On Tue, May 27, 2025 at 1:38 AM Feng Yang <yangfeng59949@xxxxxxx> wrote: > > syzbot found the following issue on https://lore.kernel.org/all/683428c7.a70a0220.29d4a0.0800.GAE@xxxxxxxxxx/ > > Related source code: > BPF_CALL_0(bpf_get_cgroup_classid_curr) > { > return __task_get_classid(current); > } > > const struct bpf_func_proto bpf_get_cgroup_classid_curr_proto = { > .func = bpf_get_cgroup_classid_curr, > .gpl_only = false, > .ret_type = RET_INTEGER, > }; > > static inline u32 __task_get_classid(struct task_struct *task) > { > return task_cls_state(task)->classid; > } > > struct cgroup_cls_state *task_cls_state(struct task_struct *p) > { > return css_cls_state(task_css_check(p, net_cls_cgrp_id, > rcu_read_lock_bh_held())); > } > > > So, do I need to move bpf_get_cgroup_classid_curr_proto back from bpf_base_func_proto, or is there a better solution? I'd try to fix that rcu_read_lock_bh_held() check. Can we use rcu_read_lock_any_held() instead? >
