From: "Michael S. Tsirkin" <mst@redhat.com>
To: Jason Wang <jasowang@redhat.com>
Cc: linux-kernel@vger.kernel.org, yechuan@huawei.com,
eperezma@redhat.com, huangzhichao@huawei.com,
stefanha@redhat.com, "Longpeng\(Mike\)" <longpeng2@huawei.com>,
virtualization@lists.linux-foundation.org
Subject: Re: [PATCH v3 2/2] vdpasim: support doorbell mapping
Date: Mon, 13 Feb 2023 07:05:21 -0500 [thread overview]
Message-ID: <20230213070446-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <CACGkMEvZsfxQW0fVdy0CpqxoWQzz6z=dYK__xFisncuSRms67A@mail.gmail.com>
On Sun, Jan 29, 2023 at 02:19:51PM +0800, Jason Wang wrote:
> On Sun, Jan 29, 2023 at 10:51 AM Longpeng(Mike) <longpeng2@huawei.com> wrote:
> >
> > From: Longpeng <longpeng2@huawei.com>
> >
> > Support doorbell mapping for vdpasim devices, then we can test the notify
> > passthrough feature even if there's no real hardware on hand.
> >
> > Allocates a dummy page which is used to emulate the notify page of the device,
> > all VQs share the same notify register that initiated to 0xffff. A periodic
> > work will check whether there're requests need to process ( the value of the
> > notify register is 0xffff or not ).
> > ---
> > drivers/vdpa/vdpa_sim/vdpa_sim.c | 65 ++++++++++++++++++++++++++++++++
> > drivers/vdpa/vdpa_sim/vdpa_sim.h | 3 ++
> > 2 files changed, 68 insertions(+)
> >
> > diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim.c b/drivers/vdpa/vdpa_sim/vdpa_sim.c
> > index b071f0d842fb..4fcfeb6e2fb8 100644
> > --- a/drivers/vdpa/vdpa_sim/vdpa_sim.c
> > +++ b/drivers/vdpa/vdpa_sim/vdpa_sim.c
> > @@ -39,6 +39,8 @@ MODULE_PARM_DESC(max_iotlb_entries,
> > #define VDPASIM_QUEUE_ALIGN PAGE_SIZE
> > #define VDPASIM_QUEUE_MAX 256
> > #define VDPASIM_VENDOR_ID 0
> > +#define VDPASIM_VRING_POLL_PERIOD 100 /* ms */
> > +#define VDPASIM_NOTIFY_DEFVAL 0xffff
> >
> > static struct vdpasim *vdpa_to_sim(struct vdpa_device *vdpa)
> > {
> > @@ -246,6 +248,28 @@ static const struct dma_map_ops vdpasim_dma_ops = {
> > static const struct vdpa_config_ops vdpasim_config_ops;
> > static const struct vdpa_config_ops vdpasim_batch_config_ops;
> >
> > +static void vdpasim_notify_work(struct work_struct *work)
> > +{
> > + struct vdpasim *vdpasim;
> > + u16 *val;
> > +
> > + vdpasim = container_of(work, struct vdpasim, notify_work.work);
> > +
> > + if (!(vdpasim->status & VIRTIO_CONFIG_S_DRIVER_OK))
> > + goto out;
> > +
> > + if (!vdpasim->running)
> > + goto out;
> > +
> > + val = (u16 *)vdpasim->notify;
> > + if (xchg(val, VDPASIM_NOTIFY_DEFVAL) != VDPASIM_NOTIFY_DEFVAL)
> > + schedule_work(&vdpasim->work);
> > +
> > +out:
> > + schedule_delayed_work(&vdpasim->notify_work,
> > + msecs_to_jiffies(VDPASIM_VRING_POLL_PERIOD));
> > +}
> > +
> > struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *dev_attr,
> > const struct vdpa_dev_set_config *config)
> > {
> > @@ -287,6 +311,13 @@ struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *dev_attr,
> > set_dma_ops(dev, &vdpasim_dma_ops);
> > vdpasim->vdpa.mdev = dev_attr->mgmt_dev;
> >
> > + INIT_DELAYED_WORK(&vdpasim->notify_work, vdpasim_notify_work);
> > +
> > + vdpasim->notify = __get_free_page(GFP_KERNEL | __GFP_ZERO);
> > + if (!vdpasim->notify)
> > + goto err_iommu;
>
> We can simply avoid the advertising notification area in this case.
>
> > + *(u16 *)vdpasim->notify = VDPASIM_NOTIFY_DEFVAL;
>
> WRITE_ONCE()?
it is just initialization so it should not matter.
> > +
> > vdpasim->config = kzalloc(dev_attr->config_size, GFP_KERNEL);
> > if (!vdpasim->config)
> > goto err_iommu;
> > @@ -498,16 +529,21 @@ static u8 vdpasim_get_status(struct vdpa_device *vdpa)
> > static void vdpasim_set_status(struct vdpa_device *vdpa, u8 status)
> > {
> > struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> > + bool started = vdpasim->status & VIRTIO_CONFIG_S_DRIVER_OK;
>
> Do we need to do the check under the vdpasim->lock?
>
> >
> > spin_lock(&vdpasim->lock);
> > vdpasim->status = status;
> > spin_unlock(&vdpasim->lock);
> > + if (!started && (status & VIRTIO_CONFIG_S_DRIVER_OK))
> > + schedule_delayed_work(&vdpasim->notify_work,
> > + msecs_to_jiffies(VDPASIM_VRING_POLL_PERIOD));
> > }
> >
> > static int vdpasim_reset(struct vdpa_device *vdpa)
> > {
> > struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> >
> > + cancel_delayed_work_sync(&vdpasim->notify_work);
>
> Do we need to do this after setting running to zero? Otherwise it's racy.
>
> Thanks
>
> > spin_lock(&vdpasim->lock);
> > vdpasim->status = 0;
> > vdpasim_do_reset(vdpasim);
> > @@ -672,11 +708,34 @@ static int vdpasim_dma_unmap(struct vdpa_device *vdpa, unsigned int asid,
> > return 0;
> > }
> >
> > +static pgprot_t vdpasim_get_vq_notification_pgprot(struct vdpa_device *vdpa,
> > + u16 qid, pgprot_t prot)
> > +{
> > + /*
> > + * We use normal RAM pages to emulate the vq notification area, so
> > + * just keep the pgprot as it mmaped.
> > + */
> > + return prot;
> > +}
> > +
> > +static struct vdpa_notification_area
> > +vdpasim_get_vq_notification(struct vdpa_device *vdpa, u16 qid)
> > +{
> > + struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> > + struct vdpa_notification_area notify;
> > +
> > + notify.addr = virt_to_phys((void *)vdpasim->notify);
> > + notify.size = PAGE_SIZE;
> > +
> > + return notify;
> > +}
> > +
> > static void vdpasim_free(struct vdpa_device *vdpa)
> > {
> > struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> > int i;
> >
> > + cancel_delayed_work_sync(&vdpasim->notify_work);
> > cancel_work_sync(&vdpasim->work);
> >
> > for (i = 0; i < vdpasim->dev_attr.nvqs; i++) {
> > @@ -693,6 +752,8 @@ static void vdpasim_free(struct vdpa_device *vdpa)
> > vhost_iotlb_free(vdpasim->iommu);
> > kfree(vdpasim->vqs);
> > kfree(vdpasim->config);
> > + if (vdpasim->notify)
> > + free_page(vdpasim->notify);
> > }
> >
> > static const struct vdpa_config_ops vdpasim_config_ops = {
> > @@ -704,6 +765,8 @@ static const struct vdpa_config_ops vdpasim_config_ops = {
> > .get_vq_ready = vdpasim_get_vq_ready,
> > .set_vq_state = vdpasim_set_vq_state,
> > .get_vq_state = vdpasim_get_vq_state,
> > + .get_vq_notification = vdpasim_get_vq_notification,
> > + .get_vq_notification_pgprot = vdpasim_get_vq_notification_pgprot,
> > .get_vq_align = vdpasim_get_vq_align,
> > .get_vq_group = vdpasim_get_vq_group,
> > .get_device_features = vdpasim_get_device_features,
> > @@ -737,6 +800,8 @@ static const struct vdpa_config_ops vdpasim_batch_config_ops = {
> > .get_vq_ready = vdpasim_get_vq_ready,
> > .set_vq_state = vdpasim_set_vq_state,
> > .get_vq_state = vdpasim_get_vq_state,
> > + .get_vq_notification = vdpasim_get_vq_notification,
> > + .get_vq_notification_pgprot = vdpasim_get_vq_notification_pgprot,
> > .get_vq_align = vdpasim_get_vq_align,
> > .get_vq_group = vdpasim_get_vq_group,
> > .get_device_features = vdpasim_get_device_features,
> > diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim.h b/drivers/vdpa/vdpa_sim/vdpa_sim.h
> > index 0e78737dcc16..0769ccbd3911 100644
> > --- a/drivers/vdpa/vdpa_sim/vdpa_sim.h
> > +++ b/drivers/vdpa/vdpa_sim/vdpa_sim.h
> > @@ -69,6 +69,9 @@ struct vdpasim {
> > bool running;
> > /* spinlock to synchronize iommu table */
> > spinlock_t iommu_lock;
> > + /* dummy notify page */
> > + unsigned long notify;
> > + struct delayed_work notify_work;
> > };
> >
> > struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *attr,
> > --
> > 2.23.0
> >
_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization
WARNING: multiple messages have this Message-ID (diff)
From: "Michael S. Tsirkin" <mst@redhat.com>
To: Jason Wang <jasowang@redhat.com>
Cc: "Longpeng(Mike)" <longpeng2@huawei.com>,
stefanha@redhat.com, sgarzare@redhat.com,
arei.gonglei@huawei.com, yechuan@huawei.com,
huangzhichao@huawei.com,
virtualization@lists.linux-foundation.org,
linux-kernel@vger.kernel.org, eperezma@redhat.com
Subject: Re: [PATCH v3 2/2] vdpasim: support doorbell mapping
Date: Mon, 13 Feb 2023 07:05:21 -0500 [thread overview]
Message-ID: <20230213070446-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <CACGkMEvZsfxQW0fVdy0CpqxoWQzz6z=dYK__xFisncuSRms67A@mail.gmail.com>
On Sun, Jan 29, 2023 at 02:19:51PM +0800, Jason Wang wrote:
> On Sun, Jan 29, 2023 at 10:51 AM Longpeng(Mike) <longpeng2@huawei.com> wrote:
> >
> > From: Longpeng <longpeng2@huawei.com>
> >
> > Support doorbell mapping for vdpasim devices, then we can test the notify
> > passthrough feature even if there's no real hardware on hand.
> >
> > Allocates a dummy page which is used to emulate the notify page of the device,
> > all VQs share the same notify register that initiated to 0xffff. A periodic
> > work will check whether there're requests need to process ( the value of the
> > notify register is 0xffff or not ).
> > ---
> > drivers/vdpa/vdpa_sim/vdpa_sim.c | 65 ++++++++++++++++++++++++++++++++
> > drivers/vdpa/vdpa_sim/vdpa_sim.h | 3 ++
> > 2 files changed, 68 insertions(+)
> >
> > diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim.c b/drivers/vdpa/vdpa_sim/vdpa_sim.c
> > index b071f0d842fb..4fcfeb6e2fb8 100644
> > --- a/drivers/vdpa/vdpa_sim/vdpa_sim.c
> > +++ b/drivers/vdpa/vdpa_sim/vdpa_sim.c
> > @@ -39,6 +39,8 @@ MODULE_PARM_DESC(max_iotlb_entries,
> > #define VDPASIM_QUEUE_ALIGN PAGE_SIZE
> > #define VDPASIM_QUEUE_MAX 256
> > #define VDPASIM_VENDOR_ID 0
> > +#define VDPASIM_VRING_POLL_PERIOD 100 /* ms */
> > +#define VDPASIM_NOTIFY_DEFVAL 0xffff
> >
> > static struct vdpasim *vdpa_to_sim(struct vdpa_device *vdpa)
> > {
> > @@ -246,6 +248,28 @@ static const struct dma_map_ops vdpasim_dma_ops = {
> > static const struct vdpa_config_ops vdpasim_config_ops;
> > static const struct vdpa_config_ops vdpasim_batch_config_ops;
> >
> > +static void vdpasim_notify_work(struct work_struct *work)
> > +{
> > + struct vdpasim *vdpasim;
> > + u16 *val;
> > +
> > + vdpasim = container_of(work, struct vdpasim, notify_work.work);
> > +
> > + if (!(vdpasim->status & VIRTIO_CONFIG_S_DRIVER_OK))
> > + goto out;
> > +
> > + if (!vdpasim->running)
> > + goto out;
> > +
> > + val = (u16 *)vdpasim->notify;
> > + if (xchg(val, VDPASIM_NOTIFY_DEFVAL) != VDPASIM_NOTIFY_DEFVAL)
> > + schedule_work(&vdpasim->work);
> > +
> > +out:
> > + schedule_delayed_work(&vdpasim->notify_work,
> > + msecs_to_jiffies(VDPASIM_VRING_POLL_PERIOD));
> > +}
> > +
> > struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *dev_attr,
> > const struct vdpa_dev_set_config *config)
> > {
> > @@ -287,6 +311,13 @@ struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *dev_attr,
> > set_dma_ops(dev, &vdpasim_dma_ops);
> > vdpasim->vdpa.mdev = dev_attr->mgmt_dev;
> >
> > + INIT_DELAYED_WORK(&vdpasim->notify_work, vdpasim_notify_work);
> > +
> > + vdpasim->notify = __get_free_page(GFP_KERNEL | __GFP_ZERO);
> > + if (!vdpasim->notify)
> > + goto err_iommu;
>
> We can simply avoid the advertising notification area in this case.
>
> > + *(u16 *)vdpasim->notify = VDPASIM_NOTIFY_DEFVAL;
>
> WRITE_ONCE()?
it is just initialization so it should not matter.
> > +
> > vdpasim->config = kzalloc(dev_attr->config_size, GFP_KERNEL);
> > if (!vdpasim->config)
> > goto err_iommu;
> > @@ -498,16 +529,21 @@ static u8 vdpasim_get_status(struct vdpa_device *vdpa)
> > static void vdpasim_set_status(struct vdpa_device *vdpa, u8 status)
> > {
> > struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> > + bool started = vdpasim->status & VIRTIO_CONFIG_S_DRIVER_OK;
>
> Do we need to do the check under the vdpasim->lock?
>
> >
> > spin_lock(&vdpasim->lock);
> > vdpasim->status = status;
> > spin_unlock(&vdpasim->lock);
> > + if (!started && (status & VIRTIO_CONFIG_S_DRIVER_OK))
> > + schedule_delayed_work(&vdpasim->notify_work,
> > + msecs_to_jiffies(VDPASIM_VRING_POLL_PERIOD));
> > }
> >
> > static int vdpasim_reset(struct vdpa_device *vdpa)
> > {
> > struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> >
> > + cancel_delayed_work_sync(&vdpasim->notify_work);
>
> Do we need to do this after setting running to zero? Otherwise it's racy.
>
> Thanks
>
> > spin_lock(&vdpasim->lock);
> > vdpasim->status = 0;
> > vdpasim_do_reset(vdpasim);
> > @@ -672,11 +708,34 @@ static int vdpasim_dma_unmap(struct vdpa_device *vdpa, unsigned int asid,
> > return 0;
> > }
> >
> > +static pgprot_t vdpasim_get_vq_notification_pgprot(struct vdpa_device *vdpa,
> > + u16 qid, pgprot_t prot)
> > +{
> > + /*
> > + * We use normal RAM pages to emulate the vq notification area, so
> > + * just keep the pgprot as it mmaped.
> > + */
> > + return prot;
> > +}
> > +
> > +static struct vdpa_notification_area
> > +vdpasim_get_vq_notification(struct vdpa_device *vdpa, u16 qid)
> > +{
> > + struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> > + struct vdpa_notification_area notify;
> > +
> > + notify.addr = virt_to_phys((void *)vdpasim->notify);
> > + notify.size = PAGE_SIZE;
> > +
> > + return notify;
> > +}
> > +
> > static void vdpasim_free(struct vdpa_device *vdpa)
> > {
> > struct vdpasim *vdpasim = vdpa_to_sim(vdpa);
> > int i;
> >
> > + cancel_delayed_work_sync(&vdpasim->notify_work);
> > cancel_work_sync(&vdpasim->work);
> >
> > for (i = 0; i < vdpasim->dev_attr.nvqs; i++) {
> > @@ -693,6 +752,8 @@ static void vdpasim_free(struct vdpa_device *vdpa)
> > vhost_iotlb_free(vdpasim->iommu);
> > kfree(vdpasim->vqs);
> > kfree(vdpasim->config);
> > + if (vdpasim->notify)
> > + free_page(vdpasim->notify);
> > }
> >
> > static const struct vdpa_config_ops vdpasim_config_ops = {
> > @@ -704,6 +765,8 @@ static const struct vdpa_config_ops vdpasim_config_ops = {
> > .get_vq_ready = vdpasim_get_vq_ready,
> > .set_vq_state = vdpasim_set_vq_state,
> > .get_vq_state = vdpasim_get_vq_state,
> > + .get_vq_notification = vdpasim_get_vq_notification,
> > + .get_vq_notification_pgprot = vdpasim_get_vq_notification_pgprot,
> > .get_vq_align = vdpasim_get_vq_align,
> > .get_vq_group = vdpasim_get_vq_group,
> > .get_device_features = vdpasim_get_device_features,
> > @@ -737,6 +800,8 @@ static const struct vdpa_config_ops vdpasim_batch_config_ops = {
> > .get_vq_ready = vdpasim_get_vq_ready,
> > .set_vq_state = vdpasim_set_vq_state,
> > .get_vq_state = vdpasim_get_vq_state,
> > + .get_vq_notification = vdpasim_get_vq_notification,
> > + .get_vq_notification_pgprot = vdpasim_get_vq_notification_pgprot,
> > .get_vq_align = vdpasim_get_vq_align,
> > .get_vq_group = vdpasim_get_vq_group,
> > .get_device_features = vdpasim_get_device_features,
> > diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim.h b/drivers/vdpa/vdpa_sim/vdpa_sim.h
> > index 0e78737dcc16..0769ccbd3911 100644
> > --- a/drivers/vdpa/vdpa_sim/vdpa_sim.h
> > +++ b/drivers/vdpa/vdpa_sim/vdpa_sim.h
> > @@ -69,6 +69,9 @@ struct vdpasim {
> > bool running;
> > /* spinlock to synchronize iommu table */
> > spinlock_t iommu_lock;
> > + /* dummy notify page */
> > + unsigned long notify;
> > + struct delayed_work notify_work;
> > };
> >
> > struct vdpasim *vdpasim_create(struct vdpasim_dev_attr *attr,
> > --
> > 2.23.0
> >
next prev parent reply other threads:[~2023-02-13 12:05 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-01-29 2:50 [PATCH v3 0/2] vdpasim: support doorbell mapping Longpeng(Mike)
2023-01-29 2:50 ` [PATCH v3 1/2] vdpa: support specify the pgprot of vq notification area Longpeng(Mike)
2023-01-29 6:03 ` Jason Wang
2023-01-29 6:03 ` Jason Wang
2023-01-29 2:50 ` [PATCH v3 2/2] vdpasim: support doorbell mapping Longpeng(Mike)
2023-01-29 6:19 ` Jason Wang
2023-01-29 6:19 ` Jason Wang
2023-02-13 12:05 ` Michael S. Tsirkin [this message]
2023-02-13 12:05 ` Michael S. Tsirkin
2023-02-14 2:05 ` Longpeng (Mike, Cloud Infrastructure Service Product Dept.)
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=20230213070446-mutt-send-email-mst@kernel.org \
--to=mst@redhat.com \
--cc=eperezma@redhat.com \
--cc=huangzhichao@huawei.com \
--cc=jasowang@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=longpeng2@huawei.com \
--cc=stefanha@redhat.com \
--cc=virtualization@lists.linux-foundation.org \
--cc=yechuan@huawei.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.