Am 18.07.2017 um 18:38 schrieb Bjorn Andersson: > On Tue 18 Jul 01:45 PDT 2017, Oleksij Rempel wrote: > >> Hallo Bjorn, >> >> On 11.07.2017 00:14, Bjorn Andersson wrote: >>> On Mon 10 Jul 07:42 PDT 2017, Oleksij Rempel wrote: >>> >>>> Signed-off-by: Oleksij Rempel <o.rem...@pengutronix.de> >>>> --- >>>> .../devicetree/bindings/remoteproc/imx-rproc.txt | 44 >>>> ++++++++++++++++++++++ >>>> 1 file changed, 44 insertions(+) >>>> create mode 100644 >>>> Documentation/devicetree/bindings/remoteproc/imx-rproc.txt >>>> >>>> diff --git a/Documentation/devicetree/bindings/remoteproc/imx-rproc.txt >>>> b/Documentation/devicetree/bindings/remoteproc/imx-rproc.txt >>>> new file mode 100644 >>>> index 000000000000..e7c61993e1b8 >>>> --- /dev/null >>>> +++ b/Documentation/devicetree/bindings/remoteproc/imx-rproc.txt >>>> @@ -0,0 +1,44 @@ >>>> +NXP iMX6SX/iMX7D Co-Processor Bindings >>>> +---------------------------------------- >>>> + >>>> +This binding provides support for ARM Cortex M4 Co-processor found on some >>>> +NXP iMX SoCs. >>>> + >>>> +Required properties: >>>> +- compatible Should be one of: >>>> + "fsl,imx7d-rproc" >>>> + "fsl,imx6sx-rproc" >>>> +- clocks Clock for co-processor (See: >>>> ../clock/clock-bindings.txt) >>>> +- syscfg Phandle to syscon block which provide access to >>> >>> This is called "syscon" in the code and the example. >> >> done. >> >>>> + System Reset Controller >>>> + >>>> +Optional properties: >>>> +- reg: Should contain the address ranges for some of >>>> internal >>>> + memory regions. >>> >>> Something less hand-wavy, like: "Should list the memory regions for the >>> remoteproc" >>> >>>> +- reg-names: Contains the corresponding names for the memory >>>> + regions. These should be named "ddr", "ocram", >>>> "ocram-s", >>>> + "ocram-epdc" or "ocram-pxp". >>> >>> Make this comment define which memory regions are required for each of >>> the systems. >> >> done. >> >>>> +Fallowing memory ranges are expected: >>>> +- For "fsl,imx7d-rproc" >>>> + <0x00900000 0x00020000> - "ocram" >>>> + <0x00920000 0x00020000> - "ocram-epdc" >>>> + <0x00940000 0x00008000> - "ocram-pxp" >>>> + <0x80000000 0x0FFF0000> - "ddr" (code area) >>>> + <0x80000000 0x60000000> - "ddr" (data area) >>> >>> There's no reason to list the actual regions in the binding >>> document. Just list the requires regions for each system. >>> >>>> +- For "fsl,imx6sx-rproc" >>>> + <0x008F8000 0x00004000> - "ocram-s" >>>> + <0x80000000 0x0FFF8000> - "ddr" (code area) >>>> + <0x80000000 0x60000000> - "ddr" (data area) >>>> + >>>> +Note: the "ddr" code and data ranges are overlapping. Code area is smaller >>>> +than data area. So this range should be carefully chosen according to >>>> your >>>> +system and application requirements. >>>> + >>> >>> This is a source of future issues as this indicates that I should have >>> reg-names list "ddr" twice. >> >> Then I will name it: >> "ddri" (incstruction/code area), >> "ddrd" (data area) >> >> unless there are other suggestions. >> > > But are you saying that it's correct that these two memory regions > should overlap?
Yes, from Cortex-m4 the same memory regions are aliased to different address ranges to provide different cache optimizations. From Cortex-A7 side - not. But on this side we don't care about meaning of it, it is just some memory region. >>> Also, as you warn about the user needing to pick the values for "ddr", >>> does that mean that it's a carveout of System RAM? >> >> Yes, it is a part of System RAM. >> > > Then there will be an associated reserved-memory region for the > region(s), you should add a label to this and use "memory-region" to get > hold of the addresses instead. Ok. Should only system memory region be assigned over reserved-memory or all of named regions? -- Regards, Oleksij
signature.asc
Description: OpenPGP digital signature
