CVE-2022-48658

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> mm: slub: fix flush_cpu_slab()/__free_slab() invocations in task context.<br /> <br /> Commit 5a836bf6b09f ("mm: slub: move flush_cpu_slab() invocations<br /> __free_slab() invocations out of IRQ context") moved all flush_cpu_slab()<br /> invocations to the global workqueue to avoid a problem related<br /> with deactivate_slab()/__free_slab() being called from an IRQ context<br /> on PREEMPT_RT kernels.<br /> <br /> When the flush_all_cpu_locked() function is called from a task context<br /> it may happen that a workqueue with WQ_MEM_RECLAIM bit set ends up<br /> flushing the global workqueue, this will cause a dependency issue.<br /> <br /> workqueue: WQ_MEM_RECLAIM nvme-delete-wq:nvme_delete_ctrl_work [nvme_core]<br /> is flushing !WQ_MEM_RECLAIM events:flush_cpu_slab<br /> WARNING: CPU: 37 PID: 410 at kernel/workqueue.c:2637<br /> check_flush_dependency+0x10a/0x120<br /> Workqueue: nvme-delete-wq nvme_delete_ctrl_work [nvme_core]<br /> RIP: 0010:check_flush_dependency+0x10a/0x120[ 453.262125] Call Trace:<br /> __flush_work.isra.0+0xbf/0x220<br /> ? __queue_work+0x1dc/0x420<br /> flush_all_cpus_locked+0xfb/0x120<br /> __kmem_cache_shutdown+0x2b/0x320<br /> kmem_cache_destroy+0x49/0x100<br /> bioset_exit+0x143/0x190<br /> blk_release_queue+0xb9/0x100<br /> kobject_cleanup+0x37/0x130<br /> nvme_fc_ctrl_free+0xc6/0x150 [nvme_fc]<br /> nvme_free_ctrl+0x1ac/0x2b0 [nvme_core]<br /> <br /> Fix this bug by creating a workqueue for the flush operation with<br /> the WQ_MEM_RECLAIM bit set.