[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <1364461215-7793-3-git-send-email-jeffrey.t.kirsher@intel.com>
Date: Thu, 28 Mar 2013 02:00:05 -0700
From: Jeff Kirsher <jeffrey.t.kirsher@...el.com>
To: davem@...emloft.net
Cc: Bruce Allan <bruce.w.allan@...el.com>, netdev@...r.kernel.org,
gospo@...hat.com, sassmann@...hat.com,
Jeff Kirsher <jeffrey.t.kirsher@...el.com>
Subject: [net-next 02/12] e1000e: slow performance between two 82579 connected via 10Mbit hub
From: Bruce Allan <bruce.w.allan@...el.com>
Two 82579 LOMs connected via a 10Mb hub experience extraordinarily low
performance. This is because 82579 is excessively aggressive on transmit
at 10Mb half-duplex and will not provide sufficient time for the link
partner to transmit. When the link partner is also 82579, the result is a
lot of collisions (and corresponding re-transmits) that cause the poor
performance. To work-around this issue, significantly increase the IPG in
the MAC to allow enough gap for the link partner to transmit and reduce the
Rx latency in the analog PHY to 0 to reduce the number of collisions.
Signed-off-by: Bruce Allan <bruce.w.allan@...el.com>
Tested-by: Jeff Pieper <jeffrey.e.pieper@...el.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@...el.com>
---
drivers/net/ethernet/intel/e1000e/defines.h | 1 +
drivers/net/ethernet/intel/e1000e/ich8lan.c | 28 ++++++++++++++++++++++++++++
drivers/net/ethernet/intel/e1000e/ich8lan.h | 1 +
3 files changed, 30 insertions(+)
diff --git a/drivers/net/ethernet/intel/e1000e/defines.h b/drivers/net/ethernet/intel/e1000e/defines.h
index a6aeb19..351c94a 100644
--- a/drivers/net/ethernet/intel/e1000e/defines.h
+++ b/drivers/net/ethernet/intel/e1000e/defines.h
@@ -236,6 +236,7 @@
#define E1000_STATUS_FUNC_SHIFT 2
#define E1000_STATUS_FUNC_1 0x00000004 /* Function 1 */
#define E1000_STATUS_TXOFF 0x00000010 /* transmission paused */
+#define E1000_STATUS_SPEED_MASK 0x000000C0
#define E1000_STATUS_SPEED_10 0x00000000 /* Speed 10Mb/s */
#define E1000_STATUS_SPEED_100 0x00000040 /* Speed 100Mb/s */
#define E1000_STATUS_SPEED_1000 0x00000080 /* Speed 1000Mb/s */
diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.c b/drivers/net/ethernet/intel/e1000e/ich8lan.c
index 1cdec5f..4f2f0f6 100644
--- a/drivers/net/ethernet/intel/e1000e/ich8lan.c
+++ b/drivers/net/ethernet/intel/e1000e/ich8lan.c
@@ -871,6 +871,34 @@ static s32 e1000_check_for_copper_link_ich8lan(struct e1000_hw *hw)
return ret_val;
}
+ /* When connected at 10Mbps half-duplex, 82579 parts are excessively
+ * aggressive resulting in many collisions. To avoid this, increase
+ * the IPG and reduce Rx latency in the PHY.
+ */
+ if ((hw->mac.type == e1000_pch2lan) && link) {
+ u32 reg;
+ reg = er32(STATUS);
+ if (!(reg & (E1000_STATUS_FD | E1000_STATUS_SPEED_MASK))) {
+ reg = er32(TIPG);
+ reg &= ~E1000_TIPG_IPGT_MASK;
+ reg |= 0xFF;
+ ew32(TIPG, reg);
+
+ /* Reduce Rx latency in analog PHY */
+ ret_val = hw->phy.ops.acquire(hw);
+ if (ret_val)
+ return ret_val;
+
+ ret_val =
+ e1000_write_emi_reg_locked(hw, I82579_RX_CONFIG, 0);
+
+ hw->phy.ops.release(hw);
+
+ if (ret_val)
+ return ret_val;
+ }
+ }
+
/* Work-around I218 hang issue */
if ((hw->adapter->pdev->device == E1000_DEV_ID_PCH_LPTLP_I218_LM) ||
(hw->adapter->pdev->device == E1000_DEV_ID_PCH_LPTLP_I218_V)) {
diff --git a/drivers/net/ethernet/intel/e1000e/ich8lan.h b/drivers/net/ethernet/intel/e1000e/ich8lan.h
index 8bf4655..00ba6c9 100644
--- a/drivers/net/ethernet/intel/e1000e/ich8lan.h
+++ b/drivers/net/ethernet/intel/e1000e/ich8lan.h
@@ -211,6 +211,7 @@
#define I82579_MSE_THRESHOLD 0x084F /* 82579 Mean Square Error Threshold */
#define I82577_MSE_THRESHOLD 0x0887 /* 82577 Mean Square Error Threshold */
#define I82579_MSE_LINK_DOWN 0x2411 /* MSE count before dropping link */
+#define I82579_RX_CONFIG 0x3412 /* Receive configuration */
#define I82579_EEE_PCS_STATUS 0x182D /* IEEE MMD Register 3.1 >> 8 */
#define I82579_EEE_CAPABILITY 0x0410 /* IEEE MMD Register 3.20 */
#define I82579_EEE_ADVERTISEMENT 0x040E /* IEEE MMD Register 7.60 */
--
1.7.11.7
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists