On Thu, 2020-07-23 at 14:47 -0600, Rob Herring wrote:
> On Sat, Jul 11, 2020 at 02:48:43PM +0800, Yong Wu wrote:
> > Some HW IP(ex: CCU) require the special iova range. That means the
> > iova got from dma_alloc_attrs for that devices must locate in his
> > special range. In this patch, we allocate a special iova_range for
> > each a special requirement and create each a iommu domain for each
> > a iova_range.
> >
> > meanwhile we still use one pagetable which support 16GB iova.
> >
> > After this patch, If the iova range of a master is over 4G, the master
> > should:
> > a) Declare its special dma_ranges in its dtsi node. For example, If we
> > preassign the iova 4G-8G for vcodec, then the vcodec dtsi node should:
> > dma-ranges = <0x1 0x0 0x1 0x0 0x1 0x0>; /* 4G ~ 8G */
>
> BTW, dma-ranges should be in the parent node of the vcodec.
But the vcodec doesn't have its special parent node. Currently the
vcodec/display dtsi like this:
soc {
ovl:{ /* display */
/*No dma-ranges property. defaultly it is 0-4G iova range. */
}
vcodec_dec: { /* decode */
dma-ranges = <0x1 0x0 0x1 0x0 0x1 0x0>; /* 4G ~ 8G*/
};
vcodec_enc: { /* encode */
dma-ranges = <0x1 0x0 0x1 0x0 0x1 0x0>; /* 4G ~ 8G*/
};
camera: {
dma-ranges = <0x2 0x0 0x2 0x0 0x1 0x0>; /* 8G ~ 12G */
};
}
If we add the parent node for vcodec, the vcodec driver flow will be
changed, and it may be incompatible with the previous dtb.
Here we don't have the actual bus concept. currently we support 16GB
dma_addr(iova) ranges. we only preassign 4-8G for vcodec, 8G-12G for
camera.
If the usage of dma-ranges here is different from the common one. then
how should I do here?
Thanks.
>
> > b) Update the dma_mask:
> > dma_set_mask_and_coherent(dev, DMA_BIT_MASK(33));
>
> This should happen for you automatically. The DMA PFN offset
> should also be 4GB here.
I may not follow here.
If the iova start at 0x1_0000_0000, phys address start at 0x4000_0000.
Do you means the dma-ranges should be <0x1 0 0x0 0x40000000 0x1 0x0>?
then dma_pfn_offset = PFN_DOWN(paddr - dma_addr) = 0xffffffff40000. this
is also ok for us. we don't call the macro regarding this
"dev->dma_pfn_offset"
The purpose that I call it here is for updating the
dev->coherent_dma_mask[1], then we could get the iova over 4GB.
[1]
https://elixir.bootlin.com/linux/v5.8-rc1/source/drivers/iommu/dma-iommu.c#L619
>
> >
> > Signed-off-by: Yong Wu <yong.wu@xxxxxxxxxxxx>
> > ---
> > drivers/iommu/mtk_iommu.c | 49 ++++++++++++++++++++++++++++++++-------
> > drivers/iommu/mtk_iommu.h | 3 ++-
> > 2 files changed, 42 insertions(+), 10 deletions(-)
