Do we need more robust archeticture protection

[laokz]

I noticed that there are many places using relocation types without #ifdef architecture protection, such as in create-diff-object.c:

	return (rela_toc && rela_toc->sym->type == STT_FUNC &&
		!rela_toc->sym->parent &&
		rela_toc->addend == (int)rela_toc->sym->sym.st_value &&
		(rela->type == R_X86_64_32S ||
		rela->type == R_PPC64_TOC16_HA ||
		rela->type == R_PPC64_TOC16_LO_DS));

All architectures relocation types use same number space. R_X86_64_32S value is 11. This value for PPC is R_PPC_REL14, for arm is R_ARM_THM_PC8. Could this break the code?

[jpoimboe]

Good point! We definitely need to fix that.

[joe-lawrence]

Just a heads up, conditionally building code around #ifdef <arch> works against (eventually) cross-building/testing kpatches. I would think the correct thing to do would be the include the source/target along with the relocation enum check.

[github-actions]

This issue has been open for 30 days with no activity and no assignee. It will be closed in 7 days unless a comment is added.

[github-actions]

This issue was closed because it was inactive for 7 days after being marked stale.