CVE-2026-64459
CriticalCVSS 9.8Exploitation Probability (EPSS)
Low risk39th percentile - higher than 39% of all known CVEs
Summary
In the Linux kernel, the tcp_ao_destroy_sock function removed the call_rcu() call, breaking RCU synchronization. During connection setup (TCP_SYN_SENT), the socket is visible to other CPUs, which may read the ao_info pointer just before it is freed, leading to a use-after-free vulnerability.
Risk Assessment
An unprivileged attacker (UID 1000) in a user namespace can trigger a kernel panic or potentially execute arbitrary code in kernel context by sending crafted TCP-AO segments.
Recommendation
Apply the patch that restores the call_rcu() call in tcp_ao_destroy_sock to ensure proper RCU synchronization. Update the kernel to a version containing this fix.
Other vulnerabilities in Linux kernel
See all- CVE-2026-90049Unknown
In the Linux kernel, a vulnerability in skb_zerocopy() was fixed: it incorrectly called skb_tx_error() on the source skb when copying frags. This completed the zerocopy uarg and cleared the SKBFL_SHARED_FRAG flag on an skb still in use by the network stack. Specifically on the OVS_ACTION_ATTR_USERSPACE path, the skb was not freed on error, which could lead to ESP data being decrypted in place where the skb did not privately own frags, potentially causing data corruption or other issues.
- CVE-2026-90046Unknown
In the Linux kernel, free_pages_nolock() uses spin_trylock() in NMI context on uniprocessor (UP) systems, which is unsafe. The bug affects the page freeing path and can crash the kernel. It is likely exploitable by local attackers for privilege escalation.
- CVE-2026-90045Unknown
In the Linux kernel USB gadget ffs driver, the io_data structure stores a pointer to the submitting task's mm_struct but does not hold a reference while async requests are pending. This can cause a use-after-free if the task exits before completion handling finishes. The fix takes a reference with mmgrab() when queuing and releases it with mmdrop() on completion.
- CVE-2026-90044Unknown
In the Linux kernel USB gadget f_fs driver, a use-after-free exists in the AIO error path. When ffs_epfile_io() fails with an error other than -EIOCBQUEUED, the io_data structure is freed, but the kiocb cancel function remains armed and points to the freed pointer. A concurrent cancel (e.g., sys_io_cancel()) can cause a use-after-free. The fix safely disarms the cancellation via kiocb->ki_complete() and returns -EIOCBQUEUED.
- CVE-2026-90043Unknown
In the Linux kernel zram module, the slot lock bit is set in the wrong position on 64-bit big-endian systems. Instead of landing in the flags field, it lands in ac_time, so with ZRAM_TRACK_ENTRY_ACTIME enabled, storing the access time wipes out the held lock bit and lets another CPU take the same slot lock. An access time value with that bit set can make the slot appear locked forever. The fix shifts the lock bit into the flags half on big-endian 64-bit.
- CVE-2026-90042Unknown
A vulnerability was found in the Linux kernel's Ceph filesystem where ceph_fname_to_usr() passed vmalloc() buffers to the fscrypt crypto API, which requires linear memory. This can cause kernel oopses, especially on non-x86 platforms. The fix adds support for vmalloc() buffers using a bounce buffer.
- CVE-2026-90041Unknown
In the Linux kernel, the HID sony driver fails to remove a controller from the device list when HID core input device registration fails. The devres-managed sony_sc is freed while its list node remains linked, leading to use-after-free when the next matching controller is handled.
- CVE-2026-90039Unknown
In the Linux kernel, NFSD admin state-revocation handlers (unlock_filesystem, NFSD_CMD_UNLOCK_FILESYSTEM, NFSD_CMD_UNLOCK_EXPORT) only check nn->nfsd_serv, which is set before nn->conf_id_hashtbl is allocated. This leads to a NULL dereference when the server has not yet started.
- CVE-2026-90038Unknown
In the Linux kernel, nfsd4_revoke_export_states() drops nn->client_lock but the held stateid reference does not pin the client. A concurrent client teardown can free the client while revoke_one_stid() still dereferences it, causing a use-after-free.
- CVE-2026-90037Unknown
In the Linux kernel, an nfs4_openowner left on nn->close_lru holds only a raw pointer to its nfs4_client. When the laundromat reaps entries, drops client_lock and calls nfs4_put_stid(), a concurrent force_expire_client() can free the client, causing a use-after-free.
Original NVD description (English source)
In the Linux kernel, the following vulnerability has been resolved: tcp: restore RCU grace period in tcp_ao_destroy_sock Commit 51e547e8c89c ("tcp: Free TCP-AO/TCP-MD5 info/keys without RCU") removed the call_rcu() callback from tcp_ao_destroy_sock(), arguing that "the destruction of info/keys is delayed until the socket destructor" and therefore "no one can discover it anymore". That argument does not hold for the call site in tcp_connect() (net/ipv4/tcp_output.c:4327-4332). At that point the socket is in TCP_SYN_SENT, has already been inserted into the inet ehash by inet_hash_connect() in tcp_v4_connect(), and is therefore very much discoverable: any softirq running tcp_v4_rcv() on another CPU can take the socket out of the ehash, walk into tcp_inbound_hash(), and load tp->ao_info via implicit RCU before bh_lock_sock_nested() is taken on the destroying CPU. The reader path then enters __tcp_ao_do_lookup() (net/ipv4/tcp_ao.c:208) which re-loads tp->ao_info via rcu_dereference_check(); the re-load can still observe the (about-to-be-freed) pointer because there is no synchronize_rcu() between rcu_assign_pointer(tp->ao_info, NULL) and tcp_ao_info_free() in tcp_ao_destroy_sock(). The captured pointer is then walked at line 223: hlist_for_each_entry_rcu(key, &ao->head, node, ...) The writer's synchronous kfree() is free to complete between the line 218 re-fetch and the line 223 hlist iteration. The slab is reused (or simply LIST_POISON1-stamped if not yet reused) and the iteration walks attacker-controlled or poison memory in softirq context. Reproducer (no debug shim, stock x86_64 v7.1-rc2 SMP+KASAN, QEMU+KVM): an unprivileged uid=1000 process inside CLONE_NEWUSER|CLONE_NEWNET installs TCP_MD5SIG + TCP_AO_ADD_KEY on a TCP socket, sprays forged TCP-AO segments toward its eventual 4-tuple via raw sockets, then calls connect(). The md5-wins reconciliation in tcp_connect() fires tcp_ao_destroy_sock(); the softirq backlog reader on the loopback NAPI path crashes on the freed ao->head.first walk: Oops: general protection fault, probably for non-canonical address 0xfbd59c000000002f KASAN: maybe wild-memory-access in range [0xdead000000000178-0xdead00000000017f] CPU: 0 UID: 1000 PID: 100 Comm: repro_userns RIP: 0010:__tcp_ao_do_lookup+0x107/0x1c0 Call Trace: <IRQ> __tcp_ao_do_lookup+0x107/0x1c0 tcp_ao_inbound_lookup.constprop.0+0x12a/0x200 tcp_inbound_ao_hash+0x5ea/0x1520 tcp_inbound_hash+0x7ce/0x1240 tcp_v4_rcv+0x1e7a/0x3e10 ... Restore the RCU grace period: re-add struct rcu_head to tcp_ao_info and replace the synchronous tcp_ao_info_free() with a call_rcu() callback. Readers that captured tp->ao_info before rcu_assign_pointer NULLed it now see the object remain valid until rcu_read_unlock(). With the patch applied the reproducer runs cleanly for 2000 iterations on the same kernel build.