Skip to contentSkip to conversation

net: phy: improve genphy_read_status

PATCH Applied Heiner Kallweit 3 participants 8 messages Mar 30, 2019 – Apr 2, 2019

Subsystems (3)
lore
Share & tools
lore

Heiner Kallweit

author maintainer
#0 lore raw

PATCH 1/1 net: phy: improve genphy_read_status

This patch improves few aspects of genphy_read_status():

- Don't initialize lpagb, it's not needed.

- Move initializing phydev->speed et al before the if clause.

- In auto-neg case, skip populating lp_advertising if we don't have a link. This avoids quite some unnecessary MDIO reads in case of phylib polling mode.

---
 drivers/net/phy/phy_device.c | 23 ++++++++---------------
 1 file changed, 8 insertions(+), 15 deletions(-)
drivers/net/phy/phy_device.c +8 -15
index 0c4cc9b54..67fc581e3 100644
@@ -1743,19 +1743,21 @@ EXPORT_SYMBOL(genphy_update_link);
  */
 int genphy_read_status(struct phy_device *phydev)
 {
-	int adv;
-	int err;
-	int lpa;
-	int lpagb = 0;
+	int adv, lpa, lpagb, err;
 
 	/* Update the link, but return if there was an error */
 	err = genphy_update_link(phydev);
 	if (err)
 		return err;
 
+	phydev->speed = SPEED_UNKNOWN;
+	phydev->duplex = DUPLEX_UNKNOWN;
+	phydev->pause = 0;
+	phydev->asym_pause = 0;
+
 	linkmode_zero(phydev->lp_advertising);
 
-	if (AUTONEG_ENABLE == phydev->autoneg) {
+	if (phydev->autoneg == AUTONEG_ENABLE && phydev->link) {
 		if (linkmode_test_bit(ETHTOOL_LINK_MODE_1000baseT_Half_BIT,
 				      phydev->supported) ||
 		    linkmode_test_bit(ETHTOOL_LINK_MODE_1000baseT_Full_BIT,
@@ -1785,14 +1787,8 @@ int genphy_read_status(struct phy_device *phydev)
 			return lpa;
 
 		mii_lpa_mod_linkmode_lpa_t(phydev->lp_advertising, lpa);
-
-		phydev->speed = SPEED_UNKNOWN;
-		phydev->duplex = DUPLEX_UNKNOWN;
-		phydev->pause = 0;
-		phydev->asym_pause = 0;
-
 		phy_resolve_aneg_linkmode(phydev);
-	} else {
+	} else if (phydev->autoneg == AUTONEG_DISABLE) {
 		int bmcr = phy_read(phydev, MII_BMCR);
 
 		if (bmcr < 0)
@@ -1809,9 +1805,6 @@ int genphy_read_status(struct phy_device *phydev)
 			phydev->speed = SPEED_100;
 		else
 			phydev->speed = SPEED_10;
-
-		phydev->pause = 0;
-		phydev->asym_pause = 0;
 	}
 
 	return 0;
Signature · 1 line
2.21.0
maintainer
#1 lore raw

drivers/net/phy/phy_device.c EXPORT_SYMBOL(genphy_update_link); · line 1760

Quoted 2 lines
drivers/net/phy/phy_device.c +1 -1
-	if (AUTONEG_ENABLE == phydev->autoneg) {
+	if (phydev->autoneg == AUTONEG_ENABLE && phydev->link) {

Hi Heiner

I don't necessary agree with placing the constant first in the comparison, but it is best practice not to change it when making additions. It makes it a little bit harder to see what the actual change was.

Andrew

author maintainer
#2 lore raw
Quoted 10 lines · Andrew Lunn

drivers/net/phy/phy_device.c EXPORT_SYMBOL(genphy_update_link); · line 1760

Quoted 2 lines
drivers/net/phy/phy_device.c +1 -1
-	if (AUTONEG_ENABLE == phydev->autoneg) {
+	if (phydev->autoneg == AUTONEG_ENABLE && phydev->link) {

Hi Heiner

I don't necessary agree with placing the constant first in the comparison, but it is best practice not to change it when making additions. It makes it a little bit harder to see what the actual change was.

I understand the point. However a patch to only change the order of the operands most likely would also be rejected as not being worth it. As a consequence we would have to live with it forever. So I think it needs some "if we touch the code anyway" situation. Or what would be the preferred way to change something like that that is in general ok and can be just a little bit improved?

Quoted 2 lines

Andrew

Heiner

maintainer
#3 lore raw
Quoted 6 lines · Heiner Kallweit
Quoted 5 lines · Andrew Lunn

drivers/net/phy/phy_device.c EXPORT_SYMBOL(genphy_update_link); · line 1760

Quoted 2 lines
drivers/net/phy/phy_device.c +1 -1
-	if (AUTONEG_ENABLE == phydev->autoneg) {
+	if (phydev->autoneg == AUTONEG_ENABLE && phydev->link) {

Hi Heiner

Quoted 8 lines
Quoted 5 lines

I don't necessary agree with placing the constant first in the comparison, but it is best practice not to change it when making additions. It makes it a little bit harder to see what the actual change was.

I understand the point. However a patch to only change the order of the operands most likely would also be rejected as not being worth it.

As i said, i don't necessarily agree with the ordering, but i don't object to it. There are reasonable arguments which it is better. So i would just leave it alone.

Andrew

maintainer
#4 lore raw
In reply to Heiner Kallweit
Quoted 3 lines

- In auto-neg case, skip populating lp_advertising if we don't have a link. This avoids quite some unnecessary MDIO reads in case of phylib polling mode.

Hi Heiner

Could it be that we don't have a link because what the partner is advertising does not intersect with what we are advertising?

Hence knowing what it is adverting is actually useful in this case?

Andrew

author maintainer
#5 lore raw
Quoted 9 lines · Andrew Lunn
Quoted 3 lines

- In auto-neg case, skip populating lp_advertising if we don't have a link. This avoids quite some unnecessary MDIO reads in case of phylib polling mode.

Hi Heiner

Could it be that we don't have a link because what the partner is advertising does not intersect with what we are advertising?

Interesting point. Yes, I think I missed that scenario.

Quoted 4 lines

Hence knowing what it is adverting is actually useful in this case?

Andrew

Heiner

author maintainer
#6 lore raw
Quoted 5 lines · Heiner Kallweit
Quoted 4 lines · Andrew Lunn
Quoted 3 lines

- In auto-neg case, skip populating lp_advertising if we don't have a link. This avoids quite some unnecessary MDIO reads in case of phylib polling mode.

Quoted 7 lines
Quoted 5 lines

Hi Heiner

Could it be that we don't have a link because what the partner is advertising does not intersect with what we are advertising?

Interesting point. Yes, I think I missed that scenario.

Just saw that Dave applied this patch. So I will send a fix to deal with this scenario properly.

Quoted 6 lines
Quoted 4 lines

Hence knowing what it is adverting is actually useful in this case?

Andrew

Heiner

maintainer Applied
#7 lore raw
In reply to Heiner Kallweit

From: Heiner Kallweit <[email protected]>
Date: Sat, 30 Mar 2019 10:22:45 +0100

Quoted 11 lines

This patch improves few aspects of genphy_read_status():

- Don't initialize lpagb, it's not needed.

- Move initializing phydev->speed et al before the if clause.

- In auto-neg case, skip populating lp_advertising if we don't have a link. This avoids quite some unnecessary MDIO reads in case of phylib polling mode.

I decided that I'm not going to push back on this patch purely on the issue of not touching the conditional ordering.

Applied, thanks Heiner.