Le 13/10/2021 à 09:09, Kees Cook a écrit :

On Mon, Oct 11, 2021 at 05:25:36PM +0200, Christophe Leroy wrote:

quoted

Behind a location, lkdtm_EXEC_RODATA() executes a real function,
not a copy of do_nothing().

So do it directly instead of using execute_location().

I don't understand this. Why does the next patch not fix this?

Well, probably it would, but it looked incorrect in my mind.

lkdtm_rodata_do_nothing() is a function which has its own function 
descriptor, it is not a copy of do_nothing().

If we use execute_location() as modified by next patch, then we will 
execute it using the function descriptor of do_nothing(). Allthough it 
most likely works (at least on powerpc as it uses the same TOC) it looks 
odd to me to do so.

Am I missing something ?

Christophe

-Kees

quoted

And fix displayed addresses by dereferencing the function descriptors.

Signed-off-by: Christophe Leroy <redacted>
---
  drivers/misc/lkdtm/perms.c | 9 ++++++++-
  1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/misc/lkdtm/perms.c b/drivers/misc/lkdtm/perms.c
index 442d60ed25ef..da16564e1ecd 100644
--- a/drivers/misc/lkdtm/perms.c
+++ b/drivers/misc/lkdtm/perms.c

@@ -153,7 +153,14 @@ void lkdtm_EXEC_VMALLOC(void)
  
  void lkdtm_EXEC_RODATA(void)
  {
-	execute_location(lkdtm_rodata_do_nothing, CODE_AS_IS);
+	pr_info("attempting ok execution at %px\n",
+		dereference_symbol_descriptor(do_nothing));
+	do_nothing();
+
+	pr_info("attempting bad execution at %px\n",
+		dereference_symbol_descriptor(lkdtm_rodata_do_nothing));
+	lkdtm_rodata_do_nothing();
+	pr_err("FAIL: func returned\n");
  }
  
  void lkdtm_EXEC_USERSPACE(void)

-- 
2.31.1