On 2.3.2009 15:34, Jiri Slaby wrote:

quoted

+static int at32_wm97xx_suspend(struct platform_device *pdev,
pm_message_t msg)
+{
+ struct at32_wm97xx *at32_wm97xx = platform_get_drvdata(pdev);
+
+ ac97c_writel(at32_wm97xx, IDR, AC97C_INT_CBEVT);
+ disable_irq(at32_wm97xx->gpio_irq);

Here, you should call disable_irq_sync() (that's what I meant by _sync
in the last post).

Ah, it's in the opposite way. This one is fine, the patch 1/2 should use 
disable_irq_nosync(), if I understand the code correctly.