soc: fsl: dpio: use a cpumask to identify which cpus are unused

The current implementation of the dpio driver uses a static next_cpu
variable to keep track of the index of the next cpu available. This
approach does not handle well unbinding and binding dpio devices in a
random order. For example, unbinding a dpio and then binding it again
with the driver, will generate the below error:

$ echo dpio.5 > /sys/bus/fsl-mc/drivers/fsl_mc_dpio/unbind
$ echo dpio.5 > /sys/bus/fsl-mc/drivers/fsl_mc_dpio/bind
[  103.946380] fsl_mc_dpio dpio.5: probe failed. Number of DPIOs exceeds
NR_CPUS.
[  103.955157] fsl_mc_dpio dpio.5: fsl_mc_driver_probe failed: -34
-bash: echo: write error: No such device

Fix this error by keeping a global cpumask of unused cpus that will be
updated at every dpaa2_dpio_[probe,remove].

Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Li Yang <leoyang.li@nxp.com>

drivers/soc/fsl/dpio/dpio-service.c

@@ -214,6 +214,19 @@ done:
 	return IRQ_HANDLED;
 }
 
+/**
+ * dpaa2_io_get_cpu() - get the cpu associated with a given DPIO object
+ *
+ * @d: the given DPIO object.
+ *
+ * Return the cpu associated with the DPIO object
+ */
+int dpaa2_io_get_cpu(struct dpaa2_io *d)
+{
+	return d->dpio_desc.cpu;
+}
+EXPORT_SYMBOL(dpaa2_io_get_cpu);
+
 /**
  * dpaa2_io_service_register() - Prepare for servicing of FQDAN or CDAN
  *                               notifications on the given DPIO service.