CVE-2025-40163

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> sched/deadline: Stop dl_server before CPU goes offline<br /> <br /> IBM CI tool reported kernel warning[1] when running a CPU removal<br /> operation through drmgr[2]. i.e "drmgr -c cpu -r -q 1"<br /> <br /> WARNING: CPU: 0 PID: 0 at kernel/sched/cpudeadline.c:219 cpudl_set+0x58/0x170<br /> NIP [c0000000002b6ed8] cpudl_set+0x58/0x170<br /> LR [c0000000002b7cb8] dl_server_timer+0x168/0x2a0<br /> Call Trace:<br /> [c000000002c2f8c0] init_stack+0x78c0/0x8000 (unreliable)<br /> [c0000000002b7cb8] dl_server_timer+0x168/0x2a0<br /> [c00000000034df84] __hrtimer_run_queues+0x1a4/0x390<br /> [c00000000034f624] hrtimer_interrupt+0x124/0x300<br /> [c00000000002a230] timer_interrupt+0x140/0x320<br /> <br /> Git bisects to: commit 4ae8d9aa9f9d ("sched/deadline: Fix dl_server getting stuck")<br /> <br /> This happens since:<br /> - dl_server hrtimer gets enqueued close to cpu offline, when<br /> kthread_park enqueues a fair task.<br /> - CPU goes offline and drmgr removes it from cpu_present_mask.<br /> - hrtimer fires and warning is hit.<br /> <br /> Fix it by stopping the dl_server before CPU is marked dead.<br /> <br /> [1]: https://lore.kernel.org/all/8218e149-7718-4432-9312-f97297c352b9@linux.ibm.com/<br /> [2]: https://github.com/ibm-power-utilities/powerpc-utils/tree/next/src/drmgr<br /> <br /> [sshegde: wrote the changelog and tested it]