Date:   Sun, 16 Jul 2023 06:35:19 +0200
From:   Alban Browaeys <alban.browaeys@...il.com>
To:     Christopher Obbard <chris.obbard@...labora.com>,
        linux-rockchip@...ts.infradead.org
Cc:     kernel@...labora.com, Akash Gajjar <Akash_Gajjar@...tor.com>,
        Conor Dooley <conor+dt@...nel.org>,
        FUKAUMI Naoki <naoki@...xa.com>,
        Heiko Stuebner <heiko@...ech.de>,
        Jagan Teki <jagan@...rulasolutions.com>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Pragnesh Patel <Pragnesh_Patel@...tor.com>,
        Rob Herring <robh+dt@...nel.org>, devicetree@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        FolkerSchwesinger <dev@...ker-schwesinger.de>
Subject: Re: [PATCH v1 1/2] arm64: dts: rockchip: Disable HS400 for eMMC on
 ROCK Pi 4

Le mercredi 05 juillet 2023 à 15:42 +0100, Christopher Obbard a écrit :
> > > > There is some instablity with some eMMC modules on ROCK Pi 4
> > > > SBCs
> > > > running
> > > > in HS400 mode. This ends up resulting in some block errors
> > > > after a
> > > > while
> > > > or after a "heavy" operation utilising the eMMC (e.g. resizing
> > > > a
> > > > filesystem). An example of these errors is as follows:

I did not report my finding to the Linux upstream back then (due to
using a non vanilla Linux kernel) but with my Armbian install I had
bisected this issue to 06653ebc0ad2e0b7d799cd71a5c2933ed2fb7a66 as the
first bad commit.
I believe it was released in 5.10.60 (the first broken version to reach
armbian was 5.10.63 from a working 5.10.43.
Since then all rk3399 I have checked have disabled hs400 (down to hs200
which is stable even with the above commits).

commit 06653ebc0ad2e0b7d799cd71a5c2933ed2fb7a66
Author: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
Date:   Thu May 20 01:12:23 2021 +0300

    regulator: core: resolve supply for boot-on/always-on regulators
    
    commit 98e48cd9283dbac0e1445ee780889f10b3d1db6a upstream.
    
    For the boot-on/always-on regulators the set_machine_constrainst() is
    called before resolving rdev->supply. Thus the code would try to enable
    rdev before enabling supplying regulator. Enforce resolving supply
    regulator before enabling rdev.
    
    Fixes: aea6cb99703e ("regulator: resolve supply after creating regulator")
    Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
    Link: https://lore.kernel.org/r/20210519221224.2868496-1-dmitry.baryshkov@linaro.org
    Signed-off-by: Mark Brown <broonie@...nel.org>
    Signed-off-by: Greg Kroah-Hartman <gregkh@...uxfoundation.org>

 drivers/regulator/core.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/regulator/core.c b/drivers/regulator/core.c
index f192bf19492ed..e20e77e4c159d 100644
--- a/drivers/regulator/core.c
+++ b/drivers/regulator/core.c
@@ -1425,6 +1425,12 @@ static int set_machine_constraints(struct regulator_dev *rdev)
 	 * and we have control then make sure it is enabled.
 	 */
 	if (rdev->constraints->always_on || rdev->constraints->boot_on) {
+		/* If we want to enable this regulator, make sure that we know
+		 * the supplying regulator.
+		 */
+		if (rdev->supply_name && !rdev->supply)
+			return -EPROBE_DEFER;
+
 		if (rdev->supply) {
 			ret = regulator_enable(rdev->supply);
 			if (ret < 0) {


My findings here:
https://forum.armbian.com/topic/18855-upgrading-to-bullseye-troubleshooting-armbian-21081/?do=findComment&comment=128793
this on a kobol helios64 rk3399 board.

I told a user to try this fix (revert commits 06653ebc0ad2e0b7d799cd71a5c2933ed2fb7a66
 and aea6cb99703e17019e025aa71643b4d3e0a24413) also for an armbian kernel on a  Nanopc-T4
 and it fixes the issue https://forum.armbian.com/topic/20002-nanopc-t4-new-kernel-2202-generates-issues-on-mmc2-and-makes-system-not-properly-working/?do=findComment&comment=138052
This above 5.16.8.





I had high expectations that the commit that fixed double init would fix the issue for good, but sadly not.
I believe this would have been the only required fix for 5.16 kernels but nowadays it is not enough a revert.

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/regulator/core.c?id=8a866d527ac0441c0eb14a991fa11358b476b11d

regulator: core: Resolve supply name earlier to prevent double-init
Previously, an unresolved regulator supply reference upon calling
regulator_register on an always-on or boot-on regulator caused
set_machine_constraints to be called twice.

This in turn may initialize the regulator twice, leading to voltage
glitches that are timing-dependent. A simple, unrelated configuration
change may be enough to hide this problem, only to be surfaced by
chance.

One such example is the SD-Card voltage regulator in a NanoPI R4S that
would not initialize reliably unless the registration flow was just
complex enough to allow the regulator to properly reset between calls.

Fix this by re-arranging regulator_register, trying resolve the
regulator's supply early enough that set_machine_constraints does not
need to be called twice.

Signed-off-by: Christian Kohlschütter <christian@...lschutter.com>
Link: https://lore.kernel.org/r/20220818124646.6005-1-christian@kohlschutter.com
Signed-off-by: Mark Brown <broonie@...nel.org>
"
story behing this patch https://kohlschuetter.github.io/blog/posts/2022/10/28/linux-nanopi-r4s/

It should have worked because basically this patch is a revert of
commit aea6cb99703e17019e025aa71643b4d3e0a24413 "regulator: resolve
supply after creating regulator" except it keep what I believe is now
dead code (ie the second set_machine_constains in "if (ret == -
EPROBE_DEFER) " is of no use now that the regulator supply is resolved
before the first set_machine_constraints call in regilator_registers.
The only code left from the 5.10.60 breakage is the EPROBE_DEFER if
regulator supply is not registered in set_machine_constrains.
But even after removing this leftover and the new EPROBE_DEFER that was
added to set_machine_constraints for "regulator that have no direct
control", I cannot get rid of the Filesystem corruption and errors with
hs400 with 6.3.

Still I have no clue why emmc regulators double init is fine on most
SoC but not rk3399.


Cheers,
Alban

Powered by blists - more mailing lists