Re: [PATCH V2 1/2] firmware: arm_scmi: Ensure that the message-id supports fastchannel

[Sibi Sankar]

On 9/5/24 18:24, Konrad Dybcio wrote:
> On 4.09.2024 4:20 PM, Cristian Marussi wrote:
> > On Wed, Sep 04, 2024 at 01:38:55PM +0100, Sudeep Holla wrote:
> > > On Wed, Sep 04, 2024 at 01:29:29PM +0200, Konrad Dybcio wrote:
> > > > On 4.09.2024 9:00 AM, Johan Hovold wrote:
> > >
> > > [...]
> > >
> > > > > Unfortunately, this patch breaks resume from suspend on the x1e80100 crd:
> > > > >
> > > > >          [   26.919676] CPU4: Booted secondary processor 0x0000010000 [0x511f0011]
> > > > >          [   26.960607] arm-scmi firmware:scmi: timed out in resp(caller: do_xfer+0x164/0x568)
> > > > >          [   26.987142] cpufreq: cpufreq_online: ->get() failed
> > > > >
> > > > > and then the machine hangs (mostly, I saw an nvme timeout message after a
> > > > > while).
> > > > >
> > > > > Make sure you test suspend as well as some of the warnings I reported
> > > > > only show up during suspend.
> > > >
> > > > Eh it looks like PERF_LEVEL_GET (msgid 8) requires the use of FC, but
> > > > the firmware fails to inform us about it through BIT(0) in attrs..
> > >
> > > Just trying to understand things better here. So the firmware expects OSPM
> > > to just use FC only for PERF_LEVEL_GET and hence doesn't implement the
> > > default/normal channel for PERF_LEVEL_GET(I assume it returns error ?)
> > > but fails to set the attribute indicating FC is available for the domain.
> >
> > Is not that FCs are optional BUT PERF_LEVEL_GET standard messages is
> > support is mandatory by the spec anyway ?
>
> So doing a bit of poking I think it's that FC is not marked as supported,
> but we need to read out the frequency from the .get_addr.. which is only
> populated if we go through fastchannel_init

On further debug it was found that the SCP was servicing the request
but mailbox had the interrupt disabled during suspend which caused the
timeout. I just re-spun the series wit hte fix. So yeah PERF_LEVEL_GET 
is expectedto work without any problems. There is no dependence on EC as
Konrad speculated. Just a straight forward case of interrupt being
disabled in the resume path.

-Sibi

> Konrad