hv: make CPU offlining prevention fine-grained

Since commit e513229b4c38 ("Drivers: hv: vmbus: prevent cpu offlining on
newer hypervisors") cpu offlining was disabled. It is still true that we
can't offline CPUs which have VMBus channels bound to them but we may have
'free' CPUs (e.v. we booted with maxcpus= parameter and onlined CPUs after
VMBus was initialized), these CPUs may be disabled without issues.

In future, we may even allow closing CPUs which have only sub-channels
assinged to them by closing these sub-channels. All devices will continue
to work.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

2 files changed, 35 insertions, 5 deletions

diff --git a/drivers/hv/hv.c b/drivers/hv/hv.c
index 8dbbed4efa45..714e1ebc834c 100644
--- a/drivers/hv/hv.c
+++ b/drivers/hv/hv.c
@@ -587,10 +587,41 @@ int hv_synic_cleanup(unsigned int cpu)
 	union hv_synic_simp simp;
 	union hv_synic_siefp siefp;
 	union hv_synic_scontrol sctrl;
+	struct vmbus_channel *channel, *sc;
+	bool channel_found = false;
+	unsigned long flags;
 
 	if (!hv_context.synic_initialized)
 		return -EFAULT;
 
+	/*
+	 * Search for channels which are bound to the CPU we're about to
+	 * cleanup. In case we find one and vmbus is still connected we need to
+	 * fail, this will effectively prevent CPU offlining. There is no way
+	 * we can re-bind channels to different CPUs for now.
+	 */
+	mutex_lock(&vmbus_connection.channel_mutex);
+	list_for_each_entry(channel, &vmbus_connection.chn_list, listentry) {
+		if (channel->target_cpu == cpu) {
+			channel_found = true;
+			break;
+		}
+		spin_lock_irqsave(&channel->lock, flags);
+		list_for_each_entry(sc, &channel->sc_list, sc_list) {
+			if (sc->target_cpu == cpu) {
+				channel_found = true;
+				break;
+			}
+		}
+		spin_unlock_irqrestore(&channel->lock, flags);
+		if (channel_found)
+			break;
+	}
+	mutex_unlock(&vmbus_connection.channel_mutex);
+
+	if (channel_found && vmbus_connection.conn_state == CONNECTED)
+		return -EBUSY;
+
 	/* Turn off clockevent device */
 	if (ms_hyperv.features & HV_X64_MSR_SYNTIMER_AVAILABLE) {
 		clockevents_unbind_device(hv_context.clk_evt[cpu], cpu);
diff --git a/drivers/hv/vmbus_drv.c b/drivers/hv/vmbus_drv.c
index a16d519b7e37..565bdd16134a 100644
--- a/drivers/hv/vmbus_drv.c
+++ b/drivers/hv/vmbus_drv.c
@@ -1008,9 +1008,6 @@ static int vmbus_bus_init(void)
 	if (ret)
 		goto err_connect;
 
-	if (vmbus_proto_version > VERSION_WIN7)
-		cpu_hotplug_disable();
-
 	/*
 	 * Only register if the crash MSRs are available
 	 */
@@ -1486,6 +1483,9 @@ static void hv_kexec_handler(void)
 {
 	hv_synic_clockevents_cleanup();
 	vmbus_initiate_unload(false);
+	vmbus_connection.conn_state = DISCONNECTED;
+	/* Make sure conn_state is set as hv_synic_cleanup checks for it */
+	mb();
 	cpuhp_remove_state(hyperv_cpuhp_online);
 	hv_cleanup(false);
 };
@@ -1498,6 +1498,7 @@ static void hv_crash_handler(struct pt_regs *regs)
 	 * doing the cleanup for current CPU only. This should be sufficient
 	 * for kdump.
 	 */
+	vmbus_connection.conn_state = DISCONNECTED;
 	hv_synic_cleanup(smp_processor_id());
 	hv_cleanup(true);
 };
@@ -1566,8 +1567,6 @@ static void __exit vmbus_exit(void)
 	cpuhp_remove_state(hyperv_cpuhp_online);
 	hv_synic_free();
 	acpi_bus_unregister_driver(&vmbus_acpi_driver);
-	if (vmbus_proto_version > VERSION_WIN7)
-		cpu_hotplug_enable();
 }