{"id":"AZL-115642","summary":"CVE-2026-99751 affecting package kernel 6.6.161.1-1","details":"In the Linux kernel, the following vulnerability has been resolved:\n\nbpf: Fix skb_pull() BUG in __bpf_redirect_neigh()\n\nThe issue:\nThis is a follow-up to the short link-layer header fixes from Eric.\nAfter vlan pop_eth, a privileged SCHED_ACT BPF program can call\nbpf_skb_change_head(skb, 1, 0) and then bpf_redirect_neigh().\n\n__bpf_redirect_neigh() calls skb_pull(skb, sizeof(*ethh)) after checking\nonly that mac_header \u003c network_header.  Nothing verifies that the MAC\nheader is long enough, that it is linear, or that the IP header the\nredirect goes on to read is linear either.\n\nA SCHED_ACT BPF program can do:\n\n    bpf_skb_change_head(skb, 1, 0);\n    bpf_redirect_neigh(lo, NULL, 0, 0);\n\nafter vlan pop_eth.  The skb then has mac_len == 1 and only one byte of\nlinear data; skb_pull(ETH_HLEN) trips the skb-\u003elen \u003c skb-\u003edata_len BUG in\n__skb_pull():\n\n    kernel BUG at ./include/linux/skbuff.h:2839!\n    RIP: skb_pull+0x40/0x50\n    Call Trace: skb_do_redirect -\u003e packet_xmit -\u003e packet_sendmsg\n    Kernel panic - not syncing: Fatal exception in interrupt\n\nThe v4 and v6 paths read ip_hdr() and ipv6_hdr() as soon as the MAC\nheader is gone.  On the AF_PACKET frame below that leaves headlen == 0\nafter the pull, so the network header sits at skb-\u003etail and the route\nlookup reads ip4h-\u003eprotocol, ip4h-\u003esaddr and ip4h-\u003edaddr past the linear\narea.\n\nThe Fix:\nUse skb_network_offset(), which is the MAC header span, instead of\nskb-\u003emac_len.  mac_len is zero for locally generated tc egress and netkit\npackets, while the span still rejects the one-byte header produced by\npop_eth + bpf_skb_change_head(skb, 1, 0).  The single check also covers\nthe mac_header \u003e= network_header test it replaces.\n\nUse pskb_may_pull() to make the Ethernet header linear before it is read\nor pulled, and to make the IP header linear in both the v4 and the v6\npath before ip_hdr()/ipv6_hdr() is read.  The header pointers are taken\nafter the pull, since pskb_may_pull() may reallocate skb-\u003ehead.\n\nConditions to recreate the bug:\nOn x86-64 with 4 KiB pages, enable CONFIG_PACKET, CONFIG_BPF_SYSCALL,\nCONFIG_NET_SCH_INGRESS, CONFIG_NET_CLS_MATCHALL, CONFIG_NET_ACT_VLAN and\nCONFIG_NET_ACT_BPF.  Run as root with BPF loading, network administration\nand raw-packet privileges.  Bring lo up and attach a SCHED_ACT program\nthat calls bpf_skb_change_head(skb, 1, 0), then returns\nbpf_redirect_neigh(skb-\u003eifindex, NULL, 0, 0):\n\n  ip link set lo up\n  tc qdisc add dev lo clsact\n  tc filter add dev lo egress matchall action vlan pop_eth pipe \\\n      action bpf object-file neigh_chghead.o section action\n\nSend an untagged ETH_P_IP frame of ETH_HLEN + 8192 bytes through an\nAF_PACKET/SOCK_RAW socket bound to lo, without PACKET_VNET_HDR or\nPACKET_QDISC_BYPASS.  Use a unicast destination MAC.  The sender's skb\nhas a 14-byte linear Ethernet prefix and a fragmented payload; pop_eth\nremoves that prefix and change_head restores only one linear byte.\nThe unfixed neighbor redirect then tries to pull 14 bytes and BUGs.\n\nSending the same frame without pop_eth and change_head leaves the IP\nheader in the frags, which is the route lookup reading past skb-\u003etail.","modified":"2026-11-21T14:17:24.117520729Z","published":"2026-11-20T06:18:13Z","upstream":["CVE-2026-99751"],"references":[{"type":"WEB","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-99751"}],"affected":[{"package":{"name":"kernel","ecosystem":"Azure Linux:3","purl":"pkg:rpm/azure-linux/kernel"},"ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"last_affected":"6.6.161.1-1"}]}],"database_specific":{"source":"https://github.com/microsoft/AzureLinuxVulnerabilityData/blob/main/osv/AZL-115642.json"}}],"schema_version":"1.9.2"}