[PATCH v2 06/10] rpi4: add a mapping for the PCIe XHCI controller MMIO registers (ARM 32bit)

[Marek Szyprowski]

Hi Matthias,

On 08.05.2020 23:26, Matthias Brugger wrote:
> Adding Tom as he is the arm maintainer.
>
> On 04/05/2020 14:45, Sylwester Nawrocki wrote:
>> From: Marek Szyprowski <m.szyprowski at samsung.com>
>>
>> Create a non-cacheable mapping for the 0x600000000 physical memory region,
>> where MMIO registers for the PCIe XHCI controller are instantiated by the
>> PCIe bridge. Due to 32bit limit in the CPU virtual address space in ARM
>> 32bit mode, this region is mapped at 0xff800000 CPU virtual address.
>>
>> Signed-off-by: Marek Szyprowski <m.szyprowski at samsung.com>
>> Signed-off-by: Sylwester Nawrocki <s.nawrocki at samsung.com>
>> ---
>> Changes since v1:
>>   - none.
>> ---
>>   arch/arm/mach-bcm283x/Kconfig             |  1 +
>>   arch/arm/mach-bcm283x/include/mach/base.h |  7 +++++
>>   arch/arm/mach-bcm283x/init.c              | 52 +++++++++++++++++++++++++++++++
>>   3 files changed, 60 insertions(+)
>>
>> diff --git a/arch/arm/mach-bcm283x/Kconfig b/arch/arm/mach-bcm283x/Kconfig
>> index 00419bf..bcb7f1d 100644
>> --- a/arch/arm/mach-bcm283x/Kconfig
>> +++ b/arch/arm/mach-bcm283x/Kconfig
>> @@ -36,6 +36,7 @@ config BCM2711_32B
>>   	select BCM2711
>>   	select ARMV7_LPAE
>>   	select CPU_V7A
>> +	select PHYS_64BIT
>>   
>>   config BCM2711_64B
>>   	bool "Broadcom BCM2711 SoC 64-bit support"
>> diff --git a/arch/arm/mach-bcm283x/include/mach/base.h b/arch/arm/mach-bcm283x/include/mach/base.h
>> index c4ae398..1d10dc9 100644
>> --- a/arch/arm/mach-bcm283x/include/mach/base.h
>> +++ b/arch/arm/mach-bcm283x/include/mach/base.h
>> @@ -6,6 +6,13 @@
>>   #ifndef _BCM283x_BASE_H_
>>   #define _BCM283x_BASE_H_
>>   
>> +#include <linux/types.h>
>> +
>>   extern unsigned long rpi_bcm283x_base;
>>   
>> +#ifdef CONFIG_ARMV7_LPAE
>> +extern void *rpi4_phys_to_virt(phys_addr_t paddr);
>> +#define phys_to_virt(x) rpi4_phys_to_virt(x)
>> +#endif
>> +
>>   #endif
>> diff --git a/arch/arm/mach-bcm283x/init.c b/arch/arm/mach-bcm283x/init.c
>> index 6a748da..5d0d160 100644
>> --- a/arch/arm/mach-bcm283x/init.c
>> +++ b/arch/arm/mach-bcm283x/init.c
>> @@ -145,6 +145,58 @@ int mach_cpu_init(void)
>>   }
>>   
>>   #ifdef CONFIG_ARMV7_LPAE
>> +
>> +#define BCM2711_RPI4_PCIE_XHCI_MMIO_VIRT	0xff800000UL
>> +
>> +void *rpi4_phys_to_virt(phys_addr_t paddr)
>> +{
>> +	if (paddr >= BCM2711_RPI4_PCIE_XHCI_MMIO_PHYS)
>> +		paddr = paddr - BCM2711_RPI4_PCIE_XHCI_MMIO_PHYS +
>> +			BCM2711_RPI4_PCIE_XHCI_MMIO_VIRT;
>> +	return (void *)(unsigned long)paddr;
> I think for this cases we have addrmap_phys_to_virt() which up to now is only
> used by powerpc.
>
>> +}
>> +
>> +static void set_section_phys(unsigned int section, phys_addr_t phys,
>> +			     enum dcache_option option)
>> +{
>> +	u64 *page_table = (u64 *)gd->arch.tlb_addr;
>> +	/* Need to set the access flag to not fault */
>> +	u64 value = TTB_SECT_AP | TTB_SECT_AF;
>> +
>> +	/* Add the page offset */
>> +	value |= (phys);
>> +
>> +	/* Add caching bits */
>> +	value |= option;
>> +
>> +	/* Set PTE */
>> +	page_table[section] = value;
>> +}
>> +
>> +static void rpi4_create_pcie_xhci_mapping(void)
>> +{
>> +	unsigned sect = BCM2711_RPI4_PCIE_XHCI_MMIO_VIRT >> MMU_SECTION_SHIFT;
>> +	phys_addr_t phys_addr = BCM2711_RPI4_PCIE_XHCI_MMIO_PHYS;
>> +	unsigned int size = BCM2711_RPI4_PCIE_XHCI_MMIO_SIZE;
>> +
>> +	while (size) {
>> +		set_section_phys(sect, phys_addr, DCACHE_OFF);
>> +		sect++;
>> +		phys_addr += MMU_SECTION_SIZE;
>> +		size -= MMU_SECTION_SIZE;
>> +	}
>> +}
> I wonder if we can't do all this in the pcie driver probe function. Maybe we can
> create a new function like mmu_set_region_dcache_behaviour_phys which allows us
> to update a mapping that's not 1:1.
>
> Tom what do you think?

In theory this could be moved to pcie probe, but then how the pcie 
driver would know WHERE in the virtual address space it should create a 
mapping for the needed MMIO area? IMHO this is something SoC or board 
specific and creating such mapping in board init is justified.

I will adapt the code to use addr_map helpers.

Best regards

-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland