Thread (15 messages) 15 messages, 5 authors, 2021-09-01

Re: [PATCH 0/2] UCSI race condition resulting in wrong port state

From: Salvatore Bonaccorso <hidden>
Date: 2021-08-20 13:01:58
Also in: lkml

Hi Greg,

On Fri, Nov 06, 2020 at 11:47:25AM +0100, Greg Kroah-Hartman wrote:
On Wed, Oct 28, 2020 at 10:10:43AM +0100, Greg Kroah-Hartman wrote:
quoted
On Fri, Oct 09, 2020 at 04:40:45PM +0200, Benjamin Berg wrote:
quoted
From: Benjamin Berg <redacted>

Hi all,

so, I kept running in an issue where the UCSI port information was saying
that power was being delivered (online: 1), while no cable was attached.

The core of the problem is that there are scenarios where UCSI change
notifications are lost. This happens because querying the changes that
happened is done using the GET_CONNECTOR_STATUS command while clearing the
bitfield happens from the separate ACK command. Any change in between will
be lost.

Note that the problem may be almost invisible in the UI as e.g. GNOME will
still show the battery as discharging. But some policies like automatic
suspend may be applied incorrectly.

Cc: Hans de Goede <redacted>
Cc: Heikki Krogerus <heikki.krogerus@linux.intel.com>

Benjamin Berg (2):
  usb: typec: ucsi: acpi: Always decode connector change information
  usb: typec: ucsi: Work around PPM losing change information
Do these need to be backported to stable kernel releases?  If so, how
far back?
Due to the lack of response, I guess they don't need to go to any stable
kernel, so will queue them up for 5.11-rc1.
At least one user in Debian (https://bugs.debian.org/992004) would be
happy to have those backported as well to the 5.10.y series (which we
will pick up).

So if Benjamin ack's this, this would be great to have in 5.10.y.

Regards,
Salvatore
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help