DPDK patches and discussions
 help / color / mirror / Atom feed
From: David Marchand <david.marchand@redhat.com>
To: Maxime Coquelin <maxime.coquelin@redhat.com>
Cc: dev <dev@dpdk.org>, Matan Azrad <matan@mellanox.com>,
	 "Xia, Chenbo" <chenbo.xia@intel.com>,
	Marvin Liu <yong.liu@intel.com>,
	 "Wang, Yinan" <yinan.wang@intel.com>,
	Thomas Monjalon <thomas@monjalon.net>,
	 "Yigit, Ferruh" <ferruh.yigit@intel.com>
Subject: Re: [dpdk-dev] [PATCH v3 3/3] net/vhost: fix interrupt mode
Date: Wed, 29 Jul 2020 13:27:28 +0200
Message-ID: <CAJFAV8yU=6aSc3Rs7ARwON9FW+Pi_pSo=mYNcBFmF2iKE9dqPQ@mail.gmail.com> (raw)
In-Reply-To: <20200729092000.233036-4-maxime.coquelin@redhat.com>

On Wed, Jul 29, 2020 at 11:20 AM Maxime Coquelin
<maxime.coquelin@redhat.com> wrote:
>
> At .new_device() time, only the first vring pair is
> now ready, other vrings are consfigured later.

configured*

>
> Problem is that when application will setup and enable
> interrupts, only the first queue pair Rx interrupt will
> be enabled.
>
> This patches fixes the issue by setting the number of
> max interrupts to the number of Rx queues that will be
> later initialized. Then, as soon as a Rx vring is ready
> and interrupt enabled by the application, it removes the
> corresponding uninitialized epoll event, and install a

installs*

> new one with the valid FD.
>
> Fixes: 604052ae5395 ("net/vhost: support queue update")
>
> Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>

It seems a bit of a hack, but I _think_ the patch is good wrt races on
epoll configuration.

We are only touching the vhost pmd, in interrupt mode.
The interrupt mode is not that frequently used (I found no usage in
opensource projects).
The vhost pmd is not used in OVS as it lags behind the vhost library
and has limitations.

So my opinion is that the risk of taking this patch rather than
reverting the changes (which is not trivial iiuc) in the vhost library
is acceptable.


One comment below:

> ---
>  drivers/net/vhost/rte_eth_vhost.c | 75 +++++++++++++++++++++++++++----
>  1 file changed, 66 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/net/vhost/rte_eth_vhost.c b/drivers/net/vhost/rte_eth_vhost.c
> index 951929c663..237785dd66 100644
> --- a/drivers/net/vhost/rte_eth_vhost.c
> +++ b/drivers/net/vhost/rte_eth_vhost.c
> @@ -5,6 +5,7 @@
>  #include <unistd.h>
>  #include <pthread.h>
>  #include <stdbool.h>
> +#include <sys/epoll.h>
>
>  #include <rte_mbuf.h>
>  #include <rte_ethdev_driver.h>
> @@ -95,6 +96,8 @@ struct vhost_queue {
>         uint16_t port;
>         uint16_t virtqueue_id;
>         struct vhost_stats stats;
> +       int intr_enable;
> +       rte_spinlock_t intr_lock;
>  };
>
>  struct pmd_internal {
> @@ -524,6 +527,45 @@ find_internal_resource(char *ifname)
>         return list;
>  }
>
> +static int
> +eth_vhost_update_intr(struct rte_eth_dev *eth_dev, uint16_t rxq_idx)
> +{
> +       struct rte_intr_handle *handle = eth_dev->intr_handle;
> +       struct rte_epoll_event rev;
> +       int epfd, ret;
> +
> +       if (handle->efds[rxq_idx] == handle->elist[rxq_idx].fd)
> +               return 0;

Feel free to ignore if this situation can not happen.

We are expecting only -1 -> valid fd transitions.
Maybe add an error log if we are in another situation?
This would indicate something quite broken.



> +
> +       VHOST_LOG(INFO, "kickfd for rxq-%d was changed, updating handler.\n",
> +                       rxq_idx);
> +
> +       /*
> +        * First remove invalid epoll event, and then isntall
> +        * the new one. May be solved with a proper API in the
> +        * future.
> +        */
> +       epfd = handle->elist[rxq_idx].epfd;
> +       rev = handle->elist[rxq_idx];
> +       ret = rte_epoll_ctl(epfd, EPOLL_CTL_DEL, rev.fd,
> +                       &handle->elist[rxq_idx]);
> +       if (ret) {
> +               VHOST_LOG(ERR, "Delete epoll event failed.\n");
> +               return ret;
> +       }
> +
> +       rev.fd = handle->efds[rxq_idx];
> +       handle->elist[rxq_idx] = rev;
> +       ret = rte_epoll_ctl(epfd, EPOLL_CTL_ADD, rev.fd,
> +                       &handle->elist[rxq_idx]);
> +       if (ret) {
> +               VHOST_LOG(ERR, "Add epoll event failed.\n");
> +               return ret;
> +       }
> +
> +       return 0;
> +}
> +
>  static int
>  eth_rxq_intr_enable(struct rte_eth_dev *dev, uint16_t qid)
>  {
> @@ -537,6 +579,11 @@ eth_rxq_intr_enable(struct rte_eth_dev *dev, uint16_t qid)
>                 return -1;
>         }
>
> +       rte_spinlock_lock(&vq->intr_lock);
> +       vq->intr_enable = 1;
> +       ret = eth_vhost_update_intr(dev, qid);
> +       rte_spinlock_unlock(&vq->intr_lock);
> +
>         ret = rte_vhost_get_vhost_vring(vq->vid, (qid << 1) + 1, &vring);
>         if (ret < 0) {
>                 VHOST_LOG(ERR, "Failed to get rxq%d's vring\n", qid);
> @@ -571,6 +618,8 @@ eth_rxq_intr_disable(struct rte_eth_dev *dev, uint16_t qid)
>         rte_vhost_enable_guest_notification(vq->vid, (qid << 1) + 1, 0);
>         rte_wmb();
>
> +       vq->intr_enable = 0;
> +
>         return 0;
>  }
>
> @@ -593,7 +642,6 @@ eth_vhost_install_intr(struct rte_eth_dev *dev)
>  {
>         struct rte_vhost_vring vring;
>         struct vhost_queue *vq;
> -       int count = 0;
>         int nb_rxq = dev->data->nb_rx_queues;
>         int i;
>         int ret;
> @@ -623,6 +671,8 @@ eth_vhost_install_intr(struct rte_eth_dev *dev)
>
>         VHOST_LOG(INFO, "Prepare intr vec\n");
>         for (i = 0; i < nb_rxq; i++) {
> +               dev->intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + i;
> +               dev->intr_handle->efds[i] = -1;
>                 vq = dev->data->rx_queues[i];
>                 if (!vq) {
>                         VHOST_LOG(INFO, "rxq-%d not setup yet, skip!\n", i);
> @@ -641,14 +691,12 @@ eth_vhost_install_intr(struct rte_eth_dev *dev)
>                                 "rxq-%d's kickfd is invalid, skip!\n", i);
>                         continue;
>                 }
> -               dev->intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + i;
>                 dev->intr_handle->efds[i] = vring.kickfd;
> -               count++;
>                 VHOST_LOG(INFO, "Installed intr vec for rxq-%d\n", i);
>         }
>
> -       dev->intr_handle->nb_efd = count;
> -       dev->intr_handle->max_intr = count + 1;
> +       dev->intr_handle->nb_efd = nb_rxq;
> +       dev->intr_handle->max_intr = nb_rxq + 1;
>         dev->intr_handle->type = RTE_INTR_HANDLE_VDEV;
>
>         return 0;
> @@ -835,6 +883,7 @@ vring_conf_update(int vid, struct rte_eth_dev *eth_dev, uint16_t vring_id)
>  {
>         struct rte_eth_conf *dev_conf = &eth_dev->data->dev_conf;
>         struct pmd_internal *internal = eth_dev->data->dev_private;
> +       struct vhost_queue *vq;
>         struct rte_vhost_vring vring;
>         int rx_idx = vring_id % 2 ? (vring_id - 1) >> 1 : -1;
>         int ret = 0;
> @@ -853,12 +902,18 @@ vring_conf_update(int vid, struct rte_eth_dev *eth_dev, uint16_t vring_id)
>                                         vring_id);
>                         return ret;
>                 }
> +               eth_dev->intr_handle->efds[rx_idx] = vring.kickfd;
>
> -               if (vring.kickfd != eth_dev->intr_handle->efds[rx_idx]) {
> -                       VHOST_LOG(INFO, "kickfd for rxq-%d was changed.\n",
> -                                         rx_idx);
> -                       eth_dev->intr_handle->efds[rx_idx] = vring.kickfd;
> +               vq = eth_dev->data->rx_queues[rx_idx];
> +               if (!vq) {
> +                       VHOST_LOG(ERR, "rxq%d is not setup yet\n", rx_idx);
> +                       return -1;
>                 }
> +
> +               rte_spinlock_lock(&vq->intr_lock);
> +               if (vq->intr_enable)
> +                       ret = eth_vhost_update_intr(eth_dev, rx_idx);
> +               rte_spinlock_unlock(&vq->intr_lock);
>         }
>
>         return ret;
> @@ -1152,6 +1207,7 @@ eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
>
>         vq->mb_pool = mb_pool;
>         vq->virtqueue_id = rx_queue_id * VIRTIO_QNUM + VIRTIO_TXQ;
> +       rte_spinlock_init(&vq->intr_lock);
>         dev->data->rx_queues[rx_queue_id] = vq;
>
>         return 0;
> @@ -1173,6 +1229,7 @@ eth_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id,
>         }
>
>         vq->virtqueue_id = tx_queue_id * VIRTIO_QNUM + VIRTIO_RXQ;
> +       rte_spinlock_init(&vq->intr_lock);
>         dev->data->tx_queues[tx_queue_id] = vq;
>
>         return 0;
> --
> 2.26.2
>



--
David Marchand


  reply	other threads:[~2020-07-29 11:27 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-29  9:19 [dpdk-dev] [PATCH v3 0/3] Fix Vhost regressions Maxime Coquelin
2020-07-29  9:19 ` [dpdk-dev] [PATCH v3 1/3] vhost: fix guest notification setting Maxime Coquelin
2020-07-29  9:19 ` [dpdk-dev] [PATCH v3 2/3] net/vhost: fix queue update Maxime Coquelin
2020-07-29  9:20 ` [dpdk-dev] [PATCH v3 3/3] net/vhost: fix interrupt mode Maxime Coquelin
2020-07-29 11:27   ` David Marchand [this message]
2020-07-29 13:19     ` Maxime Coquelin
2020-07-29 12:53   ` Maxime Coquelin
2020-07-29 13:24     ` Xia, Chenbo
2020-07-29 13:27       ` Maxime Coquelin

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='CAJFAV8yU=6aSc3Rs7ARwON9FW+Pi_pSo=mYNcBFmF2iKE9dqPQ@mail.gmail.com' \
    --to=david.marchand@redhat.com \
    --cc=chenbo.xia@intel.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@intel.com \
    --cc=matan@mellanox.com \
    --cc=maxime.coquelin@redhat.com \
    --cc=thomas@monjalon.net \
    --cc=yinan.wang@intel.com \
    --cc=yong.liu@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

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ https://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git