ACPI should depend on, not select PCI.
Otherwise, illegal configurations like X86_VOYAGER=y, PCI=y are
possible.
This patch also fixes the options select'ing ACPI to also select PCI.
Signed-off-by: Adrian Bunk <[email protected]>
---
This patch was already sent on:
- 16 Feb 2006
arch/ia64/Kconfig | 1 +
arch/x86_64/Kconfig | 1 +
drivers/acpi/Kconfig | 3 +--
3 files changed, 3 insertions(+), 2 deletions(-)
--- linux-2.6.16-rc3-mm1-full/drivers/acpi/Kconfig.old 2006-02-16 21:21:23.000000000 +0100
+++ linux-2.6.16-rc3-mm1-full/drivers/acpi/Kconfig 2006-02-16 21:21:42.000000000 +0100
@@ -10,9 +10,8 @@
config ACPI
bool "ACPI Support"
depends on IA64 || X86
+ depends on PCI
select PM
- select PCI
-
default y
---help---
Advanced Configuration and Power Interface (ACPI) support for
--- linux-2.6.16-rc3-mm1-full/arch/ia64/Kconfig.old 2006-02-16 21:21:52.000000000 +0100
+++ linux-2.6.16-rc3-mm1-full/arch/ia64/Kconfig 2006-02-16 21:22:08.000000000 +0100
@@ -73,6 +73,7 @@
config IA64_GENERIC
bool "generic"
select ACPI
+ select PCI
select NUMA
select ACPI_NUMA
help
--- linux-2.6.16-rc3-mm1-full/arch/x86_64/Kconfig.old 2006-02-16 21:22:19.000000000 +0100
+++ linux-2.6.16-rc3-mm1-full/arch/x86_64/Kconfig 2006-02-16 21:23:39.000000000 +0100
@@ -285,6 +285,7 @@
bool "ACPI NUMA detection"
depends on NUMA
select ACPI
+ select PCI
select ACPI_NUMA
default y
help
Adrian Bunk <[email protected]> wrote:
>
> ACPI should depend on, not select PCI.
>
It's surprising that there's any such linkage, actually. Is it
impossible for a non-PCI system to have ACPI?
On Sun, Mar 05, 2006 at 02:25:54PM -0800, Andrew Morton wrote:
> Adrian Bunk <[email protected]> wrote:
> >
> > ACPI should depend on, not select PCI.
> >
>
> It's surprising that there's any such linkage, actually. Is it
> impossible for a non-PCI system to have ACPI?
When ACPI was concieved (~1999 according to the spec I have handy)
ISA-only PCs were somewhat scarce. It's probably safe to say anything
that supports ACPI has PCI (For x86[64] at least, I don't know about ia64)
The spec mentions as a minimum requirement for OSPM/ACPI systems
amongst other things..
"A _PRT method for all root PCI bridges"
which pretty much sounds like it's at dependant on PCI (or a later
evolved standard like PCIE) hardware being present.
Dave
--
http://www.codemonkey.org.uk