Message-ID: <87v7ko11iw.fsf@osv.gnss.ru>
Date: Thu, 09 Oct 2025 19:29:59 +0300
From: Sergey Organov <sorganov@...il.com>
To: Fabio Estevam <festevam@...il.com>
Cc: linux-kernel@...r.kernel.org,  Russell King <linux@...linux.org.uk>,
  Ulf Hansson <ulf.hansson@...aro.org>,  Shawn Guo <shawnguo@...nel.org>,
  "Rob Herring (Arm)" <robh@...nel.org>,  Angelo Dureghello
 <angelo@...nel-space.org>
Subject: Re: ARM iMX6sx board fails to boot with kernel 6.17

Hi Fabio,

Fabio Estevam <festevam@...il.com> writes:

> Hi Sergey,
>
> On Thu, Oct 9, 2025 at 11:46 AM Sergey Organov <sorganov@...il.com> wrote:
>
>> > I tried booting 6.17 on an imx6sx-sdb board and it booted
>> > successfully:
>>
>> Now after further investigations of the issue I find that success
>> a bit surprising.
>>
>> I've found that my problem is cured by a commit in
>>
>> https://github.com/nxp-imx/linux-imx
>>
>> that works-around an issue that looks like imx6sx generic problem, and
>> should probably cause troubles on imx6sx-sdb board as well, but doesn't?
>> The commit is:
>>
>> commit 65cf3a49ab15cdbbede3c97fd6bba731a3cd366f
>> Author:     Anson Huang <b20788@...escale.com>
>> AuthorDate: Tue Aug 18 15:53:54 2015 +0800
>> Commit:     Octavian Purdila <octavian.purdila@....com>
>> CommitDate: Thu Feb 23 14:21:42 2017 +0200
>>
>>     MLK-11375-2 ARM: imx: disable cpuidle for imx6sx
>>
>>     i.MX6SX CPUIdle with ARM core power down has issue
>>     and cause system hang during kernel boot up, disable
>>     it for now and will enable it when busfreq driver
>>     is ready.
>>
>> Later (by the DAG, not by date) on the same imx_4.9.11_1.0.0_ga
>> branch at that repo, there is:
>>
>> commit 816978f83b1d8dd6ce3bd5dd62223dbfdf74bdd3
>> Author:     Anson Huang <b20788@...escale.com>
>> AuthorDate: Mon Oct 20 16:16:20 2014 +0800
>> Commit:     Octavian Purdila <octavian.purdila@....com>
>> CommitDate: Thu Feb 23 14:21:42 2017 +0200
>>
>>     MLK-9708 arm: imx: add low power idle support for i.mx6sx
>>
>> that effectively reverts the aforementioned commit while adding a lot of
>> other stuff. The resulting kernel also boots fine on my hardware.
>>
>> Curiously enough, neither of these commits are there in the latest
>> branch lf-6.12.y there (and that kernel hangs on my board as well.)
>>
>> Provided the latter commit is indeed the right fix for imx6sx support in
>> the kernel, what's the procedure to get it to the mainstream Linux
>> kernel?
>
> It seems that your observation matches what Angelo (added on Cc)
> reported a long time ago:
>
> https://community.nxp.com/t5/i-MX-Processors/imx6sx-kernel-hangs-at-rootfs-mount/td-p/964920
>
> I'd like to know if the bootloader version plays a role in the
> different behavior that we see on the imx6sx-sdb versus your board.
>
> I used U-Boot 2024.07 on my imx6sx-sdb. Which version did you use? Is
> it based on an old NXP U-Boot tree?
>
> Does using a recent bootloader make the hang disappear?

Uhh... U-Boot is very old and is heavily modified anyway. I figure our
local history starts at:

rel_imx_4.1.33_7ulp_beta
commit 43bff39513ae89c476751caff427b9855af78f7e
Author: Ye Li <ye.li@....com>
Date:   Fri Apr 14 13:44:48 2017 +0800

  MLK-14693 mx7ulp: Change PLL rate calculation to avoid div 0

that looks like NXP tree to me indeed, likely taken once upon a time
from:

https://gitlab.com/jcgit/uboot-imx.git

that was a mirror of original NXP repository 

git://git.freescale.com/imx/uboot-imx.git

That said, I wanted to get to recent kernel using working u-boot, then
maybe try to upgrade u-boot as well, but maybe even not. No reason as of
yet, especially as it's difficult for me to recover from a brick state
should boot-loader fail.

-- Sergey Organov

Powered by blists - more mailing lists