[BUG] New arm scmi check in linux-next causing rk3568 not to boot due to firmware bug

[Nicolas Frattaroli]

On Donnerstag, 5. Mai 2022 12:10:32 CEST Sudeep Holla wrote:
> > > > > On Wed, May 04, 2022 at 02:49:07PM +0200, Nicolas Frattaroli wrote:
> > > > > > Good day,
> > > > > >
> > > > > > a user on the #linux-rockchip channel on the Libera.chat IRC network
> > > > > > reported that their RK3568 was no longer getting a CPU and GPU clock
> > > > > > from scmi and consequently not booting when using linux-next. This
> > > > > > was bisected down to the following commit:
> 
> OK I missed to read the above properly earlier. If scmi probe failure is
> resulting in Linux boot failure, then that is another bug that needs fixing.
> Why does not getting CPU clock block the boot. I would like to see the boot
> logs. I considered this issue to be non-fatal and must be just ending up
> disabling all SCMI communication. But the reported issue is boot failure
> which sounds like another/different bug and I would like that to be fixed
> first before we push the workaround for the reported issue so that it is
> not ignored.
> 
> Has anyone analysed why the absence of CPU clock results in boot failure ?
> Are you running the upstream kernel itself ?
> 
> 

Hello,

I'm sorry, I seem to have misinterpreted the original user's messages
as having been a boot failure. Upon re-reading the logs, this doesn't
seem to have been explicitly mentioned. I therefore assume this wasn't
causing a failure to boot.

Sadly the user isn't in the IRC channel at this moment so I cannot ask
them further questions.

I have tested this out on my own RK3566 based platform, and found that
we get the following:

$ sudo dmesg | grep arm-scmi
[    0.247134] arm-scmi firmware:scmi: Enabled polling mode TX channel - prot_id:16
[    0.247526] arm-scmi firmware:scmi: SCMI Notifications - Core Enabled.
[    0.247760] arm-scmi firmware:scmi: Malformed reply - real_sz:8  calc_sz:4
[    0.247773] arm-scmi firmware:scmi: SCMI Protocol v2.0 'rockchip:' Firmware version 0x0
[    0.247920] arm-scmi firmware:scmi: SCMI protocol 20 not implemented
[    1.545441] arm-scmi firmware:scmi: Failed. SCMI protocol 20 not active.
[    1.562958] arm-scmi firmware:scmi: Failed. SCMI protocol 23 not active.
[    1.565676] arm-scmi firmware:scmi: Failed. SCMI protocol 22 not active.
[    2.094446] arm-scmi firmware:scmi: Failed. SCMI protocol 21 not active.
[    2.103474] arm-scmi firmware:scmi: Failed. SCMI protocol 19 not active.
[    5.586871] arm-scmi firmware:scmi: Failed. SCMI protocol 17 not active.
[    5.593178] arm-scmi firmware:scmi: Failed. SCMI protocol 21 not active.

$ sudo dmesg | grep clk
[   18.255901] panfrost fde60000.gpu: clk init failed -517
[   18.686720] panfrost fde60000.gpu: clk init failed -517

The system does boot, it's just awfully slow. This is not a boot failure,
but arguably still a pretty bad failure mode to find oneself in.

Sorry for the confusion this caused.

Regards,
Nicolas Frattaroli