Sharing dma-bufs using a driver-private interconnect

Sharing dma-bufs using a driver-private interconnect

[Thomas Hellström]

Hi!

I previously discussed this with Simona on IRC but would like to get
some feedback also from a wider audience:

We're planning to share dma-bufs using a fast interconnect in a way
similar to pcie-p2p:

The rough plan is to identify dma-bufs capable of sharing this way by
looking at the address of either the dma-buf ops and / or the
importer_ops to conclude it's a device using the same driver (or
possibly child driver) and then take special action when the dma-
addresses are obtained. Nothing visible outside of the xe driver or its
child driver.

Are there any absolute "DON'T"s or recommendations to keep in mind WRT
to this approach?

Thanks,
Thomas

Re: Sharing dma-bufs using a driver-private interconnect

[Christian König]

Hi Thomas,

On 5/16/25 10:33, Thomas Hellström wrote:
> Hi!
> 
> I previously discussed this with Simona on IRC but would like to get
> some feedback also from a wider audience:
> 
> We're planning to share dma-bufs using a fast interconnect in a way
> similar to pcie-p2p:
> 
> The rough plan is to identify dma-bufs capable of sharing this way by
> looking at the address of either the dma-buf ops and / or the
> importer_ops to conclude it's a device using the same driver (or
> possibly child driver) and then take special action when the dma-
> addresses are obtained. Nothing visible outside of the xe driver or its
> child driver.

As far as I can see that should work and we have recommended doing exactly that to multiple people in the past.
 
> Are there any absolute "DON'T"s or recommendations to keep in mind WRT
> to this approach?

My only requirement is that you write up some documentation (preferable either in drivers/dma-buf/dma-buf.c or Documentation/driver-api/dma-buf.rst) how drivers should do this, what the implications with the rest of the DMA-buf interface is etc... etc...

The goal is to nail down how to do driver private interfaces without breaking interoperability with drivers who don't participate in that.

Regards,
Christian.

> 
> Thanks,
> Thomas
> 
)

Re: [Linaro-mm-sig] Re: Sharing dma-bufs using a driver-private interconnect

[Thomas Hellström]

On Fri, 2025-05-16 at 10:44 +0200, Christian König wrote:
> Hi Thomas,
> 
> On 5/16/25 10:33, Thomas Hellström wrote:
> > Hi!
> > 
> > I previously discussed this with Simona on IRC but would like to
> > get
> > some feedback also from a wider audience:
> > 
> > We're planning to share dma-bufs using a fast interconnect in a way
> > similar to pcie-p2p:
> > 
> > The rough plan is to identify dma-bufs capable of sharing this way
> > by
> > looking at the address of either the dma-buf ops and / or the
> > importer_ops to conclude it's a device using the same driver (or
> > possibly child driver) and then take special action when the dma-
> > addresses are obtained. Nothing visible outside of the xe driver or
> > its
> > child driver.
> 
> As far as I can see that should work and we have recommended doing
> exactly that to multiple people in the past.
>  
> > Are there any absolute "DON'T"s or recommendations to keep in mind
> > WRT
> > to this approach?
> 
> My only requirement is that you write up some documentation
> (preferable either in drivers/dma-buf/dma-buf.c or
> Documentation/driver-api/dma-buf.rst) how drivers should do this,
> what the implications with the rest of the DMA-buf interface is
> etc... etc...
> 
> The goal is to nail down how to do driver private interfaces without
> breaking interoperability with drivers who don't participate in that.

Ok, Will ensure that we adhere to this.

Thanks,
Thomas


> 
> Regards,
> Christian.
> 
> > 
> > Thanks,
> > Thomas
> > 
> )
> _______________________________________________
> Linaro-mm-sig mailing list -- linaro-mm-sig@lists.linaro.org
> To unsubscribe send an email to linaro-mm-sig-leave@lists.linaro.org