[Bug 221065] ideapad_acpi: unexpected charge_types spam on Yoga Pro 7 14ASP9

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=221065

--- Comment #80 from Rong Zhang (i@rong.moe) ---
(In reply to Rick from comment #78)
> (In reply to Rong Zhang from comment #77)
> > (In reply to Rick from comment #76)
> > Hmm, interesting. I assume the pauses had existed before I introduced the
> > warning message (unexpected charge_types...), right?
> 
> Correct - this pause happens in Kernel 6.17 with no messages, when I load
> Kernel 6.19 I started getting the messages and that lead me to this thread.
> I tried Kernel 6.16 and the pause does not happen - so I believe this
> behavior was introduced in Kernel 6.17

Oh, that was when the charge_types attribute was added to ideapad-laptop.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=da8f2708f9b69707f4efeb432a18395e46b4666f

> > This sounds like a slow SMI handling due to the broken firmware.
> > 
> > Could you kindly provide more detailed information about the pause? When it
> > happened, did you notice a spike in CPU usage? Did you see any dmesg
> message
> > at that time? Did it freeze your system completely or just make things
> laggy?
> 
> The system appears to freeze solid for about 1 to 2 seconds (hard to tell
> the exact timing - but the pause feels like it is always the same duration).
> Nothing on screen changes.
> 
> When running system monitor during the event I see it freeze but when it
> restarts I don't see any spikes or anything interesting on any graph.

Did it have a power consumption graph and a CPU frequency graph? I will talk
more about monitoring later.

> 
> When I was running the "perf stat" stuff below I did notice that during the
> freeze it stopped updating and when it resumed it looked like it filled in
> all the data that was paused. Normally the perf stat seems to just produce a
> row of data on a periodic basis and when the pause happens it stops but
> after the freeze it looks like it prints a bunch of rows to catch up to
> where it should be.
> 
> There is no laggy behavior either before or after the freeze. The system
> works perfectly at all times except for the duration of the freeze. 
> 
> No message in dmesg under Kernel 6.17 (the default kernel in Mint - my daily
> driver).
> 
> The dmesg message shows up (Kernel 6.19 used for this) as soon as the freeze
> is finished (here are three consecutive events captured using "dmesg -w" so
> you can see the timing between them:
> [ 107.554265] ideapad_acpi VPC2004:00: unexpected charge_types: both [Fast]
> and [Long_Life] are enabled
> [ 501.851737] ideapad_acpi VPC2004:00: unexpected charge_types: both [Fast]
> and [Long_Life] are enabled
> [ 933.076412] ideapad_acpi VPC2004:00: unexpected charge_types: both [Fast]
> and [Long_Life] are enabled

So I speculate that some userspace utilities (probably upower or the desktop
environment?) read charge_types regularly or upon power events. Before v6.17,
ideapad-laptop didn't support it, so nothing could be read. However, since
v6.17, they can actually read it, and reading it triggers ECMT chaos upon power
events.

> 
> > 
> > Could you also monitor the SMI count, i.e.,
> > 
> >   sudo perf stat -e ls_smi_rx -I 1000
> > 
> > ..., wait for the pause to occur, and check how many SMIs have occurred
> > during the pause?
> 
> There number stayed at 0 the entire time I was testing the pause - the pause
> did not make the number increment.

Hmm, that's weird...

> Let me clarify this - After further testing: the mouse cursor continues to
> move on screen during a pause event.

Thanks a lot for the information. It's very helpful.

I suppose it could be a CPU/GPU throttle that delayed some rendering work. On
modern desktop environments, the cursor is rendered on a different layer --
that may be why it continues to move.

> For testing I was in a text editor scrolling a long document up and down
> (using mouse wheel) while moving the mouse in a circle.  When the pause
> happens the document stops scrolling and the performance monitor stops
> updating.
> 
> During the freeze the mouse cursor stayed fully responsive to my movements
> and did not seem laggy or out of the ordinary.
> 
> For reference I am running Linux Mint 22.3 - Cinnamon 64-bit - Cinnamon 6.6.7
> 
> I was also typing this comment with the ideapad_laptop kernel module enabled
> and I think the pause also allows keyboard input to buffer and not be
> ignored. (I was typing when a pause happened and I could have sworn that a
> key didn't register on screen so I hit it again and when the pause ended I
> ended up with two of the characters).

Sounds like rendering delays, too. Such rendering delays usually don't freeze
remote shells, e.g., SSH.

Could you try monitoring your laptop using s-tui? You would need to run it as
root so that it can read the power consumption. The refresh interval should be
set to 0.5s so that you can see measurements during pause events.

If possible, please SSH to your laptop from another PC and watch s-tui from the
latter so that you can immediately see what's happening during the pause.

It would also be helpful to monitor twice with s-tui: once with its stress
feature on and once with it off.

If you have enough time to mess around with pause event, could you also check
if it freezes virtual terminals (use the kernel one instead of kmscon)? I guess
it won't.

You may also try to SSH to your laptop from another PC or your smartphone, type
`echo l | sudo tee /proc/sysrq-trigger' in the shell, wait for a pause event,
and press Enter immediately. This should reveal what processes are running and
what the kernel is doing during the event. `echo w | sudo tee
/proc/sysrq-trigger' would also be helpful, as it reveals what processes are
blocked during the pause.

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

You are receiving this mail because:
You are watching the assignee of the bug.