[<prev] [next>] [day] [month] [year] [list]
Message-Id: <20081119104550ohashi-h@mail.jp.nec.com>
Date: Wed, 19 Nov 2008 10:45:50 +0900
From: ohashi-h@...dnes.nec.co.jp
To: netdev@...r.kernel.org
Subject: e1000 driver: LED indicates trouble
From: Hiroki Ohashi <ohashi-h@...dnes.nec.co.jp>
Hi all,
I faced the following situations.
The LED of NIC was indicated, even when the system was not linked
the network (auto-negotiation failed or link partner cannot
auto-negotiate).
LED should be indicate only when the network was linked.
And so, it is a problem that LED indicates.
We are using 82571EB ethernet controler and serdes interface.
I think this problem is caused by the e1000 device driver
setting LU and SLU bits in the interface and forcing the link
to go up.
I deleted this logic. (the following modification.)
Then, LED was turned off.
================================
--- e1000_hw.c.org
+++ e1000_hw.c
@@ -3069,6 +3069,7 @@
hw->autoneg_failed = 1;
return 0;
}
+#if 0
DEBUGOUT("NOT RXing /C/, disable AutoNeg and force link.\n");
/* Disable auto-negotiation in the TXCW register */
@@ -3078,6 +3079,7 @@
ctrl = E1000_READ_REG(hw, CTRL);
ctrl |= (E1000_CTRL_SLU | E1000_CTRL_FD);
E1000_WRITE_REG(hw, CTRL, ctrl);
+#endif
/* Configure Flow Control after forcing link up. */
ret_val = e1000_config_fc_after_link_up(hw);
================================
I think the e1000 driver should enable/disable force-link-up
depending on the condition of the link.
But I don't know a good way to do this.
If you have any good ideas, please let me know.
As a temporary idea, I thought of a way of adding a module parameter
to control enable/disable force-link-up.
The patch is as follows.
================================
--- drivers/net/e1000/e1000_hw.h.org
+++ drivers/net/e1000/e1000_hw.h
@@ -1459,6 +1459,7 @@ struct e1000_hw {
boolean_t leave_av_bit_off;
boolean_t kmrn_lock_loss_workaround_disabled;
boolean_t rx_needs_kicking;
+ boolean_t disable_force_link_up_when_autoneg_failed;
};
--- drivers/net/e1000/e1000_param.c.org
+++ drivers/net/e1000/e1000_param.c
@@ -218,6 +218,13 @@ E1000_PARAM(SmartPowerDownEnable, "Enabl
*/
E1000_PARAM(KumeranLockLoss, "Enable Kumeran lock loss workaround");
+/* Enable Force Link Up When AutoNeg Failed
+ *
+ * Valid Range: 0, 1
+ *
+ * Default Value: 1 (force link up enabled) */
+E1000_PARAM(EnableForceLinkOnSERDES, "Enable force link up when AutoNeg failed");
+
struct e1000_option {
enum { enable_option, range_option, list_option } type;
char *name;
@@ -596,6 +603,28 @@ e1000_check_options(struct e1000_adapter
}
#endif
}
+ { /* Enable Force Link Up When AutoNeg Failed */
+ struct e1000_option opt = {
+ .type = enable_option,
+ .name = "Enable Force Link Up When AutoNeg Failed",
+ .err = "defaulting to Enabled",
+ .def = OPTION_ENABLED
+ };
+
+ if (num_EnableForceLinkOnSERDES > bd) {
+ unsigned int force_link_up = EnableForceLinkOnSERDES [bd];
+ e1000_validate_option(&force_link_up, &opt, adapter);
+ adapter- >hw.disable_force_link_up_when_autoneg_failed = !force_link_up;
+
+ } else {
+ adapter->hw.disable_force_link_up_when_autoneg_failed = !opt.def;
+ }
+
+ }
switch (adapter->hw.media_type) {
case e1000_media_type_fiber:
--- drivers/net/e1000/e1000_hw.c.org
+++ drivers/net/e1000/e1000_hw.c
@@ -3057,7 +3057,16 @@ e1000_check_for_link(struct e1000_hw *hw
hw->autoneg_failed = 1;
return 0;
}
+
+ if (hw->disable_force_link_up_when_autoneg_failed) {
+ DEBUGOUT("Force Link Up Disabled\n");
+ return -E1000_ERR_PHY;
+ }
+else {
DEBUGOUT("NOT RXing /C/, disable AutoNeg and force link.\n");
/* Disable auto-negotiation in the TXCW register */
E1000_WRITE_REG(hw, TXCW, (hw->txcw & ~E1000_TXCW_ANE));
@@ -3066,7 +3075,7 @@ e1000_check_for_link(struct e1000_hw *hw
ctrl = E1000_READ_REG(hw, CTRL);
ctrl |= (E1000_CTRL_SLU | E1000_CTRL_FD);
E1000_WRITE_REG(hw, CTRL, ctrl);
-
+}
/* Configure Flow Control after forcing link up. */
ret_val = e1000_config_fc_after_link_up(hw);
if (ret_val) {
================================
--
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