Re: [PATCH v8 06/12] ARM: dts: Add description of System MMU of Exynos SoCs

[Marek Szyprowski <m.szyprowski@samsung.com>]

Hello,

On 8/6/2013 11:54 AM, Cho KyongHo wrote:
> > -----Original Message-----
> > From: Bartlomiej Zolnierkiewicz [mailto:b.zolnierkie@samsung.com]
> > Sent: Monday, August 05, 2013 10:10 PM
> >
> > On Monday, August 05, 2013 08:16:40 PM Cho KyongHo wrote:
> > > > -----Original Message-----
> > > > From: Bartlomiej Zolnierkiewicz [mailto:b.zolnierkie@samsung.com]
> > > > Sent: Saturday, August 03, 2013 2:14 AM
> > > >
> > > > Hi,
> > > >
> > > > On Friday, July 26, 2013 08:28:19 PM Cho KyongHo wrote:
> > > > > Signed-off-by: Cho KyongHo <pullip.cho@samsung.com>
> > > > > ---
> > > > >  .../bindings/iommu/samsung,exynos4210-sysmmu.txt   |  103 +++++++
> > > > >  arch/arm/boot/dts/exynos4.dtsi                     |  122 ++++++++
> > > > >  arch/arm/boot/dts/exynos4210.dtsi                  |   25 ++
> > > > >  arch/arm/boot/dts/exynos4x12.dtsi                  |   76 +++++
> > > > >  arch/arm/boot/dts/exynos5250.dtsi                  |  291 ++++++++++++++++++++
> > > > >  5 files changed, 617 insertions(+), 0 deletions(-)
> > > > >  create mode 100644 Documentation/devicetree/bindings/iommu/samsung,exynos4210-sysmmu.txt
> > > > >
> > > > > diff --git a/Documentation/devicetree/bindings/iommu/samsung,exynos4210-sysmmu.txt
> > > > > b/Documentation/devicetree/bindings/iommu/samsung,exynos4210-sysmmu.txt
> > > > > new file mode 100644
> > > > > index 0000000..92f0a33
> > > > > --- /dev/null
> > > > > +++ b/Documentation/devicetree/bindings/iommu/samsung,exynos4210-sysmmu.txt
> > > > > @@ -0,0 +1,103 @@
> > > > > +Samsung Exynos4210 IOMMU H/W, System MMU (System Memory Management Unit)
> > > > > +
> > > > > +Samsung's Exynos architecture contains System MMU that enables scattered
> > > > > +physical memory chunks visible as a contiguous region to DMA-capable peripheral
> > > > > +devices like MFC, FIMC, FIMD, GScaler, FIMC-IS and so forth.
> > > > > +
> > > > > +System MMU is a sort of IOMMU and support identical translation table format to
> > > > > +ARMv7 translation tables with minimum set of page properties including access
> > > > > +permissions, shareability and security protection. In addition, System MMU has
> > > > > +another capabilities like L2 TLB or block-fetch buffers to minimize translation
> > > > > +latency.
> > > > > +
> > > > > +A System MMU is dedicated to a single master peripheral device.  Thus, it is
> > > > > +important to specify the correct System MMU in the device node of its master
> > > > > +device. Whereas a System MMU is dedicated to a master device, the master device
> > > > > +may have more than one System MMU.
> > > > > +
> > > > > +Required properties:
> > > > > +- compatible: Should be "samsung,exynos4210-sysmmu"
> > > > > +- reg: A tuple of base address and size of System MMU registers.
> > > > > +- interrupt-parent: The phandle of the interrupt controller of System MMU
> > > > > +- interrupts: A tuple of numbers that indicates the interrupt source.
> > > > > +- clock-names: Should be "sysmmu" if the System MMU is needed to gate its clock.
> > > > > +               Please refer to the following documents:
> > > > > +	       Documentation/devicetree/bindings/clock/clock-bindings.txt
> > > > > +	       Documentation/devicetree/bindings/clock/exynos4-clock.txt
> > > > > +	       Documentation/devicetree/bindings/clock/exynos5250-clock.txt
> > > > > +	       Optional "master" if the clock to the System MMU is gated by
> > > > > +	       another gate clock other than "sysmmu". The System MMU driver
> > > > > +	       sets "master" the parent of "sysmmu".
> > > > > +	       Exynos4 SoCs, there needs no "master" clocks.
> > > > > +	       Exynos5 SoCs, some System MMUs must have "master" clocks.
> > > > > +- clocks: Required if the System MMU is needed to gate its clock.
> > > > > +	  Please refer to the documents listed above.
> > > > > +- samsung,power-domain: Required if the System MMU is needed to gate its power.
> > > > > +	  Please refer to the following document:
> > > > > +	  Documentation/devicetree/bindings/arm/exynos/power_domain.txt
> > > > > +
> > > > > +Required properties for the master peripheral devices:
> > > > > +- iommu: phandles to the System MMUs of the device
> > > > > +
> > > > > +Examples:
> > > > > +A System MMU is dedicated to a single master device.
> > > > > +	gsc_0:  gsc@0x13e00000 {
> > > > > +		compatible = "samsung,exynos5-gsc";
> > > > > +		reg = <0x13e00000 0x1000>;
> > > > > +		interrupts = <0 85 0>;
> > > > > +		samsung,power-domain = <&pd_gsc>;
> > > > > +		clocks = <&clock 256>;
> > > > > +		clock-names = "gscl";
> > > > > +		iommu = <&sysmmu_gsc1>;
> > > > > +	};
> > > > > +
> > > > > +	sysmmu_gsc0: sysmmu@13E80000 {
> > > > > +		compatible = "samsung,exynos4210-sysmmu";
> > > > > +		reg = <0x13E80000 0x1000>;
> > > > > +		interrupt-parent = <&combiner>;
> > > > > +		interrupt-names = "sysmmu-gsc0";
> > > > > +		interrupts = <2 0>;
> > > > > +		clock-names = "sysmmu", "master";
> > > > > +		clocks = <&clock 262>, <&clock 256>;
> > > > > +		samsung,power-domain = <&pd_gsc>;
> > > > > +		status = "ok";
> > > > > +	};
> > > > > +
> > > > > +MFC has 2 System MMUs for each port that MFC is attached. Thus it seems natural
> > > > > +to define 2 System MMUs for each port of the MFC:
> > > >
> > > > Marek Szyprowski (added to cc:) has a patch fixing MFC to create separate
> > > > mfc_l and mfc_r devices (like it was in the past). Using this patch it
> > > > would be possible to bind sysmmu_mfc_l to mfc_l device and sysmmu_mfc_r to
> > > > mfc_r device. This probably also requires adding some MFC specific handling
> > > > in a device tree node and to the new master's device PM ops (in patch #10)
> > > > as previously (in our trees) sysmmu_mfc r device was set as parent of
> > > > sysmmu_mfc_l device which in turn was a parent for main MFC device (to make
> > > > runtime Power Management work). However because MFC is the only device
> > > > requiring use of multiple System MMUs above changes would allow us (unless
> > > > I'm missing something?) to use just one System MMU device per struct
> > > > exynos_iommu_client instance (making driver a lot simpler).
> > > >
> > >
> > > Does it mean that we can make the exynos-iommu driver simpler
> > > with Marek Szyprowski's patch?
> >
> > I think so and you probably need to change MFC handling anyway because
> > MFC driver does DMA allocations per mfc_l/mfc_r devices and not per main
> > device (at least in the upstream kernels).
> >
> > [ Marek, could you please resfresh and post your patch on the list? ]
> >
> > BTW There is an additional problem with combining System MMU devices per
> > one main device - it limits available address space (which in case of MFC
> > is very limited by hardware design).
> >
> > > It is welcome but I don't think it covers all topologies of System MMU and
> > > master H/W. Those are getting more complex.
> >
> > Could you please be more specific? I know about FIMC ISP subsystem but
> > it doesn't require combining System MMUs. Are there any other examples of
> > complex System MMU + master H/W topologies?
>
> I meant a lot of System MMUs in FIMC-IS subsystem.
> A big challenge to System MMU configuration is FIMC-IS as you mentioned.
> Since I don't know how the driver of FIMC-IS controls its H/W,
> I made the System MMU driver to configure System MMUs flexibly.

Don't think that combining the IOMMU controllers together made them 
flexible.
I only see it as overengineering of the complex FIMC-IS and MFC devices 
topology.

IMHO it is much better to have a simple driver, which binds to a single 
IOMMU
controller and leave it to the driver whether to have a same virtual address
space for all parts of FIMC-IS or MFC submodules/memory ports or not. 
Just make
sure that it will be possible to attach more than one sysmmu controller 
to one
iommu domain.

With some fixes and additional initialization code I have finally 
managed to
get Your driver working with FIMC & memport-based MFC together with 
iommu based
dma-mapping on Exynos4210. I will post my patches after you post Your 
updated
version.

> > Anyway I think that System MMUs should not be combined (as it is done in
> > patch #10) and should be binded per "memport" devices and not per main
> > device (as done in this patch).
> >
>
> Ok.
> I will make it them as you advised.
> And we can make the driver better gradually.

Best regards
-- 
Marek Szyprowski
Samsung R&D Institute Poland