Skip to contentSkip to conversation

net: dsa: mt7530: fix null pointer dereferencing in port5 setup

PATCH In review Chuanhong Guo 6 participants 9 messages Apr 3, 2020 – Apr 4, 2020

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net: dsa: mt7530: fix null pointer dereferencing in port5 setup

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.
This commit fix this setup code by checking return value of of_parse_phandle before using it.

Fixes 38f790a80560 ("net: dsa: mt7530: Add support for port 5")

---

mt7530 is available as a standalone chip and we should not make it tightly coupled with a specific type of ethernet dt binding in the first place.
A proper fix is to replace this port detection logic with a dt property under mt7530 node, but that's too much for linux-stable.

 drivers/net/dsa/mt7530.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/net/dsa/mt7530.c +3 -0
index 6e91fe2f4b9a..1d53a4ebcd5a 100644
@@ -1414,6 +1414,9 @@ mt7530_setup(struct dsa_switch *ds)
 				continue;
 
 			phy_node = of_parse_phandle(mac_np, "phy-handle", 0);
+			if (!phy_node)
+				continue;
+
 			if (phy_node->parent == priv->dev->of_node->parent) {
 				ret = of_get_phy_mode(mac_np, &interface);
 				if (ret && ret != -ENODEV)
Signature · 7 lines
2.25.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Reviewed-by on v1 1/1
#1 lore raw
Quoted 9 lines · Chuanhong Guo

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.
This commit fix this setup code by checking return value of of_parse_phandle before using it.

Quoted 3 lines
Fixes 38f790a80560 ("net: dsa: mt7530: Add support for port 5")
Reviewed-by on v1 1/1
#2 lore raw
In reply to Chuanhong Guo
Quoted 9 lines · Chuanhong Guo

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.
This commit fix this setup code by checking return value of of_parse_phandle before using it.

Quoted 3 lines
Fixes 38f790a80560 ("net: dsa: mt7530: Add support for port 5")
Quoted 7 lines

---

mt7530 is available as a standalone chip and we should not make it tightly coupled with a specific type of ethernet dt binding in the first place.
A proper fix is to replace this port detection logic with a dt property under mt7530 node, but that's too much for linux-stable.

Agree, one would also wonder why the driver needs to parse parts of the Device Tree which should be done by the core DSA layer.

Signature · 1 line
Florian
Tested-by on v1 1/1
#3 lore raw
In reply to Chuanhong Guo

Quoting Chuanhong Guo <[email protected]>:

Hi Chuanhong,

Quoted 6 lines

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.

MT7530 tries to detect if 2nd GMAC is using a phy with phy-address 0 or 4. If so, switch port 5 needs to be setup so that PHY 0 or 4 is available via port 5 of the switch. Any MAC can talk to PHY 0/4 directly via port 5. This is also explained in the kernel docs mt7530.txt.

May be there are better way to detect that any node is using phy 0/4 of the switch.

Funny that I never tested this case that 2nd gmac node exits and is disabled without using port 5.

Thanks for the fix.

Greats,

René

Quoted 33 lines

This commit fix this setup code by checking return value of of_parse_phandle before using it.

Fixes 38f790a80560 ("net: dsa: mt7530: Add support for port 5")

---

mt7530 is available as a standalone chip and we should not make it tightly coupled with a specific type of ethernet dt binding in the first place.
A proper fix is to replace this port detection logic with a dt property under mt7530 node, but that's too much for linux-stable.

 drivers/net/dsa/mt7530.c | 3 +++
 1 file changed, 3 insertions(+)
drivers/net/dsa/mt7530.c +3 -1
index 6e91fe2f4b9a..1d53a4ebcd5a 100644
@@ -1414,6 +1414,9 @@ mt7530_setup(struct dsa_switch *ds)
 				continue;
 
 			phy_node = of_parse_phandle(mac_np, "phy-handle", 0);
+			if (!phy_node)
+				continue;
+
 			if (phy_node->parent == priv->dev->of_node->parent) {
 				ret = of_get_phy_mode(mac_np, &interface);
 				if (ret && ret != -ENODEV)
--

2.25.1

#4 lore raw

Hi!

Quoted 5 lines · René van Dorst

Quoting Chuanhong Guo <[email protected]>:

Hi Chuanhong,

Quoted 8 lines
Quoted 6 lines

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.

MT7530 tries to detect if 2nd GMAC is using a phy with phy-address 0 or 4.

What if the 2nd GMAC connects to an external PHY on address 0 on a different mdio-bus? This is already happening on mt7629 where the integrated PHY connected to gmac2 is on address 0 and gmac1 connects to external mt753x switch.
On mt7621, the RGMII2 is always wired to switch mac5 as well as external pins, and one should disable switch mac5 in this case or there's pin conflict.

Quoted 6 lines

If so, switch port 5 needs to be setup so that PHY 0 or 4 is available via port 5 of the switch. Any MAC can talk to PHY 0/4 directly via port 5. This is also explained in the kernel docs mt7530.txt.

May be there are better way to detect that any node is using phy 0/4 of the switch.

It should never be detected. This piece of code is to configure how RGMII2 on mt7530 is wired: PHY0/PHY4/switch MAC5/off. And it should be implemented as a configurable dt property. We should not make assumption that 2 RGMIIs always connected to the two macs on mtk_eth_soc and we should never parse parent eth dts in DSA driver.

Signature · 2 lines
Regards,
Chuanhong Guo
maintainer
#5 lore raw
Quoted 15 lines · Chuanhong Guo

Hi!

Quoted 12 lines · René van Dorst

Quoting Chuanhong Guo <[email protected]>:

Hi Chuanhong,

Quoted 6 lines

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.

Quoted 4 lines
Quoted 1 line

MT7530 tries to detect if 2nd GMAC is using a phy with phy-address 0 or 4.

What if the 2nd GMAC connects to an external PHY on address 0 on a different mdio-bus?

In general, you using a phy-handle to cover such a situation. If there is a phy-handle, just use it.

Andrew

#6 lore raw

Hi!

Quoted 7 lines · Andrew Lunn
Quoted 4 lines
Quoted 1 line

MT7530 tries to detect if 2nd GMAC is using a phy with phy-address 0 or 4.

What if the 2nd GMAC connects to an external PHY on address 0 on a different mdio-bus?

In general, you using a phy-handle to cover such a situation. If there is a phy-handle, just use it.

If it's determining where switch mac5 is wired, a phy-handle is fine. Here we are determining where exposed rgmii2 pins are wired. It can be wired to switch mac5 or skip the switch mac completely and connected to phy0/phy4.
Current driver is determining rgmii2 wiring on mt7530 using phy-handle on *another unrelated ethernet node* which doesn't sound right.

Signature · 2 lines
Regards,
Chuanhong Guo
maintainer
#7 lore raw
In reply to Chuanhong Guo

From: Chuanhong Guo <[email protected]>
Date: Fri, 3 Apr 2020 19:28:24 +0800

Quoted 9 lines

The 2nd gmac of mediatek soc ethernet may not be connected to a PHY and a phy-handle isn't always available.
Unfortunately, mt7530 dsa driver assumes that the 2nd gmac is always connected to switch port 5 and setup mt7530 according to phy address of 2nd gmac node, causing null pointer dereferencing when phy-handle isn't defined in dts.
This commit fix this setup code by checking return value of of_parse_phandle before using it.

Quoted 3 lines
Fixes 38f790a80560 ("net: dsa: mt7530: Add support for port 5")

Please do not CC: stable for networking changes, as per:

Documentation/networking/netdev-FAQ.rstq

Applied and queued up for -stable, thank you.

#8 lore raw

Hi!

Quoted 5 lines · David Miller
Quoted 1 line

Please do not CC: stable for networking changes, as per:

Documentation/networking/netdev-FAQ.rstq

Oh! I'm not aware of this doc. Thanks for pointing it out.

Signature · 2 lines
Regards,
Chuanhong Guo