Re: Possible deadlock in pcpu_freelist_pop

[Cong Wang <xiyou.wangcong@xxxxxxxxx>]

On Thu, Sep 05, 2024 at 11:40:06AM -0700, Alexei Starovoitov wrote:
> On Thu, Sep 5, 2024 at 11:37 AM Priya Bala Govindasamy <pgovind2@xxxxxxx> wrote:
> >
> > SEC("kprobe/__pcpu_freelist_pop+0x58c")
> 
> We should disallow such recursion in the verifier.
> All these "bugs" are hard to prioritize as bugs.
> When people shot themselves in the foot there will be pain.

diff --git a/kernel/bpf/percpu_freelist.c b/kernel/bpf/percpu_freelist.c
index 034cf87b54e9..14c9fcf81ac6 100644
--- a/kernel/bpf/percpu_freelist.c
+++ b/kernel/bpf/percpu_freelist.c
@@ -187,6 +187,7 @@ struct pcpu_freelist_node *__pcpu_freelist_pop(struct pcpu_freelist *s)
                return ___pcpu_freelist_pop_nmi(s);
        return ___pcpu_freelist_pop(s);
 }
+NOKPROBE_SYMBOL(__pcpu_freelist_pop);

 struct pcpu_freelist_node *pcpu_freelist_pop(struct pcpu_freelist *s)
 {