s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Thorsten Leemhuis]

Hi Mario! Did you by chance hear anything about lock-ups on resume with
mainline affecting AMD systems? I'm asking, as s2idle stopped working
for me with 6.6-rc on my Thinkpad T14 G1 (AMD). I'm as of now not sure
exactly when the problem started, as 6.6-rc1 didn't boot for me due to
another regression fixed for -rc2. But 6.6-rc2 was affected and -rc3
still is. Haven't found time for a bisection yet and will unlikely to be
so in the next two or three days, that's why I thought "ask Mario, with
a bit of luck he might be aware of a problem or even a fix".

FWIW, I ran your amd_s2idle.py script, but the machine seems to lock up
hard on resume, so I guess it's not much or help.

https://www.leemhuis.info/files/misc/s2idle_report-6.5.4-2023-09-25.txt
https://www.leemhuis.info/files/misc/s2idle_report-6.6_rc3-2023-09-25.txt

Dmesg for 6.6-rc3 (with a USB-C dock, but problem happens without it, too):
https://www.leemhuis.info/files/misc/dmesg-6.6_rc3

Ciao, Thorsten

#regzbot introduced v6.5..v6.6-rc2

Re: s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Mario Limonciello]

On 9/25/2023 05:55, Thorsten Leemhuis wrote:
> Hi Mario! Did you by chance hear anything about lock-ups on resume with
> mainline affecting AMD systems? I'm asking, as s2idle stopped working
> for me with 6.6-rc on my Thinkpad T14 G1 (AMD). 

You're the first report I've heard of this.  I've personally done s2idle 
on 6.6-rc1 + patch to fix the UCSI issue and 6.6-rc2 on a newer model 
(Lenovo Z13).

I haven't checked 6.6-rc3.

> I'm as of now not sure
> exactly when the problem started, as 6.6-rc1 didn't boot for me due to
> another regression fixed for -rc2. But 6.6-rc2 was affected and -rc3
> still is. Haven't found time for a bisection yet and will unlikely to be
> so in the next two or three days, that's why I thought "ask Mario, with
> a bit of luck he might be aware of a problem or even a fix".
> 

For any bisect steps that are failing to boot, I suggest you either 
compile without UCSI support (it shouldn't be critical) or apply

f26a679ed799 ("usb: typec: ucsi: Fix NULL pointer dereference")

> FWIW, I ran your amd_s2idle.py script, but the machine seems to lock up
> hard on resume, so I guess it's not much or help.
> 
> https://www.leemhuis.info/files/misc/s2idle_report-6.5.4-2023-09-25.txt
> https://www.leemhuis.info/files/misc/s2idle_report-6.6_rc3-2023-09-25.txt
> 

Yeah not too much useful stuff here unfortunately.  Can you check 
/var/lib/systemd/pstore to see if you got a traceback for whatever 
caused the problem?

> Dmesg for 6.6-rc3 (with a USB-C dock, but problem happens without it, too):
> https://www.leemhuis.info/files/misc/dmesg-6.6_rc3
> 
> Ciao, Thorsten
> 
> #regzbot introduced v6.5..v6.6-rc2

Being G1, can you try "amdgpu.mcbp=0" on kernel command line?

Re: s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Thorsten Leemhuis]

On 25.09.23 15:25, Mario Limonciello wrote:
> On 9/25/2023 05:55, Thorsten Leemhuis wrote:
>> Hi Mario! Did you by chance hear anything about lock-ups on resume with
>> mainline affecting AMD systems? I'm asking, as s2idle stopped working
>> for me with 6.6-rc on my Thinkpad T14 G1 (AMD). 
> You're the first report I've heard of this.

Thx for taking the time and replying.

> Yeah not too much useful stuff here unfortunately.  Can you check
> /var/lib/systemd/pstore to see if you got a traceback for whatever
> caused the problem?

Sadly nothing there.

>> Dmesg for 6.6-rc3 (with a USB-C dock, but problem happens without it,
>> too):
>> https://www.leemhuis.info/files/misc/dmesg-6.6_rc3
> 
> Being G1, can you try "amdgpu.mcbp=0" on kernel command line?

Just tried it, but didn't help.

Guess it bisection time then...

Thx again for your reply. Ciao, Thorsten

Re: s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Mario Limonciello]

On 9/25/2023 08:45, Thorsten Leemhuis wrote:
> On 25.09.23 15:25, Mario Limonciello wrote:
>> On 9/25/2023 05:55, Thorsten Leemhuis wrote:
>>> Hi Mario! Did you by chance hear anything about lock-ups on resume with
>>> mainline affecting AMD systems? I'm asking, as s2idle stopped working
>>> for me with 6.6-rc on my Thinkpad T14 G1 (AMD).
>> You're the first report I've heard of this.
> 
> Thx for taking the time and replying.
> 
>> Yeah not too much useful stuff here unfortunately.  Can you check
>> /var/lib/systemd/pstore to see if you got a traceback for whatever
>> caused the problem?
> 
> Sadly nothing there.
> 
>>> Dmesg for 6.6-rc3 (with a USB-C dock, but problem happens without it,
>>> too):
>>> https://www.leemhuis.info/files/misc/dmesg-6.6_rc3
>>
>> Being G1, can you try "amdgpu.mcbp=0" on kernel command line?
> 
> Just tried it, but didn't help.
> 
> Guess it bisection time then...
> 
> Thx again for your reply. Ciao, Thorsten

FWIW your email prompted me to double check the two other Lenovo systems 
I have on hand at my desk:
* P14s Gen 1
* L14 Gen2a

One of them had a Ubuntu userspace (20.04 IIRC) and the other a Fedora 
38.  I built 6.6-rc3 and couldn't reproduce any issues with s2idle.

So yes; bisect is the next step, thanks.

Re: s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Linux regression tracking (Thorsten Leemhuis)]

On 25.09.23 19:44, Mario Limonciello wrote:
> On 9/25/2023 08:45, Thorsten Leemhuis wrote:
>> On 25.09.23 15:25, Mario Limonciello wrote:
>>> On 9/25/2023 05:55, Thorsten Leemhuis wrote:
>>>> Hi Mario! Did you by chance hear anything about lock-ups on resume with
>>>> mainline affecting AMD systems? I'm asking, as s2idle stopped working
>>>> for me with 6.6-rc on my Thinkpad T14 G1 (AMD).
>>> You're the first report I've heard of this.
>> Thx for taking the time and replying.
> [...]
> FWIW your email prompted me to double check the two other Lenovo systems
> I have on hand at my desk:
> * P14s Gen 1
> * L14 Gen2a
> 
> One of them had a Ubuntu userspace (20.04 IIRC) and the other a Fedora
> 38.  I built 6.6-rc3 and couldn't reproduce any issues with s2idle.
> 
> So yes; bisect is the next step, thanks.

Thx again for your work. Turns out it was

Re: s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Thorsten Leemhuis]

[one more time, please ignore the previous message]

On 25.09.23 19:44, Mario Limonciello wrote:
> On 9/25/2023 08:45, Thorsten Leemhuis wrote:
>> On 25.09.23 15:25, Mario Limonciello wrote:
>>> On 9/25/2023 05:55, Thorsten Leemhuis wrote:
>>>> Hi Mario! Did you by chance hear anything about lock-ups on resume with
>>>> mainline affecting AMD systems? I'm asking, as s2idle stopped working
>>>> for me with 6.6-rc on my Thinkpad T14 G1 (AMD).
>>> You're the first report I've heard of this.
>> Thx for taking the time and replying.
> [...]
> FWIW your email prompted me to double check the two other Lenovo systems
> I have on hand at my desk:
> * P14s Gen 1
> * L14 Gen2a
> 
> One of them had a Ubuntu userspace (20.04 IIRC) and the other a Fedora
> 38.  I built 6.6-rc3 and couldn't reproduce any issues with s2idle.
> 
> So yes; bisect is the next step, thanks.

Thx again for your work.

FWIW, turned out the problem is caused by 92e24e0e57f72e ("Input:
psmouse - add delay when deactivating for SMBus mode"). Guess it's a
timing issue. Reported it in a new thread:

#regzbot dup-of:
https://lore.kernel.org/regressions/ca0109fa-c64b-43c1-a651-75b294d750a1@leemhuis.info/

Ciao, Thorsten

Re: s2idle stopped working with 6.6-rc on Thinkpad T14 G1 (AMD)

[Mario Limonciello]

On 9/27/2023 03:57, Thorsten Leemhuis wrote:
> [one more time, please ignore the previous message]
> 
> On 25.09.23 19:44, Mario Limonciello wrote:
>> On 9/25/2023 08:45, Thorsten Leemhuis wrote:
>>> On 25.09.23 15:25, Mario Limonciello wrote:
>>>> On 9/25/2023 05:55, Thorsten Leemhuis wrote:
>>>>> Hi Mario! Did you by chance hear anything about lock-ups on resume with
>>>>> mainline affecting AMD systems? I'm asking, as s2idle stopped working
>>>>> for me with 6.6-rc on my Thinkpad T14 G1 (AMD).
>>>> You're the first report I've heard of this.
>>> Thx for taking the time and replying.
>> [...]
>> FWIW your email prompted me to double check the two other Lenovo systems
>> I have on hand at my desk:
>> * P14s Gen 1
>> * L14 Gen2a
>>
>> One of them had a Ubuntu userspace (20.04 IIRC) and the other a Fedora
>> 38.  I built 6.6-rc3 and couldn't reproduce any issues with s2idle.
>>
>> So yes; bisect is the next step, thanks.
> 
> Thx again for your work.
> 
> FWIW, turned out the problem is caused by 92e24e0e57f72e ("Input:
> psmouse - add delay when deactivating for SMBus mode"). Guess it's a
> timing issue. Reported it in a new thread:
> 
> #regzbot dup-of:
> https://lore.kernel.org/regressions/ca0109fa-c64b-43c1-a651-75b294d750a1@leemhuis.info/
> 
> Ciao, Thorsten
> 
> 
> 
> 

Sounds good, thanks for letting me know.