If usb_gadget_giveback_request() is called in usb_ep_queue(),
this printer_write() is possible to cause spinlock recursion. So,
this patch adds spin_unlock() before calls usb_ep_queue() to avoid it.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@xxxxxxxxxxx>
---
The recursion can be caused by renesas_usbhs udc driver. I'm not sure
this patch is acceptable or not. So, I marked RFC on this patch.
Should we fix the udc driver? Or, should we fix the f_printer driver
like this patch?
drivers/usb/gadget/function/f_printer.c | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/drivers/usb/gadget/function/f_printer.c b/drivers/usb/gadget/function/f_printer.c
index 8df244f..ea0da35a 100644
--- a/drivers/usb/gadget/function/f_printer.c
+++ b/drivers/usb/gadget/function/f_printer.c
@@ -555,6 +555,7 @@ static void tx_complete(struct usb_ep *ep, struct usb_request *req)
size_t size; /* Amount of data in a TX request. */
size_t bytes_copied = 0;
struct usb_request *req;
+ int value;
DBG(dev, "printer_write trying to send %d bytes\n", (int)len);
@@ -634,7 +635,11 @@ static void tx_complete(struct usb_ep *ep, struct usb_request *req)
return -EAGAIN;
}
- if (usb_ep_queue(dev->in_ep, req, GFP_ATOMIC)) {
+ /* here, we unlock, and only unlock, to avoid deadlock. */
+ spin_unlock(&dev->lock);
+ value = usb_ep_queue(dev->in_ep, req, GFP_ATOMIC);
+ spin_lock(&dev->lock);
+ if (value) {
list_add(&req->list, &dev->tx_reqs);
spin_unlock_irqrestore(&dev->lock, flags);
mutex_unlock(&dev->lock_printer_io);
--
1.9.1
