CVE-2026-98377

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> vlan: require the MAC header to be present in __vlan_insert_inner_tag()<br /> <br /> __vlan_insert_inner_tag() only guarantees head room via skb_cow_head(),<br /> never that mac_len bytes of MAC header are present. Its ETH_HLEN<br /> wrappers - __vlan_insert_tag() under skb_vlan_push(), and<br /> vlan_insert_tag() under validate_xmit_vlan() on the generic transmit<br /> path - therefore rewrite the first 16 bytes at skb-&gt;data: a 12-byte<br /> memmove plus two 2-byte stores at +12 and +14. No caller supplies the<br /> bound, while the pop helpers use skb_ensure_writable()/pskb_may_pull().<br /> <br /> An IFF_TUN device has hard_header_len == 0, so packet_snd() accepts a<br /> one-byte AF_PACKET/SOCK_RAW frame. The first vlan push only sets a<br /> hwaccel tag; the next - clsact "action vlan push" or<br /> bpf_skb_vlan_push() - enters the helper with skb-&gt;len still 1. The<br /> head comes from skbuff_small_head without __GFP_ZERO, so each push<br /> drags bytes from beyond skb-&gt;tail into the frame. After three the<br /> one-byte send leaves as 13 bytes carrying 11 bytes of uninitialised<br /> slab:<br /> <br /> 0000: 5a b3 62 12 80 88 ff ff 00 b3 62 12 81<br /> `------------------------------&amp;#39;<br /> only 0x5a was sent; the rest is slab, here the top 56 bits of a<br /> linear-map address<br /> <br /> Require the MAC header the helper rewrites to be present, so such a<br /> frame is dropped rather than transmitted.