On 23 September 2015 at 12:31, Ard Biesheuvel <[email protected]> wrote: > On 18 September 2015 at 07:30, Ard Biesheuvel <[email protected]> > wrote: >> This is a follow up to the series I sent out a week ago: >> http://thread.gmane.org/gmane.comp.bios.edk2.devel/2162 >> >> These two patches only address the special case in GenFw for ARM where it >> allows sections that are not aligned to the PE/COFF global section alignment, >> to interoperate with a size optimization in RVCT that emits sections at an >> offset that is not aligned to their own alignment. This violates the PE/COFF >> spec, and complicates the creation of DXE runtime modules with 4 KB section >> alignment. >> >> As it turns out, this optimization can be disabled by passing the >> --no_legacyalign command line option to the RVCT linker, so instead of >> patching things up, this time I simply disable the optimization. >> >> Patch #1 removes the optimization from the 64-bit ELF code, since RVCT >> is 32-bit only. >> >> Patch #2 disables the optimization, and removes the implementation of >> the special case from the 32-bit ELF code. >> >> Anyone with access to RVCT able to test this? I would highly appreciate it. >> > > OK, I managed to get access to RVCT myself, and gave this a spin. > > As I had hoped, just removing the workaround in GenFw breaks the build > for RVCTLINUX, and I can get it working correctly again by adding the > --no_legacyalign linker switch.
Thanks Ard. Reviewed-by: Leif Lindholm <[email protected]> _______________________________________________ edk2-devel mailing list [email protected] https://lists.01.org/mailman/listinfo/edk2-devel
