Fix ESP32-{S2/C2/C3} radio clocks

[bugadani]

Counterpart of #679. Submitted as a different PR because I can not test these and I would prefer not delaying the S3 pull request if this one is incorrect somehow.

If some of the code doesn't seem to make sense, you are right. But I just plucked the constants out of esp-idf, without simplifying them. This is particularly visible in S2, where the code:

• disables a non-existing bluetooth (and I think the value mask overlaps with the other values)
• and sets and clears the same bits at the same time (though, esp-idf does this in two register operations, but I don't think it matters).

I'm not touching C6 and H2, those are scary (i.e. they are different enough that I can't tell at a glance whether they are wrong)

[bjoernQ]

Works fine on S2,C2,C3.

On ESP32 WIFI also works fine but the BLE example crashes immediately:

INFO - COEX Version 2.0.0
ASSERT_PARAM(134220032 0), in lld.c at line 289


Exception occured 'IllegalInstruction'
Context
PC=0x40082edb       PS=0x00060b15
0x40082edb - r_assert
    at ??:??
0x00060b15 - PS_WOE
    at ??:??
A0=0x800832c1       A1=0x3ffc2320       A2=0x00000000       A3=0x08000900       A4=0x00000000
.........

[bugadani]

I don't have a usable ESP32 at hand, unfortunately. Can you please tell if that crash is a result of this PR or one of the previous ones? Since this one mostly just modifies register values, I'm betting on the latter.

[bjoernQ]

I don't have a usable ESP32 at hand, unfortunately. Can you please tell if that crash is a result of this PR or one of the previous ones? Since this one mostly just modifies register values, I'm betting on the latter.

2f091161b40ee985a1f12ee43c412c045ae809af (the commit before 4c34c8d) still works

[bugadani]

Bizarre. Thanks, I'll investigate somehow.

[bugadani]

In the mean time, should I revert ESP32 and offer the rest up for consideration, or should we wait until some miracle happens? :)

[bjoernQ]

In the mean time, should I revert ESP32 and offer the rest up for consideration, or should we wait until some miracle happens? :)

Probably better to revert ESP32 and get the other code merged👍

[bjoernQ]

LGTM - Thanks!