> -----Original Message-----
> From: Shuang Han <hanshuang87@gmail.com>
> Sent: Wednesday, November 23, 2022 8:42 PM
> To: dev@dpdk.org
> Cc: Yang, Qiming <qiming.yang@intel.com>; Zhang, Qi Z
> <qi.z.zhang@intel.com>; Shuang Han <hanshuang87@gmail.com>
> Subject: [PATCH] net/ice: fix link status error when dev start
>
> ice_link_update get port link status and wait for one second when dev start.
> If the device need almost one second for link up, ice_link_update will enter
> the last rte_delay_ms(CHECK_INTERVAL) with link down, in the meantime
> link interrupt raised, the interrupt handler update dev->data->dev_link to up.
> After a while the rte_delay_ms(CHECK_INTERVAL) finished and update dev-
> >data->dev_link to down.
>
> Signed-off-by: Shuang Han <hanshuang87@gmail.com>
> ---
>  drivers/net/ice/ice_ethdev.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/ice/ice_ethdev.c b/drivers/net/ice/ice_ethdev.c index
> 0bc739d..316ba5a 100644
> --- a/drivers/net/ice/ice_ethdev.c
> +++ b/drivers/net/ice/ice_ethdev.c
> @@ -4013,9 +4013,9 @@ static int ice_init_rss(struct ice_pf *pf)
>                             RTE_ETH_LINK_SPEED_FIXED);
>
>  out:
> -     ice_atomic_write_link_status(dev, &link);
>       if (link.link_status == old.link_status)
>               return -1;
> +     ice_atomic_write_link_status(dev, &link);

This looks like a workaround, I think we may need some lock to prevent ice_link_update be re-entered from different thread at same time.

>
>       return 0;
>  }
> --
> 1.8.3.1