Re: [PATCH] Invalid gimple __BB# accepted due to usage of atoi -> replace atoi with stroul in c_parser_gimple_parse_bb_spec [PR114541]

[Heiko Eißfeldt]

On 12/8/24 10:17 AM, Richard Biener wrote:

Note you picked a difficult patch - as said, the atoi fix looks OK to
me, even if
it doesn't solve downstream issues.  Can you split that out so I can
approve that
separately?

Thanks for your insights. I am a bit relieved that I don't have to
handle all of the fallout.

        PR c/114541
        * gimple-parser.cc (c_parser_gimple_parse_bb_spec):
        Use strtoul with ERANGE check instead of atoi to avoid UB
        and detect invalid __BB#. The full treatment of these invalid
        values was considered out of scope for this patch.

Signed-off-by: Heiko Eißfeldt <he...@hexco.de>

diff --git a/gcc/c/gimple-parser.cc b/gcc/c/gimple-parser.cc
index 78e85d93487..b018bb6afb6 100644
--- a/gcc/c/gimple-parser.cc
+++ b/gcc/c/gimple-parser.cc
@@ -133,11 +133,21 @@ c_parser_gimple_parse_bb_spec (tree val, int *index)
 {
   if (!startswith (IDENTIFIER_POINTER (val), "__BB"))
     return false;
-  for (const char *p = IDENTIFIER_POINTER (val) + 4; *p; ++p)
-    if (!ISDIGIT (*p))
-      return false;
-  *index = atoi (IDENTIFIER_POINTER (val) + 4);
-  return *index > 0;
+
+  const char *bb = IDENTIFIER_POINTER (val) + 4;
+  if (! ISDIGIT (*bb))
+    return false;
+
+  char *pend;
+  errno = 0;
+  const unsigned long number = strtoul (bb, &pend, 10);
+  if (errno == ERANGE
+      || *pend != '\0'
+      || number > INT_MAX)
+    return false;
+
+  *index = number;
+  return true;
 }
 
 /* See if VAL is an identifier matching __BB<num> and return <num>