arch/arm/include/asm/pmu.h
+++ b/arch/arm/include/asm/pmu.h
@@ -75,7 +75,7 @@ reserve_pmu(enum arm_pmu_type device)
 }
 
 static inline int
-release_pmu(struct platform_device *pdev)
+release_pmu(enum arm_pmu_type device)
 {
 	return -ENODEV;
 }
-- 
1.7.0.4

[PATCH 2/2] ARM: perf: make name of arm_pmu_type consistent

[Mark Rutland]

Commit f12482c9 ("ARM: 6974/1: pmu: refactor reservation") changed
{release,reserve}_pmu to take an enum arm_pmu_type as a parameter, but
inconsistently named the parameter `type' or `device'. It would be nice
if these were consistent.

This patch makes use of enum arm_pmu_type consistent, always using
'device' as this is more common in usage.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
---
 arch/arm/include/asm/pmu.h |    2 +-
 arch/arm/kernel/pmu.c      |   14 +++++++-------
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/arch/arm/include/asm/pmu.h b/arch/arm/include/asm/pmu.h
index 8ae32ba..456117c 100644
--- a/arch/arm/include/asm/pmu.h
+++ b/arch/arm/include/asm/pmu.h
@@ -52,7 +52,7 @@ reserve_pmu(enum arm_pmu_type device);
  * a cookie.
  */
 extern int
-release_pmu(enum arm_pmu_type type);
+release_pmu(enum arm_pmu_type device);
 
 /**
  * init_pmu() - Initialise the PMU.
diff --git a/arch/arm/kernel/pmu.c b/arch/arm/kernel/pmu.c
index 2b70709..89cbdc5 100644
--- a/arch/arm/kernel/pmu.c
+++ b/arch/arm/kernel/pmu.c
@@ -27,22 +27,22 @@ static volatile long pmu_lock;
 static struct platform_device *pmu_devices[ARM_NUM_PMU_DEVICES];
 
 static int __devinit pmu_register(struct platform_device *pdev,
-					enum arm_pmu_type type)
+					enum arm_pmu_type device)
 {
-	if (type < 0 || type >= ARM_NUM_PMU_DEVICES) {
+	if (device < 0 || device >= ARM_NUM_PMU_DEVICES) {
 		pr_warning("received registration request for unknown "
-				"device %d\n", type);
+				"device %d\n", device);
 		return -EINVAL;
 	}
 
-	if (pmu_devices[type]) {
+	if (pmu_devices[device]) {
 		pr_warning("rejecting duplicate registration of PMU device "
-			"type %d.", type);
+			"type %d.", device);
 		return -ENOSPC;
 	}
 
-	pr_info("registered new PMU device of type %d\n", type);
-	pmu_devices[type] = pdev;
+	pr_info("registered new PMU device of type %d\n", device);
+	pmu_devices[device] = pdev;
 	return 0;
 }
 
-- 
1.7.0.4

[PATCH 2/2] ARM: perf: make name of arm_pmu_type consistent

[Will Deacon]

Mark,

On Wed, Aug 10, 2011 at 01:58:58PM +0100, Mark Rutland wrote:
> Commit f12482c9 ("ARM: 6974/1: pmu: refactor reservation") changed
> {release,reserve}_pmu to take an enum arm_pmu_type as a parameter, but
> inconsistently named the parameter `type' or `device'. It would be nice
> if these were consistent.
> 
> This patch makes use of enum arm_pmu_type consistent, always using
> 'device' as this is more common in usage.

Thanks for fixing this. I'll take care of these two patches as fixes for
3.1.

Will

[PATCH 2/2] ARM: perf: make name of arm_pmu_type consistent

[Russell King - ARM Linux]

On Wed, Aug 10, 2011 at 01:58:58PM +0100, Mark Rutland wrote:
> Commit f12482c9 ("ARM: 6974/1: pmu: refactor reservation") changed
> {release,reserve}_pmu to take an enum arm_pmu_type as a parameter, but
> inconsistently named the parameter `type' or `device'. It would be nice
> if these were consistent.
> 
> This patch makes use of enum arm_pmu_type consistent, always using
> 'device' as this is more common in usage.

This is confusing.  It's called arm_pmu_TYPE, and it's referred to as a
type in printks.  Yet you're renaming the variable to device.  That just
adds to the problem.

What is it, a type number or a device number?

If it's a device number, then the printks also need fixing, and maybe the
enumerated type name too.

[PATCH 2/2] ARM: perf: make name of arm_pmu_type consistent

[Will Deacon]

On Fri, Aug 12, 2011 at 10:06:09AM +0100, Russell King - ARM Linux wrote:
> On Wed, Aug 10, 2011 at 01:58:58PM +0100, Mark Rutland wrote:
> > Commit f12482c9 ("ARM: 6974/1: pmu: refactor reservation") changed
> > {release,reserve}_pmu to take an enum arm_pmu_type as a parameter, but
> > inconsistently named the parameter `type' or `device'. It would be nice
> > if these were consistent.
> > 
> > This patch makes use of enum arm_pmu_type consistent, always using
> > 'device' as this is more common in usage.
> 
> This is confusing.  It's called arm_pmu_TYPE, and it's referred to as a
> type in printks.  Yet you're renaming the variable to device.  That just
> adds to the problem.
> 
> What is it, a type number or a device number?
> 
> If it's a device number, then the printks also need fixing, and maybe the
> enumerated type name too.

Yes, we should uniformly use type instead of device. The only reason device
is kicking around is because the functions used to take a platform_device
rather than an arm_pmu_type.

Once I've got an updated patch from Mark, I'll update my fixes branch and
send a pull request to you Russell (I have a handful of other fixes that
have been on the list).

Cheers,

Will