[PATCH v3 52/62] arm/acpi: Prepare EFI memory descriptor for Dom0
Stefano Stabellini
stefano.stabellini at eu.citrix.com
Mon Jan 4 06:54:16 PST 2016
On Thu, 31 Dec 2015, Shannon Zhao wrote:
> On 2015/11/27 22:30, Stefano Stabellini wrote:
> > On Tue, 17 Nov 2015, shannon.zhao at linaro.org wrote:
> >> > From: Shannon Zhao <shannon.zhao at linaro.org>
> >> >
> >> > Create a few EFI memory descriptors to tell Dom0 the RAM region
> >> > information, ACPI table regions and EFI tables reserved resions.
> >> >
> >> > Signed-off-by: Parth Dixit <parth.dixit at linaro.org>
> >> > Signed-off-by: Shannon Zhao <shannon.zhao at linaro.org>
> >> > ---
> >> > xen/arch/arm/domain_build.c | 2 ++
> >> > xen/common/efi/boot.c | 48 +++++++++++++++++++++++++++++++++++++++++++++
> >> > xen/include/asm-arm/setup.h | 4 ++++
> >> > 3 files changed, 54 insertions(+)
> >> >
> >> > diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c
> >> > index 9d667ea..073c634 100644
> >> > --- a/xen/arch/arm/domain_build.c
> >> > +++ b/xen/arch/arm/domain_build.c
> >> > @@ -1737,6 +1737,8 @@ static int prepare_acpi(struct domain *d, struct kernel_info *kinfo)
> >> > acpi_map_rest_tables(d);
> >> > acpi_create_efi_system_table(d->arch.efi_acpi_gpa, d->arch.efi_acpi_table,
> >> > tbl_add);
> >> > + acpi_create_efi_mmap_table(d->arch.efi_acpi_gpa, d->arch.efi_acpi_table,
> >> > + &kinfo->mem, tbl_add);
> >> >
> >> > return 0;
> >> > }
> >> > diff --git a/xen/common/efi/boot.c b/xen/common/efi/boot.c
> >> > index 75835ae..ff2faed 100644
> >> > --- a/xen/common/efi/boot.c
> >> > +++ b/xen/common/efi/boot.c
> >> > @@ -1256,6 +1256,54 @@ void __init acpi_create_efi_system_table(paddr_t paddr, void *efi_acpi_table,
> >> > tbl_add[TBL_EFIT].start = table_addr;
> >> > tbl_add[TBL_EFIT].size = table_size;
> >> > }
> >> > +
> >> > +void __init acpi_create_efi_mmap_table(paddr_t paddr, void *efi_acpi_table,
> >> > + const struct meminfo *mem,
> >> > + struct membank tbl_add[])
> > This function shouldn't be in this file.
> >
> >
> >
> >> > +{
> >> > + EFI_MEMORY_DESCRIPTOR *memory_map;
> >> > + int i, offset;
> >> > + u8 *base_ptr;
> >> > +
> >> > + tbl_add[TBL_MMAP].start = paddr + acpi_get_table_offset(tbl_add, TBL_MMAP);
> >> > + tbl_add[TBL_MMAP].size = sizeof(EFI_MEMORY_DESCRIPTOR)
> >> > + * (mem->nr_banks + acpi_mem.nr_banks + TBL_MMAX);
> >> > +
> >> > + base_ptr = efi_acpi_table + acpi_get_table_offset(tbl_add, TBL_MMAP);
> >> > + memory_map = (EFI_MEMORY_DESCRIPTOR *)(base_ptr);
> >> > +
> >> > + offset = 0;
> >> > + for( i = 0; i < mem->nr_banks; i++, offset++ )
> >> > + {
> >> > + memory_map[offset].Type = EfiConventionalMemory;
> >> > + memory_map[offset].PhysicalStart = mem->bank[i].start;
> >> > + memory_map[offset].NumberOfPages = mem->bank[i].size/PAGE_SIZE;
> > Use PAGE_SHIFT throughout the function.
> >
> >
> >> > + memory_map[offset].Attribute = EFI_MEMORY_WB;
> >> > + }
> >> > +
> >> > + for( i = 0; i < acpi_mem.nr_banks; i++, offset++ )
> >> > + {
> >> > + memory_map[offset].Type = EfiACPIReclaimMemory;
> >> > + memory_map[offset].PhysicalStart = acpi_mem.bank[i].start;
> >> > + memory_map[offset].NumberOfPages = acpi_mem.bank[i].size/PAGE_SIZE;
> >> > + memory_map[offset].Attribute = EFI_MEMORY_WB;
> >> > + }
> >> > +
> >> > + for( i = 0; i < TBL_EFIT; i++, offset++ )
> >> > + {
> >> > + memory_map[offset].Type = EfiACPIReclaimMemory;
> >> > + memory_map[offset].PhysicalStart = tbl_add[i].start;
> >> > + memory_map[offset].NumberOfPages =tbl_add[i].size/PAGE_SIZE;
> >> > + memory_map[offset].Attribute = EFI_MEMORY_WB;
> >> > + }
> > Can't we just use a single EFI_MEMORY_DESCRIPTOR (or maybe 2, to cover
> > EfiACPIReclaimMemory and EfiReservedMemoryType regions) to describe the
> > whole efi_acpi_gpa region? Using one EFI_MEMORY_DESCRIPTOR per table,
> > and PAGE_ALIGNing each of them, looks like a waste of memory to me.
> > Unless it's an ACPI or EFI spec requirement.
> >
>
> Looking at the UEFI SPEC it says
> "EFI memory descriptors of type EfiACPIReclaimMemory and EfiACPIMemoryNVS
> must be aligned on a 4 KiB boundary and must be a multiple of 4 KiB in size"
OK, but can we use only one, 4 KiB aligned, descriptor for all tables?
More information about the linux-arm-kernel
mailing list