| Message ID | 20240619085322.66716-3-angelogioacchino.delregno@collabora.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [1/3] arm64: dts: mediatek: mt8188: Fix VPPSYS0/1 node name/compatibles | expand |
On Wed, Jun 19, 2024 at 10:53:22AM +0200, AngeloGioacchino Del Regno wrote: > Add mediatek,mt8188-gce to the list of compatibles for which the > clock-names property is not required. Because, I assume, it has some internal clock? Why do either of these things have no clock? Doesn't the internal logic require one? > > Fixes: f2b53c295620 ("dt-bindings: mailbox: mediatek,gce-mailbox: add mt8188 compatible name") > Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> > --- > .../devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml > index cef9d7601398..55d4c34aa4b4 100644 > --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml > +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml > @@ -62,7 +62,9 @@ allOf: > properties: > compatible: > contains: > - const: mediatek,mt8195-gce > + enum: > + - mediatek,mt8188-gce > + - mediatek,mt8195-gce > then: > required: > - clock-names > -- > 2.45.2 >
Il 19/06/24 19:49, Conor Dooley ha scritto: > On Wed, Jun 19, 2024 at 10:53:22AM +0200, AngeloGioacchino Del Regno wrote: >> Add mediatek,mt8188-gce to the list of compatibles for which the >> clock-names property is not required. > > Because, I assume, it has some internal clock? Why do either of these > things have no clock? Doesn't the internal logic require one? > Because there's no gce0/gce1 clock, there's only an infracfg_AO clock that is for one GCE instance, hence there's no need to require clock-names. I can't remove the clock-names requirement from the older compatibles though, because the (sorry about this word) driver (eh..) gets the clock by name for the single GCE SoCs... ...and here comes a self-NACK for this commit, I have to fix the driver and then stop requiring clock-names on all compatibles, instead of having this ugly nonsense. Self-note: gce0/gce1 clocks lookup was implemented in the driver but never used and never added to the binding - luckily. Sorry Conor, I just acknowledged that there's a better way of doing that. Thank you for making me re-read this stuff, I'll send the proper changes later today, driver change + binding change in a separate series. As for the other two commits in this series, completely unrelated to GCE, those are still fine, and are fixing dtbs_check warnings. Cheers, Angelo >> >> Fixes: f2b53c295620 ("dt-bindings: mailbox: mediatek,gce-mailbox: add mt8188 compatible name") >> Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> >> --- >> .../devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml | 4 +++- >> 1 file changed, 3 insertions(+), 1 deletion(-) >> >> diff --git a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml >> index cef9d7601398..55d4c34aa4b4 100644 >> --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml >> +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml >> @@ -62,7 +62,9 @@ allOf: >> properties: >> compatible: >> contains: >> - const: mediatek,mt8195-gce >> + enum: >> + - mediatek,mt8188-gce >> + - mediatek,mt8195-gce >> then: >> required: >> - clock-names >> -- >> 2.45.2 >>
On Thu, Jun 20, 2024 at 10:01:18AM +0200, AngeloGioacchino Del Regno wrote: > Il 19/06/24 19:49, Conor Dooley ha scritto: > > On Wed, Jun 19, 2024 at 10:53:22AM +0200, AngeloGioacchino Del Regno wrote: > > > Add mediatek,mt8188-gce to the list of compatibles for which the > > > clock-names property is not required. > > > > Because, I assume, it has some internal clock? Why do either of these > > things have no clock? Doesn't the internal logic require one? > > > > Because there's no gce0/gce1 clock, there's only an infracfg_AO clock that is > for one GCE instance, hence there's no need to require clock-names. clock-names, d'oh. I misread that completely yesterday. > I can't remove the clock-names requirement from the older compatibles though, > because the (sorry about this word) driver (eh..) gets the clock by name for > the single GCE SoCs... > > ...and here comes a self-NACK for this commit, I have to fix the driver and > then stop requiring clock-names on all compatibles, instead of having this > ugly nonsense. Is it not worth keeping the clock names, even if ugly or w/e, because things have been done that way for a while? Also, what does U-Boot do on these systems to get the clocks? > Self-note: gce0/gce1 clocks lookup was implemented in the driver but never > used and never added to the binding - luckily. > > Sorry Conor, I just acknowledged that there's a better way of doing that. > > Thank you for making me re-read this stuff, I'll send the proper changes > later today, driver change + binding change in a separate series. > > As for the other two commits in this series, completely unrelated to GCE, > those are still fine, and are fixing dtbs_check warnings.
Il 20/06/24 10:22, Conor Dooley ha scritto: > On Thu, Jun 20, 2024 at 10:01:18AM +0200, AngeloGioacchino Del Regno wrote: >> Il 19/06/24 19:49, Conor Dooley ha scritto: >>> On Wed, Jun 19, 2024 at 10:53:22AM +0200, AngeloGioacchino Del Regno wrote: >>>> Add mediatek,mt8188-gce to the list of compatibles for which the >>>> clock-names property is not required. >>> >>> Because, I assume, it has some internal clock? Why do either of these >>> things have no clock? Doesn't the internal logic require one? >>> >> >> Because there's no gce0/gce1 clock, there's only an infracfg_AO clock that is >> for one GCE instance, hence there's no need to require clock-names. > > clock-names, d'oh. I misread that completely yesterday. > >> I can't remove the clock-names requirement from the older compatibles though, >> because the (sorry about this word) driver (eh..) gets the clock by name for >> the single GCE SoCs... >> >> ...and here comes a self-NACK for this commit, I have to fix the driver and >> then stop requiring clock-names on all compatibles, instead of having this >> ugly nonsense. > > Is it not worth keeping the clock names, even if ugly or w/e, because > things have been done that way for a while? It's worth allowing clock-names, but *requiring* that is unnecessary because there is, and there will always be, only one clock...! > Also, what does U-Boot do on these systems to get the clocks? > U-Boot doesn't support GCE at all (no driver - at least upstream)...! >> Self-note: gce0/gce1 clocks lookup was implemented in the driver but never >> used and never added to the binding - luckily. >> >> Sorry Conor, I just acknowledged that there's a better way of doing that. >> >> Thank you for making me re-read this stuff, I'll send the proper changes >> later today, driver change + binding change in a separate series. >> >> As for the other two commits in this series, completely unrelated to GCE, >> those are still fine, and are fixing dtbs_check warnings.
On Thu, Jun 20, 2024 at 10:32:36AM +0200, AngeloGioacchino Del Regno wrote: > Il 20/06/24 10:22, Conor Dooley ha scritto: > > On Thu, Jun 20, 2024 at 10:01:18AM +0200, AngeloGioacchino Del Regno wrote: > > > Il 19/06/24 19:49, Conor Dooley ha scritto: > > > > On Wed, Jun 19, 2024 at 10:53:22AM +0200, AngeloGioacchino Del Regno wrote: > > > > > Add mediatek,mt8188-gce to the list of compatibles for which the > > > > > clock-names property is not required. > > > > > > > > Because, I assume, it has some internal clock? Why do either of these > > > > things have no clock? Doesn't the internal logic require one? > > > > > > > > > > Because there's no gce0/gce1 clock, there's only an infracfg_AO clock that is > > > for one GCE instance, hence there's no need to require clock-names. > > > > clock-names, d'oh. I misread that completely yesterday. > > > > > I can't remove the clock-names requirement from the older compatibles though, > > > because the (sorry about this word) driver (eh..) gets the clock by name for > > > the single GCE SoCs... > > > > > > ...and here comes a self-NACK for this commit, I have to fix the driver and > > > then stop requiring clock-names on all compatibles, instead of having this > > > ugly nonsense. > > > > Is it not worth keeping the clock names, even if ugly or w/e, because > > things have been done that way for a while? > > It's worth allowing clock-names, but *requiring* that is unnecessary because > there is, and there will always be, only one clock...! Right, dunno if I misread you earlier or misunderstood. Fighting fires at work and replying to mails mid bisection is what I am going to blame ;) > > Also, what does U-Boot do on these systems to get the clocks? > > > > U-Boot doesn't support GCE at all (no driver - at least upstream)...! Running LIFO through my mailbox today, seeing this after the other mail.. > > > > Self-note: gce0/gce1 clocks lookup was implemented in the driver but never > > > used and never added to the binding - luckily. > > > > > > Sorry Conor, I just acknowledged that there's a better way of doing that. > > > > > > Thank you for making me re-read this stuff, I'll send the proper changes > > > later today, driver change + binding change in a separate series. > > > > > > As for the other two commits in this series, completely unrelated to GCE, > > > those are still fine, and are fixing dtbs_check warnings. > >
diff --git a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml index cef9d7601398..55d4c34aa4b4 100644 --- a/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml +++ b/Documentation/devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml @@ -62,7 +62,9 @@ allOf: properties: compatible: contains: - const: mediatek,mt8195-gce + enum: + - mediatek,mt8188-gce + - mediatek,mt8195-gce then: required: - clock-names
Add mediatek,mt8188-gce to the list of compatibles for which the clock-names property is not required. Fixes: f2b53c295620 ("dt-bindings: mailbox: mediatek,gce-mailbox: add mt8188 compatible name") Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com> --- .../devicetree/bindings/mailbox/mediatek,gce-mailbox.yaml | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)