2019-12-10 10:52:42

by Hanjun Guo

Subject: [PATCH] perf/smmuv3: Remove the leftover put_cpu() in error path

In smmu_pmu_probe(), there is put_cpu() in the error path,
which is wrong because we use raw_smp_processor_id() to
get the cpu ID, not get_cpu(), remove it.

Signed-off-by: Hanjun Guo <[email protected]>
---
drivers/perf/arm_smmuv3_pmu.c | 1 -
1 file changed, 1 deletion(-)

diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
index 773128f..fd1d46a 100644
--- a/drivers/perf/arm_smmuv3_pmu.c
+++ b/drivers/perf/arm_smmuv3_pmu.c
@@ -834,7 +834,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
out_unregister:
cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
out_cpuhp_err:
- put_cpu();
return err;
}

--
1.7.12.4

2019-12-10 11:06:41

by Robin Murphy

Subject: Re: [PATCH] perf/smmuv3: Remove the leftover put_cpu() in error path

On 10/12/2019 10:46 am, Hanjun Guo wrote:
> In smmu_pmu_probe(), there is put_cpu() in the error path,
> which is wrong because we use raw_smp_processor_id() to
> get the cpu ID, not get_cpu(), remove it.

Bah, somehow that slipped through the last round of review :)

Acked-by: Robin Murphy <[email protected]>

> Signed-off-by: Hanjun Guo <[email protected]>
> ---
> drivers/perf/arm_smmuv3_pmu.c | 1 -
> 1 file changed, 1 deletion(-)
>
> diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
> index 773128f..fd1d46a 100644
> --- a/drivers/perf/arm_smmuv3_pmu.c
> +++ b/drivers/perf/arm_smmuv3_pmu.c
> @@ -834,7 +834,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
> out_unregister:
> cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
> out_cpuhp_err:
> - put_cpu();
> return err;
> }
>
>

2019-12-10 13:27:02

by Will Deacon

Subject: Re: [PATCH] perf/smmuv3: Remove the leftover put_cpu() in error path

On Tue, Dec 10, 2019 at 06:46:24PM +0800, Hanjun Guo wrote:
> In smmu_pmu_probe(), there is put_cpu() in the error path,
> which is wrong because we use raw_smp_processor_id() to
> get the cpu ID, not get_cpu(), remove it.
>
> Signed-off-by: Hanjun Guo <[email protected]>
> ---
> drivers/perf/arm_smmuv3_pmu.c | 1 -
> 1 file changed, 1 deletion(-)
>
> diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
> index 773128f..fd1d46a 100644
> --- a/drivers/perf/arm_smmuv3_pmu.c
> +++ b/drivers/perf/arm_smmuv3_pmu.c
> @@ -834,7 +834,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
> out_unregister:
> cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
> out_cpuhp_err:
> - put_cpu();
> return err;

Can we kill 'out_cpuhp_err' altogether then and just return err if we fail
to add the hotplug instance?

Will

2019-12-10 13:56:36

by Hanjun Guo

Subject: Re: [PATCH] perf/smmuv3: Remove the leftover put_cpu() in error path

On 2019/12/10 21:24, Will Deacon wrote:
> On Tue, Dec 10, 2019 at 06:46:24PM +0800, Hanjun Guo wrote:
>> In smmu_pmu_probe(), there is put_cpu() in the error path,
>> which is wrong because we use raw_smp_processor_id() to
>> get the cpu ID, not get_cpu(), remove it.
>>
>> Signed-off-by: Hanjun Guo <[email protected]>
>> ---
>> drivers/perf/arm_smmuv3_pmu.c | 1 -
>> 1 file changed, 1 deletion(-)
>>
>> diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
>> index 773128f..fd1d46a 100644
>> --- a/drivers/perf/arm_smmuv3_pmu.c
>> +++ b/drivers/perf/arm_smmuv3_pmu.c
>> @@ -834,7 +834,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
>> out_unregister:
>> cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
>> out_cpuhp_err:
>> - put_cpu();
>> return err;
>
> Can we kill 'out_cpuhp_err' altogether then and just return err if we fail
> to add the hotplug instance?

Makes sense, but I think we can go further to kill both 'out_cpuhp_err' and
'out_register' as below [1], what do you think?

Thanks
Hanjun

[1]:
diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
index fd1d46a..a5adaba 100644
--- a/drivers/perf/arm_smmuv3_pmu.c
+++ b/drivers/perf/arm_smmuv3_pmu.c
@@ -814,14 +814,15 @@ static int smmu_pmu_probe(struct platform_device *pdev)
if (err) {
dev_err(dev, "Error %d registering hotplug, PMU @%pa\n",
err, &res_0->start);
- goto out_cpuhp_err;
+ return err;
}

err = perf_pmu_register(&smmu_pmu->pmu, name, -1);
if (err) {
+ cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
dev_err(dev, "Error %d registering PMU @%pa\n",
err, &res_0->start);
- goto out_unregister;
+ return err;
}

dev_info(dev, "Registered PMU @ %pa using %d counters with %s filter settings\n",
@@ -830,11 +831,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
"Individual");

return 0;
-
-out_unregister:
- cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
-out_cpuhp_err:
- return err;
}

static int smmu_pmu_remove(struct platform_device *pdev)

2019-12-10 14:11:45

by Will Deacon

Subject: Re: [PATCH] perf/smmuv3: Remove the leftover put_cpu() in error path

On Tue, Dec 10, 2019 at 09:55:28PM +0800, Hanjun Guo wrote:
> On 2019/12/10 21:24, Will Deacon wrote:
> > On Tue, Dec 10, 2019 at 06:46:24PM +0800, Hanjun Guo wrote:
> >> In smmu_pmu_probe(), there is put_cpu() in the error path,
> >> which is wrong because we use raw_smp_processor_id() to
> >> get the cpu ID, not get_cpu(), remove it.
> >>
> >> Signed-off-by: Hanjun Guo <[email protected]>
> >> ---
> >> drivers/perf/arm_smmuv3_pmu.c | 1 -
> >> 1 file changed, 1 deletion(-)
> >>
> >> diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
> >> index 773128f..fd1d46a 100644
> >> --- a/drivers/perf/arm_smmuv3_pmu.c
> >> +++ b/drivers/perf/arm_smmuv3_pmu.c
> >> @@ -834,7 +834,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
> >> out_unregister:
> >> cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
> >> out_cpuhp_err:
> >> - put_cpu();
> >> return err;
> >
> > Can we kill 'out_cpuhp_err' altogether then and just return err if we fail
> > to add the hotplug instance?
>
> Makes sense, but I think we can go further to kill both 'out_cpuhp_err' and
> 'out_register' as below [1], what do you think?

Although that's functionally correct, I'd prefer to keep out_unregister(),
since it acts as good reminder to anybody extending this function in future
that they need to unregister the hotplug instance on failure.

Will

2019-12-11 06:23:29

by Hanjun Guo

Subject: Re: [PATCH] perf/smmuv3: Remove the leftover put_cpu() in error path

On 2019/12/10 22:10, Will Deacon wrote:
> On Tue, Dec 10, 2019 at 09:55:28PM +0800, Hanjun Guo wrote:
>> On 2019/12/10 21:24, Will Deacon wrote:
>>> On Tue, Dec 10, 2019 at 06:46:24PM +0800, Hanjun Guo wrote:
>>>> In smmu_pmu_probe(), there is put_cpu() in the error path,
>>>> which is wrong because we use raw_smp_processor_id() to
>>>> get the cpu ID, not get_cpu(), remove it.
>>>>
>>>> Signed-off-by: Hanjun Guo <[email protected]>
>>>> ---
>>>> drivers/perf/arm_smmuv3_pmu.c | 1 -
>>>> 1 file changed, 1 deletion(-)
>>>>
>>>> diff --git a/drivers/perf/arm_smmuv3_pmu.c b/drivers/perf/arm_smmuv3_pmu.c
>>>> index 773128f..fd1d46a 100644
>>>> --- a/drivers/perf/arm_smmuv3_pmu.c
>>>> +++ b/drivers/perf/arm_smmuv3_pmu.c
>>>> @@ -834,7 +834,6 @@ static int smmu_pmu_probe(struct platform_device *pdev)
>>>> out_unregister:
>>>> cpuhp_state_remove_instance_nocalls(cpuhp_state_num, &smmu_pmu->node);
>>>> out_cpuhp_err:
>>>> - put_cpu();
>>>> return err;
>>>
>>> Can we kill 'out_cpuhp_err' altogether then and just return err if we fail
>>> to add the hotplug instance?
>>
>> Makes sense, but I think we can go further to kill both 'out_cpuhp_err' and
>> 'out_register' as below [1], what do you think?
>
> Although that's functionally correct, I'd prefer to keep out_unregister(),
> since it acts as good reminder to anybody extending this function in future
> that they need to unregister the hotplug instance on failure.

OK, I will add Robin's ACK and resend.

Thanks
Hanjun