lkml.org 
[lkml]   [2021]   [Dec]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [RFC PATCH v2 4/5] dt-bindings: interrupt-controller: realtek,rtl-intc: map output lines
On Sun, 26 Dec 2021 19:59:27 +0000,
Sander Vanheule <sander@svanheule.net> wrote:
>
> Amend the binding to also require a list of parent interrupts, and an
> optional mask to specify which parent is mapped to which output.
>
> Without this information, any driver would have to make an assumption on
> which parent interrupt is connected to which output.

Why should an endpoint driver care at all?

>
> Additionally, extend (or add) the relevant descriptions to more clearly
> describe the inputs and outputs of this router.
>
> Signed-off-by: Sander Vanheule <sander@svanheule.net>
> ---
> Since it does not properly describe the hardware, I would still really
> rather get rid of "interrupt-map", even though that would mean breaking
> ABI for this binding. As we've argued before [1], that is our prefered
> solution, and would enable us to not carry more (hacky) code because of
> a mistake with the initial submission.

Again, this is too late. Broken bindings live forever.

>
> Vendors don't ship independent DT blobs for devices with this hardware,
> so the independent devicetree/kernel upgrades issue is really rather
> theoretical here. Realtek isn't driving the development of the bindings
> and associated drivers for this platform. They have their SDK and seem
> to care very little about proper kernel integration.

Any vendor can do whatever they want. You can do the same thing if you
really want to.

>
> Furthermore, there are currently no device descriptions in the kernel
> using this binding. There are in OpenWrt, but OpenWrt firmware images
> for this platform always contain both the kernel and the appended DTB,
> so there's also no breakage to worry about.

That's just one use case. Who knows who is using this stuff in a
different context? Nobody can tell.

>
> [1] https://lore.kernel.org/all/9c169aad-3c7b-2ffb-90a2-1ca791a3f411@phrozen.org/
>
> Differences with v1:
> - Don't drop the "interrupt-map" property
> - Add the "realtek,output-valid-mask" property
> ---
> .../realtek,rtl-intc.yaml | 38 ++++++++++++++++---
> 1 file changed, 33 insertions(+), 5 deletions(-)
>
> diff --git a/Documentation/devicetree/bindings/interrupt-controller/realtek,rtl-intc.yaml b/Documentation/devicetree/bindings/interrupt-controller/realtek,rtl-intc.yaml
> index 9e76fff20323..29014673c34e 100644
> --- a/Documentation/devicetree/bindings/interrupt-controller/realtek,rtl-intc.yaml
> +++ b/Documentation/devicetree/bindings/interrupt-controller/realtek,rtl-intc.yaml
> @@ -6,6 +6,10 @@ $schema: http://devicetree.org/meta-schemas/core.yaml#
>
> title: Realtek RTL SoC interrupt controller devicetree bindings
>
> +description:
> + Interrupt router for Realtek MIPS SoCs, allowing up to 32 SoC interrupts to
> + be routed to one of up to 15 parent interrupts, or left disconnected.
> +
> maintainers:
> - Birger Koblitz <mail@birger-koblitz.de>
> - Bert Vermeulen <bert@biot.com>
> @@ -22,7 +26,11 @@ properties:
> maxItems: 1
>
> interrupts:
> - maxItems: 1
> + minItems: 1
> + maxItems: 15
> + description:
> + List of parent interrupts, in the order that they are connected to this
> + interrupt router's outputs.

Is that to support multiple SoCs? I'd expect a given SoC to have a
fixed number of output interrupts.

>
> interrupt-controller: true
>
> @@ -30,7 +38,21 @@ properties:
> const: 0
>
> interrupt-map:
> - description: Describes mapping from SoC interrupts to CPU interrupts
> + description:
> + List of <soc_int parent_phandle parent_args ...> tuples, where "soc_int"
> + is the interrupt input line number as provided by this controller.
> + "parent_phandle" and "parent_args" specify which parent interrupt this
> + line should be routed to. Note that interrupt specifiers should be
> + identical to the parents specified in the "interrupts" property.
> +
> + realtek,output-valid-mask:
> + $ref: /schemas/types.yaml#/definitions/uint32
> + description:
> + Optional bit mask indicating which outputs are connected to the parent
> + interrupts. The lowest set bit indicates which output line the first
> + interrupt from "interrupts" is connected to, the second lowest set bit
> + for the second interrupt, etc. If not provided, parent interrupts will be
> + assigned sequentially to the outputs.
>
> required:
> - compatible
> @@ -39,6 +61,7 @@ required:
> - interrupt-controller
> - "#address-cells"
> - interrupt-map
> + - interrupts
>
> additionalProperties: false
>
> @@ -49,9 +72,14 @@ examples:
> #interrupt-cells = <1>;
> interrupt-controller;
> reg = <0x3000 0x20>;
> +
> + interrupt-parent = <&cpuintc>;
> + interrupts = <1>, <2>, <5>;
> + realtek,output-valid-mask = <0x13>;

What additional information does this bring? From the description
above, this is all SW configurable, so why should this be described in
the DT?

> +
> #address-cells = <0>;
> interrupt-map =
> - <31 &cpuintc 2>,
> - <30 &cpuintc 1>,
> - <29 &cpuintc 5>;
> + <31 &cpuintc 2>, /* connect to cpuintc 2 via output 1 */
> + <30 &cpuintc 1>, /* connect to cpuintc 1 via output 0 */
> + <29 &cpuintc 5>; /* connect to cpuintc 5 via output 4 */
> };

My conclusion here is that, as I stated in my initial review of this
series, you could completely ignore the 3rd field of the map, and let
the driver decide on the mapping without any extra information.

We already have plenty of crossbar-type drivers in the tree that can
mux a number of input to a number of outputs and route them
accordingly to a set of parent interrupts. None of this requires to be
described in DT.

M.

--
Without deviation from the norm, progress is not possible.

\
 
 \ /
  Last update: 2021-12-27 12:19    [W:0.181 / U:0.100 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site