Re: [PATCH v3] usb: typec: ucsi: do not register UCSI on Lenovo Legion Pro 7 16IAX10H

From: Alex

Date: Thu Nov 19 2026 - 16:24:45 EST


Hi Wei, Heikki, Greg,

I need to correct my earlier report and withdraw my Tested-by on this
patch. Please don't apply v3 as it stands.

The commit message says that with ucsi_acpi blacklisted the EC
negotiates PD on its own and charging works normally. Longer testing
shows that isn't true. With ucsi_acpi and typec_ucsi blacklisted and
confirmed not loaded, USB-C charging still drops:

- 2026-11-18: ADP0 went offline about 7.5 minutes after a cold boot,
with no suspend and no replug.
- 2026-11-19: ADP0 went offline at about 588 s uptime and came back
by itself about 5 minutes later.

When it drops, the EC itself stops seeing the adapter. ADPT (bit 7
of EC register 0xA3, which ADP0._PSR reads) clears. Unplugging and
replugging, or swapping to a different charger, often doesn't bring
it back. Swapping from my main charger (a Shargeek, 100 W or more)
to a 60 W PD charger while it was dropped changed nothing in the EC.
Sometimes it recovers by itself, and a power cycle always does. The
barrel jack is unaffected.

My August test only showed that charging stayed up across a module
reload for a few minutes. That was too short to show the blacklist
fixes anything, and I shouldn't have offered a Tested-by on that
basis. I'm sorry for the noise.

What I believe now:

- The drops happen with and without ucsi_acpi bound, so the
driver is not the cause. Skipping UCSI on this machine wouldn't
fix the problem, and it would hide the typec/PD information.
With ucsi_acpi bound on Q7CN78WW it probes cleanly, and both
ports and pd0/pd1 (PD rev 3.0) appear. Drops happen in both
configurations. I don't have enough data yet to say whether
binding UCSI changes how often they happen.
- Windows on the same machine, using the stock Microsoft UCSI
driver (UcmUcsiAcpiClient, no Lenovo filter driver), charges
reliably over USB-C. So whatever goes wrong probably depends on
something Linux does, or doesn't do, with the EC or ACPI
firmware. I don't know what that is yet.
- The PPM reports UCSI version 0x0120. While charging is broken,
CCI = 0x08000000 and CONTROL = 0, so the PPM looks idle, not
stuck.
- At the moment of the 11-19 drop, with UCSI not loaded, dmesg
shows a burst of "workqueue: acpi_os_execute_deferred hogged CPU
for >10000us" warnings. Those come from ACPI/EC event handling,
not from UCSI. I mention them only because they line up with the
drop.
- The one remaining UCSI-side issue I can see is cosmetic:
ucsi-source-psy reports voltage_now = 0 while charging. That is
unrelated to the drops.

System details:

Machine: Lenovo Legion Pro 7 16IAX10H, DMI product 83F5
CPU: Intel Core Ultra 9 275HX (Arrow Lake-HX)
BIOS: Q7CN78WW (2026-04-07), EC firmware revision 1.78.
Seen earlier on Q7CN44WW too. Lenovo's newer Q7CN83WW
only changes the lid logo LED, according to its readme.
Distro: Ubuntu 26.04.1 LTS
Kernel: 7.0.0-43-generic (Ubuntu 7.0.0-43.43, newest in the
archive); also seen on 7.0.0-42 and a 6.18 build. Not tested on
mainline yet.
Cmdline: quiet splash snd_intel_dspcfg.dsp_driver=3
mem_sleep_default=s2idle
Taint: 4160 (O + U). Out-of-tree modules: nvidia-open
595.99.02 and acpi_call 1.2.2 (DKMS), the latter used
for the EC/PPM reads above.
Firmware: linux-firmware 20260319.git217ca6e4
Userspace power: power-profiles-daemon 0.30, thermald 2.5.11,
upower 1.91.1, tlp 1.8.0 (installed, not running).
ideapad conservation_mode = 0.
Other platform drivers loaded: ideapad_laptop, lenovo_wmi_*
(gamezone, other, events, capdata), int3400_thermal,
processor_thermal_*, nvidia_wmi_ec_backlight.
I haven't yet ruled these out by blacklisting them.

I'll keep working on it and will follow up on bugzilla #221142 with
data rather than conclusions. I can test on a current mainline
kernel, and capture anything else that would help (traces, ACPI/EC
dumps, specific test sequences).

Thanks for your time on this, and again, sorry for steering you
wrong.

Alex