{"id":"CVE-2026-98503","published":"2026-10-18T16:17:01.370","lastModified":"2026-10-19T10:19:03.490","description":"In the Linux kernel, the following vulnerability has been resolved:\n\npacket: use ubuf_info completion for TX_RING packets\n\ntpacket_snd sends skbs with frags pointing into its ring slots. Slots\nare released when skb->destructor is called.\n\nA call to skb_orphan calls skb->destructor before the skb is freed.\nThis can cause the slot to be reused while still linked into the skb.\n\nSwitch to standard zerocopy completion (ubuf_info) so the slot is only\nreleased once all references to the payload are freed or copied.\nRestore skb->destructor to standard sock_wfree.\n\nThe ubuf_info completion callback can be called with a NULL skb, but\nonly from net_zcopy_put and related API, used by zerocopy implementations\nthat hold their own reference on the uarg, such as MSG_ZEROCOPY. This\nuarg is only ever completed from skb_zcopy_clear, so skb is always set.\n\nTo prevent userspace from aliasing in-flight state on shared ring\nslots, allocate tpacket_uarg per packet, rather than per slot. This\nadds a small allocation to the transmit path. Use standard kmalloc to\nallow backporting to stable kernels.\n\nThe uarg holds an sk_wmem_alloc reference, rather than an sk_refcnt\nreference. packet_free_tx_ring waits on sk_wmem_alloc before freeing\nthe ring pages. Always allocate vec->deferred for tx_ring so page-backed\nrings also wait on sk_wmem_alloc when skb_copy_ubufs drops page refs\nbefore calling tpacket_ubuf_complete.\n\nDrop the tx_ring.pg_vec test that tpacket_destruct_skb performed before\naccessing the slot. The sk_wmem_alloc reference now guarantees that the\nslot is valid. The test is also not sufficient by itself, as it reads\npg_vec without pg_vec_lock, so it can race with packet_set_ring.\n\nAs a result a slot is released when its payload is copied, which can\nbe before transmission (e.g., in skb_orphan_frags_rx). If copied\nbefore skb_tx_timestamp() is called, no slot timestamp is recorded,\nsimilar to when skb_orphan() was called early in the datapath before\nthis patch.\n\nRevert the now unused previous skb_zcopy_.._nouarg infra.\n\nDepends on commit 992cc9f94ca9 (\"net/packet: defer vmalloc TX_RING\nfree until skbs finish\").","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/2a0c9afefc0ecea2cc30ccfbbbc98182a0986191","tags":[]},{"url":"https://git.kernel.org/stable/c/9518405613863d0bf0700927a21367f5942cf058","tags":[]},{"url":"https://git.kernel.org/stable/c/e67c455d79fbf709a94965f09a2be6bf4f4ba5fd","tags":[]},{"url":"https://git.kernel.org/stable/c/eda9bcb721f15a93746a99eabda33d80b86404fd","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}