On 2016/5/12 7:24, Matt Fleming wrote: > On Wed, 11 May, at 09:35:47PM, Shannon Zhao wrote: >>> > > >>> > > Also, why do you need to setup efi.runtime_version here? Isn't that >>> > > done inside uefi_init()? >>> > > >> > I don't see any codes which setup efi.runtime_version in uefi_init(). > > Look in the EFI tree, > > https://git.kernel.org/cgit/linux/kernel/git/mfleming/efi.git/tree/drivers/firmware/efi/arm-init.c?h=next#n120 > Ah, there are some patches in EFI tree introducing this change, but they are not in kernel matser but I didn't notice this, sorry. >>> > > Here is how I would have expected this patch to look: >>> > > >>> > > - Add FDT code to find hypervisor EFI params >>> > > >>> > > - Force enable EFI_RUNTIME_SERVICES for Xen and call >>> > > xen_efi_runtime_setup() inside xen_guest_init() >>> > > >>> > > - Change arm_enable_runtime_services() to handle the case where >>> > > EFI_RUNTIME_SERVICES is already set >>> > > >>> > > Am I misunderstanding some ordering issues? >> > >> > Since xen_guest_init() and arm_enable_runtime_services() are both >> > early_initcall and the calling order is random I think. > Urgh, right, I missed that. > >> > And when we call xen_efi_runtime_setup() and setup >> > efi.runtime_version in xen_guest_init(), the efi.systab might be >> > NULL. So it needs to map the systanle explicitly before we use >> > efi.systab. > Could you explain why you need to copy efi.runtime_version instead of > letting the existing code in uefi_init() set it up? > > Also, efi.systab isn't used by xen_efi_runtime_setup() as far I can > see, not sure why you need that either? As said above, I will rebase this patch on top of the EFI next branch. Thanks, -- Shannon -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html