PCI: hv: Add hibernation support

Add suspend() and resume() functions so that Hyper-V virtual PCI devices
are handled properly when the VM hibernates and resumes from
hibernation.

Note that the suspend() function must make sure there are no pending
work items before calling vmbus_close(), since it runs in a process
context as a callback in dpm_suspend(). When it starts to run, the
channel callback hv_pci_onchannelcallback(), which runs in a tasklet
context, can be still running concurrently and scheduling new work items
onto hbus->wq in hv_pci_devices_present() and hv_pci_eject_device(), and
the work item handlers can access the vmbus channel, which can be being
closed by hv_pci_suspend(), e.g. the work item handler
pci_devices_present_work() -> new_pcichild_device() writes to the vmbus
channel.

To eliminate the race, hv_pci_suspend() disables the channel callback
tasklet, sets hbus->state to hv_pcibus_removing, and re-enables the
tasklet.  This way, when hv_pci_suspend() proceeds, it knows that no new
work item can be scheduled, and then it flushes hbus->wq and safely
closes the vmbus channel.
Signed-off-by: Dexuan Cui <decui@microsoft.com>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>

drivers/pci/controller/pci-hyperv.c

@@ -455,6 +455,7 @@ enum hv_pcibus_state {
 	hv_pcibus_init = 0,
 	hv_pcibus_probed,
 	hv_pcibus_installed,
+	hv_pcibus_removing,
 	hv_pcibus_removed,
 	hv_pcibus_maximum
 };
@@ -1681,6 +1682,23 @@ static void prepopulate_bars(struct hv_pcibus_device *hbus)
 
 	spin_lock_irqsave(&hbus->device_list_lock, flags);
 
+	/*
+	 * Clear the memory enable bit, in case it's already set. This occurs
+	 * in the suspend path of hibernation, where the device is suspended,
+	 * resumed and suspended again: see hibernation_snapshot() and
+	 * hibernation_platform_enter().
+	 *
+	 * If the memory enable bit is already set, Hyper-V sliently ignores
+	 * the below BAR updates, and the related PCI device driver can not
+	 * work, because reading from the device register(s) always returns
+	 * 0xFFFFFFFF.
+	 */
+	list_for_each_entry(hpdev, &hbus->children, list_entry) {
+		_hv_pcifront_read_config(hpdev, PCI_COMMAND, 2, &command);
+		command &= ~PCI_COMMAND_MEMORY;
+		_hv_pcifront_write_config(hpdev, PCI_COMMAND, 2, command);
+	}
+
 	/* Pick addresses for the BARs. */
 	do {
 		list_for_each_entry(hpdev, &hbus->children, list_entry) {
@@ -2107,6 +2125,12 @@ static void hv_pci_devices_present(struct hv_pcibus_device *hbus,
 	unsigned long flags;
 	bool pending_dr;
 
+	if (hbus->state == hv_pcibus_removing) {
+		dev_info(&hbus->hdev->device,
+			 "PCI VMBus BUS_RELATIONS: ignored\n");
+		return;
+	}
+
 	dr_wrk = kzalloc(sizeof(*dr_wrk), GFP_NOWAIT);
 	if (!dr_wrk)
 		return;
@@ -2223,11 +2247,19 @@ static void hv_eject_device_work(struct work_struct *work)
  */
 static void hv_pci_eject_device(struct hv_pci_dev *hpdev)
 {
+	struct hv_pcibus_device *hbus = hpdev->hbus;
+	struct hv_device *hdev = hbus->hdev;
+
+	if (hbus->state == hv_pcibus_removing) {
+		dev_info(&hdev->device, "PCI VMBus EJECT: ignored\n");
+		return;
+	}
+
 	hpdev->state = hv_pcichild_ejecting;
 	get_pcichild(hpdev);
 	INIT_WORK(&hpdev->wrk, hv_eject_device_work);
-	get_hvpcibus(hpdev->hbus);
-	queue_work(hpdev->hbus->wq, &hpdev->wrk);
+	get_hvpcibus(hbus);
+	queue_work(hbus->wq, &hpdev->wrk);
 }
 
 /**
@@ -3107,6 +3139,93 @@ static int hv_pci_remove(struct hv_device *hdev)
 	return ret;
 }
 
+static int hv_pci_suspend(struct hv_device *hdev)
+{
+	struct hv_pcibus_device *hbus = hv_get_drvdata(hdev);
+	enum hv_pcibus_state old_state;
+	int ret;
+
+	/*
+	 * hv_pci_suspend() must make sure there are no pending work items
+	 * before calling vmbus_close(), since it runs in a process context
+	 * as a callback in dpm_suspend().  When it starts to run, the channel
+	 * callback hv_pci_onchannelcallback(), which runs in a tasklet
+	 * context, can be still running concurrently and scheduling new work
+	 * items onto hbus->wq in hv_pci_devices_present() and
+	 * hv_pci_eject_device(), and the work item handlers can access the
+	 * vmbus channel, which can be being closed by hv_pci_suspend(), e.g.
+	 * the work item handler pci_devices_present_work() ->
+	 * new_pcichild_device() writes to the vmbus channel.
+	 *
+	 * To eliminate the race, hv_pci_suspend() disables the channel
+	 * callback tasklet, sets hbus->state to hv_pcibus_removing, and
+	 * re-enables the tasklet. This way, when hv_pci_suspend() proceeds,
+	 * it knows that no new work item can be scheduled, and then it flushes
+	 * hbus->wq and safely closes the vmbus channel.
+	 */
+	tasklet_disable(&hdev->channel->callback_event);
+
+	/* Change the hbus state to prevent new work items. */
+	old_state = hbus->state;
+	if (hbus->state == hv_pcibus_installed)
+		hbus->state = hv_pcibus_removing;
+
+	tasklet_enable(&hdev->channel->callback_event);
+
+	if (old_state != hv_pcibus_installed)
+		return -EINVAL;
+
+	flush_workqueue(hbus->wq);
+
+	ret = hv_pci_bus_exit(hdev, true);
+	if (ret)
+		return ret;
+
+	vmbus_close(hdev->channel);
+
+	return 0;
+}
+
+static int hv_pci_resume(struct hv_device *hdev)
+{
+	struct hv_pcibus_device *hbus = hv_get_drvdata(hdev);
+	enum pci_protocol_version_t version[1];
+	int ret;
+
+	hbus->state = hv_pcibus_init;
+
+	ret = vmbus_open(hdev->channel, pci_ring_size, pci_ring_size, NULL, 0,
+			 hv_pci_onchannelcallback, hbus);
+	if (ret)
+		return ret;
+
+	/* Only use the version that was in use before hibernation. */
+	version[0] = pci_protocol_version;
+	ret = hv_pci_protocol_negotiation(hdev, version, 1);
+	if (ret)
+		goto out;
+
+	ret = hv_pci_query_relations(hdev);
+	if (ret)
+		goto out;
+
+	ret = hv_pci_enter_d0(hdev);
+	if (ret)
+		goto out;
+
+	ret = hv_send_resources_allocated(hdev);
+	if (ret)
+		goto out;
+
+	prepopulate_bars(hbus);
+
+	hbus->state = hv_pcibus_installed;
+	return 0;
+out:
+	vmbus_close(hdev->channel);
+	return ret;
+}
+
 static const struct hv_vmbus_device_id hv_pci_id_table[] = {
 	/* PCI Pass-through Class ID */
 	/* 44C4F61D-4444-4400-9D52-802E27EDE19F */
@@ -3121,6 +3240,8 @@ static struct hv_driver hv_pci_drv = {
 	.id_table	= hv_pci_id_table,
 	.probe		= hv_pci_probe,
 	.remove		= hv_pci_remove,
+	.suspend	= hv_pci_suspend,
+	.resume		= hv_pci_resume,
 };
 
 static void __exit exit_hv_pci_drv(void)