| Message ID | 20210721140424.725744-8-maxime@cerno.tech |
|---|---|
| State | Superseded |
| Headers | show |
| Series | ARM: dts: Last round of DT schema fixes | expand |
On Wed, 21 Jul 2021 16:03:37 +0200, Maxime Ripard wrote: > The original binding was mentioning that valid values for the clocks and > clock-names property were one or two clocks from extclk, txco and lpo, > with extclk being deprecated in favor of txco. > > However, the current binding lists a valid array as extclk, txco and > lpo, with either one or two items. > > While this looks similar, it actually enforces that all the device trees > use either ["extclk"], or ["extclk", "txco"]. That doesn't make much > sense, since the two clocks are said to be equivalent, with one > superseeding the other. > > lpo is also not a valid clock anymore, and would be as the third clock > of the list, while we could have only this clock in the previous binding > (and in DTs). > > Let's rework the clock clause to allow to have either: > > - extclk, and mark it a deprecated > - txco alone > - lpo alone > - txco, lpo > > While ["extclk", "lpo"] wouldn't be valid, it wasn't found in any device > tree so it's not an issue in practice. > > Similarly, ["lpo", "txco"] is still considered invalid, but it's > generally considered as a best practice to fix the order of clocks. > > Cc: "David S. Miller" <davem@davemloft.net> > Cc: Jakub Kicinski <kuba@kernel.org> > Cc: Linus Walleij <linus.walleij@linaro.org> > Cc: netdev@vger.kernel.org > Signed-off-by: Maxime Ripard <maxime@cerno.tech> > --- > .../bindings/net/broadcom-bluetooth.yaml | 17 +++++++++++++++-- > 1 file changed, 15 insertions(+), 2 deletions(-) > Reviewed-by: Rob Herring <robh@kernel.org>
diff --git a/Documentation/devicetree/bindings/net/broadcom-bluetooth.yaml b/Documentation/devicetree/bindings/net/broadcom-bluetooth.yaml index fbdc2083bec4..5aac094fd217 100644 --- a/Documentation/devicetree/bindings/net/broadcom-bluetooth.yaml +++ b/Documentation/devicetree/bindings/net/broadcom-bluetooth.yaml @@ -50,16 +50,29 @@ properties: by interrupts and "host-wakeup" interrupt-names clocks: + minItems: 1 maxItems: 2 description: 1 or 2 clocks as defined in clock-names below, in that order clock-names: description: Names of the 1 to 2 supplied clocks - items: + oneOf: + - const: extclk + deprecated: true + description: Deprecated in favor of txco + - const: txco + description: > + external reference clock (not a standalone crystal) + - const: lpo - - const: extclk + description: > + external low power 32.768 kHz clock + + - items: + - const: txco + - const: lpo vbat-supply: description: phandle to regulator supply for VBAT
The original binding was mentioning that valid values for the clocks and clock-names property were one or two clocks from extclk, txco and lpo, with extclk being deprecated in favor of txco. However, the current binding lists a valid array as extclk, txco and lpo, with either one or two items. While this looks similar, it actually enforces that all the device trees use either ["extclk"], or ["extclk", "txco"]. That doesn't make much sense, since the two clocks are said to be equivalent, with one superseeding the other. lpo is also not a valid clock anymore, and would be as the third clock of the list, while we could have only this clock in the previous binding (and in DTs). Let's rework the clock clause to allow to have either: - extclk, and mark it a deprecated - txco alone - lpo alone - txco, lpo While ["extclk", "lpo"] wouldn't be valid, it wasn't found in any device tree so it's not an issue in practice. Similarly, ["lpo", "txco"] is still considered invalid, but it's generally considered as a best practice to fix the order of clocks. Cc: "David S. Miller" <davem@davemloft.net> Cc: Jakub Kicinski <kuba@kernel.org> Cc: Linus Walleij <linus.walleij@linaro.org> Cc: netdev@vger.kernel.org Signed-off-by: Maxime Ripard <maxime@cerno.tech> --- .../bindings/net/broadcom-bluetooth.yaml | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)