From: Johannes Berg <johannes.berg@intel.com>

It doesn't look like this could possibly work properly since
VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD is defined to 10, but the
dev->protocol_features has a bitmap. I suppose the peer this
was tested with also supported VHOST_USER_PROTOCOL_F_LOG_SHMFD,
in which case the test would always be false, but nevertheless
the code seems wrong.

Use has_feature() to fix this.

Fixes: d84599f56c82 ("libvhost-user: support host notifier")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
---
 contrib/libvhost-user/libvhost-user.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/contrib/libvhost-user/libvhost-user.c b/contrib/libvhost-user/libvhost-user.c
index 6a02eaffc672..fcf4a8a00ed2 100644
--- a/contrib/libvhost-user/libvhost-user.c
+++ b/contrib/libvhost-user/libvhost-user.c
@@ -1097,7 +1097,8 @@ bool vu_set_queue_host_notifier(VuDev *dev, VuVirtq *vq, int fd,
 
     vmsg.fd_num = fd_num;
 
-    if ((dev->protocol_features & VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD) == 0) {
+    if (!has_feature(dev->protocol_features,
+                     VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD)) {
         return false;
     }
 
-- 
2.23.0

On Tue, Sep 03, 2019 at 11:04:22PM +0300, Johannes Berg wrote:
> From: Johannes Berg <johannes.berg@intel.com>
> 
> It doesn't look like this could possibly work properly since
> VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD is defined to 10, but the
> dev->protocol_features has a bitmap. I suppose the peer this
> was tested with also supported VHOST_USER_PROTOCOL_F_LOG_SHMFD,
> in which case the test would always be false, but nevertheless
> the code seems wrong.

Ooops.. I tested `tests/vhost-user-bridge -H`. But as you
said it worked because VHOST_USER_PROTOCOL_F_LOG_SHMFD has
been negotiated. Thanks for spotting this!

> 
> Use has_feature() to fix this.
> 
> Fixes: d84599f56c82 ("libvhost-user: support host notifier")

Cc: qemu-stable@nongnu.org

> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
> ---
>  contrib/libvhost-user/libvhost-user.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/contrib/libvhost-user/libvhost-user.c b/contrib/libvhost-user/libvhost-user.c
> index 6a02eaffc672..fcf4a8a00ed2 100644
> --- a/contrib/libvhost-user/libvhost-user.c
> +++ b/contrib/libvhost-user/libvhost-user.c
> @@ -1097,7 +1097,8 @@ bool vu_set_queue_host_notifier(VuDev *dev, VuVirtq *vq, int fd,
>  
>      vmsg.fd_num = fd_num;
>  
> -    if ((dev->protocol_features & VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD) == 0) {
> +    if (!has_feature(dev->protocol_features,
> +                     VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD)) {

We have both of has_feature() and vu_has_feature() called by
other code in this file directly. Not sure which one is preferred..
Personally, I think vu_has_feature() might be better.

Thanks!
Tiwei

>          return false;
>      }
>  
> -- 
> 2.23.0
> 

On Wed, Sep 04, 2019 at 10:06:55AM +0800, Tiwei Bie wrote:
> On Tue, Sep 03, 2019 at 11:04:22PM +0300, Johannes Berg wrote:
> > From: Johannes Berg <johannes.berg@intel.com>
> > 
> > It doesn't look like this could possibly work properly since
> > VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD is defined to 10, but the
> > dev->protocol_features has a bitmap. I suppose the peer this
> > was tested with also supported VHOST_USER_PROTOCOL_F_LOG_SHMFD,
> > in which case the test would always be false, but nevertheless
> > the code seems wrong.
> 
> Ooops.. I tested `tests/vhost-user-bridge -H`. But as you
> said it worked because VHOST_USER_PROTOCOL_F_LOG_SHMFD has
> been negotiated. Thanks for spotting this!
> 
> > 
> > Use has_feature() to fix this.
> > 
> > Fixes: d84599f56c82 ("libvhost-user: support host notifier")
> 
> Cc: qemu-stable@nongnu.org
> 
> > Signed-off-by: Johannes Berg <johannes.berg@intel.com>
> > ---
> >  contrib/libvhost-user/libvhost-user.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/contrib/libvhost-user/libvhost-user.c b/contrib/libvhost-user/libvhost-user.c
> > index 6a02eaffc672..fcf4a8a00ed2 100644
> > --- a/contrib/libvhost-user/libvhost-user.c
> > +++ b/contrib/libvhost-user/libvhost-user.c
> > @@ -1097,7 +1097,8 @@ bool vu_set_queue_host_notifier(VuDev *dev, VuVirtq *vq, int fd,
> >  
> >      vmsg.fd_num = fd_num;
> >  
> > -    if ((dev->protocol_features & VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD) == 0) {
> > +    if (!has_feature(dev->protocol_features,
> > +                     VHOST_USER_PROTOCOL_F_SLAVE_SEND_FD)) {
> 
> We have both of has_feature() and vu_has_feature() called by
> other code in this file directly. Not sure which one is preferred..
> Personally, I think vu_has_feature() might be better.

Thanks for the patch introducing vu_has_protocol_feature().
This fix looks good to me. Thanks a lot!

I'm not the maintainer. But anyway, if this helps:
Reviewed-by: Tiwei Bie <tiwei.bie@intel.com>