CVE-2024-50138

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> bpf: Use raw_spinlock_t in ringbuf<br /> <br /> The function __bpf_ringbuf_reserve is invoked from a tracepoint, which<br /> disables preemption. Using spinlock_t in this context can lead to a<br /> "sleep in atomic" warning in the RT variant. This issue is illustrated<br /> in the example below:<br /> <br /> BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:48<br /> in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 556208, name: test_progs<br /> preempt_count: 1, expected: 0<br /> RCU nest depth: 1, expected: 1<br /> INFO: lockdep is turned off.<br /> Preemption disabled at:<br /> [] migrate_enable+0xc0/0x39c<br /> CPU: 7 PID: 556208 Comm: test_progs Tainted: G<br /> Hardware name: Qualcomm SA8775P Ride (DT)<br /> Call trace:<br /> dump_backtrace+0xac/0x130<br /> show_stack+0x1c/0x30<br /> dump_stack_lvl+0xac/0xe8<br /> dump_stack+0x18/0x30<br /> __might_resched+0x3bc/0x4fc<br /> rt_spin_lock+0x8c/0x1a4<br /> __bpf_ringbuf_reserve+0xc4/0x254<br /> bpf_ringbuf_reserve_dynptr+0x5c/0xdc<br /> bpf_prog_ac3d15160d62622a_test_read_write+0x104/0x238<br /> trace_call_bpf+0x238/0x774<br /> perf_call_bpf_enter.isra.0+0x104/0x194<br /> perf_syscall_enter+0x2f8/0x510<br /> trace_sys_enter+0x39c/0x564<br /> syscall_trace_enter+0x220/0x3c0<br /> do_el0_svc+0x138/0x1dc<br /> el0_svc+0x54/0x130<br /> el0t_64_sync_handler+0x134/0x150<br /> el0t_64_sync+0x17c/0x180<br /> <br /> Switch the spinlock to raw_spinlock_t to avoid this error.