Kernel moving to Linux v5.10, dropping PECI

Andrei Kartashev a.kartashev at yadro.com
Thu Feb 11 00:40:51 AEDT 2021


On Wed, 2021-02-10 at 12:43 +0000, Joel Stanley wrote:
> The openbmc kernel will move to a 5.10 based tree for Aspeed and
> Nuoton machines.
> 
>     linux-openbmc: Move to Linux 5.10
> 
>     This moves the OpenBMC kernel to a v5.10 base for both Aspeed and
>     Nuvoton. There are 125 patches in the tree, with 80 of those
> patches not
>     yet queued for merging in v5.11.
> 
>     Notably the PECI patchset has been dropped as the author, Intel,
> has
>     elected to develop it out of tree instead of submitting it for
> mainline
>     inclusion.
> 
>     https://gerrit.openbmc-project.xyz/c/openbmc/openbmc/+/40404
> 
> Regarding the PECI situation, I raised it on the list back in August.
> The conversation finished up in October with a commitment that the
> work would be done as soon as possible.
> 
> This kernel config option is enabled by machines from Facebook,
> Bytedance, HPE, Lenovo, Quanta and Supermicro. (Surprisingly Intel
> doesn't enable it on their platform?). It would be great for someone
> from one of those teams to step up and submit the PECI patchset
> upstream.

Intel enable PECI in their downstream port 
https://github.com/Intel-BMC/openbmc , where they do have downstream
fork of the kernel with PECI patches: 
https://github.com/Intel-BMC/linux
We used to branch from the fork for our x86 platform, so now it's
really tricky for us to follow upstream. We will very appropriate if
one will push PECI patches upstream, but this is still Intel's code
under development and this sounds a bit risky if someone but Intel do
upstreaming. 
So I'd like first to see Intel's position about not to upstream the
patches: what is the problem there?

> 
> In the meantime these in-tree systems will regress their PECI support
> until the patchset is submitted to mainline.
> 
> Cheers,
> 
> Joel
-- 
Best regards,
Andrei Kartashev




More information about the openbmc mailing list