Linux Audio

Check our new training course

Yocto / OpenEmbedded training

Feb 10-13, 2025
Register
Loading...
v5.9
  1// SPDX-License-Identifier: GPL-2.0-only
  2/*
  3 * Link physical devices with ACPI devices support
  4 *
  5 * Copyright (c) 2005 David Shaohua Li <shaohua.li@intel.com>
  6 * Copyright (c) 2005 Intel Corp.
  7 */
  8
 
 
  9#include <linux/acpi_iort.h>
 10#include <linux/export.h>
 11#include <linux/init.h>
 12#include <linux/list.h>
 13#include <linux/device.h>
 14#include <linux/slab.h>
 15#include <linux/rwsem.h>
 16#include <linux/acpi.h>
 17#include <linux/dma-mapping.h>
 18#include <linux/platform_device.h>
 19
 20#include "internal.h"
 21
 22#define ACPI_GLUE_DEBUG	0
 23#if ACPI_GLUE_DEBUG
 24#define DBG(fmt, ...)						\
 25	printk(KERN_DEBUG PREFIX fmt, ##__VA_ARGS__)
 26#else
 27#define DBG(fmt, ...)						\
 28do {								\
 29	if (0)							\
 30		printk(KERN_DEBUG PREFIX fmt, ##__VA_ARGS__);	\
 31} while (0)
 32#endif
 33static LIST_HEAD(bus_type_list);
 34static DECLARE_RWSEM(bus_type_sem);
 35
 36#define PHYSICAL_NODE_STRING "physical_node"
 37#define PHYSICAL_NODE_NAME_SIZE (sizeof(PHYSICAL_NODE_STRING) + 10)
 38
 39int register_acpi_bus_type(struct acpi_bus_type *type)
 40{
 41	if (acpi_disabled)
 42		return -ENODEV;
 43	if (type && type->match && type->find_companion) {
 44		down_write(&bus_type_sem);
 45		list_add_tail(&type->list, &bus_type_list);
 46		up_write(&bus_type_sem);
 47		printk(KERN_INFO PREFIX "bus type %s registered\n", type->name);
 48		return 0;
 49	}
 50	return -ENODEV;
 51}
 52EXPORT_SYMBOL_GPL(register_acpi_bus_type);
 53
 54int unregister_acpi_bus_type(struct acpi_bus_type *type)
 55{
 56	if (acpi_disabled)
 57		return 0;
 58	if (type) {
 59		down_write(&bus_type_sem);
 60		list_del_init(&type->list);
 61		up_write(&bus_type_sem);
 62		printk(KERN_INFO PREFIX "bus type %s unregistered\n",
 63		       type->name);
 64		return 0;
 65	}
 66	return -ENODEV;
 67}
 68EXPORT_SYMBOL_GPL(unregister_acpi_bus_type);
 69
 70static struct acpi_bus_type *acpi_get_bus_type(struct device *dev)
 71{
 72	struct acpi_bus_type *tmp, *ret = NULL;
 73
 74	down_read(&bus_type_sem);
 75	list_for_each_entry(tmp, &bus_type_list, list) {
 76		if (tmp->match(dev)) {
 77			ret = tmp;
 78			break;
 79		}
 80	}
 81	up_read(&bus_type_sem);
 82	return ret;
 83}
 84
 85#define FIND_CHILD_MIN_SCORE	1
 86#define FIND_CHILD_MAX_SCORE	2
 87
 88static int find_child_checks(struct acpi_device *adev, bool check_children)
 89{
 90	bool sta_present = true;
 91	unsigned long long sta;
 92	acpi_status status;
 93
 94	status = acpi_evaluate_integer(adev->handle, "_STA", NULL, &sta);
 95	if (status == AE_NOT_FOUND)
 96		sta_present = false;
 97	else if (ACPI_FAILURE(status) || !(sta & ACPI_STA_DEVICE_ENABLED))
 98		return -ENODEV;
 99
100	if (check_children && list_empty(&adev->children))
101		return -ENODEV;
102
103	/*
104	 * If the device has a _HID returning a valid ACPI/PNP device ID, it is
105	 * better to make it look less attractive here, so that the other device
106	 * with the same _ADR value (that may not have a valid device ID) can be
107	 * matched going forward.  [This means a second spec violation in a row,
108	 * so whatever we do here is best effort anyway.]
109	 */
110	return sta_present && !adev->pnp.type.platform_id ?
111			FIND_CHILD_MAX_SCORE : FIND_CHILD_MIN_SCORE;
112}
113
114struct acpi_device *acpi_find_child_device(struct acpi_device *parent,
115					   u64 address, bool check_children)
116{
117	struct acpi_device *adev, *ret = NULL;
118	int ret_score = 0;
119
120	if (!parent)
121		return NULL;
122
123	list_for_each_entry(adev, &parent->children, node) {
124		unsigned long long addr;
125		acpi_status status;
126		int score;
127
128		status = acpi_evaluate_integer(adev->handle, METHOD_NAME__ADR,
129					       NULL, &addr);
130		if (ACPI_FAILURE(status) || addr != address)
131			continue;
132
133		if (!ret) {
134			/* This is the first matching object.  Save it. */
135			ret = adev;
136			continue;
137		}
138		/*
139		 * There is more than one matching device object with the same
140		 * _ADR value.  That really is unexpected, so we are kind of
141		 * beyond the scope of the spec here.  We have to choose which
142		 * one to return, though.
143		 *
144		 * First, check if the previously found object is good enough
145		 * and return it if so.  Second, do the same for the object that
146		 * we've just found.
147		 */
148		if (!ret_score) {
149			ret_score = find_child_checks(ret, check_children);
150			if (ret_score == FIND_CHILD_MAX_SCORE)
151				return ret;
152		}
153		score = find_child_checks(adev, check_children);
154		if (score == FIND_CHILD_MAX_SCORE) {
155			return adev;
156		} else if (score > ret_score) {
157			ret = adev;
158			ret_score = score;
159		}
160	}
161	return ret;
162}
163EXPORT_SYMBOL_GPL(acpi_find_child_device);
164
165static void acpi_physnode_link_name(char *buf, unsigned int node_id)
166{
167	if (node_id > 0)
168		snprintf(buf, PHYSICAL_NODE_NAME_SIZE,
169			 PHYSICAL_NODE_STRING "%u", node_id);
170	else
171		strcpy(buf, PHYSICAL_NODE_STRING);
172}
173
174int acpi_bind_one(struct device *dev, struct acpi_device *acpi_dev)
175{
176	struct acpi_device_physical_node *physical_node, *pn;
177	char physical_node_name[PHYSICAL_NODE_NAME_SIZE];
178	struct list_head *physnode_list;
179	unsigned int node_id;
180	int retval = -EINVAL;
181
182	if (has_acpi_companion(dev)) {
183		if (acpi_dev) {
184			dev_warn(dev, "ACPI companion already set\n");
185			return -EINVAL;
186		} else {
187			acpi_dev = ACPI_COMPANION(dev);
188		}
189	}
190	if (!acpi_dev)
191		return -EINVAL;
192
193	get_device(&acpi_dev->dev);
194	get_device(dev);
195	physical_node = kzalloc(sizeof(*physical_node), GFP_KERNEL);
196	if (!physical_node) {
197		retval = -ENOMEM;
198		goto err;
199	}
200
201	mutex_lock(&acpi_dev->physical_node_lock);
202
203	/*
204	 * Keep the list sorted by node_id so that the IDs of removed nodes can
205	 * be recycled easily.
206	 */
207	physnode_list = &acpi_dev->physical_node_list;
208	node_id = 0;
209	list_for_each_entry(pn, &acpi_dev->physical_node_list, node) {
210		/* Sanity check. */
211		if (pn->dev == dev) {
212			mutex_unlock(&acpi_dev->physical_node_lock);
213
214			dev_warn(dev, "Already associated with ACPI node\n");
215			kfree(physical_node);
216			if (ACPI_COMPANION(dev) != acpi_dev)
217				goto err;
218
219			put_device(dev);
220			put_device(&acpi_dev->dev);
221			return 0;
222		}
223		if (pn->node_id == node_id) {
224			physnode_list = &pn->node;
225			node_id++;
226		}
227	}
228
229	physical_node->node_id = node_id;
230	physical_node->dev = dev;
231	list_add(&physical_node->node, physnode_list);
232	acpi_dev->physical_node_count++;
233
234	if (!has_acpi_companion(dev))
235		ACPI_COMPANION_SET(dev, acpi_dev);
236
237	acpi_physnode_link_name(physical_node_name, node_id);
238	retval = sysfs_create_link(&acpi_dev->dev.kobj, &dev->kobj,
239				   physical_node_name);
240	if (retval)
241		dev_err(&acpi_dev->dev, "Failed to create link %s (%d)\n",
242			physical_node_name, retval);
243
244	retval = sysfs_create_link(&dev->kobj, &acpi_dev->dev.kobj,
245				   "firmware_node");
246	if (retval)
247		dev_err(dev, "Failed to create link firmware_node (%d)\n",
248			retval);
249
250	mutex_unlock(&acpi_dev->physical_node_lock);
251
252	if (acpi_dev->wakeup.flags.valid)
253		device_set_wakeup_capable(dev, true);
254
255	return 0;
256
257 err:
258	ACPI_COMPANION_SET(dev, NULL);
259	put_device(dev);
260	put_device(&acpi_dev->dev);
261	return retval;
262}
263EXPORT_SYMBOL_GPL(acpi_bind_one);
264
265int acpi_unbind_one(struct device *dev)
266{
267	struct acpi_device *acpi_dev = ACPI_COMPANION(dev);
268	struct acpi_device_physical_node *entry;
269
270	if (!acpi_dev)
271		return 0;
272
273	mutex_lock(&acpi_dev->physical_node_lock);
274
275	list_for_each_entry(entry, &acpi_dev->physical_node_list, node)
276		if (entry->dev == dev) {
277			char physnode_name[PHYSICAL_NODE_NAME_SIZE];
278
279			list_del(&entry->node);
280			acpi_dev->physical_node_count--;
281
282			acpi_physnode_link_name(physnode_name, entry->node_id);
283			sysfs_remove_link(&acpi_dev->dev.kobj, physnode_name);
284			sysfs_remove_link(&dev->kobj, "firmware_node");
285			ACPI_COMPANION_SET(dev, NULL);
286			/* Drop references taken by acpi_bind_one(). */
287			put_device(dev);
288			put_device(&acpi_dev->dev);
289			kfree(entry);
290			break;
291		}
292
293	mutex_unlock(&acpi_dev->physical_node_lock);
294	return 0;
295}
296EXPORT_SYMBOL_GPL(acpi_unbind_one);
297
298static int acpi_device_notify(struct device *dev)
299{
300	struct acpi_bus_type *type = acpi_get_bus_type(dev);
301	struct acpi_device *adev;
302	int ret;
303
304	ret = acpi_bind_one(dev, NULL);
305	if (ret && type) {
306		struct acpi_device *adev;
307
308		adev = type->find_companion(dev);
309		if (!adev) {
310			DBG("Unable to get handle for %s\n", dev_name(dev));
311			ret = -ENODEV;
312			goto out;
313		}
314		ret = acpi_bind_one(dev, adev);
315		if (ret)
316			goto out;
317	}
318	adev = ACPI_COMPANION(dev);
319	if (!adev)
320		goto out;
321
322	if (dev_is_platform(dev))
323		acpi_configure_pmsi_domain(dev);
324
325	if (type && type->setup)
326		type->setup(dev);
327	else if (adev->handler && adev->handler->bind)
328		adev->handler->bind(dev);
329
330 out:
331#if ACPI_GLUE_DEBUG
332	if (!ret) {
333		struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
334
335		acpi_get_name(ACPI_HANDLE(dev), ACPI_FULL_PATHNAME, &buffer);
336		DBG("Device %s -> %s\n", dev_name(dev), (char *)buffer.pointer);
337		kfree(buffer.pointer);
338	} else
339		DBG("Device %s -> No ACPI support\n", dev_name(dev));
340#endif
341
342	return ret;
343}
344
345static int acpi_device_notify_remove(struct device *dev)
346{
347	struct acpi_device *adev = ACPI_COMPANION(dev);
348	struct acpi_bus_type *type;
349
350	if (!adev)
351		return 0;
352
353	type = acpi_get_bus_type(dev);
354	if (type && type->cleanup)
355		type->cleanup(dev);
356	else if (adev->handler && adev->handler->unbind)
357		adev->handler->unbind(dev);
358
359	acpi_unbind_one(dev);
360	return 0;
361}
362
363int acpi_platform_notify(struct device *dev, enum kobject_action action)
364{
365	switch (action) {
366	case KOBJ_ADD:
367		acpi_device_notify(dev);
368		break;
369	case KOBJ_REMOVE:
370		acpi_device_notify_remove(dev);
371		break;
372	default:
373		break;
374	}
375	return 0;
376}
v5.14.15
  1// SPDX-License-Identifier: GPL-2.0-only
  2/*
  3 * Link physical devices with ACPI devices support
  4 *
  5 * Copyright (c) 2005 David Shaohua Li <shaohua.li@intel.com>
  6 * Copyright (c) 2005 Intel Corp.
  7 */
  8
  9#define pr_fmt(fmt) "ACPI: " fmt
 10
 11#include <linux/acpi_iort.h>
 12#include <linux/export.h>
 13#include <linux/init.h>
 14#include <linux/list.h>
 15#include <linux/device.h>
 16#include <linux/slab.h>
 17#include <linux/rwsem.h>
 18#include <linux/acpi.h>
 19#include <linux/dma-mapping.h>
 20#include <linux/platform_device.h>
 21
 22#include "internal.h"
 23
 
 
 
 
 
 
 
 
 
 
 
 24static LIST_HEAD(bus_type_list);
 25static DECLARE_RWSEM(bus_type_sem);
 26
 27#define PHYSICAL_NODE_STRING "physical_node"
 28#define PHYSICAL_NODE_NAME_SIZE (sizeof(PHYSICAL_NODE_STRING) + 10)
 29
 30int register_acpi_bus_type(struct acpi_bus_type *type)
 31{
 32	if (acpi_disabled)
 33		return -ENODEV;
 34	if (type && type->match && type->find_companion) {
 35		down_write(&bus_type_sem);
 36		list_add_tail(&type->list, &bus_type_list);
 37		up_write(&bus_type_sem);
 38		pr_info("bus type %s registered\n", type->name);
 39		return 0;
 40	}
 41	return -ENODEV;
 42}
 43EXPORT_SYMBOL_GPL(register_acpi_bus_type);
 44
 45int unregister_acpi_bus_type(struct acpi_bus_type *type)
 46{
 47	if (acpi_disabled)
 48		return 0;
 49	if (type) {
 50		down_write(&bus_type_sem);
 51		list_del_init(&type->list);
 52		up_write(&bus_type_sem);
 53		pr_info("bus type %s unregistered\n", type->name);
 
 54		return 0;
 55	}
 56	return -ENODEV;
 57}
 58EXPORT_SYMBOL_GPL(unregister_acpi_bus_type);
 59
 60static struct acpi_bus_type *acpi_get_bus_type(struct device *dev)
 61{
 62	struct acpi_bus_type *tmp, *ret = NULL;
 63
 64	down_read(&bus_type_sem);
 65	list_for_each_entry(tmp, &bus_type_list, list) {
 66		if (tmp->match(dev)) {
 67			ret = tmp;
 68			break;
 69		}
 70	}
 71	up_read(&bus_type_sem);
 72	return ret;
 73}
 74
 75#define FIND_CHILD_MIN_SCORE	1
 76#define FIND_CHILD_MAX_SCORE	2
 77
 78static int find_child_checks(struct acpi_device *adev, bool check_children)
 79{
 80	bool sta_present = true;
 81	unsigned long long sta;
 82	acpi_status status;
 83
 84	status = acpi_evaluate_integer(adev->handle, "_STA", NULL, &sta);
 85	if (status == AE_NOT_FOUND)
 86		sta_present = false;
 87	else if (ACPI_FAILURE(status) || !(sta & ACPI_STA_DEVICE_ENABLED))
 88		return -ENODEV;
 89
 90	if (check_children && list_empty(&adev->children))
 91		return -ENODEV;
 92
 93	/*
 94	 * If the device has a _HID returning a valid ACPI/PNP device ID, it is
 95	 * better to make it look less attractive here, so that the other device
 96	 * with the same _ADR value (that may not have a valid device ID) can be
 97	 * matched going forward.  [This means a second spec violation in a row,
 98	 * so whatever we do here is best effort anyway.]
 99	 */
100	return sta_present && !adev->pnp.type.platform_id ?
101			FIND_CHILD_MAX_SCORE : FIND_CHILD_MIN_SCORE;
102}
103
104struct acpi_device *acpi_find_child_device(struct acpi_device *parent,
105					   u64 address, bool check_children)
106{
107	struct acpi_device *adev, *ret = NULL;
108	int ret_score = 0;
109
110	if (!parent)
111		return NULL;
112
113	list_for_each_entry(adev, &parent->children, node) {
114		unsigned long long addr;
115		acpi_status status;
116		int score;
117
118		status = acpi_evaluate_integer(adev->handle, METHOD_NAME__ADR,
119					       NULL, &addr);
120		if (ACPI_FAILURE(status) || addr != address)
121			continue;
122
123		if (!ret) {
124			/* This is the first matching object.  Save it. */
125			ret = adev;
126			continue;
127		}
128		/*
129		 * There is more than one matching device object with the same
130		 * _ADR value.  That really is unexpected, so we are kind of
131		 * beyond the scope of the spec here.  We have to choose which
132		 * one to return, though.
133		 *
134		 * First, check if the previously found object is good enough
135		 * and return it if so.  Second, do the same for the object that
136		 * we've just found.
137		 */
138		if (!ret_score) {
139			ret_score = find_child_checks(ret, check_children);
140			if (ret_score == FIND_CHILD_MAX_SCORE)
141				return ret;
142		}
143		score = find_child_checks(adev, check_children);
144		if (score == FIND_CHILD_MAX_SCORE) {
145			return adev;
146		} else if (score > ret_score) {
147			ret = adev;
148			ret_score = score;
149		}
150	}
151	return ret;
152}
153EXPORT_SYMBOL_GPL(acpi_find_child_device);
154
155static void acpi_physnode_link_name(char *buf, unsigned int node_id)
156{
157	if (node_id > 0)
158		snprintf(buf, PHYSICAL_NODE_NAME_SIZE,
159			 PHYSICAL_NODE_STRING "%u", node_id);
160	else
161		strcpy(buf, PHYSICAL_NODE_STRING);
162}
163
164int acpi_bind_one(struct device *dev, struct acpi_device *acpi_dev)
165{
166	struct acpi_device_physical_node *physical_node, *pn;
167	char physical_node_name[PHYSICAL_NODE_NAME_SIZE];
168	struct list_head *physnode_list;
169	unsigned int node_id;
170	int retval = -EINVAL;
171
172	if (has_acpi_companion(dev)) {
173		if (acpi_dev) {
174			dev_warn(dev, "ACPI companion already set\n");
175			return -EINVAL;
176		} else {
177			acpi_dev = ACPI_COMPANION(dev);
178		}
179	}
180	if (!acpi_dev)
181		return -EINVAL;
182
183	acpi_dev_get(acpi_dev);
184	get_device(dev);
185	physical_node = kzalloc(sizeof(*physical_node), GFP_KERNEL);
186	if (!physical_node) {
187		retval = -ENOMEM;
188		goto err;
189	}
190
191	mutex_lock(&acpi_dev->physical_node_lock);
192
193	/*
194	 * Keep the list sorted by node_id so that the IDs of removed nodes can
195	 * be recycled easily.
196	 */
197	physnode_list = &acpi_dev->physical_node_list;
198	node_id = 0;
199	list_for_each_entry(pn, &acpi_dev->physical_node_list, node) {
200		/* Sanity check. */
201		if (pn->dev == dev) {
202			mutex_unlock(&acpi_dev->physical_node_lock);
203
204			dev_warn(dev, "Already associated with ACPI node\n");
205			kfree(physical_node);
206			if (ACPI_COMPANION(dev) != acpi_dev)
207				goto err;
208
209			put_device(dev);
210			acpi_dev_put(acpi_dev);
211			return 0;
212		}
213		if (pn->node_id == node_id) {
214			physnode_list = &pn->node;
215			node_id++;
216		}
217	}
218
219	physical_node->node_id = node_id;
220	physical_node->dev = dev;
221	list_add(&physical_node->node, physnode_list);
222	acpi_dev->physical_node_count++;
223
224	if (!has_acpi_companion(dev))
225		ACPI_COMPANION_SET(dev, acpi_dev);
226
227	acpi_physnode_link_name(physical_node_name, node_id);
228	retval = sysfs_create_link(&acpi_dev->dev.kobj, &dev->kobj,
229				   physical_node_name);
230	if (retval)
231		dev_err(&acpi_dev->dev, "Failed to create link %s (%d)\n",
232			physical_node_name, retval);
233
234	retval = sysfs_create_link(&dev->kobj, &acpi_dev->dev.kobj,
235				   "firmware_node");
236	if (retval)
237		dev_err(dev, "Failed to create link firmware_node (%d)\n",
238			retval);
239
240	mutex_unlock(&acpi_dev->physical_node_lock);
241
242	if (acpi_dev->wakeup.flags.valid)
243		device_set_wakeup_capable(dev, true);
244
245	return 0;
246
247 err:
248	ACPI_COMPANION_SET(dev, NULL);
249	put_device(dev);
250	acpi_dev_put(acpi_dev);
251	return retval;
252}
253EXPORT_SYMBOL_GPL(acpi_bind_one);
254
255int acpi_unbind_one(struct device *dev)
256{
257	struct acpi_device *acpi_dev = ACPI_COMPANION(dev);
258	struct acpi_device_physical_node *entry;
259
260	if (!acpi_dev)
261		return 0;
262
263	mutex_lock(&acpi_dev->physical_node_lock);
264
265	list_for_each_entry(entry, &acpi_dev->physical_node_list, node)
266		if (entry->dev == dev) {
267			char physnode_name[PHYSICAL_NODE_NAME_SIZE];
268
269			list_del(&entry->node);
270			acpi_dev->physical_node_count--;
271
272			acpi_physnode_link_name(physnode_name, entry->node_id);
273			sysfs_remove_link(&acpi_dev->dev.kobj, physnode_name);
274			sysfs_remove_link(&dev->kobj, "firmware_node");
275			ACPI_COMPANION_SET(dev, NULL);
276			/* Drop references taken by acpi_bind_one(). */
277			put_device(dev);
278			acpi_dev_put(acpi_dev);
279			kfree(entry);
280			break;
281		}
282
283	mutex_unlock(&acpi_dev->physical_node_lock);
284	return 0;
285}
286EXPORT_SYMBOL_GPL(acpi_unbind_one);
287
288static int acpi_device_notify(struct device *dev)
289{
290	struct acpi_bus_type *type = acpi_get_bus_type(dev);
291	struct acpi_device *adev;
292	int ret;
293
294	ret = acpi_bind_one(dev, NULL);
295	if (ret && type) {
296		struct acpi_device *adev;
297
298		adev = type->find_companion(dev);
299		if (!adev) {
300			pr_debug("Unable to get handle for %s\n", dev_name(dev));
301			ret = -ENODEV;
302			goto out;
303		}
304		ret = acpi_bind_one(dev, adev);
305		if (ret)
306			goto out;
307	}
308	adev = ACPI_COMPANION(dev);
309	if (!adev)
310		goto out;
311
312	if (dev_is_platform(dev))
313		acpi_configure_pmsi_domain(dev);
314
315	if (type && type->setup)
316		type->setup(dev);
317	else if (adev->handler && adev->handler->bind)
318		adev->handler->bind(dev);
319
320 out:
 
321	if (!ret) {
322		struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL };
323
324		acpi_get_name(ACPI_HANDLE(dev), ACPI_FULL_PATHNAME, &buffer);
325		pr_debug("Device %s -> %s\n", dev_name(dev), (char *)buffer.pointer);
326		kfree(buffer.pointer);
327	} else {
328		pr_debug("Device %s -> No ACPI support\n", dev_name(dev));
329	}
330
331	return ret;
332}
333
334static int acpi_device_notify_remove(struct device *dev)
335{
336	struct acpi_device *adev = ACPI_COMPANION(dev);
337	struct acpi_bus_type *type;
338
339	if (!adev)
340		return 0;
341
342	type = acpi_get_bus_type(dev);
343	if (type && type->cleanup)
344		type->cleanup(dev);
345	else if (adev->handler && adev->handler->unbind)
346		adev->handler->unbind(dev);
347
348	acpi_unbind_one(dev);
349	return 0;
350}
351
352int acpi_platform_notify(struct device *dev, enum kobject_action action)
353{
354	switch (action) {
355	case KOBJ_ADD:
356		acpi_device_notify(dev);
357		break;
358	case KOBJ_REMOVE:
359		acpi_device_notify_remove(dev);
360		break;
361	default:
362		break;
363	}
364	return 0;
365}