Re: New Software Suspend Patch for testing.

[Pavel Machek <pavel@ucw.cz>]

Hi!

> The changes from the last version are not huge; perhaps most significant
> is that the dynamically allocated bitmaps (in place of pageflags) are
> implemented. This won't mean anything to most people, but Andrew
> Morton,

@@ -536,6 +519,9 @@
 	struct page *page;
 	int i;
 	int cold;
+#if CONFIG_SOFTWARE_SUSPEND
+	static unsigned int loopcount;
+#endif	
 
 	if (wait)
 		might_sleep();
@@ -589,7 +575,23 @@
 
 	/* here we're in the low on memory slow path */
 
+#if CONFIG_SOFTWARE_SUSPEND
+	loopcount=0;
+#endif
 rebalance:
+#ifdef CONFIG_SOFTWARE_SUSPEND
+	if(gfp_mask & __GFP_FAST) {
+/* when using memeat, we ask for all pages that are really free.
+   5 calls to reschedule should be sufficient to recall all of them since
+   when a page can be found, it is after only one reschedule.
+   Actually I consider this as a bug of alloc_pages, since allocating a
+   page should not hang in an endless loop when it is clear that no
+   memory is available (cbd) */	  
+		loopcount++;
+		if(loopcount > 5)
+			return NULL;
+	}
+#endif
 	if ((current->flags & (PF_MEMALLOC | PF_MEMDIE)) && !in_interrupt()) {
 		/* go through the zonelist yet again, ignoring mins */
 		for (i = 0; zones[i] != NULL; i++) {

Again this is *really really* ugly. Why do you need it? 2.5. has
special function for freeing memory.

							Pavel
-- 
When do you have heart between your knees?