Re: ipv4: distinguish EHOSTUNREACH from the ENETUNREACH

[Daniel Walker]

Hi,

It seems like commit cd0f0b is trying to add back these two errors 
values into ip_route_input_slow(). However, if you follow the code path 
further down you get to the two exit points of this function,

in net/ipv4/route.c:ip_route_input_slow()

if (rt_cache_valid(rth)) {
        skb_dst_set_noref(skb, &rth->dst);
        err = 0;
        goto out;
}

and

skb_dst_set(skb, &rth->dst);
err = 0;
goto out;

Both of these set "err" variable to 0. This effective destroys the 
return value which the patch seems to be adding. Am I missing something 
here?


Thanks,

Daniel