CVE-2026-93197 in Linux
Summary
by MITRE • 09/18/2026
In the Linux kernel, the following vulnerability has been resolved:
memcg: move LRU size accounting on reparenting instead of copying it
When a memory cgroup is offlined its LRU folios are reparented to the parent. lruvec_reparent_lru() splices the child's lists into the parent's and credits the parent with the child's per-zone lru_zone_size[], but never clears the child's copy, so the size is
copied rather than moved. lru_gen_reparent_memcg() does the same for MGLRU.
The parent is left correct, credited with exactly the folios it took over. The stale value sits on the child and nothing will correct it: folio->memcg_data now resolves to the parent, so every later update_lru_size() for those folios goes there.
Dying cgroups are not freed immediately and mem_cgroup_iter() still walks them, so shrink_lruvec() keeps being called on them. get_scan_count() reads the phantom counter through lruvec_lru_size() and the scan loop then grinds through nr[] in SWAP_CLUSTER_MAX steps against
an empty list, for as long as the dead cgroup lives. Under MGLRU the MGLRU scanner runs instead, but count_shadow_nodes() sums all of NR_LRU_LISTS through lruvec_lru_size() and over-budgets the shadow node limit just the same.
On one 251 GiB host a sweep of every mz->lru_zone_size[] found 380
counters describing folios on no list at all: 124777314 pages, 476 GiB, 1.89x the machine's RAM, across 57 cgroups. All were on memcgs with CSS_DYING set and CSS_ONLINE clear, and parent/child pairs reported byte-identical sizes.
LRU_UNEVICTABLE needs its size moved too. Its list is deliberately not spliced because lruvec_init() poisons the head - the unevictable LRU is imaginary and folios are never threaded on it - but the size is kept by lruvec_add_folio()/lruvec_del_folio() and those folios account to the parent from here on.
This depends on commit bf4ade7dbd76 ("memcg: keep folio's objcg same as its node") and must not be backported ahead of it. Without that invariant a folio's objcg can belong to another node, so a folio already spliced onto the parent's list can still resolve to the child's lruvec until the objcg's node is reparented in a later iteration of memcg_reparent_objcgs(); clearing the child's counter early then lets lruvec_del_folio() underflow it and trip the WARN_ONCE()/VM_BUG_ON() in mem_cgroup_update_lru_size().
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 09/18/2026
The Linux kernel memory control group subsystem contains a critical accounting flaw related to how Large Page Replacement List sizes are handled during cgroup reparenting operations. When a memory cgroup is taken offline, its associated LRU folios must be transferred to the parent cgroup to maintain accurate system-wide resource tracking. The existing implementation attempts this transfer by splicing the child's lists into the parent and crediting the parent with the corresponding per-zone sizes. However, the code fails to clear these size counters on the child side after the transfer is complete. This results in a duplication of accounting data where both the parent and the dying child cgroup retain identical records for the same set of memory pages. This error persists because folio metadata continues to resolve to the parent cgroup for subsequent updates, leaving the stale values trapped within the structure of the offlined child cgroup with no mechanism to automatically correct them during normal operation.
The operational impact of this double-counting is severe and manifests as significant resource exhaustion and performance degradation on systems utilizing memory cgroups extensively. Because dying cgroups are not immediately freed from memory, iteration functions continue to process them. Consequently, the LRU shrinker logic repeatedly invokes scan routines against these stale counters. The system attempts to reclaim pages based on inflated size metrics that no longer correspond to actual active lists in memory. This leads to excessive CPU consumption as the kernel performs futile scanning operations over empty or non-existent page lists. In extreme cases observed on large hosts, this bug has resulted in phantom accounting values exceeding physical RAM capacity by nearly double, affecting dozens of cgroups simultaneously and degrading overall system stability through wasted computational resources and inaccurate memory pressure signals.
From a vulnerability classification perspective, this issue aligns with CWE-665 Improper Initialization as the root cause lies in failing to properly initialize or clear state variables during object lifecycle transitions. The resulting behavior also reflects aspects of CWE-401 Missing Release of Memory after Effective Lifetime since the accounting structures retain references to resources that are no longer exclusively owned by the dying cgroup, leading to resource inflation. In terms of attack surface and mitigation strategies, this is primarily a denial-of-service vector triggered by specific workload patterns involving frequent creation and deletion of memory cgroups with high page activity. Mitigation involves applying kernel patches that ensure atomicity in reparenting operations, specifically moving rather than copying size accounting data. Security administrators should prioritize updating to kernel versions containing the fix for lruvec_reparent_lru and related functions while ensuring dependencies on prior commits regarding object group node consistency are met before deployment to avoid introducing new underflow vulnerabilities during the transition period.