Re: AMD PMF: CCP PSP fails to reinitialize after hibernation causing TEE errors

[Mario Limonciello <superm1@kernel.org>]

> Sorry for the long silence, just back from vacation.
> 
> I have now tried these three patches on 6.18-rc5:
> 
> * 0001-crypto-ccp-Add-an-S4-restore-flow.patch
> * 0001-crypto-ccp-Prevent-calling-psp_tee_process_cmd-with-.patch
> * 0002-platform-x86-amd-pmf-Prevent-TEE-errors-after-hibern.patch
> 
> The good news: The original error (TEE enact cmd) is gone after hibernate.
> The bad news:
> * I tried four hibernate cycles
> * Two of them did shut down the laptop, two did not
> * After resuming from hibernate I could not reenter hibernate, reboot
> or shutdown at all on one occasion
> * I don't have logs for any of these :(

I think you are seeing an unrelated problem those other two cycles.

I've done testing myself (on top of 6.18-rc6) and they are testing solid 
for me.  I've posted the patches to the mailing list for review.

Link: 
https://lore.kernel.org/platform-driver-x86/20251119205942.3411155-1-superm1@kernel.org/

> 
> I assume these two are unrelated:
> * On boot I now get a message "RDSEED32 is broken. Disabling
> corresponding CPUID bit." before I enter my LUKS password

Yes; this is AMD SB-7055.  There is an updated linux-firmware microcode 
available for it.

> * Twice WiFi failed to reconnect after boot ("send auth to" followed
> by "authentication timed out") - I'm 100% sure this is not an issue
> with my home setup

Yes, should be unrelated to the PMF/CCP S4 issue.

> 
> I have not tried to debug any of these yet as I'm just back at my
> laptop but I can try to look into them a bit more if there's interest.
> 
> Thanks,
> Lars

I think if you can revisit on 6.18-rc6 with some more testing we can see 
how things are going.  FYI - the second patch for psp_tee_process_cmd 
appears to be unnecessary.  I couldn't reproduce any failures that 
required it.