CVE-2025-39780

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> sched/ext: Fix invalid task state transitions on class switch<br /> <br /> When enabling a sched_ext scheduler, we may trigger invalid task state<br /> transitions, resulting in warnings like the following (which can be<br /> easily reproduced by running the hotplug selftest in a loop):<br /> <br /> sched_ext: Invalid task state transition 0 -&gt; 3 for fish[770]<br /> WARNING: CPU: 18 PID: 787 at kernel/sched/ext.c:3862 scx_set_task_state+0x7c/0xc0<br /> ...<br /> RIP: 0010:scx_set_task_state+0x7c/0xc0<br /> ...<br /> Call Trace:<br /> <br /> scx_enable_task+0x11f/0x2e0<br /> switching_to_scx+0x24/0x110<br /> scx_enable.isra.0+0xd14/0x13d0<br /> bpf_struct_ops_link_create+0x136/0x1a0<br /> __sys_bpf+0x1edd/0x2c30<br /> __x64_sys_bpf+0x21/0x30<br /> do_syscall_64+0xbb/0x370<br /> entry_SYSCALL_64_after_hwframe+0x77/0x7f<br /> <br /> This happens because we skip initialization for tasks that are already<br /> dead (with their usage counter set to zero), but we don&amp;#39;t exclude them<br /> during the scheduling class transition phase.<br /> <br /> Fix this by also skipping dead tasks during class swiching, preventing<br /> invalid task state transitions.