Re: [RFC v1 PATCH 1/2] of/pci: add of_pci_dma_configure() update dma configuration

[Murali Karicheri]

On 12/22/2014 02:43 PM, Arnd Bergmann wrote:
> On Monday 22 December 2014 12:46:12 Murali Karicheri wrote:
> > On 12/18/2014 05:29 PM, Arnd Bergmann wrote:
> > > On Thursday 18 December 2014 17:07:04 Murali Karicheri wrote:
> > > > Add of_pci_dma_configure() to allow updating the dma configuration
> > > > of the pci device using the configuration from the parent of
> > > > the root bridge device.
> > > > +	/*
> > > > +	 * Set it to coherent_dma_mask by default if the architecture
> > > > +	 * code has not set it.
> > > > +	 */
> > > > +	if (!dev->dma_mask)
> > > > +		dev->dma_mask =&dev->coherent_dma_mask;
> > > > +
> > > > +	ret = of_dma_get_range(parent_np,&dma_addr,&paddr,&size);
> > > > +	if (ret<   0) {
> > > > +		dma_addr = offset = 0;
> > > > +		size = dev->coherent_dma_mask;
> > >
> > > Can you check one thing here? I believe the size argument as returned
> > > from of_dma_get_range is inclusive (e.g. 0x100000000), while the coherent
> > > mask by definition is exlusive (e.g. 0xffffffff), so the size needs to
> > > be adapted here. I haven't checked all the code here though, so I may
> > > be wrong.
> >
> > size returned by of_dma_get_range() is inclusive as you indicated. Fromt
> > the grep of dma-ranges in arch/arm/boot/dts, I see
> >
> > arch/arm/boot/dts/keystone.dtsi:		dma-ranges =<0x80000000 0x8
> > 0x00000000 0x80000000>;
> > arch/arm/boot/dts/keystone.dtsi:			dma-ranges;
> > arch/arm/boot/dts/r8a7790.dtsi:		dma-ranges =<0x42000000 0 0x40000000 0
> > 0x40000000 0 0x80000000
> > arch/arm/boot/dts/integratorap.dts:	dma-ranges =<0x80000000 0x0
> > 0x80000000>;
> > arch/arm/boot/dts/r8a7791.dtsi:		dma-ranges =<0x42000000 0 0x40000000 0
> > 0x40000000 0 0x80000000
> >
> > So I guess I need to change the code to
> >
> >   >>  +	ret = of_dma_get_range(parent_np,&dma_addr,&paddr,&size);
> >   >>  +	if (ret<   0) {
> >   >>  +		dma_addr = offset = 0;
> >   >>  +		size = dev->coherent_dma_mask + 1;
>
> Right.
>
> > > > +	} else {
> > > > +		offset = PFN_DOWN(paddr - dma_addr);
> > > > +		dev_dbg(dev, "dma_pfn_offset(%#08lx)\n", dev->dma_pfn_offset);
> > > > +	}
> > > > +	dev->dma_pfn_offset = offset;
> > > > +
> > > > +	coherent = of_dma_is_coherent(parent_np);
> > > > +	dev_dbg(dev, "device is%sdma coherent\n",
> > > > +		coherent ? " " : " not ");
> > > > +
> > > > +	arch_setup_dma_ops(dev, dma_addr, size, NULL, coherent);
> > >
> > > Basically, I would use limit the size argument we pass into
> > > arch_setup_dma_ops to the minimum of 'size' and 'dma_mask' here,
> > > after converting into the same format. We should make sure we do the
> > > same thing for platform_device as well, so it might be better to do
> > > it inside of arch_setup_dma_ops instead.
> >
> > Do you think following changes will work?
> >
> > +++ b/arch/arm/mm/dma-mapping.c
> > @@ -2052,9 +2052,10 @@ void arch_setup_dma_ops(struct device *dev, u64
> > dma_base, u64 size,
> >                           struct iommu_ops *iommu, bool coherent)
> >    {
> >           struct dma_map_ops *dma_ops;
> > +       u64 temp_size = min((*(dev->dma_mask) + 1), size);
> >
> >           dev->archdata.dma_coherent = coherent;
> > -       if (arm_setup_iommu_dma_ops(dev, dma_base, size, iommu))
> > +       if (arm_setup_iommu_dma_ops(dev, dma_base, temp_size, iommu))
> >
> > If you agree, I will post v1 of the patch with these updates. Let me
> > know. I did some basic tests on Keystone with these changes and it works
> > fine.
>
> It's not exactly what I meant. My main point was that we need to limit
> dev->dma_mask to (size-1) here, but you are not touching that.

if you mean overriding the dev->dma_mask to min((*dev->dma_mask), 
size-1), then I am getting the error "Coherent DMA mask 0x7fffffff (pfn 
0x780000-0x800000) covers a smaller range of system memory than the DMA 
zone pfn 0x0-0x880000) when the devices on Keystone tries to set the dma 
mask. Something wrong and I need to look into the code.

> either change the two functions in which we first assign the dma_mask
> (platform and pci bus), or set it again in arch_setup_dma_ops (on each
> architecture implementing it), either way would work. Someone else
> might have a stronger opinion on that matter.
>
> For arm_setup_iommu_dma_ops, passing the original size is probably
> best.
>
> 	Arnd


--
Murali Karicheri
Linux Kernel, Texas Instruments
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/