Re: [PATCH bpf] libbpf: fix compatibility for kernels without need_wakeup

[John Fastabend <john.fastabend@gmail.com>]

Magnus Karlsson wrote:
> On Tue, Oct 8, 2019 at 9:29 PM John Fastabend <john.fastabend@gmail.com> wrote:
> >
> > Magnus Karlsson wrote:
> > > When the need_wakeup flag was added to AF_XDP, the format of the
> > > XDP_MMAP_OFFSETS getsockopt was extended. Code was added to the kernel
> > > to take care of compatibility issues arrising from running
> > > applications using any of the two formats. However, libbpf was not
> > > extended to take care of the case when the application/libbpf uses the
> > > new format but the kernel only supports the old format. This patch
> > > adds support in libbpf for parsing the old format, before the
> > > need_wakeup flag was added, and emulating a set of static need_wakeup
> > > flags that will always work for the application.
> > >
> > > Fixes: a4500432c2587cb2a ("libbpf: add support for need_wakeup flag in AF_XDP part")
> > > Reported-by: Eloy Degen <degeneloy@gmail.com>
> > > Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
> > > ---
> > >  tools/lib/bpf/xsk.c | 109 +++++++++++++++++++++++++++++++++++++---------------
> > >  1 file changed, 78 insertions(+), 31 deletions(-)
> > >
> > > diff --git a/tools/lib/bpf/xsk.c b/tools/lib/bpf/xsk.c
> > > index a902838..46f9687 100644
> > > --- a/tools/lib/bpf/xsk.c
> > > +++ b/tools/lib/bpf/xsk.c
> > > @@ -44,6 +44,25 @@
> > >   #define PF_XDP AF_XDP
> > >  #endif
> > >
> > > +#define is_mmap_offsets_v1(optlen) \
> > > +     ((optlen) == sizeof(struct xdp_mmap_offsets_v1))
> > > +
> > > +#define get_prod_off(ring) \
> > > +     (is_mmap_offsets_v1(optlen) ? \
> > > +      ((struct xdp_mmap_offsets_v1 *)&off)->ring.producer : \
> > > +      off.ring.producer)
> > > +#define get_cons_off(ring) \
> > > +     (is_mmap_offsets_v1(optlen) ? \
> > > +      ((struct xdp_mmap_offsets_v1 *)&off)->ring.consumer : \
> > > +      off.ring.consumer)
> > > +#define get_desc_off(ring) \
> > > +     (is_mmap_offsets_v1(optlen) ? \
> > > +      ((struct xdp_mmap_offsets_v1 *)&off)->ring.desc : off.ring.desc)
> > > +#define get_flags_off(ring) \
> > > +     (is_mmap_offsets_v1(optlen) ? \
> > > +      ((struct xdp_mmap_offsets_v1 *)&off)->ring.consumer + sizeof(u32) : \
> > > +      off.ring.flags)
> > > +
> >
> > It seems the only thing added was flags right? If so seems we
> > only need the last one there, get_flags_off(). I think it would
> > be a bit cleaner to just use the macros where its actually
> > needed IMO.
> 
> The flag is indeed added to the end of struct xdp_ring_offsets, but
> this struct is replicated four times in the struct xdp_mmap_offsets,
> so the added flags are present four time there at different offsets.
> This means that 3 out of the 4 prod, cons and desc variables are
> located at different offsets from the original. Do not know how I can
> get rid of these macros in this case. But it might just be me not
> seeing it, of course :-).

Not sure I like it but not seeing a cleaner solution that doesn't cause
larger changes so...

Acked-by: John Fastabend <john.fastabend.gmail.com>