[RFC PATCH v3 0/5] Quality Improvements for Rockchip-IP PCIe

[RFC PATCH v3 0/5] Quality Improvements for Rockchip-IP PCIe

[Geraldo Nascimento]

During a 30-day debugging-run fighting quirky PCIe devices on RK3399
some quality improvements began to take form and this is my attempt
at upstreaming it. It will ensure maximum chance of retraining to Gen2
5.0GT/s, on all four lanes and plus if anybody is debugging the PHY
they'll now get real values from TEST_I[3:0] for every TEST_ADDR[4:0]
without risk of locking up kernel like with present broken async
strobe TEST_WRITE.

---
V2 -> V3: correctly clean-up with standard PCIe defines as per Bjorn's
suggestion
V1 -> V2: use standard PCIe defines as suggested by Bjorn

Geraldo Nascimento (5):
  PCI: rockchip-host: Use standard PCIe defines
  PCI: rockchip: Drop unused custom registers and bitfields
  PCI: rockchip-host: Set Target Link Speed before retraining
  phy: rockchip-pcie: Enable all four lanes
  phy: rockchip-pcie: Adjust read mask and write

 drivers/pci/controller/pcie-rockchip-host.c | 48 +++++++++++----------
 drivers/pci/controller/pcie-rockchip.h      | 11 +----
 drivers/phy/rockchip/phy-rockchip-pcie.c    | 16 ++++---
 3 files changed, 36 insertions(+), 39 deletions(-)

-- 
2.49.0

[RFC PATCH v3 1/5] PCI: rockchip-host: Use standard PCIe defines

[Geraldo Nascimento]

Current code uses custom-defined register offsets
and bitfields for standard PCIe registers. Change
to using standard PCIe defines.

Suggested-By: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Geraldo Nascimento <geraldogabriel@gmail.com>
---
 drivers/pci/controller/pcie-rockchip-host.c | 44 ++++++++++-----------
 1 file changed, 22 insertions(+), 22 deletions(-)

diff --git a/drivers/pci/controller/pcie-rockchip-host.c b/drivers/pci/controller/pcie-rockchip-host.c
index b9e7a8710cf0..65653218b9ab 100644
--- a/drivers/pci/controller/pcie-rockchip-host.c
+++ b/drivers/pci/controller/pcie-rockchip-host.c
@@ -40,18 +40,18 @@ static void rockchip_pcie_enable_bw_int(struct rockchip_pcie *rockchip)
 {
 	u32 status;
 
-	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_LCS);
+	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 	status |= (PCI_EXP_LNKCTL_LBMIE | PCI_EXP_LNKCTL_LABIE);
-	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_LCS);
+	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 }
 
 static void rockchip_pcie_clr_bw_int(struct rockchip_pcie *rockchip)
 {
 	u32 status;
 
-	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_LCS);
+	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 	status |= (PCI_EXP_LNKSTA_LBMS | PCI_EXP_LNKSTA_LABS) << 16;
-	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_LCS);
+	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 }
 
 static void rockchip_pcie_update_txcredit_mui(struct rockchip_pcie *rockchip)
@@ -269,7 +269,7 @@ static void rockchip_pcie_set_power_limit(struct rockchip_pcie *rockchip)
 	scale = 3; /* 0.001x */
 	curr = curr / 1000; /* convert to mA */
 	power = (curr * 3300) / 1000; /* milliwatt */
-	while (power > PCIE_RC_CONFIG_DCR_CSPL_LIMIT) {
+	while (power > FIELD_MAX(PCI_EXP_DEVCAP_PWR_VAL)) {
 		if (!scale) {
 			dev_warn(rockchip->dev, "invalid power supply\n");
 			return;
@@ -278,10 +278,10 @@ static void rockchip_pcie_set_power_limit(struct rockchip_pcie *rockchip)
 		power = power / 10;
 	}
 
-	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_DCR);
-	status |= (power << PCIE_RC_CONFIG_DCR_CSPL_SHIFT) |
-		  (scale << PCIE_RC_CONFIG_DCR_CPLS_SHIFT);
-	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_DCR);
+	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_DEVCAP);
+	status |= FIELD_PREP(PCI_EXP_DEVCAP_PWR_VAL, power);
+	status |= FIELD_PREP(PCI_EXP_DEVCAP_PWR_SCL, scale);
+	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_DEVCAP);
 }
 
 /**
@@ -309,14 +309,14 @@ static int rockchip_pcie_host_init_port(struct rockchip_pcie *rockchip)
 	rockchip_pcie_set_power_limit(rockchip);
 
 	/* Set RC's clock architecture as common clock */
-	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_LCS);
+	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 	status |= PCI_EXP_LNKSTA_SLC << 16;
-	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_LCS);
+	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 
 	/* Set RC's RCB to 128 */
-	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_LCS);
+	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 	status |= PCI_EXP_LNKCTL_RCB;
-	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_LCS);
+	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 
 	/* Enable Gen1 training */
 	rockchip_pcie_write(rockchip, PCIE_CLIENT_LINK_TRAIN_ENABLE,
@@ -341,9 +341,9 @@ static int rockchip_pcie_host_init_port(struct rockchip_pcie *rockchip)
 		 * Enable retrain for gen2. This should be configured only after
 		 * gen1 finished.
 		 */
-		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_LCS);
+		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 		status |= PCI_EXP_LNKCTL_RL;
-		rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_LCS);
+		rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 
 		err = readl_poll_timeout(rockchip->apb_base + PCIE_CORE_CTRL,
 					 status, PCIE_LINK_IS_GEN2(status), 20,
@@ -380,15 +380,15 @@ static int rockchip_pcie_host_init_port(struct rockchip_pcie *rockchip)
 
 	/* Clear L0s from RC's link cap */
 	if (of_property_read_bool(dev->of_node, "aspm-no-l0s")) {
-		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_LINK_CAP);
-		status &= ~PCIE_RC_CONFIG_LINK_CAP_L0S;
-		rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_LINK_CAP);
+		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCAP);
+		status &= ~PCI_EXP_LNKCAP_ASPM_L0S;
+		rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCAP);
 	}
 
-	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_DCSR);
-	status &= ~PCIE_RC_CONFIG_DCSR_MPS_MASK;
-	status |= PCIE_RC_CONFIG_DCSR_MPS_256;
-	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_DCSR);
+	status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_DEVCTL);
+	status &= ~PCI_EXP_DEVCTL_PAYLOAD;
+	status |= PCI_EXP_DEVCTL_PAYLOAD_256B;
+	rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_DEVCTL);
 
 	return 0;
 err_power_off_phy:
-- 
2.49.0

[RFC PATCH v3 2/5] PCI: rockchip: Drop unused custom registers and bitfields

[Geraldo Nascimento]

Since we are now using standard PCIe defines, drop
unused custom-defined ones and instead reference
them by offset from Capabilities Register.

Suggested-By: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Geraldo Nascimento <geraldogabriel@gmail.com>
---
 drivers/pci/controller/pcie-rockchip.h | 11 +----------
 1 file changed, 1 insertion(+), 10 deletions(-)

diff --git a/drivers/pci/controller/pcie-rockchip.h b/drivers/pci/controller/pcie-rockchip.h
index 5864a20323f2..f611599988d7 100644
--- a/drivers/pci/controller/pcie-rockchip.h
+++ b/drivers/pci/controller/pcie-rockchip.h
@@ -155,16 +155,7 @@
 #define PCIE_EP_CONFIG_DID_VID		(PCIE_EP_CONFIG_BASE + 0x00)
 #define PCIE_EP_CONFIG_LCS		(PCIE_EP_CONFIG_BASE + 0xd0)
 #define PCIE_RC_CONFIG_RID_CCR		(PCIE_RC_CONFIG_BASE + 0x08)
-#define PCIE_RC_CONFIG_DCR		(PCIE_RC_CONFIG_BASE + 0xc4)
-#define   PCIE_RC_CONFIG_DCR_CSPL_SHIFT		18
-#define   PCIE_RC_CONFIG_DCR_CSPL_LIMIT		0xff
-#define   PCIE_RC_CONFIG_DCR_CPLS_SHIFT		26
-#define PCIE_RC_CONFIG_DCSR		(PCIE_RC_CONFIG_BASE + 0xc8)
-#define   PCIE_RC_CONFIG_DCSR_MPS_MASK		GENMASK(7, 5)
-#define   PCIE_RC_CONFIG_DCSR_MPS_256		(0x1 << 5)
-#define PCIE_RC_CONFIG_LINK_CAP		(PCIE_RC_CONFIG_BASE + 0xcc)
-#define   PCIE_RC_CONFIG_LINK_CAP_L0S		BIT(10)
-#define PCIE_RC_CONFIG_LCS		(PCIE_RC_CONFIG_BASE + 0xd0)
+#define PCIE_RC_CONFIG_CR		(PCIE_RC_CONFIG_BASE + 0xc0)
 #define PCIE_EP_CONFIG_LCS		(PCIE_EP_CONFIG_BASE + 0xd0)
 #define PCIE_RC_CONFIG_L1_SUBSTATE_CTRL2 (PCIE_RC_CONFIG_BASE + 0x90c)
 #define PCIE_RC_CONFIG_THP_CAP		(PCIE_RC_CONFIG_BASE + 0x274)
-- 
2.49.0

[RFC PATCH v3 3/5] PCI: rockchip-host: Set Target Link Speed before retraining

[Geraldo Nascimento]

Current code may fail Gen2 retraining if Target Link Speed
is set to 2.5 GT/s in Link Control and Status Register 2.
Set it to 5.0 GT/s accordingly.

Signed-off-by: Geraldo Nascimento <geraldogabriel@gmail.com>
---
 drivers/pci/controller/pcie-rockchip-host.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/pci/controller/pcie-rockchip-host.c b/drivers/pci/controller/pcie-rockchip-host.c
index 65653218b9ab..68634ae8caaf 100644
--- a/drivers/pci/controller/pcie-rockchip-host.c
+++ b/drivers/pci/controller/pcie-rockchip-host.c
@@ -341,6 +341,10 @@ static int rockchip_pcie_host_init_port(struct rockchip_pcie *rockchip)
 		 * Enable retrain for gen2. This should be configured only after
 		 * gen1 finished.
 		 */
+		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL2);
+		status |= FIELD_PREP(PCI_EXP_LNKCTL2_TLS, PCI_EXP_LNKCTL2_TLS_5_0GT);
+		rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL2);
+
 		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
 		status |= PCI_EXP_LNKCTL_RL;
 		rockchip_pcie_write(rockchip, status, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL);
-- 
2.49.0

[RFC PATCH v3 4/5] phy: rockchip-pcie: Enable all four lanes

[Geraldo Nascimento]

Current code enables only Lane 0 because pwr_cnt will be incremented
on first call to the function. Use for-loop to enable all 4 lanes
through GRF.

Signed-off-by: Geraldo Nascimento <geraldogabriel@gmail.com>
---
 drivers/phy/rockchip/phy-rockchip-pcie.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/drivers/phy/rockchip/phy-rockchip-pcie.c b/drivers/phy/rockchip/phy-rockchip-pcie.c
index bd44af36c67a..48bcc7d2b33b 100644
--- a/drivers/phy/rockchip/phy-rockchip-pcie.c
+++ b/drivers/phy/rockchip/phy-rockchip-pcie.c
@@ -176,11 +176,13 @@ 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));
+	for (int i=0; i < PHY_MAX_LANE_NUM; i++) {
+		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 + i));
+	}
 
 	/*
 	 * No documented timeout value for phy operation below,
-- 
2.49.0

[RFC PATCH v3 5/5] phy: rockchip-pcie: Adjust read mask and write

[Geraldo Nascimento]

Section 17.6.10 of the RK3399 TRM "PCIe PIPE PHY registers Description"
defines asynchronous strobe TEST_WRITE which should be enabled then
disabled and seems to have been copy-pasted as of current. Adjust it.
While at it, adjust read mask which should be the same as write mask.

Signed-off-by: Geraldo Nascimento <geraldogabriel@gmail.com>
---
 drivers/phy/rockchip/phy-rockchip-pcie.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/phy/rockchip/phy-rockchip-pcie.c b/drivers/phy/rockchip/phy-rockchip-pcie.c
index 48bcc7d2b33b..35d2523ee776 100644
--- a/drivers/phy/rockchip/phy-rockchip-pcie.c
+++ b/drivers/phy/rockchip/phy-rockchip-pcie.c
@@ -30,9 +30,9 @@
 #define PHY_CFG_ADDR_SHIFT    1
 #define PHY_CFG_DATA_MASK     0xf
 #define PHY_CFG_ADDR_MASK     0x3f
-#define PHY_CFG_RD_MASK       0x3ff
+#define PHY_CFG_RD_MASK       0x3f
 #define PHY_CFG_WR_ENABLE     1
-#define PHY_CFG_WR_DISABLE    1
+#define PHY_CFG_WR_DISABLE    0
 #define PHY_CFG_WR_SHIFT      0
 #define PHY_CFG_WR_MASK       1
 #define PHY_CFG_PLL_LOCK      0x10
-- 
2.49.0

Re: [RFC PATCH v3 3/5] PCI: rockchip-host: Set Target Link Speed before retraining

[Geraldo Nascimento]

On Fri, Jun 13, 2025 at 02:21:59AM -0300, Geraldo Nascimento wrote:
> Current code may fail Gen2 retraining if Target Link Speed
> is set to 2.5 GT/s in Link Control and Status Register 2.
> Set it to 5.0 GT/s accordingly.
> 
> Signed-off-by: Geraldo Nascimento <geraldogabriel@gmail.com>
> ---
>  drivers/pci/controller/pcie-rockchip-host.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/pci/controller/pcie-rockchip-host.c b/drivers/pci/controller/pcie-rockchip-host.c
> index 65653218b9ab..68634ae8caaf 100644
> --- a/drivers/pci/controller/pcie-rockchip-host.c
> +++ b/drivers/pci/controller/pcie-rockchip-host.c
> @@ -341,6 +341,10 @@ static int rockchip_pcie_host_init_port(struct rockchip_pcie *rockchip)
>  		 * Enable retrain for gen2. This should be configured only after
>  		 * gen1 finished.
>  		 */
> +		status = rockchip_pcie_read(rockchip, PCIE_RC_CONFIG_CR + PCI_EXP_LNKCTL2);
> +		status |= FIELD_PREP(PCI_EXP_LNKCTL2_TLS, PCI_EXP_LNKCTL2_TLS_5_0GT);

Although this incidentally "works" for facilitating Gen2 training
because mask clears 2.5 GT/s, this will shift the value 0x2
by the mask I think. This is clearly wrong, and instead
we should use AND clearing plus OR'ing the 5.0GT/s value.

But I wait until at least Bjorn's review to send v4 with
correction.

Geraldo Nascimento