Paul,
On Thu, Jan 24, 2019 at 10:41 PM Paul Cercueil <paul@xxxxxxxxxxxxxxx <mailto:paul@xxxxxxxxxxxxxxx>> wrote:
Hi Mathieu,
Le jeu. 24 janv. 2019 à 18:26, Mathieu Malaterre <malat@xxxxxxxxxx <mailto:malat@xxxxxxxxxx>> a
écrit :
> Paul,
>
> On Wed, Dec 12, 2018 at 11:09 PM Paul Cercueil <paul@xxxxxxxxxxxxxxx <mailto:paul@xxxxxxxxxxxxxxx>>
> wrote:
>>
>> Hi,
>>
>> Here's the version 8 and hopefully final version of my patchset,
>> which
>> adds support for the Timer/Counter Unit found in JZ47xx SoCs from
>> Ingenic.
>
> I can no longer boot my MIPS Creator CI20 with this series (merged
> opendingux/for-upstream-timer-v8).
>
> Using screen+ttyUSB, I can see messages stopping at:
>
> ...
> [ OK ] Started Cgroup management daemon.
> Starting Regular background program processing daemon...
> [ OK ] Started Regular background program processing daemon.
> Starting System Logging Service...
> Starting Provide limited super user privileges to specific
> users...
> Starting Restore /etc/resolv.conf if the system cras...s
> shut down....
> Starting WPA supplicant...
> Starting D-Bus System Message Bus...
> [ OK ] Started D-Bus System Message Bus.
>
> Nothing really stands out in the error messages. Could you suggest
> things to try out to get into a bootable state ?
I'm debugging it right now on jz4740, it seems to happen when the
clocksource
from the ingenic-timer driver is used. Is it your case? It should not
happen
if you have CONFIG_INGENIC_OST set.
Here is what I see:
$ grep CONFIG_INGENIC_OST arch/mips/configs/ci20_defconfig
CONFIG_INGENIC_OST=y
$ make O=ci20 ARCH=mips CROSS_COMPILE=mipsel-linux-gnu- ci20_defconfig
$ grep CONFIG_INGENIC_OST ci20/.config
CONFIG_INGENIC_OST=y
The setting is coming from your commit:
8f66e6b9c98f MIPS: CI20: defconfig: enable OST driver
In an attempt to solve the symptoms I even played with the clock rates
with no success:
&tcu {
/* 3 MHz for the system timer and clocksource */
assigned-clocks = <&tcu TCU_CLK_TIMER0>, <&tcu TCU_CLK_TIMER1>;
assigned-clock-rates = <750000>, <750000>;
};
>> The big change is that the timer driver has been simplified. The
>> code to
>> dynamically update the system timer or clocksource to a new channel
>> has
>> been removed. Now, the system timer and clocksource are provided as
>> children nodes in the devicetree, and the TCU channel to use for
>> these
>> is deduced from their respective memory resource. The PWM driver
>> will
>> also deduce from its memory resources whether a given PWM channel
>> can be
>> used, or is reserved for the system timers.
>>
>> Kind regards,
>> - Paul Cercueil
>>