On 4/17/22 20:24, Adam Ford wrote:
Hi,
When I just reverted all three of those, it still hangs. However,
when I removed CONFIG_CLK, it recovered. I'll examine which clocks
may be missing or if there is some other clock dependency.
Hum, clock sounds indeed more plausible cause of the hang because the
SMC calls (or the power domain driver) does not come up so early in the
process. (the other possibility is that you're running of MALLOC_F space
due to the extra clock).
The obvious other question I have would be -- does it hang if you only
enable CONFIG_CLK=y without these patches here ?
It appears that I just needed to enable the CCF, CCF composite, and
the IMX8MQ clock driver which were previously disabled by default.
With those enabled, it booted just fine.
Once it booted, I was able to start the USB system and it detected a
USB thumb drive.
I did not enable them in SPL, but I can try it if you want.
Oh, right, now you actually can bring up USB in SPL, that's a nice bonus.
Should we expand the power domain dependency to include these clocks,
or should these clocks just turn on by default if/when the 8MQ is the
target SOC?
Probably add something like this into the Kconfig file:
select CLK_IMX8MQ if IMX8MQ
for all four MX8M variants and for all SPL variants (so 8 entries total)
and that should fix the clock dependency in general.
At this point, I don't think it's a MALLOC_F issue, but I considered that.
I can mark your patch with a t-b, but I am not sure if you want to do
a V2 with these clocks enabled or do something different.
Maybe just send a subsequent patch. That's likely easier on Stefano,
since this series is already buries deep in his master-next branch.
