From 6668ed271eaefaa63e686bdfbedaeb7b8e492722 Mon Sep 17 00:00:00 2001 From: Kumar Kartikeya Dwivedi Date: Fri, 4 Sep 2026 10:43:19 +0200 Subject: selftests/bpf: Reject graph kptr use after RCU unlock Add a sleepable verifier test that loads a graph-node local kptr in an explicit RCU read-side critical section, then passes its node to bpf_rbtree_remove() after the section ends. Before the verifier fix, the stale NON_OWN_REF flag makes the node look like a live borrowed reference and the program is accepted. After the fix, the pointer is demoted without NON_OWN_REF and the graph kfunc argument is rejected. Also exercise a graph kptr loaded while a spin lock provides implicit RCU protection. The pointer must be invalidated when the lock is released, which guards the required ordering between non-owning-reference invalidation and RCU demotion. Update the existing fault-protected load test state description. The post-unlock pointer no longer carries NON_OWN_REF, but remains readable because the load is rewritten to use BPF_PROBE_MEM. Signed-off-by: Kumar Kartikeya Dwivedi Link: https://lore.kernel.org/r/20260904084325.52250-7-memxor@gmail.com Signed-off-by: Alexei Starovoitov --- include/net/cls_cgroup.h | 88 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 88 insertions(+) create mode 100644 include/net/cls_cgroup.h (limited to 'include/net/cls_cgroup.h') diff --git a/include/net/cls_cgroup.h b/include/net/cls_cgroup.h new file mode 100644 index 000000000..668aeee9b --- /dev/null +++ b/include/net/cls_cgroup.h @@ -0,0 +1,88 @@ +/* SPDX-License-Identifier: GPL-2.0-or-later */ +/* + * cls_cgroup.h Control Group Classifier + * + * Authors: Thomas Graf + */ + +#ifndef _NET_CLS_CGROUP_H +#define _NET_CLS_CGROUP_H + +#include +#include +#include +#include +#include + +#ifdef CONFIG_CGROUP_NET_CLASSID +struct cgroup_cls_state { + struct cgroup_subsys_state css; + u32 classid; +}; + +struct cgroup_cls_state *task_cls_state(struct task_struct *p); + +static inline u32 task_cls_classid(struct task_struct *p) +{ + u32 classid; + + if (in_interrupt()) + return 0; + + rcu_read_lock(); + classid = container_of(task_css(p, net_cls_cgrp_id), + struct cgroup_cls_state, css)->classid; + rcu_read_unlock(); + + return classid; +} + +static inline void sock_update_classid(struct sock_cgroup_data *skcd) +{ + u32 classid; + + classid = task_cls_classid(current); + sock_cgroup_set_classid(skcd, classid); +} + +static inline u32 __task_get_classid(struct task_struct *task) +{ + return task_cls_state(task)->classid; +} + +static inline u32 task_get_classid(const struct sk_buff *skb) +{ + u32 classid = __task_get_classid(current); + + /* Due to the nature of the classifier it is required to ignore all + * packets originating from softirq context as accessing `current' + * would lead to false results. + * + * This test assumes that all callers of dev_queue_xmit() explicitly + * disable bh. Knowing this, it is possible to detect softirq based + * calls by looking at the number of nested bh disable calls because + * softirqs always disables bh. + */ + if (softirq_count()) { + struct sock *sk = skb_to_full_sk(skb); + + /* If there is an sock_cgroup_classid we'll use that. */ + if (!sk || !sk_fullsock(sk)) + return 0; + + classid = sock_cgroup_classid(&sk->sk_cgrp_data); + } + + return classid; +} +#else /* !CONFIG_CGROUP_NET_CLASSID */ +static inline void sock_update_classid(struct sock_cgroup_data *skcd) +{ +} + +static inline u32 task_get_classid(const struct sk_buff *skb) +{ + return 0; +} +#endif /* CONFIG_CGROUP_NET_CLASSID */ +#endif /* _NET_CLS_CGROUP_H */ -- cgit v1.3.1