On Thu, Aug 25, 2016 at 09:18:30AM +0200, Maxime Coquelin wrote:

On 08/25/2016 05:03 AM, Yuanhan Liu wrote:

quoted

On Wed, Aug 24, 2016 at 10:11:57AM +0200, Maxime Coquelin wrote:

quoted

On 08/18/2016 10:48 AM, Yuanhan Liu wrote:

quoted

No need to use a pointer to store/retrieve features.

Signed-off-by: Yuanhan Liu <redacted>
---
lib/librte_vhost/vhost_user.c | 20 ++++++++------------
1 file changed, 8 insertions(+), 12 deletions(-)

diff --git a/lib/librte_vhost/vhost_user.c b/lib/librte_vhost/vhost_user.c
index ef4a0c1..eee99e9 100644
--- a/lib/librte_vhost/vhost_user.c
+++ b/lib/librte_vhost/vhost_user.c

@@ -155,23 +155,22 @@ vhost_user_reset_owner(struct virtio_net *dev)

/*
* The features that we support are requested.
*/
-static int
-vhost_user_get_features(uint64_t *pu)
+static uint64_t
+vhost_user_get_features(void)
{
-	*pu = VHOST_FEATURES;
-	return 0;
+	return VHOST_FEATURES;
}

This is not the topic of this series, but I wonder if it
could make sense to be able to override supported features
at device init time.

Not quite sure I understood it correctly: is rte_vhost_feature_disable()
the answer you are looking for?

Not really.

I meant a per-device supported features, and something you could set
also via the vhost PMD options, without needing to recompile.

I see. I think my following proposal would fix it then.

    http://dpdk.org/ml/archives/dev/2016-June/040021.html

Note you might want to follow the discussion, as it mentioned
per-device config later.

Another note, it lacked a fancy name at last discussion. For
now, I think I come up with a better one: --dpdk-env :)

	--yliu

But maybe it would make more sense to do it a guest level?

Maxime