From: Jason Wang <jasowang@redhat.com>
To: Michal Privoznik <mprivozn@redhat.com>,
"Michael S. Tsirkin" <mst@redhat.com>
Cc: Pankaj Gupta <pagupta@redhat.com>,
qemu-devel@nongnu.org, aliguori@amazon.com
Subject: Re: [Qemu-devel] [PATCH] net: Remove vhostforce option in addition to vhost parameter
Date: Thu, 28 May 2015 14:43:11 +0800 [thread overview]
Message-ID: <5566B8FF.7060905@redhat.com> (raw)
In-Reply-To: <5566B572.1020008@redhat.com>
On 05/28/2015 02:28 PM, Michal Privoznik wrote:
> On 28.05.2015 05:36, Jason Wang wrote:
>>
>> On 05/28/2015 11:21 AM, Jason Wang wrote:
>>> On 05/27/2015 07:57 PM, Michael S. Tsirkin wrote:
>>>> On Wed, May 27, 2015 at 04:45:34PM +0800, Jason Wang wrote:
>>>>>> On 05/27/2015 02:26 PM, Pankaj Gupta wrote:
>>>>>>>> Ping.
>>>>>>>>
>>>>>>>> Can I get any suggestions on this patch.
>>>>>>>>
>>>>>>>> Best regards,
>>>>>>>> Pankaj
>>>>>>>>
>>>>>>>>>> vhostforce was added to enable vhost when
>>>>>>>>>> guest don't have MSI-X support.
>>>>>>>>>> Now, we have scenarios like DPDK in Guest which dont use
>>>>>>>>>> interrupts and still use vhost. Also, performance of guests
>>>>>>>>>> without MSI-X support is getting less popular.
>>>>>>>>>>
>>>>>>>>>> Its OK to remove this extra option and enable vhost
>>>>>>>>>> on the basis of vhost=ON/OFF.
>>>>>>>>>> Done basic testing with vhost on/off for latest guests
>>>>>>>>>> and old guests(non-msix).
>>>>>>>>>>
>>>>>>>>>> Signed-off-by: Pankaj Gupta <pagupta@redhat.com>
>>>>>> Looks good. Two questions:
>>>>>>
>>>>>> - Did libvirt use this? if not, we may want to drop vhostfore option
>>>>>> completely.
>>>> Yes, it did.
>>>>
>>> For vhost-user, vhostforce is mandatory. But how about tap? Looks like
>>> it was not used, and I could not even find any option in
>>> http://libvirt.org/formatdomain.html.
>>>
>> CC Michal for the answer.
>>
> No, vhostforce is not used by libvirt at all. Which brings up
> interesting question: for vhost-user libvirt constructs merely the
> follwing cmd line:
>
> -chardev socket,id=charnet0,path=/tmp/vhost0.sock,server \
> -netdev type=vhost-user,id=hostnet0,chardev=charnet0 \
>
> How does vhostforce fit in? I suppose the cmd line is working even
> without libvirt passing vhostforce. Should it do so?
Without this patch, it works with normal virtio-net drivers but may not
work for guest pmd. Since pmd does not use MSI-X interrupts, so qemu
won't start vhost_net. So libvirt needs vhostforce for vhost-user.
With this patch, libvirt does not need to care about this.
>
> Michal
>
next prev parent reply other threads:[~2015-05-28 6:43 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-03-09 6:00 [Qemu-devel] [PATCH] net: Remove vhostforce option in addition to vhost parameter Pankaj Gupta
2015-05-27 6:26 ` Pankaj Gupta
2015-05-27 8:45 ` Jason Wang
2015-05-27 11:57 ` Michael S. Tsirkin
2015-05-28 3:21 ` Jason Wang
2015-05-28 3:36 ` Jason Wang
2015-05-28 6:28 ` Michal Privoznik
2015-05-28 6:43 ` Jason Wang [this message]
2015-05-27 12:00 ` Michael S. Tsirkin
2015-05-29 4:59 ` Pankaj Gupta
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5566B8FF.7060905@redhat.com \
--to=jasowang@redhat.com \
--cc=aliguori@amazon.com \
--cc=mprivozn@redhat.com \
--cc=mst@redhat.com \
--cc=pagupta@redhat.com \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).