drivers: hv: Increase the number of VCPUs supported in the guest

The current code arbirarily limited the number of CPUs the guest could have.
Change that so that we can support the maximum number of CPUs the guest can
support. While we use NR_CPUS to size the per-cpu state all we are allocating
based on NR_CPUS are the  pointers to per-cpu state that will be allocatted in
the context of the initializing CPU. This patch triggers a checkpatch warning
for the usage of NR_CPU and since all we are allocating a couple of pointers
per CPU, it should be ok.

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/hv/hv.c

@@ -155,9 +155,9 @@ int hv_init(void)
 	union hv_x64_msr_hypercall_contents hypercall_msr;
 	void *virtaddr = NULL;
 
-	memset(hv_context.synic_event_page, 0, sizeof(void *) * MAX_NUM_CPUS);
+	memset(hv_context.synic_event_page, 0, sizeof(void *) * NR_CPUS);
 	memset(hv_context.synic_message_page, 0,
-	       sizeof(void *) * MAX_NUM_CPUS);
+	       sizeof(void *) * NR_CPUS);
 
 	if (!query_hypervisor_presence())
 		goto cleanup;

drivers/hv/hyperv_vmbus.h

@@ -457,7 +457,6 @@ static const uuid_le VMBUS_SERVICE_ID = {
 	},
 };
 
-#define MAX_NUM_CPUS	32
 
 
 struct hv_input_signal_event_buffer {
@@ -483,8 +482,8 @@ struct hv_context {
 	/* 8-bytes aligned of the buffer above */
 	struct hv_input_signal_event *signal_event_param;
 
-	void *synic_message_page[MAX_NUM_CPUS];
-	void *synic_event_page[MAX_NUM_CPUS];
+	void *synic_message_page[NR_CPUS];
+	void *synic_event_page[NR_CPUS];
 };
 
 extern struct hv_context hv_context;