Re: [tipc-discussion] [PATCH RFC] tipc: improve bundle algorithm

[Jon M. <jon...@er...>]

> -----Original Message-----
> From: Tuong Lien <tuo...@de...>
> Sent: 26-Aug-19 07:46
> To: Jon Maloy <jon...@er...>; ma...@do...;
> yin...@wi...; tip...@li...
> Subject: [PATCH RFC] tipc: improve bundle algorithm
> 
> Signed-off-by: Tuong Lien <tuo...@de...>
> ---
>  net/tipc/link.c | 29 ++++++++++++++++++-----------  net/tipc/msg.c  |  4 ++--
>  2 files changed, 20 insertions(+), 13 deletions(-)


[...]

> @@ -544,7 +544,7 @@ bool tipc_msg_make_bundle(struct sk_buff **skb,
> struct tipc_msg *msg,
>  	tipc_msg_init(msg_prevnode(msg), bmsg, MSG_BUNDLER, 0,
>  		      INT_H_SIZE, dnode);
>  	if (msg_isdata(msg))
> -		msg_set_importance(bmsg, TIPC_CRITICAL_IMPORTANCE);
> +		msg_set_importance(bmsg, msg_importance(msg));
>  	else
>  		msg_set_importance(bmsg, TIPC_SYSTEM_IMPORTANCE);

The if-clause is not necessary any more. You can assign the importance of the inner message directly to the bundle.
You can do that here, or inside the "tipc_make_bundle" branch of tipc_link_xmit().
Otherwise I think this is a smart, although not very elegant, solution to our problem.
Maybe you could steal some of the log text from my first suggestion to this patch? I think that describes the problem well.

Acked-by: Jon

///jon


>  	msg_set_seqno(bmsg, msg_seqno(msg));
> --
> 2.13.7