[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #8 from Armin Wolf (W_Armin@gmx.de) ---
Yes, i need the output of acpidump to check what kind of WMI interface the AMW0
device exposes.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #9 from raoul (ein4rth@gmail.com) ---
Created attachment 308203
  --> https://bugzilla.kernel.org/attachment.cgi?id=308203&action=edit
acpidump and report

Sorry for the delay, here's the acpidump and a new s2idle report (generated
after disconnecting all devices and disabling keyboard wakeup as suggested by
Mario)

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #10 from raoul (ein4rth@gmail.com) ---
BTW out of curiosity I've tried a script to dynamically unload/load i8042 on
suspend (https://github.com/mistine/i8042-keyboard-fix-linux), but the problem
remains.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #11 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
Your new report looks "better" actually.  I see you're getting into hardware
sleep each time.  If you run a longer cycle (say 60s), I bet you get higher
residency too.

Is the keyboard "working" after you've resumed?  With all the errors listed for
it:

atkbd serio0: Failed to deactivate keyboard on isa0060/serio0
atkbd serio0: Failed to enable keyboard on isa0060/serio0
atkbd serio0: Spurious ACK on isa0060/serio0. Some program might be trying to
access hardware directly.

Other observations:
 * AMW0 notify only happens on the first cycle.  
 * Spurious ack events happen on all the cycles.
 * You've added i8042.noreset and i8042.nomux to kernel command line.

I would say drop the kernel i8042 parameters and just turn off the wakeup for
u8042 and try to run a few 60s cycles and see how things look.

I'll let Armin comment on the AMW0 device from the acpidump.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #12 from Armin Wolf (W_Armin@gmx.de) ---
The AMW0 device seems to be associated with the custom EC interface found on
devices manufactured by Uniwill (see
https://github.com/Wer-Wolf/uniwill-laptop). The notification was likely caused
by pressing a hotkey button.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #13 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
@raoul, did you press any buttons?  my understanding was the event looked
spurious.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #14 from raoul (ein4rth@gmail.com) ---
I didn't press anything after starting the s2idle command. I can run it again
or do other tests if needed.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #15 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
Can you try my suggestions from #11?

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #16 from raoul (ein4rth@gmail.com) ---
Created attachment 308205
  --> https://bugzilla.kernel.org/attachment.cgi?id=308205&action=edit
report with 60s cycles

Sure, here's a report for 60s duration per cycle, 5s wait, 5 cycles

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #17 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---

I still see the keyboard wakeup marked enabled.  Did you forget to disable it
before suspending?

│ AT Translated Set 2 keyboard [serio0]: enabled

Also I still see the extra setting on kernel command line:

/proc/cmdline: nvme_load=YES nowatchdog i8042.reset i8042.nomux
systemd.machine_id=0fbf90921f964c578cfe50596ce020e0

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

raoul (ein4rth@gmail.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #308205|0                           |1
        is obsolete|                            |

--- Comment #18 from raoul (ein4rth@gmail.com) ---
Created attachment 308206
  --> https://bugzilla.kernel.org/attachment.cgi?id=308206&action=edit
report with 60s cycles

Sorry, you're right. Rerun s2idle with 60s cycles.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #19 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
Everything looks solid for hardware sleep now.

Did you rmmod i8042 or did you disable wakeup?  It /looks/ like you might have
done rmmod.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #20 from raoul (ein4rth@gmail.com) ---
Created attachment 308209
  --> https://bugzilla.kernel.org/attachment.cgi?id=308209&action=edit
reports with and without wakeup disabled

No I didn't rmmod and I made sure to remove that i8042-keyboard-fix-linux
script I tried before. Here's another 2 tests I've just run, one before
disabling wakeups and one after disabling.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #21 from raoul (ein4rth@gmail.com) ---
Manually pressing the power button to suspend now actually works. I still get
those "spurious ack" you mentioned, and an ucsi_acpi error "failed to re-enable
notifications (-110)", but at least I can work.

So this has nothing to do with amdgpu... Should I file a bug report somewhere
else?

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

Mario Limonciello (AMD) (mario.limonciello@amd.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
          Component|x86-64                      |Input Devices
            Product|Platform Specific/Hardware  |Drivers

--- Comment #22 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
OK thanks.  It looks to me that this is most likely a problem with your
firmware.  
You can pass this over to linux-input guys to see if they agree.

But I feel you should report this to your manufacturer.

If the input guys don't find anything wrong with the driver something we "can"
do in Linux is default that i8042 wakeup to disabled, which is something we
have done for a number of other vendors with various firmware problems.

https://github.com/torvalds/linux/blob/v6.15/drivers/platform/x86/amd/pmc/pmc-quirks.c#L223

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

Mario Limonciello (AMD) (mario.limonciello@amd.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|mario.limonciello@amd.com   |drivers_input-devices@kerne
                   |                            |l-bugs.osdl.org

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #23 from raoul (ein4rth@gmail.com) ---
Thanks so much for your help!

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #24 from raoul (ein4rth@gmail.com) ---
Just reported both to https://lore.kernel.org/linux-input/ and to PCSpecialist.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #25 from raoul (ein4rth@gmail.com) ---
I haven't heard from the linux-input devs, but PCSpecialist replied they don't
care about Linux, so no chance for a bios fix. This keyboard also seems to
cause issues like lag after a while while I'm typing, but I couldn't find any
message in dmesg except for those ACPI errors.

Anyway - since disabling the wakeup fixed suspend with a consistent 88%
hardware sleep, I guess defaulting i8042 to disabled for this hardware could be
a good idea. I noticed other manufacturers share the same hardware (usually
Tuxedo and others release theirs after PCSpecialist), so maybe they're going to
find a fix in the future.

Thanks again, Mario.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #26 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
Created attachment 308225
  --> https://bugzilla.kernel.org/attachment.cgi?id=308225&action=edit
workaround patch

Well thanks for trying with the vendor, that's unfortunate to hear.  Let's try
the workaround that disables it automatically.

Can you confirm this helps (make sure you take any modifications to i8042 via
udev rules or kernel command line out of the picture)

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #27 from raoul (ein4rth@gmail.com) ---
Just to be sure before I compile, is the new .ident line in the patch correct?
(it says MECHREVO)

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #28 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
Whoops! Totally copy paste error. It should be harmless though.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #29 from raoul (ein4rth@gmail.com) ---
It works for me!

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

Mario Limonciello (AMD) (mario.limonciello@amd.com) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |CODE_FIX

--- Comment #30 from Mario Limonciello (AMD) (mario.limonciello@amd.com) ---
W/A is landed.  Will be in 6.16-rc3.

https://github.com/torvalds/linux/commit/9ba75ccad85708c5a484637dccc1fc59295b0a83

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.

[Bug 220116] amd_pmc - Last suspend didn't reach deepest state

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=220116

--- Comment #31 from raoul (ein4rth@gmail.com) ---
Thank you.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching someone on the CC list of the bug.