概要

In the Linux kernel, the following vulnerability has been resolved:memcg: protect concurrent access to mem_cgroup_idrCommit 73f576c04b94 ( mm: memcontrol: fix cgroup creation failure aftermany small jobs ) decoupled the memcg IDs from the CSS ID space to fix thecgroup creation failures. It introduced IDR to maintain the memcg IDspace. The IDR depends on external synchronization mechanisms formodifications. For the mem_cgroup_idr, the idr_alloc() and idr_replace()happen within css callback and thus are protected through cgroup_mutexfrom concurrent modifications. However idr_remove() for mem_cgroup_idrwas not protected against concurrency and can be run concurrently fordifferent memcgs when they hit their refcnt to zero. Fix that.We have been seeing list_lru based kernel crashes at a low frequency inour fleet for a long time. These crashes were in different part oflist_lru code including list_lru_add(), list_lru_del() and reparentingcode. Upon further inspection, it looked like for a given object (dentryand inode), the super_block s list_lru didn t have list_lru_one for thememcg of that object. The initial suspicions were either the object isnot allocated through kmem_cache_alloc_lru() or somehowmemcg_list_lru_alloc() failed to allocate list_lru_one() for a memcg butreturned success. No evidence were found for these cases.Looking more deeply, we started seeing situations where valid memcg s idis not present in mem_cgroup_idr and in some cases multiple valid memcgshave same id and mem_cgroup_idr is pointing to one of them. So, the mostreasonable explanation is that these situations can happen due to racebetween multiple idr_remove() calls or race betweenidr_alloc()/idr_replace() and idr_remove(). These races are causingmultiple memcgs to acquire the same ID and then offlining of one of themwould cleanup list_lrus on the system for all of them. Later access fromother memcgs to the list_lru cause crashes due to missing list_lru_one.

CVSS v3 指标

安全公告

影响产品