CVE-2025-21920

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> vlan: enforce underlying device type<br /> <br /> Currently, VLAN devices can be created on top of non-ethernet devices.<br /> <br /> Besides the fact that it doesn&amp;#39;t make much sense, this also causes a<br /> bug which leaks the address of a kernel function to usermode.<br /> <br /> When creating a VLAN device, we initialize GARP (garp_init_applicant)<br /> and MRP (mrp_init_applicant) for the underlying device.<br /> <br /> As part of the initialization process, we add the multicast address of<br /> each applicant to the underlying device, by calling dev_mc_add.<br /> <br /> __dev_mc_add uses dev-&gt;addr_len to determine the length of the new<br /> multicast address.<br /> <br /> This causes an out-of-bounds read if dev-&gt;addr_len is greater than 6,<br /> since the multicast addresses provided by GARP and MRP are only 6<br /> bytes long.<br /> <br /> This behaviour can be reproduced using the following commands:<br /> <br /> ip tunnel add gretest mode ip6gre local ::1 remote ::2 dev lo<br /> ip l set up dev gretest<br /> ip link add link gretest name vlantest type vlan id 100<br /> <br /> Then, the following command will display the address of garp_pdu_rcv:<br /> <br /> ip maddr show | grep 01:80:c2:00:00:21<br /> <br /> Fix the bug by enforcing the type of the underlying device during VLAN<br /> device initialization.