Merge tag 'libnvdimm-fixes-5.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm

Pull libnvdimm fixes from Dan Williams:
 "A collection of locking and async operations fixes for v5.3-rc2. These
  had been soaking in a branch targeting the merge window, but missed
  due to a regression hunt. This fixed up version has otherwise been in
  -next this past week with no reported issues.

  In order to gain confidence in the locking changes the pull also
  includes a debug / instrumentation patch to enable lockdep coverage
  for libnvdimm subsystem operations that depend on the device_lock for
  exclusion. As mentioned in the changelog it is a hack, but it works
  and documents the locking expectations of the sub-system in a way that
  others can use lockdep to verify. The driver core touches got an ack
  from Greg.

  Summary:

   - Fix duplicate device_unregister() calls (multiple threads competing
     to do unregister work when scheduling device removal from a sysfs
     attribute of the self-same device).

   - Fix badblocks registration order bug. Ensure region badblocks are
     initialized in advance of namespace registration.

   - Fix a deadlock between the bus lock and probe operations.

   - Export device-core infrastructure to coordinate async operations
     via the device ->dead state.

   - Add device-core infrastructure to validate device_lock() usage with
     lockdep"

* tag 'libnvdimm-fixes-5.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm:
  driver-core, libnvdimm: Let device subsystems add local lockdep coverage
  libnvdimm/bus: Fix wait_nvdimm_bus_probe_idle() ABBA deadlock
  libnvdimm/bus: Stop holding nvdimm_bus_list_mutex over __nd_ioctl()
  libnvdimm/bus: Prepare the nd_ioctl() path to be re-entrant
  libnvdimm/region: Register badblocks before namespaces
  libnvdimm/bus: Prevent duplicate device_unregister() calls
  drivers/base: Introduce kill_device()

drivers/acpi/nfit/core.c

@@ -1282,7 +1282,7 @@ static ssize_t hw_error_scrub_store(struct device *dev,
 	if (rc)
 		return rc;
 
-	device_lock(dev);
+	nfit_device_lock(dev);
 	nd_desc = dev_get_drvdata(dev);
 	if (nd_desc) {
 		struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
@@ -1299,7 +1299,7 @@ static ssize_t hw_error_scrub_store(struct device *dev,
 			break;
 		}
 	}
-	device_unlock(dev);
+	nfit_device_unlock(dev);
 	if (rc)
 		return rc;
 	return size;
@@ -1319,7 +1319,7 @@ static ssize_t scrub_show(struct device *dev,
 	ssize_t rc = -ENXIO;
 	bool busy;
 
-	device_lock(dev);
+	nfit_device_lock(dev);
 	nd_desc = dev_get_drvdata(dev);
 	if (!nd_desc) {
 		device_unlock(dev);
@@ -1339,7 +1339,7 @@ static ssize_t scrub_show(struct device *dev,
 	}
 
 	mutex_unlock(&acpi_desc->init_mutex);
-	device_unlock(dev);
+	nfit_device_unlock(dev);
 	return rc;
 }
 
@@ -1356,14 +1356,14 @@ static ssize_t scrub_store(struct device *dev,
 	if (val != 1)
 		return -EINVAL;
 
-	device_lock(dev);
+	nfit_device_lock(dev);
 	nd_desc = dev_get_drvdata(dev);
 	if (nd_desc) {
 		struct acpi_nfit_desc *acpi_desc = to_acpi_desc(nd_desc);
 
 		rc = acpi_nfit_ars_rescan(acpi_desc, ARS_REQ_LONG);
 	}
-	device_unlock(dev);
+	nfit_device_unlock(dev);
 	if (rc)
 		return rc;
 	return size;
@@ -1749,9 +1749,9 @@ static void acpi_nvdimm_notify(acpi_handle handle, u32 event, void *data)
 	struct acpi_device *adev = data;
 	struct device *dev = &adev->dev;
 
-	device_lock(dev->parent);
+	nfit_device_lock(dev->parent);
 	__acpi_nvdimm_notify(dev, event);
-	device_unlock(dev->parent);
+	nfit_device_unlock(dev->parent);
 }
 
 static bool acpi_nvdimm_has_method(struct acpi_device *adev, char *method)
@@ -3457,8 +3457,8 @@ static int acpi_nfit_flush_probe(struct nvdimm_bus_descriptor *nd_desc)
 	struct device *dev = acpi_desc->dev;
 
 	/* Bounce the device lock to flush acpi_nfit_add / acpi_nfit_notify */
-	device_lock(dev);
-	device_unlock(dev);
+	nfit_device_lock(dev);
+	nfit_device_unlock(dev);
 
 	/* Bounce the init_mutex to complete initial registration */
 	mutex_lock(&acpi_desc->init_mutex);
@@ -3602,8 +3602,8 @@ void acpi_nfit_shutdown(void *data)
 	 * acpi_nfit_ars_rescan() submissions have had a chance to
 	 * either submit or see ->cancel set.
 	 */
-	device_lock(bus_dev);
-	device_unlock(bus_dev);
+	nfit_device_lock(bus_dev);
+	nfit_device_unlock(bus_dev);
 
 	flush_workqueue(nfit_wq);
 }
@@ -3746,9 +3746,9 @@ EXPORT_SYMBOL_GPL(__acpi_nfit_notify);
 
 static void acpi_nfit_notify(struct acpi_device *adev, u32 event)
 {
-	device_lock(&adev->dev);
+	nfit_device_lock(&adev->dev);
 	__acpi_nfit_notify(&adev->dev, adev->handle, event);
-	device_unlock(&adev->dev);
+	nfit_device_unlock(&adev->dev);
 }
 
 static const struct acpi_device_id acpi_nfit_ids[] = {

drivers/acpi/nfit/nfit.h

@@ -312,6 +312,30 @@ static inline struct acpi_nfit_desc *to_acpi_desc(
 	return container_of(nd_desc, struct acpi_nfit_desc, nd_desc);
 }
 
+#ifdef CONFIG_PROVE_LOCKING
+static inline void nfit_device_lock(struct device *dev)
+{
+	device_lock(dev);
+	mutex_lock(&dev->lockdep_mutex);
+}
+
+static inline void nfit_device_unlock(struct device *dev)
+{
+	mutex_unlock(&dev->lockdep_mutex);
+	device_unlock(dev);
+}
+#else
+static inline void nfit_device_lock(struct device *dev)
+{
+	device_lock(dev);
+}
+
+static inline void nfit_device_unlock(struct device *dev)
+{
+	device_unlock(dev);
+}
+#endif
+
 const guid_t *to_nfit_uuid(enum nfit_uuids id);
 int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *nfit, acpi_size sz);
 void acpi_nfit_shutdown(void *data);

drivers/base/core.c

@@ -1663,6 +1663,9 @@ void device_initialize(struct device *dev)
 	kobject_init(&dev->kobj, &device_ktype);
 	INIT_LIST_HEAD(&dev->dma_pools);
 	mutex_init(&dev->mutex);
+#ifdef CONFIG_PROVE_LOCKING
+	mutex_init(&dev->lockdep_mutex);
+#endif
 	lockdep_set_novalidate_class(&dev->mutex);
 	spin_lock_init(&dev->devres_lock);
 	INIT_LIST_HEAD(&dev->devres_head);
@@ -2211,6 +2214,24 @@ void put_device(struct device *dev)
 }
 EXPORT_SYMBOL_GPL(put_device);
 
+bool kill_device(struct device *dev)
+{
+	/*
+	 * Require the device lock and set the "dead" flag to guarantee that
+	 * the update behavior is consistent with the other bitfields near
+	 * it and that we cannot have an asynchronous probe routine trying
+	 * to run while we are tearing out the bus/class/sysfs from
+	 * underneath the device.
+	 */
+	lockdep_assert_held(&dev->mutex);
+
+	if (dev->p->dead)
+		return false;
+	dev->p->dead = true;
+	return true;
+}
+EXPORT_SYMBOL_GPL(kill_device);
+
 /**
  * device_del - delete device from system.
  * @dev: device.
@@ -2230,15 +2251,8 @@ void device_del(struct device *dev)
 	struct kobject *glue_dir = NULL;
 	struct class_interface *class_intf;
 
-	/*
-	 * Hold the device lock and set the "dead" flag to guarantee that
-	 * the update behavior is consistent with the other bitfields near
-	 * it and that we cannot have an asynchronous probe routine trying
-	 * to run while we are tearing out the bus/class/sysfs from
-	 * underneath the device.
-	 */
 	device_lock(dev);
-	dev->p->dead = true;
+	kill_device(dev);
 	device_unlock(dev);
 
 	/* Notify clients of device removal.  This call must come

drivers/nvdimm/btt_devs.c

@@ -62,14 +62,14 @@ static ssize_t sector_size_store(struct device *dev,
 	struct nd_btt *nd_btt = to_nd_btt(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	rc = nd_size_select_store(dev, buf, &nd_btt->lbasize,
 			btt_lbasize_supported);
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc ? rc : len;
 }
@@ -91,11 +91,11 @@ static ssize_t uuid_store(struct device *dev,
 	struct nd_btt *nd_btt = to_nd_btt(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	rc = nd_uuid_store(dev, &nd_btt->uuid, buf, len);
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc ? rc : len;
 }
@@ -120,13 +120,13 @@ static ssize_t namespace_store(struct device *dev,
 	struct nd_btt *nd_btt = to_nd_btt(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	rc = nd_namespace_store(dev, &nd_btt->ndns, buf, len);
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }
@@ -138,14 +138,14 @@ static ssize_t size_show(struct device *dev,
 	struct nd_btt *nd_btt = to_nd_btt(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	if (dev->driver)
 		rc = sprintf(buf, "%llu\n", nd_btt->size);
 	else {
 		/* no size to convey if the btt instance is disabled */
 		rc = -ENXIO;
 	}
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }

drivers/nvdimm/bus.c

@@ -26,7 +26,7 @@
 
 int nvdimm_major;
 static int nvdimm_bus_major;
-static struct class *nd_class;
+struct class *nd_class;
 static DEFINE_IDA(nd_ida);
 
 static int to_nd_device_type(struct device *dev)
@@ -73,7 +73,7 @@ static void nvdimm_bus_probe_end(struct nvdimm_bus *nvdimm_bus)
 {
 	nvdimm_bus_lock(&nvdimm_bus->dev);
 	if (--nvdimm_bus->probe_active == 0)
-		wake_up(&nvdimm_bus->probe_wait);
+		wake_up(&nvdimm_bus->wait);
 	nvdimm_bus_unlock(&nvdimm_bus->dev);
 }
 
@@ -91,7 +91,10 @@ static int nvdimm_bus_probe(struct device *dev)
 			dev->driver->name, dev_name(dev));
 
 	nvdimm_bus_probe_start(nvdimm_bus);
+	debug_nvdimm_lock(dev);
 	rc = nd_drv->probe(dev);
+	debug_nvdimm_unlock(dev);
+
 	if (rc == 0)
 		nd_region_probe_success(nvdimm_bus, dev);
 	else
@@ -113,8 +116,11 @@ static int nvdimm_bus_remove(struct device *dev)
 	struct nvdimm_bus *nvdimm_bus = walk_to_nvdimm_bus(dev);
 	int rc = 0;
 
-	if (nd_drv->remove)
+	if (nd_drv->remove) {
+		debug_nvdimm_lock(dev);
 		rc = nd_drv->remove(dev);
+		debug_nvdimm_unlock(dev);
+	}
 	nd_region_disable(nvdimm_bus, dev);
 
 	dev_dbg(&nvdimm_bus->dev, "%s.remove(%s) = %d\n", dev->driver->name,
@@ -140,7 +146,7 @@ static void nvdimm_bus_shutdown(struct device *dev)
 
 void nd_device_notify(struct device *dev, enum nvdimm_event event)
 {
-	device_lock(dev);
+	nd_device_lock(dev);
 	if (dev->driver) {
 		struct nd_device_driver *nd_drv;
 
@@ -148,7 +154,7 @@ void nd_device_notify(struct device *dev, enum nvdimm_event event)
 		if (nd_drv->notify)
 			nd_drv->notify(dev, event);
 	}
-	device_unlock(dev);
+	nd_device_unlock(dev);
 }
 EXPORT_SYMBOL(nd_device_notify);
 
@@ -296,7 +302,7 @@ static void nvdimm_bus_release(struct device *dev)
 	kfree(nvdimm_bus);
 }
 
-static bool is_nvdimm_bus(struct device *dev)
+bool is_nvdimm_bus(struct device *dev)
 {
 	return dev->release == nvdimm_bus_release;
 }
@@ -341,7 +347,7 @@ struct nvdimm_bus *nvdimm_bus_register(struct device *parent,
 		return NULL;
 	INIT_LIST_HEAD(&nvdimm_bus->list);
 	INIT_LIST_HEAD(&nvdimm_bus->mapping_list);
-	init_waitqueue_head(&nvdimm_bus->probe_wait);
+	init_waitqueue_head(&nvdimm_bus->wait);
 	nvdimm_bus->id = ida_simple_get(&nd_ida, 0, 0, GFP_KERNEL);
 	if (nvdimm_bus->id < 0) {
 		kfree(nvdimm_bus);
@@ -426,6 +432,9 @@ static int nd_bus_remove(struct device *dev)
 	list_del_init(&nvdimm_bus->list);
 	mutex_unlock(&nvdimm_bus_list_mutex);
 
+	wait_event(nvdimm_bus->wait,
+			atomic_read(&nvdimm_bus->ioctl_active) == 0);
+
 	nd_synchronize();
 	device_for_each_child(&nvdimm_bus->dev, NULL, child_unregister);
 
@@ -547,13 +556,38 @@ EXPORT_SYMBOL(nd_device_register);
 
 void nd_device_unregister(struct device *dev, enum nd_async_mode mode)
 {
+	bool killed;
+
 	switch (mode) {
 	case ND_ASYNC:
+		/*
+		 * In the async case this is being triggered with the
+		 * device lock held and the unregistration work needs to
+		 * be moved out of line iff this is thread has won the
+		 * race to schedule the deletion.
+		 */
+		if (!kill_device(dev))
+			return;
+
 		get_device(dev);
 		async_schedule_domain(nd_async_device_unregister, dev,
 				&nd_async_domain);
 		break;
 	case ND_SYNC:
+		/*
+		 * In the sync case the device is being unregistered due
+		 * to a state change of the parent. Claim the kill state
+		 * to synchronize against other unregistration requests,
+		 * or otherwise let the async path handle it if the
+		 * unregistration was already queued.
+		 */
+		nd_device_lock(dev);
+		killed = kill_device(dev);
+		nd_device_unlock(dev);
+
+		if (!killed)
+			return;
+
 		nd_synchronize();
 		device_unregister(dev);
 		break;
@@ -859,10 +893,12 @@ void wait_nvdimm_bus_probe_idle(struct device *dev)
 	do {
 		if (nvdimm_bus->probe_active == 0)
 			break;
-		nvdimm_bus_unlock(&nvdimm_bus->dev);
-		wait_event(nvdimm_bus->probe_wait,
+		nvdimm_bus_unlock(dev);
+		nd_device_unlock(dev);
+		wait_event(nvdimm_bus->wait,
 				nvdimm_bus->probe_active == 0);
-		nvdimm_bus_lock(&nvdimm_bus->dev);
+		nd_device_lock(dev);
+		nvdimm_bus_lock(dev);
 	} while (true);
 }
 
@@ -945,20 +981,19 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 		int read_only, unsigned int ioctl_cmd, unsigned long arg)
 {
 	struct nvdimm_bus_descriptor *nd_desc = nvdimm_bus->nd_desc;
-	static char out_env[ND_CMD_MAX_ENVELOPE];
-	static char in_env[ND_CMD_MAX_ENVELOPE];
 	const struct nd_cmd_desc *desc = NULL;
 	unsigned int cmd = _IOC_NR(ioctl_cmd);
 	struct device *dev = &nvdimm_bus->dev;
 	void __user *p = (void __user *) arg;
+	char *out_env = NULL, *in_env = NULL;
 	const char *cmd_name, *dimm_name;
 	u32 in_len = 0, out_len = 0;
 	unsigned int func = cmd;
 	unsigned long cmd_mask;
 	struct nd_cmd_pkg pkg;
 	int rc, i, cmd_rc;
+	void *buf = NULL;
 	u64 buf_len = 0;
-	void *buf;
 
 	if (nvdimm) {
 		desc = nd_cmd_dimm_desc(cmd);
@@ -989,7 +1024,7 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 		case ND_CMD_ARS_START:
 		case ND_CMD_CLEAR_ERROR:
 		case ND_CMD_CALL:
-			dev_dbg(&nvdimm_bus->dev, "'%s' command while read-only.\n",
+			dev_dbg(dev, "'%s' command while read-only.\n",
 					nvdimm ? nvdimm_cmd_name(cmd)
 					: nvdimm_bus_cmd_name(cmd));
 			return -EPERM;
@@ -998,6 +1033,9 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 		}
 
 	/* process an input envelope */
+	in_env = kzalloc(ND_CMD_MAX_ENVELOPE, GFP_KERNEL);
+	if (!in_env)
+		return -ENOMEM;
 	for (i = 0; i < desc->in_num; i++) {
 		u32 in_size, copy;
 
@@ -1005,14 +1043,17 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 		if (in_size == UINT_MAX) {
 			dev_err(dev, "%s:%s unknown input size cmd: %s field: %d\n",
 					__func__, dimm_name, cmd_name, i);
-			return -ENXIO;
+			rc = -ENXIO;
+			goto out;
 		}
-		if (in_len < sizeof(in_env))
-			copy = min_t(u32, sizeof(in_env) - in_len, in_size);
+		if (in_len < ND_CMD_MAX_ENVELOPE)
+			copy = min_t(u32, ND_CMD_MAX_ENVELOPE - in_len, in_size);
 		else
 			copy = 0;
-		if (copy && copy_from_user(&in_env[in_len], p + in_len, copy))
-			return -EFAULT;
+		if (copy && copy_from_user(&in_env[in_len], p + in_len, copy)) {
+			rc = -EFAULT;
+			goto out;
+		}
 		in_len += in_size;
 	}
 
@@ -1024,6 +1065,12 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 	}
 
 	/* process an output envelope */
+	out_env = kzalloc(ND_CMD_MAX_ENVELOPE, GFP_KERNEL);
+	if (!out_env) {
+		rc = -ENOMEM;
+		goto out;
+	}
+
 	for (i = 0; i < desc->out_num; i++) {
 		u32 out_size = nd_cmd_out_size(nvdimm, cmd, desc, i,
 				(u32 *) in_env, (u32 *) out_env, 0);
@@ -1032,15 +1079,18 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 		if (out_size == UINT_MAX) {
 			dev_dbg(dev, "%s unknown output size cmd: %s field: %d\n",
 					dimm_name, cmd_name, i);
-			return -EFAULT;
+			rc = -EFAULT;
+			goto out;
 		}
-		if (out_len < sizeof(out_env))
-			copy = min_t(u32, sizeof(out_env) - out_len, out_size);
+		if (out_len < ND_CMD_MAX_ENVELOPE)
+			copy = min_t(u32, ND_CMD_MAX_ENVELOPE - out_len, out_size);
 		else
 			copy = 0;
 		if (copy && copy_from_user(&out_env[out_len],
-					p + in_len + out_len, copy))
-			return -EFAULT;
+					p + in_len + out_len, copy)) {
+			rc = -EFAULT;
+			goto out;
+		}
 		out_len += out_size;
 	}
 
@@ -1048,19 +1098,23 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 	if (buf_len > ND_IOCTL_MAX_BUFLEN) {
 		dev_dbg(dev, "%s cmd: %s buf_len: %llu > %d\n", dimm_name,
 				cmd_name, buf_len, ND_IOCTL_MAX_BUFLEN);
-		return -EINVAL;
+		rc = -EINVAL;
+		goto out;
 	}
 
 	buf = vmalloc(buf_len);
-	if (!buf)
-		return -ENOMEM;
+	if (!buf) {
+		rc = -ENOMEM;
+		goto out;
+	}
 
 	if (copy_from_user(buf, p, buf_len)) {
 		rc = -EFAULT;
 		goto out;
 	}
 
-	nvdimm_bus_lock(&nvdimm_bus->dev);
+	nd_device_lock(dev);
+	nvdimm_bus_lock(dev);
 	rc = nd_cmd_clear_to_send(nvdimm_bus, nvdimm, func, buf);
 	if (rc)
 		goto out_unlock;
@@ -1075,39 +1129,24 @@ static int __nd_ioctl(struct nvdimm_bus *nvdimm_bus, struct nvdimm *nvdimm,
 		nvdimm_account_cleared_poison(nvdimm_bus, clear_err->address,
 				clear_err->cleared);
 	}
-	nvdimm_bus_unlock(&nvdimm_bus->dev);
 
 	if (copy_to_user(p, buf, buf_len))
 		rc = -EFAULT;
 
-	vfree(buf);
-	return rc;
-
- out_unlock:
-	nvdimm_bus_unlock(&nvdimm_bus->dev);
- out:
+out_unlock:
+	nvdimm_bus_unlock(dev);
+	nd_device_unlock(dev);
+out:
+	kfree(in_env);
+	kfree(out_env);
 	vfree(buf);
 	return rc;
 }
 
-static long nd_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
-{
-	long id = (long) file->private_data;
-	int rc = -ENXIO, ro;
-	struct nvdimm_bus *nvdimm_bus;
-
-	ro = ((file->f_flags & O_ACCMODE) == O_RDONLY);
-	mutex_lock(&nvdimm_bus_list_mutex);
-	list_for_each_entry(nvdimm_bus, &nvdimm_bus_list, list) {
-		if (nvdimm_bus->id == id) {
-			rc = __nd_ioctl(nvdimm_bus, NULL, ro, cmd, arg);
-			break;
-		}
-	}
-	mutex_unlock(&nvdimm_bus_list_mutex);
-
-	return rc;
-}
+enum nd_ioctl_mode {
+	BUS_IOCTL,
+	DIMM_IOCTL,
+};
 
 static int match_dimm(struct device *dev, void *data)
 {
@@ -1122,31 +1161,62 @@ static int match_dimm(struct device *dev, void *data)
 	return 0;
 }
 
-static long nvdimm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+static long nd_ioctl(struct file *file, unsigned int cmd, unsigned long arg,
+		enum nd_ioctl_mode mode)
+
 {
-	int rc = -ENXIO, ro;
-	struct nvdimm_bus *nvdimm_bus;
+	struct nvdimm_bus *nvdimm_bus, *found = NULL;
+	long id = (long) file->private_data;
+	struct nvdimm *nvdimm = NULL;
+	int rc, ro;
 
 	ro = ((file->f_flags & O_ACCMODE) == O_RDONLY);
 	mutex_lock(&nvdimm_bus_list_mutex);
 	list_for_each_entry(nvdimm_bus, &nvdimm_bus_list, list) {
-		struct device *dev = device_find_child(&nvdimm_bus->dev,
-				file->private_data, match_dimm);
-		struct nvdimm *nvdimm;
+		if (mode == DIMM_IOCTL) {
+			struct device *dev;
 
+			dev = device_find_child(&nvdimm_bus->dev,
+					file->private_data, match_dimm);
 			if (!dev)
 				continue;
-
 			nvdimm = to_nvdimm(dev);
-		rc = __nd_ioctl(nvdimm_bus, nvdimm, ro, cmd, arg);
-		put_device(dev);
+			found = nvdimm_bus;
+		} else if (nvdimm_bus->id == id) {
+			found = nvdimm_bus;
+		}
+
+		if (found) {
+			atomic_inc(&nvdimm_bus->ioctl_active);
 			break;
 		}
+	}
 	mutex_unlock(&nvdimm_bus_list_mutex);
 
+	if (!found)
+		return -ENXIO;
+
+	nvdimm_bus = found;
+	rc = __nd_ioctl(nvdimm_bus, nvdimm, ro, cmd, arg);
+
+	if (nvdimm)
+		put_device(&nvdimm->dev);
+	if (atomic_dec_and_test(&nvdimm_bus->ioctl_active))
+		wake_up(&nvdimm_bus->wait);
+
 	return rc;
 }
 
+static long bus_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	return nd_ioctl(file, cmd, arg, BUS_IOCTL);
+}
+
+static long dimm_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
+{
+	return nd_ioctl(file, cmd, arg, DIMM_IOCTL);
+}
+
 static int nd_open(struct inode *inode, struct file *file)
 {
 	long minor = iminor(inode);
@@ -1158,16 +1228,16 @@ static int nd_open(struct inode *inode, struct file *file)
 static const struct file_operations nvdimm_bus_fops = {
 	.owner = THIS_MODULE,
 	.open = nd_open,
-	.unlocked_ioctl = nd_ioctl,
-	.compat_ioctl = nd_ioctl,
+	.unlocked_ioctl = bus_ioctl,
+	.compat_ioctl = bus_ioctl,
 	.llseek = noop_llseek,
 };
 
 static const struct file_operations nvdimm_fops = {
 	.owner = THIS_MODULE,
 	.open = nd_open,
-	.unlocked_ioctl = nvdimm_ioctl,
-	.compat_ioctl = nvdimm_ioctl,
+	.unlocked_ioctl = dimm_ioctl,
+	.compat_ioctl = dimm_ioctl,
 	.llseek = noop_llseek,
 };
 

drivers/nvdimm/core.c

@@ -246,7 +246,7 @@ static int nd_uuid_parse(struct device *dev, u8 *uuid_out, const char *buf,
  *
  * Enforce that uuids can only be changed while the device is disabled
  * (driver detached)
- * LOCKING: expects device_lock() is held on entry
+ * LOCKING: expects nd_device_lock() is held on entry
  */
 int nd_uuid_store(struct device *dev, u8 **uuid_out, const char *buf,
 		size_t len)
@@ -347,15 +347,15 @@ static DEVICE_ATTR_RO(provider);
 
 static int flush_namespaces(struct device *dev, void *data)
 {
-	device_lock(dev);
-	device_unlock(dev);
+	nd_device_lock(dev);
+	nd_device_unlock(dev);
 	return 0;
 }
 
 static int flush_regions_dimms(struct device *dev, void *data)
 {
-	device_lock(dev);
-	device_unlock(dev);
+	nd_device_lock(dev);
+	nd_device_unlock(dev);
 	device_for_each_child(dev, NULL, flush_namespaces);
 	return 0;
 }

drivers/nvdimm/dimm_devs.c

@@ -484,12 +484,12 @@ static ssize_t security_store(struct device *dev,
 	 * done while probing is idle and the DIMM is not in active use
 	 * in any region.
 	 */
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	rc = __security_store(dev, buf, len);
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }

drivers/nvdimm/namespace_devs.c

@@ -410,7 +410,7 @@ static ssize_t alt_name_store(struct device *dev,
 	struct nd_region *nd_region = to_nd_region(dev->parent);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	rc = __alt_name_store(dev, buf, len);
@@ -418,7 +418,7 @@ static ssize_t alt_name_store(struct device *dev,
 		rc = nd_namespace_label_update(nd_region, dev);
 	dev_dbg(dev, "%s(%zd)\n", rc < 0 ? "fail " : "", rc);
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc < 0 ? rc : len;
 }
@@ -1077,7 +1077,7 @@ static ssize_t size_store(struct device *dev,
 	if (rc)
 		return rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	rc = __size_store(dev, val);
@@ -1103,7 +1103,7 @@ static ssize_t size_store(struct device *dev,
 	dev_dbg(dev, "%llx %s (%d)\n", val, rc < 0 ? "fail" : "success", rc);
 
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc < 0 ? rc : len;
 }
@@ -1286,7 +1286,7 @@ static ssize_t uuid_store(struct device *dev,
 	} else
 		return -ENXIO;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	if (to_ndns(dev)->claim)
@@ -1302,7 +1302,7 @@ static ssize_t uuid_store(struct device *dev,
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc < 0 ? rc : len;
 }
@@ -1376,7 +1376,7 @@ static ssize_t sector_size_store(struct device *dev,
 	} else
 		return -ENXIO;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	if (to_ndns(dev)->claim)
 		rc = -EBUSY;
@@ -1387,7 +1387,7 @@ static ssize_t sector_size_store(struct device *dev,
 	dev_dbg(dev, "result: %zd %s: %s%s", rc, rc < 0 ? "tried" : "wrote",
 			buf, buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc ? rc : len;
 }
@@ -1502,9 +1502,9 @@ static ssize_t holder_show(struct device *dev,
 	struct nd_namespace_common *ndns = to_ndns(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	rc = sprintf(buf, "%s\n", ndns->claim ? dev_name(ndns->claim) : "");
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }
@@ -1541,7 +1541,7 @@ static ssize_t holder_class_store(struct device *dev,
 	struct nd_region *nd_region = to_nd_region(dev->parent);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	rc = __holder_class_store(dev, buf);
@@ -1549,7 +1549,7 @@ static ssize_t holder_class_store(struct device *dev,
 		rc = nd_namespace_label_update(nd_region, dev);
 	dev_dbg(dev, "%s(%zd)\n", rc < 0 ? "fail " : "", rc);
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc < 0 ? rc : len;
 }
@@ -1560,7 +1560,7 @@ static ssize_t holder_class_show(struct device *dev,
 	struct nd_namespace_common *ndns = to_ndns(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	if (ndns->claim_class == NVDIMM_CCLASS_NONE)
 		rc = sprintf(buf, "\n");
 	else if ((ndns->claim_class == NVDIMM_CCLASS_BTT) ||
@@ -1572,7 +1572,7 @@ static ssize_t holder_class_show(struct device *dev,
 		rc = sprintf(buf, "dax\n");
 	else
 		rc = sprintf(buf, "<unknown>\n");
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }
@@ -1586,7 +1586,7 @@ static ssize_t mode_show(struct device *dev,
 	char *mode;
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	claim = ndns->claim;
 	if (claim && is_nd_btt(claim))
 		mode = "safe";
@@ -1599,7 +1599,7 @@ static ssize_t mode_show(struct device *dev,
 	else
 		mode = "raw";
 	rc = sprintf(buf, "%s\n", mode);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }
@@ -1703,8 +1703,8 @@ struct nd_namespace_common *nvdimm_namespace_common_probe(struct device *dev)
 		 * Flush any in-progess probes / removals in the driver
 		 * for the raw personality of this namespace.
 		 */
-		device_lock(&ndns->dev);
-		device_unlock(&ndns->dev);
+		nd_device_lock(&ndns->dev);
+		nd_device_unlock(&ndns->dev);
 		if (ndns->dev.driver) {
 			dev_dbg(&ndns->dev, "is active, can't bind %s\n",
 					dev_name(dev));

drivers/nvdimm/nd-core.h

@@ -9,6 +9,7 @@
 #include <linux/sizes.h>
 #include <linux/mutex.h>
 #include <linux/nd.h>
+#include "nd.h"
 
 extern struct list_head nvdimm_bus_list;
 extern struct mutex nvdimm_bus_list_mutex;
@@ -17,10 +18,11 @@ extern struct workqueue_struct *nvdimm_wq;
 
 struct nvdimm_bus {
 	struct nvdimm_bus_descriptor *nd_desc;
-	wait_queue_head_t probe_wait;
+	wait_queue_head_t wait;
 	struct list_head list;
 	struct device dev;
 	int id, probe_active;
+	atomic_t ioctl_active;
 	struct list_head mapping_list;
 	struct mutex reconfig_mutex;
 	struct badrange badrange;
@@ -181,4 +183,71 @@ ssize_t nd_namespace_store(struct device *dev,
 		struct nd_namespace_common **_ndns, const char *buf,
 		size_t len);
 struct nd_pfn *to_nd_pfn_safe(struct device *dev);
+bool is_nvdimm_bus(struct device *dev);
+
+#ifdef CONFIG_PROVE_LOCKING
+extern struct class *nd_class;
+
+enum {
+	LOCK_BUS,
+	LOCK_NDCTL,
+	LOCK_REGION,
+	LOCK_DIMM = LOCK_REGION,
+	LOCK_NAMESPACE,
+	LOCK_CLAIM,
+};
+
+static inline void debug_nvdimm_lock(struct device *dev)
+{
+	if (is_nd_region(dev))
+		mutex_lock_nested(&dev->lockdep_mutex, LOCK_REGION);
+	else if (is_nvdimm(dev))
+		mutex_lock_nested(&dev->lockdep_mutex, LOCK_DIMM);
+	else if (is_nd_btt(dev) || is_nd_pfn(dev) || is_nd_dax(dev))
+		mutex_lock_nested(&dev->lockdep_mutex, LOCK_CLAIM);
+	else if (dev->parent && (is_nd_region(dev->parent)))
+		mutex_lock_nested(&dev->lockdep_mutex, LOCK_NAMESPACE);
+	else if (is_nvdimm_bus(dev))
+		mutex_lock_nested(&dev->lockdep_mutex, LOCK_BUS);
+	else if (dev->class && dev->class == nd_class)
+		mutex_lock_nested(&dev->lockdep_mutex, LOCK_NDCTL);
+	else
+		dev_WARN(dev, "unknown lock level\n");
+}
+
+static inline void debug_nvdimm_unlock(struct device *dev)
+{
+	mutex_unlock(&dev->lockdep_mutex);
+}
+
+static inline void nd_device_lock(struct device *dev)
+{
+	device_lock(dev);
+	debug_nvdimm_lock(dev);
+}
+
+static inline void nd_device_unlock(struct device *dev)
+{
+	debug_nvdimm_unlock(dev);
+	device_unlock(dev);
+}
+#else
+static inline void nd_device_lock(struct device *dev)
+{
+	device_lock(dev);
+}
+
+static inline void nd_device_unlock(struct device *dev)
+{
+	device_unlock(dev);
+}
+
+static inline void debug_nvdimm_lock(struct device *dev)
+{
+}
+
+static inline void debug_nvdimm_unlock(struct device *dev)
+{
+}
+#endif
 #endif /* __ND_CORE_H__ */

drivers/nvdimm/pfn_devs.c

@@ -67,7 +67,7 @@ static ssize_t mode_store(struct device *dev,
 	struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
 	ssize_t rc = 0;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	if (dev->driver)
 		rc = -EBUSY;
@@ -89,7 +89,7 @@ static ssize_t mode_store(struct device *dev,
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc ? rc : len;
 }
@@ -132,14 +132,14 @@ static ssize_t align_store(struct device *dev,
 	struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	rc = nd_size_select_store(dev, buf, &nd_pfn->align,
 			nd_pfn_supported_alignments());
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc ? rc : len;
 }
@@ -161,11 +161,11 @@ static ssize_t uuid_store(struct device *dev,
 	struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	rc = nd_uuid_store(dev, &nd_pfn->uuid, buf, len);
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc ? rc : len;
 }
@@ -190,13 +190,13 @@ static ssize_t namespace_store(struct device *dev,
 	struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	rc = nd_namespace_store(dev, &nd_pfn->ndns, buf, len);
 	dev_dbg(dev, "result: %zd wrote: %s%s", rc, buf,
 			buf[len - 1] == '\n' ? "" : "\n");
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }
@@ -208,7 +208,7 @@ static ssize_t resource_show(struct device *dev,
 	struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	if (dev->driver) {
 		struct nd_pfn_sb *pfn_sb = nd_pfn->pfn_sb;
 		u64 offset = __le64_to_cpu(pfn_sb->dataoff);
@@ -222,7 +222,7 @@ static ssize_t resource_show(struct device *dev,
 		/* no address to convey if the pfn instance is disabled */
 		rc = -ENXIO;
 	}
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }
@@ -234,7 +234,7 @@ static ssize_t size_show(struct device *dev,
 	struct nd_pfn *nd_pfn = to_nd_pfn_safe(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	if (dev->driver) {
 		struct nd_pfn_sb *pfn_sb = nd_pfn->pfn_sb;
 		u64 offset = __le64_to_cpu(pfn_sb->dataoff);
@@ -250,7 +250,7 @@ static ssize_t size_show(struct device *dev,
 		/* no size to convey if the pfn instance is disabled */
 		rc = -ENXIO;
 	}
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }

drivers/nvdimm/pmem.c

@@ -522,8 +522,8 @@ static int nd_pmem_remove(struct device *dev)
 		nvdimm_namespace_detach_btt(to_nd_btt(dev));
 	else {
 		/*
-		 * Note, this assumes device_lock() context to not race
-		 * nd_pmem_notify()
+		 * Note, this assumes nd_device_lock() context to not
+		 * race nd_pmem_notify()
 		 */
 		sysfs_put(pmem->bb_state);
 		pmem->bb_state = NULL;

drivers/nvdimm/region.c

@@ -34,17 +34,6 @@ static int nd_region_probe(struct device *dev)
 	if (rc)
 		return rc;
 
-	rc = nd_region_register_namespaces(nd_region, &err);
-	if (rc < 0)
-		return rc;
-
-	ndrd = dev_get_drvdata(dev);
-	ndrd->ns_active = rc;
-	ndrd->ns_count = rc + err;
-
-	if (rc && err && rc == err)
-		return -ENODEV;
-
 	if (is_nd_pmem(&nd_region->dev)) {
 		struct resource ndr_res;
 
@@ -60,6 +49,17 @@ static int nd_region_probe(struct device *dev)
 		nvdimm_badblocks_populate(nd_region, &nd_region->bb, &ndr_res);
 	}
 
+	rc = nd_region_register_namespaces(nd_region, &err);
+	if (rc < 0)
+		return rc;
+
+	ndrd = dev_get_drvdata(dev);
+	ndrd->ns_active = rc;
+	ndrd->ns_count = rc + err;
+
+	if (rc && err && rc == err)
+		return -ENODEV;
+
 	nd_region->btt_seed = nd_btt_create(nd_region);
 	nd_region->pfn_seed = nd_pfn_create(nd_region);
 	nd_region->dax_seed = nd_dax_create(nd_region);
@@ -102,7 +102,7 @@ static int nd_region_remove(struct device *dev)
 	nvdimm_bus_unlock(dev);
 
 	/*
-	 * Note, this assumes device_lock() context to not race
+	 * Note, this assumes nd_device_lock() context to not race
 	 * nd_region_notify()
 	 */
 	sysfs_put(nd_region->bb_state);

drivers/nvdimm/region_devs.c

@@ -331,7 +331,7 @@ static ssize_t set_cookie_show(struct device *dev,
 	 * the v1.1 namespace label cookie definition. To read all this
 	 * data we need to wait for probing to settle.
 	 */
-	device_lock(dev);
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	if (nd_region->ndr_mappings) {
@@ -348,7 +348,7 @@ static ssize_t set_cookie_show(struct device *dev,
 		}
 	}
 	nvdimm_bus_unlock(dev);
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	if (rc)
 		return rc;
@@ -424,10 +424,12 @@ static ssize_t available_size_show(struct device *dev,
 	 * memory nvdimm_bus_lock() is dropped, but that's userspace's
 	 * problem to not race itself.
 	 */
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	available = nd_region_available_dpa(nd_region);
 	nvdimm_bus_unlock(dev);
+	nd_device_unlock(dev);
 
 	return sprintf(buf, "%llu\n", available);
 }
@@ -439,10 +441,12 @@ static ssize_t max_available_extent_show(struct device *dev,
 	struct nd_region *nd_region = to_nd_region(dev);
 	unsigned long long available = 0;
 
+	nd_device_lock(dev);
 	nvdimm_bus_lock(dev);
 	wait_nvdimm_bus_probe_idle(dev);
 	available = nd_region_allocatable_dpa(nd_region);
 	nvdimm_bus_unlock(dev);
+	nd_device_unlock(dev);
 
 	return sprintf(buf, "%llu\n", available);
 }
@@ -561,12 +565,12 @@ static ssize_t region_badblocks_show(struct device *dev,
 	struct nd_region *nd_region = to_nd_region(dev);
 	ssize_t rc;
 
-	device_lock(dev);
+	nd_device_lock(dev);
 	if (dev->driver)
 		rc = badblocks_show(&nd_region->bb, buf, 0);
 	else
 		rc = -ENXIO;
-	device_unlock(dev);
+	nd_device_unlock(dev);
 
 	return rc;
 }

include/linux/device.h

@@ -915,6 +915,8 @@ struct dev_links_info {
  * 		This identifies the device type and carries type-specific
  * 		information.
  * @mutex:	Mutex to synchronize calls to its driver.
+ * @lockdep_mutex: An optional debug lock that a subsystem can use as a
+ * 		peer lock to gain localized lockdep coverage of the device_lock.
  * @bus:	Type of bus device is on.
  * @driver:	Which driver has allocated this
  * @platform_data: Platform data specific to the device.
@@ -998,6 +1000,9 @@ struct device {
 					   core doesn't touch it */
 	void		*driver_data;	/* Driver data, set and get with
 					   dev_set_drvdata/dev_get_drvdata */
+#ifdef CONFIG_PROVE_LOCKING
+	struct mutex		lockdep_mutex;
+#endif
 	struct mutex		mutex;	/* mutex to synchronize calls to
 					 * its driver.
 					 */
@@ -1383,6 +1388,7 @@ extern int (*platform_notify_remove)(struct device *dev);
  */
 extern struct device *get_device(struct device *dev);
 extern void put_device(struct device *dev);
+extern bool kill_device(struct device *dev);
 
 #ifdef CONFIG_DEVTMPFS
 extern int devtmpfs_create_node(struct device *dev);