Use EPERM instead of EIO when try setting irq affinity fails

Kernel commit eb29369fa543e ("genirq/proc: Change the return value for
set affinity permission error") changed the error number from EIO to
EPERM when a specific irq failed to set affinity from userspace.

In addition, users have complained about EIO is misleading when
setting affinity fails, however it just meaning "user cannot set
affinity", so EPERM is better for that.

This patch will follow the kernel change, to unify the behaviour of
irqbalance from the kernels which have not integrated the kernel
commit mentioned above.

Signed-off-by: Tao Liu <ltao@redhat.com>

activate.c

@@ -95,6 +95,8 @@ static void activate_mapping(struct irq_info *info, void *data __attribute__((un
 	info->moved = 0; /*migration is done*/
 	return;
 error:
+	/* Use EPERM as the explaination for EIO */
+	errsave = (errsave == EIO) ? EPERM : errsave;
 	log(TO_ALL, LOG_WARNING,
 		"Cannot change IRQ %i affinity: %s\n",
 		info->irq, strerror(errsave));