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

[Mario Limonciello <superm1@kernel.org>]

On 10/26/25 12:40 PM, Mario Limonciello wrote:
> 
> 
> On 10/26/25 11:49 AM, Shyam Sundar S K wrote:
>> + Mario
>>
>> On 10/26/2025 22:16, Lars Francke wrote:
>>> Hi,
>>>
>>> thank you Mario for the super quick response.
>>> It arrived so fast I missed it.
>>>
>>> On Sun, Oct 26, 2025 at 3:17 PM Shyam Sundar S K
>>> <Shyam-sundar.S-k@amd.com> wrote:
>>>>> After resuming from hibernation I get this log line once a second:
>>>>>    amd-pmf AMDI0105:00: TEE enact cmd failed. err: ffff000e, ret:0
>>>>
>>>> Can you please share full dmesg log? Perhaps with both ccp and amd_tee
>>>> drivers enabled with dynamic debug.
>>>
>>> I have (with AI help, but reviewed manually) created a shell script to
>>> collect the necessary information. I have attached the full dmesg file
>>> to this mail but am not sure if that is proper etiquette here?
>>>
>>> I put up the script & more debug information in a dedicated
>>> repository: https://github.com/lfrancke/amd-pmf-hibernate/tree/main
>>>
>>>>> Right after hibernation and before the first of those errors I get:
>>>>>    ccp 0000:c3:00.2: tee: command 0x5 timed out, disabling PSP
>>>>>
> 
> This comes from tee_wait_cmd_completion().  Looking at the suspend/ 
> resume handling code from ccp.ko, I notice that we don't actually have 
> any special handling for the TEE ring for S4.  Normally that ring is 
> initialized at bootup (by sending PSP_CMD_TEE_RING_INIT to PSP).
> 
> During the S4 resume flow I don't expect that this has happened and is 
> needed explicitly in the restore() call.
> 
> Can you try this attached patch which inserts that flow to see if it helps?
> 
> Note mostly to myself: I need to double check flows but I think restore 
> also needs a pm_hibernate_is_recovering() check in this case too if this 
> works.
> 
>>>>> On boot it looks good:
>>>>>    ccp 0000:c3:00.2: psp enabled
>>>>
>>>> Do you see this message after hibernation exit?
>>>
>>> Yes. On a fresh boot it works without issues but after the first
>>> hibernate I get these.
>>>
>>>
>>>>> Commit 11e298f3548a6fe5e6ad78f811abfba15e6ebbc1 from 2024 has
>>>>> more or less exactly my error message but it doesn't seem
>>>>> to be fixed for this case.
>>>>> https://lore.kernel.org/all/20240216064112.962582-2-Shyam-sundar.S- 
>>>>> k@amd.com/
>>>>
>>>> This was a case fixed for S2Idle cases, but I seem to understand that
>>>> you are attempting hibernate here, right?
>>>
>>> Correct. I use hibernate.
>>>
>>> The issue I'm really trying to debug is a separate one (I'll report
>>> separately): Hibernate doesn't reliably turn off the laptop. Image is
>>> being written successfully but I have to press the power button for
>>> 4s.
> 
> Potential theory on root cause is that it's the same regression reported 
> here [1].
> 
> If so the root cause is [2], and the fix is [3].
> 
> Link:
> https://lore.kernel.org/linux-pm/a2563b83-e5d7-4d02- 
> a317-54d51e718d32@kernel.org/T/ 
> #m19966ea259837e4b02c6d0a55d67acfe15fc64ae [1]
> Link: https://git.kernel.org/torvalds/c/469d80a3712c6  [2]

Sorry - minor correction.  Root cause is [4].
> Link: https://lore.kernel.org/linux-pm/20251026033115.436448-1- 
> superm1@kernel.org/ [3]

Link: https://git.kernel.org/torvalds/c/449c9c02537a [4]