mbox series

[v2,0/4] arm64: rockchip: Fix PCIe ep-gpios requirement and Add Nanopi M4B

Message ID 20201118071724.4866-1-wens@kernel.org
Headers show
Series arm64: rockchip: Fix PCIe ep-gpios requirement and Add Nanopi M4B | expand

Message

Chen-Yu Tsai Nov. 18, 2020, 7:17 a.m. UTC 
From: Chen-Yu Tsai <wens@csie.org>

Hi everyone,

This is v2 of my Nanopi M4B series. Changes since v1 include:

  - Rewrite subject of patch 1 to match existing convention and reference
    'ep-gpios' DT property instead of the 'ep_gpio' field
 
This series mainly adds support for the new Nanopi M4B, which is a newer
variant of the Nanopi M4.

The differences against the original Nanopi M4 that are common with the
other M4V2 revision include:

  - microphone header removed
  - power button added
  - recovery button added

Additional changes specific to the M4B:

  - USB 3.0 hub removed; board now has 2x USB 3.0 type-A ports and 2x
    USB 2.0 ports
  - ADB toggle switch added; this changes the top USB 3.0 host port to
    a peripheral port
  - Type-C port no longer supports data or PD
  - WiFi/Bluetooth combo chip switched to AP6256, which supports BT 5.0
    but only 1T1R (down from 2T2R) for WiFi

While working on this, I found that for the M4 family, the PCIe reset
pin (from the M.2 expansion board) was not wired to the SoC. Only the
NanoPC T4 has this wired. This ended up in patches 1 and 3.

Patch 1 makes ep_gpio in the Rockchip PCIe driver optional. This property
is optional in the DT binding, so this just makes the driver adhere to
the binding.

Patch 2 adds a new compatible string for the new board.

Patch 3 moves the ep-gpios property of the pcie controller from the
common nanopi4.dtsi file to the nanopc-t4.dts file.

Patch 4 adds a new device tree file for the new board. It includes the
original device tree for the M4, and then lists the differences.

Given that patch 3 would make PCIe unusable without patch 1, I suggest
merging patch 1 through the PCI tree as a fix for 5.10, and the rest
for 5.11 through the Rockchip tree.

Please have a look. The changes are mostly trivial.


Regards
ChenYu

Chen-Yu Tsai (4):
  PCI: rockchip: Make 'ep-gpios' DT property optional
  dt-bindings: arm: rockchip: Add FriendlyARM NanoPi M4B
  arm64: dts: rockchip: nanopi4: Move ep-gpios property to nanopc-t4
  arm64: dts: rockchip: rk3399: Add NanoPi M4B

 .../devicetree/bindings/arm/rockchip.yaml     |  1 +
 arch/arm64/boot/dts/rockchip/Makefile         |  1 +
 .../boot/dts/rockchip/rk3399-nanopc-t4.dts    |  1 +
 .../boot/dts/rockchip/rk3399-nanopi-m4b.dts   | 52 +++++++++++++++++++
 .../boot/dts/rockchip/rk3399-nanopi4.dtsi     |  1 -
 drivers/pci/controller/pcie-rockchip.c        |  2 +-
 6 files changed, 56 insertions(+), 2 deletions(-)
 create mode 100644 arch/arm64/boot/dts/rockchip/rk3399-nanopi-m4b.dts

Comments

Heiko Stuebner Nov. 18, 2020, 8:49 a.m. UTC | #1 
Am Mittwoch, 18. November 2020, 08:17:21 CET schrieb Chen-Yu Tsai:
> From: Chen-Yu Tsai <wens@csie.org>
> 
> The Rockchip PCIe controller DT binding clearly states that 'ep-gpios' is
> an optional property. And indeed there are boards that don't require it.
> 
> Make the driver follow the binding by using devm_gpiod_get_optional()
> instead of devm_gpiod_get().
> 
> Fixes: e77f847df54c ("PCI: rockchip: Add Rockchip PCIe controller support")
> Fixes: 956cd99b35a8 ("PCI: rockchip: Separate common code from RC driver")
> Fixes: 964bac9455be ("PCI: rockchip: Split out rockchip_pcie_parse_dt() to parse DT")
> Signed-off-by: Chen-Yu Tsai <wens@csie.org>

Reviewed-by: Heiko Stuebner <heiko@sntech.de


I'll pick up patches 2-4 separately, after giving Rob a chance to look at
the simple binding.


Heiko

> ---
> Changes since v1:
> 
>   - Rewrite subject to match existing convention and reference
>     'ep-gpios' DT property instead of the 'ep_gpio' field
> ---
>  drivers/pci/controller/pcie-rockchip.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/pci/controller/pcie-rockchip.c b/drivers/pci/controller/pcie-rockchip.c
> index 904dec0d3a88..c95950e9004f 100644
> --- a/drivers/pci/controller/pcie-rockchip.c
> +++ b/drivers/pci/controller/pcie-rockchip.c
> @@ -118,7 +118,7 @@ int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
>  	}
>  
>  	if (rockchip->is_rc) {
> -		rockchip->ep_gpio = devm_gpiod_get(dev, "ep", GPIOD_OUT_HIGH);
> +		rockchip->ep_gpio = devm_gpiod_get_optional(dev, "ep", GPIOD_OUT_HIGH);
>  		if (IS_ERR(rockchip->ep_gpio)) {
>  			dev_err(dev, "missing ep-gpios property in node\n");
>  			return PTR_ERR(rockchip->ep_gpio);
>
Chen-Yu Tsai Dec. 7, 2020, 3:30 a.m. UTC | #2 
Ping

On Wed, Nov 18, 2020 at 4:49 PM Heiko Stübner <heiko@sntech.de> wrote:
>
> Am Mittwoch, 18. November 2020, 08:17:21 CET schrieb Chen-Yu Tsai:
> > From: Chen-Yu Tsai <wens@csie.org>
> >
> > The Rockchip PCIe controller DT binding clearly states that 'ep-gpios' is
> > an optional property. And indeed there are boards that don't require it.
> >
> > Make the driver follow the binding by using devm_gpiod_get_optional()
> > instead of devm_gpiod_get().
> >
> > Fixes: e77f847df54c ("PCI: rockchip: Add Rockchip PCIe controller support")
> > Fixes: 956cd99b35a8 ("PCI: rockchip: Separate common code from RC driver")
> > Fixes: 964bac9455be ("PCI: rockchip: Split out rockchip_pcie_parse_dt() to parse DT")
> > Signed-off-by: Chen-Yu Tsai <wens@csie.org>
>
> Reviewed-by: Heiko Stuebner <heiko@sntech.de

It's been close to three weeks since this was sent.
Any chance we can get this into v5.10 or v5.11?


Regards
ChenYu

> I'll pick up patches 2-4 separately, after giving Rob a chance to look at
> the simple binding.
>
>
> Heiko
>
> > ---
> > Changes since v1:
> >
> >   - Rewrite subject to match existing convention and reference
> >     'ep-gpios' DT property instead of the 'ep_gpio' field
> > ---
> >  drivers/pci/controller/pcie-rockchip.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/pci/controller/pcie-rockchip.c b/drivers/pci/controller/pcie-rockchip.c
> > index 904dec0d3a88..c95950e9004f 100644
> > --- a/drivers/pci/controller/pcie-rockchip.c
> > +++ b/drivers/pci/controller/pcie-rockchip.c
> > @@ -118,7 +118,7 @@ int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
> >       }
> >
> >       if (rockchip->is_rc) {
> > -             rockchip->ep_gpio = devm_gpiod_get(dev, "ep", GPIOD_OUT_HIGH);
> > +             rockchip->ep_gpio = devm_gpiod_get_optional(dev, "ep", GPIOD_OUT_HIGH);
> >               if (IS_ERR(rockchip->ep_gpio)) {
> >                       dev_err(dev, "missing ep-gpios property in node\n");
> >                       return PTR_ERR(rockchip->ep_gpio);
> >
>
>
>
>
Rob Herring (Arm) Dec. 7, 2020, 2:11 p.m. UTC | #3 
On Wed, Nov 18, 2020 at 1:17 AM Chen-Yu Tsai <wens@kernel.org> wrote:
>
> From: Chen-Yu Tsai <wens@csie.org>
>
> The Rockchip PCIe controller DT binding clearly states that 'ep-gpios' is
> an optional property. And indeed there are boards that don't require it.
>
> Make the driver follow the binding by using devm_gpiod_get_optional()
> instead of devm_gpiod_get().
>
> Fixes: e77f847df54c ("PCI: rockchip: Add Rockchip PCIe controller support")
> Fixes: 956cd99b35a8 ("PCI: rockchip: Separate common code from RC driver")
> Fixes: 964bac9455be ("PCI: rockchip: Split out rockchip_pcie_parse_dt() to parse DT")
> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> ---
> Changes since v1:
>
>   - Rewrite subject to match existing convention and reference
>     'ep-gpios' DT property instead of the 'ep_gpio' field
> ---
>  drivers/pci/controller/pcie-rockchip.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/controller/pcie-rockchip.c b/drivers/pci/controller/pcie-rockchip.c
> index 904dec0d3a88..c95950e9004f 100644
> --- a/drivers/pci/controller/pcie-rockchip.c
> +++ b/drivers/pci/controller/pcie-rockchip.c
> @@ -118,7 +118,7 @@ int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
>         }
>
>         if (rockchip->is_rc) {
> -               rockchip->ep_gpio = devm_gpiod_get(dev, "ep", GPIOD_OUT_HIGH);
> +               rockchip->ep_gpio = devm_gpiod_get_optional(dev, "ep", GPIOD_OUT_HIGH);
>                 if (IS_ERR(rockchip->ep_gpio)) {
>                         dev_err(dev, "missing ep-gpios property in node\n");

You should drop the error message. What it says is now never the
reason for the error and it could most likely be a deferred probe
which you don't want an error message for.

>                         return PTR_ERR(rockchip->ep_gpio);
> --
> 2.29.1
>
Chen-Yu Tsai Dec. 12, 2020, 3:18 p.m. UTC | #4 
On Mon, Dec 7, 2020 at 10:11 PM Rob Herring <robh@kernel.org> wrote:
>
> On Wed, Nov 18, 2020 at 1:17 AM Chen-Yu Tsai <wens@kernel.org> wrote:
> >
> > From: Chen-Yu Tsai <wens@csie.org>
> >
> > The Rockchip PCIe controller DT binding clearly states that 'ep-gpios' is
> > an optional property. And indeed there are boards that don't require it.
> >
> > Make the driver follow the binding by using devm_gpiod_get_optional()
> > instead of devm_gpiod_get().
> >
> > Fixes: e77f847df54c ("PCI: rockchip: Add Rockchip PCIe controller support")
> > Fixes: 956cd99b35a8 ("PCI: rockchip: Separate common code from RC driver")
> > Fixes: 964bac9455be ("PCI: rockchip: Split out rockchip_pcie_parse_dt() to parse DT")
> > Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> > ---
> > Changes since v1:
> >
> >   - Rewrite subject to match existing convention and reference
> >     'ep-gpios' DT property instead of the 'ep_gpio' field
> > ---
> >  drivers/pci/controller/pcie-rockchip.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/pci/controller/pcie-rockchip.c b/drivers/pci/controller/pcie-rockchip.c
> > index 904dec0d3a88..c95950e9004f 100644
> > --- a/drivers/pci/controller/pcie-rockchip.c
> > +++ b/drivers/pci/controller/pcie-rockchip.c
> > @@ -118,7 +118,7 @@ int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
> >         }
> >
> >         if (rockchip->is_rc) {
> > -               rockchip->ep_gpio = devm_gpiod_get(dev, "ep", GPIOD_OUT_HIGH);
> > +               rockchip->ep_gpio = devm_gpiod_get_optional(dev, "ep", GPIOD_OUT_HIGH);
> >                 if (IS_ERR(rockchip->ep_gpio)) {
> >                         dev_err(dev, "missing ep-gpios property in node\n");
>
> You should drop the error message. What it says is now never the
> reason for the error and it could most likely be a deferred probe
> which you don't want an error message for.

What about switching to dev_err_probe() instead?

That way deferred probe errors get silenced (or get a better debug message),
and error messages for other issues, such as miswritten gpio properties are
still produced.

ChenYu

> >                         return PTR_ERR(rockchip->ep_gpio);
> > --
> > 2.29.1
> >
Rob Herring (Arm) Dec. 14, 2020, 2:17 p.m. UTC | #5 
On Sat, Dec 12, 2020 at 9:18 AM Chen-Yu Tsai <wens@kernel.org> wrote:
>
> On Mon, Dec 7, 2020 at 10:11 PM Rob Herring <robh@kernel.org> wrote:
> >
> > On Wed, Nov 18, 2020 at 1:17 AM Chen-Yu Tsai <wens@kernel.org> wrote:
> > >
> > > From: Chen-Yu Tsai <wens@csie.org>
> > >
> > > The Rockchip PCIe controller DT binding clearly states that 'ep-gpios' is
> > > an optional property. And indeed there are boards that don't require it.
> > >
> > > Make the driver follow the binding by using devm_gpiod_get_optional()
> > > instead of devm_gpiod_get().
> > >
> > > Fixes: e77f847df54c ("PCI: rockchip: Add Rockchip PCIe controller support")
> > > Fixes: 956cd99b35a8 ("PCI: rockchip: Separate common code from RC driver")
> > > Fixes: 964bac9455be ("PCI: rockchip: Split out rockchip_pcie_parse_dt() to parse DT")
> > > Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> > > ---
> > > Changes since v1:
> > >
> > >   - Rewrite subject to match existing convention and reference
> > >     'ep-gpios' DT property instead of the 'ep_gpio' field
> > > ---
> > >  drivers/pci/controller/pcie-rockchip.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/pci/controller/pcie-rockchip.c b/drivers/pci/controller/pcie-rockchip.c
> > > index 904dec0d3a88..c95950e9004f 100644
> > > --- a/drivers/pci/controller/pcie-rockchip.c
> > > +++ b/drivers/pci/controller/pcie-rockchip.c
> > > @@ -118,7 +118,7 @@ int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
> > >         }
> > >
> > >         if (rockchip->is_rc) {
> > > -               rockchip->ep_gpio = devm_gpiod_get(dev, "ep", GPIOD_OUT_HIGH);
> > > +               rockchip->ep_gpio = devm_gpiod_get_optional(dev, "ep", GPIOD_OUT_HIGH);
> > >                 if (IS_ERR(rockchip->ep_gpio)) {
> > >                         dev_err(dev, "missing ep-gpios property in node\n");
> >
> > You should drop the error message. What it says is now never the
> > reason for the error and it could most likely be a deferred probe
> > which you don't want an error message for.
>
> What about switching to dev_err_probe() instead?
>
> That way deferred probe errors get silenced (or get a better debug message),
> and error messages for other issues, such as miswritten gpio properties are
> still produced.

I guess, but those errors should be in the subsystem code IMO rather
than every driver.

Rob
Robin Murphy Dec. 14, 2020, 3:18 p.m. UTC | #6 
On 2020-11-18 07:17, Chen-Yu Tsai wrote:
> From: Chen-Yu Tsai <wens@csie.org>
> 
> Only the NanoPC T4 hs the PCIe reset pin routed to the SoC. For the
> NanoPi M4 family, no such signal is routed to the expansion header on
> the base board.
> 
> As the schematics for the expansion board were not released, it is
> unclear how this is handled, but the likely answer is that the signal
> is always pulled high.
> 
> Move the ep-gpios property from the common nanopi4.dtsi file to the
> board level nanopc-t4.dts file. This makes the nanopi-m4 lack ep-gpios,
> matching the board design.
> 
> A companion patch "PCI: rockchip: make ep_gpio optional" for the Linux
> driver is required, as the driver currently requires the property to be
> present.

I concur that this is a more correct description per the schematics (the 
SOM-RK3399 Dev Kit carrier board is the only other thing showing PERST# 
wired up like this), and whatever the M4 hats are doing they clearly 
aren't doing it with GPIO2_A4 either way.

Reviewed-by: Robin Murphy <robin.murphy@arm.com>

> Fixes: e7a095908227 ("arm64: dts: rockchip: Add devicetree for NanoPC-T4")
> Signed-off-by: Chen-Yu Tsai <wens@csie.org>
> ---
>   arch/arm64/boot/dts/rockchip/rk3399-nanopc-t4.dts | 1 +
>   arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi  | 1 -
>   2 files changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/boot/dts/rockchip/rk3399-nanopc-t4.dts b/arch/arm64/boot/dts/rockchip/rk3399-nanopc-t4.dts
> index e0d75617bb7e..452728b82e42 100644
> --- a/arch/arm64/boot/dts/rockchip/rk3399-nanopc-t4.dts
> +++ b/arch/arm64/boot/dts/rockchip/rk3399-nanopc-t4.dts
> @@ -95,6 +95,7 @@ map3 {
>   };
>   
>   &pcie0 {
> +	ep-gpios = <&gpio2 RK_PA4 GPIO_ACTIVE_HIGH>;
>   	num-lanes = <4>;
>   	vpcie3v3-supply = <&vcc3v3_sys>;
>   };
> diff --git a/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi b/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi
> index 76a8b40a93c6..48ed4aaa37f3 100644
> --- a/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi
> +++ b/arch/arm64/boot/dts/rockchip/rk3399-nanopi4.dtsi
> @@ -504,7 +504,6 @@ &pcie_phy {
>   };
>   
>   &pcie0 {
> -	ep-gpios = <&gpio2 RK_PA4 GPIO_ACTIVE_HIGH>;
>   	max-link-speed = <2>;
>   	num-lanes = <2>;
>   	vpcie0v9-supply = <&vcca0v9_s3>;
>