x86/memtest: Shorten time for tests

By just reversing the order memtest is using the test patterns,
an additional round to zero the memory is not necessary.

This might save up to a second or even more for setups which are
doing tests on every boot.

Signed-off-by: Alexander Holler <holler@ahsoftware.de>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1361029097-8308-1-git-send-email-holler@ahsoftware.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/mm/memtest.c

@@ -9,6 +9,7 @@
 #include <linux/memblock.h>
 
 static u64 patterns[] __initdata = {
+	/* The first entry has to be 0 to leave memtest with zeroed memory */
 	0,
 	0xffffffffffffffffULL,
 	0x5555555555555555ULL,
@@ -110,15 +111,8 @@ void __init early_memtest(unsigned long start, unsigned long end)
 		return;
 
 	printk(KERN_INFO "early_memtest: # of tests: %d\n", memtest_pattern);
-	for (i = 0; i < memtest_pattern; i++) {
+	for (i = memtest_pattern-1; i < UINT_MAX; --i) {
 		idx = i % ARRAY_SIZE(patterns);
 		do_one_pass(patterns[idx], start, end);
 	}
-
-	if (idx > 0) {
-		printk(KERN_INFO "early_memtest: wipe out "
-		       "test pattern from memory\n");
-		/* additional test with pattern 0 will do this */
-		do_one_pass(0, start, end);
-	}
 }