Patchwork NET/PHY: Eliminate the forced speed reduction algorithm

login
register
mail settings
Submitter Kirill Kapranov
Date Feb. 25, 2013, 12:25 p.m.
Message ID <1361795118.3980.108.camel@kkk.nita.ru>
Download mbox | patch
Permalink /patch/222914/
State Changes Requested
Delegated to: David Miller
Headers show

Comments

Kirill Kapranov - Feb. 25, 2013, 12:25 p.m.
From Kirill Kapranov  <kkk@nita.ru>,<kapranoff@inbox.ru>

NET/PHY: Eliminate the forced speed reduction algorithm.
In case of fixed speed set up for a NIC (e.g. ethtool -s eth0 autoneg off speed 100 duplex full) with an ethernet cable plugged off, the mentioned algorithm slows down a NIC speed, so further hooking up a cable does not lead to "link" state.
Signed-off-by: Kirill Kapranov <kkk@nita.ru>,<kapranoff@inbox.ru>
---
The purpose of the introduced patch is deletion of the forced speed reduction algorithm realization from the driver module "phy".
The above mentioned algorithm works in the following way: if the PHY detected unlink line state (connector plugged off), NIC speed is decreased step-by-step in the sequence:
100 full duplex
100 half duplex
10 full duplex
10 half duplex
with the latency circa 10 s per step, and stops at 10-HD value.
I have looked up RFC-802.3, and found, that the mentioned algorithm is neither quoted nor described. AFAIK, no one RFC describe the mentioned algorithm, so it may be a witty invention of the developer(s).

In the case of the fixed speed and duplex set, with the autonegotiation off, for a NIC (e.g. # ethtool -s eth0 autoneg off speed 100 duplex full) with ethernet cable plugged off, mentioned algorithm slows down NIC speed, so when ethernet connector is plugged in, connection will be inoperative: an ethernet switch will try to connect with 100/full (e.g.), a NIC will stay at 10/half.
Thus, this algorithm is destructive for the fixed speed/duplex mode (with autonegotiation off).

In the AUTO mode, the mentioned algorithm is inessential. The autonegotiation procedure works fine regardless an speed/duplex settings at the moment of connector hooking up.
Thus, there is no point in using of this algorithm in driver.

Tested at 2.6.38.7, applicable up to for 3.0.4. 



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller - Feb. 25, 2013, 8:59 p.m.
From: Kirill Kapranov <kapranoff@inbox.ru>
Date: Mon, 25 Feb 2013 16:25:18 +0400


To be completely honest with you, I'm really getting tired of
making corrections to this one simple patch submission.

> From Kirill Kapranov  <kkk@nita.ru>,<kapranoff@inbox.ru>

It makes no sense to mention multiple email addresses in your
authorship line, reduce it to one.

If you cannot do things like this correctly on your own, have
automated tools (such as git) generate these patch emails for you.

> NET/PHY: Eliminate the forced speed reduction algorithm.

You do not need to mention the commit header line again, all
this does is make more work for me as I have to edit it out.
To one in the Subject line is sufficient, and you should not
use all-CAPS, and also the subsystem indication is not correct.
Your subject line should be something like:

"[PATCH] phy: Eliminate the forced speed reduction algorithm"

That is, remove the "NET/" part, leave "PHY" and make it all
lowercase.

> In case of fixed speed set up for a NIC (e.g. ethtool -s eth0 autoneg off speed 100 duplex full) with an ethernet cable plugged off, the mentioned algorithm slows down a NIC speed, so further hooking up a cable does not lead to "link" state.

I also had to formate this paragraph down to 80 column lines instead
of one long one.

> Signed-off-by: Kirill Kapranov <kkk@nita.ru>,<kapranoff@inbox.ru>

Again, using two email address is not correct, just use one.

Please, do yourself a huge favor, look at how other people submit
accepted patches on this list.  Learn by example rather than making
many mistakes trying to figure things out purely on your own.

> I have looked up RFC-802.3, and found, that the mentioned algorithm is neither quoted nor described. AFAIK, no one RFC describe the mentioned algorithm, so it may be a witty invention of the developer(s).

At the time that autonegiation was a new or non-existing feature, this
approach of stepping down the link parameters trying different
settings one-by-one was an absolute necessity.  It's probably not
needed anymore in modern times.

> Tested at 2.6.38.7, applicable up to for 3.0.4. 

This patch does not apply to current upstream sources at all.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

--- linux/drivers/net/phy/phy.c.orig	2011-05-22 02:13:59.000000000 +0400
+++ linux/drivers/net/phy/phy.c	2012-04-28 12:49:37.000000000 +0400
@@ -457,34 +457,6 @@  void phy_stop_machine(struct phy_device
 }
 
 /**
- * phy_force_reduction - reduce PHY speed/duplex settings by one step
- * @phydev: target phy_device struct
- *
- * Description: Reduces the speed/duplex settings by one notch,
- *   in this order--
- *   1000/FULL, 1000/HALF, 100/FULL, 100/HALF, 10/FULL, 10/HALF.
- *   The function bottoms out at 10/HALF.
- */
-static void phy_force_reduction(struct phy_device *phydev)
-{
-	int idx;
-
-	idx = phy_find_setting(phydev->speed, phydev->duplex);
-	
-	idx++;
-
-	idx = phy_find_valid(idx, phydev->supported);
-
-	phydev->speed = settings[idx].speed;
-	phydev->duplex = settings[idx].duplex;
-
-	pr_info("Trying %d/%s\n", phydev->speed,
-			DUPLEX_FULL == phydev->duplex ?
-			"FULL" : "HALF");
-}
-
-
-/**
  * phy_error - enter HALTED state for this PHY device
  * @phydev: target phy_device struct
  *
@@ -814,30 +786,12 @@  void phy_state_machine(struct work_struc
 				phydev->adjust_link(phydev->attached_dev);
 
 			} else if (0 == phydev->link_timeout--) {
-				int idx;
 
 				needs_aneg = 1;
 				/* If we have the magic_aneg bit,
 				 * we try again */
 				if (phydev->drv->flags & PHY_HAS_MAGICANEG)
 					break;
-
-				/* The timer expired, and we still
-				 * don't have a setting, so we try
-				 * forcing it until we find one that
-				 * works, starting from the fastest speed,
-				 * and working our way down */
-				idx = phy_find_valid(0, phydev->supported);
-
-				phydev->speed = settings[idx].speed;
-				phydev->duplex = settings[idx].duplex;
-
-				phydev->autoneg = AUTONEG_DISABLE;
-
-				pr_info("Trying %d/%s\n", phydev->speed,
-						DUPLEX_FULL ==
-						phydev->duplex ?
-						"FULL" : "HALF");
 			}
 			break;
 		case PHY_NOLINK:
@@ -863,7 +817,6 @@  void phy_state_machine(struct work_struc
 				netif_carrier_on(phydev->attached_dev);
 			} else {
 				if (0 == phydev->link_timeout--) {
-					phy_force_reduction(phydev);
 					needs_aneg = 1;
 				}
 			}