On 23 February 2025 09:53:07 GMT, Ingo Molnar <mi...@kernel.org> wrote: > >* David Woodhouse <dw...@infradead.org> wrote: > >> From: David Woodhouse <d...@amazon.co.uk> >> >> There are some failure modes which lead to triple-faults in the >> relocate_kernel function, which is fairly much undebuggable for normal >> mortals. >> >> Adding a GDT in the relocate_kernel environment is step 1 towards being >> able to catch faults and do something more useful. >> >> Signed-off-by: David Woodhouse <d...@amazon.co.uk> >> --- >> arch/x86/kernel/relocate_kernel_64.S | 27 +++++++++++++++++++++++++++ >> 1 file changed, 27 insertions(+) >> >> diff --git a/arch/x86/kernel/relocate_kernel_64.S >> b/arch/x86/kernel/relocate_kernel_64.S >> index af2cd06ff318..c62f03808f18 100644 >> --- a/arch/x86/kernel/relocate_kernel_64.S >> +++ b/arch/x86/kernel/relocate_kernel_64.S >> @@ -39,6 +39,18 @@ SYM_DATA(kexec_pa_table_page, .quad 0) >> SYM_DATA(kexec_pa_swap_page, .quad 0) >> SYM_DATA_LOCAL(pa_backup_pages_map, .quad 0) >> >> +#ifdef CONFIG_KEXEC_DEBUG >> + .balign 16 >> +SYM_DATA_START_LOCAL(kexec_debug_gdt) >> + .word kexec_debug_gdt_end - kexec_debug_gdt - 1 >> + .long 0 >> + .word 0 >> + .quad 0x00cf9a000000ffff /* __KERNEL32_CS */ >> + .quad 0x00af9a000000ffff /* __KERNEL_CS */ >> + .quad 0x00cf92000000ffff /* __KERNEL_DS */ >> +SYM_DATA_END_LABEL(kexec_debug_gdt, SYM_L_LOCAL, kexec_debug_gdt_end) >> +#endif /* CONFIG_KEXEC_DEBUG */ > >Yeah, so is there any reason (other than paranoia) why the early-early >GDT and IDT shouldn't be unconditional? There's many ways for such an >approach to bitrot, it's much better to not hide it behind a >default-disabled debug option... > >Some of the other bits, like the hard-coded serial debugging >assumptions, probably need to be behind the debug option - but much of >the new debug mechanism looks safe and generic and can be always-on, >IMHO. > >This would also throw regressions back into the face of whoever manages >to introduce them, ideally. ;-) > >Thanks, > > Ingo
Makes sense to me. I was just trying to be as unobtrusive as possible. In a test branch where I was trying to fix up the objtool vs. CFI pain, I did move the IDT/GDT setup entirely into the ASM code and remove the C code which clears them (before the call into relocate_kernel() which might now trap if we remove the __nocfi hack). I never did get objtool to tolerate both clang and GCC builds though. I think even the serial output (tied as it is to earlyprintk setup) could reasonably be enabled by default too.
