[PATCH 0/2] bus: CCN/CCI: Fix use of smp_processor_id()

[PATCH 0/2] bus: CCN/CCI: Fix use of smp_processor_id()

[Marc Zyngier]

I've just noticed that both the CCI and CCN drivers have a small
buglet in that they call smp_processor_id() from preemptible context,
which is frown upon (having just booted a 4.13 kernel with
DEBUG_PREEMPT on my Seattle, I was surprised to be greeted with a nice
backtrace...).

I've tested the CCN patch on the same Seatle box, but I've only
compile-tested the equivalent CCI patch (which is obviously correct --
famous last words...).

Thanks,

	M.

Marc Zyngier (2):
  bus: arm-ccn: Fix use of smp_processor_id() in preemptible context
  bus: arm-cci: Fix use of smp_processor_id() in preemptible context

 drivers/bus/arm-cci.c | 3 ++-
 drivers/bus/arm-ccn.c | 3 ++-
 2 files changed, 4 insertions(+), 2 deletions(-)

-- 
2.11.0

[PATCH 1/2] bus: arm-ccn: Fix use of smp_processor_id() in preemptible context

[Marc Zyngier]

Booting a DEBUG_PREEMPT enabled kernel on a CCN-based system
results in the following splat:

[...]
arm-ccn e8000000.ccn: No access to interrupts, using timer.
BUG: using smp_processor_id() in preemptible [00000000] code: swapper/0/1
caller is debug_smp_processor_id+0x1c/0x28
CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.13.0 #6111
Hardware name: AMD Seattle/Seattle, BIOS 17:08:23 Jun 26 2017
Call trace:
[<ffff000008089e78>] dump_backtrace+0x0/0x278
[<ffff00000808a22c>] show_stack+0x24/0x30
[<ffff000008bc3bc4>] dump_stack+0x8c/0xb0
[<ffff00000852b534>] check_preemption_disabled+0xfc/0x100
[<ffff00000852b554>] debug_smp_processor_id+0x1c/0x28
[<ffff000008551bd8>] arm_ccn_probe+0x358/0x4f0
[...]

as we use smp_processor_id() in the wrong context. Turning this
into a get_cpu()/put_cpu() does the trick.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
---
 drivers/bus/arm-ccn.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/bus/arm-ccn.c b/drivers/bus/arm-ccn.c
index e8c6946fed9d..8502f1767024 100644
--- a/drivers/bus/arm-ccn.c
+++ b/drivers/bus/arm-ccn.c
@@ -1297,7 +1297,8 @@ static int arm_ccn_pmu_init(struct arm_ccn *ccn)
 	}
 
 	/* Pick one CPU which we will use to collect data from CCN... */
-	cpumask_set_cpu(smp_processor_id(), &ccn->dt.cpu);
+	cpumask_set_cpu(get_cpu(), &ccn->dt.cpu);
+	put_cpu();
 
 	/* Also make sure that the overflow interrupt is handled by this CPU */
 	if (ccn->irq) {
-- 
2.11.0

[PATCH 2/2] bus: arm-cci: Fix use of smp_processor_id() in preemptible context

[Marc Zyngier]

The ARM CCI driver seem to be using smp_processor_id() in a
preemptible context, which is likely to make a DEBUG_PREMPT
kernel scream at boot time.

Turning this into a get_cpu()/put_cpu() should make it behave.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
---
 drivers/bus/arm-cci.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/bus/arm-cci.c b/drivers/bus/arm-cci.c
index c49da15d9790..e2873789972a 100644
--- a/drivers/bus/arm-cci.c
+++ b/drivers/bus/arm-cci.c
@@ -1755,7 +1755,8 @@ static int cci_pmu_probe(struct platform_device *pdev)
 	raw_spin_lock_init(&cci_pmu->hw_events.pmu_lock);
 	mutex_init(&cci_pmu->reserve_mutex);
 	atomic_set(&cci_pmu->active_events, 0);
-	cpumask_set_cpu(smp_processor_id(), &cci_pmu->cpus);
+	cpumask_set_cpu(get_cpu(), &cci_pmu->cpus);
+	put_cpu();
 
 	ret = cci_pmu_init(cci_pmu, pdev);
 	if (ret)
-- 
2.11.0

Re: [PATCH 1/2] bus: arm-ccn: Fix use of smp_processor_id() in preemptible context

[Mark Rutland]

On Mon, Sep 04, 2017 at 06:30:40PM +0100, Marc Zyngier wrote:
> Booting a DEBUG_PREEMPT enabled kernel on a CCN-based system
> results in the following splat:
> 
> [...]
> arm-ccn e8000000.ccn: No access to interrupts, using timer.
> BUG: using smp_processor_id() in preemptible [00000000] code: swapper/0/1
> caller is debug_smp_processor_id+0x1c/0x28
> CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.13.0 #6111
> Hardware name: AMD Seattle/Seattle, BIOS 17:08:23 Jun 26 2017
> Call trace:
> [<ffff000008089e78>] dump_backtrace+0x0/0x278
> [<ffff00000808a22c>] show_stack+0x24/0x30
> [<ffff000008bc3bc4>] dump_stack+0x8c/0xb0
> [<ffff00000852b534>] check_preemption_disabled+0xfc/0x100
> [<ffff00000852b554>] debug_smp_processor_id+0x1c/0x28
> [<ffff000008551bd8>] arm_ccn_probe+0x358/0x4f0
> [...]
> 
> as we use smp_processor_id() in the wrong context. Turning this
> into a get_cpu()/put_cpu() does the trick.
> 
> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
> ---
>  drivers/bus/arm-ccn.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/bus/arm-ccn.c b/drivers/bus/arm-ccn.c
> index e8c6946fed9d..8502f1767024 100644
> --- a/drivers/bus/arm-ccn.c
> +++ b/drivers/bus/arm-ccn.c
> @@ -1297,7 +1297,8 @@ static int arm_ccn_pmu_init(struct arm_ccn *ccn)
>  	}
>  
>  	/* Pick one CPU which we will use to collect data from CCN... */
> -	cpumask_set_cpu(smp_processor_id(), &ccn->dt.cpu);
> +	cpumask_set_cpu(get_cpu(), &ccn->dt.cpu);
> +	put_cpu();

I think this isn't quite right, and we need to keep a CPU online until
we've registerd the hotplug callbacks. Otherwise, we could race with a
concurrent hotplug, and not migrate events appropriately.

Thanks,
Mark.

Re: [PATCH 1/2] bus: arm-ccn: Fix use of smp_processor_id() in preemptible context

[Marc Zyngier]

On 04/09/17 18:37, Mark Rutland wrote:
> On Mon, Sep 04, 2017 at 06:30:40PM +0100, Marc Zyngier wrote:
>> Booting a DEBUG_PREEMPT enabled kernel on a CCN-based system
>> results in the following splat:
>>
>> [...]
>> arm-ccn e8000000.ccn: No access to interrupts, using timer.
>> BUG: using smp_processor_id() in preemptible [00000000] code: swapper/0/1
>> caller is debug_smp_processor_id+0x1c/0x28
>> CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.13.0 #6111
>> Hardware name: AMD Seattle/Seattle, BIOS 17:08:23 Jun 26 2017
>> Call trace:
>> [<ffff000008089e78>] dump_backtrace+0x0/0x278
>> [<ffff00000808a22c>] show_stack+0x24/0x30
>> [<ffff000008bc3bc4>] dump_stack+0x8c/0xb0
>> [<ffff00000852b534>] check_preemption_disabled+0xfc/0x100
>> [<ffff00000852b554>] debug_smp_processor_id+0x1c/0x28
>> [<ffff000008551bd8>] arm_ccn_probe+0x358/0x4f0
>> [...]
>>
>> as we use smp_processor_id() in the wrong context. Turning this
>> into a get_cpu()/put_cpu() does the trick.
>>
>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
>> ---
>>  drivers/bus/arm-ccn.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/bus/arm-ccn.c b/drivers/bus/arm-ccn.c
>> index e8c6946fed9d..8502f1767024 100644
>> --- a/drivers/bus/arm-ccn.c
>> +++ b/drivers/bus/arm-ccn.c
>> @@ -1297,7 +1297,8 @@ static int arm_ccn_pmu_init(struct arm_ccn *ccn)
>>  	}
>>  
>>  	/* Pick one CPU which we will use to collect data from CCN... */
>> -	cpumask_set_cpu(smp_processor_id(), &ccn->dt.cpu);
>> +	cpumask_set_cpu(get_cpu(), &ccn->dt.cpu);
>> +	put_cpu();
> 
> I think this isn't quite right, and we need to keep a CPU online until
> we've registerd the hotplug callbacks. Otherwise, we could race with a
> concurrent hotplug, and not migrate events appropriately.
Right, I see. I'll respin that.

Thanks,

	M.
-- 
Jazz is not dead. It just smells funny...