Re: Fwd: Logitech G915 Wireless Keyboard acts weird on 6.6.0

From: Mavroudis Chatzilazaridis <mavchatz@protonmail.com>
To: Jiri Kosina <jikos@kernel.org>, Bagas Sanjaya <bagasdotme@gmail.com>
Cc: "Linux Kernel Mailing List" <linux-kernel@vger.kernel.org>,
	"Linux Regressions" <regressions@lists.linux.dev>,
	"Linux Input Devices" <linux-input@vger.kernel.org>,
	"Thorsten Leemhuis" <regressions@leemhuis.info>,
	"Filipe Laíns" <lains@riseup.net>,
	"Bastien Nocera" <hadess@hadess.net>,
	LinuxCat <masch77.linuxcat@gmail.com>,
	Marcelo <mmbossoni@gmail.com>, "Takashi Iwai" <tiwai@suse.de>,
	"Hans de Goede" <hdegoede@redhat.com>,
	"Linus Torvalds" <torvalds@linux-foundation.org>
Subject: Re: Fwd: Logitech G915 Wireless Keyboard acts weird on 6.6.0
Date: Tue, 21 Nov 2023 14:15:31 +0000	[thread overview]
Message-ID: <4071999d-77ca-4a75-a9e0-d8ccac0dada1@protonmail.com> (raw)
In-Reply-To: <nycvar.YFH.7.76.2311211435050.29220@cbobk.fhfr.pm>

On 2023-11-21 15:37, Jiri Kosina wrote:
> On Tue, 21 Nov 2023, Bagas Sanjaya wrote:
> 
>> Hi Thorsten and all,
>>
>> On Thu, Nov 02, 2023 at 09:11:42PM +0700, Bagas Sanjaya wrote:
>>> Hi,
>>>
>>> I notice a regression report on Bugzilla [1]. Quoting from it:
>>>
>>>> Hello,
>>>> After upgrading from 6.5.9 to 6.6.0, my keyboard started acting really weird in its wireless mode, key actions sent are completely wrong, see video attached.
>>>>
>>>> Most keys are perceived as either E, 3 or F7, with F8 and <, as well.
>>>>
>>>> Modifier keys (CTRL, ALT, ALT GR, Shift and Super) are working normally, as well as media control keys (pause/play, previous, next, mute and sound up/down).
>>>>
>>>> The keyboard works as expected if it's wired.
>>>
>>> Another reporter bisected the regression:
>>>
>>>> Bisected to
>>>>
>>>> 9d1bd9346241cd6963b58da7ffb7ed303285f684 is the first bad commit
>>>> commit 9d1bd9346241cd6963b58da7ffb7ed303285f684
>>>> Author: Mavroudis Chatzilazaridis <mavchatz@protonmail.com>
>>>> Date: Sun Jul 16 18:23:44 2023 +0000
>>>>
>>>> HID: logitech-dj: Add support for a new lightspeed receiver iteration
>>>>
>>>> The lightspeed receiver for the Pro X Superlight uses 13 byte mouse reports
>>>> without a report id. The workaround for such cases has been adjusted to
>>>> handle these larger packets.
>>>>
>>>> The device now reports the status of its battery in wireless mode and
>>>> libratbag now recognizes the device and it can be configured with Piper.
>>>>
>>>> https://github.com/libratbag/libratbag/pull/1122
>>>>
>>>> Co-developed-by: Filipe Laíns <lains@riseup.net>
>>>> Signed-off-by: Filipe Laíns <lains@riseup.net>
>>>> Signed-off-by: Mavroudis Chatzilazaridis <mavchatz@protonmail.com>
>>>> Reviewed-by: Bastien Nocera <hadess@hadess.net>
>>>> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
>>>>
>>>> drivers/hid/hid-ids.h | 1 +
>>>> drivers/hid/hid-logitech-dj.c | 11 ++++++++---
>>>> 2 files changed, 9 insertions(+), 3 deletions(-)
>>>
>>> See Bugzilla for the full thread.
>>>
>>> Anyway, I'm adding this regression to regzbot:
>>>
>>> #regzbot introduced: 9d1bd9346241cd https://bugzilla.kernel.org/show_bug.cgi?id=218094
>>> #regzbot title: Logitech G915 Wireless Keyboard key event only detects few key codes
>>> #regzbot link: https://streamable.com/ac6l8u
>>>
>>
>> There's no reply from culprit author nor linux-input people (did they miss
>> this regression?). And on Bugzilla, other reporters replied that reverting
>> the culprit fixed the regression.
>>
>> FYI, there's similar Bugzilla report on [1].
> 
> As there was no reaction from Mavroudis in order to figure out why he is
> not observing the issues the other reporters do and what to do to fix
> those, I already do have revert in my queue for -rc3.
> 
> My first guess would be that the extra buttons in the extended report are
> not properly reflected in the emulated report descriptor, but that
> wouldn't explain why it worked for the author of the commit.
> 
> So revert it is, and once Marvoudis resurfaces, we can try again for some
> of later releases.
> 
> --
> Jiri Kosina
> SUSE Labs
> 

Apologies for not requesting a revert earlier, I was away on vacation 
and didn't realize it would take me this long to get to it.

As I mentioned in this bugzilla report [0], I think the extra buttons 
are handled by vendor defined bytes at the end of the descriptor, so it 
boils down to the emulated descriptor not being updated.
This patch worked for me as the Superlight X that I own only has two 
side buttons. In addition, I was unaware of how many devices this would 
affect.

For now let's make sure this commit is reverted so that people have 
working devices and I will get to fixing the issues that showed up.

Once again, apologies for this.

[0] https://bugzilla.kernel.org/show_bug.cgi?id=218172#c9