Re: kernel error at led trigger "phy0tpt"

From: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
To: Bagas Sanjaya <bagasdotme@gmail.com>,
	Tobias Dahms <dahms.tobias@web.de>,
	Sean Wang <sean.wang@mediatek.com>
Cc: stable@vger.kernel.org, Pavel Machek <pavel@ucw.cz>,
	Lee Jones <lee@kernel.org>, LKML <linux-kernel@vger.kernel.org>,
	linux-leds@vger.kernel.org, linux-wireless@vger.kernel.org,
	Linux regressions mailing list <regressions@lists.linux.dev>
Subject: Re: kernel error at led trigger "phy0tpt"
Date: Mon, 27 Mar 2023 10:23:39 +0200	[thread overview]
Message-ID: <8b07ead5-f105-da86-e7da-ee49616f7c1d@collabora.com> (raw)
In-Reply-To: <d14fb08c-70e3-4cc7-caf9-87e73eab9194@gmail.com>

Il 26/03/23 15:23, Bagas Sanjaya ha scritto:
> On 3/26/23 02:20, Tobias Dahms wrote:
>> Hello,
>>
>> the bisection gives following result:
>> --------------------------------------------------------------------
>> 18c7deca2b812537aa4d928900e208710f1300aa is the first bad commit
>> commit 18c7deca2b812537aa4d928900e208710f1300aa
>> Author: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
>> Date:   Tue May 17 12:47:08 2022 +0200
>>
>>      soc: mediatek: pwrap: Use readx_poll_timeout() instead of custom
>> function
>>
>>      Function pwrap_wait_for_state() is a function that polls an address
>>      through a helper function, but this is the very same operation that
>>      the readx_poll_timeout macro means to do.
>>      Convert all instances of calling pwrap_wait_for_state() to instead
>>      use the read_poll_timeout macro.
>>
>>      Signed-off-by: AngeloGioacchino Del Regno
>> <angelogioacchino.delregno@collabora.com>
>>      Reviewed-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>
>>      Tested-by: Nícolas F. R. A. Prado <nfraprado@collabora.com>
>>      Link:
>> https://lore.kernel.org/r/20220517104712.24579-2-angelogioacchino.delregno@collabora.com
>>      Signed-off-by: Matthias Brugger <matthias.bgg@gmail.com>
>>
>>   drivers/soc/mediatek/mtk-pmic-wrap.c | 60
>> ++++++++++++++++++++----------------
>>   1 file changed, 33 insertions(+), 27 deletions(-)
>> --------------------------------------------------------------------
>>
> 
> OK, I'm updating the regression status:
> 
> #regzbot introduced: 18c7deca2b8125
> 
> And for replying, don't top-post, but rather reply inline with
> appropriate context instead; hence I cut the replied context.
> 

There are two possible solutions to that, specifically, either:
  1. Change readx_poll_timeout() to readx_poll_timeout_atomic(); or
  2. Fix the mt6323-led driver so that this operation gets done
     out of atomic context, which is IMO the option to prefer.

Ideas?

Regards,
Angelo