On 01/19/2018 09:16 PM, xypron.glpk@gmx.de wrote:
From: Heinrich Schuchardt xypron.glpk@gmx.de
The calling convention for the entry point of an EFI image is always 'asmlinkage'.
Signed-off-by: Heinrich Schuchardt xypron.glpk@gmx.de
v4 rebase according to https://github.com/agraf/efi_next v3 Use efi_handle_t as type of the image handle. v2 no change
lib/efi_loader/efi_boottime.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/lib/efi_loader/efi_boottime.c b/lib/efi_loader/efi_boottime.c index 7c61dfb3a7..324abe4d48 100644 --- a/lib/efi_loader/efi_boottime.c +++ b/lib/efi_loader/efi_boottime.c @@ -1530,7 +1530,8 @@ static efi_status_t EFIAPI efi_start_image(efi_handle_t image_handle, unsigned long *exit_data_size, s16 **exit_data) {
- ulong (*entry)(efi_handle_t image_handle, struct efi_system_table *st);
- asmlinkage ulong (*entry)(efi_handle_t image_handle,
struct efi_system_table *st);
Alex,
could you once again carefully review this change.
Have a look at the definition of EFIAPI in include/efi.h
In cmd/bootefi.c we assume the entry point is asmlinkage. In lib/efi_loader/helloworld.c we assume it is EFIAPI.
The definition of EFIAPI depends on CONFIG_EFI_STUB_64BIT, which is only defined in configs/qemu-x86_efi_payload64_defconfig.
Why should the definition of EFIAPI depend on whether we are consuming or offering the API? Shouldn't EFIAPI have the same definition when compiling qemu-x86_64_defconfig?
Am I right that the entry point should in cmd/bootefi.c, helloworld.c, efi_start_image() should always be defined as
EFIAPI efi_status_t (*entry)(efi_handle_t image_handle
and further when compiling for x86_64 we should always define EFIAPI as
__attribute__((ms_abi))
and on other systems as
asmlinkage
Best regards
Heinrich
struct efi_loaded_image *info = image_handle; efi_status_t ret;