Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220518095646.47548-1-sunilvl@ventanamicro.com>
 <20220518095646.47548-2-sunilvl@ventanamicro.com> <CAMj1kXF2N+0zRCNg0Aq9Y8Y35FUaQ1V+swzY__y86sayKW7PPA@mail.gmail.com>
 <04ed20b1-e312-dfc2-c314-02750c076681@canonical.com>
In-Reply-To: <04ed20b1-e312-dfc2-c314-02750c076681@canonical.com>
From:   Ard Biesheuvel <ardb@kernel.org>
Date:   Thu, 19 May 2022 10:13:46 +0200
Message-ID: <CAMj1kXEX0DWprb7cv_js1=M1J1hSXLBa=cOOjVo3tTVyGT0siQ@mail.gmail.com>
Subject: Re: [PATCH V4 1/1] riscv/efi_stub: Add support for RISCV_EFI_BOOT_PROTOCOL
To:     Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Cc:     Sunil V L <sunilvl@ventanamicro.com>,
        Palmer Dabbelt <palmer@dabbelt.com>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        Albert Ou <aou@eecs.berkeley.edu>,
        Ilias Apalodimas <ilias.apalodimas@linaro.org>,
        Atish Patra <atishp@rivosinc.com>,
        Anup Patel <apatel@ventanamicro.com>,
        Jessica Clarke <jrtc27@jrtc27.com>,
        Abner Chang <abner.chang@hpe.com>,
        linux-efi <linux-efi@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        linux-riscv <linux-riscv@lists.infradead.org>,
        Palmer Dabbelt <palmer@rivosinc.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, 19 May 2022 at 08:17, Heinrich Schuchardt
<heinrich.schuchardt@canonical.com> wrote:
>
>
>
> On 5/18/22 23:03, Ard Biesheuvel wrote:
> > On Wed, 18 May 2022 at 11:57, Sunil V L <sunilvl@ventanamicro.com> wrote:
> >>
> >> This patch adds the support for getting the boot hart ID in
> >> Linux EFI stub using RISCV_EFI_BOOT_PROTOCOL. This protocol
> >> is preferred method over existing DT based solution since it
> >> works irrespective of DT or ACPI.
> >>
> >> The specification of the protocol is hosted at:
> >> https://github.com/riscv-non-isa/riscv-uefi
> >>
> >> Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
> >> Acked-by: Palmer Dabbelt <palmer@rivosinc.com>
> >> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
> >> ---
> >>   drivers/firmware/efi/libstub/efistub.h    |  7 ++++++
> >>   drivers/firmware/efi/libstub/riscv-stub.c | 29 +++++++++++++++++++----
> >>   include/linux/efi.h                       |  1 +
> >>   3 files changed, 32 insertions(+), 5 deletions(-)
> >>
> >> diff --git a/drivers/firmware/efi/libstub/efistub.h b/drivers/firmware/efi/libstub/efistub.h
> >> index edb77b0621ea..aced62a0907e 100644
> >> --- a/drivers/firmware/efi/libstub/efistub.h
> >> +++ b/drivers/firmware/efi/libstub/efistub.h
> >> @@ -720,6 +720,13 @@ union efi_tcg2_protocol {
> >>          } mixed_mode;
> >>   };
> >>
> >> +struct riscv_efi_boot_protocol {
> >> +       u64 revision;
> >> +
> >> +       efi_status_t (__efiapi * get_boot_hartid)(struct riscv_efi_boot_protocol *this,
> >> +                                                 size_t *boot_hartid);
> >
> > size_t is not a EFI type, and your spec uses UINTN here, which is
> > equivalent to 'unsigned long'. However, jump_kernel_func() takes a
> > unsigned int for the hartid. Please clean this up.
>
> unsigned long and size_t have the same number of bits. This seems to be
> a question of taste where we should follow the maintainer.
>

We use unsigned long wherever the UEFI spec uses UINTN. This is not a
matter of taste, really.

> jump_kernel_func() assuming boot hart ID to be an unsigned int is not in
> line with the RISC-V ISA which allows to use all xlen bits.
>
> >
> >
> >> +};
> >> +
> >>   typedef union efi_load_file_protocol efi_load_file_protocol_t;
> >>   typedef union efi_load_file_protocol efi_load_file2_protocol_t;
> >>
> >> diff --git a/drivers/firmware/efi/libstub/riscv-stub.c b/drivers/firmware/efi/libstub/riscv-stub.c
> >> index 9c460843442f..012504f6f9a4 100644
> >> --- a/drivers/firmware/efi/libstub/riscv-stub.c
> >> +++ b/drivers/firmware/efi/libstub/riscv-stub.c
> >> @@ -23,7 +23,7 @@
> >>
> >>   typedef void __noreturn (*jump_kernel_func)(unsigned int, unsigned long);
> >>
> >> -static u32 hartid;
> >> +static size_t hartid;
> >>
> >
> > and here
> >
> >>   static int get_boot_hartid_from_fdt(void)
> >>   {
> >> @@ -47,14 +47,33 @@ static int get_boot_hartid_from_fdt(void)
> >>          return 0;
> >>   }
> >>
> >> +static efi_status_t get_boot_hartid_from_efi(void)
> >> +{
> >> +       efi_guid_t boot_protocol_guid = RISCV_EFI_BOOT_PROTOCOL_GUID;
> >> +       efi_status_t status;
> >> +       struct riscv_efi_boot_protocol *boot_protocol;
> >> +
> >> +       status = efi_bs_call(locate_protocol, &boot_protocol_guid, NULL,
> >> +                            (void **)&boot_protocol);
> >> +       if (status == EFI_SUCCESS) {
> >> +               status = efi_call_proto(boot_protocol,
> >> +                                       get_boot_hartid, &hartid);
>
> A lot of the kernel code seems to be unfit to handle hart IDs exceeding
> INT_MAX (e.g. sbi_cpu_is_stopped()). Until this is fixed we have to
> treat hartid > INT_MAX as an error.
>

This is an issue in the core kernel code, not in the EFI stub. As you
pointed out, the ISA implies UINTN / unsigned long here, and if the
startup code cannot deal with that, it can be fixed separately.