Re: [PATCH 1/3] Misc cleanups

[Patrick McHardy <kaber@trash.net>]

Jan Engelhardt wrote:
> On Jun 11 2007 16:17, Patrick McHardy wrote:
> 
>>>@@ -249,7 +249,7 @@ clusterip_hashfn(struct sk_buff *skb, st
>>> 	case IPPROTO_SCTP:
>>> 	case IPPROTO_DCCP:
>>> 	case IPPROTO_ICMP:
>>>-		ports = (void *)iph+iph->ihl*4;
>>>+		ports = (const void *)iph+iph->ihl*4;
>>
>>
>>This one is not neccessary.
> 
> 
> *Are* you sure? iph is a const struct iphdr *, so adding (iph->ihl*4)
> will add sizeof(const struct iphdr *) * (iph->ihl*4) bytes. Which is
> four times more (eight on 64-bit) than actually wanted. C basics.


I'm talking about the const. But OK, I missed that iph is already const,
so the const cast is cleaner. I'll add it back.

>>>@@ -88,7 +88,7 @@ masquerade_target(struct sk_buff **pskb,
>>> 		return NF_ACCEPT;
>>> 
>>> 	mr = targinfo;
>>>-	rt = (struct rtable *)(*pskb)->dst;
>>>+	rt = (const struct rtable *)(*pskb)->dst;
>>
>>
>>This one neither.
> 
> 
> Uhm? ->dst is of type "struct dst_entry", while rt is of type struct rtable.

I'm just talking about casting to const.

> 
> [...]
> 
> I think something is wrong. What is, actually? The "const"? Well 
> yeah I added that for great consistency.


Its unnecessary, but OK in places where the pointer you cast was
already const (since its ugly to cast const pointers to non-const),
in all other casts its just unnecessary.

>>>@@ -443,8 +449,8 @@ hashlimit_match(const struct sk_buff *sk
>>> 		unsigned int protoff,
>>> 		bool *hotdrop)
>>> {
>>>-	struct xt_hashlimit_info *r =
>>>-		((struct xt_hashlimit_info *)matchinfo)->u.master;
>>>+	const struct xt_hashlimit_info *r =
>>>+		((const struct xt_hashlimit_info *)matchinfo)->u.master;
>>
>>
>>This one is fine I guess since matchinfo is also const. Same for
>>all other matches.
>>
> 
> 
> And here it's ok? I don't get it. De-de-referencing a const item does 
> not always need to give a const item, hence, all is fine.
> Mind you, I compile tested it, and it was warning-free, as before.
> Care to explain?


See above.