Re: [PATCH net] net: sched: fix initialiser warning in sch_frag.c

["Russell King (Oracle)" <linux@armlinux.org.uk>]

On Mon, Sep 20, 2021 at 11:08:53AM +0200, Davide Caratti wrote:
> On Sun, Sep 19, 2021 at 11:40:33PM +0100, Russell King (Oracle) wrote:
> > Debian gcc 10.2.1 complains thusly:
> > 
> > net/sched/sch_frag.c:93:10: warning: missing braces around initializer [-Wmissing-braces]
> >    struct rtable sch_frag_rt = { 0 };
> >           ^
> > net/sched/sch_frag.c:93:10: warning: (near initialization for 'sch_frag_rt.dst') [-Wmissing-braces]
> > 
> > Fix it by removing the unnecessary '0' initialiser, leaving the
> > braces.
> 
> hello Russell, thanks a lot for reporting!
>  
> > diff --git a/net/sched/sch_frag.c b/net/sched/sch_frag.c
> > index 8c06381391d6..ab359d63287c 100644
> > --- a/net/sched/sch_frag.c
> > +++ b/net/sched/sch_frag.c
> > @@ -90,7 +90,7 @@ static int sch_fragment(struct net *net, struct sk_buff *skb,
> >  	}
> >  
> >  	if (skb_protocol(skb, true) == htons(ETH_P_IP)) {
> > -		struct rtable sch_frag_rt = { 0 };
> > +		struct rtable sch_frag_rt = { };
> 
> this surely fixes the -Wmissing-braces, but then -Wpedantic
> would complain about usage of GNU extension (I just tried on godbolt
> with x86_64 gcc 11.2):
> 
> warning: ISO C forbids empty initializer braces [-Wpedantic]
> 
> While we are fixing this, probably the best thing is to initialize the
> 'dst' struct member  to 0: in my understanding this should be sufficient
> to let the compiler fill all the struct members with 0.
> 
> Oh, and I might have inserted a similar thing in openvswitch kernel
> module (see [1]), if you agree I will send a patch that fixes this as
> well. WDYT?

ISO C may forbid it, but the kernel build uses -std=gnu89 - which is
c89 with GNU extensions. One of the GNU extensions is to allow the
empty initialiser, which means "initialise all members of this struct
to zero".

However, as I say, this was found using gcc 4.9.4 under 5.14, where
4.9.4 is a permissable compiler. However, under 5.15-rc it is no
longer so the patch should not be applied to development kernels.
It leaves the question open whether it should be fixed in stable or
not, since stable kernels _do_ permit gcc 4.9.4.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!