Linux: Clear CONFIG_INITRAMFS_SOURCE if BR2_TARGET_ROOTFS_INITRAMFS is not set

This patch makes sure CONFIG_INITRAMFS_SOURCE option gets reset
when BR2_TARGET_ROOTFS_INITRAMFS is not set. Some kernel defconfigs, 
for example for ARC processors, set CONFIG_INITRAMFS_SOURCE value, which cause such error:
"./scripts/gen_initramfs_list.sh: Cannot open '../../arc_initramfs_hs/'"

Signed-off-by: Evgeniy Didin <didin@synopsys.com> 
CC: Alexey Brodkin <abrodkin@synopsys.com>
Cc: arc-buildroot@synopsys.com
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

---
 linux/linux.mk | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

>>>>> "Evgeniy" == Evgeniy Didin <Evgeniy.Didin@synopsys.com> writes:

 > This patch makes sure CONFIG_INITRAMFS_SOURCE option gets reset
 > when BR2_TARGET_ROOTFS_INITRAMFS is not set. Some kernel defconfigs, 
 > for example for ARC processors, set CONFIG_INITRAMFS_SOURCE value, which cause such error:
 > "./scripts/gen_initramfs_list.sh: Cannot open '../../arc_initramfs_hs/'"

I'm not sure. I really would like to keep our "magic" .config fixups to
a minimum as they are kind of hidden and force specific use cases
without any options of disabling it (E.G. perhaps somebody has a
prebuilt initramfs they would like to include even though buildroot
builds the real rootfs (so BR2_TARGET_ROOTFS_INITRAMFS is not set). With
this change, this is no longer possible.

Couldn't you fix these arc specific defconfig issues with a small
kconfig fragment instead?

 > Signed-off-by: Evgeniy Didin <didin@synopsys.com> 
 > CC: Alexey Brodkin <abrodkin@synopsys.com>
 > Cc: arc-buildroot@synopsys.com
 > Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>

 > ---
 >  linux/linux.mk | 3 ++-
 >  1 file changed, 2 insertions(+), 1 deletion(-)

 > diff --git a/linux/linux.mk b/linux/linux.mk
 > index 9c2aa77..0eee315 100644
 > --- a/linux/linux.mk
 > +++ b/linux/linux.mk
 > @@ -269,7 +269,8 @@ define LINUX_KCONFIG_FIXUP_CMDS
 >  		touch $(BINARIES_DIR)/rootfs.cpio
 >  		$(call KCONFIG_SET_OPT,CONFIG_INITRAMFS_SOURCE,"$${BR_BINARIES_DIR}/rootfs.cpio",$(@D)/.config)
 >  		$(call KCONFIG_SET_OPT,CONFIG_INITRAMFS_ROOT_UID,0,$(@D)/.config)
 > -		$(call KCONFIG_SET_OPT,CONFIG_INITRAMFS_ROOT_GID,0,$(@D)/.config))
 > +		$(call KCONFIG_SET_OPT,CONFIG_INITRAMFS_ROOT_GID,0,$(@D)/.config),
 > +		$(call KCONFIG_SET_OPT,CONFIG_INITRAMFS_SOURCE,"",$(@D)/.config))
 >  	$(if $(BR2_ROOTFS_DEVICE_CREATION_STATIC),,
 >  		$(call KCONFIG_ENABLE_OPT,CONFIG_DEVTMPFS,$(@D)/.config)
 >  		$(call KCONFIG_ENABLE_OPT,CONFIG_DEVTMPFS_MOUNT,$(@D)/.config))
 > -- 
 > 2.9.3

 > _______________________________________________
 > buildroot mailing list
 > buildroot@busybox.net
 > http://lists.busybox.net/mailman/listinfo/buildroot

Hi Peter,

On Mon, 2017-11-13 at 15:03 +0100, Peter Korsgaard wrote:
> > 

> > > 

> > > > 

> > > > > 

> > > > > > 

> > > > > > "Evgeniy" == Evgeniy Didin <Evgeniy.Didin@synopsys.com> writes:

> 

>  > This patch makes sure CONFIG_INITRAMFS_SOURCE option gets reset

>  > when BR2_TARGET_ROOTFS_INITRAMFS is not set. Some kernel defconfigs, 

>  > for example for ARC processors, set CONFIG_INITRAMFS_SOURCE value, which cause such error:

>  > "./scripts/gen_initramfs_list.sh: Cannot open '../../arc_initramfs_hs/'"

> 

> I'm not sure. I really would like to keep our "magic" .config fixups to

> a minimum as they are kind of hidden and force specific use cases

> without any options of disabling it (E.G. perhaps somebody has a

> prebuilt initramfs they would like to include even though buildroot

> builds the real rootfs (so BR2_TARGET_ROOTFS_INITRAMFS is not set). With

> this change, this is no longer possible.


Well a couple of things I was thinking about:
1. This would be [in my view] just a complimentary option to setting BR's rootfs.
2. Usage of prebuilt initramf with BR might be quite a tricky thing because
   different toolchain could be used with different ABI, libc etc so chances
   are quite high that user-space stuff built with BR will be incompatible with
   contents of prebuilt rootfs. But I guess we cannot stop people from doing that.

> Couldn't you fix these arc specific defconfig issues with a small

> kconfig fragment instead?


So maybe this is indeed a bit better solution even though we know nobody
who'd do such a trick with pre-built initramfs.

-Alexey

>>>>> "Alexey" == Alexey Brodkin <Alexey.Brodkin@synopsys.com> writes:

Hi,

 > Well a couple of things I was thinking about:
 > 1. This would be [in my view] just a complimentary option to setting BR's rootfs.
 > 2. Usage of prebuilt initramf with BR might be quite a tricky thing because
 >    different toolchain could be used with different ABI, libc etc so chances
 >    are quite high that user-space stuff built with BR will be incompatible with
 >    contents of prebuilt rootfs. But I guess we cannot stop people from doing that.

Correct, but that is in reality not so difficult to do as the kernel ABI
is stable, and most architectures have stable ABIs. Libc doesn't enter
the question as the initramfs typically will not use files from the
rootfs or the other way around.


 >> Couldn't you fix these arc specific defconfig issues with a small
 >> kconfig fragment instead?

 > So maybe this is indeed a bit better solution even though we know nobody
 > who'd do such a trick with pre-built initramfs.

I do ;) I am using an external initramfs instead of built into the
kernel though, but that is just a minor detail.