MIME-Version: 1.0
References: <20211205215846.153703-1-aouledameur@baylibre.com> <20211205215846.153703-2-aouledameur@baylibre.com>
In-Reply-To: <20211205215846.153703-2-aouledameur@baylibre.com>
From:   Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Date:   Mon, 6 Dec 2021 22:19:43 +0100
Message-ID: <CAFBinCDZ2nTg2V_6qxok_y+wXHTVJ6XyyvON59TLJZ9B3fKY3Q@mail.gmail.com>
Subject: Re: [PATCH v4 1/3] phy: amlogic: phy-meson-gxl-usb2: fix shared reset
 controller use
To:     Amjad Ouled-Ameur <aouledameur@baylibre.com>,
        p.zabel@pengutronix.de
Cc:     khilman@baylibre.com, balbi@kernel.org, jbrunet@baylibre.com,
        linux-amlogic@lists.infradead.org, linux-kernel@vger.kernel.org,
        linux-usb@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Amjad,

On Sun, Dec 5, 2021 at 10:59 PM Amjad Ouled-Ameur
<aouledameur@baylibre.com> wrote:
>
> Use reset_control_rearm() call if an error occurs in case
> phy_meson_gxl_usb2_init() fails after reset() has been called ; or in case
> phy_meson_gxl_usb2_exit() is called i.e the resource is no longer used
> and the reset line may be triggered again by other devices.
>
> reset_control_rearm() keeps use of triggered_count sane in the reset
> framework. Therefore, use of reset_control_reset() on shared reset line
> should be balanced with reset_control_rearm().
>
> Signed-off-by: Amjad Ouled-Ameur <aouledameur@baylibre.com>
> Reported-by: Jerome Brunet <jbrunet@baylibre.com>
> ---
> changes since v3:
> - Remove unnecessary reset_control_rearm() after reset_control_reset()
> failure.
I double-checked your patch in v3 and Philipp was right:
reset_control_rearm() should not be right after reset_control_reset().
However, I think reset_control_rearm() is still needed
phy_meson_gxl_usb2_init() whenever clk_prepare_enable() fails.

So my suggestion is to add reset_control_rearm() in
phy_meson_gxl_usb2_init() if clk_prepare_enable() fails so we are
resetting the ref-count for the reset line (just like
phy_meson_gxl_usb2_exit() does).
The difference compared to the previous version is that the
reset_control_rearm() call needs to be placed a few lines down.


Thank you!
Martin