On Tuesday, July 12, 2016 3:32:59 PM CEST Wan Zongshun wrote:

On 2016?07?11? 23:36, Arnd Bergmann wrote:

quoted

On Sunday, July 10, 2016 3:27:23 PM CEST Wan Zongshun wrote:

quoted

+config NUC900_TIMER
+        bool "Clocksource timer for nuc900 platform" if COMPILE_TEST
+        depends on ARM
+        select CLKSRC_OF if OF
+        select CLKSRC_MMIO
+        help
+          Enables the clocksource for the NUC900 platform.

I have put this patch into my randconfig build system and found that
it lacks a dependency:

diff --git a/drivers/clocksource/Kconfig b/drivers/clocksource/Kconfig
index e18ef32776a3..59b9251eef37 100644
--- a/drivers/clocksource/Kconfig
+++ b/drivers/clocksource/Kconfig

@@ -523,7 +523,7 @@ config CLKSRC_ST_LPC

  config NUC900_TIMER
          bool "Clocksource timer for nuc900 platform" if COMPILE_TEST
-        depends on ARM
+        depends on ARM && GENERIC_CLOCKEVENTS
          select CLKSRC_OF if OF
          select CLKSRC_MMIO
          help

So this patch, I still need submit or you have merged it?

Ideally the driver should get submitted through the clocksource
maintainer tree. I have not applied it to any git tree that I
plan to send anywhere.

quoted

Also the init function has changed its return type in linux-next:

quoted

+static void __init nuc970_timer_of_init(struct device_node *node)

This now needs to return an error code or we get:

../include/linux/of.h:1004:20: error: comparison of distinct pointer types lacks a cast [-Werror]
         .data = (fn == (fn_type)NULL) ? fn : fn  }

Daniel Lezcano seems to have implemented a migration strategy, but I
can't see what you are supposed to do here, since the
CLOCKSOURCE_OF_DECLARE_RET macro is no longer part of linux-next.

Wait for Daniel's comments? or what should I do now?

Yes, let's see what he says. I guess from the timing, this will probably
have to wait for linux-4.9 anyway, and then we have no problem because the
API change will make it into 4.8.

	Arnd