Bluetooth USB: fix kernel panic during suspend - 2.6.24
PATCH 1/1 Bluetooth USB: fix kernel panic during suspend - 2.6.24
The kernel panics when hci_usb_tx_complete() calls _urb_unlink() on an _urb which has been previously been removed from a list. This occurs during suspend while audio is being streamed to a bluetooth headset.
The panic occurs because hci_usb_suspend() dequeues _urb and then calls usb_kill_urb() - instead it should put _urb on the killed list first before killing the urb.
Also added a spin_lock around the list_add operation.
--- drivers/bluetooth/hci_usb.c | 4 +++- 1 files changed, 3 insertions(+), 1 deletions(-)
drivers/bluetooth/hci_usb.c +0 -0
@@ -1029,8 +1029,10 @@ static int hci_usb_suspend(struct usb_interface
*intf, pm_message_t message)
while ((_urb = _urb_dequeue(q))) {
/* reset queue since _urb_dequeue sets it to NULL */ _urb->queue = q;
- usb_kill_urb(&_urb->urb);
+ spin_lock_irqsave(&q->lock, flags);
list_add(&_urb->list, &killed);
+ spin_unlock_irqrestore(&q->lock, flags);
+ usb_kill_urb(&_urb->urb);
}
spin_lock_irqsave(&q->lock, flags);
Quoted 6 lines · Colin Ian King
The kernel panics when
hci_usb_tx_complete()calls_urb_unlink()on an_urbwhich has been previously been removed from a list. This occurs during suspend while audio is being streamed to a bluetooth headset.
Quoted 7 lines
The panic occurs because
hci_usb_suspend()dequeues_urband then callsusb_kill_urb()- instead it should put_urbon the killed list first before killing the urb.Also added a
spin_lockaround thelist_addoperation.Signed-off-by Colin Ian King <[email protected]>
Is this still an issue on a supported-by-the-community kernel (.27, .28, .29)?
thanks,
greg k-h
Marcel Holtmann
maintainerHi Greg,
Quoted 9 lines
Quoted 9 lines
The kernel panics when
hci_usb_tx_complete()calls_urb_unlink()on an_urbwhich has been previously been removed from a list. This occurs during suspend while audio is being streamed to a bluetooth headset.The panic occurs because
hci_usb_suspend()dequeues_urband then callsusb_kill_urb()- instead it should put_urbon the killed list first before killing the urb.
Quoted 6 lines
Quoted 3 lines
Also added a
spin_lockaround thelist_addoperation.Signed-off-by Colin Ian King <[email protected]>Is this still an issue on a supported-by-the-community kernel (.27, .28, .29)?
starting with 2.6.29 the hci_usb driver has been replaced by btusb.
Regards
Marcel