x86: Introduce REX prefix helper for kprobes

Fold some small ifdefs into a helper function.

Signed-off-by Harvey Harrison <[email protected]>

---
Masami, Ingo, I had this left in some unsent kprobes unification work. Depends on your tastes, but does reduce ifdefs and is a bit better about self-documenting the REX prefix on X86_64.

If I find places that could also use this I'll try to find a suitable header any stick a static inline there instead. Otherwise static to kprobes.c is probably more appropriate for now.

 arch/x86/kernel/kprobes.c |   27 +++++++++++++++++++--------
 1 files changed, 19 insertions(+), 8 deletions(-)

arch/x86/kernel/kprobes.c +19 -8

index 4e33329..b1804e4 100644
@@ -171,6 +171,19 @@ static void __kprobes set_jmp_op(void *from, void *to)
 }
 
 /*
+ * Check for the REX prefix which can only exist on X86_64
+ * X86_32 always returns 0
+ */
+static int __kprobes is_REX_prefix(kprobe_opcode_t *insn)
+{
+#ifdef CONFIG_X86_64
+	if ((*insn & 0xf0) == 0x40)
+		return 1;
+#endif
+	return 0;
+}
+
+/*
  * Returns non-zero if opcode is boostable.
  * RIP relative instructions are adjusted at copying time in 64 bits mode
  */

@@ -239,14 +252,14 @@ static int __kprobes is_IF_modifier(kprobe_opcode_t *insn)
 	case 0x9d:		/* popf/popfd */
 		return 1;
 	}
-#ifdef CONFIG_X86_64
+
 	/*
-	 * on 64 bit x86, 0x40-0x4f are prefixes so we need to look
+	 * on X86_64, 0x40-0x4f are REX prefixes so we need to look
 	 * at the next byte instead.. but of course not recurse infinitely
 	 */
-	if (*insn  >= 0x40 && *insn <= 0x4f)
+	if (is_REX_prefix(insn))
 		return is_IF_modifier(++insn);
-#endif
+
 	return 0;
 }
 

@@ -284,7 +297,7 @@ static void __kprobes fix_riprel(struct kprobe *p)
 	}
 
 	/* Skip REX instruction prefix.  */
-	if ((*insn & 0xf0) == 0x40)
+	if (is_REX_prefix(insn))
 		++insn;
 
 	if (*insn == 0x0f) {

@@ -748,11 +761,9 @@ static void __kprobes resume_execution(struct kprobe *p,
 	unsigned long orig_ip = (unsigned long)p->addr;
 	kprobe_opcode_t *insn = p->ainsn.insn;
 
-#ifdef CONFIG_X86_64
 	/*skip the REX prefix*/
-	if (*insn >= 0x40 && *insn <= 0x4f)
+	if (is_REX_prefix(insn))
 		insn++;
-#endif
 
 	regs->flags &= ~X86_EFLAGS_TF;
 	switch (*insn) {

Hi Harvey,

Quoted 7 lines · Harvey Harrison

Fold some small ifdefs into a helper function.

Signed-off-by Harvey Harrison <[email protected]>

---
Masami, Ingo, I had this left in some unsent kprobes unification work. Depends on your tastes, but does reduce ifdefs and is a bit better about self-documenting the REX prefix on X86_64.

Basically, I think it is good idea.
Could you use a macro same as the stack_addr() macro, like as below?

#defile is_REX_prefix(insn) ((insn & 0xf0) == 0x40))

This is just a bit checker, so I think a macro is better to do that.

Quoted 6 lines

Could you use a macro same as the stack_addr() macro, like as below?

#defile is_REX_prefix(insn) ((insn & 0xf0) == 0x40))

This is just a bit checker, so I think a macro is better to do that.

Why is a macro better than an inline, and why the odd mIXed case?

-hpa

(in case of CONFIG_X86_64)
static inline int is_rex_prefix(int op)

About the name, I just used the previous inline function name.

Quoted 1 line

Why is a macro better than an inline, and why the odd mIXed case?

I thought we can use macro because it just check a bit mask. And if we use this as a macro, it will be defined in #ifdef block at the top of kprobes.c. It is simple in this case. I know the inline is better than the macro, it can check the type of arguments. If you would like to use inline, how about this?

Quoted 14 lines · Masami Hiramatsu

Hi Harvey,

Quoted 7 lines · Harvey Harrison

Fold some small ifdefs into a helper function.

Signed-off-by Harvey Harrison <[email protected]>

---
Masami, Ingo, I had this left in some unsent kprobes unification work. Depends on your tastes, but does reduce ifdefs and is a bit better about self-documenting the REX prefix on X86_64.

Basically, I think it is good idea.
Could you use a macro same as the stack_addr() macro, like as below?

Quoted 4 lines

#defile is_REX_prefix(insn) ((insn & 0xf0) == 0x40))

This is just a bit checker, so I think a macro is better to do that.

Why is a macro better than an inline, and why the odd mIXed case?

Fold some small ifdefs into a helper function.

Signed-off-by Harvey Harrison <[email protected]>