Re: [tipc-discussion] [PATCH RFC 2/2] tipc: improve message bundling algorithm

[Tuong L. T. <tuo...@de...>]

Hi Ying,

Agree, it's hard to trace...
I've changed the way we approach, will post it as a new patch, please take a
look from there!
Thanks a lot!

BR/Tuong

-----Original Message-----
From: Xue, Ying <Yin...@wi...> 
Sent: Friday, October 11, 2019 9:52 PM
To: Tuong Lien <tuo...@de...>;
tip...@li...; jon...@er...;
ma...@do...
Subject: RE: [PATCH RFC 2/2] tipc: improve message bundling algorithm

I can recognize this is a good improvement except that the following switch
cases of return values of tipc_msg_try_bundle() are not very friendly for
code reader. Although I do understand their real meanings, I have to spend
time checking its context back and forth. At least we should the meaningless
hard code case numbers or we try to change return value numbers of
tipc_msg_try_bundle().

+		n = tipc_msg_try_bundle(&l->backlog[imp].target_bskb, skb,
+					mtu - INT_H_SIZE,
+					l->addr);
+		switch (n) {
+		case 0:
+			break;
+		case 1:
+			__skb_queue_tail(backlogq, skb);
 			l->backlog[imp].len++;
-			l->stats.sent_bundled++;
+			continue;
+		case 2:
 			l->stats.sent_bundles++;
+			l->stats.sent_bundled++;
+		default:
+			kfree_skb(skb);
+			l->stats.sent_bundled++;
 			continue;