net: phy: don't crash in phy_read/_write_mmd without a PHY driver
Subsystems (3)
PATCH 1/1 net: phy: don't crash in phy_read/_write_mmd without a PHY driver
From: Alex Marginean <[email protected]>
The APIs can be used by Ethernet drivers without actually loading a PHY driver. This may become more widespread in the future with 802.3z compatible MAC PCS devices being locally driven by the MAC driver when configuring for a PHY mode with in-band negotiation.
Check that drv is not NULL before reading from it.
---
If this hasn't been reported by now I assume it wasn't an issue so far. So I've targeted the patch for net-next and not provided a Fixes: tag.
drivers/net/phy/phy-core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/phy/phy-core.c +2 -2
@@ -387,7 +387,7 @@ int __phy_read_mmd(struct phy_device *phydev, int devad, u32 regnum)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->read_mmd) { +if (phydev->drv && phydev->drv->read_mmd) { val = phydev->drv->read_mmd (phydev, devad, regnum); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff );
@@ -444,7 +444,7 @@ int __phy_write_mmd(struct phy_device *phydev, int devad, u32 regnum, u16 val)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->write_mmd) { +if (phydev->drv && phydev->drv->write_mmd) { ret = phydev->drv->write_mmd (phydev, devad, regnum, val); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff );
Signature · 1 line
2.17.1
Quoted 8 lines · Vladimir Oltean
From: Alex Marginean <[email protected]>
The APIs can be used by Ethernet drivers without actually loading a PHY driver. This may become more widespread in the future with 802.3z compatible MAC PCS devices being locally driven by the MAC driver when configuring for a PHY mode with in-band negotiation.
Quoted 4 lines
Check that drv is not NULL before reading from it.
Signed-off-by Alex Marginean <[email protected]>---
Ugh. Can I just add here:
without resending?
Quoted 29 lines
If this hasn't been reported by now I assume it wasn't an issue so far. So I've targeted the patch for net-next and not provided a Fixes: tag.
drivers/net/phy/phy-core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/net/phy/phy-core.c +2
-3@@ -387,7 +387,7 @@ int __phy_read_mmd(struct phy_device *phydev, int devad, u32 regnum)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->read_mmd) { +if (phydev->drv && phydev->drv->read_mmd) { val = phydev->drv->read_mmd (phydev, devad, regnum); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff );@@ -444,7 +444,7 @@ int __phy_write_mmd(struct phy_device *phydev, int devad, u32 regnum, u16 val)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->write_mmd) { +if (phydev->drv && phydev->drv->write_mmd) { ret = phydev->drv->write_mmd (phydev, devad, regnum, val); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff ); --
Quoted 2 lines
2.17.1
Sorry,
-Vladimir
Quoted 11 lines · Vladimir Oltean
Quoted 10 lines · Vladimir Oltean
From: Alex Marginean <[email protected]>
The APIs can be used by Ethernet drivers without actually loading a PHY driver. This may become more widespread in the future with 802.3z compatible MAC PCS devices being locally driven by the MAC driver when configuring for a PHY mode with in-band negotiation.
Check that drv is not NULL before reading from it.
Quoted 8 lines
Quoted 2 lines
Signed-off-by Alex Marginean <[email protected]>---
Ugh. Can I just add here:
Signed-off-by Vladimir Oltean <[email protected]>without resending?
That's a tag that patchwork understands, so yes, that works:
http://patchwork.ozlabs.org/patch/1224348/mbox/
Quoted 36 lines
Quoted 30 lines
If this hasn't been reported by now I assume it wasn't an issue so far. So I've targeted the patch for net-next and not provided a Fixes: tag.
drivers/net/phy/phy-core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/net/phy/phy-core.c +2
-3@@ -387,7 +387,7 @@ int __phy_read_mmd(struct phy_device *phydev, int devad, u32 regnum)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->read_mmd) { +if (phydev->drv && phydev->drv->read_mmd) { val = phydev->drv->read_mmd (phydev, devad, regnum); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff );@@ -444,7 +444,7 @@ int __phy_write_mmd(struct phy_device *phydev, int devad, u32 regnum, u16 val)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->write_mmd) { +if (phydev->drv && phydev->drv->write_mmd) { ret = phydev->drv->write_mmd (phydev, devad, regnum, val); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff ); --Quoted 2 lines
2.17.1
Sorry,
-Vladimir
Signature · 1 line
Florian
Heiner Kallweit
maintainerQuoted 9 lines · Vladimir Oltean
From: Alex Marginean <[email protected]>
The APIs can be used by Ethernet drivers without actually loading a PHY driver. This may become more widespread in the future with 802.3z compatible MAC PCS devices being locally driven by the MAC driver when configuring for a PHY mode with in-band negotiation.
Check that drv is not NULL before reading from it.
If there's no dedicated PHY driver, then the genphy driver is bound. I think therefore we don't face issues with the current code. But the change looks reasonable.
Quoted 31 lines
Signed-off-by Alex Marginean <[email protected]>---
If this hasn't been reported by now I assume it wasn't an issue so far. So I've targeted the patch for net-next and not provided a Fixes: tag.drivers/net/phy/phy-core.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/net/phy/phy-core.c +2
-2@@ -387,7 +387,7 @@ int __phy_read_mmd(struct phy_device *phydev, int devad, u32 regnum)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->read_mmd) { +if (phydev->drv && phydev->drv->read_mmd) { val = phydev->drv->read_mmd (phydev, devad, regnum); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff );@@ -444,7 +444,7 @@ int __phy_write_mmd(struct phy_device *phydev, int devad, u32 regnum, u16 val)if (regnum > (u16 )~0 || devad >32 )return -EINVAL ; -if (phydev->drv->write_mmd) { +if (phydev->drv && phydev->drv->write_mmd) { ret = phydev->drv->write_mmd (phydev, devad, regnum, val); }else if (phydev->is_c45) {u32 addr =MII_ADDR_C45 | (devad <<16 ) | (regnum &0xffff );
Quoted 8 lines · Heiner Kallweit
Quoted 7 lines · Vladimir Oltean
From: Alex Marginean <[email protected]>
The APIs can be used by Ethernet drivers without actually loading a PHY driver. This may become more widespread in the future with 802.3z compatible MAC PCS devices being locally driven by the MAC driver when configuring for a PHY mode with in-band negotiation.
Quoted 5 lines
Quoted 2 lines
Check that drv is not NULL before reading from it.
If there's no dedicated PHY driver, then the genphy driver is bound. I think therefore we don't face issues with the current code. But the change looks reasonable.
You can unbind the driver from sysfs and observe the problem.
Signature · 1 line
Florian
From: Vladimir Oltean <[email protected]>
Date: Thu, 16 Jan 2020 19:46:28 +0200
Quoted 10 lines
From: Alex Marginean <[email protected]>
The APIs can be used by Ethernet drivers without actually loading a PHY driver. This may become more widespread in the future with 802.3z compatible MAC PCS devices being locally driven by the MAC driver when configuring for a PHY mode with in-band negotiation.
Check that drv is not NULL before reading from it.
Signed-off-by Alex Marginean <[email protected]>
Applied, thank you.