On Thu, Jul 31, 2014 at 5:30 AM, Tony Lindgren tony@atomide.com wrote:
- Olof Johansson olof@lixom.net [140731 10:25]:
On Thu, Jul 31, 2014 at 12:21 AM, Olof Johansson olof@lixom.net wrote:
On Wed, Jul 30, 2014 at 11:39 PM, Olof's autobooter build@lixom.net wrote:
Full logs are available at http://arm-soc.lixom.net/bootlogs/arm-soc/v3.16-rc6-1009-g709032a/
Failed boards:
bbb arm-multi_v7_defconfig : FAILED 1:06.42 omap5uevm arm-multi_v7_defconfig : FAILED 1:15.77 omap5uevm arm-multi_lpae_defconfig : FAILED 1:13.66 pandaes arm-multi_v7_defconfig : FAILED 2:13.70This happened because ARCH_HISI wasn't selected, and the hisilicon platforms were the ones who used to bring in PINCTRL_SINGLE through a select.
Either we should remove the select from HISI, or we should add it to OMAP and others that need it. It's sort of silly to try to boot without pinctrl on these platforms, so selecting does make some sense...
Btw, I've made sure hisi is now selected, so the immediate need for this has been resolved. Still, the fragility remains.
Please let's not select drivers in Kconfig. That eventually leads into chasing of dependencies by adding more selects and randconfig build failures. Best to just remove the select for hisi and add it to the defconfigs instead.
For the pinctrl, it is possible to configure all the pins in the bootloader and not touch them at all in the kernel. Of course this will rule out some PM use cases that require dynamic pin configuration. But still no need to enforce it, the pinctrl just won't do anything if not selected.
Common practice on pretty much every other platform disagrees.
In reality, is there really ever a case where you're not going to want pinctrl enabled on OMAP?
It's possible to set it up in the bootloader mostly the right way, but it isn't done so today, and we do have a policy to not rely on the bootloader always getting everything right. Obviously FW doesn't get it right given the breakage seen when it accidentally got turned off.
-Olof