
On Tue, 25 Feb 2020 at 09:59, Chang, Abner (HPS SW/FW Technologist) abner.chang@hpe.com wrote:
-----Original Message----- From: Ard Biesheuvel [mailto:ard.biesheuvel@linaro.org] Sent: Tuesday, February 25, 2020 4:48 PM To: Chang, Abner (HPS SW/FW Technologist) abner.chang@hpe.com Cc: Atish Patra atishp@atishpatra.org; Heinrich Schuchardt xypron.glpk@gmx.de; Atish Patra atish.patra@wdc.com; U-Boot Mailing List u-boot@lists.denx.de; Alexander Graf agraf@csgraf.de; Anup Patel anup.patel@wdc.com; Bin Meng bmeng.cn@gmail.com; Joe Hershberger joe.hershberger@ni.com; Loic Pallardy loic.pallardy@st.com; Lukas Auer lukas.auer@aisec.fraunhofer.de; Marek BehĂșn marek.behun@nic.cz; Marek Vasut marek.vasut@gmail.com; Patrick Delaunay patrick.delaunay@st.com; Peng Fan peng.fan@nxp.com; Philippe Reynes philippe.reynes@softathome.com; Simon Glass sjg@chromium.org; Simon Goldschmidt simon.k.r.goldschmidt@gmail.com; Stefano Babic sbabic@denx.de; Thierry Reding treding@nvidia.com; Tom Rini trini@konsulko.com; leif@nuviainc.com; Schaefer, Daniel (DualStudy) daniel.schaefer@hpe.com Subject: Re: [RFC PATCH 0/1] Add boot hartid to a Device tree
On Tue, 25 Feb 2020 at 09:28, Chang, Abner (HPS SW/FW Technologist) abner.chang@hpe.com wrote:
-----Original Message----- From: Atish Patra [mailto:atishp@atishpatra.org]
<snip header soup>
On Mon, Feb 24, 2020 at 3:35 PM Ard Biesheuvel ard.biesheuvel@linaro.org wrote:
On Tue, 25 Feb 2020 at 00:22, Heinrich Schuchardt xypron.glpk@gmx.de
wrote:
On 2/24/20 11:19 PM, Atish Patra wrote: > The RISC-V booting protocol requires the hart id to be present in
"a0"
> register. This is not a problem for bootm/booti commands as > they directly jump to Linux kernel. However, bootefi jumps to > a EFI boot stub code in Linux kernel which acts a loader and > jumps to real Linux after terminating the boot services. This > boot stub code has to be aware of the boot hart id so that it > can set it in "a0" before jumping to Linux kernel. Currently, > UEFI protocol doesn't have any mechanism to pass the boot hart > id to an EFI executable. We should keep it this way as this is > a RISC-V specific requirement rather than a UEFI requirement. > Out of the all
possible options, device tree seemed to be the best choice to do this job.
> The detailed discussion can be found in the following thread. > > INVALID URI REMOVED >
abs.org_patch_1233664_&d=DwIBaQ&c=C5b8zRQO1miGmBeVZ2LFWg&r=_S
N6FZB
>
N4Vgi4Ulkskz6qU3NYRO03nHp9P7Z5q59A3E&m=J8GY_HS3fV_cJH9duXP739y
0hDK
> 3qfHGNx2Dpcf-UBY&s=iVpRlpTOME_A-
O5STNbXXawkW24gxy2yi56Q8AtZ2bI&e=
The above mentioned patch is obsoleted by the new suggestion.
Thanks for pointing that out to avoid confusion.
> > This patch updates the device tree in arch_fixup_fdt() which > is common for all booting commands. As a result, the DT > modification doesn't require any efi related arch specific > functions and all DT related modifications are contained at > one place. However, the hart id node will be available for > Linux even if the kernel is booted using
bootm command.
> > If that is not acceptable, we can always move the code to an > efi specific function.
Does a related Linux patch already exist?
Yes. But in my local tree ;). It will be included in RISC-V EFI stub support series which I am planning to post in a couple of days.
How about EDK2?
RISC-V is not supported at all yet in EDK2.
The EDK2 patches are out there and reviewed. I guess it will be available in mainline EDK2 pretty soon.
Yes, currently we are working on edk2 CI testing for RISCV64 arch. We
hope edk2 RISC-V port could be in mainstream in Mar.
Excellent! Is this core support? Or do you have a platform implemented as well that can be upstreamed?
Yes we do have platform implementations to be upstreamed, below is the latest status of RISC-V edk2 port. We will have to update status later because we just merged OpenSBI tag 0.6 to edk2 RISC-V. https://github.com/riscv/riscv-uefi-edk2-docs
Good to know! I saw some patches going by on the mailing list, but it is hard to derive the current state of affairs from that.
I'm glad to see you did not make the same mistake we made on ARM and omit the PEI phase entirely.
What I did notice is the use of APRIORI PEI and APRIORI DXE sections in your platform descriptions. I recommend you try to avoid that, as it is a maintenance burden going forward: instead, please use dummy protocols and NULL library class resolutions if you need to make generic components depend on platform specific protocols. Also, please document this - the APRIORI section does not explain *why* you have to circumvent the ordinary dependency tree based module dispatch.