Re: [PATCH net-next v6 1/2] nfc: llcp_core: Hold a ref to llcp_local->dev when holding a ref to llcp_local

[Siddh Raman Pant]

On Mon, 18 Dec 2023 15:09:00 +0530, Krzysztof Kozlowski wrote:
> On 17/12/2023 14:11, Siddh Raman Pant wrote:
> >  static struct nfc_llcp_sock *nfc_llcp_sock_get(struct nfc_llcp_local *local,
> > @@ -959,8 +974,18 @@ static void nfc_llcp_recv_connect(struct nfc_llcp_local *local,
> >  	}
> >  
> >  	new_sock = nfc_llcp_sock(new_sk);
> > -	new_sock->dev = local->dev;
> > +
> >  	new_sock->local = nfc_llcp_local_get(local);
> > +	if (!new_sock->local) {
> > +		reason = LLCP_DM_REJ;
> > +		release_sock(&sock->sk);
> > +		sock_put(&sock->sk);
> > +		sock_put(&new_sock->sk);
> 
> Why is this needed? Which part earlier gets the reference?

Thanks for pointing out. sk_init sets refcount to 1. Actually on a
further look, the next line shouldn't be there as nfc_llcp_sock_free()
is already called in sk->sk_destruct (== llcp_sock_destruct()), which
is called via __sk_destruct().

As sock_put() -> sk_free() -> __sk_destruct() -> sk_prot_free(),
so we need to put.

TBH really don't know why nfc_llcp_sock_free() is not static.

> > +		nfc_llcp_sock_free(new_sock);
> 
> This order is still wrong. Unwinding is almost always done in reversed
> order, for good reasons. Why do you unwind in other order?

Oops, extremely sorry about that :( I reverted back to wrong ordering
from an older local commit and didn't check.

I'll send the fixed one.

Thanks,
Siddh