[dpdk-dev] [RFC] avoid testpmd only working in XEN

[Christian Ehrhardt]

Hi,
I expected more of a discussion fallout as there would be so different
approaches to solve it (with the explicit xs_open stuff I mentioned, but
that I consider overengineering).
I had this running a while now and I can only say for me it works.
A few internal people working with Xen acked it as sufficient - but only by
looking at the code.

Since I got neither objections nor any other negative feedback so far I'll
just rewrite the comment and submit it as patch.

Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd

On Tue, Mar 1, 2016 at 5:32 PM, Christian Ehrhardt <
christian.ehrhardt at canonical.com> wrote:

> With LIBRTE_PMD_XENVIRT enabled testpmd is built in a way to ONLY work
> in XEN environments.
> It will surface as:
>    PMD: gntalloc: ioctl error
>    EAL: Error - exiting with code: 1
>      Cause: Creation of mbuf pool for socket 0 failed
>
> There could be a complex solution checking if this is a Xen env via
> something like:
>    #include <xenstore.h>
>    struct xs_handle *xsh = NULL;
>    xsh = xs_open(XS_OPEN_READONLY);
>    check for xsh being NULL ...
>
> But I see no reason to go so complex and instead the patch goes the
> simple and more readable way of just falling back to the normal
> allocation if the xen based allocation failed.
> ---
>  app/test-pmd/testpmd.c | 33 ++++++++++++++++-----------------
>  1 file changed, 16 insertions(+), 17 deletions(-)
>
> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
> index 1319917..b008df3 100644
> --- a/app/test-pmd/testpmd.c
> +++ b/app/test-pmd/testpmd.c
> @@ -410,7 +410,7 @@ mbuf_pool_create(uint16_t mbuf_seg_size, unsigned
> nb_mbuf,
>                  unsigned int socket_id)
>  {
>         char pool_name[RTE_MEMPOOL_NAMESIZE];
> -       struct rte_mempool *rte_mp;
> +       struct rte_mempool *rte_mp = NULL;
>         uint32_t mb_size;
>
>         mb_size = sizeof(struct rte_mbuf) + mbuf_seg_size;
> @@ -423,24 +423,23 @@ mbuf_pool_create(uint16_t mbuf_seg_size, unsigned
> nb_mbuf,
>                 rte_pktmbuf_pool_init, NULL,
>                 rte_pktmbuf_init, NULL,
>                 socket_id, 0);
> -
> -
> -
> -#else
> -       if (mp_anon != 0)
> -               rte_mp = mempool_anon_create(pool_name, nb_mbuf, mb_size,
> -                                   (unsigned) mb_mempool_cache,
> -                                   sizeof(struct
> rte_pktmbuf_pool_private),
> -                                   rte_pktmbuf_pool_init, NULL,
> -                                   rte_pktmbuf_init, NULL,
> -                                   socket_id, 0);
> -       else
> -               /* wrapper to rte_mempool_create() */
> -               rte_mp = rte_pktmbuf_pool_create(pool_name, nb_mbuf,
> -                       mb_mempool_cache, 0, mbuf_seg_size, socket_id);
> -
>  #endif
>
> +       /* if the former XEN allocation failed fall back to normal
> allocation */
> +       if (rte_mp == NULL) {
> +               if (mp_anon != 0)
> +                       rte_mp = mempool_anon_create(pool_name, nb_mbuf,
> +                                       mb_size, (unsigned)
> mb_mempool_cache,
> +                                       sizeof(struct
> rte_pktmbuf_pool_private),
> +                                       rte_pktmbuf_pool_init, NULL,
> +                                       rte_pktmbuf_init, NULL,
> +                                       socket_id, 0);
> +               else
> +                       /* wrapper to rte_mempool_create() */
> +                       rte_mp = rte_pktmbuf_pool_create(pool_name,
> nb_mbuf,
> +                               mb_mempool_cache, 0, mbuf_seg_size,
> socket_id);
> +       }
> +
>         if (rte_mp == NULL) {
>                 rte_exit(EXIT_FAILURE, "Creation of mbuf pool for socket
> %u "
>                                                 "failed\n", socket_id);
> --
> 2.7.0
>
>