| Message ID | 20211223110755.22722-4-zajec5@gmail.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | nvmem: support more NVMEM cells variants | expand |
| Context | Check | Description |
|---|---|---|
| netdev/tree_selection | success | Not a local patch |
On Thu, Dec 23, 2021 at 7:08 AM Rafał Miłecki <zajec5@gmail.com> wrote: > > From: Rafał Miłecki <rafal@milecki.pl> > > Not every NVMEM has predefined cells at hardcoded addresses. Some > devices store cells in internal structs and custom formats. Referencing > such cells is still required to let other bindings use them. > > Modify binding to require "reg" xor "label". The later one can be used > to match "dynamic" NVMEM cells by their names. 'label' is supposed to correspond to a sticker on a port or something human identifiable. It generally should be something optional to making the OS functional. Yes, there are already some abuses of that, but this case is too far for me. Rob
On 23.12.2021 22:18, Rob Herring wrote: > On Thu, Dec 23, 2021 at 7:08 AM Rafał Miłecki <zajec5@gmail.com> wrote: >> >> From: Rafał Miłecki <rafal@milecki.pl> >> >> Not every NVMEM has predefined cells at hardcoded addresses. Some >> devices store cells in internal structs and custom formats. Referencing >> such cells is still required to let other bindings use them. >> >> Modify binding to require "reg" xor "label". The later one can be used >> to match "dynamic" NVMEM cells by their names. > > 'label' is supposed to correspond to a sticker on a port or something > human identifiable. It generally should be something optional to > making the OS functional. Yes, there are already some abuses of that, > but this case is too far for me. Good to learn that! "name" is special & not allowed I think. Any suggestion what to use? I'm not native. What about "title"? Or maybe "term", "entity", "tag"?
On Thu, Dec 23, 2021 at 10:58:56PM +0100, Rafał Miłecki wrote: > On 23.12.2021 22:18, Rob Herring wrote: > > On Thu, Dec 23, 2021 at 7:08 AM Rafał Miłecki <zajec5@gmail.com> wrote: > > > > > > From: Rafał Miłecki <rafal@milecki.pl> > > > > > > Not every NVMEM has predefined cells at hardcoded addresses. Some > > > devices store cells in internal structs and custom formats. Referencing > > > such cells is still required to let other bindings use them. > > > > > > Modify binding to require "reg" xor "label". The later one can be used > > > to match "dynamic" NVMEM cells by their names. > > > > 'label' is supposed to correspond to a sticker on a port or something > > human identifiable. It generally should be something optional to > > making the OS functional. Yes, there are already some abuses of that, > > but this case is too far for me. > > Good to learn that! > > "name" is special & not allowed I think. It's the node name essentially. Why is using node names not sufficient? Do you have some specific examples? Rob
On 4.01.2022 21:16, Rob Herring wrote: > On Thu, Dec 23, 2021 at 10:58:56PM +0100, Rafał Miłecki wrote: >> On 23.12.2021 22:18, Rob Herring wrote: >>> On Thu, Dec 23, 2021 at 7:08 AM Rafał Miłecki <zajec5@gmail.com> wrote: >>>> >>>> From: Rafał Miłecki <rafal@milecki.pl> >>>> >>>> Not every NVMEM has predefined cells at hardcoded addresses. Some >>>> devices store cells in internal structs and custom formats. Referencing >>>> such cells is still required to let other bindings use them. >>>> >>>> Modify binding to require "reg" xor "label". The later one can be used >>>> to match "dynamic" NVMEM cells by their names. >>> >>> 'label' is supposed to correspond to a sticker on a port or something >>> human identifiable. It generally should be something optional to >>> making the OS functional. Yes, there are already some abuses of that, >>> but this case is too far for me. >> >> Good to learn that! >> >> "name" is special & not allowed I think. > > It's the node name essentially. Why is using node names not sufficient? > Do you have some specific examples? I tried to explain in [PATCH 1/5] dt-bindings: nvmem: add "label" property to allow more flexible cells names that some vendors come with fancy names that can't fit node names. Broadcom's NVRAM examples: 0:macaddr 1:macaddr 2:macaddr 0:ccode 1:ccode 2:ccode 0:regrev
On 4.01.2022 21:50, Rafał Miłecki wrote: > On 4.01.2022 21:16, Rob Herring wrote: >> On Thu, Dec 23, 2021 at 10:58:56PM +0100, Rafał Miłecki wrote: >>> On 23.12.2021 22:18, Rob Herring wrote: >>>> On Thu, Dec 23, 2021 at 7:08 AM Rafał Miłecki <zajec5@gmail.com> wrote: >>>>> >>>>> From: Rafał Miłecki <rafal@milecki.pl> >>>>> >>>>> Not every NVMEM has predefined cells at hardcoded addresses. Some >>>>> devices store cells in internal structs and custom formats. Referencing >>>>> such cells is still required to let other bindings use them. >>>>> >>>>> Modify binding to require "reg" xor "label". The later one can be used >>>>> to match "dynamic" NVMEM cells by their names. >>>> >>>> 'label' is supposed to correspond to a sticker on a port or something >>>> human identifiable. It generally should be something optional to >>>> making the OS functional. Yes, there are already some abuses of that, >>>> but this case is too far for me. >>> >>> Good to learn that! >>> >>> "name" is special & not allowed I think. >> >> It's the node name essentially. Why is using node names not sufficient? >> Do you have some specific examples? > > I tried to explain in > [PATCH 1/5] dt-bindings: nvmem: add "label" property to allow more flexible cells names > that some vendors come with fancy names that can't fit node names. > > Broadcom's NVRAM examples: > 0:macaddr > 1:macaddr > 2:macaddr > 0:ccode > 1:ccode > 2:ccode > 0:regrev In other words I'd like to have something like: nvram@1eff0000 { compatible = "brcm,nvram"; reg = <0x1eff0000 0x10000>; mac: cell-0 { label = "1:macaddr"; }; }; ethernet@1000 { compatible = "brcm,ethernet"; reg = <0x1000 0x1000>; nvmem-cells = <&mac>; nvmem-cell-names = "mac-address"; };
On Tue, Jan 04, 2022 at 09:56:01PM +0100, Rafał Miłecki wrote: > On 4.01.2022 21:50, Rafał Miłecki wrote: > > On 4.01.2022 21:16, Rob Herring wrote: > > > On Thu, Dec 23, 2021 at 10:58:56PM +0100, Rafał Miłecki wrote: > > > > On 23.12.2021 22:18, Rob Herring wrote: > > > > > On Thu, Dec 23, 2021 at 7:08 AM Rafał Miłecki <zajec5@gmail.com> wrote: > > > > > > > > > > > > From: Rafał Miłecki <rafal@milecki.pl> > > > > > > > > > > > > Not every NVMEM has predefined cells at hardcoded addresses. Some > > > > > > devices store cells in internal structs and custom formats. Referencing > > > > > > such cells is still required to let other bindings use them. > > > > > > > > > > > > Modify binding to require "reg" xor "label". The later one can be used > > > > > > to match "dynamic" NVMEM cells by their names. > > > > > > > > > > 'label' is supposed to correspond to a sticker on a port or something > > > > > human identifiable. It generally should be something optional to > > > > > making the OS functional. Yes, there are already some abuses of that, > > > > > but this case is too far for me. > > > > > > > > Good to learn that! > > > > > > > > "name" is special & not allowed I think. > > > > > > It's the node name essentially. Why is using node names not sufficient? > > > Do you have some specific examples? > > > > I tried to explain in > > [PATCH 1/5] dt-bindings: nvmem: add "label" property to allow more flexible cells names > > that some vendors come with fancy names that can't fit node names. I still don't see the issue. Why do you need 'more flexible cells names'? What problem does that solve? > > Broadcom's NVRAM examples: > > 0:macaddr > > 1:macaddr > > 2:macaddr > > 0:ccode > > 1:ccode > > 2:ccode > > 0:regrev > > In other words I'd like to have something like: > > nvram@1eff0000 { > compatible = "brcm,nvram"; > reg = <0x1eff0000 0x10000>; > > mac: cell-0 { > label = "1:macaddr"; > }; > }; > > ethernet@1000 { > compatible = "brcm,ethernet"; > reg = <0x1000 0x1000>; > nvmem-cells = <&mac>; > nvmem-cell-names = "mac-address"; > }; How does 'label' help here? Note there's some other efforts around multiple mac addresses and how to interpret the nvmem data. Maybe that helps solve your problem. Rob
diff --git a/Documentation/devicetree/bindings/nvmem/nvmem.yaml b/Documentation/devicetree/bindings/nvmem/nvmem.yaml index 3392405ee010..83154df25c27 100644 --- a/Documentation/devicetree/bindings/nvmem/nvmem.yaml +++ b/Documentation/devicetree/bindings/nvmem/nvmem.yaml @@ -43,6 +43,12 @@ patternProperties: "@[0-9a-f]+(,[0-7])?$": type: object + description: | + NVMEM cell - a part of NVMEM containing one specific information. + + Cells can be fully defined by a binding or stored in NVMEM device specific + data and just referenced in DT by a name (label). + properties: reg: maxItems: 1 @@ -64,8 +70,11 @@ patternProperties: description: Size in bit within the address range specified by reg. - required: - - reg + oneOf: + - required: + - reg + - required: + - label additionalProperties: true