diff mbox

Fix ICE on very large functions in ira-costs.c (PR middle-end/79399)

Message ID 20170207211722.GV1849@tucnak
State New
Headers show

Commit Message

Jakub Jelinek Feb. 7, 2017, 9:17 p.m. UTC 
Hi!

As mentioned in the PR, on a huge function init_costs attempts to allocate
124 * 17380077 bytes, 124 is max_struct_costs_size and 17380077
is cost_elements_num.  The problem is that the multiplication is done in
int type, so it overflows and we attempt to allocate 0xffffffff8074aacc
bytes.  Other similar multiplications in ira-costs.c use size_t as type
of at least one operand, so it works properly.  The following patch
fixes it by making sure struct_costs_size and max_struct_costs_size have
size_t type.  The first hunk shouldn't result in more memory being used,
on 32-bit host it is not any change and the field is followed by
a pointer, so on 64-bit hosts there has been padding that we now use.

Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?

2017-02-07  Jakub Jelinek  <jakub@redhat.com>
	    Richard Biener  <rguenther@suse.de>

	PR middle-end/79399
	* ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
	type from int to size_t.
	* ira-costs.c (struct_costs_size): Change type from int to size_t.


	Jakub

Comments

Vladimir Makarov Feb. 7, 2017, 9:43 p.m. UTC | #1 
On 02/07/2017 04:17 PM, Jakub Jelinek wrote:
> Hi!
>
> As mentioned in the PR, on a huge function init_costs attempts to allocate
> 124 * 17380077 bytes, 124 is max_struct_costs_size and 17380077
> is cost_elements_num.  The problem is that the multiplication is done in
> int type, so it overflows and we attempt to allocate 0xffffffff8074aacc
> bytes.  Other similar multiplications in ira-costs.c use size_t as type
> of at least one operand, so it works properly.  The following patch
> fixes it by making sure struct_costs_size and max_struct_costs_size have
> size_t type.  The first hunk shouldn't result in more memory being used,
> on 32-bit host it is not any change and the field is followed by
> a pointer, so on 64-bit hosts there has been padding that we now use.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
>
Sure, Jakub.
diff mbox

Patch

--- gcc/ira-int.h.jj	2017-01-01 12:45:39.000000000 +0100
+++ gcc/ira-int.h	2017-02-07 09:29:32.694102809 +0100
@@ -782,7 +782,7 @@  struct target_ira_int {
 
   /* Initialized once.  It is a maximal possible size of the allocated
      struct costs.  */
-  int x_max_struct_costs_size;
+  size_t x_max_struct_costs_size;
 
   /* Allocated and initialized once, and used to initialize cost values
      for each insn.  */
--- gcc/ira-costs.c.jj	2017-01-16 12:28:35.000000000 +0100
+++ gcc/ira-costs.c	2017-02-07 11:44:26.534868418 +0100
@@ -74,7 +74,7 @@  static struct costs *costs;
 static struct costs *total_allocno_costs;
 
 /* It is the current size of struct costs.  */
-static int struct_costs_size;
+static size_t struct_costs_size;
 
 /* Return pointer to structure containing costs of allocno or pseudo
    with given NUM in array ARR.  */