Re: [PATCH v4 3/5] net/tcp: Disable TCP-MD5 static key on tcp_md5sig_info destruction

On 11/19/22 03:18, Jakub Kicinski wrote:

On Tue, 15 Nov 2022 21:19:03 +0000 Dmitry Safonov wrote:

quoted

+	if (!rcu_dereference_protected(tp->md5sig_info, lockdep_sock_is_held(sk))) {
+		if (tcp_md5sig_info_add(sk, sk_gfp_mask(sk, GFP_ATOMIC)))
+			return -ENOMEM;
+
+		if (!static_key_fast_inc_not_negative(&tcp_md5_needed.key.key)) {
+			struct tcp_md5sig_info *md5sig = tp->md5sig_info;

I don't think sparse will be able to deduce that ->md5sig_info access
is safe here, so could you wrap it up as well?

Sure, that sounds logical to do.

Maybe it wouldn't be 
the worst move to provide a sk_rcu_dereference() or rcu_dereference_sk()
or some such wrapper.

More importantly tho - was the merging part for this patches discussed?
They don't apply to net-next.

They apply over linux-next as there's a change [1] in
linux-tip/locking/core on which the patches set based.

Could the way forward be through linux-tip tree, or that might create
net conflicts?

I'll send v5 with the trivial change to rcu_dereference_protected()
mentioned above.

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=locking/core&id=d0c006402e7941558e5283ae434e2847c7999378

Thanks,
          Dmitry