Skip to contentSkip to conversation

net: ftgmac100: fix potential NULL pointer access in ftgmac100_phy_disconnect

PATCH Applied Heiner Kallweit 3 participants 5 messages Jul 30, 2025 – Aug 5, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net: ftgmac100: fix potential NULL pointer access in ftgmac100_phy_disconnect

After the call to phy_disconnect() netdev->phydev is reset to NULL. So fixed_phy_unregister() would be called with a NULL pointer as argument. Therefore cache the phy_device before this call.

Fixes e24a6c874601 ("net: ftgmac100: Get link speed and duplex for NC-SI")
---
 drivers/net/ethernet/faraday/ftgmac100.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/net/ethernet/faraday/ftgmac100.c +4 -3
index 5d0c09068..a863f7841 100644
@@ -1750,16 +1750,17 @@ static int ftgmac100_setup_mdio(struct net_device *netdev)
 static void ftgmac100_phy_disconnect(struct net_device *netdev)
 {
 	struct ftgmac100 *priv = netdev_priv(netdev);
+	struct phy_device *phydev = netdev->phydev;
 
-	if (!netdev->phydev)
+	if (!phydev)
 		return;
 
-	phy_disconnect(netdev->phydev);
+	phy_disconnect(phydev);
 	if (of_phy_is_fixed_link(priv->dev->of_node))
 		of_phy_deregister_fixed_link(priv->dev->of_node);
 
 	if (priv->use_ncsi)
-		fixed_phy_unregister(netdev->phydev);
+		fixed_phy_unregister(phydev);
 }
 
 static void ftgmac100_destroy_mdio(struct net_device *netdev)
Signature · 1 line
2.50.1
#1 lore raw
Quoted 1 line · Heiner Kallweit

After the call to phy_disconnect() netdev->phydev is reset to NULL.

phy_disconnect() in its flow does not set phydev to NULL, if anywhere it happens in of_phy_deregister_fixed_link(), which already calls fixed_phy_unregister() before setting phydev to NULL

From my understanding (which very much could be wrong) of ftgmac100_probe(), these two cases are mutually exclusive. The device either uses NCSI or will use a phy based on the DT "fixed-link" property

Quoted 1 line

So fixed_phy_unregister() would be called with a NULL pointer as argument.

Given my analysis above, I don't think this case is possible.

Best regards,
Dawid

Quoted 34 lines

Therefore cache the phy_device before this call.

Fixes e24a6c874601 ("net: ftgmac100: Get link speed and duplex for NC-SI")
---
  drivers/net/ethernet/faraday/ftgmac100.c | 7 ++++---
  1 file changed, 4 insertions(+), 3 deletions(-)
drivers/net/ethernet/faraday/ftgmac100.c +4 -3
index 5d0c09068..a863f7841 100644
@@ -1750,16 +1750,17 @@ static int ftgmac100_setup_mdio(struct net_device *netdev)
  static void ftgmac100_phy_disconnect(struct net_device *netdev)
  {
  	struct ftgmac100 *priv = netdev_priv(netdev);
+	struct phy_device *phydev = netdev->phydev;
 
-	if (!netdev->phydev)
+	if (!phydev)
  		return;
 
-	phy_disconnect(netdev->phydev);
+	phy_disconnect(phydev);
  	if (of_phy_is_fixed_link(priv->dev->of_node))
  		of_phy_deregister_fixed_link(priv->dev->of_node);
 
  	if (priv->use_ncsi)
-		fixed_phy_unregister(netdev->phydev);
+		fixed_phy_unregister(phydev);
  }
 
  static void ftgmac100_destroy_mdio(struct net_device *netdev)
#2 lore raw
Quoted 4 lines · Dawid Osuchowski
Quoted 1 line · Heiner Kallweit

After the call to phy_disconnect() netdev->phydev is reset to NULL.

phy_disconnect() in its flow does not set phydev to NULL, if anywhere it happens in of_phy_deregister_fixed_link(), which already calls fixed_phy_unregister() before setting phydev to NULL

I can't follow this argumentation. Look at phy_detach(), there we have:
if (phydev->attached_dev)

	phydev->attached_dev->phydev = NULL;

So netdev->phydev is NULL after the call to phy_disconnect, provided that the PHY was attached to the netdev before. If use_ncsi is true, then fixed_phy_unregister() will be called with a NULL argument.

This is independent of whether of_phy_is_fixed_link() is true or not. Very likely it's false in the NCSI case.

Quoted 40 lines

From my understanding (which very much could be wrong) of ftgmac100_probe(), these two cases are mutually exclusive. The device either uses NCSI or will use a phy based on the DT "fixed-link" property

Quoted 1 line

So fixed_phy_unregister() would be called with a NULL pointer as argument.

Given my analysis above, I don't think this case is possible.

Best regards,
Dawid

Quoted 30 lines

Therefore cache the phy_device before this call.

Fixes e24a6c874601 ("net: ftgmac100: Get link speed and duplex for NC-SI")

---
  drivers/net/ethernet/faraday/ftgmac100.c | 7 ++++---

  1 file changed, 4 insertions(+), 3 deletions(-)
drivers/net/ethernet/faraday/ftgmac100.c +4 -1
index 5d0c09068..a863f7841 100644
@@ -1750,16 +1750,17 @@ static int ftgmac100_setup_mdio(struct net_device *netdev)
   static void ftgmac100_phy_disconnect(struct net_device *netdev)
   {
       struct ftgmac100 *priv = netdev_priv(netdev);
+    struct phy_device *phydev = netdev->phydev;
   -    if (!netdev->phydev)
+    if (!phydev)
           return;
   -    phy_disconnect(netdev->phydev);
+    phy_disconnect(phydev);
       if (of_phy_is_fixed_link(priv->dev->of_node))
           of_phy_deregister_fixed_link(priv->dev->of_node);
         if (priv->use_ncsi)
-        fixed_phy_unregister(netdev->phydev);
+        fixed_phy_unregister(phydev);
   }
     static void ftgmac100_destroy_mdio(struct net_device *netdev)
Reviewed-by
#3 lore raw
Quoted 11 lines · Heiner Kallweit
Quoted 4 lines · Dawid Osuchowski
Quoted 1 line · Heiner Kallweit

After the call to phy_disconnect() netdev->phydev is reset to NULL.

phy_disconnect() in its flow does not set phydev to NULL, if anywhere it happens in of_phy_deregister_fixed_link(), which already calls fixed_phy_unregister() before setting phydev to NULL

I can't follow this argumentation. Look at phy_detach(), there we have:
if (phydev->attached_dev)

	phydev->attached_dev->phydev = NULL;

So netdev->phydev is NULL after the call to phy_disconnect, provided that the PHY was attached to the netdev before. If use_ncsi is true, then fixed_phy_unregister() will be called with a NULL argument.

You are absolutely correct. Somehow I missed that and I apologize for the confusion on my side.

With that cleared up:

Thanks,
Dawid

Quoted 45 lines

This is independent of whether of_phy_is_fixed_link() is true or not. Very likely it's false in the NCSI case.

Quoted 40 lines

From my understanding (which very much could be wrong) of ftgmac100_probe(), these two cases are mutually exclusive. The device either uses NCSI or will use a phy based on the DT "fixed-link" property

Quoted 1 line

So fixed_phy_unregister() would be called with a NULL pointer as argument.

Given my analysis above, I don't think this case is possible.

Best regards,
Dawid

Quoted 30 lines

Therefore cache the phy_device before this call.

Fixes e24a6c874601 ("net: ftgmac100: Get link speed and duplex for NC-SI")
---
   drivers/net/ethernet/faraday/ftgmac100.c | 7 ++++---
   1 file changed, 4 insertions(+), 3 deletions(-)
drivers/net/ethernet/faraday/ftgmac100.c +4 -1
index 5d0c09068..a863f7841 100644
@@ -1750,16 +1750,17 @@ static int ftgmac100_setup_mdio(struct net_device *netdev)
   static void ftgmac100_phy_disconnect(struct net_device *netdev)
   {
       struct ftgmac100 *priv = netdev_priv(netdev);
+    struct phy_device *phydev = netdev->phydev;
   -    if (!netdev->phydev)
+    if (!phydev)
           return;
   -    phy_disconnect(netdev->phydev);
+    phy_disconnect(phydev);
       if (of_phy_is_fixed_link(priv->dev->of_node))
           of_phy_deregister_fixed_link(priv->dev->of_node);
         if (priv->use_ncsi)
-        fixed_phy_unregister(netdev->phydev);
+        fixed_phy_unregister(phydev);
   }
     static void ftgmac100_destroy_mdio(struct net_device *netdev)
#4 lore raw
In reply to Heiner Kallweit
This patch was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 9 lines · On Wed, 30 Jul 2025 22:23:23 +0200 you wrote:

After the call to phy_disconnect() netdev->phydev is reset to NULL. So fixed_phy_unregister() would be called with a NULL pointer as argument. Therefore cache the phy_device before this call.

Fixes e24a6c874601 ("net: ftgmac100: Get link speed and duplex for NC-SI")

[...]

Here is the summary with links:
- [net] net: ftgmac100: fix potential NULL pointer access in ftgmac100_phy_disconnect

    https://git.kernel.org/netdev/net/c/e88fbc30dda1

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html