On Fri, Mar 17, 2017 at 10:14:56AM -0700, Yongseok Koh wrote:

Hi Nelio,

On Fri, Mar 17, 2017 at 02:11:43AM -0700, Nélio Laranjeiro wrote:

quoted

On Thu, Mar 16, 2017 at 03:40:56PM -0700, Yongseok Koh wrote:

quoted

Currently mlx5_dev_rss_reta_update() just updates tables in the host,
therefore it isn't immediately effective until restarting the device by
calling mlx5_dev_stop()/mlx5_dev_start() to update the changes in the
device side. This patch adds rebuilding the device-specific datastructure
and applying it to the device right away.

Signed-off-by: Yongseok Koh <redacted>
---
 drivers/net/mlx5/mlx5_rss.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/net/mlx5/mlx5_rss.c b/drivers/net/mlx5/mlx5_rss.c
index 0702f1a63..30e59faa5 100644
--- a/drivers/net/mlx5/mlx5_rss.c
+++ b/drivers/net/mlx5/mlx5_rss.c

@@ -357,8 +357,11 @@ mlx5_dev_rss_reta_update(struct rte_eth_dev *dev,
       int ret;
       struct priv *priv = dev->data->dev_private;

+     mlx5_dev_stop(dev);
       priv_lock(priv);
       ret = priv_dev_rss_reta_update(priv, reta_conf, reta_size);
       priv_unlock(priv);
+     if (!ret)
+             ret = (unsigned int)mlx5_dev_start(dev);
       return -ret;
 }
--

2.11.0

Hi Yongseok,

I don't understand why you need the cast for the returned value of
mlx5_dev_start() as it already returns an int and your final variable is
also an int.

For some reason, in mlx5 PMD code, priv_* calls return positive error numbers
and corresponding mlx5_* APIs return negative values by adding (-) sign the the
return value from priv_* calls.

This is due to the IOCTL calls which returns only positives errno, we
decided to keep only positives errors internally and negate them in when
they are returned to DPDK.

To follow this tacit rule in the existing code, I casted the return value to
unsigned.

I see more this cast as something confusing than helping, knowing that
it could be avoided with something like:

 if (ret)
 	return -ret;
 return mlx5_dev_start(dev);

Regards,

-- 
Nélio Laranjeiro
6WIND