Re: [PATCH v4 1/6] of: iommu: add ptr to OF node arg to of_iommu_configure()

[Will Deacon <will.deacon-5wv7dgnIgG8@public.gmane.org>]

On Wed, Jan 28, 2015 at 12:23:03PM +0000, Laurent Pinchart wrote:
> On Wednesday 28 January 2015 11:33:00 Will Deacon wrote:
> > On Mon, Jan 26, 2015 at 06:49:01PM +0000, Murali Karicheri wrote:
> > > On 01/25/2015 08:32 AM, Laurent Pinchart wrote:
> > >> On Friday 23 January 2015 17:32:34 Murali Karicheri wrote:
> > >>> Function of_iommu_configure() is called from of_dma_configure() to
> > >>> setup iommu ops using DT property. This API is currently used for
> > >>> platform devices for which DMA configuration (including iommu ops)
> > >>> may come from device's parent. To extend this functionality for PCI
> > >>> devices, this API need to take a parent node ptr as an argument
> > >>> instead of assuming device's parent. This is needed since for PCI, the
> > >>> dma configuration may be defined in the DT node of the root bus
> > >>> bridge's parent device. Currently only dma-range is used for PCI and
> > >>> iommu is not supported. So return error if the device is PCI.
> > >>> 
> > >>> Cc: Joerg Roedel<joro-zLv9SwRftAIdnm+yROfE0A@public.gmane.org>
> > >>> Cc: Grant Likely<grant.likely-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
> > >>> Cc: Rob Herring<robh+dt-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>
> > >>> Cc: Bjorn Helgaas<bhelgaas-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
> > >>> Cc: Will Deacon<will.deacon-5wv7dgnIgG8@public.gmane.org>
> > >>> Cc: Russell King<linux-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org>
> > >>> Cc: Arnd Bergmann<arnd-r2nGTMty4D4@public.gmane.org>
> > >>> Cc: Suravee Suthikulpanit<Suravee.Suthikulpanit-5C7GfCeVMHo@public.gmane.org>
> > >>> 
> > >>> Signed-off-by: Murali Karicheri<m-karicheri2-l0cyMroinI0@public.gmane.org>
> > >>> ---
> > >>> 
> > >>>   drivers/iommu/of_iommu.c |   10 ++++++++--
> > >>>   drivers/of/platform.c    |    2 +-
> > >>>   include/linux/of_iommu.h |    6 ++++--
> > >>>   3 files changed, 13 insertions(+), 5 deletions(-)
> > >>> 
> > >>> diff --git a/drivers/iommu/of_iommu.c b/drivers/iommu/of_iommu.c
> > >>> index af1dc6a..439235b 100644
> > >>> --- a/drivers/iommu/of_iommu.c
> > >>> +++ b/drivers/iommu/of_iommu.c
> > >>> @@ -133,19 +133,25 @@ struct iommu_ops *of_iommu_get_ops(struct
> > >>> device_node *np)
> > >>>   	return ops;
> > >>>  }
> > >>> 
> > >>> -struct iommu_ops *of_iommu_configure(struct device *dev)
> > >>> +struct iommu_ops *of_iommu_configure(struct device *dev,
> > >>> +				     struct device_node *iommu_np)
> > >>>  {
> > >>>   	struct of_phandle_args iommu_spec;
> > >>>   	struct device_node *np;
> > >>>   	struct iommu_ops *ops = NULL;
> > >>>   	int idx = 0;
> > >>> 
> > >>> +	if (dev_is_pci(dev)) {
> > >>> +		dev_err(dev, "iommu is currently not supported for PCI\n");
> > >>> +		return NULL;
> > >>> +	}
> > >>> +
> > >>>   	/*
> > >>>   	 * We don't currently walk up the tree looking for a parent IOMMU.
> > >>>   	 * See the `Notes:' section of
> > >>>   	 * Documentation/devicetree/bindings/iommu/iommu.txt
> > >>>   	 */
> > >> 
> > >> Wouldn't it be better to fix this rather than passing the device node
> > >> pointer to the function ? The solution would be more generic.
> > > 
> > > Will Deacon (Copied here) is working on this as we alteady discussed in
> > > this thread. So it will be addressed by him in another series.
> > 
> > Well, "working on this" equates to "has it somewhere near the bottom of
> > a very long list" :)
> 
> What's your opinion on this patch then, do you think adding the iommu_np 
> argument makes sense as a temporary hack, or should we instead walk up the 
> tree looking for an iommus attribute in parent nodes ? I don't expect that to 
> be insanely difficult to code.

If walking up the tree is useful, then I think we should do that and update
the Documentation to reflect the new behaviour. The only reason that I
didn't code that in of_iommu_configure originally is because I didn't want
to go against the binding spec for the initial version, especially as we
didn't have a reason to look at parent nodes.

The only thing to double-check is that we don't break any existing users
of the binding with this change, but I don't think that we do.

Will