From: "Michael S. Tsirkin" <mst@redhat.com>
To: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
Cc: virtualization@lists.linux-foundation.org,
netdev@vger.kernel.org, Jason Wang <jasowang@redhat.com>,
"David S. Miller" <davem@davemloft.net>,
Jakub Kicinski <kuba@kernel.org>,
Alexei Starovoitov <ast@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Jesper Dangaard Brouer <hawk@kernel.org>,
John Fastabend <john.fastabend@gmail.com>,
bpf@vger.kernel.org
Subject: Re: [PATCH v2 07/12] virtio: queue_reset: pci: support VIRTIO_F_RING_RESET
Date: Thu, 20 Jan 2022 05:55:14 -0500 [thread overview]
Message-ID: <20220120055227-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <20220120064303.106639-8-xuanzhuo@linux.alibaba.com>
On Thu, Jan 20, 2022 at 02:42:58PM +0800, Xuan Zhuo wrote:
> This patch implements virtio pci support for QUEUE RESET.
>
> Performing reset on a queue is divided into two steps:
>
> 1. reset_vq: reset one vq
> 2. enable_reset_vq: re-enable the reset queue
>
> In the first step, these tasks will be completed:
> 1. notify the hardware queue to reset
> 2. recycle the buffer from vq
> 3. delete the vq
>
> When deleting a vq, vp_del_vq() will be called to release all the memory
> of the vq. But this does not affect the process of deleting vqs, because
> that is based on the queue to release all the vqs. During this process,
> the vq has been removed from the queue.
>
> When deleting vq, info and vq will be released, and I save msix_vec in
> vp_dev->vqs[queue_index]. When re-enable, the current msix_vec can be
> reused. And based on intx_enabled to determine which method to use to
> enable this queue.
>
> Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
There's something I don't understand here. It looks like
you assume that when you reset a queue, you also
reset the mapping from queue to event vector.
The spec does not say it should, and I don't think it's
useful to extend spec to do it - we already have a simple
way to tweak the mapping.
Avoid doing that, and things will be much easier, with no need
to interact with a transport, won't they?
> ---
> drivers/virtio/virtio_pci_common.c | 49 ++++++++++++++++++++
> drivers/virtio/virtio_pci_common.h | 4 ++
> drivers/virtio/virtio_pci_modern.c | 73 ++++++++++++++++++++++++++++++
> 3 files changed, 126 insertions(+)
>
> diff --git a/drivers/virtio/virtio_pci_common.c b/drivers/virtio/virtio_pci_common.c
> index 5afe207ce28a..28b5ffde4621 100644
> --- a/drivers/virtio/virtio_pci_common.c
> +++ b/drivers/virtio/virtio_pci_common.c
> @@ -464,6 +464,55 @@ int vp_find_vqs(struct virtio_device *vdev, unsigned nvqs,
> return vp_find_vqs_intx(vdev, nvqs, vqs, callbacks, names, ctx);
> }
>
> +#define VQ_IS_DELETED(vp_dev, idx) ((unsigned long)vp_dev->vqs[idx] & 1)
> +#define VQ_RESET_MSIX_VEC(vp_dev, idx) ((unsigned long)vp_dev->vqs[idx] >> 2)
> +#define VQ_RESET_MARK(msix_vec) ((void *)(long)((msix_vec << 2) + 1))
> +
> +void vp_del_reset_vq(struct virtio_device *vdev, u16 queue_index)
> +{
> + struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> + struct virtio_pci_vq_info *info;
> + u16 msix_vec;
> +
> + info = vp_dev->vqs[queue_index];
> +
> + msix_vec = info->msix_vector;
> +
> + /* delete vq */
> + vp_del_vq(info->vq);
> +
> + /* Mark the vq has been deleted, and save the msix_vec. */
> + vp_dev->vqs[queue_index] = VQ_RESET_MARK(msix_vec);
> +}
> +
> +struct virtqueue *vp_enable_reset_vq(struct virtio_device *vdev,
> + int queue_index,
> + vq_callback_t *callback,
> + const char *name,
> + const bool ctx)
> +{
> + struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> + struct virtqueue *vq;
> + u16 msix_vec;
> +
> + if (!VQ_IS_DELETED(vp_dev, queue_index))
> + return ERR_PTR(-EPERM);
> +
> + msix_vec = VQ_RESET_MSIX_VEC(vp_dev, queue_index);
> +
> + if (vp_dev->intx_enabled)
> + vq = vp_setup_vq(vdev, queue_index, callback, name, ctx,
> + VIRTIO_MSI_NO_VECTOR);
> + else
> + vq = vp_enable_vq_msix(vdev, queue_index, callback, name, ctx,
> + msix_vec);
> +
> + if (IS_ERR(vq))
> + vp_dev->vqs[queue_index] = VQ_RESET_MARK(msix_vec);
> +
> + return vq;
> +}
> +
> const char *vp_bus_name(struct virtio_device *vdev)
> {
> struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> diff --git a/drivers/virtio/virtio_pci_common.h b/drivers/virtio/virtio_pci_common.h
> index 23f6c5c678d5..96c13b1398f8 100644
> --- a/drivers/virtio/virtio_pci_common.h
> +++ b/drivers/virtio/virtio_pci_common.h
> @@ -115,6 +115,10 @@ int vp_find_vqs(struct virtio_device *vdev, unsigned nvqs,
> struct virtqueue *vqs[], vq_callback_t *callbacks[],
> const char * const names[], const bool *ctx,
> struct irq_affinity *desc);
> +void vp_del_reset_vq(struct virtio_device *vdev, u16 queue_index);
> +struct virtqueue *vp_enable_reset_vq(struct virtio_device *vdev, int queue_index,
> + vq_callback_t *callback, const char *name,
> + const bool ctx);
> const char *vp_bus_name(struct virtio_device *vdev);
>
> /* Setup the affinity for a virtqueue:
> diff --git a/drivers/virtio/virtio_pci_modern.c b/drivers/virtio/virtio_pci_modern.c
> index 5455bc041fb6..fbf87239c920 100644
> --- a/drivers/virtio/virtio_pci_modern.c
> +++ b/drivers/virtio/virtio_pci_modern.c
> @@ -34,6 +34,9 @@ static void vp_transport_features(struct virtio_device *vdev, u64 features)
> if ((features & BIT_ULL(VIRTIO_F_SR_IOV)) &&
> pci_find_ext_capability(pci_dev, PCI_EXT_CAP_ID_SRIOV))
> __virtio_set_bit(vdev, VIRTIO_F_SR_IOV);
> +
> + if (features & BIT_ULL(VIRTIO_F_RING_RESET))
> + __virtio_set_bit(vdev, VIRTIO_F_RING_RESET);
> }
>
> /* virtio config->finalize_features() implementation */
> @@ -176,6 +179,72 @@ static void vp_reset(struct virtio_device *vdev)
> vp_disable_cbs(vdev);
> }
>
> +static int vp_modern_reset_vq(struct virtio_device *vdev, u16 queue_index,
> + vq_reset_callback_t *callback, void *data)
> +{
> + struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> + struct virtio_pci_modern_device *mdev = &vp_dev->mdev;
> + struct virtio_pci_vq_info *info;
> + u16 msix_vec;
> + void *buf;
> +
> + if (!virtio_has_feature(vdev, VIRTIO_F_RING_RESET))
> + return -ENOENT;
> +
> + vp_modern_set_queue_reset(mdev, queue_index);
> +
> + /* After write 1 to queue reset, the driver MUST wait for a read of
> + * queue reset to return 1.
> + */
> + while (vp_modern_get_queue_reset(mdev, queue_index) != 1)
> + msleep(1);
> +
> + info = vp_dev->vqs[queue_index];
> + msix_vec = info->msix_vector;
> +
> + /* Disable VQ callback. */
> + if (vp_dev->per_vq_vectors && msix_vec != VIRTIO_MSI_NO_VECTOR)
> + disable_irq(pci_irq_vector(vp_dev->pci_dev, msix_vec));
> +
> + while ((buf = virtqueue_detach_unused_buf(info->vq)) != NULL)
> + callback(vdev, buf, data);
> +
> + vp_del_reset_vq(vdev, queue_index);
> +
> + return 0;
> +}
> +
> +static struct virtqueue *vp_modern_enable_reset_vq(struct virtio_device *vdev,
> + u16 queue_index,
> + vq_callback_t *callback,
> + const char *name,
> + const bool *ctx)
> +{
> + struct virtio_pci_device *vp_dev = to_vp_device(vdev);
> + struct virtio_pci_modern_device *mdev = &vp_dev->mdev;
> + struct virtqueue *vq;
> + u16 msix_vec;
> +
> + if (!virtio_has_feature(vdev, VIRTIO_F_RING_RESET))
> + return ERR_PTR(-ENOENT);
> +
> + /* check queue reset status */
> + if (vp_modern_get_queue_reset(mdev, queue_index) != 1)
> + return ERR_PTR(-EBUSY);
> +
> + vq = vp_enable_reset_vq(vdev, queue_index, callback, name, ctx);
> + if (IS_ERR(vq))
> + return vq;
> +
> + vp_modern_set_queue_enable(&vp_dev->mdev, vq->index, true);
> +
> + msix_vec = vp_dev->vqs[queue_index]->msix_vector;
> + if (vp_dev->per_vq_vectors && msix_vec != VIRTIO_MSI_NO_VECTOR)
> + enable_irq(pci_irq_vector(vp_dev->pci_dev, msix_vec));
> +
> + return vq;
> +}
> +
> static u16 vp_config_vector(struct virtio_pci_device *vp_dev, u16 vector)
> {
> return vp_modern_config_vector(&vp_dev->mdev, vector);
> @@ -395,6 +464,8 @@ static const struct virtio_config_ops virtio_pci_config_nodev_ops = {
> .set_vq_affinity = vp_set_vq_affinity,
> .get_vq_affinity = vp_get_vq_affinity,
> .get_shm_region = vp_get_shm_region,
> + .reset_vq = vp_modern_reset_vq,
> + .enable_reset_vq = vp_modern_enable_reset_vq,
> };
>
> static const struct virtio_config_ops virtio_pci_config_ops = {
> @@ -413,6 +484,8 @@ static const struct virtio_config_ops virtio_pci_config_ops = {
> .set_vq_affinity = vp_set_vq_affinity,
> .get_vq_affinity = vp_get_vq_affinity,
> .get_shm_region = vp_get_shm_region,
> + .reset_vq = vp_modern_reset_vq,
> + .enable_reset_vq = vp_modern_enable_reset_vq,
> };
>
> /* the PCI probing function */
> --
> 2.31.0
next prev parent reply other threads:[~2022-01-20 10:55 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-01-20 6:42 [PATCH v2 00/12] virtio pci support VIRTIO_F_RING_RESET Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 01/12] virtio: pci: struct virtio_pci_common_cfg add queue_notify_data Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 02/12] virtio: queue_reset: add VIRTIO_F_RING_RESET Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 03/12] virtio: queue_reset: struct virtio_config_ops add callbacks for queue_reset Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 04/12] virtio: queue_reset: pci: update struct virtio_pci_common_cfg and option functions Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 05/12] virito: queue_reset: pci: move the per queue irq logic from vp_del_vqs to vp_del_vq Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 06/12] virtio: queue_reset: pci: add independent function to enable msix vq Xuan Zhuo
2022-01-20 6:42 ` [PATCH v2 07/12] virtio: queue_reset: pci: support VIRTIO_F_RING_RESET Xuan Zhuo
2022-01-20 10:55 ` Michael S. Tsirkin [this message]
[not found] ` <1642679180.4063296-1-xuanzhuo@linux.alibaba.com>
2022-01-20 15:03 ` Michael S. Tsirkin
[not found] ` <1642731779.2471316-1-xuanzhuo@linux.alibaba.com>
2022-01-21 10:22 ` Michael S. Tsirkin
2022-01-21 10:22 ` Michael S. Tsirkin
[not found] ` <1642760793.1188169-1-xuanzhuo@linux.alibaba.com>
2022-01-21 13:19 ` Michael S. Tsirkin
[not found] ` <1642774171.933696-1-xuanzhuo@linux.alibaba.com>
2022-01-21 15:28 ` Michael S. Tsirkin
[not found] ` <1642779265.2774203-1-xuanzhuo@linux.alibaba.com>
2022-01-21 15:45 ` Michael S. Tsirkin
2022-01-20 6:42 ` [PATCH v2 08/12] virtio: queue_reset: add helper Xuan Zhuo
2022-01-20 6:43 ` [PATCH v2 09/12] virtio_net: virtnet_tx_timeout() fix style Xuan Zhuo
2022-01-20 6:43 ` [PATCH v2 10/12] virtio_net: virtnet_tx_timeout() stop ref sq->vq Xuan Zhuo
2022-01-20 6:43 ` [PATCH v2 11/12] virtio_net: split free_unused_bufs() Xuan Zhuo
2022-01-20 6:43 ` [PATCH v2 12/12] virtio-net: support pair disable/enable Xuan Zhuo
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=20220120055227-mutt-send-email-mst@kernel.org \
--to=mst@redhat.com \
--cc=ast@kernel.org \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=hawk@kernel.org \
--cc=jasowang@redhat.com \
--cc=john.fastabend@gmail.com \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=virtualization@lists.linux-foundation.org \
--cc=xuanzhuo@linux.alibaba.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).