Hi Markus,
On 2019-01-24 15:27, Markus Reichl wrote:
> Am 24.01.19 um 13:11 schrieb Marek Szyprowski:
>> On 2019-01-24 11:40, Krzysztof Kozlowski wrote:
>>> On Thu, 24 Jan 2019 at 10:10, Markus Reichl <m.reichl@xxxxxxxxxxxxx> wrote:
>>>> Am 23.01.19 um 19:23 schrieb Markus Reichl:
>>>>> Am 23.01.19 um 16:59 schrieb Ulf Hansson:
>>>>>> On Wed, 23 Jan 2019 at 16:37, Markus Reichl <m.reichl@xxxxxxxxxxxxx> wrote:
>>>>>>> Am 23.01.19 um 16:08 schrieb Ulf Hansson:
>>>>>>>> On Wed, 23 Jan 2019 at 14:30, Markus Reichl <m.reichl@xxxxxxxxxxxxx> wrote:
>>>>>>>>> Hi,
>>>>>>>>>
>>>>>>>>> with all recent kernel versions Reboot hangs on both my Odroid-X2 boards.
>>>>>>>>> Boot and root partition reside on eMMC.
>>>>>>>>> Last lines on ttySAC1:
>>>>>>>>> [ 39.423093] systemd-shutdown[1]: Rebooting.
>>>>>>>>> [ ▒
>>>>>>>>>
>>>>>>>> Thanks for reporting!
>>>>>>>>
>>>>>>>>> Bisecting points to:
>>>>>>>>>
>>>>>>>>> d97a1e5d7cd2b5b0edc02a40fe6897b710c9e10f is the first bad commit
>>>>>>>>> commit d97a1e5d7cd2b5b0edc02a40fe6897b710c9e10f
>>>>>>>>> Author: Srinivas Kandagatla <srinivas.kandagatla@xxxxxxxxxx>
>>>>>>>>> Date: Thu Apr 14 14:02:16 2016 +0100
>>>>>>>>>
>>>>>>>>> mmc: pwrseq: convert to proper platform device
>>>>>>>> Hmm, looking at history from git log gives me a hint about that it's
>>>>>>>> the dts that may have the wrong pinctrl for the pwrseq node used for
>>>>>>>> eMMC.
>>>>>>>>
>>>>>>>> commit 225da7e65a03 ("ARM: dts: add eMMC reset line for
>>>>>>>> exynos4412-odroid-common") was introduced in v4.1, which adds the
>>>>>>>> pwrseq node containing a pinctrl. However, at that point pinctrl was
>>>>>>>> not a supported DT binding, thus not used by the mmc core.
>>>>>>>>
>>>>>>>> It's not until the commit you point to above, for when we start
>>>>>>>> parsing for pinctrl, via driver core.
>>>>>>>>
>>>>>>>> So as a test, could you try the patch below, to see if that helps.
>>>>>>> Thanks Uffe,
>>>>>>>
>>>>>>> this fixes reboot.
>>>>>>> Tested with kernel v4.20.2.
>>>>>>> Tested-by: Markus Reichl <m.reichl@xxxxxxxxxxxxx>
>>>>>> Thanks Markus!
>>>>>>
>>>>>> However, it seems like we need some Exynos guys to confirm that this
>>>>>> is an acceptable solution. Or if there is something else we need to
>>>>>> do. I have looped in Marek, to what his opinion is.
>>>>> I did further tests:
>>>>> With an additional SD-Card, even only partly pushed in and therefore
>>>>> SD-detect pin operated reboot still hangs.
>>>>> Without SD-Card reboot works fine as tested above.
>>>> I'm using exynos_defconfig for testing but devfreq must be disabled in kernel config otherwise reboot still doesn't work most of the time.
>>>> Now got it rebooting even with an SD-Card plugged in.
>>> Unfortunately I cannot reproduce it as I use SD card for my Odroid U3
>>> board. Make Marek can also take a look. Anyway, I will need some
>>> rationale behind the patch. :)
>> Huh, it took a while to find the source of the problem. I didn't notice
>> it for so many releases, because I always thought that reboot is broken
>> because of the devfreq issue (which finally was fixed in v5.0-rc3). I
>> will send a proper fix in a few minutes.
> This fixes reboot with devfreq disabled.
> With devfreq enabled in v5.0-rc3 reboot still hangs.
Well, you are right. I thought that devfreq suspend related patches
resurrected recently fixed that issue too, but I was wrong. v5.0-rc3
requires one more patch to fix reboot issue, see:
https://patchwork.kernel.org/patch/10781431/
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland
