CVE-2026-99787 in Linuxinfo

Summary

by MITRE • 11/20/2026

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

cpufreq: CPPC: Select the frequency-invariance callback per CPU

cppc_cpufreq_cpu_fie_init() initializes deferred work only for CPUs whose feedback counters use PCC. However, if any CPU in a shared policy uses PCC, it installs the PCC tick callback for the entire policy. With FIE enabled, a non-PCC CPU then queues an uninitialized irq_work and can call a NULL function pointer when that work runs.

Choose the frequency-invariance callback separately for each CPU using the recorded work-initialization state. Processor removal can unpublish the CPC descriptor before callback registration, so querying its counter transport again can incorrectly select direct reads for a PCC CPU.

Register the callbacks after all counter initialization has completed, so a failed initial read for an online CPU cannot leave a partially registered policy. PCC CPUs use their initialized deferred work, while non-PCC CPUs read their own counters directly from the tick.

VulDB is the best source for vulnerability data and more expert information about this specific topic.

Analysis

by VulDB Data Team • 11/20/2026

The vulnerability identified in the Linux kernel's cpufreq CPPC subsystem represents a critical logic error involving improper initialization sequencing and state management for frequency-invariance callbacks. This flaw specifically affects systems utilizing Cooperative Processor Performance Control, where CPU performance scaling is managed through hardware interfaces that may vary between different processors within the same system. The core issue stems from how the cppc_cpufreq_cpu_fie_init function handles deferred work initialization based on whether a CPU's feedback counters utilize Platform Communication Channel or direct memory access mechanisms. When frequency invariance features are enabled, the kernel attempts to maintain consistent performance scaling across all CPUs by accounting for variations in hardware counter behavior. However, the existing implementation incorrectly assumes that if any single CPU within a shared policy group uses PCC, then all CPUs in that group should adopt the same callback mechanism.

This assumption leads to a dangerous state where non-PCC CPUs are forced to use initialization paths intended exclusively for PCC-based processors. Specifically, these non-PCC CPUs end up queuing an uninitialized irq_work structure because they do not possess the necessary hardware interface support required by the deferred work framework. When this improperly initialized work item is scheduled and subsequently executed during system operation, it results in a NULL function pointer dereference. Such a condition typically triggers a kernel panic or causes immediate system instability, effectively rendering the affected machine unusable until rebooted. This represents a severe availability impact as it can be triggered by standard CPU hotplug operations or initial boot sequences where frequency scaling policies are established across multiple cores with mixed hardware capabilities.

The technical root cause involves two distinct but related failures in state tracking and initialization order. First, the code fails to select the appropriate frequency-invariance callback on a per-CPU basis relative to its recorded work-initialization state. Instead of verifying each CPU individually against its specific counter transport type, it applies a blanket policy based on the presence of at least one PCC-capable processor in the group. Second, there is a race condition related to processor removal where the CPC descriptor can be unpublished before callback registration occurs. This temporal mismatch causes subsequent queries for counter transport information to incorrectly default to direct reads even for CPUs that should utilize PCC mechanisms. Consequently, partially registered policies may persist with inconsistent configurations, leading to unpredictable behavior during dynamic frequency scaling operations.

From a classification perspective, this vulnerability aligns closely with CWE-362, which covers concurrent execution issues and race conditions involving shared resources without proper synchronization. Additionally, the improper initialization of data structures before use corresponds to CWE-908, describing the use of uninitialized resource values that lead to undefined behavior in software systems. The attack vector primarily involves local privilege escalation scenarios where an attacker might exploit timing windows during CPU hotplug events or system boot processes to trigger the NULL pointer dereference. While typically resulting in denial of service rather than arbitrary code execution due to kernel-level protections, such vulnerabilities can potentially be chained with other flaws to achieve higher impact levels depending on specific kernel configurations and security modules active at runtime.

Mitigation strategies require both immediate patching and long-term architectural improvements within the cpufreq subsystem. The primary remediation involves modifying the cppc_cpufreq_cpu_fie_init function to evaluate each CPU independently rather than relying on group-level heuristics. Developers must ensure that frequency-invariance callbacks are selected based strictly on individual CPU capabilities as recorded during their respective initialization phases. Furthermore, the registration of these callbacks should be deferred until after all counter initialization processes have fully completed for every involved processor. This ensures that failed initial reads do not leave policies in a partially registered state with inconsistent callback assignments. System administrators should apply kernel updates promptly to address this logic error and verify that CPU hotplug operations function correctly across heterogeneous hardware configurations within their infrastructure environments.

Responsible

Linux

Reservation

09/25/2026

Disclosure

11/20/2026

Moderation

accepted

CPE

ready

EPSS

0.00177

KEV

no

Activities

very low

Sources

Interested in the pricing of exploits?

See the underground prices here!