Re: [PATCH net-next 08/10] tls: rx: use async as an in-out argument

[Jakub Kicinski <kuba@kernel.org>]

On Mon, 25 Apr 2022 10:19:45 +0300 Gal Pressman wrote:
> On 11/04/2022 22:19, Jakub Kicinski wrote:
> > Propagating EINPROGRESS thru multiple layers of functions is
> > error prone. Use darg->async as an in/out argument, like we
> > use darg->zc today. On input it tells the code if async is
> > allowed, on output if it took place.
> >
> > Signed-off-by: Jakub Kicinski <kuba@kernel.org>  
> 
> I know this is not much to go on, but this patch broke our tls workflows
> when device offload is enabled.
> I'm still looking into it, but maybe you have an idea what might have
> went wrong?

Oof right, sorry. When packet is already decrypted by HW we'll skip 
the decrypt completely and leave async to whatever it was at input.

Something like this?

--->8---------

diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c
index ddbe05ec5489..80094528eadb 100644
--- a/net/tls/tls_sw.c
+++ b/net/tls/tls_sw.c
@@ -1562,6 +1562,7 @@ static int decrypt_skb_update(struct sock *sk, struct sk_buff *skb,
 
 	if (tlm->decrypted) {
 		darg->zc = false;
+		darg->async = false;
 		return 0;
 	}
 
@@ -1572,6 +1573,7 @@ static int decrypt_skb_update(struct sock *sk, struct sk_buff *skb,
 		if (err > 0) {
 			tlm->decrypted = 1;
 			darg->zc = false;
+			darg->async = false;
 			goto decrypt_done;
 		}
 	}