On Mon, Oct 21, 2024 at 6:20 PM Paolo Abeni <pabeni@xxxxxxxxxx> wrote:
>
> On 10/15/24 16:07, Menglong Dong wrote:
> > diff --git a/include/net/ip_fib.h b/include/net/ip_fib.h
> > index 90ff815f212b..b3f7a1562140 100644
> > --- a/include/net/ip_fib.h
> > +++ b/include/net/ip_fib.h
> > @@ -452,13 +452,16 @@ int __fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
> > dscp_t dscp, int oif, struct net_device *dev,
> > struct in_device *idev, u32 *itag);
> >
> > -static inline int
> > +static inline enum skb_drop_reason
> > fib_validate_source(struct sk_buff *skb, __be32 src, __be32 dst,
> > dscp_t dscp, int oif, struct net_device *dev,
> > struct in_device *idev, u32 *itag)
> > {
> > - return __fib_validate_source(skb, src, dst, dscp, oif, dev, idev,
> > - itag);
> > + int err = __fib_validate_source(skb, src, dst, dscp, oif, dev, idev,
> > + itag);
> > + if (err < 0)
> > + return -err;
> > + return SKB_NOT_DROPPED_YET;
> > }
>
> It looks like the code churn in patch 1 is not needed??? You could just
> define here a fib_validate_source_reason() helper doing the above, and
> replace fib_validate_source with the the new helper as needed. Would
> that work?
>
Of course, that works fine. I'm just trying to find a graceful way
for this part. Defining a fib_validate_source_reason() here looks
nice too, and we can ignore the 1st patch. I'll do it this way in
the next version.
Thanks!
Menglong Dong
> > @@ -1785,9 +1785,10 @@ static int __mkroute_input(struct sk_buff *skb, const struct fib_result *res,
> > return -EINVAL;
> > }
> >
> > - err = fib_validate_source(skb, saddr, daddr, dscp, FIB_RES_OIF(*res),
> > - in_dev->dev, in_dev, &itag);
> > + err = __fib_validate_source(skb, saddr, daddr, dscp, FIB_RES_OIF(*res),
> > + in_dev->dev, in_dev, &itag);
> > if (err < 0) {
> > + err = -EINVAL;
> > ip_handle_martian_source(in_dev->dev, in_dev, skb, daddr,
> > saddr);
>
> I'm sorry for not noticing this issue before, but must preserve (at
> least) the -EXDEV error code from the unpatched version or RP Filter MIB
> accounting in ip_rcv_finish_core() will be fooled.
>
> Thanks,
>
> Paolo
>
