[PATCH v2 -next 1/4] RISC-V: ACPI: Enhance acpi_os_ioremap with MMIO remapping

[Alexandre Ghiti]

Hi Sunil,

On Wed, Sep 27, 2023 at 7:00 PM Sunil V L <sunilvl at ventanamicro.com> wrote:
>
> Enhance the acpi_os_ioremap() to support opregions in MMIO space. Also,
> have strict checks using EFI memory map to allow remapping the RAM similar
> to arm64.
>
> Signed-off-by: Sunil V L <sunilvl at ventanamicro.com>
> Reviewed-by: Andrew Jones <ajones at ventanamicro.com>
> ---
>  arch/riscv/Kconfig       |  1 +
>  arch/riscv/kernel/acpi.c | 87 +++++++++++++++++++++++++++++++++++++++-
>  2 files changed, 86 insertions(+), 2 deletions(-)
>
> diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> index d607ab0f7c6d..ac039cf8af7a 100644
> --- a/arch/riscv/Kconfig
> +++ b/arch/riscv/Kconfig
> @@ -39,6 +39,7 @@ config RISCV
>         select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
>         select ARCH_HAS_UBSAN_SANITIZE_ALL
>         select ARCH_HAS_VDSO_DATA
> +       select ARCH_KEEP_MEMBLOCK

Shouldn't we restrict this to ACPI?

>         select ARCH_OPTIONAL_KERNEL_RWX if ARCH_HAS_STRICT_KERNEL_RWX
>         select ARCH_OPTIONAL_KERNEL_RWX_DEFAULT
>         select ARCH_STACKWALK
> diff --git a/arch/riscv/kernel/acpi.c b/arch/riscv/kernel/acpi.c
> index 56cb2c986c48..e619edc8b0cc 100644
> --- a/arch/riscv/kernel/acpi.c
> +++ b/arch/riscv/kernel/acpi.c
> @@ -14,9 +14,10 @@
>   */
>
>  #include <linux/acpi.h>
> +#include <linux/efi.h>
>  #include <linux/io.h>
> +#include <linux/memblock.h>
>  #include <linux/pci.h>
> -#include <linux/efi.h>
>
>  int acpi_noirq = 1;            /* skip ACPI IRQ initialization */
>  int acpi_disabled = 1;
> @@ -217,7 +218,89 @@ void __init __acpi_unmap_table(void __iomem *map, unsigned long size)
>
>  void __iomem *acpi_os_ioremap(acpi_physical_address phys, acpi_size size)
>  {
> -       return (void __iomem *)memremap(phys, size, MEMREMAP_WB);
> +       efi_memory_desc_t *md, *region = NULL;
> +       pgprot_t prot;
> +
> +       if (WARN_ON_ONCE(!efi_enabled(EFI_MEMMAP)))
> +               return NULL;
> +
> +       for_each_efi_memory_desc(md) {
> +               u64 end = md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT);
> +
> +               if (phys < md->phys_addr || phys >= end)
> +                       continue;
> +
> +               if (phys + size > end) {
> +                       pr_warn(FW_BUG "requested region covers multiple EFI memory regions\n");
> +                       return NULL;
> +               }
> +               region = md;
> +               break;
> +       }
> +
> +       /*
> +        * It is fine for AML to remap regions that are not represented in the
> +        * EFI memory map at all, as it only describes normal memory, and MMIO
> +        * regions that require a virtual mapping to make them accessible to
> +        * the EFI runtime services.
> +        */
> +       prot = PAGE_KERNEL_IO;
> +       if (region) {
> +               switch (region->type) {
> +               case EFI_LOADER_CODE:
> +               case EFI_LOADER_DATA:
> +               case EFI_BOOT_SERVICES_CODE:
> +               case EFI_BOOT_SERVICES_DATA:
> +               case EFI_CONVENTIONAL_MEMORY:
> +               case EFI_PERSISTENT_MEMORY:
> +                       if (memblock_is_map_memory(phys) ||
> +                           !memblock_is_region_memory(phys, size)) {
> +                               pr_warn(FW_BUG "requested region covers kernel memory\n");
> +                               return NULL;
> +                       }
> +
> +                       /*
> +                        * Mapping kernel memory is permitted if the region in
> +                        * question is covered by a single memblock with the
> +                        * NOMAP attribute set: this enables the use of ACPI
> +                        * table overrides passed via initramfs.
> +                        * This particular use case only requires read access.
> +                        */
> +                       fallthrough;
> +
> +               case EFI_RUNTIME_SERVICES_CODE:
> +                       /*
> +                        * This would be unusual, but not problematic per se,
> +                        * as long as we take care not to create a writable
> +                        * mapping for executable code.
> +                        */
> +                       prot = PAGE_KERNEL_RO;
> +                       break;
> +
> +               case EFI_ACPI_RECLAIM_MEMORY:
> +                       /*
> +                        * ACPI reclaim memory is used to pass firmware tables
> +                        * and other data that is intended for consumption by
> +                        * the OS only, which may decide it wants to reclaim
> +                        * that memory and use it for something else. We never
> +                        * do that, but we usually add it to the linear map
> +                        * anyway, in which case we should use the existing
> +                        * mapping.
> +                        */
> +                       if (memblock_is_map_memory(phys))
> +                               return (void __iomem *)__va(phys);
> +                       fallthrough;
> +
> +               default:
> +                       if (region->attribute & EFI_MEMORY_WB)
> +                               prot = PAGE_KERNEL;
> +                       else if ((region->attribute & EFI_MEMORY_WC) ||
> +                                (region->attribute & EFI_MEMORY_WT))
> +                               prot = pgprot_writecombine(PAGE_KERNEL);

I have to ask: why is write-through mapped to write-combined here?

> +               }
> +       }
> +
> +       return ioremap_prot(phys, size, pgprot_val(prot));
>  }
>
>  #ifdef CONFIG_PCI
> --
> 2.39.2
>

Like Andrew said in v1, too bad we can't merge that with arm64 instead
of duplicating.

But otherwise, you can add:

Reviewed-by: Alexandre Ghiti <alexghiti at rivosinc.com>

Thanks,

Alex