| Message ID | 1441977482-29215-1-git-send-email-jsarha@ti.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Fri, Sep 11, 2015 at 04:18:02PM +0300, Jyri Sarha wrote: > The updated binding provides a way to set clock-ID and direction > parameters for DAI drivers set_sysclk() call back. > I proposed something similar about a year ago, but Mark rejected that > at the time. This RFC is to start that discussion again. This time > before I do any code changes. What's the use case again? Can we address this by converting the relevant drivers to the clock API (or improving their clock API support)?
On 09/19/15 21:42, Mark Brown wrote: > On Fri, Sep 11, 2015 at 04:18:02PM +0300, Jyri Sarha wrote: > >> The updated binding provides a way to set clock-ID and direction >> parameters for DAI drivers set_sysclk() call back. > >> I proposed something similar about a year ago, but Mark rejected that >> at the time. This RFC is to start that discussion again. This time >> before I do any code changes. > > What's the use case again? Can we address this by converting the > relevant drivers to the clock API (or improving their clock API > support)? > Sorry, I forgot to reply this earlier. The reason why we need this is the way McASP driver uses and provides clocks for different purposes. The most pressing need is to be able to select if we want to use some external clock pin as an input for McASP clock divider that produces the i2s bit-clock or if we want to use McASP's internal clock source. There are several other things this binding would allow us, and others with flexible i2s HW, to do. Some TI codecs would also benefit from a flexible way of describing the used clock configuration, but Peter know that part better. I tried to make the binding as flexible and generic as possible. But I do not currently see any immediate need for more than one set_sysclk() call per dai. I just did not see any reason to not allow it either. Best regards, Jyri
On Mon, Sep 28, 2015 at 09:49:35PM +0300, Jyri Sarha wrote: > On 09/19/15 21:42, Mark Brown wrote: > >What's the use case again? Can we address this by converting the > >relevant drivers to the clock API (or improving their clock API > >support)? > Sorry, I forgot to reply this earlier. The reason why we need this is the > way McASP driver uses and provides clocks for different purposes. The most > pressing need is to be able to select if we want to use some external clock > pin as an input for McASP clock divider that produces the i2s bit-clock or > if we want to use McASP's internal clock source. > There are several other things this binding would allow us, and others with > flexible i2s HW, to do. Some TI codecs would also benefit from a flexible > way of describing the used clock configuration, but Peter know that part > better. > I tried to make the binding as flexible and generic as possible. But I do > not currently see any immediate need for more than one set_sysclk() call per > dai. I just did not see any reason to not allow it either. This explains why you want to do this but what about the clock API portion of the question - it would be good to move the ASoC clocking more into the clock API, this would help integration with wider clock trees.
diff --git a/Documentation/devicetree/bindings/sound/simple-card.txt b/Documentation/devicetree/bindings/sound/simple-card.txt index cf3979e..d10bf2d 100644 --- a/Documentation/devicetree/bindings/sound/simple-card.txt +++ b/Documentation/devicetree/bindings/sound/simple-card.txt @@ -76,6 +76,7 @@ Optional CPU/CODEC subnodes properties: - dai-tdm-slot-num : Please refer to tdm-slot.txt. - dai-tdm-slot-width : Please refer to tdm-slot.txt. - clocks / system-clock-frequency : specify subnode's clock if needed. + it can be specified via "clocks" if system has clock node (= common clock), or "system-clock-frequency" (if system doens't support common clock) @@ -83,7 +84,21 @@ Optional CPU/CODEC subnodes properties: enabled with clk_prepare_enable() in dai startup() and disabled with clk_disable_unprepare() in dai - shutdown(). + shutdown(). "system-clock-frequency" + can also be an array if more than one + clock is described. +- clock-ids : An array of clock ID integers, + preferrably defined in DT header. + Each entry corresponds to the same + index postion first in "clocks" and + after the end of clocks array to + "system-clock-frequency" array. +- clock-dirs : An array of integers describing + clock directions: CLK_IN (= 0) or + OUT (= 1). Entries in the array + refer to clocks in the same way as + in clock-ids property. + Example 1 - single DAI link:
The updated binding provides a way to set clock-ID and direction parameters for DAI drivers set_sysclk() call back. Signed-off-by: Jyri Sarha <jsarha@ti.com> --- I proposed something similar about a year ago, but Mark rejected that at the time. This RFC is to start that discussion again. This time before I do any code changes. Best regards, Jyri Documentation/devicetree/bindings/sound/simple-card.txt | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-)