net: phy: Eliminate unnecessary soft
Subsystems (3)
cover net: phy: Eliminate unnecessary soft
Hi all,
This patch series eliminates unnecessary software resets of the PHY. This should hopefully not break anybody's hardware; but I would appreciate testing to make sure this is is the case.
Sorry for this long email list, I wanted to make sure I reached out to all people who made changes to the Marvell PHY driver.
Thank you!
Florian Fainelli (2):
net: phy: Stop with excessive soft reset
net: phy: marvell: Avoid unnecessary soft reset
drivers/net/phy/marvell.c | 63 ++++++++++++------------------------ drivers/net/phy/phy_device.c | 2 -- 2 files changed, 21 insertions(+), 44 deletions(-)
Signature · 1 line
2.17.1
PATCH 1/2 net: phy: Stop with excessive soft reset
While consolidating the PHY reset in phy_init_hw() an unconditionaly BMCR soft-reset I became quite trigger happy with those. This was later on deactivated for the Generic PHY driver on the premise that a prior software entity (e.g: bootloader) might have applied workarounds in commit 0878fff1f42c ("net: phy: Do not perform software reset for Generic PHY").
Since we have a hook to wire-up a soft_reset callback, just use that and get rid of the call to genphy_soft_reset() entirely. This speeds up initialization and link establishment for most PHYs out there that do not require a reset.
--- drivers/net/phy/phy_device.c | 2 -- 1 file changed, 2 deletions(-)
drivers/net/phy/phy_device.c +0 -2
@@ -880,8 +880,6 @@ int phy_init_hw(struct phy_device *phydev)if (phydev->drv->soft_reset) ret = phydev->drv->soft_reset (phydev); -else - ret =genphy_soft_reset (phydev);if (ret <0 )return ret;
Signature · 1 line
2.17.1
PATCH 2/2 net: phy: marvell: Avoid unnecessary soft reset
In reply to Florian FainelliThe BMCR.RESET bit on Marvell PHYs appears to be acting as a commit operation, so we try to hit that bit which disrupts the link, only when an actual register programming required a change.
Determine from marvell_set_polarity()'s return code whether the register value was changed and if it was, propagate that to the logic that hits the software reset bit. This avoids doing unnecessary soft reset if the PHY is configured in the same state it was previously.
Since m88e1111_config_aneg() likely just hit the software reset bit blindly without checking whether this is necessary, eliminate that function and make it use the generic marvell_config_aneg() function instead.
--- drivers/net/phy/marvell.c | 63 +++++++++++++-------------------------- 1 file changed, 21 insertions(+), 42 deletions(-)
drivers/net/phy/marvell.c +21 -42
@@ -265,7 +265,7 @@ static int marvell_set_polarity(struct phy_device *phydev, int polarity)return err; } -return 0 ; +return val != reg; } static int marvell_set_downshift(struct phy_device *phydev, bool enable,
@@ -287,12 +287,15 @@ static int marvell_set_downshift(struct phy_device *phydev, bool enable,static int marvell_config_aneg (struct phy_device *phydev) { +int changed =0 ;int err; err =marvell_set_polarity (phydev, phydev->mdix_ctrl);if (err <0 )return err; + changed = err; + err =phy_write (phydev,MII_M1111_PHY_LED_CONTROL ,MII_M1111_PHY_LED_DIRECT );if (err <0 )
@@ -302,7 +305,7 @@ static int marvell_config_aneg(struct phy_device *phydev)if (err <0 )return err; -if (phydev->autoneg !=AUTONEG_ENABLE ) { +if (phydev->autoneg !=AUTONEG_ENABLE || changed) {/* A write to speed/duplex bits (that is performed by * genphy_config_aneg() call above) must be followed by * a software reset. Otherwise, the write has no effect.
@@ -350,42 +353,6 @@ static int m88e1101_config_aneg(struct phy_device *phydev)return marvell_config_aneg (phydev); } -static int m88e1111_config_aneg (struct phy_device *phydev) -{ -int err; - -/* The Marvell PHY has an errata which requires -* that certain registers get written in order -* to restart autonegotiation -*/ - err =genphy_soft_reset (phydev); - - err =marvell_set_polarity (phydev, phydev->mdix_ctrl); -if (err <0 ) -return err; - - err =phy_write (phydev,MII_M1111_PHY_LED_CONTROL , -MII_M1111_PHY_LED_DIRECT ); -if (err <0 ) -return err; - - err =genphy_config_aneg (phydev); -if (err <0 ) -return err; - -if (phydev->autoneg !=AUTONEG_ENABLE ) { -/* A write to speed/duplex bits (that is performed by -* genphy_config_aneg() call above) must be followed by -* a software reset. Otherwise, the write has no effect. -*/ - err =genphy_soft_reset (phydev); -if (err <0 ) -return err; - } - -return 0 ; -} -#ifdef CONFIG_OF_MDIO /* Set and/or override some configuration registers based on the * marvell,reg-init property stored in the of_node for the phydev.
@@ -479,6 +446,7 @@ static int m88e1121_config_aneg_rgmii_delays(struct phy_device *phydev)static int m88e1121_config_aneg (struct phy_device *phydev) { +int changed =0 ;int err =0 ;if (phy_interface_is_rgmii (phydev)) {
@@ -487,15 +455,26 @@ static int m88e1121_config_aneg(struct phy_device *phydev)return err; } - err =genphy_soft_reset (phydev); + err =marvell_set_polarity (phydev, phydev->mdix_ctrl);if (err <0 )return err; - err =marvell_set_polarity (phydev, phydev->mdix_ctrl); + changed = err; + + err =genphy_config_aneg (phydev);if (err <0 )return err; -return genphy_config_aneg (phydev); +if (phydev->autoneg !=AUTONEG_ENABLE || changed) { +/* A software reset is used to ensure a "commit" of the +* changes is done. +*/ + err =genphy_soft_reset (phydev); +if (err <0 ) +return err; + } + +return 0 ; }static int m88e1318_config_aneg (struct phy_device *phydev)
@@ -2067,7 +2046,7 @@ static struct phy_driver marvell_drivers[] = { .flags = PHY_HAS_INTERRUPT, .probe = marvell_probe, .config_init = &m88e1111_config_init, - .config_aneg = &m88e1111_config_aneg, + .config_aneg = &marvell_config_aneg, .read_status = &marvell_read_status, .ack_interrupt = &marvell_ack_interrupt, .config_intr = &marvell_config_intr,
Signature · 1 line
2.17.1
Quoted 3 lines
This patch series eliminates unnecessary software resets of the PHY. This should hopefully not break anybody's hardware; but I would appreciate testing to make sure this is is the case.
Tested with Marvell 88E6161 and Marvell 88E6352 switches (which use Marvell PHY driver.)
Andrew Lunn
maintainerQuoted 8 lines · Florian Fainelli
Hi all,
This patch series eliminates unnecessary software resets of the PHY. This should hopefully not break anybody's hardware; but I would appreciate testing to make sure this is is the case.
Sorry for this long email list, I wanted to make sure I reached out to all people who made changes to the Marvell PHY driver.
Hi Florian
I want to test this, but i'm battling other issues at the moment and not got around to it.
But i see it has been tested by two people. So i think we should merge it, and see if anybody screams.
Andrew
Quoted 8 lines · Florian Fainelli
Hi all,
This patch series eliminates unnecessary software resets of the PHY. This should hopefully not break anybody's hardware; but I would appreciate testing to make sure this is is the case.
Sorry for this long email list, I wanted to make sure I reached out to all people who made changes to the Marvell PHY driver.
I got around to testing this at last. Not too much testing, just booting it on five different systems using Marvell and Micrel PHYs. They all got link as expected.
Andrew
Hi,
Quoted 11 lines · Florian Fainelli
Hi all,
This patch series eliminates unnecessary software resets of the PHY. This should hopefully not break anybody's hardware; but I would appreciate testing to make sure this is is the case.
Sorry for this long email list, I wanted to make sure I reached out to all people who made changes to the Marvell PHY driver.
Thank you!
Quoted 11 lines
Florian Fainelli (2):
net: phy: Stop with excessive soft reset
net: phy: marvell: Avoid unnecessary soft resetdrivers/net/phy/marvell.c | 63 ++++++++++++------------------------ drivers/net/phy/phy_device.c | 2 -- 2 files changed, 21 insertions(+), 44 deletions(-)Signature · 1 line
2.17.1
I tested your patches on our board with a Marvell 88E1510. Looks good!
Clemens