Re: [PATCH net-next v2 08/12] selftests: ncdevmem: Use YNL to enable TCP header split

[Stanislav Fomichev <stfomichev@gmail.com>]

On 10/03, Mina Almasry wrote:
> On Mon, Sep 30, 2024 at 10:18 AM Stanislav Fomichev <sdf@fomichev.me> wrote:
> >
> > In the next patch the hard-coded queue numbers are gonna be removed.
> > So introduce some initial support for ethtool YNL and use
> > it to enable header split.
> >
> > Also, tcp-data-split requires latest ethtool which is unlikely
> > to be present in the distros right now.
> >
> > (ideally, we should not shell out to ethtool at all).
> >
> > Cc: Mina Almasry <almasrymina@google.com>
> > Signed-off-by: Stanislav Fomichev <sdf@fomichev.me>
> > ---
> >  tools/testing/selftests/net/Makefile   |  2 +-
> >  tools/testing/selftests/net/ncdevmem.c | 43 ++++++++++++++++++++++++--
> >  2 files changed, 42 insertions(+), 3 deletions(-)
> >
> > diff --git a/tools/testing/selftests/net/Makefile b/tools/testing/selftests/net/Makefile
> > index 649f1fe0dc46..9c970e96ed33 100644
> > --- a/tools/testing/selftests/net/Makefile
> > +++ b/tools/testing/selftests/net/Makefile
> > @@ -112,7 +112,7 @@ TEST_INCLUDES := forwarding/lib.sh
> >  include ../lib.mk
> >
> >  # YNL build
> > -YNL_GENS := netdev
> > +YNL_GENS := ethtool netdev
> >  include ynl.mk
> >
> >  $(OUTPUT)/epoll_busy_poll: LDLIBS += -lcap
> > diff --git a/tools/testing/selftests/net/ncdevmem.c b/tools/testing/selftests/net/ncdevmem.c
> > index 48cbf057fde7..a1fa818c8229 100644
> > --- a/tools/testing/selftests/net/ncdevmem.c
> > +++ b/tools/testing/selftests/net/ncdevmem.c
> > @@ -28,10 +28,12 @@
> >  #include <linux/netlink.h>
> >  #include <linux/genetlink.h>
> >  #include <linux/netdev.h>
> > +#include <linux/ethtool_netlink.h>
> >  #include <time.h>
> >  #include <net/if.h>
> >
> >  #include "netdev-user.h"
> > +#include "ethtool-user.h"
> >  #include <ynl.h>
> >
> >  #define PAGE_SHIFT 12
> > @@ -217,8 +219,42 @@ static int reset_flow_steering(void)
> >
> >  static int configure_headersplit(bool on)
> >  {
> > -       return run_command("sudo ethtool -G %s tcp-data-split %s >&2", ifname,
> > -                          on ? "on" : "off");
> > +       struct ethtool_rings_set_req *req;
> > +       struct ynl_error yerr;
> > +       struct ynl_sock *ys;
> > +       int ret;
> > +
> > +       ys = ynl_sock_create(&ynl_ethtool_family, &yerr);
> > +       if (!ys) {
> > +               fprintf(stderr, "YNL: %s\n", yerr.msg);
> > +               return -1;
> > +       }
> > +
> > +       req = ethtool_rings_set_req_alloc();
> > +       ethtool_rings_set_req_set_header_dev_index(req, ifindex);
> > +       ethtool_rings_set_req_set_tcp_data_split(req, on ? 2 : 0);
> 
> I'm guessing 2 is explicit on? 1 being auto probably? A comment would
> be nice, but that's just a nit.

Sure, will add.

> > +       ret = ethtool_rings_set(ys, req);
> > +       if (ret < 0)
> > +               fprintf(stderr, "YNL failed: %s\n", ys->err.msg);
> 
> Don't you wanna return ret; here on error?

Good point. Will add 'if (ret == 0)' to the get request.

> > +       ethtool_rings_set_req_free(req);
> > +
> > +       {
> > +               struct ethtool_rings_get_req *req;
> > +               struct ethtool_rings_get_rsp *rsp;
> > +
> 
> I'm guessing you're creating a new scope to re-declare req/rsp? To be
> honest it's a bit weird style I haven't seen anywhere else. I would
> prefer get_req and get_rsp instead, but this is a nit.

SG.

> > +               req = ethtool_rings_get_req_alloc();
> > +               ethtool_rings_get_req_set_header_dev_index(req, ifindex);
> > +               rsp = ethtool_rings_get(ys, req);
> > +               ethtool_rings_get_req_free(req);
> > +               if (rsp)
> > +                       fprintf(stderr, "TCP header split: %d\n",
> > +                               rsp->tcp_data_split);
> 
> I would prefer to cehck that rsp->tcp_data_split == 2 for 'on' and ==
> 0 for 'off' instead of just printing and relying on the user to notice
> the mismatch in the logs.

Will do, thanks!
 
> Consider addressing the feedback, but mostly nits/improvements that
> can be done later, so:
> 
> Reviewed-by: Mina Almasry <almasrymina@google.com>