CVE-2026-72139
CriticalCVSS 9.8Exploitation Probability (EPSS)
Low risk45th percentile - higher than 45% of all known CVEs
Summary
A vulnerability was found in the Linux kernel's tcp_connect() function related to TCP and MD5 authentication. In one code branch, the tcp_md5sig_info structure and tcp_md5sig_key entries are freed without respecting the RCU grace period, causing a race condition with softirq readers. This leads to a use-after-free of MD5 key memory.
Risk Assessment
An attacker could exploit this vulnerability to cause a kernel panic (denial of service) or potentially escalate privileges within the kernel, compromising system integrity and availability.
Recommendation
Immediately update the Linux kernel to a patched version that defers freeing via kfree_rcu() and uses hlist_del_rcu plus kfree_rcu for MD5 keys. Monitor vendor security advisories and apply the relevant patch.
Other vulnerabilities in Linux kernel
See all- CVE-2026-98164Unknown
In the Linux kernel, KVM x86/mmu's kvm_gfn_is_write_tracked() only checks the supplied memslot, but page tracking is per-address-space and shadow pages are shared across address spaces. With SMM, a GFN can be write-tracked in one address space and appear untracked in another. The fix checks the supplied slot first, then the slot for the other address space, preventing mmu_try_to_unsync_pages() from marking an upper-level shadow page unsync and triggering a BUG in pte_list_remove().
- CVE-2026-98048Unknown
In the Linux kernel's BPF subsystem, mark_fastcall_pattern_for_call() must ensure that a matched "spill; call; fill" instruction series is not interrupted by a jump. Otherwise the rewrite applied by bpf_remove_fastcall_spills_fills() is unsound. The fix records instructions targeted by jumps in insn_aux_data[*].jump_target and uses this flag to stop growing a pattern.
- CVE-2026-98047Unknown
In the Linux kernel's BPF subsystem, in_rbtree_lock_required_cb() only checks the innermost verifier frame, so rbtree callback restrictions disappear in a nested static subprogram call frame. The subprogram can unlock the tree, remove and drop the node being compared, then relock, leading to freed memory being linked into the tree.
- CVE-2026-98046Unknown
In the Linux kernel's BPF subsystem, bpf_btf_find_by_name_kind() returns a new BTF object fd through __btf_new_fd(), which reaches anon_inode_getfd() that can sleep while allocating or expanding the current task fd table. The helper prototype does not set might_sleep, so the verifier allows the helper in non-sleepable contexts such as BPF timer callbacks.
- CVE-2026-98045Unknown
In the Linux kernel, a BPF verifier vulnerability affects stack helpers that may block on filesystem reads (e.g., resolving build IDs) but were not marked as sleepable. The verifier could still allow these helpers from non-sleepable regions such as RCU or preemption-disabled sections.
- CVE-2026-98044Unknown
In the Linux kernel, the BPF verifier mishandled legacy packet-load instructions (BPF_LD_ABS/BPF_LD_IND) reached from callbacks, triggering a verifier bug warning and an -EFAULT on BPF_PROG_LOAD. A privileged program loader could trigger this issue.
- CVE-2026-98043Unknown
In the Linux kernel, the BPF verifier incorrectly inferred that a pointer with an unbounded offset is non-NULL based solely on its type. As a result, a BPF program could pass verification while a NULL pointer dereference occurred at runtime.
- CVE-2026-98042Unknown
In the Linux kernel, the BPF verifier could resurrect a scalar id dropped by collect_linked_regs() because snapshots of compared registers were taken before linked registers were collected. This could lead to range inconsistencies and precision propagation issues.
- CVE-2026-98041High
In the Linux kernel, the BPF verifier incorrectly predicted the outcome of pointer vs zero comparisons in JMP32 instructions, failing to distinguish BPF_JMP from BPF_JMP32 comparisons. This led to incorrect inference that a jump is always taken.
- CVE-2026-98040Unknown
In the Linux kernel, the BPF verifier did not mark the zero register as precise for a register-form NULL check. As a result, one path was pruned and the program could dereference a zero pointer at runtime.
Original NVD description (English source)
In the Linux kernel, the following vulnerability has been resolved: tcp: defer md5sig_info kfree past RCU grace period in tcp_connect The md5+ao reconciliation in tcp_connect() (net/ipv4/tcp_output.c) has two symmetric branches: if (needs_md5) { tcp_ao_destroy_sock(sk, false); } else if (needs_ao) { tcp_clear_md5_list(sk); kfree(rcu_replace_pointer(tp->md5sig_info, NULL, ...)); } Both branches free a per-socket auth-info object while the socket is in TCP_SYN_SENT and is already on the inet ehash (inserted by inet_hash_connect() in tcp_v4_connect()). Both branches are reachable by softirq RX-path readers that load the corresponding info pointer via implicit RCU before bh_lock_sock_nested() is taken. The needs_md5 branch is fixed in the prior patch by re-introducing the call_rcu() free in tcp_ao_destroy_sock(): the equivalent per-key loop runs inside tcp_ao_info_free_rcu(), the RCU callback, so by the time it frees each tcp_ao_key all softirq readers that captured the container have already completed rcu_read_unlock(). The needs_ao branch is not symmetric in the same way. The container free can be deferred via kfree_rcu(md5sig, rcu) -- struct tcp_md5sig_info already has the required rcu member (include/net/tcp.h:1999-2002), and the rest of the tree already does this in the tcp_md5sig_info_add() rollback paths (net/ipv4/tcp_ipv4.c:1410, 1436). But the per-key teardown is done by tcp_clear_md5_list() in process context BEFORE the container's RCU grace period: it walks &md5sig->head and frees each tcp_md5sig_key with bare hlist_del + kfree. A concurrent softirq reader in __tcp_md5_do_lookup() / __tcp_md5_do_lookup_exact() (tcp_ipv4.c:1253, 1298) walks the same list via hlist_for_each_entry_rcu() and races with that bare kfree on the keys themselves -- a per-key slab use-after-free of the same class as the TCP-AO bug, on the same race window. Fix this in two halves: 1. Convert the bare kfree() in tcp_connect() to kfree_rcu() so the md5sig_info container joins the rest of the md5sig lifecycle. The local-variable lift is mechanical and required because kfree_rcu() is a macro that expects an lvalue. 2. Make tcp_clear_md5_list() RCU-safe by replacing hlist_del + kfree(key) with hlist_del_rcu + kfree_rcu(key, rcu). struct tcp_md5sig_key already carries the rcu member (include/net/tcp.h:1995) and tcp_md5_do_del() (net/ipv4/tcp_ipv4.c:1456) already uses kfree_rcu, so this restores the lifecycle invariant the rest of the file follows rather than introducing a one-off. The other caller of tcp_clear_md5_list() is tcp_md5_destruct_sock() (net/ipv4/tcp.c:412), which runs from the sock destructor when the socket is already unhashed and unreachable; the extra grace period there is unnecessary but harmless. Making the helper unconditionally RCU-safe is the cleaner contract. The needs_ao branch is not reachable by the userns reproducer used to demonstrate the AO-side splat (the repro installs both keys but ends up in the needs_md5 branch because the connect peer matches the MD5 key, not the AO key); however the symmetric race exists and a maintainer touching this code should not have to think about which branch escapes RCU and which one does not. [also credits to Qihang, who found that this races with tcp-diag]
Vulnerability data from NVD (NIST) · CISA KEV · EPSS