Le ven. 11 août 2023, 02:17, Pedro Miguel Justo <pm...@texair.net> a écrit :
> Hello! > > It seems as a recent change has caused a regression on Itanium (ia64), > rendering the platform unbootable. GRUB will crash very early with: > > > ``` > Welcome to GRUB! > > 7 0 0x00006B 0x000000000000001E unexpected trap > 7 0 0x000066 0x000000000000001E trap taken, number in ext PE > 7 0 0x00003C 0x0000000000005A00 trap taken, offset in ext PE > ``` > > I have bisected the issue and it resulted into the following change: > > ``` > 06edd40db76bb78457ac26156ed5f7b62381bbe8 is the first bad commit > commit 06edd40db76bb78457ac26156ed5f7b62381bbe8 > Author: Oliver Steffen <ostef...@redhat.com> > Date: Fri May 26 13:35:43 2023 +0200 > > guid: Unify GUID types > > There are 3 implementations of a GUID in GRUB. Replace them with > a common one, placed in types.h. > > It uses the "packed" flavor of the GUID structs, the alignment > attribute > is dropped, since it is not required. > > Signed-off-by: Oliver Steffen <ostef...@redhat.com> > Reviewed-by: Daniel Kiper <daniel.ki...@oracle.com> > ``` > > The issue is most likely related with an unaligned memory access exception > resulting from may global GUID variables being changed from > `grub_efi_guid_t` (which are decorated with `__attribute__ ((aligned(8)))`) > to the new and unified `grub_guid_t` type (now decorated with > `__attribute__ ((packed))`). > > Do you concur? > That makes perfect sense. Thank you for the analysis. When passing GUID to EFI we need it to be aligned. I propose to partially revert the commit and keep efi_guid either completely separate or to include common guid type as the only structure in it. > > Pedro > > > _______________________________________________ > Grub-devel mailing list > Grub-devel@gnu.org > https://lists.gnu.org/mailman/listinfo/grub-devel >
_______________________________________________ Grub-devel mailing list Grub-devel@gnu.org https://lists.gnu.org/mailman/listinfo/grub-devel
