Re: [PATCH 2/7] iommu/arm-smmu: add support for multi-master iommu groups

Re: [PATCH 2/7] iommu/arm-smmu: add support for multi-master iommu groups

[Joerg Roedel]

On Thu, Sep 04, 2014 at 05:50:00PM +0100, Will Deacon wrote:
>  static struct arm_smmu_master_cfg *
> -find_smmu_master_cfg(struct arm_smmu_device *smmu, struct device *dev)
> +find_smmu_master_cfg(struct device *dev)
>  {
> -	struct arm_smmu_master *master;
> -
> -	if (dev_is_pci(dev))
> -		return dev->archdata.iommu;
> -
> -	master = find_smmu_master(smmu, dev->of_node);
> -	return master ? &master->cfg : NULL;
> +	struct iommu_group *group = iommu_group_get(dev);
> +	return group ? iommu_group_get_iommudata(group) : NULL;

Using iommu_group_get requires the call-sites to do an iommu_group_put
in the end, but I don't see the call-sites updated.


	Joerg

Re: [PATCH 2/7] iommu/arm-smmu: add support for multi-master iommu groups

[Will Deacon]

Hi Joerg,

On Mon, Sep 15, 2014 at 04:22:57PM +0100, Joerg Roedel wrote:
> On Thu, Sep 04, 2014 at 05:50:00PM +0100, Will Deacon wrote:
> >  static struct arm_smmu_master_cfg *
> > -find_smmu_master_cfg(struct arm_smmu_device *smmu, struct device *dev)
> > +find_smmu_master_cfg(struct device *dev)
> >  {
> > -	struct arm_smmu_master *master;
> > -
> > -	if (dev_is_pci(dev))
> > -		return dev->archdata.iommu;
> > -
> > -	master = find_smmu_master(smmu, dev->of_node);
> > -	return master ? &master->cfg : NULL;
> > +	struct iommu_group *group = iommu_group_get(dev);
> > +	return group ? iommu_group_get_iommudata(group) : NULL;
> 
> Using iommu_group_get requires the call-sites to do an iommu_group_put
> in the end, but I don't see the call-sites updated.

Well spotted, thanks. Given that we have a reference on the group thanks to
iommu_group_alloc, I think it's better just having the put in this helper.

diff below. Unfortunately, I'm at a conference this week and so testing is
going to be difficult. I'll test early next week and try to get you a pull
request later that week for 3.18.

Will

--->8

diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
index d2d8cdaf..939242c 100644
--- a/drivers/iommu/arm-smmu.c
+++ b/drivers/iommu/arm-smmu.c
@@ -474,8 +474,15 @@ static struct arm_smmu_master *find_smmu_master(struct arm_smmu_device *smmu,
 static struct arm_smmu_master_cfg *
 find_smmu_master_cfg(struct device *dev)
 {
+       struct arm_smmu_master_cfg *cfg = NULL;
        struct iommu_group *group = iommu_group_get(dev);
-       return group ? iommu_group_get_iommudata(group) : NULL;
+
+       if (group) {
+               cfg = iommu_group_get_iommudata(group);
+               iommu_group_put(group);
+       }
+
+       return cfg;
 }
 

Re: [PATCH 5/7] iommu/arm-smmu: don't bother truncating the s1 output size to VA_BITS

[Andreas Schwab]

Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org> writes:

> diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> index ecad700cd4f4..2a7e3331b93a 100644
> --- a/drivers/iommu/arm-smmu.c
> +++ b/drivers/iommu/arm-smmu.c
> @@ -24,7 +24,7 @@
>   *	- v7/v8 long-descriptor format
>   *	- Non-secure access to the SMMU
>   *	- 4k and 64k pages, with contiguous pte hints.
> - *	- Up to 42-bit addressing (dependent on VA_BITS)
> + *	- Up to 48-bit addressing (dependent on VA_BITS)

Does that mean that the dependency on !ARM_SMMU for ARM64_VA_BITS_48 is
obsolete?

Andreas.

-- 
Andreas Schwab, schwab-Td1EMuHUCqxL1ZNQvxDV9g@public.gmane.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [PATCH 5/7] iommu/arm-smmu: don't bother truncating the s1 output size to VA_BITS

[Will Deacon]

On Mon, Dec 01, 2014 at 08:39:18PM +0000, Andreas Schwab wrote:
> Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org> writes:
> 
> > diff --git a/drivers/iommu/arm-smmu.c b/drivers/iommu/arm-smmu.c
> > index ecad700cd4f4..2a7e3331b93a 100644
> > --- a/drivers/iommu/arm-smmu.c
> > +++ b/drivers/iommu/arm-smmu.c
> > @@ -24,7 +24,7 @@
> >   *	- v7/v8 long-descriptor format
> >   *	- Non-secure access to the SMMU
> >   *	- 4k and 64k pages, with contiguous pte hints.
> > - *	- Up to 42-bit addressing (dependent on VA_BITS)
> > + *	- Up to 48-bit addressing (dependent on VA_BITS)
> 
> Does that mean that the dependency on !ARM_SMMU for ARM64_VA_BITS_48 is
> obsolete?

No, you need my other patch series for that (we don't handle stage-2
properly without it):

  http://lists.infradead.org/pipermail/linux-arm-kernel/2014-November/306786.html

Will