SM8750: fix suspend wedge from ucsi USB role on charger unplug#3039
SM8750: fix suspend wedge from ucsi USB role on charger unplug#3039aanze wants to merge 1 commit into
Conversation
|
0509 is confirmed (manual inspection), 0510 is debatable (and has potential bugs such as use-after-free). I'd split the patches into different commits. |
|
Sure, after all it's a matter of choice, but without 0510, device will absolutely wedge if suspended then uplugged from any usb data capable peripheral |
|
I don't have an issue with 0510's intentions, only with the current implementation, which seems incomplete. Namely:
I understand "experiment" is there meaning "trial", but it's an interesting choice of words regardless... Knowing the pattern-matchey nature of transformers I couldn't help it but wonder about a possible "subtext", and no big deal at all, as I see it Fable gave it a measure of effort sufficient for ruling out some things while testing other hypotheses, and now it needs the correction and robustness pass. |
|
Thanks for the input, will work on a cleaner 0510 |
Add a SM8750 kernel patch reverting the u_role gating introduced by
0c8ee850572 ("usb: typec: ucsi: Add UCSI_USB4_IMPLIES_USB quirk for
X1E80100"). Without that quirk the USB data role is never cleared for
a charger-only partner and stays attached across suspend, keeping the
USB power island up; the firmware power-collapse then hard-hangs the
SoC when the cable is unplugged while suspended. Device-tested on the
AYN Odin 3.
032fba4 to
f0703a0
Compare
|
Done — split as suggested: this PR is now 0509 only. 0510 is fully reworked in #3049: per-instance state in struct ucsi, synchronous unregister in ucsi_unregister() (the PDR/SSR path), role writes under con->lock + a suspend flag closing the workqueue race, and no blind restore — the role is re-derived from a fresh GET_CONNECTOR_STATUS at resume. Re-tested on device, including dock kept attached across suspend |
2 participants
Summary
(AYN Odin 3) when the USB cable is unplugged while the device is
suspended — both for chargers (patch 0509) and for data-capable
partners such as docks (patch 0510).
Testing
and flashed it to an AYN Odin 3. Reproduced the original hang on
stock 7.1.x, then ran the patched kernel through several
suspend/unplug cycles with a charger and with a dock.
unplug while asleep hard-hangs the SoC (dead screen, runs hot,
forced reset needed) on every 7.1.x; 7.0.x is fine. After — suspends
and resumes normally with the cable or dock unplugged during
suspend, over several cycles each.
Additional Context
UCSI_USB4_IMPLIES_USB quirk for X1E80100") gated the "partner has no
USB data -> USB_ROLE_NONE" decision behind the UCSI_USB4_IMPLIES_USB
quirk. SM8750's pmic_glink UCSI never sets that quirk, so the USB
role is no longer cleared for a charger-only partner and stays
attached across suspend, keeping the USB power island up; the
firmware power-collapse then hangs on the cable event.
Patch 0510 extends the same fix to data-capable partners: the role
is forced to USB_ROLE_NONE at suspend entry and restored at resume,
so the power island can collapse with a dock attached too.
AI Usage
Did you use AI tools to help write this code? YES