Christophe Leroy wrote:

Le 29/08/2022 à 07:52, Sathvika Vasireddy a écrit :

quoted

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 4c466acdc70d..dc05cd23c233 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig

@@ -237,6 +237,7 @@ config PPC
  	select HAVE_MOD_ARCH_SPECIFIC
  	select HAVE_NMI				if PERF_EVENTS || (PPC64 && PPC_BOOK3S)
  	select HAVE_OPTPROBES
+	select HAVE_OBJTOOL			if PPC32 || MPROFILE_KERNEL

Why restrict it to MPROFILE_KERNEL ? In your RFC it was for all PPC64.

Recent discussion on the list shows new problem with recordmcount, see 
https://lore.kernel.org/all/MW5PR84MB184250EA1CAE04497C1E7CE9AB769@MW5PR84MB1842.NAMPRD84.PROD.OUTLOOK.COM/ (local)

Those ones are with ppc64 big endian, so objtool would be welcome here 
as well.

I don't think adding support for objtool on ppc64 elfv1 is a good idea. 
While it might solve the immediate issue with recordmcount, I worry that 
the function descriptors and dot symbols in ppc64 elfv1 might throw up 
issues in the future causing maintenance overhead.


- Naveen