[PATCH v1 1/2] net/e1000: better clock synchronization workaround

[Anatoly Burakov]

From: Vitaly Lifshits <vitaly.lifshits at intel.com>

There was a workaround for clock synchronization issue already in place,
however this workaround has a high power penalty. Fix the issue by
reconfiguring the k1 exit timeout, thereby avoiding the side effect of high
power draw.

Fixes: 90f456cea0c2 ("net/e1000/base: correct clock synchronization")

Signed-off-by: Vitaly Lifshits <vitaly.lifshits at intel.com>
Signed-off-by: Anatoly Burakov <anatoly.burakov at intel.com>
---
 drivers/net/intel/e1000/base/e1000_ich8lan.c | 82 +++++++++++++++++---
 drivers/net/intel/e1000/base/e1000_ich8lan.h |  4 +
 2 files changed, 74 insertions(+), 12 deletions(-)

diff --git a/drivers/net/intel/e1000/base/e1000_ich8lan.c b/drivers/net/intel/e1000/base/e1000_ich8lan.c
index 6b7eb245a6..ece61650f7 100644
--- a/drivers/net/intel/e1000/base/e1000_ich8lan.c
+++ b/drivers/net/intel/e1000/base/e1000_ich8lan.c
@@ -284,6 +284,43 @@ STATIC void e1000_toggle_lanphypc_pch_lpt(struct e1000_hw *hw)
 	}
 }
 
+/**
+ * e1000_reconfigure_k1_exit_timeout
+ * @hw: pointer to the HW structure
+ *
+ * Reconfigure K1 exit timeout as a workaround to the PHY sycndhronization issue
+ * on MTL, LNL, PTL and WCL.
+ *
+ * Assuming PHY semaphore is taken prior to this function call.
+ **/
+STATIC s32 e1000_reconfigure_k1_exit_timeout(struct e1000_hw *hw)
+{
+	s32 ret_val = E1000_SUCCESS;
+	u32 fextnvm12;
+	u16 phy_timeout;
+
+	DEBUGFUNC("e1000_reconfigure_k1_exit_timeout");
+
+	if (hw->mac.type < e1000_pch_mtp)
+		return E1000_SUCCESS;
+
+	fextnvm12 = E1000_READ_REG(hw, E1000_FEXTNVM12);
+	fextnvm12 |= (1 << 23);
+	fextnvm12 &= ~((1 << 22));
+	E1000_WRITE_REG(hw, E1000_FEXTNVM12, fextnvm12);
+
+	msec_delay_irq(1);
+
+	ret_val = hw->phy.ops.read_reg_locked(hw, E1000_PHY_TIMEOUTS_REG,
+					      &phy_timeout);
+	phy_timeout &= ~E1000_PHY_TIMEOUTS_K1_EXIT_TO_MASK;
+	phy_timeout |= 0xF00;
+	ret_val = hw->phy.ops.write_reg_locked(hw, E1000_PHY_TIMEOUTS_REG,
+					       phy_timeout);
+
+	return ret_val;
+}
+
 /**
  *  e1000_init_phy_workarounds_pchlan - PHY initialization workarounds
  *  @hw: pointer to the HW structure
@@ -332,6 +369,14 @@ STATIC s32 e1000_init_phy_workarounds_pchlan(struct e1000_hw *hw)
 	case e1000_pch_adp:
 	case e1000_pch_mtp:
 	case e1000_pch_ptp:
+		/* Due to clock synchronization issue on MTL and above prior to
+		 * disabling k1 it is required to disable P0s state and
+		 * reconfigure PHY k1 exit timeout. At this point the PHY might
+		 * be inaccessible so don't propagate the failure.
+		 */
+		if (hw->mac.type >= e1000_pch_mtp)
+			e1000_reconfigure_k1_exit_timeout(hw);
+
 		if (e1000_phy_is_accessible_pchlan(hw))
 			break;
 
@@ -417,6 +462,16 @@ STATIC s32 e1000_init_phy_workarounds_pchlan(struct e1000_hw *hw)
 		ret_val = hw->phy.ops.check_reset_block(hw);
 		if (ret_val)
 			ERROR_REPORT("ME blocked access to PHY after reset\n");
+
+		if (hw->mac.type >= e1000_pch_mtp) {
+			ret_val = hw->phy.ops.acquire(hw);
+			if (ret_val) {
+				DEBUGOUT("Failed to reconfigure K1 exit timeout\n");
+				goto out;
+			}
+			ret_val = e1000_reconfigure_k1_exit_timeout(hw);
+			hw->phy.ops.release(hw);
+		}
 	}
 
 out:
@@ -2463,18 +2518,6 @@ s32 e1000_configure_k1_ich8lan(struct e1000_hw *hw, bool k1_enable)
 
 	DEBUGFUNC("e1000_configure_k1_ich8lan");
 
-	/* Due to clock synchronization issue on MTL and above prior to
-	 * disabling k1 it is required to disable P0s state
-	 */
-	if ((!k1_enable) && (hw->mac.type >= e1000_pch_mtp)) {
-		u32 fextnvm12 = E1000_READ_REG(hw, E1000_FEXTNVM12);
-		fextnvm12 |= (1 << 23);
-		fextnvm12 &= ~((1 << 22));
-		E1000_WRITE_REG(hw, E1000_FEXTNVM12, fextnvm12);
-
-		usec_delay(100);
-	}
-
 	ret_val = e1000_read_kmrn_reg_locked(hw, E1000_KMRNCTRLSTA_K1_CONFIG,
 					     &kmrn_reg);
 	if (ret_val)
@@ -5086,6 +5129,21 @@ STATIC s32 e1000_init_hw_ich8lan(struct e1000_hw *hw)
 
 	e1000_initialize_hw_bits_ich8lan(hw);
 
+
+	if (hw->mac.type >= e1000_pch_mtp) {
+		ret_val = hw->phy.ops.acquire(hw);
+		if (ret_val)
+			return ret_val;
+
+		ret_val = e1000_reconfigure_k1_exit_timeout(hw);
+		hw->phy.ops.release(hw);
+
+		if (ret_val) {
+			DEBUGOUT("Error reconfiguring PHY K1 exit timeout\n");
+			return ret_val;
+		}
+	}
+
 	/* Initialize identification LED */
 	ret_val = mac->ops.id_led_init(hw);
 	/* An error is not fatal and we should not stop init due to this */
diff --git a/drivers/net/intel/e1000/base/e1000_ich8lan.h b/drivers/net/intel/e1000/base/e1000_ich8lan.h
index fbe118723c..f2ba910ea6 100644
--- a/drivers/net/intel/e1000/base/e1000_ich8lan.h
+++ b/drivers/net/intel/e1000/base/e1000_ich8lan.h
@@ -218,6 +218,10 @@
 #define I217_PLL_CLOCK_GATE_REG	PHY_REG(772, 28)
 #define I217_PLL_CLOCK_GATE_MASK	0x07FF
 
+/* PHY Timeouts */
+#define E1000_PHY_TIMEOUTS_REG		PHY_REG(770, 21)
+#define E1000_PHY_TIMEOUTS_K1_EXIT_TO_MASK	0x0FC0
+
 #define SW_FLAG_TIMEOUT		1000 /* SW Semaphore flag timeout in ms */
 
 /* Inband Control */
-- 
2.47.1