On Wed, Aug 3, 2016 at 10:11 PM, David Ahern <[email protected]> wrote:
> On 8/3/16 1:57 PM, Geert Uytterhoeven wrote:
>>> +static void vrf_ip6_input_dst(struct sk_buff *skb, struct net_device
>>> *vrf_dev,
>>> > + int ifindex)
>>> > +{
>>> > + const struct ipv6hdr *iph = ipv6_hdr(skb);
>>> > + struct flowi6 fl6 = {
>>> > + .daddr = iph->daddr,
>>> > + .saddr = iph->saddr,
>>> > + .flowlabel = ip6_flowinfo(iph),
>> The above assignment causes the following compiler warning with
>> m68k-linux-gnu-gcc-4.1:
>>
>> drivers/net/vrf.c: In function ‘vrf_ip6_input_dst’:
>> drivers/net/vrf.c:870: warning: initialized field with
>> side-effects overwritten
>> drivers/net/vrf.c:870: warning: (near initialization for ‘fl6’)
>>
>> Unfortunately I have no idea what it means, nor do I see what's wrong
>> with the code.
>
> no idea. Fields are initialized once and left and right data types are the
> same.
>
> Can you remove one line at a time? Line 870 is ".flowi6_proto =
> iph->nexthdr," but all of the flowi6 macros are unique references to unique
> fields in flowi_common. The flowlabel line you point out is a unique field as
> well.
The only thing that seems to matter is assigning the result of the call to
ip6_flowinfo() to .flowlabel. Assigning a constant makes the warning go away.
Yeah, the 870 line number is funny, as it doesn't point to the offending line.
> Can you run pahole on file that did compile? e.g.,
>
> pahole -C 'flowi6' net/ipv6/route.o
>
> and get the common struct too:
>
> pahole -C 'flowi_common' net/ipv6/route.o
No output. Perhaps pahole doesn't play well with cross-compiling?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
