Add LS1046A clockgen driver.

[dgr_semihalf.com]

Summary

Driver provides probe and attach functions for LS1046A clockgen and passes
configuration information to QorIQ clockgen class. May be used as
reference implementation for different QorIQ clockgen devices.

[manu]

This is a comment for all the LS1046A reviews: Please add a new SOC_NXP_LS1046A option to sys/conf/options.arm64 and make the drivers depend on that in sys/conf/files.arm64
Thanks

[dgr_semihalf.com]

I modified the driver to use the modified QorIQ clockgen classes.
Also added dependency on SOC_NXP_LS104A.

[manu]

sys/arm64/qoriq/clk/ls1046a_clkgen.c

173 ↗

(On Diff #71196)

this is the same id as the hwaccel1, is that intentional ?
If the clocks aren't exposed in the DTS an id of 0 is perfectly fine.

213 ↗

(On Diff #71196)

Why does this fixed clock needs a init function to be registered ?

[dgr_semihalf.com]

sys/arm64/qoriq/clk/ls1046a_clkgen.c

173 ↗

(On Diff #71196)

No, that was not intentional.
Most of the clocks are exposed in dts, but due to using two clock cells in dts the custom ofw_mapper function finds clock based on name instead of id.

213 ↗

(On Diff #71196)

I used init function here as the way FMAN nodes work is specific to SoC, so different way of creating this node may be needed.

[dgr_semihalf.com]

Hi, if you have any more comments or remarks, please let me know. Thanks.

[mmel]

Imho, It will be better to wait with this until we find consensus with D24351, right ?

sys/arm64/qoriq/clk/ls1046a_clkgen.c

59 ↗

(On Diff #71196)

There is no reason to obfuscate clock node name in this way, please use names from TRM

126 ↗

(On Diff #71196)

huh!. Use nitems() :)

173 ↗

(On Diff #71196)

again this is wrong way. Simply encode both cells into one id and use it.

[dgr_semihalf.com]

Modified to work with new version of D24351

[mmel]

The LS family has too many members and can share 99% of the code. It seems unnecessary for me to have the option for each individual variant. How about 'SOC_NXP_LS' ?

[dgr_semihalf.com]

I'll change it in all the patches to SOC_NXP_LS.

[manu]

In D24352#549462, @mmel wrote:

The LS family has too many members and can share 99% of the code. It seems unnecessary for me to have the option for each individual variant. How about 'SOC_NXP_LS' ?

That might be fine for now but what about the future ?

[mw]

@mmel @manu Please agree on the desired option naming as it affects all submitted patches. Once done, I'll commit.

[manu]

Just go with NXP_LS, if there is any problem later we'll handle it at this moment.

[mw]

Ok, thanks.

[mmel]

from IRC:
There is at least 11 variants of QoriQ arm64 based SoCs. This is too much for per-SOC options.
We can always add a SOC_NXP_LS_BLAH and !it if one soc differs

[dgr_semihalf.com]

Replace SOC_NXP_LS1046 with SOC_NXP_LS.