Linux Kernel
CVE-2024-26864
HIGH
Severity by source
AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
AC:H reflects the requirement to engineer a bhash2 memory allocation failure at a specific code point; all other metrics align with local low-privilege kernel memory corruption.
Primary rating from Vendor (416baaa9-dc9f-4396-8d5f-8c081fb06d67).
CVSS VectorVendor: 416baaa9-dc9f-4396-8d5f-8c081fb06d67
Lifecycle Timeline
7DescriptionCVE.org
In the Linux kernel, the following vulnerability has been resolved:
tcp: Fix refcnt handling in __inet_hash_connect().
syzbot reported a warning in sk_nulls_del_node_init_rcu().
The commit 66b60b0c8c4a ("dccp/tcp: Unhash sk from ehash for tb2 alloc failure after check_estalblished().") tried to fix an issue that an unconnected socket occupies an ehash entry when bhash2 allocation fails.
In such a case, we need to revert changes done by check_established(), which does not hold refcnt when inserting socket into ehash.
So, to revert the change, we need to __sk_nulls_add_node_rcu() instead of sk_nulls_add_node_rcu().
Otherwise, sock_put() will cause refcnt underflow and leak the socket.
[0]: WARNING: CPU: 0 PID: 23948 at include/net/sock.h:799 sk_nulls_del_node_init_rcu+0x166/0x1a0 include/net/sock.h:799 Modules linked in: CPU: 0 PID: 23948 Comm: syz-executor.2 Not tainted 6.8.0-rc6-syzkaller-00159-gc055fc00c07b #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/25/2024 RIP: 0010:sk_nulls_del_node_init_rcu+0x166/0x1a0 include/net/sock.h:799 Code: e8 7f 71 c6 f7 83 fb 02 7c 25 e8 35 6d c6 f7 4d 85 f6 0f 95 c0 5b 41 5c 41 5d 41 5e 41 5f 5d c3 cc cc cc cc e8 1b 6d c6 f7 90 <0f> 0b 90 eb b2 e8 10 6d c6 f7 4c 89 e7 be 04 00 00 00 e8 63 e7 d2 RSP: 0018:ffffc900032d7848 EFLAGS: 00010246 RAX: ffffffff89cd0035 RBX: 0000000000000001 RCX: 0000000000040000 RDX: ffffc90004de1000 RSI: 000000000003ffff RDI: 0000000000040000 RBP: 1ffff1100439ac26 R08: ffffffff89ccffe3 R09: 1ffff1100439ac28 R10: dffffc0000000000 R11: ffffed100439ac29 R12: ffff888021cd6140 R13: dffffc0000000000 R14: ffff88802a9bf5c0 R15: ffff888021cd6130 FS: 00007f3b823f16c0(0000) GS:ffff8880b9400000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00007f3b823f0ff8 CR3: 000000004674a000 CR4: 00000000003506f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: <TASK> __inet_hash_connect+0x140f/0x20b0 net/ipv4/inet_hashtables.c:1139 dccp_v6_connect+0xcb9/0x1480 net/dccp/ipv6.c:956 __inet_stream_connect+0x262/0xf30 net/ipv4/af_inet.c:678 inet_stream_connect+0x65/0xa0 net/ipv4/af_inet.c:749 __sys_connect_file net/socket.c:2048 [inline] __sys_connect+0x2df/0x310 net/socket.c:2065 __do_sys_connect net/socket.c:2075 [inline] __se_sys_connect net/socket.c:2072 [inline] __x64_sys_connect+0x7a/0x90 net/socket.c:2072 do_syscall_64+0xf9/0x240 entry_SYSCALL_64_after_hwframe+0x6f/0x77 RIP: 0033:0x7f3b8167dda9 Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 e1 20 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 b0 ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007f3b823f10c8 EFLAGS: 00000246 ORIG_RAX: 000000000000002a RAX: ffffffffffffffda RBX: 00007f3b817abf80 RCX: 00007f3b8167dda9 RDX: 000000000000001c RSI: 0000000020000040 RDI: 0000000000000003 RBP: 00007f3b823f1120 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000001 R13: 000000000000000b R14: 00007f3b817abf80 R15: 00007ffd3beb57b8 </TASK>
AnalysisAI
Reference count underflow in Linux kernel's __inet_hash_connect() corrupts socket lifecycle state, potentially enabling local privilege escalation via use-after-free. The bug was introduced by a prior fix (commit 66b60b0c8c4a) in the TCP/DCCP networking stack: when bhash2 allocation fails after check_established(), the error-revert path calls sk_nulls_add_node_rcu() - which bumps refcnt - instead of __sk_nulls_add_node_rcu(), so the subsequent sock_put() underflows the counter, leaking the socket object. No public exploit exists and this is not in CISA KEV, but the local CVSS 7.8 (H/H/H impact) reflects the kernel memory corruption potential arising from the refcnt mismanagement, confirmed by syzbot on kernel 6.8.0-rc6.
Technical ContextAI
The vulnerability resides in net/ipv4/inet_hashtables.c within __inet_hash_connect(), which manages TCP and DCCP socket insertion into the kernel's established hash table (ehash) during connection setup. The function check_established() inserts a socket into ehash without holding an additional reference count. A prior fix (commit 66b60b0c8c4a) added an ehash revert path for bhash2 allocation failures, but incorrectly used sk_nulls_add_node_rcu() - which increments refcnt - rather than __sk_nulls_add_node_rcu() which does not. When sock_put() is called to release the spuriously incremented reference, the counter underflows, producing a dangling socket pointer. Affected products are Linux kernel versions across multiple stable branches per CPE cpe:2.3:o:linux:linux_kernel:*:*:*:*:*:*:*:*. The NVD CWE tag is CWE-200 (Information Disclosure), which appears misclassified - the root cause is a reference counting error (structurally closer to CWE-911) with memory safety consequences that go beyond disclosure.
RemediationAI
Apply the upstream kernel stable patches: 04d9d1fc428ac9f581d55118d67e0cb546701feb, 1b20e61d36f490319d3gbdedd410155232ab5190, 856baaa100cd288d3685eedae9a129c996e7e755, 86d9b040421bbd26425f5a3edc226f57ecdecbfe, and ad105cde6b261b8b05ec872fe7d1987417d7fe5a, all accessible at https://git.kernel.org/stable/c/. Linux distribution vendors are expected to backport these fixes; check vendor-specific kernel security channels for patched package versions. No configuration option exists to disable the affected code path without eliminating TCP/DCCP networking, so patching is the only definitive fix. As a compensating control where patching is delayed, restrict local user access using seccomp profiles or LSM policies (AppArmor/SELinux) to deny the connect() syscall for untrusted workloads - this prevents triggering the vulnerable code path but may break legitimate socket use in constrained environments.
Same weakness CWE-200 – Information Exposure
View allSame technique Information Disclosure
View allShare
External POC / Exploit Code
Leaving vuln.today