From: Maxime Coquelin <maxime.coquelin@redhat.com>
To: "Wang, Yinan" <yinan.wang@intel.com>,
Adrian Moreno <amorenoz@redhat.com>,
"dev@dpdk.org" <dev@dpdk.org>,
"Xia, Chenbo" <chenbo.xia@intel.com>
Cc: "Fu, Patrick" <patrick.fu@intel.com>,
"stable@dpdk.org" <stable@dpdk.org>,
"Wang, Zhihong" <zhihong.wang@intel.com>,
"Xu, Qian Q" <qian.q.xu@intel.com>
Subject: Re: [dpdk-dev] [PATCH v2 3/3] virtio-user: set status on virtio-user reconnect
Date: Thu, 22 Oct 2020 09:13:43 +0200 [thread overview]
Message-ID: <59a07add-84e4-b111-4fc0-f8a491fe86b0@redhat.com> (raw)
In-Reply-To: <BYAPR11MB2648B4E5F12AC48B0ED1ACBF8F1D0@BYAPR11MB2648.namprd11.prod.outlook.com>
Hi Yinan,
On 10/22/20 6:01 AM, Wang, Yinan wrote:
> Hi Maxime/ Adrian,
>
> Thanks for the patch. we can launch vhost-user with client mode with this fix patch.
> But still fail to get throughput with basic vhost/virtio-user server mode loopback test. This is another problem which introduced by 57912824615fd7787a48a7b18e40661466.
> Bugzilla: https://bugs.dpdk.org/show_bug.cgi?id=541
Thanks for reporting the issue, I will look at it this morning.
BTW, this virtio-user server mode is broken by design, we should really
fix it.
For example, it assumes features to be supported by the backend before
the negotiation took place:
https://git.dpdk.org/dpdk/tree/drivers/net/virtio/virtio_user/virtio_user_dev.c#n513
As part of my rework, I suggest we implement the same behaviour as QEMU
does, which will be reliable and consistent with QEMU. It means that if
server mode is enabled in device command-line, the driver waits until
the socket is ready.
Chenbo, Adrian, what do you think?
Thanks,
Maxime
> BR,
> Yinan
>> -----Original Message-----
>> From: Maxime Coquelin <maxime.coquelin@redhat.com>
>> Sent: 2020年10月21日 0:43
>> To: Adrian Moreno <amorenoz@redhat.com>; dev@dpdk.org
>> Cc: Wang, Yinan <yinan.wang@intel.com>; Fu, Patrick <patrick.fu@intel.com>;
>> stable@dpdk.org; Xia, Chenbo <chenbo.xia@intel.com>; Wang, Zhihong
>> <zhihong.wang@intel.com>
>> Subject: Re: [PATCH v2 3/3] virtio-user: set status on virtio-user reconnect
>>
>>
>>
>> On 10/20/20 5:20 PM, Adrian Moreno wrote:
>>> Newer vhost-user backends will rely on SET_STATUS to start the device
>>> so this required to support them.
>>>
>>> Fixes: 57912824615f ("net/virtio-user: support vhost status setting")
>>> Cc: maxime.coquelin@redhat.com
>>> Cc: stable@dpdk.org
>>>
>>> Signed-off-by: Adrian Moreno <amorenoz@redhat.com>
>>> ---
>>> drivers/net/virtio/virtio_user_ethdev.c | 14 ++++++++++++--
>>> 1 file changed, 12 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/net/virtio/virtio_user_ethdev.c
>> b/drivers/net/virtio/virtio_user_ethdev.c
>>> index e870fb2ff..d8bea4537 100644
>>> --- a/drivers/net/virtio/virtio_user_ethdev.c
>>> +++ b/drivers/net/virtio/virtio_user_ethdev.c
>>> @@ -78,6 +78,13 @@ virtio_user_server_reconnect(struct virtio_user_dev
>> *dev)
>>> return -1;
>>>
>>> dev->vhostfd = connectfd;
>>> +
>>> + vtpci_reset(hw);
>>> +
>>> + vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_ACK);
>>> +
>>> + vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_DRIVER);
>>> +
>>> if (dev->ops->send_request(dev, VHOST_USER_GET_FEATURES,
>>> &dev->device_features) < 0) {
>>> PMD_INIT_LOG(ERR, "get_features failed: %s",
>>> @@ -111,6 +118,8 @@ virtio_user_server_reconnect(struct virtio_user_dev
>> *dev)
>>>
>>> dev->features &= dev->device_features;
>>>
>>> + vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_FEATURES_OK);
>>> +
>>> /* For packed ring, resetting queues is required in reconnection. */
>>> if (vtpci_packed_queue(hw) &&
>>> (vtpci_get_status(hw) & VIRTIO_CONFIG_STATUS_DRIVER_OK)) {
>>> @@ -119,8 +128,9 @@ virtio_user_server_reconnect(struct virtio_user_dev
>> *dev)
>>> virtio_user_reset_queues_packed(eth_dev);
>>> }
>>>
>>> - ret = virtio_user_start_device(dev);
>>> - if (ret < 0)
>>> + /* Start the device */
>>> + vtpci_set_status(hw, VIRTIO_CONFIG_STATUS_DRIVER_OK);
>>> + if (!dev->started)
>>> return -1;
>>>
>>> if (dev->queue_pairs > 1) {
>>>
>>
>> Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>
>>
>> Thanks,
>> Maxime
>
next prev parent reply other threads:[~2020-10-22 7:14 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-20 15:20 [dpdk-dev] [PATCH v2 0/3] virtio-user: fix virtio-user server mode Adrian Moreno
2020-10-20 15:20 ` [dpdk-dev] [PATCH v2 1/3] virtio-user: fix backend selection if stat fails Adrian Moreno
2020-10-20 15:26 ` Maxime Coquelin
2020-10-20 16:52 ` Kevin Traynor
2020-10-20 15:20 ` [dpdk-dev] [PATCH v2 2/3] virtio_user: don't set/get_status until FEATURES_OK Adrian Moreno
2020-10-20 16:17 ` Maxime Coquelin
2020-10-20 15:20 ` [dpdk-dev] [PATCH v2 3/3] virtio-user: set status on virtio-user reconnect Adrian Moreno
2020-10-20 16:42 ` Maxime Coquelin
2020-10-22 4:01 ` Wang, Yinan
2020-10-22 7:13 ` Maxime Coquelin [this message]
2020-10-22 7:37 ` Xia, Chenbo
2020-10-22 8:12 ` Adrian Moreno
2020-10-22 8:32 ` Maxime Coquelin
2020-10-23 5:21 ` Jiang, YuX
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=59a07add-84e4-b111-4fc0-f8a491fe86b0@redhat.com \
--to=maxime.coquelin@redhat.com \
--cc=amorenoz@redhat.com \
--cc=chenbo.xia@intel.com \
--cc=dev@dpdk.org \
--cc=patrick.fu@intel.com \
--cc=qian.q.xu@intel.com \
--cc=stable@dpdk.org \
--cc=yinan.wang@intel.com \
--cc=zhihong.wang@intel.com \
/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).