iommu: arm-smmu: handle client iommu translation fault handlers

Message ID 20181010144407.23336-1-stanimir.varbanov@linaro.org
State New
Headers show
Series
  • iommu: arm-smmu: handle client iommu translation fault handlers
Related show

Commit Message

Stanimir Varbanov Oct. 10, 2018, 2:44 p.m.
Call iommu client translation fault handler(s).

Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>

---
 drivers/iommu/arm-smmu.c | 3 +++
 1 file changed, 3 insertions(+)

-- 
2.17.1

Comments

Will Deacon Oct. 10, 2018, 5:08 p.m. | #1
On Wed, Oct 10, 2018 at 05:44:07PM +0300, Stanimir Varbanov wrote:
> Call iommu client translation fault handler(s).

> 

> Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>

> ---

>  drivers/iommu/arm-smmu.c | 3 +++

>  1 file changed, 3 insertions(+)

> 

> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c

> index 066f4c8daf4e..02a8aab0cc59 100644

> --- a/drivers/iommu/arm-smmu.c

> +++ b/drivers/iommu/arm-smmu.c

> @@ -568,6 +568,9 @@ static irqreturn_t arm_smmu_context_fault(int irq, void *dev)

>  			    fsr, iova, fsynr, cfg->cbndx);

>  

>  	writel(fsr, cb_base + ARM_SMMU_CB_FSR);

> +

> +	report_iommu_fault(domain, smmu->dev, iova, 0);


Who's using this? Generally, the callback here isn't so useful because it's
called in irq context. We also don't enable stalling since 3714ce1d6655.

We previously had fairly detailed discussions on the list with Rob Clark
about alternative ways to design this, but I don't remember where it ended
up.

Will
Stanimir Varbanov Oct. 11, 2018, 12:26 p.m. | #2
Hi Will,

On 10/10/2018 08:08 PM, Will Deacon wrote:
> On Wed, Oct 10, 2018 at 05:44:07PM +0300, Stanimir Varbanov wrote:

>> Call iommu client translation fault handler(s).

>>

>> Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>

>> ---

>>  drivers/iommu/arm-smmu.c | 3 +++

>>  1 file changed, 3 insertions(+)

>>

>> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c

>> index 066f4c8daf4e..02a8aab0cc59 100644

>> --- a/drivers/iommu/arm-smmu.c

>> +++ b/drivers/iommu/arm-smmu.c

>> @@ -568,6 +568,9 @@ static irqreturn_t arm_smmu_context_fault(int irq, void *dev)

>>  			    fsr, iova, fsynr, cfg->cbndx);

>>  

>>  	writel(fsr, cb_base + ARM_SMMU_CB_FSR);

>> +

>> +	report_iommu_fault(domain, smmu->dev, iova, 0);

> 

> Who's using this? Generally, the callback here isn't so useful because it's


The idea was to use it from video driver to dump few multimedia
interconnect registers which will give us some more info about the fault.

> called in irq context. We also don't enable stalling since 3714ce1d6655.

> 

> We previously had fairly detailed discussions on the list with Rob Clark

> about alternative ways to design this, but I don't remember where it ended

> up.


It will be useful if you point me to the discussion. Thanks!

-- 
regards,
Stan
Will Deacon Oct. 19, 2018, 5:44 p.m. | #3
On Thu, Oct 11, 2018 at 03:26:57PM +0300, Stanimir Varbanov wrote:
> On 10/10/2018 08:08 PM, Will Deacon wrote:

> > On Wed, Oct 10, 2018 at 05:44:07PM +0300, Stanimir Varbanov wrote:

> >> Call iommu client translation fault handler(s).

> >>

> >> Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>

> >> ---

> >>  drivers/iommu/arm-smmu.c | 3 +++

> >>  1 file changed, 3 insertions(+)

> >>

> >> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c

> >> index 066f4c8daf4e..02a8aab0cc59 100644

> >> --- a/drivers/iommu/arm-smmu.c

> >> +++ b/drivers/iommu/arm-smmu.c

> >> @@ -568,6 +568,9 @@ static irqreturn_t arm_smmu_context_fault(int irq, void *dev)

> >>  			    fsr, iova, fsynr, cfg->cbndx);

> >>  

> >>  	writel(fsr, cb_base + ARM_SMMU_CB_FSR);

> >> +

> >> +	report_iommu_fault(domain, smmu->dev, iova, 0);

> > 

> > Who's using this? Generally, the callback here isn't so useful because it's

> 

> The idea was to use it from video driver to dump few multimedia

> interconnect registers which will give us some more info about the fault.

> 

> > called in irq context. We also don't enable stalling since 3714ce1d6655.

> > 

> > We previously had fairly detailed discussions on the list with Rob Clark

> > about alternative ways to design this, but I don't remember where it ended

> > up.

> 

> It will be useful if you point me to the discussion. Thanks!


There's more, but some stuff I managed to dig up is here:

https://lists.linuxfoundation.org/pipermail/iommu/2017-January/019931.html
https://lists.linuxfoundation.org/pipermail/iommu/2017-September/024143.html

Will

Patch

diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
index 066f4c8daf4e..02a8aab0cc59 100644
--- a/drivers/iommu/arm-smmu.c
+++ b/drivers/iommu/arm-smmu.c
@@ -568,6 +568,9 @@  static irqreturn_t arm_smmu_context_fault(int irq, void *dev)
 			    fsr, iova, fsynr, cfg->cbndx);
 
 	writel(fsr, cb_base + ARM_SMMU_CB_FSR);
+
+	report_iommu_fault(domain, smmu->dev, iova, 0);
+
 	return IRQ_HANDLED;
 }