Skip to main content
Vulnerability Database/CVE-2026-98028

CVE-2026-98028: Linux Kernel Privilege Escalation Vulnerability

CVE-2026-98028 is a privilege escalation vulnerability in the Linux kernel affecting the nfp network driver. This flaw causes list corruption when rule reprogramming fails. This article covers technical details, affected versions, impact assessment, and mitigation strategies.

Published:

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

CVE-2026-98028 Overview

CVE-2026-98028 affects the Linux kernel's Netronome Flow Processor (nfp) Ethernet driver. The flaw lives in nfp_net_fs_add(), which replaces an existing flow steering rule by removing the old hardware entry, decrementing nn->fs.count, and programming the new rule. When nfp_net_fs_add_hw() fails, the old entry remains on nn->fs.list because only the success path reaches list_replace(). The list then holds one more entry than nn->fs.count reports, and the driver advertises a rule whose hardware state has already been torn down.

Critical Impact

Userspace receives a persistent -EMSGSIZE from ETHTOOL_GRXCLSRLALL because the buffer sized via ETHTOOL_GRXCLSRLCNT is one entry short, breaking ethtool flow-classifier management on affected adapters.

Affected Products

  • Linux kernel networking subsystem — Netronome Flow Processor (nfp) Ethernet driver
  • Stable kernel branches referenced in the upstream fix commits
  • Systems using nfp adapters with ethtool receive flow classifier (RX NFC) rules

Discovery Timeline

  • 2026-09-25 - CVE-2026-98028 published to NVD
  • 2026-09-25 - Last updated in NVD database

Technical Details for CVE-2026-98028

Vulnerability Analysis

The nfp driver maintains a software list nn->fs.list and a counter nn->fs.count that must stay synchronized with the hardware's flow steering table. nfp_net_fs_add() implements a replace operation that proceeds in stages: delete the current hardware entry, decrement the counter, program the replacement, and then swap the list node via list_replace().

If nfp_net_fs_add_hw() fails after the hardware delete, the function returns without executing list_replace(). The stale list node remains linked while the counter and hardware state have already advanced. The driver now reports a rule count that is lower than the number of entries the GRXCLSRLALL walk will produce.

Userspace tools such as ethtool first call ETHTOOL_GRXCLSRLCNT to size an allocation, then call ETHTOOL_GRXCLSRLALL to enumerate the rules. Because the walk is bounded, the undersized buffer causes the kernel to return -EMSGSIZE on every subsequent enumeration. Nothing resyncs the counter, so the condition persists across the lifetime of the interface.

Root Cause

The root cause is incomplete error handling in nfp_net_fs_add(). The replace sequence does not roll back the counter or remove the stale list entry when the hardware reprogramming step fails. The resulting mismatch between nn->fs.count and the actual contents of nn->fs.list is a state consistency defect in the driver.

Attack Vector

This defect is reachable through standard ethtool flow steering operations on an nfp interface. A privileged local user replacing an existing RX classifier rule can trigger the mismatch when the hardware programming path returns an error. The impact is a persistent denial of service against ethtool rule enumeration; no memory corruption occurs because the GRXCLSRLALL walk is bounded. Earlier logic without the bound would have written one u32 past the allocation.

See the upstream fix commits for the exact code paths: Kernel Git Commit 108bb21, Kernel Git Commit 552e528, Kernel Git Commit 603b700, and Kernel Git Commit edaec27.

Detection Methods for CVE-2026-98028

Indicators of Compromise

  • Persistent -EMSGSIZE errors returned by ethtool -n <iface> or ethtool --show-nfc <iface> on interfaces backed by the nfp driver.
  • Discrepancy between the rule count reported by ETHTOOL_GRXCLSRLCNT and the number of entries present when walking nn->fs.list.
  • Failed nfp_net_fs_add_hw() events in kernel logs correlated with subsequent ethtool enumeration failures.

Detection Strategies

  • Inventory running kernels and identify systems that have not yet picked up the stable backports referenced in the upstream commits.
  • Audit configuration management for scripts that add or replace nfp RX classifier rules and capture their return codes.
  • Correlate dmesg output for nfp driver errors with telemetry on ethtool invocations to spot the mismatch pattern.

Monitoring Recommendations

  • Alert on repeated -EMSGSIZE return codes from monitoring agents that call ethtool to inventory flow classifier rules.
  • Monitor kernel ring buffer messages originating from the nfp driver, especially around flow steering programming failures.
  • Track kernel package versions across the fleet to confirm patched builds are deployed to hosts with nfp adapters.

How to Mitigate CVE-2026-98028

Immediate Actions Required

  • Apply the stable kernel updates that contain the upstream fixes referenced by commits 108bb21, 552e528, 603b700, and edaec27.
  • Reboot affected hosts after patching to load the corrected nfp driver.
  • Restart or reinitialize affected interfaces to resync nn->fs.count with nn->fs.list on systems that cannot be patched immediately.

Patch Information

The fix updates nfp_net_fs_add() so that the replaced rule is removed from nn->fs.list when nfp_net_fs_add_hw() fails, keeping the list and counter consistent. Backports are available across supported stable branches via the referenced git.kernel.org commits. Distribution vendors will ship the fix through their standard kernel update channels.

Workarounds

  • Avoid using ethtool replace operations on nfp RX classifier rules on unpatched systems; delete and re-add rules instead to limit exposure to the failing code path.
  • Reload the nfp driver or bring the interface down and up to clear the inconsistent state when -EMSGSIZE errors appear.
  • Restrict privileged access to hosts with nfp adapters so that only trusted operators can modify flow steering configuration.
bash
# Verify kernel version and reload the nfp interface to clear inconsistent state
uname -r
ip link set dev <iface> down
modprobe -r nfp
modprobe nfp
ip link set dev <iface> up
ethtool -n <iface>

Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

Sources: National Vulnerability Database, CISA KEV, Exploit-DB

Experience the Most Advanced Cybersecurity Platform

See how the world’s most intelligent, autonomous cybersecurity platform can protect your organization today and into the future.