hi alan:
Great thanks for your such careful review.
On 2012年03月29日 03:13, Alan Stern wrote:
On Wed, 28 Mar 2012, Lan Tianyu wrote:
Move child's pointer to the struct usb_hub_port since the child device
is directly associated with the port. Provide usb_get_hub_child_device()
to get child's pointer.
--- a/drivers/usb/core/devices.c
+++ b/drivers/usb/core/devices.c
@@ -590,7 +590,9 @@ static ssize_t usb_device_dump(char __user **buffer, size_t *nbytes,
/* Now look at all of this device's children. */
for (chix = 0; chix< usbdev->maxchild; chix++) {
- struct usb_device *childdev = usbdev->children[chix];
+ struct usb_device *childdev;
+ if (usb_get_hub_child_device(usbdev, chix + 1,&childdev)< 0)
+ continue;
Whitespace issue: The blank line should come before the new executable
statement, not after it.
@@ -1475,11 +1477,12 @@ bool usb_device_is_owned(struct usb_device *udev)
static void recursively_mark_NOTATTACHED(struct usb_device *udev)
{
+ struct usb_hub *hub = hdev_to_hub(udev);
I'm not sure this is good. You are calling hdev_to_hub without knowing
beforehand that udev really is a hub -- in fact, most of the time udev
won't be a hub. The routine wasn't meant to be used that way. Have
you checked that hdev_to_hub won't crash when the argument isn't a hub?
What happens if the active configuration has no interfaces?
About this, I have an idea to add a check of hdev->maxchild. I have
verified that hdev->maxchild only is set in the hub_configure() and
hub_disconnect(). So we can identify whether the udev is a usb hub or
not through hdev->maxchild. Does this make sense?
diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
index c70c170..d29a5dc 100644
--- a/drivers/usb/core/hub.c
+++ b/drivers/usb/core/hub.c
@@ -177,7 +177,7 @@ static inline char *portspeed(struct usb_hub *hub, int
portstatus)
/* Note that hdev or one of its children must be locked! */
static struct usb_hub *hdev_to_hub(struct usb_device *hdev)
{
- if (!hdev || !hdev->actconfig)
+ if (!hdev || !hdev->actconfig || hdev->maxchild == 0)
return NULL;
return usb_get_intfdata(hdev->actconfig->interface[0]);
}
@@ -4147,3 +4151,15 @@ void usb_queue_reset_device(struct usb_interface *iface)
schedule_work(&iface->reset_ws);
}
EXPORT_SYMBOL_GPL(usb_queue_reset_device);
+
+int usb_get_hub_child_device(struct usb_device *hdev, int port1,
+ struct usb_device **child)
+{
+ struct usb_hub *hub = hdev_to_hub(hdev);
+
+ if (port1> hdev->maxchild || port1< 1)
+ return -EINVAL;
+ *child = hub->port_data[port1 - 1].child;
+ return 0;
+}
This new routine is more complicated that it needs to be. You don't
have to return an error code; none of the callers make use of it.
Just return the child pointer, and return NULL if there's an error.
Alan Stern
--
Best Regards
Tianyu Lan
linux kernel enabling team
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
