On 9.12.2024 5:23 AM, Raj Kumar Bhagat wrote: > On 12/6/2024 4:19 PM, Konrad Dybcio wrote: >> On 6.12.2024 5:34 AM, Raj Kumar Bhagat wrote: >>> On 12/5/2024 11:12 PM, Konrad Dybcio wrote: >>>> On 3.12.2024 10:18 AM, Raj Kumar Bhagat wrote: >>>>> On 11/4/2024 7:46 PM, Konrad Dybcio wrote: >>>>>> On 15.10.2024 8:26 PM, Raj Kumar Bhagat wrote: >>>>>>> The Ath2k AHB device (IPQ5332) firmware requests BDF_MEM_REGION_TYPE >>>>>>> memory during QMI memory requests. This memory is part of the >>>>>>> HOST_DDR_REGION_TYPE. Therefore, add the BDF memory address to the >>>>>>> hardware parameter and provide this memory address to the firmware >>>>>>> during QMI memory requests. >>>>>> >>>>>> Sounds like something to put in the device tree, no? >>>>>> >>>>> >>>>> This BDF memory address is the RAM offset. We did add this in device tree in >>>>> version 1. This is removed from device tree in v2 based on the review comment that >>>>> DT should not store RAM offset. >>>>> >>>>> refer below link: >>>>> Link: https://lore.kernel.org/all/f8cd9c3d-47e1-4709-9334-78e4790acef0@xxxxxxxxxx/ >>>> >>>> Right, I think this could be something under /reserved-memory instead >>>> >>> >>> Thanks for the suggestion. However, the BDF_MEM_REGION_TYPE is already within the >>> memory reserved for HOST_DDR_REGION_TYPE through /reserved-memory. Therefore, reserving >>> the memory for BDF_MEM_REGION_TYPE again in the Device Tree (DT) will cause a warning >>> for 'overlapping memory reservation'. >> >> Then you can grab a handle to it with of_reserved_mem_lookup() >> and of_reserved_mem_device_init_by_idx() >> > > The memory HOST_DDR_REGION_TYPE is a bigger memory around 43MB, while the memory > BDF_MEM_REGION_TYPE is smaller around 256KB within HOST_DDR_REGION_TYPE, Using the > above mentioned API we still have to store the offset in ath12k to point at memory > BDF_MEM_REGION_TYPE from the start of HOST_DDR_REGION_TYPE. That's still way better than hardcoding platform specifics in the common driver Konrad
