diff mbox

Fix PR56344

Message ID 20130226182733.GG25197@redhat.com
State New
Headers show

Commit Message

Marek Polacek Feb. 26, 2013, 6:27 p.m. UTC 
This "fixes" PR56344 by prohibiting passing arguments by value
of size >= 2^30 bytes.  Probably no sane programmer would want to 
do that, but it's better to issue an error than to segfault.
This would be a good opportunity to use __builtin_expect, but
we don't use that much in the codebase...

Regtested/bootstrapped on x86_64-linux, ok for trunk?

Do we need a testcase for this (compiling it is quite slow)?

2013-02-26  Marek Polacek  <polacek@redhat.com>

	PR middle-end/56344
	* calls.c (expand_call): Disallow passing huge arguments
	by value.


	Marek

Comments

Joseph Myers Feb. 26, 2013, 11:17 p.m. UTC | #1 
On Tue, 26 Feb 2013, Marek Polacek wrote:

> +	      /* We don't allow passing huge (> 2^30 B) arguments
> +	         by value.  It would cause an overflow later on.  */
> +	      if (adjusted_args_size.constant >= (1 << 30))
> +	        {
> +	          error ("passing too large argument on stack");
> +		  continue;

This should be sorry () not error (), as a compiler limitation rather than 
a defect in the user's program.  (And is input_location set to something 
useful here so the diagnostic points to the argument in question rather 
than e.g. to the end of the function containing the problem call?)
diff mbox

Patch

--- gcc/calls.c.mp	2013-02-26 17:04:33.159555349 +0100
+++ gcc/calls.c	2013-02-26 18:50:54.864084545 +0100
@@ -3037,6 +3037,14 @@  expand_call (tree exp, rtx target, int i
 	    {
 	      rtx before_arg = get_last_insn ();
 
+	      /* We don't allow passing huge (> 2^30 B) arguments
+	         by value.  It would cause an overflow later on.  */
+	      if (adjusted_args_size.constant >= (1 << 30))
+	        {
+	          error ("passing too large argument on stack");
+		  continue;
+		}
+
 	      if (store_one_arg (&args[i], argblock, flags,
 				 adjusted_args_size.var != 0,
 				 reg_parm_stack_space)