Skip to article
Datapath technology background for Linux server patch planning and verification
Back to News
CYBERSECURITY • Published November 2, 2026 • 8 min read • Source: Atlas report and primary vendor advisories

CVE-2026-98432: When Is a Linux Server Patched?

Review the Linux TCP flaw, vendor patch status, and container-host exposure. Datapath explains how to verify an active fix before closing a patch ticket.

Dan J Sturdivant, Vice President at Datapath

By

Vice President

cybersecuritydata securitymanaged IT

Key takeaways

  • CVE-2026-98432 affects Linux kernel TCP handling; verify the distribution, running kernel, and vendor fix rather than relying on a version headline.
  • A demonstrated research exploit and confirmed attacks in the wild are different evidence; network exposure and configuration still matter.
  • Container nodes and cloud VMs need an accountable patch plan, with evidence that the supported fix is active before the ticket closes.

Original source

Atlas report and primary vendor advisories

What should a business do about CVE-2026-98432?

Identify Linux systems that may run an affected kernel, check their distribution’s current advisory, and assign a safe update and activation plan. At Datapath, we recommend treating a patch ticket as complete only when the responsible team can show that the supported fix is active and important services still work. Downloading a package is not that evidence.

Source checks: November 2, 2026. The originating Atlas report brings attention to the Linux IPv4 TCP use-after-free and the gap between upstream fixes and distribution delivery. This briefing uses primary kernel, researcher, and vendor notices for the details below. Vendor status can change after this snapshot; recheck the linked advisory for the exact product and kernel flavor before making a change.

What is the vulnerability, in plain English?

A use-after-free occurs when software releases a piece of memory while another operation can still use an old reference to it. Think of a shared work item being discarded before a colleague finishes reading it. In a kernel, that ownership error can have consequences beyond one application because the kernel manages the host’s resources and networking.

The Linux kernel fix addresses IPv4 socket-option memory used during TCP connection handling. Its explanation describes concurrent operations that can still read the options after they have been freed. The change defers freeing the memory until the protected readers are finished.

The official CVE record, assigned by the Linux kernel CNA, rates the flaw Critical at CVSS 3.1 9.8. The record traces the affected upstream code to Linux 4.4; assess distribution builds against their own advisories and backports. That assessment warrants urgent attention; it does not establish that every Linux deployment has the same reachable services, configuration, or exploitability. Vendors may use different severity assessments for their products.

Does the research show every Linux server can be taken over?

No. Anthropic’s original research describes a laboratory exploit that achieved remote root command execution on Linux servers with a multipath TCP listener. That is evidence of a demonstrated exploit under the reported conditions, not a report that attackers have compromised every affected installation.

The Go Security Team’s notice says Linux listening sockets can use MPTCP by default starting with Go 1.24, subject to program settings and explicit socket configuration. It also warns that the underlying bug is not specific to MPTCP and that disabling it program by program is fragile. Inventory actual listeners and configurations; a language-version label or absence of one known exploit prerequisite does not prove a host is unaffected.

AWS’s security bulletin, reviewed November 2, says AWS is unaware of exploitation in the wild. That is a dated statement from AWS, not universal proof that no exploitation has occurred. Keep laboratory results, public vulnerability information, and confirmed incident evidence separate when briefing leadership.

The principal IPv4 issue should also not be turned into a blanket claim that IPv6 is safe. The upstream patch discussion identifies a separate IPv6 request-migration concern. Follow the applicable vendor guidance for the actual network configuration rather than treating a protocol label as a complete risk assessment.

Which systems should go into the exposure inventory?

Include internet-facing servers, internal services reachable by untrusted peers, cloud VMs, Kubernetes or other container nodes, and vendor appliances that run a relevant Linux kernel. Record the operating-system release, kernel flavor, running kernel, installed kernel package, application owner, exposure, and the advisory being used to judge status. An application inventory alone may not identify the host responsible for the vulnerable networking code.

Docker’s container documentation explains that containers share a kernel, while a VM has its own. For ordinary Linux containers, rebuilding the application image does not update the node or VM kernel underneath it. Coordinate with the platform owner instead of assigning the same host-level remediation separately to every application team.

Managed cloud services need a clear division of responsibility too. Confirm who maintains the underlying host and what maintenance evidence the provider offers. AWS’s bulletin notes that Network Load Balancer TCP listeners pass IP options through; do not assume the presence of a cloud load balancer removes this exposure.

What do vendor trackers show as of November 2?

Use the vendor’s package and support channel for the precise release. Distribution kernels can carry backported fixes without adopting the upstream kernel’s version number. The table is a checked snapshot of selected products, not a declaration that every variant in a distribution has identical status.

Vendor or distributionObserved advisory statusWhat to verify for your system
UbuntuMain linux rows for 26.10, 26.04, 24.04 and 22.04 show vulnerable/work in progress.Check the exact kernel flavor, including cloud and hardware-enablement packages. A work-in-progress version is not confirmation of a released fix.
Red HatThe current tracker lists November 1 advisories for RHEL 8, 9 and 10: RHSA-2026:93414, RHSA-2026:93412 and RHSA-2026:93413 respectively.Match the exact product, release and kernel flavor to its advisory. The tracker still lists affected variants, including OpenShift/CoreOS entries; a RHEL fix does not establish that every layered product is patched.
AlmaLinuxIts October 30 announcement describes fixed kernels for 8, 9 and 10 in testing. This historical announcement alone does not confirm the current production-repository status.Verify the current production errata and enabled repository before selecting a package; do not assume a testing build is approved for production.
DebianSecurity packages include bookworm 6.1.190-1 and trixie 6.12.113-1 as fixed.Confirm the security update for the relevant source package is installed and active; ordinary release-package rows may differ.
Amazon LinuxThe tracker marks its listed Amazon Linux 2 and 2023 kernel families fixed.Match the installed package and advisory; this does not mean every customer VM has already applied the update.

For self-maintained upstream kernels, the CVE record identifies fixed releases including 5.10.272, 5.15.223, 6.1.190, 6.6.159, 6.12.113, 6.18.56, 7.2.10 and 7.3. Those branch boundaries are not a substitute for a distribution’s backport record. An older-looking version string can include a vendor fix, while a downloaded new package may not be the kernel currently running.

What evidence should close the patch ticket?

Plan the change around the application’s availability requirements. Identify service dependencies, confirm a recovery path, schedule the necessary maintenance, and use the vendor-supported update route. For clustered platforms, agree how nodes will be maintained while preserving required capacity. These are planning steps, not a reason to postpone an exposed system indefinitely.

After installing a fixed kernel, boot into that kernel through the approved maintenance process and verify what is running. If a vendor-supported live-patching route is proposed, confirm that it explicitly covers this CVE and the deployed kernel; do not assume live-patch coverage exists. Validate important application and network behavior after the change.

Keep the advisory, affected asset, fixed package, activation evidence, validation result, owner, and completion time together. If a host remains pending, record the reason, responsible person, interim controls, and next review. Our incident response runbook guidance explains how a lean team can establish escalation and recovery handoffs before an urgent problem.

What if a supported patch is not available yet?

Review interim controls with the vendor and the responsible network or platform team. Red Hat’s bulletin says traffic filtering can reduce exposure but cannot fully prevent exploitation and is not a substitute for patching. It also warns that controls must be reviewed and tested against operational requirements before deployment.

Choose vendor-reviewed controls for your environment. Network paths, container routing, existing listeners, and application requirements differ. Reducing unnecessary access may help, but an untested change can disrupt a service or leave an important path uncovered. Give temporary controls an owner and an expiry or review trigger, and revisit them when the supported fix becomes available.

How can Datapath help organize the response?

Datapath helps turn vulnerability findings into accountable operational work. Our managed cybersecurity services include monitoring, alert triage, response coordination, vulnerability remediation, and leadership reporting. Our co-managed IT services provide a practical route for coordinating with an internal IT team; confirm the scope of support for your environment.

Bring the affected-host inventory, vendor advisories, maintenance constraints, and the services those hosts support. We can help prioritize the review, clarify ownership, and connect remediation with verification and response planning. The goal is evidence that the approved fix is active, with remaining exceptions visible to the people responsible for them.

Need help turning a Linux advisory into a patch plan?

Talk with Datapath about host inventory, exposure priorities, maintenance ownership, and verification evidence.

Discuss your patch priorities

Frequently asked questions

Does a container image update fix this host-kernel vulnerability?

Not by itself. Ordinary Linux containers share the kernel of their node or VM. The responsible platform team must verify the relevant host-kernel fix and its activation, alongside any separate application updates.

Does an upstream fixed version prove my distribution is patched?

No. Check the exact vendor package, release and kernel flavor, including backported fixes. Then confirm the supported fix is active on the host. Installed packages and the currently running kernel can differ.

Does a working research exploit prove attacks are occurring in the wild?

No. A laboratory exploit demonstrates the reported conditions and outcome. Confirmed attacks require separate incident evidence. Attribute dated vendor statements about exploitation and keep monitoring for new information.

Disclaimer: This news summary is intended for informational and marketing purposes only, and nothing presented here is contractually binding or necessarily the final opinion of the authors.

Need to turn industry change into an IT plan?

Datapath can help translate security, compliance, and infrastructure signals into practical next steps for your organization.

Book an IT Consultation