On Wed, Jul 17, 2024 at 03:02:23PM +0530, Balasubrmanian, Vignesh wrote:

On 7/13/2024 4:12 PM, Thomas Gleixner wrote:

quoted

Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.


On Fri, Jul 12 2024 at 15:16, Vignesh Balasubramanian wrote:

quoted

diff --git a/arch/x86/include/asm/elf.h b/arch/x86/include/asm/elf.h
index 1fb83d477..cad37090b 100644
--- a/arch/x86/include/asm/elf.h
+++ b/arch/x86/include/asm/elf.h

@@ -13,6 +13,15 @@
  #include <asm/auxvec.h>
  #include <asm/fsgsbase.h>

+struct xfeat_component {
+     u32 type;
+     u32 size;
+     u32 offset;
+     u32 flags;
+} __packed;
+
+_Static_assert(sizeof(struct xfeat_component)%4 == 0, "xfeat_component is not aligned");

This struct is only used in xstate.c and asm/elf.h is not a UAPI
header. So what's the point of declaring it in the header instead of
xtsate.c?

If this needs to provided for user space consumption, then it want's to
be in a UAPI header, no?

Our initial idea is to pass the "struct xfeat_component" through "glibc".
is "include/uapi/linux/elf.h" proper header to add this ?

I'd rather not put arch-specific things in the main UAPI elf.h file
unless there is a good reason.

I couldn't see any proper header inside "arch/x86/include/uapi/asm/".

Other architectures have arch/*/include/uapi/asm/elf.h, so it may be
time to add one for x86 too. For this to be UAPI, I would want to see
more explicit namespacing, e.g. struct x86_xfeat_component, etc.

-Kees

-- 
Kees Cook