Re: [PATCH net] bpf, sockmap: Restore sk_prot ops when psock is removed from sockmap

[Stanislav Fomichev <stfomichev@gmail.com>]

On 03/05, Cong Wang wrote:
> On Wed, Mar 05, 2025 at 10:02:34PM +0800, Dong Chenchen wrote:
> > WARNING: CPU: 0 PID: 6558 at net/core/sock_map.c:1703 sock_map_close+0x3c4/0x480
> > Modules linked in:
> > CPU: 0 UID: 0 PID: 6558 Comm: syz-executor.14 Not tainted 6.14.0-rc5+ #238
> > RIP: 0010:sock_map_close+0x3c4/0x480
> > Call Trace:
> >  <TASK>
> >  inet_release+0x144/0x280
> >  __sock_release+0xb8/0x270
> >  sock_close+0x1e/0x30
> >  __fput+0x3c6/0xb30
> >  __fput_sync+0x7b/0x90
> >  __x64_sys_close+0x90/0x120
> >  do_syscall_64+0x5d/0x170
> >  entry_SYSCALL_64_after_hwframe+0x76/0x7e
> > 
> > The root cause is:
> > sock_hash_update_common
> >   sock_map_unref
> >     sock_map_del_link
> >       psock->psock_update_sk_prot(sk, psock, false);
> > 	//false won't restore proto
> >     sk_psock_put
> >        rcu_assign_sk_user_data(sk, NULL);
> > inet_release
> >   sk->sk_prot->close
> >     sock_map_close
> >       WARN(sk->sk_prot->close == sock_map_close)
> > 
> > When psock is removed from sockmap, sock_map_del_link() still set
> > sk->sk_prot to bpf proto instead of restore it (for incorrect restore
> > value). sock release will triger warning of sock_map_close() for
> > recurse after psock drop.
> 
> But sk_psock_drop() restores it with sk_psock_restore_proto() after the
> psock reference count goes to zero. So how could the above happen?

[..]
 
> By the way, it would be perfect if you could add a test case for it 
> together with this patch (a followup patch is fine too).

There is tools/testing/selftests/bpf/test_maps.c that gets broken by this
patch:

	Failed map_fd_msg update sockmap -16

---
pw-bot: cr