[RFC PATCH v3 0/7] Reconstruct rockchip's PCIe and PCIe-PHY driver for per-lane PHY model

[RFC PATCH v3 0/7] Reconstruct rockchip's PCIe and PCIe-PHY driver for per-lane PHY model

[Shawn Lin]

This patchset is trying to reconstruct PCIe and PCIe-PHY driver
for rockchip platform in order to support per-lane PHY mode. And
we could idle the inactive lane(s) finally.

We deprecate the legacy PHY mode but the code could still
support it in order not to break backware compatibility of DTB. And I
organize the patches carefully so that we don't introduce git-bisect
issue.

Note that I carry on Jeffy's test tag from v2 as he has locally tested
my v3 patchset.


Changes in v3:
- kill rockchip_pcie_manipulate_phys and related stuff
- use phys array
- improve the commit msg
- remove unnecessary forward declaration
- keep mutex inside struct rockchip_pcie_phy
- fix wrong check of args number
- move de-idle lanes after deasserting the reset
- use cached lanes_map to avoid powering off inactive
  lanes twice
- rename the commit tile

Changes in v2:
- deprecate legacy PHY model
- improve rockchip_pcie_phy_of_xlate
- fix wrong calculation of pwr_cnt and add new init_cnt
- add internal locking
- introduce per-lane data to simply the code

Shawn Lin (7):
  PCI: rockchip: split out rockchip_pcie_get_phys
  PCI: rockchip: introduce per-lanes PHYs support
  phy: rockcip-pcie: reconstruct driver to support per-lane PHYs
  PCI: rockchip: idle the inactive PHY(s)
  arm64: dts: rockchip: convert PCIe to use per-lane PHYs for rk3339
  dt-bindings: PCI: rockchip: convert to use per-lane PHY model
  dt-bindings: phy: convert to use per-lane Rockchip PCIe PHY

 .../devicetree/bindings/pci/rockchip-pcie.txt      |  25 ++++-
 .../devicetree/bindings/phy/rockchip-pcie-phy.txt  |   7 +-
 arch/arm64/boot/dts/rockchip/rk3399.dtsi           |   8 +-
 drivers/pci/host/pcie-rockchip.c                   | 123 ++++++++++++++++----
 drivers/phy/rockchip/phy-rockchip-pcie.c           | 124 ++++++++++++++++++---
 5 files changed, 245 insertions(+), 42 deletions(-)

-- 
1.9.1

[RFC PATCH v3 1/7] PCI: rockchip: split out rockchip_pcie_get_phys

[Shawn Lin]

We plan to introduce per-lanes PHY, so split out new
function to make it easy in the future. No functional
change intended.

Signed-off-by: Shawn Lin <shawn.lin-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
Tested-by: Jeffy Chen <jeffy.chen-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
---

Changes in v3: None
Changes in v2: None

 drivers/pci/host/pcie-rockchip.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c
index 5acf869..6632a51 100644
--- a/drivers/pci/host/pcie-rockchip.c
+++ b/drivers/pci/host/pcie-rockchip.c
@@ -853,6 +853,19 @@ static void rockchip_pcie_legacy_int_handler(struct irq_desc *desc)
 	chained_irq_exit(chip, desc);
 }
 
+static int rockchip_pcie_get_phys(struct rockchip_pcie *rockchip)
+{
+	struct device *dev = rockchip->dev;
+
+	rockchip->phy = devm_phy_get(dev, "pcie-phy");
+	if (IS_ERR(rockchip->phy)) {
+		if (PTR_ERR(rockchip->phy) != -EPROBE_DEFER)
+			dev_err(dev, "missing phy\n");
+		return PTR_ERR(rockchip->phy);
+	}
+
+	return 0;
+}
 
 /**
  * rockchip_pcie_parse_dt - Parse Device Tree
@@ -883,12 +896,8 @@ static int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
 	if (IS_ERR(rockchip->apb_base))
 		return PTR_ERR(rockchip->apb_base);
 
-	rockchip->phy = devm_phy_get(dev, "pcie-phy");
-	if (IS_ERR(rockchip->phy)) {
-		if (PTR_ERR(rockchip->phy) != -EPROBE_DEFER)
-			dev_err(dev, "missing phy\n");
+	if (rockchip_pcie_get_phys(rockchip))
 		return PTR_ERR(rockchip->phy);
-	}
 
 	rockchip->lanes = 1;
 	err = of_property_read_u32(node, "num-lanes", &rockchip->lanes);
-- 
1.9.1


--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [RFC PATCH v3 1/7] PCI: rockchip: split out rockchip_pcie_get_phys

[Brian Norris]

Hi Shawn,

Some nitpicks :)

On Tue, Jul 18, 2017 at 03:56:57PM +0800, Shawn Lin wrote:
> We plan to introduce per-lanes PHY, so split out new
> function to make it easy in the future. No functional
> change intended.
> 
> Signed-off-by: Shawn Lin <shawn.lin-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
> Tested-by: Jeffy Chen <jeffy.chen-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
> ---
> 
> Changes in v3: None
> Changes in v2: None
> 
>  drivers/pci/host/pcie-rockchip.c | 19 ++++++++++++++-----
>  1 file changed, 14 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c
> index 5acf869..6632a51 100644
> --- a/drivers/pci/host/pcie-rockchip.c
> +++ b/drivers/pci/host/pcie-rockchip.c
> @@ -853,6 +853,19 @@ static void rockchip_pcie_legacy_int_handler(struct irq_desc *desc)
>  	chained_irq_exit(chip, desc);
>  }
>  
> +static int rockchip_pcie_get_phys(struct rockchip_pcie *rockchip)
> +{
> +	struct device *dev = rockchip->dev;
> +
> +	rockchip->phy = devm_phy_get(dev, "pcie-phy");
> +	if (IS_ERR(rockchip->phy)) {
> +		if (PTR_ERR(rockchip->phy) != -EPROBE_DEFER)
> +			dev_err(dev, "missing phy\n");
> +		return PTR_ERR(rockchip->phy);
> +	}
> +
> +	return 0;
> +}

You promptly relocate this entire function in the next patch. Would be
nice if you picked one place and kept it there. (I believe this is
partly because of how the previous revision had some more complicated
functions added nearby that were related. But you've killed that now.)

>  
>  /**
>   * rockchip_pcie_parse_dt - Parse Device Tree
> @@ -883,12 +896,8 @@ static int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
>  	if (IS_ERR(rockchip->apb_base))
>  		return PTR_ERR(rockchip->apb_base);
>  
> -	rockchip->phy = devm_phy_get(dev, "pcie-phy");
> -	if (IS_ERR(rockchip->phy)) {
> -		if (PTR_ERR(rockchip->phy) != -EPROBE_DEFER)
> -			dev_err(dev, "missing phy\n");
> +	if (rockchip_pcie_get_phys(rockchip))
>  		return PTR_ERR(rockchip->phy);

In the next patch you (correctly) change this to actually capture the
return code from rockchip_pcie_get_phys(). Seems like you should just do
that in this patch.

Brian

> -	}
>  
>  	rockchip->lanes = 1;
>  	err = of_property_read_u32(node, "num-lanes", &rockchip->lanes);
> -- 
> 1.9.1
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[RFC PATCH v3 2/7] PCI: rockchip: introduce per-lanes PHYs support

[Shawn Lin]

We distinguish the legacy PHY with the newer per-lane
PHYs by adding legacy_phy flag. Note that the legacy phy
is still the first option to be searched in order not to
break the backward compatibility of DTB.

Signed-off-by: Shawn Lin <shawn.lin-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
Tested-by: Jeffy Chen <jeffy.chen-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
---

Changes in v3:
- kill rockchip_pcie_manipulate_phys and related stuff
- use phys array
- improve the commit msg

Changes in v2: None

 drivers/pci/host/pcie-rockchip.c | 100 +++++++++++++++++++++++++++------------
 1 file changed, 69 insertions(+), 31 deletions(-)

diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c
index 6632a51..4cc6aec 100644
--- a/drivers/pci/host/pcie-rockchip.c
+++ b/drivers/pci/host/pcie-rockchip.c
@@ -47,6 +47,7 @@
 #define HIWORD_UPDATE_BIT(val)		HIWORD_UPDATE(val, val)
 
 #define ENCODE_LANES(x)			((((x) >> 1) & 3) << 4)
+#define MAX_LANE_NUM			4
 
 #define PCIE_CLIENT_BASE		0x0
 #define PCIE_CLIENT_CONFIG		(PCIE_CLIENT_BASE + 0x00)
@@ -210,7 +211,8 @@
 struct rockchip_pcie {
 	void	__iomem *reg_base;		/* DT axi-base */
 	void	__iomem *apb_base;		/* DT apb-base */
-	struct	phy *phy;
+	bool    legacy_phy;
+	struct  phy *phys[MAX_LANE_NUM];
 	struct	reset_control *core_rst;
 	struct	reset_control *mgmt_rst;
 	struct	reset_control *mgmt_sticky_rst;
@@ -507,6 +509,46 @@ static void rockchip_pcie_set_power_limit(struct rockchip_pcie *rockchip)
 	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_DCR);
 }
 
+static int rockchip_pcie_get_phys(struct rockchip_pcie *rockchip)
+{
+	struct device *dev = rockchip->dev;
+	struct phy *phy;
+	char *name;
+	u32 i;
+
+	rockchip->phys[0] = devm_phy_get(dev, "pcie-phy");
+	if (IS_ERR(rockchip->phys[0])) {
+		if (PTR_ERR(rockchip->phys[0]) == -EPROBE_DEFER)
+			return PTR_ERR(rockchip->phys[0]);
+		dev_dbg(dev, "missing legacy phy, and search for per-lane PHY\n");
+	} else {
+		rockchip->legacy_phy = true;
+		dev_warn(dev, "legacy phy model is deprecated!\n");
+		return 0;
+	}
+
+	for (i = 0; i < MAX_LANE_NUM; i++) {
+		name = kasprintf(GFP_KERNEL, "pcie-phy-%u", i);
+		if (!name)
+			return -ENOMEM;
+
+		phy = devm_of_phy_get(rockchip->dev,
+				      rockchip->dev->of_node, name);
+		kfree(name);
+
+		if (IS_ERR(phy)) {
+			if (PTR_ERR(phy) != -EPROBE_DEFER)
+				dev_err(dev, "missing phy for lane %d: %ld\n",
+					i, PTR_ERR(phy));
+			return PTR_ERR(phy);
+		}
+
+		rockchip->phys[i] = phy;
+	}
+
+	return 0;
+}
+
 /**
  * rockchip_pcie_init_port - Initialize hardware
  * @rockchip: PCIe port information
@@ -514,7 +556,7 @@ static void rockchip_pcie_set_power_limit(struct rockchip_pcie *rockchip)
 static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
 {
 	struct device *dev = rockchip->dev;
-	int err;
+	int err, i;
 	u32 status;
 
 	gpiod_set_value(rockchip->ep_gpio, 0);
@@ -537,10 +579,12 @@ static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
 		return err;
 	}
 
-	err = phy_init(rockchip->phy);
-	if (err < 0) {
-		dev_err(dev, "fail to init phy, err %d\n", err);
-		return err;
+	for (i = 0; i < MAX_LANE_NUM; i++) {
+		err = phy_init(rockchip->phys[i]);
+		if (err) {
+			dev_err(dev, "init phy err %d\n", err);
+			return err;
+		}
 	}
 
 	err = reset_control_assert(rockchip->core_rst);
@@ -602,10 +646,12 @@ static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
 			    PCIE_CLIENT_MODE_RC,
 			    PCIE_CLIENT_CONFIG);
 
-	err = phy_power_on(rockchip->phy);
-	if (err) {
-		dev_err(dev, "fail to power on phy, err %d\n", err);
-		return err;
+	for (i = 0; i < MAX_LANE_NUM; i++) {
+		err = phy_power_on(rockchip->phys[i]);
+		if (err) {
+			dev_err(dev, "power on phy err %d\n", err);
+			return err;
+		}
 	}
 
 	/*
@@ -853,20 +899,6 @@ static void rockchip_pcie_legacy_int_handler(struct irq_desc *desc)
 	chained_irq_exit(chip, desc);
 }
 
-static int rockchip_pcie_get_phys(struct rockchip_pcie *rockchip)
-{
-	struct device *dev = rockchip->dev;
-
-	rockchip->phy = devm_phy_get(dev, "pcie-phy");
-	if (IS_ERR(rockchip->phy)) {
-		if (PTR_ERR(rockchip->phy) != -EPROBE_DEFER)
-			dev_err(dev, "missing phy\n");
-		return PTR_ERR(rockchip->phy);
-	}
-
-	return 0;
-}
-
 /**
  * rockchip_pcie_parse_dt - Parse Device Tree
  * @rockchip: PCIe port information
@@ -896,8 +928,9 @@ static int rockchip_pcie_parse_dt(struct rockchip_pcie *rockchip)
 	if (IS_ERR(rockchip->apb_base))
 		return PTR_ERR(rockchip->apb_base);
 
-	if (rockchip_pcie_get_phys(rockchip))
-		return PTR_ERR(rockchip->phy);
+	err = rockchip_pcie_get_phys(rockchip);
+	if (err)
+		return err;
 
 	rockchip->lanes = 1;
 	err = of_property_read_u32(node, "num-lanes", &rockchip->lanes);
@@ -1282,7 +1315,7 @@ static int rockchip_pcie_wait_l2(struct rockchip_pcie *rockchip)
 static int __maybe_unused rockchip_pcie_suspend_noirq(struct device *dev)
 {
 	struct rockchip_pcie *rockchip = dev_get_drvdata(dev);
-	int ret;
+	int ret, i;
 
 	/* disable core and cli int since we don't need to ack PME_ACK */
 	rockchip_pcie_write(rockchip, (PCIE_CLIENT_INT_CLI << 16) |
@@ -1295,8 +1328,10 @@ static int __maybe_unused rockchip_pcie_suspend_noirq(struct device *dev)
 		return ret;
 	}
 
-	phy_power_off(rockchip->phy);
-	phy_exit(rockchip->phy);
+	for (i = 0; i < MAX_LANE_NUM; i++) {
+		phy_power_off(rockchip->phys[i]);
+		phy_exit(rockchip->phys[i]);
+	}
 
 	clk_disable_unprepare(rockchip->clk_pcie_pm);
 	clk_disable_unprepare(rockchip->hclk_pcie);
@@ -1532,14 +1567,17 @@ static int rockchip_pcie_remove(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
 	struct rockchip_pcie *rockchip = dev_get_drvdata(dev);
+	int i;
 
 	pci_stop_root_bus(rockchip->root_bus);
 	pci_remove_root_bus(rockchip->root_bus);
 	pci_unmap_iospace(rockchip->io);
 	irq_domain_remove(rockchip->irq_domain);
 
-	phy_power_off(rockchip->phy);
-	phy_exit(rockchip->phy);
+	for (i = 0; i < MAX_LANE_NUM; i++) {
+		phy_power_off(rockchip->phys[i]);
+		phy_exit(rockchip->phys[i]);
+	}
 
 	clk_disable_unprepare(rockchip->clk_pcie_pm);
 	clk_disable_unprepare(rockchip->hclk_pcie);
-- 
1.9.1


--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [RFC PATCH v3 2/7] PCI: rockchip: introduce per-lanes PHYs support

[Brian Norris]

Hi,

On Tue, Jul 18, 2017 at 03:56:58PM +0800, Shawn Lin wrote:
> We distinguish the legacy PHY with the newer per-lane
> PHYs by adding legacy_phy flag. Note that the legacy phy
> is still the first option to be searched in order not to
> break the backward compatibility of DTB.
> 
> Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
> Tested-by: Jeffy Chen <jeffy.chen@rock-chips.com>
> ---
> 
> Changes in v3:
> - kill rockchip_pcie_manipulate_phys and related stuff
> - use phys array
> - improve the commit msg
> 
> Changes in v2: None
> 
>  drivers/pci/host/pcie-rockchip.c | 100 +++++++++++++++++++++++++++------------
>  1 file changed, 69 insertions(+), 31 deletions(-)
> 
> diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c
> index 6632a51..4cc6aec 100644
> --- a/drivers/pci/host/pcie-rockchip.c
> +++ b/drivers/pci/host/pcie-rockchip.c

...

> @@ -537,10 +579,12 @@ static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
>  		return err;
>  	}
>  
> -	err = phy_init(rockchip->phy);
> -	if (err < 0) {
> -		dev_err(dev, "fail to init phy, err %d\n", err);
> -		return err;
> +	for (i = 0; i < MAX_LANE_NUM; i++) {
> +		err = phy_init(rockchip->phys[i]);
> +		if (err) {
> +			dev_err(dev, "init phy err %d\n", err);

Include the PHY index in the error message?

> +			return err;
> +		}
>  	}
>  
>  	err = reset_control_assert(rockchip->core_rst);
> @@ -602,10 +646,12 @@ static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
>  			    PCIE_CLIENT_MODE_RC,
>  			    PCIE_CLIENT_CONFIG);
>  
> -	err = phy_power_on(rockchip->phy);
> -	if (err) {
> -		dev_err(dev, "fail to power on phy, err %d\n", err);
> -		return err;
> +	for (i = 0; i < MAX_LANE_NUM; i++) {
> +		err = phy_power_on(rockchip->phys[i]);
> +		if (err) {
> +			dev_err(dev, "power on phy err %d\n", err);

Same?

> +			return err;
> +		}
>  	}
>  
>  	/*

...

Brian

[RFC PATCH v3 3/7] phy: rockcip-pcie: reconstruct driver to support per-lane PHYs

[Shawn Lin]

This patch reconstructs the whole driver to support per-lane
PHYs. Note that we could also support the legacy PHY if you
don't provide argument to our of_xlate.

Signed-off-by: Shawn Lin <shawn.lin-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
Tested-by: Jeffy Chen <jeffy.chen-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
---

Changes in v3:
- remove unnecessary forward declaration
- keep mutex inside struct rockchip_pcie_phy
- fix wrong check of args number
- move de-idle lanes after deasserting the reset

Changes in v2:
- deprecate legacy PHY model
- improve rockchip_pcie_phy_of_xlate
- fix wrong calculation of pwr_cnt and add new init_cnt
- add internal locking
- introduce per-lane data to simply the code

 drivers/phy/rockchip/phy-rockchip-pcie.c | 124 +++++++++++++++++++++++++++----
 1 file changed, 110 insertions(+), 14 deletions(-)

diff --git a/drivers/phy/rockchip/phy-rockchip-pcie.c b/drivers/phy/rockchip/phy-rockchip-pcie.c
index 6904633..b5005a5 100644
--- a/drivers/phy/rockchip/phy-rockchip-pcie.c
+++ b/drivers/phy/rockchip/phy-rockchip-pcie.c
@@ -73,10 +73,39 @@ struct rockchip_pcie_data {
 struct rockchip_pcie_phy {
 	struct rockchip_pcie_data *phy_data;
 	struct regmap *reg_base;
+	struct phy_pcie_instance {
+		struct phy *phy;
+		u32 index;
+	} phys[PHY_MAX_LANE_NUM];
+	struct mutex pcie_mutex;
 	struct reset_control *phy_rst;
 	struct clk *clk_pciephy_ref;
+	int pwr_cnt;
+	int init_cnt;
 };
 
+static inline
+struct rockchip_pcie_phy *to_pcie_phy(struct phy_pcie_instance *inst)
+{
+	return container_of(inst, struct rockchip_pcie_phy,
+					phys[inst->index]);
+}
+
+static struct phy *rockchip_pcie_phy_of_xlate(struct device *dev,
+					      struct of_phandle_args *args)
+{
+	struct rockchip_pcie_phy *rk_phy = dev_get_drvdata(dev);
+
+	if (args->args_count == 0)
+		return rk_phy->phys[0].phy;
+
+	if (WARN_ON(args->args[0] >= PHY_MAX_LANE_NUM))
+		return ERR_PTR(-ENODEV);
+
+	return rk_phy->phys[args->args[0]].phy;
+}
+
+
 static inline void phy_wr_cfg(struct rockchip_pcie_phy *rk_phy,
 			      u32 addr, u32 data)
 {
@@ -116,29 +145,59 @@ static inline u32 phy_rd_cfg(struct rockchip_pcie_phy *rk_phy,
 
 static int rockchip_pcie_phy_power_off(struct phy *phy)
 {
-	struct rockchip_pcie_phy *rk_phy = phy_get_drvdata(phy);
+	struct phy_pcie_instance *inst = phy_get_drvdata(phy);
+	struct rockchip_pcie_phy *rk_phy = to_pcie_phy(inst);
 	int err = 0;
 
+	mutex_lock(&rk_phy->pcie_mutex);
+
+	regmap_write(rk_phy->reg_base,
+		     rk_phy->phy_data->pcie_laneoff,
+		     HIWORD_UPDATE(PHY_LANE_IDLE_OFF,
+				   PHY_LANE_IDLE_MASK,
+				   PHY_LANE_IDLE_A_SHIFT + inst->index));
+
+	if (--rk_phy->pwr_cnt)
+		goto err_out;
+
 	err = reset_control_assert(rk_phy->phy_rst);
 	if (err) {
 		dev_err(&phy->dev, "assert phy_rst err %d\n", err);
-		return err;
+		goto err_restore;
 	}
 
+err_out:
+	mutex_unlock(&rk_phy->pcie_mutex);
 	return 0;
+
+err_restore:
+	++rk_phy->pwr_cnt;
+	regmap_write(rk_phy->reg_base,
+		     rk_phy->phy_data->pcie_laneoff,
+		     HIWORD_UPDATE(!PHY_LANE_IDLE_OFF,
+				   PHY_LANE_IDLE_MASK,
+				   PHY_LANE_IDLE_A_SHIFT + inst->index));
+	mutex_unlock(&rk_phy->pcie_mutex);
+	return err;
 }
 
 static int rockchip_pcie_phy_power_on(struct phy *phy)
 {
-	struct rockchip_pcie_phy *rk_phy = phy_get_drvdata(phy);
+	struct phy_pcie_instance *inst = phy_get_drvdata(phy);
+	struct rockchip_pcie_phy *rk_phy = to_pcie_phy(inst);
 	int err = 0;
 	u32 status;
 	unsigned long timeout;
 
+	mutex_lock(&rk_phy->pcie_mutex);
+
+	if (rk_phy->pwr_cnt++)
+		goto err_out;
+
 	err = reset_control_deassert(rk_phy->phy_rst);
 	if (err) {
 		dev_err(&phy->dev, "deassert phy_rst err %d\n", err);
-		return err;
+		goto err_pwr_cnt;
 	}
 
 	regmap_write(rk_phy->reg_base, rk_phy->phy_data->pcie_conf,
@@ -146,6 +205,12 @@ static int rockchip_pcie_phy_power_on(struct phy *phy)
 				   PHY_CFG_ADDR_MASK,
 				   PHY_CFG_ADDR_SHIFT));
 
+	regmap_write(rk_phy->reg_base,
+		     rk_phy->phy_data->pcie_laneoff,
+		     HIWORD_UPDATE(!PHY_LANE_IDLE_OFF,
+				   PHY_LANE_IDLE_MASK,
+				   PHY_LANE_IDLE_A_SHIFT + inst->index));
+
 	/*
 	 * No documented timeout value for phy operation below,
 	 * so we make it large enough here. And we use loop-break
@@ -214,18 +279,29 @@ static int rockchip_pcie_phy_power_on(struct phy *phy)
 		goto err_pll_lock;
 	}
 
+err_out:
+	mutex_unlock(&rk_phy->pcie_mutex);
 	return 0;
 
 err_pll_lock:
 	reset_control_assert(rk_phy->phy_rst);
+err_pwr_cnt:
+	--rk_phy->pwr_cnt;
+	mutex_unlock(&rk_phy->pcie_mutex);
 	return err;
 }
 
 static int rockchip_pcie_phy_init(struct phy *phy)
 {
-	struct rockchip_pcie_phy *rk_phy = phy_get_drvdata(phy);
+	struct phy_pcie_instance *inst = phy_get_drvdata(phy);
+	struct rockchip_pcie_phy *rk_phy = to_pcie_phy(inst);
 	int err = 0;
 
+	mutex_lock(&rk_phy->pcie_mutex);
+
+	if (rk_phy->init_cnt++)
+		goto err_out;
+
 	err = clk_prepare_enable(rk_phy->clk_pciephy_ref);
 	if (err) {
 		dev_err(&phy->dev, "Fail to enable pcie ref clock.\n");
@@ -238,20 +314,33 @@ static int rockchip_pcie_phy_init(struct phy *phy)
 		goto err_reset;
 	}
 
-	return err;
+err_out:
+	mutex_unlock(&rk_phy->pcie_mutex);
+	return 0;
 
 err_reset:
+
 	clk_disable_unprepare(rk_phy->clk_pciephy_ref);
 err_refclk:
+	--rk_phy->init_cnt;
+	mutex_unlock(&rk_phy->pcie_mutex);
 	return err;
 }
 
 static int rockchip_pcie_phy_exit(struct phy *phy)
 {
-	struct rockchip_pcie_phy *rk_phy = phy_get_drvdata(phy);
+	struct phy_pcie_instance *inst = phy_get_drvdata(phy);
+	struct rockchip_pcie_phy *rk_phy = to_pcie_phy(inst);
+
+	mutex_lock(&rk_phy->pcie_mutex);
+
+	if (--rk_phy->init_cnt)
+		goto err_init_cnt;
 
 	clk_disable_unprepare(rk_phy->clk_pciephy_ref);
 
+err_init_cnt:
+	mutex_unlock(&rk_phy->pcie_mutex);
 	return 0;
 }
 
@@ -283,10 +372,10 @@ static int rockchip_pcie_phy_probe(struct platform_device *pdev)
 {
 	struct device *dev = &pdev->dev;
 	struct rockchip_pcie_phy *rk_phy;
-	struct phy *generic_phy;
 	struct phy_provider *phy_provider;
 	struct regmap *grf;
 	const struct of_device_id *of_id;
+	int i;
 
 	grf = syscon_node_to_regmap(dev->parent->of_node);
 	if (IS_ERR(grf)) {
@@ -305,6 +394,8 @@ static int rockchip_pcie_phy_probe(struct platform_device *pdev)
 	rk_phy->phy_data = (struct rockchip_pcie_data *)of_id->data;
 	rk_phy->reg_base = grf;
 
+	mutex_init(&rk_phy->pcie_mutex);
+
 	rk_phy->phy_rst = devm_reset_control_get(dev, "phy");
 	if (IS_ERR(rk_phy->phy_rst)) {
 		if (PTR_ERR(rk_phy->phy_rst) != -EPROBE_DEFER)
@@ -319,14 +410,19 @@ static int rockchip_pcie_phy_probe(struct platform_device *pdev)
 		return PTR_ERR(rk_phy->clk_pciephy_ref);
 	}
 
-	generic_phy = devm_phy_create(dev, dev->of_node, &ops);
-	if (IS_ERR(generic_phy)) {
-		dev_err(dev, "failed to create PHY\n");
-		return PTR_ERR(generic_phy);
+	for (i = 0; i < PHY_MAX_LANE_NUM; i++) {
+		rk_phy->phys[i].phy = devm_phy_create(dev, dev->of_node, &ops);
+		if (IS_ERR(rk_phy->phys[i].phy)) {
+			dev_err(dev, "failed to create PHY%d\n", i);
+			return PTR_ERR(rk_phy->phys[i].phy);
+		}
+		rk_phy->phys[i].index = i;
+		phy_set_drvdata(rk_phy->phys[i].phy, &rk_phy->phys[i]);
 	}
 
-	phy_set_drvdata(generic_phy, rk_phy);
-	phy_provider = devm_of_phy_provider_register(dev, of_phy_simple_xlate);
+	platform_set_drvdata(pdev, rk_phy);
+	phy_provider = devm_of_phy_provider_register(dev,
+					rockchip_pcie_phy_of_xlate);
 
 	return PTR_ERR_OR_ZERO(phy_provider);
 }
-- 
1.9.1


--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [RFC PATCH v3 0/7] Reconstruct rockchip's PCIe and PCIe-PHY driver for per-lane PHY model

[Brian Norris]

On Tue, Jul 18, 2017 at 03:56:56PM +0800, Shawn Lin wrote:
> 
> This patchset is trying to reconstruct PCIe and PCIe-PHY driver
> for rockchip platform in order to support per-lane PHY mode. And
> we could idle the inactive lane(s) finally.
> 
> We deprecate the legacy PHY mode but the code could still
> support it in order not to break backware compatibility of DTB. And I
> organize the patches carefully so that we don't introduce git-bisect
> issue.
> 
> Note that I carry on Jeffy's test tag from v2 as he has locally tested
> my v3 patchset.

I have a few nits for the patchset (will reply shortly), but otherwise
(if we're *really* going to go with this approach; I still would
appreciate a reply from Kishon on the original patchset, where I
suggested alternatives) at least the code is much better this time
around. So for the whole thing:

Reviewed-by: Brian Norris <briannorris-F7+t8E8rja9g9hUCZPvPmw@public.gmane.org>

> Changes in v3:
> - kill rockchip_pcie_manipulate_phys and related stuff
> - use phys array
> - improve the commit msg
> - remove unnecessary forward declaration
> - keep mutex inside struct rockchip_pcie_phy
> - fix wrong check of args number
> - move de-idle lanes after deasserting the reset
> - use cached lanes_map to avoid powering off inactive
>   lanes twice
> - rename the commit tile
> 
> Changes in v2:
> - deprecate legacy PHY model
> - improve rockchip_pcie_phy_of_xlate
> - fix wrong calculation of pwr_cnt and add new init_cnt
> - add internal locking
> - introduce per-lane data to simply the code
> 
> Shawn Lin (7):
>   PCI: rockchip: split out rockchip_pcie_get_phys
>   PCI: rockchip: introduce per-lanes PHYs support
>   phy: rockcip-pcie: reconstruct driver to support per-lane PHYs
>   PCI: rockchip: idle the inactive PHY(s)
>   arm64: dts: rockchip: convert PCIe to use per-lane PHYs for rk3339
>   dt-bindings: PCI: rockchip: convert to use per-lane PHY model
>   dt-bindings: phy: convert to use per-lane Rockchip PCIe PHY
> 
>  .../devicetree/bindings/pci/rockchip-pcie.txt      |  25 ++++-
>  .../devicetree/bindings/phy/rockchip-pcie-phy.txt  |   7 +-
>  arch/arm64/boot/dts/rockchip/rk3399.dtsi           |   8 +-
>  drivers/pci/host/pcie-rockchip.c                   | 123 ++++++++++++++++----
>  drivers/phy/rockchip/phy-rockchip-pcie.c           | 124 ++++++++++++++++++---
>  5 files changed, 245 insertions(+), 42 deletions(-)
> 
> -- 
> 1.9.1
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [RFC PATCH v3 0/7] Reconstruct rockchip's PCIe and PCIe-PHY driver for per-lane PHY model

[Shawn Lin]

Hi Brian,

On 2017/7/19 4:29, Brian Norris wrote:
> On Tue, Jul 18, 2017 at 03:56:56PM +0800, Shawn Lin wrote:
>>
>> This patchset is trying to reconstruct PCIe and PCIe-PHY driver
>> for rockchip platform in order to support per-lane PHY mode. And
>> we could idle the inactive lane(s) finally.
>>
>> We deprecate the legacy PHY mode but the code could still
>> support it in order not to break backware compatibility of DTB. And I
>> organize the patches carefully so that we don't introduce git-bisect
>> issue.
>>
>> Note that I carry on Jeffy's test tag from v2 as he has locally tested
>> my v3 patchset.
> 
> I have a few nits for the patchset (will reply shortly), but otherwise
> (if we're *really* going to go with this approach; I still would
> appreciate a reply from Kishon on the original patchset, where I
> suggested alternatives) at least the code is much better this time
> around. So for the whole thing:
> 
> Reviewed-by: Brian Norris <briannorris@chromium.org>

Thanks for your review and Jeffy's test!

I probably will respin my v4 soon and wait for Kishon's final decision.
If Kishon prefer to ack your alternative solution, I'm fine to commit
some other bits as your alternatives seem a much simpler way.

> 
>> Changes in v3:
>> - kill rockchip_pcie_manipulate_phys and related stuff
>> - use phys array
>> - improve the commit msg
>> - remove unnecessary forward declaration
>> - keep mutex inside struct rockchip_pcie_phy
>> - fix wrong check of args number
>> - move de-idle lanes after deasserting the reset
>> - use cached lanes_map to avoid powering off inactive
>>    lanes twice
>> - rename the commit tile
>>
>> Changes in v2:
>> - deprecate legacy PHY model
>> - improve rockchip_pcie_phy_of_xlate
>> - fix wrong calculation of pwr_cnt and add new init_cnt
>> - add internal locking
>> - introduce per-lane data to simply the code
>>
>> Shawn Lin (7):
>>    PCI: rockchip: split out rockchip_pcie_get_phys
>>    PCI: rockchip: introduce per-lanes PHYs support
>>    phy: rockcip-pcie: reconstruct driver to support per-lane PHYs
>>    PCI: rockchip: idle the inactive PHY(s)
>>    arm64: dts: rockchip: convert PCIe to use per-lane PHYs for rk3339
>>    dt-bindings: PCI: rockchip: convert to use per-lane PHY model
>>    dt-bindings: phy: convert to use per-lane Rockchip PCIe PHY
>>
>>   .../devicetree/bindings/pci/rockchip-pcie.txt      |  25 ++++-
>>   .../devicetree/bindings/phy/rockchip-pcie-phy.txt  |   7 +-
>>   arch/arm64/boot/dts/rockchip/rk3399.dtsi           |   8 +-
>>   drivers/pci/host/pcie-rockchip.c                   | 123 ++++++++++++++++----
>>   drivers/phy/rockchip/phy-rockchip-pcie.c           | 124 ++++++++++++++++++---
>>   5 files changed, 245 insertions(+), 42 deletions(-)
>>
>> -- 
>> 1.9.1
>>
>>
> 
> 
> 

[RFC PATCH v3 4/7] PCI: rockchip: idle the inactive PHY(s)

[Shawn Lin]

Check the status of all lanes and idle the inactive one(s).

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Tested-by: Jeffy Chen <jeffy.chen@rock-chips.com>
---

Changes in v3:
- use cached lanes_map to avoid powering off inactive
  lanes twice

Changes in v2: None

 drivers/pci/host/pcie-rockchip.c | 36 ++++++++++++++++++++++++++++++++++--
 1 file changed, 34 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c
index 4cc6aec..d73223f 100644
--- a/drivers/pci/host/pcie-rockchip.c
+++ b/drivers/pci/host/pcie-rockchip.c
@@ -15,6 +15,7 @@
  * (at your option) any later version.
  */
 
+#include <linux/bitrev.h>
 #include <linux/clk.h>
 #include <linux/delay.h>
 #include <linux/gpio/consumer.h>
@@ -112,6 +113,9 @@
 #define   PCIE_CORE_TXCREDIT_CFG1_MUI_SHIFT	16
 #define   PCIE_CORE_TXCREDIT_CFG1_MUI_ENCODE(x) \
 		(((x) >> 3) << PCIE_CORE_TXCREDIT_CFG1_MUI_SHIFT)
+#define PCIE_CORE_LANE_MAP             (PCIE_CORE_CTRL_MGMT_BASE + 0x200)
+#define   PCIE_CORE_LANE_MAP_MASK              0x0000000f
+#define   PCIE_CORE_LANE_MAP_REVERSE           BIT(16)
 #define PCIE_CORE_INT_STATUS		(PCIE_CORE_CTRL_MGMT_BASE + 0x20c)
 #define   PCIE_CORE_INT_PRFPE			BIT(0)
 #define   PCIE_CORE_INT_CRFPE			BIT(1)
@@ -229,6 +233,7 @@ struct rockchip_pcie {
 	struct	regulator *vpcie0v9; /* 0.9V power supply */
 	struct	gpio_desc *ep_gpio;
 	u32	lanes;
+	u8      lanes_map;
 	u8	root_bus_nr;
 	int	link_gen;
 	struct	device *dev;
@@ -301,6 +306,17 @@ static int rockchip_pcie_valid_device(struct rockchip_pcie *rockchip,
 	return 1;
 }
 
+static void rockchip_pcie_lane_map(struct rockchip_pcie *rockchip)
+{
+	u32 val = rockchip_pcie_read(rockchip, PCIE_CORE_LANE_MAP);
+
+	rockchip->lanes_map = val & PCIE_CORE_LANE_MAP_MASK;
+
+	/* The link may be using a reverse-indexed mapping. */
+	if (val & PCIE_CORE_LANE_MAP_REVERSE)
+		rockchip->lanes_map = bitrev8(rockchip->lanes_map) >> 4;
+}
+
 static int rockchip_pcie_rd_own_conf(struct rockchip_pcie *rockchip,
 				     int where, int size, u32 *val)
 {
@@ -737,6 +753,18 @@ static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
 			  PCIE_CORE_PL_CONF_LANE_SHIFT);
 	dev_dbg(dev, "current link width is x%d\n", status);
 
+	if (!rockchip->legacy_phy) {
+		/*  power off unused lane(s) */
+		rockchip_pcie_lane_map(rockchip);
+		for (i = 0; i < MAX_LANE_NUM; i++) {
+			if (rockchip->lanes_map & BIT(i))
+				continue;
+
+			dev_dbg(dev, "idling lane %d\n", i);
+			phy_power_off(rockchip->phys[i]);
+		}
+	}
+
 	rockchip_pcie_write(rockchip, ROCKCHIP_VENDOR_ID,
 			    PCIE_CORE_CONFIG_VENDOR);
 	rockchip_pcie_write(rockchip,
@@ -1329,7 +1357,9 @@ static int __maybe_unused rockchip_pcie_suspend_noirq(struct device *dev)
 	}
 
 	for (i = 0; i < MAX_LANE_NUM; i++) {
-		phy_power_off(rockchip->phys[i]);
+		/* inactive lane is already powered off */
+		if (rockchip->lanes_map & BIT(i))
+			phy_power_off(rockchip->phys[i]);
 		phy_exit(rockchip->phys[i]);
 	}
 
@@ -1575,7 +1605,9 @@ static int rockchip_pcie_remove(struct platform_device *pdev)
 	irq_domain_remove(rockchip->irq_domain);
 
 	for (i = 0; i < MAX_LANE_NUM; i++) {
-		phy_power_off(rockchip->phys[i]);
+		/* inactive lane is already powered off */
+		if (rockchip->lanes_map & BIT(i))
+			phy_power_off(rockchip->phys[i]);
 		phy_exit(rockchip->phys[i]);
 	}
 
-- 
1.9.1

Re: [RFC PATCH v3 4/7] PCI: rockchip: idle the inactive PHY(s)

[Brian Norris]

Hi,

On Tue, Jul 18, 2017 at 03:57:00PM +0800, Shawn Lin wrote:
> Check the status of all lanes and idle the inactive one(s).
> 
> Signed-off-by: Shawn Lin <shawn.lin-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
> Tested-by: Jeffy Chen <jeffy.chen-TNX95d0MmH7DzftRWevZcw@public.gmane.org>
> ---
> 
> Changes in v3:
> - use cached lanes_map to avoid powering off inactive
>   lanes twice
> 
> Changes in v2: None
> 
>  drivers/pci/host/pcie-rockchip.c | 36 ++++++++++++++++++++++++++++++++++--
>  1 file changed, 34 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/pci/host/pcie-rockchip.c b/drivers/pci/host/pcie-rockchip.c
> index 4cc6aec..d73223f 100644
> --- a/drivers/pci/host/pcie-rockchip.c
> +++ b/drivers/pci/host/pcie-rockchip.c
> @@ -15,6 +15,7 @@
>   * (at your option) any later version.
>   */
>  
> +#include <linux/bitrev.h>
>  #include <linux/clk.h>
>  #include <linux/delay.h>
>  #include <linux/gpio/consumer.h>
> @@ -112,6 +113,9 @@
>  #define   PCIE_CORE_TXCREDIT_CFG1_MUI_SHIFT	16
>  #define   PCIE_CORE_TXCREDIT_CFG1_MUI_ENCODE(x) \
>  		(((x) >> 3) << PCIE_CORE_TXCREDIT_CFG1_MUI_SHIFT)
> +#define PCIE_CORE_LANE_MAP             (PCIE_CORE_CTRL_MGMT_BASE + 0x200)
> +#define   PCIE_CORE_LANE_MAP_MASK              0x0000000f
> +#define   PCIE_CORE_LANE_MAP_REVERSE           BIT(16)
>  #define PCIE_CORE_INT_STATUS		(PCIE_CORE_CTRL_MGMT_BASE + 0x20c)
>  #define   PCIE_CORE_INT_PRFPE			BIT(0)
>  #define   PCIE_CORE_INT_CRFPE			BIT(1)
> @@ -229,6 +233,7 @@ struct rockchip_pcie {
>  	struct	regulator *vpcie0v9; /* 0.9V power supply */
>  	struct	gpio_desc *ep_gpio;
>  	u32	lanes;
> +	u8      lanes_map;
>  	u8	root_bus_nr;
>  	int	link_gen;
>  	struct	device *dev;
> @@ -301,6 +306,17 @@ static int rockchip_pcie_valid_device(struct rockchip_pcie *rockchip,
>  	return 1;
>  }
>  
> +static void rockchip_pcie_lane_map(struct rockchip_pcie *rockchip)
> +{
> +	u32 val = rockchip_pcie_read(rockchip, PCIE_CORE_LANE_MAP);
> +
> +	rockchip->lanes_map = val & PCIE_CORE_LANE_MAP_MASK;
> +
> +	/* The link may be using a reverse-indexed mapping. */
> +	if (val & PCIE_CORE_LANE_MAP_REVERSE)
> +		rockchip->lanes_map = bitrev8(rockchip->lanes_map) >> 4;

This might just be a matter of taste, but it seems to make more sense
for this function to just return a u8, and the caller can assign it to
'rockchip->lane_map'.

> +}
> +
>  static int rockchip_pcie_rd_own_conf(struct rockchip_pcie *rockchip,
>  				     int where, int size, u32 *val)
>  {
> @@ -737,6 +753,18 @@ static int rockchip_pcie_init_port(struct rockchip_pcie *rockchip)
>  			  PCIE_CORE_PL_CONF_LANE_SHIFT);
>  	dev_dbg(dev, "current link width is x%d\n", status);
>  
> +	if (!rockchip->legacy_phy) {
> +		/*  power off unused lane(s) */
> +		rockchip_pcie_lane_map(rockchip);

My above comment would make it a lot clearer that the above line...

> +		for (i = 0; i < MAX_LANE_NUM; i++) {
> +			if (rockchip->lanes_map & BIT(i))

...is determining the behavior here.

> +				continue;
> +
> +			dev_dbg(dev, "idling lane %d\n", i);
> +			phy_power_off(rockchip->phys[i]);
> +		}
> +	}
> +
>  	rockchip_pcie_write(rockchip, ROCKCHIP_VENDOR_ID,
>  			    PCIE_CORE_CONFIG_VENDOR);
>  	rockchip_pcie_write(rockchip,

...

Brian
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[RFC PATCH v3 5/7] arm64: dts: rockchip: convert PCIe to use per-lane PHYs for rk3339

[Shawn Lin]

Convert all RK3399 platforms to use per-lane PHY model in
order to save more power by idling the unused lane(s).

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Tested-by: Jeffy Chen <jeffy.chen@rock-chips.com>
---

Changes in v3: None
Changes in v2: None

 arch/arm64/boot/dts/rockchip/rk3399.dtsi | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/arch/arm64/boot/dts/rockchip/rk3399.dtsi b/arch/arm64/boot/dts/rockchip/rk3399.dtsi
index 69c56f7..5b78ce1 100644
--- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi
@@ -238,8 +238,10 @@
 		linux,pci-domain = <0>;
 		max-link-speed = <1>;
 		msi-map = <0x0 &its 0x0 0x1000>;
-		phys = <&pcie_phy>;
-		phy-names = "pcie-phy";
+		phys = <&pcie_phy 0>, <&pcie_phy 1>,
+		       <&pcie_phy 2>, <&pcie_phy 3>;
+		phy-names = "pcie-phy-0", "pcie-phy-1",
+			    "pcie-phy-2", "pcie-phy-3";
 		ranges = <0x83000000 0x0 0xfa000000 0x0 0xfa000000 0x0 0x1e00000
 			  0x81000000 0x0 0xfbe00000 0x0 0xfbe00000 0x0 0x100000>;
 		resets = <&cru SRST_PCIE_CORE>, <&cru SRST_PCIE_MGMT>,
@@ -1295,7 +1297,7 @@
 			compatible = "rockchip,rk3399-pcie-phy";
 			clocks = <&cru SCLK_PCIEPHY_REF>;
 			clock-names = "refclk";
-			#phy-cells = <0>;
+			#phy-cells = <1>;
 			resets = <&cru SRST_PCIEPHY>;
 			reset-names = "phy";
 			status = "disabled";
-- 
1.9.1

[RFC PATCH v3 6/7] dt-bindings: PCI: rockchip: convert to use per-lane PHY model

[Shawn Lin]

Deprecate legacy PHY model and encourage to use per-lane PHY
model.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Acked-by: Rob Herring <robh@kernel.org>
---

Changes in v3: None
Changes in v2: None

 .../devicetree/bindings/pci/rockchip-pcie.txt      | 25 ++++++++++++++++++++--
 1 file changed, 23 insertions(+), 2 deletions(-)

diff --git a/Documentation/devicetree/bindings/pci/rockchip-pcie.txt b/Documentation/devicetree/bindings/pci/rockchip-pcie.txt
index 1453a73..41bced4 100644
--- a/Documentation/devicetree/bindings/pci/rockchip-pcie.txt
+++ b/Documentation/devicetree/bindings/pci/rockchip-pcie.txt
@@ -19,8 +19,6 @@ Required properties:
 	- "pm"
 - msi-map: Maps a Requester ID to an MSI controller and associated
 	msi-specifier data. See ./pci-msi.txt
-- phys: From PHY bindings: Phandle for the Generic PHY for PCIe.
-- phy-names:  MUST be "pcie-phy".
 - interrupts: Three interrupt entries must be specified.
 - interrupt-names: Must include the following names
 	- "sys"
@@ -42,6 +40,18 @@ Required properties:
 	interrupt source. The value must be 1.
 - interrupt-map-mask and interrupt-map: standard PCI properties
 
+Required properties for legacy PHY model (deprecated):
+- phys: From PHY bindings: Phandle for the Generic PHY for PCIe.
+- phy-names:  MUST be "pcie-phy".
+
+Required properties for per-lane PHY model (preferred):
+- phys: Must contain an phandle to a PHY for each entry in phy-names.
+- phy-names: Must include 4 entries for all 4 lanes even if some of
+  them won't be used for your cases. Entries are of the form "pcie-phy-N":
+  where N ranges from 0 to 3.
+  (see example below and you MUST also refer to ../phy/rockchip-pcie-phy.txt
+  for changing the #phy-cells of phy node to support it)
+
 Optional Property:
 - aspm-no-l0s: RC won't support ASPM L0s. This property is needed if
 	using 24MHz OSC for RC's PHY.
@@ -95,6 +105,7 @@ pcie0: pcie@f8000000 {
 		 <&cru SRST_PCIE_PM>, <&cru SRST_P_PCIE>, <&cru SRST_A_PCIE>;
 	reset-names = "core", "mgmt", "mgmt-sticky", "pipe",
 		      "pm", "pclk", "aclk";
+	/* deprecated legacy PHY model */
 	phys = <&pcie_phy>;
 	phy-names = "pcie-phy";
 	pinctrl-names = "default";
@@ -111,3 +122,13 @@ pcie0: pcie@f8000000 {
 		#interrupt-cells = <1>;
 	};
 };
+
+pcie0: pcie@f8000000 {
+	...
+
+	/* preferred per-lane PHY model */
+	phys = <&pcie_phy 0>, <&pcie_phy 1>, <&pcie_phy 2>, <&pcie_phy 3>;
+	phy-names = "pcie-phy-0", "pcie-phy-1", "pcie-phy-2", "pcie-phy-3";
+
+	...
+};
-- 
1.9.1

[RFC PATCH v3 7/7] dt-bindings: phy: convert to use per-lane Rockchip PCIe PHY

[Shawn Lin]

This patch deprecate the legacy PCIe PHY and encourage user
to use per-lane PHY mode by setting #phy-cells to 1.

Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>

Acked-by: Rob Herring <robh@kernel.org>
---

Changes in v3:
- rename the commit tile

Changes in v2: None

 Documentation/devicetree/bindings/phy/rockchip-pcie-phy.txt | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/Documentation/devicetree/bindings/phy/rockchip-pcie-phy.txt b/Documentation/devicetree/bindings/phy/rockchip-pcie-phy.txt
index 0f6222a..b496042 100644
--- a/Documentation/devicetree/bindings/phy/rockchip-pcie-phy.txt
+++ b/Documentation/devicetree/bindings/phy/rockchip-pcie-phy.txt
@@ -3,7 +3,6 @@ Rockchip PCIE PHY
 
 Required properties:
  - compatible: rockchip,rk3399-pcie-phy
- - #phy-cells: must be 0
  - clocks: Must contain an entry in clock-names.
 	See ../clocks/clock-bindings.txt for details.
  - clock-names: Must be "refclk"
@@ -11,6 +10,12 @@ Required properties:
 	See ../reset/reset.txt for details.
  - reset-names: Must be "phy"
 
+Required properties for legacy PHY mode (deprecated):
+ - #phy-cells: must be 0
+
+Required properties for per-lane PHY mode (preferred):
+ - #phy-cells: must be 1
+
 Example:
 
 grf: syscon@ff770000 {
-- 
1.9.1