Re: [PATCH] Improve dead code elimination with -fsanitize=address (PR84307)

[Paolo Bonzini <bonzini@gnu.org>]

On 09/02/2018 17:40, Richard Biener wrote:
> On February 9, 2018 5:08:24 PM GMT+01:00, Paolo Bonzini <bonzini@gnu.org> wrote:
>> Another possibility which I considered but did not implement is to mark
>> the UNPOISON calls so that they do not cause the parameter to escape.
> 
> I'd do this, thus assign proper fnspec attributes to the asan functions. 

Hmm, actually that might be as simple as fixing a typo:

diff --git a/gcc/internal-fn.def b/gcc/internal-fn.def
index 5970d0e..15d6151 100644
--- a/gcc/internal-fn.def
+++ b/gcc/internal-fn.def
@@ -255,8 +255,8 @@ DEF_INTERNAL_FN (UBSAN_PTR, ECF_LEAF | ECF_NOTHROW, ".R.")
 DEF_INTERNAL_FN (UBSAN_OBJECT_SIZE, ECF_LEAF | ECF_NOTHROW, NULL)
 DEF_INTERNAL_FN (ABNORMAL_DISPATCHER, ECF_NORETURN, NULL)
 DEF_INTERNAL_FN (BUILTIN_EXPECT, ECF_CONST | ECF_LEAF | ECF_NOTHROW, NULL)
-DEF_INTERNAL_FN (ASAN_CHECK, ECF_TM_PURE | ECF_LEAF | ECF_NOTHROW, ".R...")
-DEF_INTERNAL_FN (ASAN_MARK, ECF_LEAF | ECF_NOTHROW, ".R..")
+DEF_INTERNAL_FN (ASAN_CHECK, ECF_TM_PURE | ECF_LEAF | ECF_NOTHROW, "..R..")
+DEF_INTERNAL_FN (ASAN_MARK, ECF_LEAF | ECF_NOTHROW, "..W.")
 DEF_INTERNAL_FN (ASAN_POISON, ECF_LEAF | ECF_NOTHROW | ECF_NOVOPS, NULL)
 DEF_INTERNAL_FN (ASAN_POISON_USE, ECF_LEAF | ECF_NOTHROW | ECF_NOVOPS, NULL)
 DEF_INTERNAL_FN (ADD_OVERFLOW, ECF_CONST | ECF_LEAF | ECF_NOTHROW, NULL)

which indeed fixes the testcase and seems not to break asan.exp.
'W' is needed to avoid breaking the pr78541.c testcase, and I think it
makes sense since ASAN_MARK is "writing" the state of the object
(in the test case FRE moves a dereference across a poisoning).

I'll look at it next week.  Someone maybe should take a look at ubsan
fnspecs too.

Paolo