mbox series

[vhost,v7,00/11] virtio core prepares for AF_XDP

Message ID 20230425073613.8839-1-xuanzhuo@linux.alibaba.com (mailing list archive)
Headers show
Series virtio core prepares for AF_XDP | expand

Message

Xuan Zhuo April 25, 2023, 7:36 a.m. UTC
## About DMA APIs

Now, virtio may can not work with DMA APIs when virtio features do not have
VIRTIO_F_ACCESS_PLATFORM.

1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just
   work with the "real" devices.
2. I tried to let xsk support callballs to get phy address from virtio-net
   driver as the dma address. But the maintainers of xsk may want to use dma-buf
   to replace the DMA APIs. I think that may be a larger effort. We will wait
   too long.

So rethinking this, firstly, we can support premapped-dma only for devices with
VIRTIO_F_ACCESS_PLATFORM. In the case of af-xdp, if the users want to use it,
they have to update the device to support VIRTIO_F_RING_RESET, and they can also
enable the device's VIRTIO_F_ACCESS_PLATFORM feature by the way.

Thanks for the help from Christoph.

=================

XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero
copy feature of xsk (XDP socket) needs to be supported by the driver. The
performance of zero copy is very good.

ENV: Qemu with vhost.

                   vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS
-----------------------------|---------------|------------------|------------
xmit by sockperf:     90%    |   100%        |                  |  318967
xmit by xsk:          100%   |   30%         |   33%            | 1192064
recv by sockperf:     100%   |   68%         |   100%           |  692288
recv by xsk:          100%   |   33%         |   43%            |  771670

Before achieving the function of Virtio-Net, we also have to let virtio core
support these features:

1. virtio core support premapped
2. virtio core support reset per-queue
3. introduce DMA APIs to virtio core

Please review.

Thanks.

v7:
 1. virtqueue_dma_dev() return NULL when virtio is without DMA API.

v6:
 1. change the size of the flags to u32.

v5:
 1. fix for error handler
 2. add flags to record internal dma mapping

v4:
 1. rename map_inter to dma_map_internal
 2. fix: Excess function parameter 'vq' description in 'virtqueue_dma_dev'

v3:
 1. add map_inter to struct desc state to reocrd whether virtio core do dma map

v2:
 1. based on sgs[0]->dma_address to judgment is premapped
 2. based on extra.addr to judgment to do unmap for no-indirect desc
 3. based on indir_desc to judgment to do unmap for indirect desc
 4. rename virtqueue_get_dma_dev to virtqueue_dma_dev

v1:
 1. expose dma device. NO introduce the api for dma and sync
 2. split some commit for review.

Xuan Zhuo (11):
  virtio_ring: split: separate dma codes
  virtio_ring: packed: separate dma codes
  virtio_ring: packed-indirect: separate dma codes
  virtio_ring: split: support premapped
  virtio_ring: packed: support premapped
  virtio_ring: packed-indirect: support premapped
  virtio_ring: update document for virtqueue_add_*
  virtio_ring: introduce virtqueue_dma_dev()
  virtio_ring: correct the expression of the description of
    virtqueue_resize()
  virtio_ring: separate the logic of reset/enable from virtqueue_resize
  virtio_ring: introduce virtqueue_reset()

 drivers/virtio/virtio_ring.c | 352 +++++++++++++++++++++++++----------
 include/linux/virtio.h       |   4 +
 2 files changed, 259 insertions(+), 97 deletions(-)

--
2.32.0.3.g01195cf9f

Comments

Michael S. Tsirkin April 25, 2023, 7:51 a.m. UTC | #1
On Tue, Apr 25, 2023 at 03:36:02PM +0800, Xuan Zhuo wrote:
> ## About DMA APIs
> 
> Now, virtio may can not work with DMA APIs when virtio features do not have
> VIRTIO_F_ACCESS_PLATFORM.
> 
> 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just
>    work with the "real" devices.
> 2. I tried to let xsk support callballs to get phy address from virtio-net
>    driver as the dma address. But the maintainers of xsk may want to use dma-buf
>    to replace the DMA APIs. I think that may be a larger effort. We will wait
>    too long.
> 
> So rethinking this, firstly, we can support premapped-dma only for devices with
> VIRTIO_F_ACCESS_PLATFORM. In the case of af-xdp, if the users want to use it,
> they have to update the device to support VIRTIO_F_RING_RESET, and they can also
> enable the device's VIRTIO_F_ACCESS_PLATFORM feature by the way.

I don't understand this last sentence. If you think ring
reset can change device features then the answer is no, it can't.

If you are saying device has to set VIRTIO_F_ACCESS_PLATFORM to
benefit from this work, that's fine at least as a first approach.
Note that setting VIRTIO_F_ACCESS_PLATFORM breaks old guests
(it's a secirity boundary), e.g. it is not available for
transitional devices.
So to support transitional devices, we might want to find another way to
address this down the road, but as a first step, I agree just going with
DMA is fine.


> Thanks for the help from Christoph.
> 
> =================
> 
> XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero
> copy feature of xsk (XDP socket) needs to be supported by the driver. The
> performance of zero copy is very good.
> 
> ENV: Qemu with vhost.
> 
>                    vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS
> -----------------------------|---------------|------------------|------------
> xmit by sockperf:     90%    |   100%        |                  |  318967
> xmit by xsk:          100%   |   30%         |   33%            | 1192064
> recv by sockperf:     100%   |   68%         |   100%           |  692288
> recv by xsk:          100%   |   33%         |   43%            |  771670
> 
> Before achieving the function of Virtio-Net, we also have to let virtio core
> support these features:
> 
> 1. virtio core support premapped
> 2. virtio core support reset per-queue
> 3. introduce DMA APIs to virtio core
> 
> Please review.
> 
> Thanks.
> 
> v7:
>  1. virtqueue_dma_dev() return NULL when virtio is without DMA API.
> 
> v6:
>  1. change the size of the flags to u32.
> 
> v5:
>  1. fix for error handler
>  2. add flags to record internal dma mapping
> 
> v4:
>  1. rename map_inter to dma_map_internal
>  2. fix: Excess function parameter 'vq' description in 'virtqueue_dma_dev'
> 
> v3:
>  1. add map_inter to struct desc state to reocrd whether virtio core do dma map
> 
> v2:
>  1. based on sgs[0]->dma_address to judgment is premapped
>  2. based on extra.addr to judgment to do unmap for no-indirect desc
>  3. based on indir_desc to judgment to do unmap for indirect desc
>  4. rename virtqueue_get_dma_dev to virtqueue_dma_dev
> 
> v1:
>  1. expose dma device. NO introduce the api for dma and sync
>  2. split some commit for review.
> 
> Xuan Zhuo (11):
>   virtio_ring: split: separate dma codes
>   virtio_ring: packed: separate dma codes
>   virtio_ring: packed-indirect: separate dma codes
>   virtio_ring: split: support premapped
>   virtio_ring: packed: support premapped
>   virtio_ring: packed-indirect: support premapped
>   virtio_ring: update document for virtqueue_add_*
>   virtio_ring: introduce virtqueue_dma_dev()
>   virtio_ring: correct the expression of the description of
>     virtqueue_resize()
>   virtio_ring: separate the logic of reset/enable from virtqueue_resize
>   virtio_ring: introduce virtqueue_reset()
> 
>  drivers/virtio/virtio_ring.c | 352 +++++++++++++++++++++++++----------
>  include/linux/virtio.h       |   4 +
>  2 files changed, 259 insertions(+), 97 deletions(-)
> 
> --
> 2.32.0.3.g01195cf9f
Xuan Zhuo April 25, 2023, 8:05 a.m. UTC | #2
On Tue, 25 Apr 2023 03:51:47 -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> On Tue, Apr 25, 2023 at 03:36:02PM +0800, Xuan Zhuo wrote:
> > ## About DMA APIs
> >
> > Now, virtio may can not work with DMA APIs when virtio features do not have
> > VIRTIO_F_ACCESS_PLATFORM.
> >
> > 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just
> >    work with the "real" devices.
> > 2. I tried to let xsk support callballs to get phy address from virtio-net
> >    driver as the dma address. But the maintainers of xsk may want to use dma-buf
> >    to replace the DMA APIs. I think that may be a larger effort. We will wait
> >    too long.
> >
> > So rethinking this, firstly, we can support premapped-dma only for devices with
> > VIRTIO_F_ACCESS_PLATFORM. In the case of af-xdp, if the users want to use it,
> > they have to update the device to support VIRTIO_F_RING_RESET, and they can also
> > enable the device's VIRTIO_F_ACCESS_PLATFORM feature by the way.
>
> I don't understand this last sentence. If you think ring
> reset can change device features then the answer is no, it can't.


Sorry, I should remove "by the way".


>
> If you are saying device has to set VIRTIO_F_ACCESS_PLATFORM to
> benefit from this work, that's fine at least as a first approach.
> Note that setting VIRTIO_F_ACCESS_PLATFORM breaks old guests
> (it's a secirity boundary), e.g. it is not available for
> transitional devices.
> So to support transitional devices, we might want to find another way to
> address this down the road,

Maybe dma-buf is a way. I'll look into it, especially some practice on xsk.

> but as a first step, I agree just going with
> DMA is fine.


Thanks.


>
>
> > Thanks for the help from Christoph.
> >
> > =================
> >
> > XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero
> > copy feature of xsk (XDP socket) needs to be supported by the driver. The
> > performance of zero copy is very good.
> >
> > ENV: Qemu with vhost.
> >
> >                    vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS
> > -----------------------------|---------------|------------------|------------
> > xmit by sockperf:     90%    |   100%        |                  |  318967
> > xmit by xsk:          100%   |   30%         |   33%            | 1192064
> > recv by sockperf:     100%   |   68%         |   100%           |  692288
> > recv by xsk:          100%   |   33%         |   43%            |  771670
> >
> > Before achieving the function of Virtio-Net, we also have to let virtio core
> > support these features:
> >
> > 1. virtio core support premapped
> > 2. virtio core support reset per-queue
> > 3. introduce DMA APIs to virtio core
> >
> > Please review.
> >
> > Thanks.
> >
> > v7:
> >  1. virtqueue_dma_dev() return NULL when virtio is without DMA API.
> >
> > v6:
> >  1. change the size of the flags to u32.
> >
> > v5:
> >  1. fix for error handler
> >  2. add flags to record internal dma mapping
> >
> > v4:
> >  1. rename map_inter to dma_map_internal
> >  2. fix: Excess function parameter 'vq' description in 'virtqueue_dma_dev'
> >
> > v3:
> >  1. add map_inter to struct desc state to reocrd whether virtio core do dma map
> >
> > v2:
> >  1. based on sgs[0]->dma_address to judgment is premapped
> >  2. based on extra.addr to judgment to do unmap for no-indirect desc
> >  3. based on indir_desc to judgment to do unmap for indirect desc
> >  4. rename virtqueue_get_dma_dev to virtqueue_dma_dev
> >
> > v1:
> >  1. expose dma device. NO introduce the api for dma and sync
> >  2. split some commit for review.
> >
> > Xuan Zhuo (11):
> >   virtio_ring: split: separate dma codes
> >   virtio_ring: packed: separate dma codes
> >   virtio_ring: packed-indirect: separate dma codes
> >   virtio_ring: split: support premapped
> >   virtio_ring: packed: support premapped
> >   virtio_ring: packed-indirect: support premapped
> >   virtio_ring: update document for virtqueue_add_*
> >   virtio_ring: introduce virtqueue_dma_dev()
> >   virtio_ring: correct the expression of the description of
> >     virtqueue_resize()
> >   virtio_ring: separate the logic of reset/enable from virtqueue_resize
> >   virtio_ring: introduce virtqueue_reset()
> >
> >  drivers/virtio/virtio_ring.c | 352 +++++++++++++++++++++++++----------
> >  include/linux/virtio.h       |   4 +
> >  2 files changed, 259 insertions(+), 97 deletions(-)
> >
> > --
> > 2.32.0.3.g01195cf9f
>
Michael S. Tsirkin April 25, 2023, 8:13 a.m. UTC | #3
On Tue, Apr 25, 2023 at 04:05:03PM +0800, Xuan Zhuo wrote:
> On Tue, 25 Apr 2023 03:51:47 -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > On Tue, Apr 25, 2023 at 03:36:02PM +0800, Xuan Zhuo wrote:
> > > ## About DMA APIs
> > >
> > > Now, virtio may can not work with DMA APIs when virtio features do not have
> > > VIRTIO_F_ACCESS_PLATFORM.
> > >
> > > 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just
> > >    work with the "real" devices.
> > > 2. I tried to let xsk support callballs to get phy address from virtio-net
> > >    driver as the dma address. But the maintainers of xsk may want to use dma-buf
> > >    to replace the DMA APIs. I think that may be a larger effort. We will wait
> > >    too long.
> > >
> > > So rethinking this, firstly, we can support premapped-dma only for devices with
> > > VIRTIO_F_ACCESS_PLATFORM. In the case of af-xdp, if the users want to use it,
> > > they have to update the device to support VIRTIO_F_RING_RESET, and they can also
> > > enable the device's VIRTIO_F_ACCESS_PLATFORM feature by the way.
> >
> > I don't understand this last sentence. If you think ring
> > reset can change device features then the answer is no, it can't.
> 
> 
> Sorry, I should remove "by the way".
> 
> 
> >
> > If you are saying device has to set VIRTIO_F_ACCESS_PLATFORM to
> > benefit from this work, that's fine at least as a first approach.
> > Note that setting VIRTIO_F_ACCESS_PLATFORM breaks old guests
> > (it's a secirity boundary), e.g. it is not available for
> > transitional devices.
> > So to support transitional devices, we might want to find another way to
> > address this down the road,
> 
> Maybe dma-buf is a way. I'll look into it, especially some practice on xsk.
> 
> > but as a first step, I agree just going with
> > DMA is fine.
> 
> 
> Thanks.

Pls do make sure to disable the feature when !VIRTIO_F_ACCESS_PLATFORM
though.

> 
> >
> >
> > > Thanks for the help from Christoph.
> > >
> > > =================
> > >
> > > XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero
> > > copy feature of xsk (XDP socket) needs to be supported by the driver. The
> > > performance of zero copy is very good.
> > >
> > > ENV: Qemu with vhost.
> > >
> > >                    vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS
> > > -----------------------------|---------------|------------------|------------
> > > xmit by sockperf:     90%    |   100%        |                  |  318967
> > > xmit by xsk:          100%   |   30%         |   33%            | 1192064
> > > recv by sockperf:     100%   |   68%         |   100%           |  692288
> > > recv by xsk:          100%   |   33%         |   43%            |  771670
> > >
> > > Before achieving the function of Virtio-Net, we also have to let virtio core
> > > support these features:
> > >
> > > 1. virtio core support premapped
> > > 2. virtio core support reset per-queue
> > > 3. introduce DMA APIs to virtio core
> > >
> > > Please review.
> > >
> > > Thanks.
> > >
> > > v7:
> > >  1. virtqueue_dma_dev() return NULL when virtio is without DMA API.
> > >
> > > v6:
> > >  1. change the size of the flags to u32.
> > >
> > > v5:
> > >  1. fix for error handler
> > >  2. add flags to record internal dma mapping
> > >
> > > v4:
> > >  1. rename map_inter to dma_map_internal
> > >  2. fix: Excess function parameter 'vq' description in 'virtqueue_dma_dev'
> > >
> > > v3:
> > >  1. add map_inter to struct desc state to reocrd whether virtio core do dma map
> > >
> > > v2:
> > >  1. based on sgs[0]->dma_address to judgment is premapped
> > >  2. based on extra.addr to judgment to do unmap for no-indirect desc
> > >  3. based on indir_desc to judgment to do unmap for indirect desc
> > >  4. rename virtqueue_get_dma_dev to virtqueue_dma_dev
> > >
> > > v1:
> > >  1. expose dma device. NO introduce the api for dma and sync
> > >  2. split some commit for review.
> > >
> > > Xuan Zhuo (11):
> > >   virtio_ring: split: separate dma codes
> > >   virtio_ring: packed: separate dma codes
> > >   virtio_ring: packed-indirect: separate dma codes
> > >   virtio_ring: split: support premapped
> > >   virtio_ring: packed: support premapped
> > >   virtio_ring: packed-indirect: support premapped
> > >   virtio_ring: update document for virtqueue_add_*
> > >   virtio_ring: introduce virtqueue_dma_dev()
> > >   virtio_ring: correct the expression of the description of
> > >     virtqueue_resize()
> > >   virtio_ring: separate the logic of reset/enable from virtqueue_resize
> > >   virtio_ring: introduce virtqueue_reset()
> > >
> > >  drivers/virtio/virtio_ring.c | 352 +++++++++++++++++++++++++----------
> > >  include/linux/virtio.h       |   4 +
> > >  2 files changed, 259 insertions(+), 97 deletions(-)
> > >
> > > --
> > > 2.32.0.3.g01195cf9f
> >
Xuan Zhuo April 25, 2023, 8:21 a.m. UTC | #4
On Tue, 25 Apr 2023 04:13:09 -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> On Tue, Apr 25, 2023 at 04:05:03PM +0800, Xuan Zhuo wrote:
> > On Tue, 25 Apr 2023 03:51:47 -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > > On Tue, Apr 25, 2023 at 03:36:02PM +0800, Xuan Zhuo wrote:
> > > > ## About DMA APIs
> > > >
> > > > Now, virtio may can not work with DMA APIs when virtio features do not have
> > > > VIRTIO_F_ACCESS_PLATFORM.
> > > >
> > > > 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just
> > > >    work with the "real" devices.
> > > > 2. I tried to let xsk support callballs to get phy address from virtio-net
> > > >    driver as the dma address. But the maintainers of xsk may want to use dma-buf
> > > >    to replace the DMA APIs. I think that may be a larger effort. We will wait
> > > >    too long.
> > > >
> > > > So rethinking this, firstly, we can support premapped-dma only for devices with
> > > > VIRTIO_F_ACCESS_PLATFORM. In the case of af-xdp, if the users want to use it,
> > > > they have to update the device to support VIRTIO_F_RING_RESET, and they can also
> > > > enable the device's VIRTIO_F_ACCESS_PLATFORM feature by the way.
> > >
> > > I don't understand this last sentence. If you think ring
> > > reset can change device features then the answer is no, it can't.
> >
> >
> > Sorry, I should remove "by the way".
> >
> >
> > >
> > > If you are saying device has to set VIRTIO_F_ACCESS_PLATFORM to
> > > benefit from this work, that's fine at least as a first approach.
> > > Note that setting VIRTIO_F_ACCESS_PLATFORM breaks old guests
> > > (it's a secirity boundary), e.g. it is not available for
> > > transitional devices.
> > > So to support transitional devices, we might want to find another way to
> > > address this down the road,
> >
> > Maybe dma-buf is a way. I'll look into it, especially some practice on xsk.
> >
> > > but as a first step, I agree just going with
> > > DMA is fine.
> >
> >
> > Thanks.
>
> Pls do make sure to disable the feature when !VIRTIO_F_ACCESS_PLATFORM
> though.

If you refer to the implementation inside virtio-net, this feature will depend
on the return of virtqueue_dma_dev().

But virtqueue_dma_dev() depends "use_dma_api". When xen_domain() is true and
!VIRTIO_F_ACCESS_PLATFORM, the "use_dma_api" is true.

So what kind of situation do you mean?

Thanks.

>
> >
> > >
> > >
> > > > Thanks for the help from Christoph.
> > > >
> > > > =================
> > > >
> > > > XDP socket(AF_XDP) is an excellent bypass kernel network framework. The zero
> > > > copy feature of xsk (XDP socket) needs to be supported by the driver. The
> > > > performance of zero copy is very good.
> > > >
> > > > ENV: Qemu with vhost.
> > > >
> > > >                    vhost cpu | Guest APP CPU |Guest Softirq CPU | PPS
> > > > -----------------------------|---------------|------------------|------------
> > > > xmit by sockperf:     90%    |   100%        |                  |  318967
> > > > xmit by xsk:          100%   |   30%         |   33%            | 1192064
> > > > recv by sockperf:     100%   |   68%         |   100%           |  692288
> > > > recv by xsk:          100%   |   33%         |   43%            |  771670
> > > >
> > > > Before achieving the function of Virtio-Net, we also have to let virtio core
> > > > support these features:
> > > >
> > > > 1. virtio core support premapped
> > > > 2. virtio core support reset per-queue
> > > > 3. introduce DMA APIs to virtio core
> > > >
> > > > Please review.
> > > >
> > > > Thanks.
> > > >
> > > > v7:
> > > >  1. virtqueue_dma_dev() return NULL when virtio is without DMA API.
> > > >
> > > > v6:
> > > >  1. change the size of the flags to u32.
> > > >
> > > > v5:
> > > >  1. fix for error handler
> > > >  2. add flags to record internal dma mapping
> > > >
> > > > v4:
> > > >  1. rename map_inter to dma_map_internal
> > > >  2. fix: Excess function parameter 'vq' description in 'virtqueue_dma_dev'
> > > >
> > > > v3:
> > > >  1. add map_inter to struct desc state to reocrd whether virtio core do dma map
> > > >
> > > > v2:
> > > >  1. based on sgs[0]->dma_address to judgment is premapped
> > > >  2. based on extra.addr to judgment to do unmap for no-indirect desc
> > > >  3. based on indir_desc to judgment to do unmap for indirect desc
> > > >  4. rename virtqueue_get_dma_dev to virtqueue_dma_dev
> > > >
> > > > v1:
> > > >  1. expose dma device. NO introduce the api for dma and sync
> > > >  2. split some commit for review.
> > > >
> > > > Xuan Zhuo (11):
> > > >   virtio_ring: split: separate dma codes
> > > >   virtio_ring: packed: separate dma codes
> > > >   virtio_ring: packed-indirect: separate dma codes
> > > >   virtio_ring: split: support premapped
> > > >   virtio_ring: packed: support premapped
> > > >   virtio_ring: packed-indirect: support premapped
> > > >   virtio_ring: update document for virtqueue_add_*
> > > >   virtio_ring: introduce virtqueue_dma_dev()
> > > >   virtio_ring: correct the expression of the description of
> > > >     virtqueue_resize()
> > > >   virtio_ring: separate the logic of reset/enable from virtqueue_resize
> > > >   virtio_ring: introduce virtqueue_reset()
> > > >
> > > >  drivers/virtio/virtio_ring.c | 352 +++++++++++++++++++++++++----------
> > > >  include/linux/virtio.h       |   4 +
> > > >  2 files changed, 259 insertions(+), 97 deletions(-)
> > > >
> > > > --
> > > > 2.32.0.3.g01195cf9f
> > >
>
Michael S. Tsirkin April 25, 2023, 12:40 p.m. UTC | #5
On Tue, Apr 25, 2023 at 04:21:53PM +0800, Xuan Zhuo wrote:
> On Tue, 25 Apr 2023 04:13:09 -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > On Tue, Apr 25, 2023 at 04:05:03PM +0800, Xuan Zhuo wrote:
> > > On Tue, 25 Apr 2023 03:51:47 -0400, "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > > > On Tue, Apr 25, 2023 at 03:36:02PM +0800, Xuan Zhuo wrote:
> > > > > ## About DMA APIs
> > > > >
> > > > > Now, virtio may can not work with DMA APIs when virtio features do not have
> > > > > VIRTIO_F_ACCESS_PLATFORM.
> > > > >
> > > > > 1. I tried to let DMA APIs return phy address by virtio-device. But DMA APIs just
> > > > >    work with the "real" devices.
> > > > > 2. I tried to let xsk support callballs to get phy address from virtio-net
> > > > >    driver as the dma address. But the maintainers of xsk may want to use dma-buf
> > > > >    to replace the DMA APIs. I think that may be a larger effort. We will wait
> > > > >    too long.
> > > > >
> > > > > So rethinking this, firstly, we can support premapped-dma only for devices with
> > > > > VIRTIO_F_ACCESS_PLATFORM. In the case of af-xdp, if the users want to use it,
> > > > > they have to update the device to support VIRTIO_F_RING_RESET, and they can also
> > > > > enable the device's VIRTIO_F_ACCESS_PLATFORM feature by the way.
> > > >
> > > > I don't understand this last sentence. If you think ring
> > > > reset can change device features then the answer is no, it can't.
> > >
> > >
> > > Sorry, I should remove "by the way".
> > >
> > >
> > > >
> > > > If you are saying device has to set VIRTIO_F_ACCESS_PLATFORM to
> > > > benefit from this work, that's fine at least as a first approach.
> > > > Note that setting VIRTIO_F_ACCESS_PLATFORM breaks old guests
> > > > (it's a secirity boundary), e.g. it is not available for
> > > > transitional devices.
> > > > So to support transitional devices, we might want to find another way to
> > > > address this down the road,
> > >
> > > Maybe dma-buf is a way. I'll look into it, especially some practice on xsk.
> > >
> > > > but as a first step, I agree just going with
> > > > DMA is fine.
> > >
> > >
> > > Thanks.
> >
> > Pls do make sure to disable the feature when !VIRTIO_F_ACCESS_PLATFORM
> > though.
> 
> If you refer to the implementation inside virtio-net, this feature will depend
> on the return of virtqueue_dma_dev().
> 
> But virtqueue_dma_dev() depends "use_dma_api". When xen_domain() is true and
> !VIRTIO_F_ACCESS_PLATFORM, the "use_dma_api" is true.
> 
> So what kind of situation do you mean?
> 
> Thanks.

E.g. a legacy device.