From: Nikita Kalyazin <n.kalyazin@samsung.com>
To: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Cc: dev@dpdk.org, "Michael S. Tsirkin" <mst@redhat.com>
Subject: Re: [dpdk-dev] [PATCH v5 resend 07/12] virtio: resolve for control queue
Date: Thu, 08 Oct 2015 18:32:05 +0300 [thread overview]
Message-ID: <20151008153205.GB21145@kalyazin.rnd.samsung.ru> (raw)
In-Reply-To: <20150921063647.GQ2339@yliu-dev.sh.intel.com>
Hi Yuanhan,
As I understand, the dead loop happened here (virtio_send_command):
while (vq->vq_used_cons_idx == vq->vq_ring.used->idx) {
rte_rmb();
usleep(100);
}
Could you explain why wrong config reading caused that and how correct reading helps to avoid?
--
Best regards,
Nikita Kalyazin,
n.kalyazin@samsung.com
Software Engineer
Virtualization Group
Samsung R&D Institute Russia
Tel: +7 (495) 797-25-00 #3816
Tel: +7 (495) 797-25-03
Office #1501, 12-1, Dvintsev str.,
Moscow, 127018, Russia
On Mon, Sep 21, 2015 at 02:36:47PM +0800, Yuanhan Liu wrote:
> On Sun, Sep 20, 2015 at 12:21:14PM +0300, Michael S. Tsirkin wrote:
> > On Fri, Sep 18, 2015 at 11:10:56PM +0800, Yuanhan Liu wrote:
> > > From: Changchun Ouyang <changchun.ouyang@intel.com>
> > >
> > > Fix the max virtio queue pair read issue.
> > >
> > > Control queue can't work for vhost-user mulitple queue mode,
> > > so introduce a counter to void the dead loop when polling
> > > the control queue.
> > >
> > > Signed-off-by: Changchun Ouyang <changchun.ouyang@intel.com>
> > > Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
> >
> > Per virtio spec, the multiqueue feature depends on control queue -
> > what do you mean when you say it can't work?
> >
> > > ---
> > > drivers/net/virtio/virtio_ethdev.c | 12 +++++++-----
> > > 1 file changed, 7 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/drivers/net/virtio/virtio_ethdev.c b/drivers/net/virtio/virtio_ethdev.c
> > > index 465d3cd..b2f4120 100644
> > > --- a/drivers/net/virtio/virtio_ethdev.c
> > > +++ b/drivers/net/virtio/virtio_ethdev.c
> > > @@ -1162,7 +1162,6 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
> > > struct virtio_hw *hw = eth_dev->data->dev_private;
> > > struct virtio_net_config *config;
> > > struct virtio_net_config local_config;
> > > - uint32_t offset_conf = sizeof(config->mac);
> > > struct rte_pci_device *pci_dev;
> > >
> > > RTE_BUILD_BUG_ON(RTE_PKTMBUF_HEADROOM < sizeof(struct virtio_net_hdr));
> > > @@ -1222,7 +1221,9 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
> > > config = &local_config;
> > >
> > > if (vtpci_with_feature(hw, VIRTIO_NET_F_STATUS)) {
> > > - offset_conf += sizeof(config->status);
> > > + vtpci_read_dev_config(hw,
> > > + offsetof(struct virtio_net_config, status),
> > > + &config->status, sizeof(config->status));
> > > } else {
> > > PMD_INIT_LOG(DEBUG,
> > > "VIRTIO_NET_F_STATUS is not supported");
> > > @@ -1230,15 +1231,16 @@ eth_virtio_dev_init(struct rte_eth_dev *eth_dev)
> > > }
> > >
> > > if (vtpci_with_feature(hw, VIRTIO_NET_F_MQ)) {
> > > - offset_conf += sizeof(config->max_virtqueue_pairs);
> > > + vtpci_read_dev_config(hw,
> > > + offsetof(struct virtio_net_config, max_virtqueue_pairs),
> > > + &config->max_virtqueue_pairs,
> > > + sizeof(config->max_virtqueue_pairs));
> > > } else {
> > > PMD_INIT_LOG(DEBUG,
> > > "VIRTIO_NET_F_MQ is not supported");
> > > config->max_virtqueue_pairs = 1;
> > > }
> > >
> > > - vtpci_read_dev_config(hw, 0, (uint8_t *)config, offset_conf);
> > > -
> > > hw->max_rx_queues =
> > > (VIRTIO_MAX_RX_QUEUES < config->max_virtqueue_pairs) ?
> > > VIRTIO_MAX_RX_QUEUES : config->max_virtqueue_pairs;
> >
> >
> > Does the patch actually do what the commit log says?
>
> Sorry, the commit log is wrong as you said.
>
> It was actually a bug in our code, which happens to be revealed when
> MQ is enabled. The old code adjusts the config bytes we want to read
> depending on what kind of features we have, but we later cast the
> entire buf we read with "struct virtio_net_config", which is obviously
> wrong.
>
> The right way to go is to read related config bytes when corresponding
> feature is set, which is exactly what this patch does.
>
> > It seems tobe about reading the device confing,
> > not breaking out of a loop ...
>
> It's just a (bad) side effect of getting the vritio_net_config wrongly:
> the wrong config causes a dead loop in our code.
>
> And sorry for the buggy commit log, will fix it next version.
>
> Thanks.
>
> --yliu
next prev parent reply other threads:[~2015-10-08 15:32 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-09-18 15:10 [dpdk-dev] [PATCH v5 resend 00/12] vhost-user multiple queues enabling Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 01/12] vhost-user: add protocol features support Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 02/12] vhost-user: add VHOST_USER_GET_QUEUE_NUM message Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 03/12] vhost: vring queue setup for multiple queue support Yuanhan Liu
2015-09-20 13:58 ` Marcel Apfelbaum
2015-09-21 2:06 ` Yuanhan Liu
2015-09-21 17:56 ` Marcel Apfelbaum
2015-09-22 7:31 ` Yuanhan Liu
2015-09-22 8:10 ` Marcel Apfelbaum
2015-09-22 8:34 ` Yuanhan Liu
2015-09-22 8:47 ` Marcel Apfelbaum
2015-09-22 9:21 ` Yuanhan Liu
2015-09-22 10:06 ` Marcel Apfelbaum
2015-09-22 14:22 ` Yuanhan Liu
2015-09-22 14:51 ` Marcel Apfelbaum
2015-09-23 3:46 ` Yuanhan Liu
2015-09-24 9:51 ` Marcel Apfelbaum
2015-09-21 9:07 ` Michael S. Tsirkin
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 04/12] vhost: rxtx: prepare work " Yuanhan Liu
2015-09-20 9:29 ` Michael S. Tsirkin
2015-09-21 2:25 ` Yuanhan Liu
2015-09-21 9:04 ` Michael S. Tsirkin
2015-09-22 2:54 ` Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 05/12] vhost: add VHOST_USER_SET_VRING_ENABLE message Yuanhan Liu
2015-09-20 9:37 ` Michael S. Tsirkin
2015-09-21 2:22 ` Yuanhan Liu
2015-09-21 9:02 ` Michael S. Tsirkin
2015-09-22 2:21 ` Yuanhan Liu
2015-09-23 8:43 ` Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 06/12] vhost-user: handle VHOST_USER_RESET_OWNER correctly Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 07/12] virtio: resolve for control queue Yuanhan Liu
2015-09-20 9:21 ` Michael S. Tsirkin
2015-09-21 6:36 ` Yuanhan Liu
2015-10-08 15:32 ` Nikita Kalyazin [this message]
2015-10-08 20:51 ` Steffen Bauch
2015-10-09 7:11 ` Nikita Kalyazin
2015-10-12 8:39 ` Yuanhan Liu
2015-10-12 9:31 ` Xie, Huawei
2015-10-12 9:56 ` Xie, Huawei
2015-10-12 20:58 ` Steffen Bauch
2015-10-13 9:54 ` Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 08/12] vhost-user: enable vhost-user multiple queue Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 09/12] vhost: add API bind a virtq to a specific core Yuanhan Liu
2015-09-18 15:10 ` [dpdk-dev] [PATCH v5 resend 10/12] ixgbe: support VMDq RSS in non-SRIOV environment Yuanhan Liu
2015-09-18 15:11 ` [dpdk-dev] [PATCH v5 resend 11/12] examples/vhost: demonstrate the usage of vhost mq feature Yuanhan Liu
2015-09-18 15:11 ` [dpdk-dev] [PATCH v5 resend 12/12] examples/vhost: add per queue stats Yuanhan Liu
2015-09-20 11:48 ` [dpdk-dev] [PATCH v5 resend 00/12] vhost-user multiple queues enabling Marcel Apfelbaum
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=20151008153205.GB21145@kalyazin.rnd.samsung.ru \
--to=n.kalyazin@samsung.com \
--cc=dev@dpdk.org \
--cc=mst@redhat.com \
--cc=yuanhan.liu@linux.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).