[Letux-kernel] OMAP5: Debugging kernel not starting...

H. Nikolaus Schaller hns at goldelico.com
Tue Sep 11 18:26:22 CEST 2018


Hi,

> Am 11.09.2018 um 15:07 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
> 
> Hi Tero,
> 
> After kernel has booted, there is a single difference
> in CM_ABE_CLKSTCTRL.
> 
> The relevant bit is: CLKACTIVITY_DPLL_ABE_X2_CLK
> 
> After reboot, there are bigger differences not wiped
> out by u-boot.
> 
> So the question is what disturbs this clock,
> especially as CLKACTIVITY_DPLL_ABE_X2_ seems
> to say that the clock is active (on Pyra) but
> not on EVM (I had expected the reverse to be more
> likely).
> 
> CM_ABE_CLKSTCTRL remains 0x803 on Pyra and 0x3 on EVM

Well, if we see the same setup by U-Boot after power-on
it must be kernel + DTS. Or is this a false conclusion?

So I started to deconfigure kernel subsystems completely:

1. all twl6040 / abe / aess
2. all OMAP3 & OMAP4
3. all wireless (wl1837 uses some external functional clocks)
4. all sound

And I have checked U-Boot and DTS sources but did not find
Pyra specific overwrites for any system clock or abe or dpll
related registers.

Then I did rm /lib/modules (the most critical ones are configured
into the kernel so that I can still boot), and interestingly the
CLKACTIVITY_DPLL_ABE_X2_CLK is now the same (0x3) on Pyra and
OMAP5EVM. And "reboot" works.

So I can now simply modprobe blacklist them one after the other
to find out which one is conflicting.

BR,
Nikolaus



More information about the Letux-kernel mailing list