Re: [PATCH v45 0/7] MCTP over PCC

[Adam Young]

On 7/21/26 13:52, Adam Young wrote:
> From: Linux Bot <linuxbot@xxxxxxxxxxxxxxxxxxx>
>
> THis patch sereis is long lived, and the main
> changelog will saty with the MCTP over PCC patch.
> Recent changes in AI code review haved shown required
> fixes in the PC mailbox layer. The main changes to this
> patch series are the addition of those patches.
>
> Adam Young (7):
>    mailbox/pcc.c: shmem map/unmap startup/teardown
>    mailbox/pcc.c: ignore errors on type 4 channels.
>    mailbox/pcc.c: report errors for PCC clients
>    mailbox/pcc.c:  add query channel function
>    mctp pcc: Implement MCTP over PCC Transport
>    synchronize IRQ before releasing shared memory
>    wrap pchan->chan_in_use  in READ/WRITE_ONCE
>
>   MAINTAINERS                 |   5 +
>   drivers/mailbox/pcc.c       | 112 ++++++---
>   drivers/net/mctp/Kconfig    |  15 ++
>   drivers/net/mctp/Makefile   |   1 +
>   drivers/net/mctp/mctp-pcc.c | 466 ++++++++++++++++++++++++++++++++++++
>   include/acpi/pcc.h          |   9 +
>   6 files changed, 576 insertions(+), 32 deletions(-)
>   create mode 100644 drivers/net/mctp/mctp-pcc.c

I am seeking feedback from both the PCC and the Networking teams on how 
to make progress here.  The MCTP Driver can be made functional with no 
changes to the PCC layer,  but it will have potential race conditions.  
Those will not be triggerable on the known hardware.  Can this driver go 
in as is, and we pursue the PCC fixes afterwards?

If not, the PCC fixes are blocking a driver that has been through a 
couple years worth of  code review, and should be prioritized. I put 
them all into this one patch series to simplify my own organization, and 
to provide a unified narrative for the networking reviewers.  However, 
the PCC code tends to be reviewed in linux-arm, and this organization is 
making it harder for people on that team.  I have been missing messages 
between reviewers.

At this point, I know that at least one of the patches in the PCC 
portion of this series needs to be rebased upon work by Sudeep Holla and 
Huisong Li.  I am reluctant to resubmit them right now for feat of 
generating unnecessary churn.  There code also obviates another patch in 
this series, which shows that we are addressing similar issues.