Vhost-user early adopter feedback

Vhost-user early adopter feedback

[Benoît Canet]

Hello Xie,

As promized I integrated your vhost-user patchset from january in my vswitch.

I just tried it, it works pretty well.

I just had a minor bug with rte_vhost_driver_register taking ownership of the
string patch pointer too late. I freed it out of habit just after registering in the
caller and when ifname[IFNAMESIZ] was written the pointer was used for a new string I
allocated later. Maybe an early strdup() would fix this.

The last patch of your new version is really a great idea since it will
simplify a lot the socket creation and management code.

Best regards

Benoît

Re: Vhost-user early adopter feedback

[Xie, Huawei]

On 2/18/2015 3:59 PM, Benoît Canet wrote:
> Hello Xie,
>
> As promized I integrated your vhost-user patchset from january in my vswitch.
>
> I just tried it, it works pretty well.
>
> I just had a minor bug with rte_vhost_driver_register taking ownership of the
> string patch pointer too late. I freed it out of habit just after registering in the
> caller and when ifname[IFNAMESIZ] was written the pointer was used for a new string I
> allocated later. Maybe an early strdup() would fix this.
Thanks.
Do you mean we duplicate a string from the first parameter path, like
vserver->path = strdup(path) ?
If this was the case, it was ever in my mind. We would do this if
necessary.
>
> The last patch of your new version is really a great idea since it will
> simplify a lot the socket creation and management code.
>
> Best regards
>
> Benoît
>
>

Re: Vhost-user early adopter feedback

[Benoît Canet]

The Wednesday 25 Feb 2015 à 07:46:56 (+0000), Xie, Huawei wrote :
> On 2/18/2015 3:59 PM, Benoît Canet wrote:
> > Hello Xie,
> >
> > As promized I integrated your vhost-user patchset from january in my vswitch.
> >
> > I just tried it, it works pretty well.
> >
> > I just had a minor bug with rte_vhost_driver_register taking ownership of the
> > string patch pointer too late. I freed it out of habit just after registering in the
> > caller and when ifname[IFNAMESIZ] was written the pointer was used for a new string I
> > allocated later. Maybe an early strdup() would fix this.
> Thanks.
> Do you mean we duplicate a string from the first parameter path, like
> vserver->path = strdup(path) ?

Yes I was thinking about this.

Best regards

Benoît

> If this was the case, it was ever in my mind. We would do this if
> necessary.

> >
> > The last patch of your new version is really a great idea since it will
> > simplify a lot the socket creation and management code.
> >
> > Best regards
> >
> > Benoît
> >
> >
>