Re: [net-next] seg6: using DSCP of inner IPv4 packets

[Ahmed Abdelsalam <ahabdels@gmail.com>]

I will refactor the code of this function and submit a new patch.
Ahmed

On 31/07/2020 01:44, David Miller wrote:
> From: Ahmed Abdelsalam <ahabdels@gmail.com>
> Date: Tue, 28 Jul 2020 12:20:44 +0000
> 
>> This patch allows copying the DSCP from inner IPv4 header to the
>> outer IPv6 header, when doing SRv6 Encapsulation.
>>
>> This allows forwarding packet across the SRv6 fabric based on their
>> original traffic class.
>>
>> Signed-off-by: Ahmed Abdelsalam <ahabdels@gmail.com>
> 
> The conditionals in this function are now a mess.
> 
>> -	inner_hdr = ipv6_hdr(skb);
>> +	if (skb->protocol == htons(ETH_P_IPV6))
>> +		inner_hdr = ipv6_hdr(skb);
>> +	else
>> +		inner_ipv4_hdr = ip_hdr(skb);
>> +
> 
> You assume that if skb->protocol is not ipv6 then it is ipv4.
> 
>> @@ -138,6 +143,10 @@ int seg6_do_srh_encap(struct sk_buff *skb, struct ipv6_sr_hdr *osrh, int proto)
>>   		ip6_flow_hdr(hdr, ip6_tclass(ip6_flowinfo(inner_hdr)),
>>   			     flowlabel);
>>   		hdr->hop_limit = inner_hdr->hop_limit;
>> +	} else if (skb->protocol == htons(ETH_P_IP)) {
>> +		ip6_flow_hdr(hdr, inner_ipv4_hdr->tos, flowlabel);
>> +		hdr->hop_limit = inner_ipv4_hdr->ttl;
>> +		memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
>>   	} else {
>>   		ip6_flow_hdr(hdr, 0, flowlabel);
>>   		hdr->hop_limit = ip6_dst_hoplimit(skb_dst(skb));
> 
> But this code did not make that assumption at all.
> 
> Only one of the two can be correct.
> 
> The conditional assignment is also very ugly, you have two pointers
> conditionally initialized.  The compiler is going to have a hard time
> figuring out that each pointer is only used in the code path where it
> is guaranteed to be initialiazed.
> 
> And it can't do that, as far as the compiler knows, skb->protocol can
> change between those two locations.  It MUST assume that can happen if
> there are any functions calls whatsoever between these two code points.
> 
> This function has to be sanitized, with better handling of access to
> the inner protocol header values, before I am willing to apply this.
>