Re: [PATCH net] vlan: prevent cross-netns MII register writes
From: Florian Fainelli
Date: Thu Nov 12 2026 - 14:53:48 EST
On 11/12/26 08:15, Andrew Lunn wrote:
On Thu, Nov 12, 2026 at 07:56:24AM -0800, Jakub Kicinski wrote:
On Fri, 6 Nov 2026 12:46:39 +0000 Hungyu Lin wrote:
vlan_dev_ioctl() forwards SIOCSMIIREG to the underlying real
device without checking whether the VLAN device and real device
belong to the same network namespace.
A VLAN device can be moved into another network namespace while
its real device remains in the host namespace. This allows a
process with the required privileges in the VLAN device's
namespace to modify PHY registers on the host's real device.
Reject SIOCSMIIREG when the two devices belong to different
network namespaces. Preserve MII reads across namespaces and
allow MII writes when both devices belong to the same namespace.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Hungyu Lin <dennylin0707@xxxxxxxxx>
diff --git a/net/8021q/vlan_dev.c b/net/8021q/vlan_dev.c
index bca0a3693636..0d0904103302 100644
--- a/net/8021q/vlan_dev.c
+++ b/net/8021q/vlan_dev.c
@@ -416,9 +416,12 @@ static int vlan_dev_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
ifrr.ifr_ifru = ifr->ifr_ifru;
switch (cmd) {
+ case SIOCSMIIREG:
+ if (!net_eq(dev_net(dev), dev_net(real_dev)))
+ break;
+ fallthrough;
case SIOCGMIIPHY:
case SIOCGMIIREG:
- case SIOCSMIIREG:
err = dev_eth_ioctl(real_dev, &ifrr, cmd);
break;
}
Andrew, Florian, any idea if we could delete this completely?
Why would anyone want to run MII/PHY commands on a VLAN?
It does seem questionable.
Agreed, it does not quite make sense to me to forward these ioctl() calls. Looks like this predates the import of Linux-2.6.12-rc2 into git, need to do some more history digging to find out why this was done.
Hungyu, do any other virtual devices forward these ioctl calls to the
physical device below? I assume they have the same problem. So we need
a fix for them all.
Andrew
--
Florian