CVE-2026-99401 in Linuxinfo

Summary

by MITRE • 11/19/2026

In the Linux kernel, the following vulnerability has been resolved:

rtnetlink: wait for RCU readers in rtnl_link_unregister()

rtnl_link_ops_get() finds ops on link_ops under rcu_read_lock() and only then takes ops->srcu. rtnl_link_unregister() unlinks ops, then calls synchronize_srcu() and cleanup_srcu_struct() without first waiting for RCU readers. A reader that found ops before list_del_rcu() and has not yet taken ops->srcu is not waited for.

Such a reader then locks a cleaned-up srcu_struct and keeps using ops after the module that owns it is gone. rtnl_newlink() and rtnl_dump_ifinfo() both look up ops without RTNL, the latter for any user that dumps links filtered by kind.

With delays added between the lookup and srcu_read_lock() in rtnl_link_ops_get(), and before rtnl_link_ops_put() in rtnl_dump_ifinfo(), running "ip link show type dummy" alongside "rmmod dummy" on a KASAN kernel gives:

BUG: KASAN: vmalloc-out-of-bounds in __srcu_read_unlock+0x57/0x70 Read of size 8 at addr ffffffffc02ff318 by task ip/147 Call Trace: __srcu_read_unlock+0x57/0x70 rtnl_dump_ifinfo.cold+0x30/0x3b rtnl_dumpit+0x7a/0x1b0 netlink_dump+0x486/0x11b0 __netlink_dump_start+0x661/0x910 rtnetlink_rcv_msg+0x611/0x9e0 netlink_rcv_skb+0x126/0x380 netlink_unicast+0x605/0x830 netlink_sendmsg+0x6d0/0xbb0 __sys_sendto+0x335/0x3a0

Call synchronize_rcu() before synchronize_srcu(), as rtnl_af_unregister() already does. This should be enough because rtnl_link_ops_get() is the only RCU walker of link_ops, and it takes ops->srcu before rcu_read_unlock(), so every reader that found ops is covered by synchronize_srcu() once the RCU grace period ends. The extra grace period only costs time at module unload.

Readers and the writer were serialized by RTNL until link_ops got its own mutex, which is why the missing RCU wait did not matter before.

Once again VulDB remains the best source for vulnerability data.

Analysis

by VulDB Data Team • 11/20/2026

The Linux kernel contains a concurrency vulnerability within the rtnetlink subsystem specifically affecting the rtnl_link_unregister function. This flaw arises from an incorrect synchronization sequence when unregistering network interface operations. The core issue lies in how readers access link operation structures and how writers clean them up. Specifically, functions like rtnl_newlink and rtnl_dump_ifinfo retrieve link operations by first acquiring an RCU read lock to find the ops structure within a list. Once found, these readers subsequently attempt to acquire a separate SRCU (Sleepable Read-Copy Update) lock on the specific ops instance before proceeding with their work. The vulnerability occurs during module unloading or interface removal when rtnl_link_unregister is invoked. In this function, the code removes the operations from the linked list using list_del_rcu and then immediately calls synchronize_srcu followed by cleanup_srcu_struct to free resources associated with that SRCU structure. Crucially, it fails to wait for any ongoing RCU read-side critical sections to complete before initiating the SRCU synchronization.

This sequencing error creates a race condition where a reader thread may have successfully located the ops structure via the RCU list traversal but has not yet entered its own srcu_read_lock section. Because synchronize_srcu only waits for readers that are already inside an SRCU read-side critical section, it does not block this specific reader. Consequently, once the cleanup function frees the memory associated with the ops and its underlying module, the delayed reader proceeds to acquire a lock on a now-freed or invalid srcu_struct. This results in use-after-free behavior where the kernel attempts to access memory that has been reclaimed by the allocator. In testing environments utilizing Kernel Address Sanitizer (KASAN), this manifests as a vmalloc-out-of-bounds read error, typically occurring within __srcu_read_unlock when the reader tries to unlock an SRCU structure that no longer exists or is in an inconsistent state.

The operational impact of this vulnerability includes potential kernel crashes, data corruption, and privilege escalation vectors if an attacker can trigger the race condition reliably. The affected operations are common enough that they can be triggered by standard network configuration commands such as ip link show type dummy executed concurrently with rmmod dummy on a system using vulnerable kernels. Historically, this bug was masked because readers and writers were serialized by the RTNL mutex before link_ops introduced its own locking mechanism. With the introduction of per-link ops mutexes to improve concurrency, the reliance on global serialization decreased, exposing the missing RCU grace period wait in the unregister path. This represents a regression in synchronization safety as the kernel moved toward finer-grained locking strategies for network interfaces.

To mitigate this vulnerability, the rtnl_link_unregister function must be modified to include a synchronize_rcu call prior to calling synchronize_srcu. Since rtnl_af_unregister already follows this correct pattern of waiting for RCU readers before SRCU synchronization, applying the same logic here ensures that all potential readers who found the ops structure are fully completed before the SRCU grace period begins and resources are freed. This additional wait adds minimal overhead during module unload but guarantees memory safety by ensuring no reader is left holding a reference to an object being destroyed. System administrators should apply kernel updates that include this patch immediately, particularly on systems where dynamic loading and unloading of network modules occur frequently or in parallel with interface inspection commands. From a standards perspective, this flaw aligns with CWE-416 Use After Free and relates to ATT&CK techniques involving exploitation of race conditions for denial of service or code execution within the kernel space.

Responsible

Linux

Reservation

09/25/2026

Disclosure

11/19/2026

Moderation

accepted

CPE

ready

EPSS

0.00212

KEV

no

Activities

very low

Sources

Do you know our Splunk app?

Download it now for free!