Re: [RFC v4 4/4] xhci: rework command timeout and cancellation,

[Mathias Nyman]

On 03/21/2014 11:35 AM, Mathias Nyman wrote:

> +void xhci_handle_command_timeout(unsigned long data)
> +{
> +	struct xhci_hcd *xhci;
> +	int ret;
> +	unsigned long flags;
> +	xhci = (struct xhci_hcd *) data;
> +	/* mark this command to be cancelled */
> +	spin_lock_irqsave(&xhci->lock, flags);
> +	if (xhci->current_cmd)
> +		xhci->current_cmd->status = COMP_CMD_ABORT;
> +	spin_unlock_irqrestore(&xhci->lock, flags);
> +
> +	ret = xhci_abort_cmd_ring(xhci);
> +
> +	if (ret) {
> +		xhci_err(xhci, "Abort command ring failed\n");
> +		if (unlikely(ret == -ESHUTDOWN)) {
> +			xhci_cleanup_command_queue(xhci);
> +			usb_hc_died(xhci_to_hcd(xhci)->primary_hcd);
> +			xhci_dbg(xhci, "xHCI host controller is dead.\n");
> +		}
> +	}
> +	return;
> +}
> +

After some more testing and fault injection it turns out that 
xhci_abort_cmd_ring() returns 0 if command ring is already stopped. In 
this case the command submitter will still be left hanging, and khubd 
wait forever.

So one more round to fix this is still needed.

-Mathias

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/