Message-ID: <e3814c81-c74d-7087-e87d-12dcb49e6444@samsung.com>
Date:   Mon, 20 Feb 2023 10:01:59 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:102.0)
        Gecko/20100101 Thunderbird/102.7.2
Subject: Re: [RFC v1 0/4] Simplify regulator supply resolution code by
 offloading to driver core
Content-Language: en-US
To:     Saravana Kannan <saravanak@google.com>,
        Liam Girdwood <lgirdwood@gmail.com>,
        Mark Brown <broonie@kernel.org>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Geert Uytterhoeven <geert@linux-m68k.org>,
        Bjorn Andersson <andersson@kernel.org>,
        Sudeep Holla <sudeep.holla@arm.com>,
        Tony Lindgren <tony@atomide.com>,
        Doug Anderson <dianders@chromium.org>,
        Guenter Roeck <linux@roeck-us.net>,
        Luca Weiss <luca.weiss@fairphone.com>, kernel-team@android.com,
        linux-kernel@vger.kernel.org
From:   Marek Szyprowski <m.szyprowski@samsung.com>
In-Reply-To: <20230218083252.2044423-1-saravanak@google.com>
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset="utf-8"
CMS-TYPE: 201P
References: <CGME20230218083300eucas1p28c7c584877b8914a3b88904690be82f6@eucas1p2.samsung.com>
        <20230218083252.2044423-1-saravanak@google.com>
Precedence: bulk

Hi Saravana,

On 18.02.2023 09:32, Saravana Kannan wrote:
> Hi Mark/Liam,
>
> This series is just an RFC to see if you agree with where this is going.
> Please point out bugs, but don't bother with a proper code review.
>
> The high level idea is to not reimplement what driver core can already
> handle for us and use it to do some of the work. Instead of trying to
> resolve supplies from all different code paths and bits and pieces of
> the tree, we just build it from the root to the leaves by using deferred
> probing to sequence things in the right order.
>
> The last patch is the main one. Rest of them are just setting up for it.
>
> I believe there's room for further simplification but this is what I
> could whip up as a quick first draft that shows the high level idea.
> I'll probably need some help with getting a better understanding of why
> things are done in a specific order in regulator_register() before I
> could attempt simplifying things further.
>
> Ideally, regulator_register() would just have DT parsing, init data
> struct sanity checks and adding the regulator device and then we move
> everything else to into the probe function that's guaranteed to run only
> after the supply has been resolved/ready to resolve.
>
> fw_devlink/device links should further optimize the flow and also allow
> us to simplify some of the guarantees and address some of the existing
> FIXMEs. But this patch series is NOT dependent on fw_devlink or device
> links.
>
> Any thoughts on where this is going?
>
> I've tested this on one hardware I have and it works and nothing is
> broken. But the regulator tree in my hardware isn't that complicated or
> deep. The regulators are also added mostly in the right order (due to
> existing fw_devlink). So if you agree with the idea, the next step is to
> ask people to give it a test.
>
> Also, it's based on driver-core-next since that's what I had synced up
> and had a working baseline. I'll rebase it on the regulator tree when I
> go from RFC -> PATCH.

I've applied this patchset on top of linux next-20230220 and gave it a 
try on my test farm, as it revealed a few issues related to regulator 
initialization in the past. It looks that handling of some corner cases 
is missing, because this patchset introduced a regression on Samsung 
Snow/Peach-Pit/Peach-Pi Chromebooks, as well as Hardkernel's Odroid-M1 
board. It looks that the issue is common - PHY devices don't probe 
properly. This is an output from Odroid-M1 board 
(arch/arm64/boot/dts/rockchip/rk3568-odroid-m1.dts):

# cat /sys/kernel/debug/devices_deferred 2>/dev/null
fd8c0000.usb    platform: wait for supplier host-port
fe830000.phy
fe8a0000.usb2phy        rockchip-usb2phy: failed to create phy
fe8b0000.usb2phy        rockchip-usb2phy: failed to create phy
3c0800000.pcie  rockchip-dw-pcie: failed to get vpcie3v3 regulator
fcc00000.usb    platform: wait for supplier otg-port
fd000000.usb    platform: wait for supplier host-port
fd800000.usb    platform: wait for supplier otg-port
fd840000.usb    platform: wait for supplier otg-port
fd880000.usb    platform: wait for supplier host-port
fe820000.phy

If you need any additional tests on the mentioned boards, let me know.


> Thanks,
> Saravana
>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Geert Uytterhoeven <geert@linux-m68k.org>
> Cc: Marek Szyprowski <m.szyprowski@samsung.com>
> Cc: Bjorn Andersson <andersson@kernel.org>
> Cc: Sudeep Holla <sudeep.holla@arm.com>
> Cc: Tony Lindgren <tony@atomide.com>
> Cc: Doug Anderson <dianders@chromium.org>
> Cc: Guenter Roeck <linux@roeck-us.net>
> Cc: Luca Weiss <luca.weiss@fairphone.com>
>
> Saravana Kannan (4):
>    regulator: core: Add regulator devices to bus instead of class
>    regulator: core: Add sysfs class backward compatibility
>    regulator: core: Probe regulator devices
>    regulator: core: Move regulator supply resolving to the probe function
>
>   drivers/regulator/core.c         | 102 +++++++++++++++++++------------
>   drivers/regulator/internal.h     |   2 +-
>   drivers/regulator/of_regulator.c |   2 +-
>   3 files changed, 64 insertions(+), 42 deletions(-)
>
Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland