Re: [PATCH v2 1/4] dt-bindings: soundwire: add slave bindings

From: Pierre-Louis Bossart
Date: Thu Aug 08 2019 - 11:58:53 EST



+++ b/Documentation/devicetree/bindings/soundwire/slave.txt
@@ -0,0 +1,46 @@
+SoundWire slave device bindings.
+
+SoundWire is a 2-pin multi-drop interface with data and clock line.
+It facilitates development of low cost, efficient, high performance systems.
+
+SoundWire slave devices:
+Every SoundWire controller node can contain zero or more child nodes
+representing slave devices on the bus. Every SoundWire slave device is
+uniquely determined by the enumeration address containing 5 fields:
+SoundWire Version, Instance ID, Manufacturer ID, Part ID and Class ID
+for a device. Addition to below required properties, child nodes can
+have device specific bindings.

In case the controller supports multiple links, what's the encoding then?
in the MIPI DisCo spec there is a linkId field in the _ADR encoding that helps identify which link the Slave device is connected to

+
+Required property for SoundWire child node if it is present:
+- compatible: "sdwVER,MFD,PID,CID". The textual representation of
+ SoundWire Enumeration address comprising SoundWire
+ Version, Manufacturer ID, Part ID and Class ID,
+ shall be in lower-case hexadecimal with leading
+ zeroes suppressed.
+ Version number '0x10' represents SoundWire 1.0
+ Version number '0x11' represents SoundWire 1.1
+ ex: "sdw10,0217,2010,0"
+
+- sdw-instance-id: Should be ('Instance ID') from SoundWire
+ Enumeration Address. Instance ID is for the cases
+ where multiple Devices of the same type or Class
+ are attached to the bus.

so it is actually required if you have a single Slave device? Or is it only required when you have more than 1 device of the same type?

FWIW in the MIPI DisCo spec we kept the instanceID as part of the _ADR, so it's implicitly mandatory (and ignored by the bus if there is only one device of the same time)

+
+SoundWire example for Qualcomm's SoundWire controller:
+
+soundwire@c2d0000 {
+ compatible = "qcom,soundwire-v1.5.0"
+ reg = <0x0c2d0000 0x2000>;
+
+ spkr_left:wsa8810-left{
+ compatible = "sdw10,0217,2010,0";
+ sdw-instance-id = <1>;
+ ...
+ };
+
+ spkr_right:wsa8810-right{
+ compatible = "sdw10,0217,2010,0";
+ sdw-instance-id = <2>;

Isn't the MIPI encoding reported in the Dev_ID0..5 registers 0-based?

+ ...
+ };
+};


And now that I think of it, wouldn't it be simpler for everyone if we aligned on that MIPI DisCo public spec? e.g. you'd have one property with a 64-bit number that follows the MIPI spec. No special encoding necessary for device tree cases, your DT blob would use this:

soundwire@c2d0000 {
compatible = "qcom,soundwire-v1.5.0"
reg = <0x0c2d0000 0x2000>;

spkr_left:wsa8810-left{
compatible = "sdw0000100217201000"
}

spkr_right:wsa8810-right{
compatible = "sdw0000100217201100"
}
}

We could use parentheses if it makes people happier, but the information from the MIPI DisCo spec can be used as is, and provide a means for spec changes via reserved bits.