soc: bcm: brcmstb: fix ARM build errors

Message ID 20170907135136.2923775-1-arnd@arndb.de
State New
Headers show
Series
  • soc: bcm: brcmstb: fix ARM build errors
Related show

Commit Message

Arnd Bergmann Sept. 7, 2017, 1:50 p.m.
When building the new PM driver on older ARM architectures, we can
run into one of two build errors:

drivers/soc/bcm/brcmstb/pm/pm-arm.c: In function 'brcmstb_do_pmsm_power_down':
drivers/soc/bcm/brcmstb/pm/pm-arm.c:334:2: error: implicit declaration of function 'wfi' [-Werror=implicit-function-declaration]

drivers/soc/bcm/brcmstb/pm/pm-arm.o: In function `brcmstb_pm_s3_finish':
pm-arm.c:(.text+0x860): undefined reference to `cpu_resume'

The first one requires at least ARMv6K, the second one requires the CPU
suspend/resume logic which is not available on some of the older CPUs,
and needs to be selected explicitly.

This adds the extra Kconfig statements to enforce this, limiting
compilation on ARM to the STB platform that is guaranteed to be
ARMv7. We could enable compile-testing for other ARMv7 platforms,
but there seems to be little value as an allmodconfig kernel
already includes it.

Acked-by: Florian Fainelli <f.fainelli@gmail.com>

Signed-off-by: Arnd Bergmann <arnd@arndb.de>

---
 drivers/soc/bcm/brcmstb/Kconfig | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

-- 
2.9.0

Comments

Arnd Bergmann Sept. 7, 2017, 2:24 p.m. | #1
On Thu, Sep 7, 2017 at 3:50 PM, Arnd Bergmann <arnd@arndb.de> wrote:
> When building the new PM driver on older ARM architectures, we can

> run into one of two build errors:

>

> drivers/soc/bcm/brcmstb/pm/pm-arm.c: In function 'brcmstb_do_pmsm_power_down':

> drivers/soc/bcm/brcmstb/pm/pm-arm.c:334:2: error: implicit declaration of function 'wfi' [-Werror=implicit-function-declaration]

>

> drivers/soc/bcm/brcmstb/pm/pm-arm.o: In function `brcmstb_pm_s3_finish':

> pm-arm.c:(.text+0x860): undefined reference to `cpu_resume'

>

> The first one requires at least ARMv6K, the second one requires the CPU

> suspend/resume logic which is not available on some of the older CPUs,

> and needs to be selected explicitly.

>

> This adds the extra Kconfig statements to enforce this, limiting

> compilation on ARM to the STB platform that is guaranteed to be

> ARMv7. We could enable compile-testing for other ARMv7 platforms,

> but there seems to be little value as an allmodconfig kernel

> already includes it.

>

> Acked-by: Florian Fainelli <f.fainelli@gmail.com>

> Signed-off-by: Arnd Bergmann <arnd@arndb.de>


It seems we always compile the pm/ subdirectory regardless of CONFIG_BRCMSTB_PM,
so we also needs this change:

--- a/drivers/soc/bcm/brcmstb/Makefile
+++ b/drivers/soc/bcm/brcmstb/Makefile
@@ -1,2 +1,2 @@
 obj-y                          += common.o biuctrl.o
-obj-y                          += pm/
+obj-$(CONFIG_BRCMSTB_PM)       += pm/

I'll resubmit the combined patch after some more testing.

        Arnd
Florian Fainelli Sept. 7, 2017, 3:34 p.m. | #2
On 09/07/2017 07:24 AM, Arnd Bergmann wrote:
> On Thu, Sep 7, 2017 at 3:50 PM, Arnd Bergmann <arnd@arndb.de> wrote:

>> When building the new PM driver on older ARM architectures, we can

>> run into one of two build errors:

>>

>> drivers/soc/bcm/brcmstb/pm/pm-arm.c: In function 'brcmstb_do_pmsm_power_down':

>> drivers/soc/bcm/brcmstb/pm/pm-arm.c:334:2: error: implicit declaration of function 'wfi' [-Werror=implicit-function-declaration]

>>

>> drivers/soc/bcm/brcmstb/pm/pm-arm.o: In function `brcmstb_pm_s3_finish':

>> pm-arm.c:(.text+0x860): undefined reference to `cpu_resume'

>>

>> The first one requires at least ARMv6K, the second one requires the CPU

>> suspend/resume logic which is not available on some of the older CPUs,

>> and needs to be selected explicitly.

>>

>> This adds the extra Kconfig statements to enforce this, limiting

>> compilation on ARM to the STB platform that is guaranteed to be

>> ARMv7. We could enable compile-testing for other ARMv7 platforms,

>> but there seems to be little value as an allmodconfig kernel

>> already includes it.

>>

>> Acked-by: Florian Fainelli <f.fainelli@gmail.com>

>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

> 

> It seems we always compile the pm/ subdirectory regardless of CONFIG_BRCMSTB_PM,

> so we also needs this change:

> 

> --- a/drivers/soc/bcm/brcmstb/Makefile

> +++ b/drivers/soc/bcm/brcmstb/Makefile

> @@ -1,2 +1,2 @@

>  obj-y                          += common.o biuctrl.o

> -obj-y                          += pm/

> +obj-$(CONFIG_BRCMSTB_PM)       += pm/

> 

> I'll resubmit the combined patch after some more testing.


Ah yes, that was correct in the original pull request, but I messed up
the conflict resolution once dpfe.c was moved out of this directory into
drivers/memory/ sorry about that. Do you want to fold these two patches
into 6274ca43eba1b2ebcac96f3c9b41fbc4945c0af0?
-- 
Florian

Patch

diff --git a/drivers/soc/bcm/brcmstb/Kconfig b/drivers/soc/bcm/brcmstb/Kconfig
index d05bfce82e71..d36f6e03c1a6 100644
--- a/drivers/soc/bcm/brcmstb/Kconfig
+++ b/drivers/soc/bcm/brcmstb/Kconfig
@@ -4,6 +4,7 @@  config BRCMSTB_PM
 	bool "Support suspend/resume for STB platforms"
 	default y
 	depends on PM
-	depends on ARM || BMIPS_GENERIC
+	depends on ARCH_BRCMSTB || BMIPS_GENERIC
+	select ARM_CPU_SUSPEND if ARM
 
 endif # SOC_BRCMSTB