[PATCH] Add 'dma-noncoherent' property as a standard property

[PATCH] Add 'dma-noncoherent' property as a standard property

[Heiko Stuebner]

Similar to the 'dma-coherent' property, some architectures may be coherent
by default with some devices being non-coherent. Add a property for this
case tool.

Signed-off-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
---
Companion patch to RISC-V Zicbom noncoherent dma support
https://lore.kernel.org/linux-riscv/20220619203212.3604485-1-heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org/

 source/chapter2-devicetree-basics.rst | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/source/chapter2-devicetree-basics.rst b/source/chapter2-devicetree-basics.rst
index 47d2fad..bdd7b9f 100644
--- a/source/chapter2-devicetree-basics.rst
+++ b/source/chapter2-devicetree-basics.rst
@@ -832,6 +832,19 @@ Description:
    coherent DMA operations. Some architectures have coherent DMA by default
    and this property is not applicable.
 
+dma-noncoherent
+~~~~~~~~~~~~
+
+Property name: ``dma-noncoherent``
+
+Value type: ``<empty>``
+
+Description:
+   For architectures which are by default coherent for I/O, the
+   *dma-noncoherent* property is used to indicate a device is not capable of
+   coherent DMA operations. Some architectures have non-coherent DMA by
+   default and this property is not applicable.
+
 name (deprecated)
 ~~~~~~~~~~~~~~~~~
 
-- 
2.35.1

Re: [PATCH] Add 'dma-noncoherent' property as a standard property

[Rob Herring]

On Sun, Jun 19, 2022 at 2:37 PM Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org> wrote:
>
> Similar to the 'dma-coherent' property, some architectures may be coherent
> by default with some devices being non-coherent. Add a property for this
> case tool.

Looks fine, but I also need a schema.

Yeah, that's redundant unfortunately. Maybe someday someone will care
enough to generate the spec from the schemas.

>
> Signed-off-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
> ---
> Companion patch to RISC-V Zicbom noncoherent dma support
> https://lore.kernel.org/linux-riscv/20220619203212.3604485-1-heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org/
>
>  source/chapter2-devicetree-basics.rst | 13 +++++++++++++
>  1 file changed, 13 insertions(+)
>
> diff --git a/source/chapter2-devicetree-basics.rst b/source/chapter2-devicetree-basics.rst
> index 47d2fad..bdd7b9f 100644
> --- a/source/chapter2-devicetree-basics.rst
> +++ b/source/chapter2-devicetree-basics.rst
> @@ -832,6 +832,19 @@ Description:
>     coherent DMA operations. Some architectures have coherent DMA by default
>     and this property is not applicable.
>
> +dma-noncoherent
> +~~~~~~~~~~~~
> +
> +Property name: ``dma-noncoherent``
> +
> +Value type: ``<empty>``
> +
> +Description:
> +   For architectures which are by default coherent for I/O, the
> +   *dma-noncoherent* property is used to indicate a device is not capable of
> +   coherent DMA operations. Some architectures have non-coherent DMA by
> +   default and this property is not applicable.
> +
>  name (deprecated)
>  ~~~~~~~~~~~~~~~~~
>
> --
> 2.35.1
>

Re: [PATCH] Add 'dma-noncoherent' property as a standard property

[Heiko Stübner]

Am Montag, 27. Juni 2022, 17:14:45 CEST schrieb Rob Herring:
> On Sun, Jun 19, 2022 at 2:37 PM Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org> wrote:
> >
> > Similar to the 'dma-coherent' property, some architectures may be coherent
> > by default with some devices being non-coherent. Add a property for this
> > case tool.
> 
> Looks fine, but I also need a schema.
> 
> Yeah, that's redundant unfortunately. Maybe someday someone will care
> enough to generate the spec from the schemas.

No worries and sorry I didn't think of where the schema comes from
in the first place. I'll do the relevant schema work as well.

Heiko


> > Signed-off-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
> > ---
> > Companion patch to RISC-V Zicbom noncoherent dma support
> > https://lore.kernel.org/linux-riscv/20220619203212.3604485-1-heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org/
> >
> >  source/chapter2-devicetree-basics.rst | 13 +++++++++++++
> >  1 file changed, 13 insertions(+)
> >
> > diff --git a/source/chapter2-devicetree-basics.rst b/source/chapter2-devicetree-basics.rst
> > index 47d2fad..bdd7b9f 100644
> > --- a/source/chapter2-devicetree-basics.rst
> > +++ b/source/chapter2-devicetree-basics.rst
> > @@ -832,6 +832,19 @@ Description:
> >     coherent DMA operations. Some architectures have coherent DMA by default
> >     and this property is not applicable.
> >
> > +dma-noncoherent
> > +~~~~~~~~~~~~
> > +
> > +Property name: ``dma-noncoherent``
> > +
> > +Value type: ``<empty>``
> > +
> > +Description:
> > +   For architectures which are by default coherent for I/O, the
> > +   *dma-noncoherent* property is used to indicate a device is not capable of
> > +   coherent DMA operations. Some architectures have non-coherent DMA by
> > +   default and this property is not applicable.
> > +
> >  name (deprecated)
> >  ~~~~~~~~~~~~~~~~~
> >
> > --
> > 2.35.1
> >
> 

Re: [PATCH] Add 'dma-noncoherent' property as a standard property

[Heiko Stübner]

Hi Rob,

Am Montag, 27. Juni 2022, 17:51:51 CEST schrieb Heiko Stübner:
> Am Montag, 27. Juni 2022, 17:14:45 CEST schrieb Rob Herring:
> > On Sun, Jun 19, 2022 at 2:37 PM Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org> wrote:
> > >
> > > Similar to the 'dma-coherent' property, some architectures may be coherent
> > > by default with some devices being non-coherent. Add a property for this
> > > case tool.
> > 
> > Looks fine, but I also need a schema.
> > 
> > Yeah, that's redundant unfortunately. Maybe someday someone will care
> > enough to generate the spec from the schemas.
> 
> No worries and sorry I didn't think of where the schema comes from
> in the first place. I'll do the relevant schema work as well.

Done in https://github.com/devicetree-org/dt-schema/pull/78 .

I wasn't sure about the "development model" of dt-schema [0] but there
were already github pull-requests pending, so I've added another one
for the new property :-) .

Heiko


[0] The devicetree-specification documentation said mailinglist, though
I also see pull-requests there now that I was looking at github.

> > > Signed-off-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
> > > ---
> > > Companion patch to RISC-V Zicbom noncoherent dma support
> > > https://lore.kernel.org/linux-riscv/20220619203212.3604485-1-heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org/
> > >
> > >  source/chapter2-devicetree-basics.rst | 13 +++++++++++++
> > >  1 file changed, 13 insertions(+)
> > >
> > > diff --git a/source/chapter2-devicetree-basics.rst b/source/chapter2-devicetree-basics.rst
> > > index 47d2fad..bdd7b9f 100644
> > > --- a/source/chapter2-devicetree-basics.rst
> > > +++ b/source/chapter2-devicetree-basics.rst
> > > @@ -832,6 +832,19 @@ Description:
> > >     coherent DMA operations. Some architectures have coherent DMA by default
> > >     and this property is not applicable.
> > >
> > > +dma-noncoherent
> > > +~~~~~~~~~~~~
> > > +
> > > +Property name: ``dma-noncoherent``
> > > +
> > > +Value type: ``<empty>``
> > > +
> > > +Description:
> > > +   For architectures which are by default coherent for I/O, the
> > > +   *dma-noncoherent* property is used to indicate a device is not capable of
> > > +   coherent DMA operations. Some architectures have non-coherent DMA by
> > > +   default and this property is not applicable.
> > > +
> > >  name (deprecated)
> > >  ~~~~~~~~~~~~~~~~~
> > >
> > > --
> > > 2.35.1
> > >
> > 
> 
> 

Re: [PATCH] Add 'dma-noncoherent' property as a standard property

[Heiko Stuebner]

Hi Rob,

Am Dienstag, 28. Juni 2022, 13:22:31 CEST schrieb Heiko Stübner:
> Am Montag, 27. Juni 2022, 17:51:51 CEST schrieb Heiko Stübner:
> > Am Montag, 27. Juni 2022, 17:14:45 CEST schrieb Rob Herring:
> > > On Sun, Jun 19, 2022 at 2:37 PM Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org> wrote:
> > > >
> > > > Similar to the 'dma-coherent' property, some architectures may be coherent
> > > > by default with some devices being non-coherent. Add a property for this
> > > > case tool.
> > > 
> > > Looks fine, but I also need a schema.
> > > 
> > > Yeah, that's redundant unfortunately. Maybe someday someone will care
> > > enough to generate the spec from the schemas.
> > 
> > No worries and sorry I didn't think of where the schema comes from
> > in the first place. I'll do the relevant schema work as well.
> 
> Done in https://github.com/devicetree-org/dt-schema/pull/78 .
> 
> I wasn't sure about the "development model" of dt-schema [0] but there
> were already github pull-requests pending, so I've added another one
> for the new property :-) .

so as the dt-schema patch did go through the github-pull,. I've converted
this patch as well, so there is now a matching pull-request on github for
the spec addition.

Heiko