2014-02-14 14:54 GMT+01:00 Josh Boyer <jwboyer(a)fedoraproject.org>:
On Fri, Feb 14, 2014 at 8:38 AM, Nicolas Chauvet
<kwizart(a)gmail.com>
wrote:
> This prevent rtc to appear too late on the Toshiba AC100
> déc. 31 19:00:01 localhost kernel: drivers/rtc/hctosys.c: unable to open
rtc device (rtc0)
> ...
> déc. 31 19:00:01 localhost kernel: tps6586x-rtc tps6586x-rtc: rtc core:
registered tps6586x-rtc as rtc0
>
> The powered rtc is accesseded via i2c so this requires IC2_TEGRA=y
> in order to work
Shouldn't the rtc module have a dependency on I2C and do a
request_module on it? Building it into the kernel works, but only
because the i2c-tegra module is doing a subsys_initcall instead of
just plain module_init. Seems somewhat tenuous.
My understanding is that lsmod output doesn't show any dependency of the
rtc driver upon ic2_tegra or tps6586_regulator, but all theses modules are
appropriately loaded and hwclock -rf /dev/rtc0 shows the right time after a
cold boot with the current kernel.
Adding theses modules within the initramfs only allows the rtc driver to be
loaded bellow 1sec too late, and as soon as there is no option for the
kernel to delay the rtc clock read, the rtc driver cannot win the race when
built as a module.
There are probably others rtc modules in that situation. But I don't know
the how they are wired...
Looking at the i2c-tegra driver, I think it should be OK to built in
as it seems to have the proper device tree matches in place. I'm
curious if you tested this on any of the other boards that are built
in the multiboard armv7 kernels though.
Testing with this scratch build:
http://kojipkgs.fedoraproject.org//work/tasks/3307/6523307/kernel-3.14.0-...
on a wandboard quad but I have issue with fedora kernel package 3.14.rc2 or
3.14-rc3, I need to find a serial to inverstigate this boot time issue.
Nicolas (kwizart)