{"id":"CVE-2026-98570","published":"2026-10-18T16:17:09.233","lastModified":"2026-10-19T10:19:15.053","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nnet: ipv6: keep room for the mac header in dst_dev_overhead()\n\nThe seg6, ioam6 and rpl lwtunnels size their skb_cow_head() request as\nthe length they are about to push plus dst_dev_overhead(), then push the\nnew headers and rebuild the mac header below them with\nskb_mac_header_rebuild().  That rebuild needs skb->mac_len of headroom,\nbut dst_dev_overhead() leaves LL_RESERVED_SPACE() of the egress device,\n16 bytes for plain Ethernet.\n\nWhere the mac header is longer than that, as it is on ingress through a\nVLAN device with reorder_hdr off, the rebuild runs out of room:\nskb_set_mac_header(skb, -skb->mac_len) computes a negative offset,\nstores it unchecked in the u16 skb->mac_header, and the memmove that\nfollows writes skb->mac_len bytes about 64 KB past skb->head.\nForwarding plain ping6 traffic through such a device reproduces it on\nall five seg6 encapsulation modes and on the rpl and ioam6 inline paths;\nskb->mac_header comes back as 65534 on a 704-byte head.\n\nReturn the larger of the two.  The helper already returns skb->mac_len\nwhen it has no dst, so this only makes the other branch agree, and it\ncovers every caller rather than each call site in turn.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/36c07127924f11fc5741a780abcaaf808fe9ef13","tags":[]},{"url":"https://git.kernel.org/stable/c/87cd6b717e4069dca34e0866e57eaeb44d3b173e","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}