On Sunday, March 22, 2015 09:32:48 PM Julien Grall wrote:

On 22/03/2015 21:49, Rafael J. Wysocki wrote:

quoted

On Sunday, March 22, 2015 09:05:21 PM Julien Grall wrote:

quoted

Hello,

On 21/03/2015 12:09, Naresh Bhat wrote:

quoted

      From 268dcdafa34a690e2f99c0784ca33a6d2352ecf5 Mon Sep 17 00:00:00 2001
     From: Hanjun Guo <hanjun.guo at linaro.org <mailto:hanjun.guo@linaro.org>>
     Date: Sat, 21 Mar 2015 14:43:54 +0800
     Subject: [PATCH] XEN / ACPI: Make XEN ACPI depend on X86

     When ACPI is enabled on ARM64, XEN ACPI will also compiled
     into the kernel, but XEN ACPI is x86 dependent, so introduce
     CONFIG_XEN_ACPI to make it depend on x86 before XEN ACPI is
     functional on ARM64.

     Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org
     <mailto:hanjun.guo@linaro.org>>
     ---
       drivers/xen/Kconfig  | 4 ++++
       drivers/xen/Makefile | 2 +-
       2 files changed, 5 insertions(+), 1 deletion(-)

     diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig
     index b812462..a31cd29 100644
     --- a/drivers/xen/Kconfig
     +++ b/drivers/xen/Kconfig
     @@ -253,4 +253,8 @@ config XEN_EFI
           def_bool y
           depends on X86_64 && EFI

     +config XEN_ACPI
     +    def_bool y
     +    depends on X86 && ACPI
     +
       endmenu
     diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile
     index 2ccd359..f4622ab 100644
     --- a/drivers/xen/Makefile
     +++ b/drivers/xen/Makefile
     @@ -13,7 +13,7 @@ CFLAGS_efi.o                += -fshort-wchar

       dom0-$(CONFIG_PCI) += pci.o
       dom0-$(CONFIG_USB_SUPPORT) += dbgp.o
     -dom0-$(CONFIG_ACPI) += acpi.o $(xen-pad-y)
     +dom0-$(CONFIG_XEN_ACPI) += acpi.o $(xen-pad-y)
       xen-pad-$(CONFIG_X86) += xen-acpi-pad.o
       dom0-$(CONFIG_X86) += pcpu.o
       obj-$(CONFIG_XEN_DOM0)            += $(dom0-y)

[..]

quoted

AFAIK,  There is already a kernel patch exists to fix this issue.  I
think  Julien or Parth is a right person to ask.  Hence I am CCed Julien
Grall too.

The ACPI support for Xen is not ready. So I think avoiding to compile
drivers/xen/acpi.c on ARM64/ARM seems the better solution for now.

Although, rather than introducing a new CONFIG option, I would use the
same trick we use within the Makefile to avoid hotplug.c on ARM/ARM64.

ifeq ($(filter y, $(CONFIG_ARM) $(CONFIG_ARM64)), )
dom0-$(CONFIG_ACPI) += acpi.o $(xen-pad-y)
endif

Well, is avoiding an extra CONFIG_ option worth the ugliness of this?

When the support of ACPI for Xen will come, the CONFIG_ option will be 
an alias to CONFIG_XEN.

In this case the CONFIG_ option won't bring much improvement to the code 
and add an extra indirection.

The "ugliness" option has, at least, the advantage to be tiny and 
self-contained.

Oh well, not really.  You're moving a config-time check to compile time
which means that it will be done every time this Makefile is executed
and for all architectures that execute it.  Not nice.

Also I think that ia64 is missing from the list, but I may be wrong.

Not to mention the fact that the dependency will be rather difficult to find
for tools like xconfig ...


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.