[alsa-devel] ASoC: About graph node issue without 'reg' property
Hello Morimoto-san, Mark,
I found that your commit b6f3fc005a2c ("ASoC: simple-card-utils: fixup asoc_simple_card_get_dai_id() counting") replaced with of_graph_parse_endpoint(), that needs 'reg' property in each port that has an endpoint.
Currently devicetrees for UniPhier SoCs have also audio graph nodes [1], and these graph nodes include an endpoint for each port. Although the port doesn't have 'reg' property, it has worked fine before. And after applying the commit, it doesn't work.
It is written in Documentation/devicetree/bindings/graph.txt that common scheme is to include the 'reg' property.
Maybe I think that we'd better add 'reg' properties to each port, however, there might be same matter in other SoCs, so I'm not sure whether your patch breaks the backword compatibility or our devicetrees are wrong.
I'd be glad if I could hear something about this issue from you.
Thank you,
[1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch...
--- Best Regards, Kunihiko Hayashi
On Thu, Dec 13, 2018 at 07:39:45PM +0900, Kunihiko Hayashi wrote:
Maybe I think that we'd better add 'reg' properties to each port, however, there might be same matter in other SoCs, so I'm not sure whether your patch breaks the backword compatibility or our devicetrees are wrong.
I'd be glad if I could hear something about this issue from you.
There's an ongoing thread between Morimoto-san and Tony Lindgren around the graph card including this issue, there Morimoto-san proposed some code which would let the reg property be optional. Device trees should still be updated but that way we wouldn't actually break anything which would be good.
On Thu, Dec 13, 2018 at 04:24:28PM +0000, Mark Brown wrote:
On Thu, Dec 13, 2018 at 07:39:45PM +0900, Kunihiko Hayashi wrote:
Maybe I think that we'd better add 'reg' properties to each port, however, there might be same matter in other SoCs, so I'm not sure whether your patch breaks the backword compatibility or our devicetrees are wrong.
I'd be glad if I could hear something about this issue from you.
There's an ongoing thread between Morimoto-san and Tony Lindgren around the graph card including this issue, there Morimoto-san proposed some code which would let the reg property be optional. Device trees should still be updated but that way we wouldn't actually break anything which would be good.
There's a patch now - "[PATCH][RFC] ASoC: simple-card-utils: fixup asoc_simple_card_get_dai_id() ID method".
Hi Hayashi-san
Sorry for my buggy patch. I and Tony could confirm that my posted patch can solve this issue. I will re-post it again as v3 patch-series. I'm happy if you can test it (and give it Tested-by).
https://patchwork.kernel.org/patch/10727943/
[1 <text/plain; us-ascii (quoted-printable)>] On Thu, Dec 13, 2018 at 04:24:28PM +0000, Mark Brown wrote:
On Thu, Dec 13, 2018 at 07:39:45PM +0900, Kunihiko Hayashi wrote:
Maybe I think that we'd better add 'reg' properties to each port, however, there might be same matter in other SoCs, so I'm not sure whether your patch breaks the backword compatibility or our devicetrees are wrong.
I'd be glad if I could hear something about this issue from you.
There's an ongoing thread between Morimoto-san and Tony Lindgren around the graph card including this issue, there Morimoto-san proposed some code which would let the reg property be optional. Device trees should still be updated but that way we wouldn't actually break anything which would be good.
There's a patch now - "[PATCH][RFC] ASoC: simple-card-utils: fixup asoc_simple_card_get_dai_id() ID method". [2 signature.asc <application/pgp-signature (7bit)>] No public key for 24D68B725D5487D0 created at 2018-12-14T03:27:38+0900 using RSA
Hi Morimoto-san, Mark,
On Fri, 14 Dec 2018 09:05:02 +0900 kuninori.morimoto.gx@renesas.com wrote:
Hi Hayashi-san
Sorry for my buggy patch. I and Tony could confirm that my posted patch can solve this issue. I will re-post it again as v3 patch-series. I'm happy if you can test it (and give it Tested-by).
Thank you for your great effort, and sorry my cross-post before your RFC patch because I didn't catch it.
I confirmed your RFC patch solved the issue. I'll also test your v3 patch-series on my environment.
[1 <text/plain; us-ascii (quoted-printable)>] On Thu, Dec 13, 2018 at 04:24:28PM +0000, Mark Brown wrote:
On Thu, Dec 13, 2018 at 07:39:45PM +0900, Kunihiko Hayashi wrote:
Maybe I think that we'd better add 'reg' properties to each port, however, there might be same matter in other SoCs, so I'm not sure whether your patch breaks the backword compatibility or our devicetrees are wrong.
I'd be glad if I could hear something about this issue from you.
There's an ongoing thread between Morimoto-san and Tony Lindgren around the graph card including this issue, there Morimoto-san proposed some code which would let the reg property be optional. Device trees should still be updated but that way we wouldn't actually break anything which would be good.
Yes. It's desiable to add the reg property in spite of the fixes. I'll update our devicetrees.
Thank you,
There's a patch now - "[PATCH][RFC] ASoC: simple-card-utils: fixup asoc_simple_card_get_dai_id() ID method". [2 signature.asc <application/pgp-signature (7bit)>] No public key for 24D68B725D5487D0 created at 2018-12-14T03:27:38+0900 using RSA
--- Best Regards, Kunihiko Hayashi
Hi Hayashi-san
There's an ongoing thread between Morimoto-san and Tony Lindgren around the graph card including this issue, there Morimoto-san proposed some code which would let the reg property be optional. Device trees should still be updated but that way we wouldn't actually break anything which would be good.
Yes. It's desiable to add the reg property in spite of the fixes. I'll update our devicetrees.
I'm not OF-graph guy, but having reg is optional, not mandatory I think. Yeah, my previous patch was not good for it :(
Best regards --- Kuninori Morimoto
participants (3)
-
Kunihiko Hayashi -
Kuninori Morimoto -
Mark Brown