[dpdk-dev] [PATCH] eal: restrict ctrl threads to startup cpu affinity

David Marchand david.marchand at redhat.com
Wed Feb 13 21:21:20 CET 2019


On Wed, Feb 13, 2019 at 5:14 PM David Marchand <david.marchand at redhat.com>
wrote:

> Spawning the ctrl threads on anything that is not part of the eal
> coremask is not that polite to the rest of the system.
>
> Rather than introduce yet another eal options for this, let's take
> the startup cpu affinity as a reference and remove the eal coremask
> from it.
> If no cpu is left, then we default to the master core.
>
> The cpuset is computed once at init before the original cpu affinity.
>

Need to fix this last sentence...


> Fixes: d651ee4919cd ("eal: set affinity for control threads")
> Signed-off-by: David Marchand <david.marchand at redhat.com>
> ---
>  lib/librte_eal/common/eal_common_options.c | 28
> ++++++++++++++++++++++++++++
>  lib/librte_eal/common/eal_common_thread.c  | 21 ++++-----------------
>  lib/librte_eal/common/eal_internal_cfg.h   |  3 +++
>  3 files changed, 35 insertions(+), 17 deletions(-)
>
> diff --git a/lib/librte_eal/common/eal_common_options.c
> b/lib/librte_eal/common/eal_common_options.c
> index 6c96f45..b766252 100644
> --- a/lib/librte_eal/common/eal_common_options.c
> +++ b/lib/librte_eal/common/eal_common_options.c
> @@ -1360,6 +1361,31 @@ static int xdigit2val(unsigned char c)
>         cfg->lcore_count -= removed;
>  }
>
> +static void
> +compute_ctrl_threads_cpuset(struct internal_config *internal_cfg)
> +{
> +       rte_cpuset_t *cpuset = &internal_cfg->ctrl_cpuset;
> +       rte_cpuset_t default_set;
> +       unsigned int lcore_id;
> +
> +       for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
> +               if (eal_cpu_detected(lcore_id) &&
> +                               rte_lcore_has_role(lcore_id, ROLE_OFF)) {
> +                       CPU_SET(lcore_id, cpuset);
> +               }
> +       }
> +
> +       if (pthread_getaffinity_np(pthread_self(), sizeof(rte_cpuset_t),
> +                               &default_set) < 0)
> +               CPU_ZERO(&default_set);
> +
> +       CPU_AND(cpuset, cpuset, &default_set);
>

CPU_AND is different on Freebsd.

     *CPU*_*AND*(*cpuset*_*t* **dst*, *cpuset*_*t* **src*);

     The *CPU*_*AND*() macro removes CPUs absent from *src* from
*dst*.	 (It is	the
     *cpuset(9)*
<https://www.freebsd.org/cgi/man.cgi?query=cpuset&sektion=9&apropos=0&manpath=FreeBSD+11.0-RELEASE+and+Ports>
equivalent of the scalar: *dst* &=	*src*.)  *CPU*_*AND*_*ATOMIC*()	is
     similar, with the same atomic semantics as	*CPU*_*OR*_*ATOMIC*().

Will fix in v2.

-- 
David Marchand


More information about the dev mailing list