On Monday, December 09, 2013 04:15:15 PM Bjorn Helgaas wrote: > On Fri, Dec 06, 2013 at 02:53:00PM +0000, Zheng, Lv wrote: > > Hi, > > > > > From: Zheng, Lv > > > Sent: Friday, December 06, 2013 4:52 PM > > > To: Wysocki, Rafael J; Brown, Len > > > > > > In Linux kernel, ACPICA is wrapped and safely exported by CONFIG_ACPI. So > > > all external modules should depend on CONFIG_ACPI rather than using ACPICA > > > header directly for stubbing. But if we moves <acpi/acpi.h> inclusions > > > into "#ifdef CONFIG_ACPI", build breakge can help to detect wrong ACPICA > > > dependent modules. > > > > > > One of the build breakage is: > > > arch/x86/pci/mmconfig-shared.c:389:1: error: unknown type name 'acpi_status' > > > arch/x86/pci/mmconfig-shared.c:389:47: warning: 'struct acpi_resource' declared inside parameter list [enabled by default] > > > arch/x86/pci/mmconfig-shared.c:389:47: warning: its scope is only this definition or declaration, which is probably not what you want > > > [enabled by default] > > > arch/x86/pci/mmconfig-shared.c: In function 'check_mcfg_resource': > > > arch/x86/pci/mmconfig-shared.c:392:33: error: storage size of 'address' isn't known > > > arch/x86/pci/mmconfig-shared.c:393:2: error: unknown type name 'acpi_status' > > > arch/x86/pci/mmconfig-shared.c:395:9: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:395:19: error: 'ACPI_RESOURCE_TYPE_FIXED_MEMORY32' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c:395:19: note: each undeclared identifier is reported only once for each function it appears in > > > arch/x86/pci/mmconfig-shared.c:397:8: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:399:11: error: 'AE_OK' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c:400:35: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:401:33: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:402:19: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:404:11: error: 'AE_CTRL_TERMINATE' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c:407:10: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:407:20: error: 'ACPI_RESOURCE_TYPE_ADDRESS32' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c:408:10: error: dereferencing pointer to incomplete type > > > arch/x86/pci/mmconfig-shared.c:408:20: error: 'ACPI_RESOURCE_TYPE_ADDRESS64' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c:411:2: error: implicit declaration of function 'acpi_resource_to_address64' [-Werror=implicit- > > > function-declaration] > > > arch/x86/pci/mmconfig-shared.c:412:2: error: implicit declaration of function 'ACPI_FAILURE' [-Werror=implicit-function-declaration] > > > arch/x86/pci/mmconfig-shared.c:414:31: error: 'ACPI_MEMORY_RANGE' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c:392:33: warning: unused variable 'address' [-Wunused-variable] > > > arch/x86/pci/mmconfig-shared.c: At top level: > > > arch/x86/pci/mmconfig-shared.c:425:1: error: unknown type name 'acpi_status' > > > arch/x86/pci/mmconfig-shared.c:425:41: error: unknown type name 'acpi_handle' > > > arch/x86/pci/mmconfig-shared.c: In function 'is_acpi_reserved': > > > arch/x86/pci/mmconfig-shared.c:447:2: error: implicit declaration of function 'acpi_get_devices' [-Werror=implicit-function- > > > declaration] > > > arch/x86/pci/mmconfig-shared.c:447:30: error: 'find_mboard_resource' undeclared (first use in this function) > > > arch/x86/pci/mmconfig-shared.c: At top level: > > > arch/x86/pci/mmconfig-shared.c:389:20: warning: 'check_mcfg_resource' defined but not used [-Wunused-function] > > > > Let me say something about this error messages. Someone may face difficulties to reproduce it. > > Currently, this message cannot be seen in any kernel build. > > This is because this patch series doesn't include a fix to remove <acpi/acpi.h> inclusion from linux/sfi_acpi.h. > > To achieve this, I have offered 2 options: > > 1. Redefining ACPICA table stuff for CONFIG_ACPI=n environment, or > > 2. Implementing stubs for all ACPICA functions/globals/macros and introducing <linux/acpica.h> (where acpi/acpi.h is included) as a top level header. > > 3. Other solutions... > > > > If we choose solution 1, then this issue is exposed by a build test where CONFIG_ACPI=n and CONFIG_SFI=y. > > Since we haven't decided what to do for the table structures and it is really not a good approach to achieve the build safety for mmconfig-share.c by relying on link-stage optimization rather than compile-stage referencing. IMO, we need to sort it out, thus I make it a part of this patchset. > > > > Well, if we choose solution 2, this patch can only help to reduce the size of the kernel image. > > OK, I assume you will make sure one of these fixes gets merged before the > change that causes the breakage, so bisection continues to work. > Therefore, I think it's best if both changes are merged through the same > tree, and I won't apply this via the PCI tree. Let me know if you need me > to do something. We're doing our best not to break compilation between the patches in this series. And I'll take care of it. :-) Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-acpi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
