Re: [PATCH bpf v3 3/5] bpf, sockmap: Fix af_unix iter deadlock

["Jiayuan Chen" <jiayuan.chen@linux.dev>]

March 6, 2026 at 22:06, "Michal Luczaj" <mhal@rbox.co mailto:mhal@rbox.co?to=%22Michal%20Luczaj%22%20%3Cmhal%40rbox.co%3E > wrote:


> 
> On 3/6/26 07:04, Jiayuan Chen wrote:
> 
> > 
> > On 3/6/26 7:30 AM, Michal Luczaj wrote:
> > 
> > > 
> > > @@ -3729,15 +3729,14 @@ static int bpf_iter_unix_seq_show(struct seq_file *seq, void *v)
> > >  struct bpf_prog *prog;
> > >  struct sock *sk = v;
> > >  uid_t uid;
> > >  - bool slow;
> > >  int ret;
> > >  
> > >  if (v == SEQ_START_TOKEN)
> > >  return 0;
> > >  
> > >  - slow = lock_sock_fast(sk);
> > >  + lock_sock(sk);
> > >  
> > >  - if (unlikely(sk_unhashed(sk))) {
> > >  + if (unlikely(sock_flag(sk, SOCK_DEAD))) {
> > >  ret = SEQ_SKIP;
> > >  goto unlock;
> > >  }
> > > 
> >  
> >  
> >  Switching to lock_sock() fixes the deadlock, but it does not provide mutual
> >  exclusion with unix_release_sock(), which uses unix_state_lock() exclusively
> >  and does not touch lock_sock() at all. So a dying socket can still reach the
> >  BPF prog concurrently with unix_release_sock() running on another CPU.
> > 
> That's right. Note that although the socket is dying, iter holds a
> reference to it, so the socket is far from being freed (as in: memory
> released).
> 
> > 
> > Both SOCK_DEAD and the clearing of unix_peer(sk) happen under
> >  unix_state_lock() in unix_release_sock(). Without taking unix_state_lock()
> >  before the SOCK_DEAD check, there is a window:
> > 
> >  iter unix_release_sock()
> >  --- lock_sock(sk)
> >  SOCK_DEAD == 0 (check passes)
> >  unix_state_lock(sk)
> >  unix_peer(sk) = NULL
> >  sock_set_flag(sk, SOCK_DEAD)
> >  unix_state_unlock(sk)
> >  BPF prog runs
> >  → accesses unix_peer(sk) == NULL → crash
> > 
> >  This was not raised in the v2 discussion.
> > 
> It was raised in v1[1]. Conclusion was that bpf prog bytecode directly
> accessing unix_peer(sk) is not an issue; bpf machinery will handle any
> faults. That said, should a "bad" value of unix_peer(sk) end up as a
> parameter of a bpf helper, yes, that is a well known[2] problem (that have
> a solution unrelated to this series).
> 
> [1]:
> https://lore.kernel.org/bpf/6de6f1bf-c8ee-4dfb-9b8c-f89185946630@linux.dev/
> [2]:
> https://lore.kernel.org/bpf/CAADnVQK_93g_KkNFYXSr8ZvA1fYh4hoFRJCJFPS-zs4ox0HhAA@mail.gmail.com/
>

Thanks for letting me know.