Re: [RFC][PATCH 2/6] mm: stack based kmap_atomic
From: KAMEZAWA Hiroyuki <hidden>
Date: 2010-08-24 07:29:37
Also in:
linux-mm, lkml
On Thu, 19 Aug 2010 22:13:19 +0200 Peter Zijlstra [off-list ref] wrote:
Keep the current interface but ignore the KM_type and use a stack based approach. Signed-off-by: Peter Zijlstra <redacted> ---
+DECLARE_PER_CPU(int, __kmap_atomic_idx);
+
+static inline int kmap_atomic_idx_push(void)
+{
+ int idx = __get_cpu_var(__kmap_atomic_idx)++;
+#ifdef CONFIG_DEBUG_HIGHMEM
+ BUG_ON(idx > KM_TYPE_NR);
+#endif
+ return idx;
+}
+
+static inline int kmap_atomic_idx_pop(void)
+{
+ int idx = --__get_cpu_var(__kmap_atomic_idx);
+#ifdef CONFIG_DEBUG_HIGHMEM
+ BUG_ON(idx < 0);
+#endif
+ return idx;
+}
+
#else /* CONFIG_HIGHMEM */I may don't understand anything... Is irq already disabled ? And Is it documented that kmap_atomic shouln't be used under NMI or something special interrupts ? Thanks, -Kame