x86: c_p_a() fix: reorder TLB / cache flushes to follow Intel recommendation

Intel recommends to first flush the TLBs and then the caches
on caching attribute changes. c_p_a() previously did it the
other way round. Reorder that.

The procedure is still not fully compliant to the Intel documentation
because Intel recommends a all CPU synchronization step between
the TLB flushes and the cache flushes.

However on all new Intel CPUs this is now meaningless anyways
because they support Self-Snoop and can skip the cache flush
step anyway.

[ mingo@elte.hu: decoupled from clflush and ported it to x86.git ]
Signed-off-by: NAndi Kleen <ak@suse.de>
Acked-by: NJan Beulich <jbeulich@novell.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>

arch/x86/mm/pageattr_32.c

@@ -87,6 +87,12 @@ static void flush_kernel_map(void *arg)
 	struct list_head *lh = (struct list_head *)arg;
 	struct page *p;
 
+	/*
+	 * Flush all to work around Errata in early athlons regarding
+	 * large page flushing.
+	 */
+	__flush_tlb_all();
+
 	/* High level code is not ready for clflush yet */
 	if (0 && cpu_has_clflush) {
 		list_for_each_entry(p, lh, lru)
@@ -95,12 +101,6 @@ static void flush_kernel_map(void *arg)
 		if (boot_cpu_data.x86_model >= 4)
 			wbinvd();
 	}
-
-	/*
-	 * Flush all to work around Errata in early athlons regarding
-	 * large page flushing.
-	 */
-	__flush_tlb_all();
 }
 
 static void set_pmd_pte(pte_t *kpte, unsigned long address, pte_t pte)

arch/x86/mm/pageattr_64.c

@@ -82,6 +82,8 @@ static void flush_kernel_map(void *arg)
 	struct list_head *l = (struct list_head *)arg;
 	struct page *pg;
 
+	__flush_tlb_all();
+
 	/* When clflush is available always use it because it is
 	   much cheaper than WBINVD. */
 	/* clflush is still broken. Disable for now. */
@@ -94,7 +96,6 @@ static void flush_kernel_map(void *arg)
 			clflush_cache_range(addr, PAGE_SIZE);
 		}
 	}
-	__flush_tlb_all();
 }
 
 static inline void flush_map(struct list_head *l)