vt6655: Fix a possible sleep-in-atomic bug in vt6655_suspend
authorJia-Ju Bai <baijiaju1990@163.com>
Mon, 9 Oct 2017 08:45:55 +0000 (16:45 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 18 Oct 2017 14:42:28 +0000 (16:42 +0200)
The driver may sleep under a spinlock, and the function call path is:
vt6655_suspend (acquire the spinlock)
  pci_set_power_state
    __pci_start_power_transition (drivers/pci/pci.c)
      msleep --> may sleep

To fix it, pci_set_power_state is called without having a spinlock.

This bug is found by my static analysis tool and my code review.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/vt6655/device_main.c

index 9fcf2e223f719d1ea12d27858cefcb83008f0bc4..1123b4f1e1d66987981687445129347c119fad14 100644 (file)
@@ -1693,10 +1693,11 @@ static int vt6655_suspend(struct pci_dev *pcid, pm_message_t state)
        MACbShutdown(priv);
 
        pci_disable_device(pcid);
-       pci_set_power_state(pcid, pci_choose_state(pcid, state));
 
        spin_unlock_irqrestore(&priv->lock, flags);
 
+       pci_set_power_state(pcid, pci_choose_state(pcid, state));
+
        return 0;
 }