[PATCH] mm/migrate: Mark unmap_and_move() "noinline" to avoid ICE in gcc 4.7.3

Andrew Morton [off-list ref] writes:

On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier [off-list ref] wrote:

quoted

quoted

-static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page,
-			unsigned long private, struct page *page, int force,
-			enum migrate_mode mode)
+static noinline int unmap_and_move(new_page_t get_new_page,
+				   free_page_t put_new_page,
+				   unsigned long private, struct page *page,
+				   int force, enum migrate_mode mode)
 {
 	int rc = 0;
 	int *result = NULL;

Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of
the parameters on the stack, which is not going to help performance
either (not that this would be useful on 32bit ARM anyway...).

Any chance you could make this dependent on some compiler detection
mechanism?

With my arm compiler (gcc-4.4.4) the patch makes no difference -
unmap_and_move() isn't being inlined anyway.

How does this look?

Kevin, could you please retest?  I might have fat-fingered something...

Your patch on top of Geert's still compiles fine for me with gcc-4.7.3.
However, I'm not sure how specific we can be on the versions.  

/me goes to test a few more compilers...   OK...

ICE: 4.7.1, 4.7.3, 4.8.3
OK: 4.6.3, 4.9.2, 4.9.3

The diff below[2] on top of yours compiles fine here and at least covers
the compilers I *know* to trigger the ICE.

Kevin


[1]

diff --git a/mm/migrate.c b/mm/migrate.c
index 25fd7f6291de..6e15ae3248e0 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c

@@ -901,10 +901,10 @@ out:
 }
 
 /*
- * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move().  Work around
+ * gcc 4.7 and 4.8 on arm gets an ICE when inlining unmap_and_move().  Work around
  * it.
  */
-#if GCC_VERSION == 40703 && defined(CONFIG_ARM)
+#if (GCC_VERSION >= 40700 && GCC_VERSION < 40900) && defined(CONFIG_ARM)
 #define ICE_noinline noinline
 #else
 #define ICE_noinline