On 1/18/17 8:41 AM, Jan Beulich wrote: >>>> On 18.01.17 at 14:33, <car...@cardoe.com> wrote: >> On 1/18/17 4:52 AM, Jan Beulich wrote: >>>>>> On 17.01.17 at 21:07, <car...@cardoe.com> wrote: >>>> @@ -686,6 +683,10 @@ paddr_t __init efi_multiboot2(EFI_HANDLE ImageHandle, >>>> EFI_SYSTEM_TABLE *SystemTa >>>> setup_efi_pci(); >>>> efi_variables(); >>>> >>>> + /* This is the maximum size of our trampoline + our low memory stack >>>> */ >>>> + cfg.size = max_t(UINTN, 64 << 10, >>>> + (trampoline_end - trampoline_start) + 4096); >>> >>> Considering the consuming code further up, I don't understand the >>> reason for the addition of 4096 here. And if there is a reason, I'm >>> pretty sure you actually mean PAGE_SIZE. >> >> You are correct. Given that the assembly is hardcoded at 64k there is no >> reason for this. I had kicked around doing a similar max() call in >> assembly instead of hardcoding the value but didn't do it. So I should >> just remove this. > > Well, I don't mind the max() (albeit it's not very useful, especially > if trampoline size would ever get pretty close to 64k). And as said > in reply to the earlier version - I think this would better be checked > at build time (see the various ASSERT()s at the end of xen.lds.S). > > Jan > Thank you. I had looked for BUILD_ASSERT() and didn't see it so I didn't know what mechanism we had which is why I didn't add it.
-- Doug Goldstein
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel