On Wed, Sep 06, 2017 at 11:22:48AM +0200, Åukasz Majewski wrote:
Here is the routine that I understood from the code:
1) asoc_simple_card_parse_clk_cpu(dev, cpu, dai_link, cpu_dai);
=> asoc_simple_card_parse_clk(dev, cpu, // cpu node in sound{} [1]
dai_link->cpu_of_node, // node ssi2 [2]
cpu_dai, dai_link->cpu_dai_name);
==> 1.1) devm_get_clk_from_child(dev, node, NULL); // [1]
==> 1.2) of_property_read_u32(node, "system-clock-frequency", &val)// [1]
==> 1.3) devm_get_clk_from_child(dev, dai_of_node, NULL); // [2]
For the cpu routine, it first checks for clock property under cpu
node of simple-card, then for "system-clock-frequency" in the cpu
node of simple-card, and finally looks for clock property in ssi2
node.
-----> dev: sound
-----> clk node: /soc/aips-bus@02000000/spba-bus@02000000/ssi@0202c000
-----> Clk asignment
And this clock is taken from this node. It looks like ipg clock for ssi...
This makes sense now. The devm_get_clk_from_child() in 1.3) fetched
the first clock of ssi2 -- ipg clock.
The problem is with the "lack" of clock nodes/properties at
dailink_master: cpu {
sound-dai = <&ssi2>;
clock = <&SSSS>;
system-clock-frequency = <XXXX>;
};
This is the right solution based on current simple-card driver. For
SSI (having two clocks), you have to specify the baud clock in the
cpu node like that. I believe this is what the simple-card designer
expected users to do since the cpu node is the first place that the
driver tries to look at.
I think that the proper solution would be to add check for:
freq < sysclk/5 in fsl_ssi_set_dai_sysclk() and return -ENOTSUPP to
make the simple-audo-card driver happy (and not introducing
regressions).
As I said in the first place, adding another check in set_sysclk()
is not that essential but seems to be plausible to me. So I am okay
if you really want to have that.