Re: [RFC PATCH 5/5] RISC-V: Add EFI stub support.

[Atish Patra]

On Wed, 2020-02-26 at 08:28 +0100, Ard Biesheuvel wrote:
> On Wed, 26 Feb 2020 at 02:10, Atish Patra <atish.patra@xxxxxxx>
> wrote:
> > Add a RISC-V architecture specific stub code that actually copies
> > the
> > actual kernel image to a valid address and jump to it after boot
> > services
> > are terminated. Enable UEFI related kernel configs as well for
> > RISC-V.
> > 
> > Signed-off-by: Atish Patra <atish.patra@xxxxxxx>
> > ---
> >  arch/riscv/Kconfig                        |  20 ++++
> >  arch/riscv/Makefile                       |   1 +
> >  arch/riscv/configs/defconfig              |   1 +
> >  drivers/firmware/efi/libstub/Makefile     |   8 ++
> >  drivers/firmware/efi/libstub/riscv-stub.c | 135
> > ++++++++++++++++++++++
> >  5 files changed, 165 insertions(+)
> >  create mode 100644 drivers/firmware/efi/libstub/riscv-stub.c
> > 
> > diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> > index 42c122170cfd..68b1d565e51d 100644
> > --- a/arch/riscv/Kconfig
> > +++ b/arch/riscv/Kconfig
> > @@ -372,10 +372,30 @@ config CMDLINE_FORCE
> > 
> >  endchoice
> > 
> > +config EFI_STUB
> > +       bool
> > +
> > +config EFI
> > +       bool "UEFI runtime support"
> > +       depends on OF
> > +       select LIBFDT
> > +       select UCS2_STRING
> > +       select EFI_PARAMS_FROM_FDT
> > +       select EFI_STUB
> > +       select EFI_GENERIC_ARCH_STUB
> > +       default y
> > +       help
> > +         This option provides support for runtime services
> > provided
> > +         by UEFI firmware (such as non-volatile variables,
> > realtime
> > +          clock, and platform reset). A UEFI stub is also provided
> > to
> > +         allow the kernel to be booted as an EFI application. This
> > +         is only useful on systems that have UEFI firmware.
> > +
> >  endmenu
> > 
> >  menu "Power management options"
> > 
> >  source "kernel/power/Kconfig"
> > +source "drivers/firmware/Kconfig"
> > 
> >  endmenu
> > diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
> > index b9009a2fbaf5..0afaa89ba9ad 100644
> > --- a/arch/riscv/Makefile
> > +++ b/arch/riscv/Makefile
> > @@ -78,6 +78,7 @@ head-y := arch/riscv/kernel/head.o
> >  core-y += arch/riscv/
> > 
> >  libs-y += arch/riscv/lib/
> > +core-$(CONFIG_EFI_STUB) +=
> > $(objtree)/drivers/firmware/efi/libstub/lib.a
> > 
> >  PHONY += vdso_install
> >  vdso_install:
> > diff --git a/arch/riscv/configs/defconfig
> > b/arch/riscv/configs/defconfig
> > index e2ff95cb3390..0a5d3578f51e 100644
> > --- a/arch/riscv/configs/defconfig
> > +++ b/arch/riscv/configs/defconfig
> > @@ -125,3 +125,4 @@ CONFIG_DEBUG_BLOCK_EXT_DEVT=y
> >  # CONFIG_FTRACE is not set
> >  # CONFIG_RUNTIME_TESTING_MENU is not set
> >  CONFIG_MEMTEST=y
> > +CONFIG_EFI=y
> > diff --git a/drivers/firmware/efi/libstub/Makefile
> > b/drivers/firmware/efi/libstub/Makefile
> > index 2c5b76787126..38facb61745b 100644
> > --- a/drivers/firmware/efi/libstub/Makefile
> > +++ b/drivers/firmware/efi/libstub/Makefile
> > @@ -21,6 +21,8 @@ cflags-$(CONFIG_ARM64)                := $(subst
> > $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
> >  cflags-$(CONFIG_ARM)           := $(subst
> > $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
> >                                    -fno-builtin -fpic \
> >                                    $(call cc-option,-mno-single-
> > pic-base)
> > +cflags-$(CONFIG_RISCV)         := $(subst
> > $(CC_FLAGS_FTRACE),,$(KBUILD_CFLAGS)) \
> > +                                  -fpic
> > 
> >  cflags-$(CONFIG_EFI_GENERIC_ARCH_STUB) +=
> > -I$(srctree)/scripts/dtc/libfdt
> > 
> > @@ -55,6 +57,7 @@ lib-$(CONFIG_EFI_GENERIC_ARCH_STUB)           +=
> > efi-stub.o fdt.o string.o \
> >  lib-$(CONFIG_ARM)              += arm32-stub.o
> >  lib-$(CONFIG_ARM64)            += arm64-stub.o
> >  lib-$(CONFIG_X86)              += x86-stub.o
> > +lib-$(CONFIG_RISCV)            += riscv-stub.o
> >  CFLAGS_arm32-stub.o            := -DTEXT_OFFSET=$(TEXT_OFFSET)
> >  CFLAGS_arm64-stub.o            := -DTEXT_OFFSET=$(TEXT_OFFSET)
> > 
> > @@ -79,6 +82,11 @@ STUBCOPY_FLAGS-$(CONFIG_ARM64)       += --
> > prefix-alloc-sections=.init \
> >                                    --prefix-symbols=__efistub_
> >  STUBCOPY_RELOC-$(CONFIG_ARM64) := R_AARCH64_ABS
> > 
> > +STUBCOPY_FLAGS-$(CONFIG_RISCV) += --prefix-alloc-sections=.init \
> > +                                  --prefix-symbols=__efistub_
> > +STUBCOPY_RELOC-$(CONFIG_RISCV) := R_RISCV_HI20
> > +
> > +
> >  $(obj)/%.stub.o: $(obj)/%.o FORCE
> >         $(call if_changed,stubcopy)
> > 
> > diff --git a/drivers/firmware/efi/libstub/riscv-stub.c
> > b/drivers/firmware/efi/libstub/riscv-stub.c
> > new file mode 100644
> > index 000000000000..3935b29ea93a
> > --- /dev/null
> > +++ b/drivers/firmware/efi/libstub/riscv-stub.c
> > @@ -0,0 +1,135 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (C) 2013, 2014 Linaro Ltd;  <roy.franz@xxxxxxxxxx>
> > + * Copyright (C) 2020 Western Digital Corporation or its
> > affiliates.
> > + *
> > + * This file implements the EFI boot stub for the RISC-V kernel.
> > + * Adapted from ARM64 version at
> > drivers/firmware/efi/libstub/arm64-stub.c.
> > + */
> > +
> > +#include <linux/efi.h>
> > +#include <linux/libfdt.h>
> > +#include <linux/libfdt_env.h>
> > +#include <asm/efi.h>
> > +#include <asm/sections.h>
> > +
> > +#include "efistub.h"
> > +/*
> > + * RISCV requires the kernel image to placed TEXT_OFFSET bytes
> > beyond a 2 MB
> > + * aligned base for 64 bit and 4MB for 32 bit.
> > + */
> > +#if IS_ENABLED(CONFIG_64BIT)
> 
> You can use #ifdef here
> 

ok.

> > +#define MIN_KIMG_ALIGN SZ_2M
> > +#else
> > +#define MIN_KIMG_ALIGN SZ_4M
> > +#endif
> > +/*
> > + * TEXT_OFFSET ensures that we don't overwrite the firmware that
> > probably sits
> > + * at the beginning of the DRAM.
> > + */
> 
> Ugh. Really? On an EFI system, that memory should be reserved in some
> way, we shouldn't be able to stomp on it like that.
> 

Currently, we reserve the initial 128KB for run time firmware(only
openSBI for now, EDK2 later) by using PMP (physical memory protection).
Any acess to that region from supervisor mode (i.e. U-Boot) will result
in a fault.

Is it mandatory for UEFI to reserve the beginning of the DRAM ? 

> > +#define TEXT_OFFSET MIN_KIMG_ALIGN
> > +
> > +typedef __attribute__((noreturn)) void
> > (*jump_kernel_func)(unsigned int,
> > +                                                          unsigned
> > long);
> > +
> > +efi_status_t check_platform_features(void)
> > +{
> > +       return EFI_SUCCESS;
> > +}
> > +
> > +u64 get_boot_hartid_from_fdt(unsigned long fdt)
> 
> static
> 
> > +{
> > +       int chosen_node, len;
> > +       const fdt64_t *prop;
> > +       uint64_t hartid = U64_MAX;
> > +
> > +       chosen_node = fdt_path_offset((void *)fdt, "/chosen");
> > +       if (chosen_node < 0)
> > +               return hartid;
> 
> Just return U64_MAX here
> 
> > +       prop = fdt_getprop((void *)fdt, chosen_node, "efi-boot-
> > hartid", &len);
> 
> Please call this 'boot-hartid' not 'efi-boot-hartid' as the hartid
> value is independent of whether you boot via EFI or not.
> 
> > +       if (!prop || len != sizeof(u64))
> > +               return hartid;
> > +
> 
> Return U64_MAX
> 
> > +       hartid = fdt64_to_cpu(*prop);
> > +
> 
> and just return the swabbed value, so you can get rid of the local
> var.
> 

Fixed all the above issues. I changed it to u32 as u64 won't work on 32
bit systems.

> > +       return hartid;
> > +}
> > +
> > +/*
> > + * Jump to real kernel here with following constraints.
> > + * 1. MMU should be disabled.
> > + * 2. a0 should contain hartid
> > + * 3. a1 should DT address
> > + */
> > +void __noreturn efi_enter_kernel(unsigned long entrypoint,
> > unsigned long fdt)
> 
> This prototype has changed, and now includes the size of the fdt in
> param 3.
> 

Ahh yes. Fixed.

> > +{
> > +       unsigned long kernel_entry = entrypoint + _start_kernel -
> > _start;
> 
> stext_offset ? It has a terrible name though, and I'll probably
> propose to change it at some point, for all arches. But you can still
> use it here.
> 

Sure. I updated it with stext_offset.

> > +       jump_kernel_func jump_kernel = (void (*)(unsigned int,
> > unsigned long))kernel_entry;
> > +       u64 hartid = get_boot_hartid_from_fdt(fdt);
> > +
> > +       if (hartid == U64_MAX)
> > +               /* We can not use panic or BUG at this point */
> > +               __asm__ __volatile__ ("ebreak");
> > +       /* Disable MMU */
> > +       csr_write(CSR_SATP, 0);
> > +       jump_kernel(hartid, fdt);
> > +}
> > +
> > +efi_status_t handle_kernel_image(unsigned long *image_addr,
> > +                                unsigned long *image_size,
> > +                                unsigned long *reserve_addr,
> > +                                unsigned long *reserve_size,
> > +                                unsigned long dram_base,
> > +                                efi_loaded_image_t *image)
> > +{
> > +       efi_status_t status;
> > +       unsigned long kernel_size, kernel_memsize = 0;
> > +       unsigned long preferred_offset;
> > +
> > +       /*
> > +        * The preferred offset of the kernel Image is TEXT_OFFSET
> > bytes beyond
> > +        * a KIMG_ALIGN aligned base.
> > +        */
> > +       preferred_offset = round_up(dram_base, MIN_KIMG_ALIGN) +
> > TEXT_OFFSET;
> > +
> > +       kernel_size = _edata - _start;
> > +       kernel_memsize = kernel_size + (_end - _edata);
> > +
> > +       /*
> > +        * Try a straight allocation at the preferred offset.
> > +        * This will work around the issue where, if dram_base ==
> > 0x0,
> > +        * efi_low_alloc() refuses to allocate at 0x0 (to prevent
> > the
> > +        * address of the allocation to be mistaken for a FAIL
> > return
> > +        * value or a NULL pointer). It will also ensure that, on
> > +        * platforms where the [dram_base, dram_base + TEXT_OFFSET)
> > +        * interval is partially occupied by the firmware (like on
> > APM
> > +        * Mustang), we can still place the kernel at the address
> > +        * 'dram_base + TEXT_OFFSET'.
> 
> Better drop this entire last sentence (unless it is relevant, but
> then
> rework it to drop the APM Mustang reference)
> 

As stated above, RISC-V firmware occupies [dram_base, dram_base +
128K). That's why I thought this comment is useful. I should have
removed the mustand reference. I will update it.

> > +        */
> > +       if (*image_addr == preferred_offset)
> > +               return EFI_SUCCESS;
> > +
> > +       *image_addr = *reserve_addr = preferred_offset;
> > +       *reserve_size = round_up(kernel_memsize, EFI_ALLOC_ALIGN);
> > +
> > +       status = efi_bs_call(allocate_pages, EFI_ALLOCATE_ADDRESS,
> > +                               EFI_LOADER_DATA,
> > +                               *reserve_size / EFI_PAGE_SIZE,
> > +                               (efi_physical_addr_t
> > *)reserve_addr);
> > +
> > +       if (status != EFI_SUCCESS) {
> > +               *reserve_size = kernel_memsize + TEXT_OFFSET;
> > +               status = efi_low_alloc(*reserve_size,
> > MIN_KIMG_ALIGN,
> > +                                      reserve_addr);
> > +
> > +               if (status != EFI_SUCCESS) {
> > +                       pr_efi_err("Failed to relocate kernel\n");
> > +                       *reserve_size = 0;
> > +                       return status;
> > +               }
> > +               *image_addr = *reserve_addr + TEXT_OFFSET;
> > +       }
> > +       memcpy((void *)*image_addr, image->image_base,
> > kernel_size);
> > +
> > +       return EFI_SUCCESS;
> > +}
> > --
> > 2.24.0
> > 

-- 
Regards,
Atish