aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorRafael J. Wysocki <rafael.j.wysocki@intel.com>2013-06-28 06:58:05 -0400
committerRafael J. Wysocki <rafael.j.wysocki@intel.com>2013-06-28 06:58:05 -0400
commita204dbc61b7f4cb1a7e2cb3ad057b135164782da (patch)
treef82151c04a30f49c3dee8926d184575ad2e7b1e2
parent45e00374db944b1c12987b501bcaa279b3e36d93 (diff)
parent08f502c1c343031f0d126bd00e87dede38269d12 (diff)
Merge branch 'acpi-hotplug'
* acpi-hotplug: ACPI: Do not use CONFIG_ACPI_HOTPLUG_MEMORY_MODULE ACPI / cpufreq: Add ACPI processor device IDs to acpi-cpufreq Memory hotplug: Move alternative function definitions to header ACPI / processor: Fix potential NULL pointer dereference in acpi_processor_add() Memory hotplug / ACPI: Simplify memory removal ACPI / scan: Add second pass of companion offlining to hot-remove code Driver core / MM: Drop offline_memory_block() ACPI / processor: Pass processor object handle to acpi_bind_one() ACPI: Drop removal_type field from struct acpi_device Driver core / memory: Simplify __memory_block_change_state() ACPI / processor: Initialize per_cpu(processors, pr->id) properly CPU: Fix sysfs cpu/online of offlined CPUs Driver core: Introduce offline/online callbacks for memory blocks ACPI / memhotplug: Bind removable memory blocks to ACPI device nodes ACPI / processor: Use common hotplug infrastructure ACPI / hotplug: Use device offline/online for graceful hot-removal Driver core: Use generic offline/online for CPU offline/online Driver core: Add offline/online device operations
Diffstat
-rw-r--r--Documentation/ABI/testing/sysfs-devices-online20
-rw-r--r--Documentation/ABI/testing/sysfs-firmware-acpi10
-rw-r--r--drivers/acpi/Makefile1
-rw-r--r--drivers/acpi/acpi_memhotplug.c62
-rw-r--r--drivers/acpi/acpi_processor.c494
-rw-r--r--drivers/acpi/glue.c6
-rw-r--r--drivers/acpi/internal.h5
-rw-r--r--drivers/acpi/processor_driver.c812
-rw-r--r--drivers/acpi/scan.c120
-rw-r--r--drivers/acpi/sysfs.c31
-rw-r--r--drivers/base/core.c130
-rw-r--r--drivers/base/cpu.c101
-rw-r--r--drivers/base/memory.c114
-rw-r--r--drivers/cpufreq/acpi-cpufreq.c7
-rw-r--r--include/acpi/acpi_bus.h9
-rw-r--r--include/acpi/processor.h5
-rw-r--r--include/linux/acpi.h3
-rw-r--r--include/linux/device.h21
-rw-r--r--include/linux/memory_hotplug.h14
-rw-r--r--mm/memory_hotplug.c81
20 files changed, 1118 insertions, 928 deletions
diff --git a/Documentation/ABI/testing/sysfs-devices-online b/Documentation/ABI/testing/sysfs-devices-online
new file mode 100644
index 000000000000..f990026c0740
--- /dev/null
+++ b/Documentation/ABI/testing/sysfs-devices-online
@@ -0,0 +1,20 @@
1What: /sys/devices/.../online
2Date: April 2013
3Contact: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
4Description:
5 The /sys/devices/.../online attribute is only present for
6 devices whose bus types provide .online() and .offline()
7 callbacks. The number read from it (0 or 1) reflects the value
8 of the device's 'offline' field. If that number is 1 and '0'
9 (or 'n', or 'N') is written to this file, the device bus type's
10 .offline() callback is executed for the device and (if
11 successful) its 'offline' field is updated accordingly. In
12 turn, if that number is 0 and '1' (or 'y', or 'Y') is written to
13 this file, the device bus type's .online() callback is executed
14 for the device and (if successful) its 'offline' field is
15 updated as appropriate.
16
17 After a successful execution of the bus type's .offline()
18 callback the device cannot be used for any purpose until either
19 it is removed (i.e. device_del() is called for it), or its bus
20 type's .online() is exeucted successfully.
diff --git a/Documentation/ABI/testing/sysfs-firmware-acpi b/Documentation/ABI/testing/sysfs-firmware-acpi
index ce9bee98b43b..b4436cca97a8 100644
--- a/Documentation/ABI/testing/sysfs-firmware-acpi
+++ b/Documentation/ABI/testing/sysfs-firmware-acpi
@@ -44,6 +44,16 @@ Description:
44 or 0 (unset). Attempts to write any other values to it will 44 or 0 (unset). Attempts to write any other values to it will
45 cause -EINVAL to be returned. 45 cause -EINVAL to be returned.
46 46
47What: /sys/firmware/acpi/hotplug/force_remove
48Date: May 2013
49Contact: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
50Description:
51 The number in this file (0 or 1) determines whether (1) or not
52 (0) the ACPI subsystem will allow devices to be hot-removed even
53 if they cannot be put offline gracefully (from the kernel's
54 viewpoint). That number can be changed by writing a boolean
55 value to this file.
56
47What: /sys/firmware/acpi/interrupts/ 57What: /sys/firmware/acpi/interrupts/
48Date: February 2008 58Date: February 2008
49Contact: Len Brown <lenb@kernel.org> 59Contact: Len Brown <lenb@kernel.org>
diff --git a/drivers/acpi/Makefile b/drivers/acpi/Makefile
index 536562c626a2..d07771bc3d8c 100644
--- a/drivers/acpi/Makefile
+++ b/drivers/acpi/Makefile
@@ -34,6 +34,7 @@ acpi-$(CONFIG_ACPI_SLEEP) += proc.o
34acpi-y += bus.o glue.o 34acpi-y += bus.o glue.o
35acpi-y += scan.o 35acpi-y += scan.o
36acpi-y += resource.o 36acpi-y += resource.o
37acpi-y += acpi_processor.o
37acpi-y += processor_core.o 38acpi-y += processor_core.o
38acpi-y += ec.o 39acpi-y += ec.o
39acpi-$(CONFIG_ACPI_DOCK) += dock.o 40acpi-$(CONFIG_ACPI_DOCK) += dock.o
diff --git a/drivers/acpi/acpi_memhotplug.c b/drivers/acpi/acpi_memhotplug.c
index 5e6301e94920..c711d1144044 100644
--- a/drivers/acpi/acpi_memhotplug.c
+++ b/drivers/acpi/acpi_memhotplug.c
@@ -28,6 +28,7 @@
28 */ 28 */
29 29
30#include <linux/acpi.h> 30#include <linux/acpi.h>
31#include <linux/memory.h>
31#include <linux/memory_hotplug.h> 32#include <linux/memory_hotplug.h>
32 33
33#include "internal.h" 34#include "internal.h"
@@ -166,13 +167,50 @@ static int acpi_memory_check_device(struct acpi_memory_device *mem_device)
166 return 0; 167 return 0;
167} 168}
168 169
170static unsigned long acpi_meminfo_start_pfn(struct acpi_memory_info *info)
171{
172 return PFN_DOWN(info->start_addr);
173}
174
175static unsigned long acpi_meminfo_end_pfn(struct acpi_memory_info *info)
176{
177 return PFN_UP(info->start_addr + info->length-1);
178}
179
180static int acpi_bind_memblk(struct memory_block *mem, void *arg)
181{
182 return acpi_bind_one(&mem->dev, (acpi_handle)arg);
183}
184
185static int acpi_bind_memory_blocks(struct acpi_memory_info *info,
186 acpi_handle handle)
187{
188 return walk_memory_range(acpi_meminfo_start_pfn(info),
189 acpi_meminfo_end_pfn(info), (void *)handle,
190 acpi_bind_memblk);
191}
192
193static int acpi_unbind_memblk(struct memory_block *mem, void *arg)
194{
195 acpi_unbind_one(&mem->dev);
196 return 0;
197}
198
199static void acpi_unbind_memory_blocks(struct acpi_memory_info *info,
200 acpi_handle handle)
201{
202 walk_memory_range(acpi_meminfo_start_pfn(info),
203 acpi_meminfo_end_pfn(info), NULL, acpi_unbind_memblk);
204}
205
169static int acpi_memory_enable_device(struct acpi_memory_device *mem_device) 206static int acpi_memory_enable_device(struct acpi_memory_device *mem_device)
170{ 207{
208 acpi_handle handle = mem_device->device->handle;
171 int result, num_enabled = 0; 209 int result, num_enabled = 0;
172 struct acpi_memory_info *info; 210 struct acpi_memory_info *info;
173 int node; 211 int node;
174 212
175 node = acpi_get_node(mem_device->device->handle); 213 node = acpi_get_node(handle);
176 /* 214 /*
177 * Tell the VM there is more memory here... 215 * Tell the VM there is more memory here...
178 * Note: Assume that this function returns zero on success 216 * Note: Assume that this function returns zero on success
@@ -203,6 +241,12 @@ static int acpi_memory_enable_device(struct acpi_memory_device *mem_device)
203 if (result && result != -EEXIST) 241 if (result && result != -EEXIST)
204 continue; 242 continue;
205 243
244 result = acpi_bind_memory_blocks(info, handle);
245 if (result) {
246 acpi_unbind_memory_blocks(info, handle);
247 return -ENODEV;
248 }
249
206 info->enabled = 1; 250 info->enabled = 1;
207 251
208 /* 252 /*
@@ -227,12 +271,11 @@ static int acpi_memory_enable_device(struct acpi_memory_device *mem_device)
227 return 0; 271 return 0;
228} 272}
229 273
230static int acpi_memory_remove_memory(struct acpi_memory_device *mem_device) 274static void acpi_memory_remove_memory(struct acpi_memory_device *mem_device)
231{ 275{
232 int result = 0, nid; 276 acpi_handle handle = mem_device->device->handle;
233 struct acpi_memory_info *info, *n; 277 struct acpi_memory_info *info, *n;
234 278 int nid = acpi_get_node(handle);
235 nid = acpi_get_node(mem_device->device->handle);
236 279
237 list_for_each_entry_safe(info, n, &mem_device->res_list, list) { 280 list_for_each_entry_safe(info, n, &mem_device->res_list, list) {
238 if (!info->enabled) 281 if (!info->enabled)
@@ -240,15 +283,12 @@ static int acpi_memory_remove_memory(struct acpi_memory_device *mem_device)
240 283
241 if (nid < 0) 284 if (nid < 0)
242 nid = memory_add_physaddr_to_nid(info->start_addr); 285 nid = memory_add_physaddr_to_nid(info->start_addr);
243 result = remove_memory(nid, info->start_addr, info->length);
244 if (result)
245 return result;
246 286
287 acpi_unbind_memory_blocks(info, handle);
288 remove_memory(nid, info->start_addr, info->length);
247 list_del(&info->list); 289 list_del(&info->list);
248 kfree(info); 290 kfree(info);
249 } 291 }
250
251 return result;
252} 292}
253 293
254static void acpi_memory_device_free(struct acpi_memory_device *mem_device) 294static void acpi_memory_device_free(struct acpi_memory_device *mem_device)
@@ -300,7 +340,7 @@ static int acpi_memory_device_add(struct acpi_device *device,
300 if (result) { 340 if (result) {
301 dev_err(&device->dev, "acpi_memory_enable_device() error\n"); 341 dev_err(&device->dev, "acpi_memory_enable_device() error\n");
302 acpi_memory_device_free(mem_device); 342 acpi_memory_device_free(mem_device);
303 return -ENODEV; 343 return result;
304 } 344 }
305 345
306 dev_dbg(&device->dev, "Memory device configured by ACPI\n"); 346 dev_dbg(&device->dev, "Memory device configured by ACPI\n");
diff --git a/drivers/acpi/acpi_processor.c b/drivers/acpi/acpi_processor.c
new file mode 100644
index 000000000000..e9b01e35ac37
--- /dev/null
+++ b/drivers/acpi/acpi_processor.c
@@ -0,0 +1,494 @@
1/*
2 * acpi_processor.c - ACPI processor enumeration support
3 *
4 * Copyright (C) 2001, 2002 Andy Grover <andrew.grover@intel.com>
5 * Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com>
6 * Copyright (C) 2004 Dominik Brodowski <linux@brodo.de>
7 * Copyright (C) 2004 Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
8 * Copyright (C) 2013, Intel Corporation
9 * Rafael J. Wysocki <rafael.j.wysocki@intel.com>
10 *
11 * This program is free software; you can redistribute it and/or modify it
12 * under the terms of the GNU General Public License version 2 as published
13 * by the Free Software Foundation.
14 */
15
16#include <linux/acpi.h>
17#include <linux/device.h>
18#include <linux/kernel.h>
19#include <linux/module.h>
20#include <linux/pci.h>
21
22#include <acpi/processor.h>
23
24#include <asm/cpu.h>
25
26#include "internal.h"
27
28#define _COMPONENT ACPI_PROCESSOR_COMPONENT
29
30ACPI_MODULE_NAME("processor");
31
32DEFINE_PER_CPU(struct acpi_processor *, processors);
33EXPORT_PER_CPU_SYMBOL(processors);
34
35/* --------------------------------------------------------------------------
36 Errata Handling
37 -------------------------------------------------------------------------- */
38
39struct acpi_processor_errata errata __read_mostly;
40EXPORT_SYMBOL_GPL(errata);
41
42static int acpi_processor_errata_piix4(struct pci_dev *dev)
43{
44 u8 value1 = 0;
45 u8 value2 = 0;
46
47
48 if (!dev)
49 return -EINVAL;
50
51 /*
52 * Note that 'dev' references the PIIX4 ACPI Controller.
53 */
54
55 switch (dev->revision) {
56 case 0:
57 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4 A-step\n"));
58 break;
59 case 1:
60 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4 B-step\n"));
61 break;
62 case 2:
63 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4E\n"));
64 break;
65 case 3:
66 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4M\n"));
67 break;
68 default:
69 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found unknown PIIX4\n"));
70 break;
71 }
72
73 switch (dev->revision) {
74
75 case 0: /* PIIX4 A-step */
76 case 1: /* PIIX4 B-step */
77 /*
78 * See specification changes #13 ("Manual Throttle Duty Cycle")
79 * and #14 ("Enabling and Disabling Manual Throttle"), plus
80 * erratum #5 ("STPCLK# Deassertion Time") from the January
81 * 2002 PIIX4 specification update. Applies to only older
82 * PIIX4 models.
83 */
84 errata.piix4.throttle = 1;
85
86 case 2: /* PIIX4E */
87 case 3: /* PIIX4M */
88 /*
89 * See erratum #18 ("C3 Power State/BMIDE and Type-F DMA
90 * Livelock") from the January 2002 PIIX4 specification update.
91 * Applies to all PIIX4 models.
92 */
93
94 /*
95 * BM-IDE
96 * ------
97 * Find the PIIX4 IDE Controller and get the Bus Master IDE
98 * Status register address. We'll use this later to read
99 * each IDE controller's DMA status to make sure we catch all
100 * DMA activity.
101 */
102 dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
103 PCI_DEVICE_ID_INTEL_82371AB,
104 PCI_ANY_ID, PCI_ANY_ID, NULL);
105 if (dev) {
106 errata.piix4.bmisx = pci_resource_start(dev, 4);
107 pci_dev_put(dev);
108 }
109
110 /*
111 * Type-F DMA
112 * ----------
113 * Find the PIIX4 ISA Controller and read the Motherboard
114 * DMA controller's status to see if Type-F (Fast) DMA mode
115 * is enabled (bit 7) on either channel. Note that we'll
116 * disable C3 support if this is enabled, as some legacy
117 * devices won't operate well if fast DMA is disabled.
118 */
119 dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
120 PCI_DEVICE_ID_INTEL_82371AB_0,
121 PCI_ANY_ID, PCI_ANY_ID, NULL);
122 if (dev) {
123 pci_read_config_byte(dev, 0x76, &value1);
124 pci_read_config_byte(dev, 0x77, &value2);
125 if ((value1 & 0x80) || (value2 & 0x80))
126 errata.piix4.fdma = 1;
127 pci_dev_put(dev);
128 }
129
130 break;
131 }
132
133 if (errata.piix4.bmisx)
134 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
135 "Bus master activity detection (BM-IDE) erratum enabled\n"));
136 if (errata.piix4.fdma)
137 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
138 "Type-F DMA livelock erratum (C3 disabled)\n"));
139
140 return 0;
141}
142
143static int acpi_processor_errata(struct acpi_processor *pr)
144{
145 int result = 0;
146 struct pci_dev *dev = NULL;
147
148
149 if (!pr)
150 return -EINVAL;
151
152 /*
153 * PIIX4
154 */
155 dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
156 PCI_DEVICE_ID_INTEL_82371AB_3, PCI_ANY_ID,
157 PCI_ANY_ID, NULL);
158 if (dev) {
159 result = acpi_processor_errata_piix4(dev);
160 pci_dev_put(dev);
161 }
162
163 return result;
164}
165
166/* --------------------------------------------------------------------------
167 Initialization
168 -------------------------------------------------------------------------- */
169
170#ifdef CONFIG_ACPI_HOTPLUG_CPU
171static int acpi_processor_hotadd_init(struct acpi_processor *pr)
172{
173 unsigned long long sta;
174 acpi_status status;
175 int ret;
176
177 status = acpi_evaluate_integer(pr->handle, "_STA", NULL, &sta);
178 if (ACPI_FAILURE(status) || !(sta & ACPI_STA_DEVICE_PRESENT))
179 return -ENODEV;
180
181 ret = acpi_map_lsapic(pr->handle, &pr->id);
182 if (ret)
183 return ret;
184
185 ret = arch_register_cpu(pr->id);
186 if (ret) {
187 acpi_unmap_lsapic(pr->id);
188 return ret;
189 }
190
191 /*
192 * CPU got hot-added, but cpu_data is not initialized yet. Set a flag
193 * to delay cpu_idle/throttling initialization and do it when the CPU
194 * gets online for the first time.
195 */
196 pr_info("CPU%d has been hot-added\n", pr->id);
197 pr->flags.need_hotplug_init = 1;
198 return 0;
199}
200#else
201static inline int acpi_processor_hotadd_init(struct acpi_processor *pr)
202{
203 return -ENODEV;
204}
205#endif /* CONFIG_ACPI_HOTPLUG_CPU */
206
207static int acpi_processor_get_info(struct acpi_device *device)
208{
209 union acpi_object object = { 0 };
210 struct acpi_buffer buffer = { sizeof(union acpi_object), &object };
211 struct acpi_processor *pr = acpi_driver_data(device);
212 int cpu_index, device_declaration = 0;
213 acpi_status status = AE_OK;
214 static int cpu0_initialized;
215
216 if (num_online_cpus() > 1)
217 errata.smp = TRUE;
218
219 acpi_processor_errata(pr);
220
221 /*
222 * Check to see if we have bus mastering arbitration control. This
223 * is required for proper C3 usage (to maintain cache coherency).
224 */
225 if (acpi_gbl_FADT.pm2_control_block && acpi_gbl_FADT.pm2_control_length) {
226 pr->flags.bm_control = 1;
227 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
228 "Bus mastering arbitration control present\n"));
229 } else
230 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
231 "No bus mastering arbitration control\n"));
232
233 if (!strcmp(acpi_device_hid(device), ACPI_PROCESSOR_OBJECT_HID)) {
234 /* Declared with "Processor" statement; match ProcessorID */
235 status = acpi_evaluate_object(pr->handle, NULL, NULL, &buffer);
236 if (ACPI_FAILURE(status)) {
237 dev_err(&device->dev,
238 "Failed to evaluate processor object (0x%x)\n",
239 status);
240 return -ENODEV;
241 }
242
243 /*
244 * TBD: Synch processor ID (via LAPIC/LSAPIC structures) on SMP.
245 * >>> 'acpi_get_processor_id(acpi_id, &id)' in
246 * arch/xxx/acpi.c
247 */
248 pr->acpi_id = object.processor.proc_id;
249 } else {
250 /*
251 * Declared with "Device" statement; match _UID.
252 * Note that we don't handle string _UIDs yet.
253 */
254 unsigned long long value;
255 status = acpi_evaluate_integer(pr->handle, METHOD_NAME__UID,
256 NULL, &value);
257 if (ACPI_FAILURE(status)) {
258 dev_err(&device->dev,
259 "Failed to evaluate processor _UID (0x%x)\n",
260 status);
261 return -ENODEV;
262 }
263 device_declaration = 1;
264 pr->acpi_id = value;
265 }
266 cpu_index = acpi_get_cpuid(pr->handle, device_declaration, pr->acpi_id);
267
268 /* Handle UP system running SMP kernel, with no LAPIC in MADT */
269 if (!cpu0_initialized && (cpu_index == -1) &&
270 (num_online_cpus() == 1)) {
271 cpu_index = 0;
272 }
273
274 cpu0_initialized = 1;
275
276 pr->id = cpu_index;
277
278 /*
279 * Extra Processor objects may be enumerated on MP systems with
280 * less than the max # of CPUs. They should be ignored _iff
281 * they are physically not present.
282 */
283 if (pr->id == -1) {
284 int ret = acpi_processor_hotadd_init(pr);
285 if (ret)
286 return ret;
287 }
288 /*
289 * On some boxes several processors use the same processor bus id.
290 * But they are located in different scope. For example:
291 * \_SB.SCK0.CPU0
292 * \_SB.SCK1.CPU0
293 * Rename the processor device bus id. And the new bus id will be
294 * generated as the following format:
295 * CPU+CPU ID.
296 */
297 sprintf(acpi_device_bid(device), "CPU%X", pr->id);
298 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Processor [%d:%d]\n", pr->id,
299 pr->acpi_id));
300
301 if (!object.processor.pblk_address)
302 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "No PBLK (NULL address)\n"));
303 else if (object.processor.pblk_length != 6)
304 dev_err(&device->dev, "Invalid PBLK length [%d]\n",
305 object.processor.pblk_length);
306 else {
307 pr->throttling.address = object.processor.pblk_address;
308 pr->throttling.duty_offset = acpi_gbl_FADT.duty_offset;
309 pr->throttling.duty_width = acpi_gbl_FADT.duty_width;
310
311 pr->pblk = object.processor.pblk_address;
312
313 /*
314 * We don't care about error returns - we just try to mark
315 * these reserved so that nobody else is confused into thinking
316 * that this region might be unused..
317 *
318 * (In particular, allocating the IO range for Cardbus)
319 */
320 request_region(pr->throttling.address, 6, "ACPI CPU throttle");
321 }
322
323 /*
324 * If ACPI describes a slot number for this CPU, we can use it to
325 * ensure we get the right value in the "physical id" field
326 * of /proc/cpuinfo
327 */
328 status = acpi_evaluate_object(pr->handle, "_SUN", NULL, &buffer);
329 if (ACPI_SUCCESS(status))
330 arch_fix_phys_package_id(pr->id, object.integer.value);
331
332 return 0;
333}
334
335/*
336 * Do not put anything in here which needs the core to be online.
337 * For example MSR access or setting up things which check for cpuinfo_x86
338 * (cpu_data(cpu)) values, like CPU feature flags, family, model, etc.
339 * Such things have to be put in and set up by the processor driver's .probe().
340 */
341static DEFINE_PER_CPU(void *, processor_device_array);
342
343static int __cpuinit acpi_processor_add(struct acpi_device *device,
344 const struct acpi_device_id *id)
345{
346 struct acpi_processor *pr;
347 struct device *dev;
348 int result = 0;
349
350 pr = kzalloc(sizeof(struct acpi_processor), GFP_KERNEL);
351 if (!pr)
352 return -ENOMEM;
353
354 if (!zalloc_cpumask_var(&pr->throttling.shared_cpu_map, GFP_KERNEL)) {
355 result = -ENOMEM;
356 goto err_free_pr;
357 }
358
359 pr->handle = device->handle;
360 strcpy(acpi_device_name(device), ACPI_PROCESSOR_DEVICE_NAME);
361 strcpy(acpi_device_class(device), ACPI_PROCESSOR_CLASS);
362 device->driver_data = pr;
363
364 result = acpi_processor_get_info(device);
365 if (result) /* Processor is not physically present or unavailable */
366 return 0;
367
368#ifdef CONFIG_SMP
369 if (pr->id >= setup_max_cpus && pr->id != 0)
370 return 0;
371#endif
372
373 BUG_ON(pr->id >= nr_cpu_ids);
374
375 /*
376 * Buggy BIOS check.
377 * ACPI id of processors can be reported wrongly by the BIOS.
378 * Don't trust it blindly
379 */
380 if (per_cpu(processor_device_array, pr->id) != NULL &&
381 per_cpu(processor_device_array, pr->id) != device) {
382 dev_warn(&device->dev,
383 "BIOS reported wrong ACPI id %d for the processor\n",
384 pr->id);
385 /* Give up, but do not abort the namespace scan. */
386 goto err;
387 }
388 /*
389 * processor_device_array is not cleared on errors to allow buggy BIOS
390 * checks.
391 */
392 per_cpu(processor_device_array, pr->id) = device;
393 per_cpu(processors, pr->id) = pr;
394
395 dev = get_cpu_device(pr->id);
396 if (!dev) {
397 result = -ENODEV;
398 goto err;
399 }
400
401 result = acpi_bind_one(dev, pr->handle);
402 if (result)
403 goto err;
404
405 pr->dev = dev;
406 dev->offline = pr->flags.need_hotplug_init;
407
408 /* Trigger the processor driver's .probe() if present. */
409 if (device_attach(dev) >= 0)
410 return 1;
411
412 dev_err(dev, "Processor driver could not be attached\n");
413 acpi_unbind_one(dev);
414
415 err:
416 free_cpumask_var(pr->throttling.shared_cpu_map);
417 device->driver_data = NULL;
418 per_cpu(processors, pr->id) = NULL;
419 err_free_pr:
420 kfree(pr);
421 return result;
422}
423
424#ifdef CONFIG_ACPI_HOTPLUG_CPU
425/* --------------------------------------------------------------------------
426 Removal
427 -------------------------------------------------------------------------- */
428
429static void acpi_processor_remove(struct acpi_device *device)
430{
431 struct acpi_processor *pr;
432
433 if (!device || !acpi_driver_data(device))
434 return;
435
436 pr = acpi_driver_data(device);
437 if (pr->id >= nr_cpu_ids)
438 goto out;
439
440 /*
441 * The only reason why we ever get here is CPU hot-removal. The CPU is
442 * already offline and the ACPI device removal locking prevents it from
443 * being put back online at this point.
444 *
445 * Unbind the driver from the processor device and detach it from the
446 * ACPI companion object.
447 */
448 device_release_driver(pr->dev);
449 acpi_unbind_one(pr->dev);
450
451 /* Clean up. */
452 per_cpu(processor_device_array, pr->id) = NULL;
453 per_cpu(processors, pr->id) = NULL;
454 try_offline_node(cpu_to_node(pr->id));
455
456 /* Remove the CPU. */
457 get_online_cpus();
458 arch_unregister_cpu(pr->id);
459 acpi_unmap_lsapic(pr->id);
460 put_online_cpus();
461
462 out:
463 free_cpumask_var(pr->throttling.shared_cpu_map);
464 kfree(pr);
465}
466#endif /* CONFIG_ACPI_HOTPLUG_CPU */
467
468/*
469 * The following ACPI IDs are known to be suitable for representing as
470 * processor devices.
471 */
472static const struct acpi_device_id processor_device_ids[] = {
473
474 { ACPI_PROCESSOR_OBJECT_HID, },
475 { ACPI_PROCESSOR_DEVICE_HID, },
476
477 { }
478};
479
480static struct acpi_scan_handler __refdata processor_handler = {
481 .ids = processor_device_ids,
482 .attach = acpi_processor_add,
483#ifdef CONFIG_ACPI_HOTPLUG_CPU
484 .detach = acpi_processor_remove,
485#endif
486 .hotplug = {
487 .enabled = true,
488 },
489};
490
491void __init acpi_processor_init(void)
492{
493 acpi_scan_add_handler_with_hotplug(&processor_handler, "processor");
494}
diff --git a/drivers/acpi/glue.c b/drivers/acpi/glue.c
index 40a84cc6740c..9783f400d857 100644
--- a/drivers/acpi/glue.c
+++ b/drivers/acpi/glue.c
@@ -105,7 +105,7 @@ acpi_handle acpi_get_child(acpi_handle parent, u64 address)
105} 105}
106EXPORT_SYMBOL(acpi_get_child); 106EXPORT_SYMBOL(acpi_get_child);
107 107
108static int acpi_bind_one(struct device *dev, acpi_handle handle) 108int acpi_bind_one(struct device *dev, acpi_handle handle)
109{ 109{
110 struct acpi_device *acpi_dev; 110 struct acpi_device *acpi_dev;
111 acpi_status status; 111 acpi_status status;
@@ -188,8 +188,9 @@ static int acpi_bind_one(struct device *dev, acpi_handle handle)
188 kfree(physical_node); 188 kfree(physical_node);
189 goto err; 189 goto err;
190} 190}
191EXPORT_SYMBOL_GPL(acpi_bind_one);
191 192
192static int acpi_unbind_one(struct device *dev) 193int acpi_unbind_one(struct device *dev)
193{ 194{
194 struct acpi_device_physical_node *entry; 195 struct acpi_device_physical_node *entry;
195 struct acpi_device *acpi_dev; 196 struct acpi_device *acpi_dev;
@@ -238,6 +239,7 @@ err:
238 dev_err(dev, "Oops, 'acpi_handle' corrupt\n"); 239 dev_err(dev, "Oops, 'acpi_handle' corrupt\n");
239 return -EINVAL; 240 return -EINVAL;
240} 241}
242EXPORT_SYMBOL_GPL(acpi_unbind_one);
241 243
242static int acpi_platform_notify(struct device *dev) 244static int acpi_platform_notify(struct device *dev)
243{ 245{
diff --git a/drivers/acpi/internal.h b/drivers/acpi/internal.h
index c610a76d92c4..520073ba36b4 100644
--- a/drivers/acpi/internal.h
+++ b/drivers/acpi/internal.h
@@ -33,6 +33,7 @@ static inline void acpi_pci_slot_init(void) { }
33void acpi_pci_root_init(void); 33void acpi_pci_root_init(void);
34void acpi_pci_link_init(void); 34void acpi_pci_link_init(void);
35void acpi_pci_root_hp_init(void); 35void acpi_pci_root_hp_init(void);
36void acpi_processor_init(void);
36void acpi_platform_init(void); 37void acpi_platform_init(void);
37int acpi_sysfs_init(void); 38int acpi_sysfs_init(void);
38#ifdef CONFIG_ACPI_CONTAINER 39#ifdef CONFIG_ACPI_CONTAINER
@@ -51,6 +52,8 @@ void acpi_memory_hotplug_init(void);
51static inline void acpi_memory_hotplug_init(void) {} 52static inline void acpi_memory_hotplug_init(void) {}
52#endif 53#endif
53 54
55extern bool acpi_force_hot_remove;
56
54void acpi_sysfs_add_hotplug_profile(struct acpi_hotplug_profile *hotplug, 57void acpi_sysfs_add_hotplug_profile(struct acpi_hotplug_profile *hotplug,
55 const char *name); 58 const char *name);
56int acpi_scan_add_handler_with_hotplug(struct acpi_scan_handler *handler, 59int acpi_scan_add_handler_with_hotplug(struct acpi_scan_handler *handler,
@@ -81,6 +84,8 @@ void acpi_init_device_object(struct acpi_device *device, acpi_handle handle,
81 int type, unsigned long long sta); 84 int type, unsigned long long sta);
82void acpi_device_add_finalize(struct acpi_device *device); 85void acpi_device_add_finalize(struct acpi_device *device);
83void acpi_free_pnp_ids(struct acpi_device_pnp *pnp); 86void acpi_free_pnp_ids(struct acpi_device_pnp *pnp);
87int acpi_bind_one(struct device *dev, acpi_handle handle);
88int acpi_unbind_one(struct device *dev);
84 89
85/* -------------------------------------------------------------------------- 90/* --------------------------------------------------------------------------
86 Power Resource 91 Power Resource
diff --git a/drivers/acpi/processor_driver.c b/drivers/acpi/processor_driver.c
index c266cdc11784..d93963f1e8f4 100644
--- a/drivers/acpi/processor_driver.c
+++ b/drivers/acpi/processor_driver.c
@@ -1,11 +1,13 @@
1/* 1/*
2 * acpi_processor.c - ACPI Processor Driver ($Revision: 71 $) 2 * processor_driver.c - ACPI Processor Driver
3 * 3 *
4 * Copyright (C) 2001, 2002 Andy Grover <andrew.grover@intel.com> 4 * Copyright (C) 2001, 2002 Andy Grover <andrew.grover@intel.com>
5 * Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com> 5 * Copyright (C) 2001, 2002 Paul Diefenbaugh <paul.s.diefenbaugh@intel.com>
6 * Copyright (C) 2004 Dominik Brodowski <linux@brodo.de> 6 * Copyright (C) 2004 Dominik Brodowski <linux@brodo.de>
7 * Copyright (C) 2004 Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com> 7 * Copyright (C) 2004 Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
8 * - Added processor hotplug support 8 * - Added processor hotplug support
9 * Copyright (C) 2013, Intel Corporation
10 * Rafael J. Wysocki <rafael.j.wysocki@intel.com>
9 * 11 *
10 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 12 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
11 * 13 *
@@ -24,52 +26,29 @@
24 * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA. 26 * 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
25 * 27 *
26 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 28 * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
27 * TBD:
28 * 1. Make # power states dynamic.
29 * 2. Support duty_cycle values that span bit 4.
30 * 3. Optimize by having scheduler determine business instead of
31 * having us try to calculate it here.
32 * 4. Need C1 timing -- must modify kernel (IRQ handler) to get this.
33 */ 29 */
34 30
35#include <linux/kernel.h> 31#include <linux/kernel.h>
36#include <linux/module.h> 32#include <linux/module.h>
37#include <linux/init.h> 33#include <linux/init.h>
38#include <linux/types.h>
39#include <linux/pci.h>
40#include <linux/pm.h>
41#include <linux/cpufreq.h> 34#include <linux/cpufreq.h>
42#include <linux/cpu.h> 35#include <linux/cpu.h>
43#include <linux/dmi.h>
44#include <linux/moduleparam.h>
45#include <linux/cpuidle.h> 36#include <linux/cpuidle.h>
46#include <linux/slab.h> 37#include <linux/slab.h>
47#include <linux/acpi.h> 38#include <linux/acpi.h>
48#include <linux/memory_hotplug.h> 39
49
50#include <asm/io.h>
51#include <asm/cpu.h>
52#include <asm/delay.h>
53#include <asm/uaccess.h>
54#include <asm/processor.h>
55#include <asm/smp.h>
56#include <asm/acpi.h>
57
58#include <acpi/acpi_bus.h>
59#include <acpi/acpi_drivers.h>
60#include <acpi/processor.h> 40#include <acpi/processor.h>
61 41
42#include "internal.h"
43
62#define PREFIX "ACPI: " 44#define PREFIX "ACPI: "
63 45
64#define ACPI_PROCESSOR_CLASS "processor"
65#define ACPI_PROCESSOR_DEVICE_NAME "Processor"
66#define ACPI_PROCESSOR_FILE_INFO "info" 46#define ACPI_PROCESSOR_FILE_INFO "info"
67#define ACPI_PROCESSOR_FILE_THROTTLING "throttling" 47#define ACPI_PROCESSOR_FILE_THROTTLING "throttling"
68#define ACPI_PROCESSOR_FILE_LIMIT "limit" 48#define ACPI_PROCESSOR_FILE_LIMIT "limit"
69#define ACPI_PROCESSOR_NOTIFY_PERFORMANCE 0x80 49#define ACPI_PROCESSOR_NOTIFY_PERFORMANCE 0x80
70#define ACPI_PROCESSOR_NOTIFY_POWER 0x81 50#define ACPI_PROCESSOR_NOTIFY_POWER 0x81
71#define ACPI_PROCESSOR_NOTIFY_THROTTLING 0x82 51#define ACPI_PROCESSOR_NOTIFY_THROTTLING 0x82
72#define ACPI_PROCESSOR_DEVICE_HID "ACPI0007"
73 52
74#define ACPI_PROCESSOR_LIMIT_USER 0 53#define ACPI_PROCESSOR_LIMIT_USER 0
75#define ACPI_PROCESSOR_LIMIT_THERMAL 1 54#define ACPI_PROCESSOR_LIMIT_THERMAL 1
@@ -81,12 +60,8 @@ MODULE_AUTHOR("Paul Diefenbaugh");
81MODULE_DESCRIPTION("ACPI Processor Driver"); 60MODULE_DESCRIPTION("ACPI Processor Driver");
82MODULE_LICENSE("GPL"); 61MODULE_LICENSE("GPL");
83 62
84static int acpi_processor_add(struct acpi_device *device); 63static int acpi_processor_start(struct device *dev);
85static int acpi_processor_remove(struct acpi_device *device); 64static int acpi_processor_stop(struct device *dev);
86static void acpi_processor_notify(struct acpi_device *device, u32 event);
87static acpi_status acpi_processor_hotadd_init(struct acpi_processor *pr);
88static int acpi_processor_handle_eject(struct acpi_processor *pr);
89static int acpi_processor_start(struct acpi_processor *pr);
90 65
91static const struct acpi_device_id processor_device_ids[] = { 66static const struct acpi_device_id processor_device_ids[] = {
92 {ACPI_PROCESSOR_OBJECT_HID, 0}, 67 {ACPI_PROCESSOR_OBJECT_HID, 0},
@@ -95,295 +70,24 @@ static const struct acpi_device_id processor_device_ids[] = {
95}; 70};
96MODULE_DEVICE_TABLE(acpi, processor_device_ids); 71MODULE_DEVICE_TABLE(acpi, processor_device_ids);
97 72
98static struct acpi_driver acpi_processor_driver = { 73static struct device_driver acpi_processor_driver = {
99 .name = "processor", 74 .name = "processor",
100 .class = ACPI_PROCESSOR_CLASS, 75 .bus = &cpu_subsys,
101 .ids = processor_device_ids, 76 .acpi_match_table = processor_device_ids,
102 .ops = { 77 .probe = acpi_processor_start,
103 .add = acpi_processor_add, 78 .remove = acpi_processor_stop,
104 .remove = acpi_processor_remove,
105 .notify = acpi_processor_notify,
106 },
107}; 79};
108 80
109#define INSTALL_NOTIFY_HANDLER 1 81static void acpi_processor_notify(acpi_handle handle, u32 event, void *data)
110#define UNINSTALL_NOTIFY_HANDLER 2
111
112DEFINE_PER_CPU(struct acpi_processor *, processors);
113EXPORT_PER_CPU_SYMBOL(processors);
114
115struct acpi_processor_errata errata __read_mostly;
116
117/* --------------------------------------------------------------------------
118 Errata Handling
119 -------------------------------------------------------------------------- */
120
121static int acpi_processor_errata_piix4(struct pci_dev *dev)
122{ 82{
123 u8 value1 = 0; 83 struct acpi_device *device = data;
124 u8 value2 = 0;
125
126
127 if (!dev)
128 return -EINVAL;
129
130 /*
131 * Note that 'dev' references the PIIX4 ACPI Controller.
132 */
133
134 switch (dev->revision) {
135 case 0:
136 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4 A-step\n"));
137 break;
138 case 1:
139 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4 B-step\n"));
140 break;
141 case 2:
142 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4E\n"));
143 break;
144 case 3:
145 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found PIIX4M\n"));
146 break;
147 default:
148 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Found unknown PIIX4\n"));
149 break;
150 }
151
152 switch (dev->revision) {
153
154 case 0: /* PIIX4 A-step */
155 case 1: /* PIIX4 B-step */
156 /*
157 * See specification changes #13 ("Manual Throttle Duty Cycle")
158 * and #14 ("Enabling and Disabling Manual Throttle"), plus
159 * erratum #5 ("STPCLK# Deassertion Time") from the January
160 * 2002 PIIX4 specification update. Applies to only older
161 * PIIX4 models.
162 */
163 errata.piix4.throttle = 1;
164
165 case 2: /* PIIX4E */
166 case 3: /* PIIX4M */
167 /*
168 * See erratum #18 ("C3 Power State/BMIDE and Type-F DMA
169 * Livelock") from the January 2002 PIIX4 specification update.
170 * Applies to all PIIX4 models.
171 */
172
173 /*
174 * BM-IDE
175 * ------
176 * Find the PIIX4 IDE Controller and get the Bus Master IDE
177 * Status register address. We'll use this later to read
178 * each IDE controller's DMA status to make sure we catch all
179 * DMA activity.
180 */
181 dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
182 PCI_DEVICE_ID_INTEL_82371AB,
183 PCI_ANY_ID, PCI_ANY_ID, NULL);
184 if (dev) {
185 errata.piix4.bmisx = pci_resource_start(dev, 4);
186 pci_dev_put(dev);
187 }
188
189 /*
190 * Type-F DMA
191 * ----------
192 * Find the PIIX4 ISA Controller and read the Motherboard
193 * DMA controller's status to see if Type-F (Fast) DMA mode
194 * is enabled (bit 7) on either channel. Note that we'll
195 * disable C3 support if this is enabled, as some legacy
196 * devices won't operate well if fast DMA is disabled.
197 */
198 dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
199 PCI_DEVICE_ID_INTEL_82371AB_0,
200 PCI_ANY_ID, PCI_ANY_ID, NULL);
201 if (dev) {
202 pci_read_config_byte(dev, 0x76, &value1);
203 pci_read_config_byte(dev, 0x77, &value2);
204 if ((value1 & 0x80) || (value2 & 0x80))
205 errata.piix4.fdma = 1;
206 pci_dev_put(dev);
207 }
208
209 break;
210 }
211
212 if (errata.piix4.bmisx)
213 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
214 "Bus master activity detection (BM-IDE) erratum enabled\n"));
215 if (errata.piix4.fdma)
216 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
217 "Type-F DMA livelock erratum (C3 disabled)\n"));
218
219 return 0;
220}
221
222static int acpi_processor_errata(struct acpi_processor *pr)
223{
224 int result = 0;
225 struct pci_dev *dev = NULL;
226
227
228 if (!pr)
229 return -EINVAL;
230
231 /*
232 * PIIX4
233 */
234 dev = pci_get_subsys(PCI_VENDOR_ID_INTEL,
235 PCI_DEVICE_ID_INTEL_82371AB_3, PCI_ANY_ID,
236 PCI_ANY_ID, NULL);
237 if (dev) {
238 result = acpi_processor_errata_piix4(dev);
239 pci_dev_put(dev);
240 }
241
242 return result;
243}
244
245/* --------------------------------------------------------------------------
246 Driver Interface
247 -------------------------------------------------------------------------- */
248
249static int acpi_processor_get_info(struct acpi_device *device)
250{
251 acpi_status status = 0;
252 union acpi_object object = { 0 };
253 struct acpi_buffer buffer = { sizeof(union acpi_object), &object };
254 struct acpi_processor *pr; 84 struct acpi_processor *pr;
255 int cpu_index, device_declaration = 0;
256 static int cpu0_initialized;
257
258 pr = acpi_driver_data(device);
259 if (!pr)
260 return -EINVAL;
261
262 if (num_online_cpus() > 1)
263 errata.smp = TRUE;
264
265 acpi_processor_errata(pr);
266
267 /*
268 * Check to see if we have bus mastering arbitration control. This
269 * is required for proper C3 usage (to maintain cache coherency).
270 */
271 if (acpi_gbl_FADT.pm2_control_block && acpi_gbl_FADT.pm2_control_length) {
272 pr->flags.bm_control = 1;
273 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
274 "Bus mastering arbitration control present\n"));
275 } else
276 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
277 "No bus mastering arbitration control\n"));
278
279 if (!strcmp(acpi_device_hid(device), ACPI_PROCESSOR_OBJECT_HID)) {
280 /* Declared with "Processor" statement; match ProcessorID */
281 status = acpi_evaluate_object(pr->handle, NULL, NULL, &buffer);
282 if (ACPI_FAILURE(status)) {
283 dev_err(&device->dev,
284 "Failed to evaluate processor object (0x%x)\n",
285 status);
286 return -ENODEV;
287 }
288
289 /*
290 * TBD: Synch processor ID (via LAPIC/LSAPIC structures) on SMP.
291 * >>> 'acpi_get_processor_id(acpi_id, &id)' in
292 * arch/xxx/acpi.c
293 */
294 pr->acpi_id = object.processor.proc_id;
295 } else {
296 /*
297 * Declared with "Device" statement; match _UID.
298 * Note that we don't handle string _UIDs yet.
299 */
300 unsigned long long value;
301 status = acpi_evaluate_integer(pr->handle, METHOD_NAME__UID,
302 NULL, &value);
303 if (ACPI_FAILURE(status)) {
304 dev_err(&device->dev,
305 "Failed to evaluate processor _UID (0x%x)\n",
306 status);
307 return -ENODEV;
308 }
309 device_declaration = 1;
310 pr->acpi_id = value;
311 }
312 cpu_index = acpi_get_cpuid(pr->handle, device_declaration, pr->acpi_id);
313
314 /* Handle UP system running SMP kernel, with no LAPIC in MADT */
315 if (!cpu0_initialized && (cpu_index == -1) &&
316 (num_online_cpus() == 1)) {
317 cpu_index = 0;
318 }
319
320 cpu0_initialized = 1;
321
322 pr->id = cpu_index;
323
324 /*
325 * Extra Processor objects may be enumerated on MP systems with
326 * less than the max # of CPUs. They should be ignored _iff
327 * they are physically not present.
328 */
329 if (pr->id == -1) {
330 if (ACPI_FAILURE(acpi_processor_hotadd_init(pr)))
331 return -ENODEV;
332 }
333 /*
334 * On some boxes several processors use the same processor bus id.
335 * But they are located in different scope. For example:
336 * \_SB.SCK0.CPU0
337 * \_SB.SCK1.CPU0
338 * Rename the processor device bus id. And the new bus id will be
339 * generated as the following format:
340 * CPU+CPU ID.
341 */
342 sprintf(acpi_device_bid(device), "CPU%X", pr->id);
343 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "Processor [%d:%d]\n", pr->id,
344 pr->acpi_id));
345
346 if (!object.processor.pblk_address)
347 ACPI_DEBUG_PRINT((ACPI_DB_INFO, "No PBLK (NULL address)\n"));
348 else if (object.processor.pblk_length != 6)
349 dev_err(&device->dev, "Invalid PBLK length [%d]\n",
350 object.processor.pblk_length);
351 else {
352 pr->throttling.address = object.processor.pblk_address;
353 pr->throttling.duty_offset = acpi_gbl_FADT.duty_offset;
354 pr->throttling.duty_width = acpi_gbl_FADT.duty_width;
355
356 pr->pblk = object.processor.pblk_address;
357
358 /*
359 * We don't care about error returns - we just try to mark
360 * these reserved so that nobody else is confused into thinking
361 * that this region might be unused..
362 *
363 * (In particular, allocating the IO range for Cardbus)
364 */
365 request_region(pr->throttling.address, 6, "ACPI CPU throttle");
366 }
367
368 /*
369 * If ACPI describes a slot number for this CPU, we can use it
370 * ensure we get the right value in the "physical id" field
371 * of /proc/cpuinfo
372 */
373 status = acpi_evaluate_object(pr->handle, "_SUN", NULL, &buffer);
374 if (ACPI_SUCCESS(status))
375 arch_fix_phys_package_id(pr->id, object.integer.value);
376
377 return 0;
378}
379
380static DEFINE_PER_CPU(void *, processor_device_array);
381
382static void acpi_processor_notify(struct acpi_device *device, u32 event)
383{
384 struct acpi_processor *pr = acpi_driver_data(device);
385 int saved; 85 int saved;
386 86
87 if (device->handle != handle)
88 return;
89
90 pr = acpi_driver_data(device);
387 if (!pr) 91 if (!pr)
388 return; 92 return;
389 93
@@ -420,55 +124,62 @@ static void acpi_processor_notify(struct acpi_device *device, u32 event)
420 return; 124 return;
421} 125}
422 126
423static int acpi_cpu_soft_notify(struct notifier_block *nfb, 127static __cpuinit int __acpi_processor_start(struct acpi_device *device);
424 unsigned long action, void *hcpu) 128
129static int __cpuinit acpi_cpu_soft_notify(struct notifier_block *nfb,
130 unsigned long action, void *hcpu)
425{ 131{
426 unsigned int cpu = (unsigned long)hcpu; 132 unsigned int cpu = (unsigned long)hcpu;
427 struct acpi_processor *pr = per_cpu(processors, cpu); 133 struct acpi_processor *pr = per_cpu(processors, cpu);
134 struct acpi_device *device;
135
136 if (!pr || acpi_bus_get_device(pr->handle, &device))
137 return NOTIFY_DONE;
428 138
429 if (action == CPU_ONLINE && pr) { 139 if (action == CPU_ONLINE) {
430 /* CPU got physically hotplugged and onlined the first time: 140 /*
431 * Initialize missing things 141 * CPU got physically hotplugged and onlined for the first time:
142 * Initialize missing things.
432 */ 143 */
433 if (pr->flags.need_hotplug_init) { 144 if (pr->flags.need_hotplug_init) {
145 int ret;
146
434 pr_info("Will online and init hotplugged CPU: %d\n", 147 pr_info("Will online and init hotplugged CPU: %d\n",
435 pr->id); 148 pr->id);
436 WARN(acpi_processor_start(pr), "Failed to start CPU:"
437 " %d\n", pr->id);
438 pr->flags.need_hotplug_init = 0; 149 pr->flags.need_hotplug_init = 0;
439 /* Normal CPU soft online event */ 150 ret = __acpi_processor_start(device);
151 WARN(ret, "Failed to start CPU: %d\n", pr->id);
440 } else { 152 } else {
153 /* Normal CPU soft online event. */
441 acpi_processor_ppc_has_changed(pr, 0); 154 acpi_processor_ppc_has_changed(pr, 0);
442 acpi_processor_hotplug(pr); 155 acpi_processor_hotplug(pr);
443 acpi_processor_reevaluate_tstate(pr, action); 156 acpi_processor_reevaluate_tstate(pr, action);
444 acpi_processor_tstate_has_changed(pr); 157 acpi_processor_tstate_has_changed(pr);
445 } 158 }
446 } 159 } else if (action == CPU_DEAD) {
447 if (action == CPU_DEAD && pr) { 160 /* Invalidate flag.throttling after the CPU is offline. */
448 /* invalidate the flag.throttling after one CPU is offline */
449 acpi_processor_reevaluate_tstate(pr, action); 161 acpi_processor_reevaluate_tstate(pr, action);
450 } 162 }
451 return NOTIFY_OK; 163 return NOTIFY_OK;
452} 164}
453 165
454static struct notifier_block acpi_cpu_notifier = 166static struct notifier_block __refdata acpi_cpu_notifier =
455{ 167{
456 .notifier_call = acpi_cpu_soft_notify, 168 .notifier_call = acpi_cpu_soft_notify,
457}; 169};
458 170
459/* 171static __cpuinit int __acpi_processor_start(struct acpi_device *device)
460 * acpi_processor_start() is called by the cpu_hotplug_notifier func:
461 * acpi_cpu_soft_notify(). Getting it __cpuinit{data} is difficult, the
462 * root cause seem to be that acpi_processor_uninstall_hotplug_notify()
463 * is in the module_exit (__exit) func. Allowing acpi_processor_start()
464 * to not be in __cpuinit section, but being called from __cpuinit funcs
465 * via __ref looks like the right thing to do here.
466 */
467static __ref int acpi_processor_start(struct acpi_processor *pr)
468{ 172{
469 struct acpi_device *device = per_cpu(processor_device_array, pr->id); 173 struct acpi_processor *pr = acpi_driver_data(device);
174 acpi_status status;
470 int result = 0; 175 int result = 0;
471 176
177 if (!pr)
178 return -ENODEV;
179
180 if (pr->flags.need_hotplug_init)
181 return 0;
182
472#ifdef CONFIG_CPU_FREQ 183#ifdef CONFIG_CPU_FREQ
473 acpi_processor_ppc_has_changed(pr, 0); 184 acpi_processor_ppc_has_changed(pr, 0);
474 acpi_processor_load_module(pr); 185 acpi_processor_load_module(pr);
@@ -506,462 +217,95 @@ static __ref int acpi_processor_start(struct acpi_processor *pr)
506 goto err_remove_sysfs_thermal; 217 goto err_remove_sysfs_thermal;
507 } 218 }
508 219
509 return 0; 220 status = acpi_install_notify_handler(device->handle, ACPI_DEVICE_NOTIFY,
221 acpi_processor_notify, device);
222 if (ACPI_SUCCESS(status))
223 return 0;
510 224
511err_remove_sysfs_thermal: 225 sysfs_remove_link(&pr->cdev->device.kobj, "device");
226 err_remove_sysfs_thermal:
512 sysfs_remove_link(&device->dev.kobj, "thermal_cooling"); 227 sysfs_remove_link(&device->dev.kobj, "thermal_cooling");
513err_thermal_unregister: 228 err_thermal_unregister:
514 thermal_cooling_device_unregister(pr->cdev); 229 thermal_cooling_device_unregister(pr->cdev);
515err_power_exit: 230 err_power_exit:
516 acpi_processor_power_exit(pr); 231 acpi_processor_power_exit(pr);
517
518 return result; 232 return result;
519} 233}
520 234
521/* 235static int __cpuinit acpi_processor_start(struct device *dev)
522 * Do not put anything in here which needs the core to be online.
523 * For example MSR access or setting up things which check for cpuinfo_x86
524 * (cpu_data(cpu)) values, like CPU feature flags, family, model, etc.
525 * Such things have to be put in and set up above in acpi_processor_start()
526 */
527static int __cpuinit acpi_processor_add(struct acpi_device *device)
528{ 236{
529 struct acpi_processor *pr = NULL; 237 struct acpi_device *device;
530 int result = 0;
531 struct device *dev;
532
533 pr = kzalloc(sizeof(struct acpi_processor), GFP_KERNEL);
534 if (!pr)
535 return -ENOMEM;
536
537 if (!zalloc_cpumask_var(&pr->throttling.shared_cpu_map, GFP_KERNEL)) {
538 result = -ENOMEM;
539 goto err_free_pr;
540 }
541
542 pr->handle = device->handle;
543 strcpy(acpi_device_name(device), ACPI_PROCESSOR_DEVICE_NAME);
544 strcpy(acpi_device_class(device), ACPI_PROCESSOR_CLASS);
545 device->driver_data = pr;
546
547 result = acpi_processor_get_info(device);
548 if (result) {
549 /* Processor is physically not present */
550 return 0;
551 }
552
553#ifdef CONFIG_SMP
554 if (pr->id >= setup_max_cpus && pr->id != 0)
555 return 0;
556#endif
557
558 BUG_ON(pr->id >= nr_cpu_ids);
559
560 /*
561 * Buggy BIOS check
562 * ACPI id of processors can be reported wrongly by the BIOS.
563 * Don't trust it blindly
564 */
565 if (per_cpu(processor_device_array, pr->id) != NULL &&
566 per_cpu(processor_device_array, pr->id) != device) {
567 dev_warn(&device->dev,
568 "BIOS reported wrong ACPI id %d for the processor\n",
569 pr->id);
570 result = -ENODEV;
571 goto err_free_cpumask;
572 }
573 per_cpu(processor_device_array, pr->id) = device;
574
575 per_cpu(processors, pr->id) = pr;
576
577 dev = get_cpu_device(pr->id);
578 if (sysfs_create_link(&device->dev.kobj, &dev->kobj, "sysdev")) {
579 result = -EFAULT;
580 goto err_clear_processor;
581 }
582 238
583 /* 239 if (acpi_bus_get_device(ACPI_HANDLE(dev), &device))
584 * Do not start hotplugged CPUs now, but when they 240 return -ENODEV;
585 * are onlined the first time
586 */
587 if (pr->flags.need_hotplug_init)
588 return 0;
589
590 result = acpi_processor_start(pr);
591 if (result)
592 goto err_remove_sysfs;
593
594 return 0;
595 241
596err_remove_sysfs: 242 return __acpi_processor_start(device);
597 sysfs_remove_link(&device->dev.kobj, "sysdev");
598err_clear_processor:
599 /*
600 * processor_device_array is not cleared to allow checks for buggy BIOS
601 */
602 per_cpu(processors, pr->id) = NULL;
603err_free_cpumask:
604 free_cpumask_var(pr->throttling.shared_cpu_map);
605err_free_pr:
606 kfree(pr);
607 return result;
608} 243}
609 244
610static int acpi_processor_remove(struct acpi_device *device) 245static int acpi_processor_stop(struct device *dev)
611{ 246{
612 struct acpi_processor *pr = NULL; 247 struct acpi_device *device;
248 struct acpi_processor *pr;
613 249
250 if (acpi_bus_get_device(ACPI_HANDLE(dev), &device))
251 return 0;
614 252
615 if (!device || !acpi_driver_data(device)) 253 acpi_remove_notify_handler(device->handle, ACPI_DEVICE_NOTIFY,
616 return -EINVAL; 254 acpi_processor_notify);
617 255
618 pr = acpi_driver_data(device); 256 pr = acpi_driver_data(device);
619 257 if (!pr)
620 if (pr->id >= nr_cpu_ids) 258 return 0;
621 goto free;
622
623 if (device->removal_type == ACPI_BUS_REMOVAL_EJECT) {
624 if (acpi_processor_handle_eject(pr))
625 return -EINVAL;
626 }
627 259
628 acpi_processor_power_exit(pr); 260 acpi_processor_power_exit(pr);
629 261
630 sysfs_remove_link(&device->dev.kobj, "sysdev");
631
632 if (pr->cdev) { 262 if (pr->cdev) {
633 sysfs_remove_link(&device->dev.kobj, "thermal_cooling"); 263 sysfs_remove_link(&device->dev.kobj, "thermal_cooling");
634 sysfs_remove_link(&pr->cdev->device.kobj, "device"); 264 sysfs_remove_link(&pr->cdev->device.kobj, "device");
635 thermal_cooling_device_unregister(pr->cdev); 265 thermal_cooling_device_unregister(pr->cdev);
636 pr->cdev = NULL; 266 pr->cdev = NULL;
637 } 267 }
638
639 per_cpu(processors, pr->id) = NULL;
640 per_cpu(processor_device_array, pr->id) = NULL;
641 try_offline_node(cpu_to_node(pr->id));
642
643free:
644 free_cpumask_var(pr->throttling.shared_cpu_map);
645 kfree(pr);
646
647 return 0; 268 return 0;
648} 269}
649 270
650#ifdef CONFIG_ACPI_HOTPLUG_CPU
651/****************************************************************************
652 * Acpi processor hotplug support *
653 ****************************************************************************/
654
655static int is_processor_present(acpi_handle handle)
656{
657 acpi_status status;
658 unsigned long long sta = 0;
659
660
661 status = acpi_evaluate_integer(handle, "_STA", NULL, &sta);
662
663 if (ACPI_SUCCESS(status) && (sta & ACPI_STA_DEVICE_PRESENT))
664 return 1;
665
666 /*
667 * _STA is mandatory for a processor that supports hot plug
668 */
669 if (status == AE_NOT_FOUND)
670 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
671 "Processor does not support hot plug\n"));
672 else
673 ACPI_EXCEPTION((AE_INFO, status,
674 "Processor Device is not present"));
675 return 0;
676}
677
678static void acpi_processor_hotplug_notify(acpi_handle handle,
679 u32 event, void *data)
680{
681 struct acpi_device *device = NULL;
682 struct acpi_eject_event *ej_event = NULL;
683 u32 ost_code = ACPI_OST_SC_NON_SPECIFIC_FAILURE; /* default */
684 acpi_status status;
685 int result;
686
687 acpi_scan_lock_acquire();
688
689 switch (event) {
690 case ACPI_NOTIFY_BUS_CHECK:
691 case ACPI_NOTIFY_DEVICE_CHECK:
692 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
693 "Processor driver received %s event\n",
694 (event == ACPI_NOTIFY_BUS_CHECK) ?
695 "ACPI_NOTIFY_BUS_CHECK" : "ACPI_NOTIFY_DEVICE_CHECK"));
696
697 if (!is_processor_present(handle))
698 break;
699
700 if (!acpi_bus_get_device(handle, &device))
701 break;
702
703 result = acpi_bus_scan(handle);
704 if (result) {
705 acpi_handle_err(handle, "Unable to add the device\n");
706 break;
707 }
708 result = acpi_bus_get_device(handle, &device);
709 if (result) {
710 acpi_handle_err(handle, "Missing device object\n");
711 break;
712 }
713 ost_code = ACPI_OST_SC_SUCCESS;
714 break;
715
716 case ACPI_NOTIFY_EJECT_REQUEST:
717 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
718 "received ACPI_NOTIFY_EJECT_REQUEST\n"));
719
720 if (acpi_bus_get_device(handle, &device)) {
721 acpi_handle_err(handle,
722 "Device don't exist, dropping EJECT\n");
723 break;
724 }
725 if (!acpi_driver_data(device)) {
726 acpi_handle_err(handle,
727 "Driver data is NULL, dropping EJECT\n");
728 break;
729 }
730
731 ej_event = kmalloc(sizeof(*ej_event), GFP_KERNEL);
732 if (!ej_event) {
733 acpi_handle_err(handle, "No memory, dropping EJECT\n");
734 break;
735 }
736
737 get_device(&device->dev);
738 ej_event->device = device;
739 ej_event->event = ACPI_NOTIFY_EJECT_REQUEST;
740 /* The eject is carried out asynchronously. */
741 status = acpi_os_hotplug_execute(acpi_bus_hot_remove_device,
742 ej_event);
743 if (ACPI_FAILURE(status)) {
744 put_device(&device->dev);
745 kfree(ej_event);
746 break;
747 }
748 goto out;
749
750 default:
751 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
752 "Unsupported event [0x%x]\n", event));
753
754 /* non-hotplug event; possibly handled by other handler */
755 goto out;
756 }
757
758 /* Inform firmware that the hotplug operation has completed */
759 (void) acpi_evaluate_hotplug_ost(handle, event, ost_code, NULL);
760
761 out:
762 acpi_scan_lock_release();
763}
764
765static acpi_status is_processor_device(acpi_handle handle)
766{
767 struct acpi_device_info *info;
768 char *hid;
769 acpi_status status;
770
771 status = acpi_get_object_info(handle, &info);
772 if (ACPI_FAILURE(status))
773 return status;
774
775 if (info->type == ACPI_TYPE_PROCESSOR) {
776 kfree(info);
777 return AE_OK; /* found a processor object */
778 }
779
780 if (!(info->valid & ACPI_VALID_HID)) {
781 kfree(info);
782 return AE_ERROR;
783 }
784
785 hid = info->hardware_id.string;
786 if ((hid == NULL) || strcmp(hid, ACPI_PROCESSOR_DEVICE_HID)) {
787 kfree(info);
788 return AE_ERROR;
789 }
790
791 kfree(info);
792 return AE_OK; /* found a processor device object */
793}
794
795static acpi_status
796processor_walk_namespace_cb(acpi_handle handle,
797 u32 lvl, void *context, void **rv)
798{
799 acpi_status status;
800 int *action = context;
801
802 status = is_processor_device(handle);
803 if (ACPI_FAILURE(status))
804 return AE_OK; /* not a processor; continue to walk */
805
806 switch (*action) {
807 case INSTALL_NOTIFY_HANDLER:
808 acpi_install_notify_handler(handle,
809 ACPI_SYSTEM_NOTIFY,
810 acpi_processor_hotplug_notify,
811 NULL);
812 break;
813 case UNINSTALL_NOTIFY_HANDLER:
814 acpi_remove_notify_handler(handle,
815 ACPI_SYSTEM_NOTIFY,
816 acpi_processor_hotplug_notify);
817 break;
818 default:
819 break;
820 }
821
822 /* found a processor; skip walking underneath */
823 return AE_CTRL_DEPTH;
824}
825
826static acpi_status acpi_processor_hotadd_init(struct acpi_processor *pr)
827{
828 acpi_handle handle = pr->handle;
829
830 if (!is_processor_present(handle)) {
831 return AE_ERROR;
832 }
833
834 if (acpi_map_lsapic(handle, &pr->id))
835 return AE_ERROR;
836
837 if (arch_register_cpu(pr->id)) {
838 acpi_unmap_lsapic(pr->id);
839 return AE_ERROR;
840 }
841
842 /* CPU got hot-plugged, but cpu_data is not initialized yet
843 * Set flag to delay cpu_idle/throttling initialization
844 * in:
845 * acpi_processor_add()
846 * acpi_processor_get_info()
847 * and do it when the CPU gets online the first time
848 * TBD: Cleanup above functions and try to do this more elegant.
849 */
850 pr_info("CPU %d got hotplugged\n", pr->id);
851 pr->flags.need_hotplug_init = 1;
852
853 return AE_OK;
854}
855
856static int acpi_processor_handle_eject(struct acpi_processor *pr)
857{
858 if (cpu_online(pr->id))
859 cpu_down(pr->id);
860
861 get_online_cpus();
862 /*
863 * The cpu might become online again at this point. So we check whether
864 * the cpu has been onlined or not. If the cpu became online, it means
865 * that someone wants to use the cpu. So acpi_processor_handle_eject()
866 * returns -EAGAIN.
867 */
868 if (unlikely(cpu_online(pr->id))) {
869 put_online_cpus();
870 pr_warn("Failed to remove CPU %d, because other task "
871 "brought the CPU back online\n", pr->id);
872 return -EAGAIN;
873 }
874 arch_unregister_cpu(pr->id);
875 acpi_unmap_lsapic(pr->id);
876 put_online_cpus();
877 return (0);
878}
879#else
880static acpi_status acpi_processor_hotadd_init(struct acpi_processor *pr)
881{
882 return AE_ERROR;
883}
884static int acpi_processor_handle_eject(struct acpi_processor *pr)
885{
886 return (-EINVAL);
887}
888#endif
889
890static
891void acpi_processor_install_hotplug_notify(void)
892{
893#ifdef CONFIG_ACPI_HOTPLUG_CPU
894 int action = INSTALL_NOTIFY_HANDLER;
895 acpi_walk_namespace(ACPI_TYPE_ANY,
896 ACPI_ROOT_OBJECT,
897 ACPI_UINT32_MAX,
898 processor_walk_namespace_cb, NULL, &action, NULL);
899#endif
900 register_hotcpu_notifier(&acpi_cpu_notifier);
901}
902
903static
904void acpi_processor_uninstall_hotplug_notify(void)
905{
906#ifdef CONFIG_ACPI_HOTPLUG_CPU
907 int action = UNINSTALL_NOTIFY_HANDLER;
908 acpi_walk_namespace(ACPI_TYPE_ANY,
909 ACPI_ROOT_OBJECT,
910 ACPI_UINT32_MAX,
911 processor_walk_namespace_cb, NULL, &action, NULL);
912#endif
913 unregister_hotcpu_notifier(&acpi_cpu_notifier);
914}
915
916/* 271/*
917 * We keep the driver loaded even when ACPI is not running. 272 * We keep the driver loaded even when ACPI is not running.
918 * This is needed for the powernow-k8 driver, that works even without 273 * This is needed for the powernow-k8 driver, that works even without
919 * ACPI, but needs symbols from this driver 274 * ACPI, but needs symbols from this driver
920 */ 275 */
921 276
922static int __init acpi_processor_init(void) 277static int __init acpi_processor_driver_init(void)
923{ 278{
924 int result = 0; 279 int result = 0;
925 280
926 if (acpi_disabled) 281 if (acpi_disabled)
927 return 0; 282 return 0;
928 283
929 result = acpi_bus_register_driver(&acpi_processor_driver); 284 result = driver_register(&acpi_processor_driver);
930 if (result < 0) 285 if (result < 0)
931 return result; 286 return result;
932 287
933 acpi_processor_syscore_init(); 288 acpi_processor_syscore_init();
934 289 register_hotcpu_notifier(&acpi_cpu_notifier);
935 acpi_processor_install_hotplug_notify();
936
937 acpi_thermal_cpufreq_init(); 290 acpi_thermal_cpufreq_init();
938
939 acpi_processor_ppc_init(); 291 acpi_processor_ppc_init();
940
941 acpi_processor_throttling_init(); 292 acpi_processor_throttling_init();
942
943 return 0; 293 return 0;
944} 294}
945 295
946static void __exit acpi_processor_exit(void) 296static void __exit acpi_processor_driver_exit(void)
947{ 297{
948 if (acpi_disabled) 298 if (acpi_disabled)
949 return; 299 return;
950 300
951 acpi_processor_ppc_exit(); 301 acpi_processor_ppc_exit();
952
953 acpi_thermal_cpufreq_exit(); 302 acpi_thermal_cpufreq_exit();
954 303 unregister_hotcpu_notifier(&acpi_cpu_notifier);
955 acpi_processor_uninstall_hotplug_notify();
956
957 acpi_processor_syscore_exit(); 304 acpi_processor_syscore_exit();
958 305 driver_unregister(&acpi_processor_driver);
959 acpi_bus_unregister_driver(&acpi_processor_driver);
960
961 return;
962} 306}
963 307
964module_init(acpi_processor_init); 308module_init(acpi_processor_driver_init);
965module_exit(acpi_processor_exit); 309module_exit(acpi_processor_driver_exit);
966 310
967MODULE_ALIAS("processor"); 311MODULE_ALIAS("processor");
diff --git a/drivers/acpi/scan.c b/drivers/acpi/scan.c
index 27da63061e11..db118b1ad3e8 100644
--- a/drivers/acpi/scan.c
+++ b/drivers/acpi/scan.c
@@ -27,6 +27,12 @@ extern struct acpi_device *acpi_root;
27 27
28#define ACPI_IS_ROOT_DEVICE(device) (!(device)->parent) 28#define ACPI_IS_ROOT_DEVICE(device) (!(device)->parent)
29 29
30/*
31 * If set, devices will be hot-removed even if they cannot be put offline
32 * gracefully (from the kernel's standpoint).
33 */
34bool acpi_force_hot_remove;
35
30static const char *dummy_hid = "device"; 36static const char *dummy_hid = "device";
31 37
32static LIST_HEAD(acpi_device_list); 38static LIST_HEAD(acpi_device_list);
@@ -120,12 +126,78 @@ acpi_device_modalias_show(struct device *dev, struct device_attribute *attr, cha
120} 126}
121static DEVICE_ATTR(modalias, 0444, acpi_device_modalias_show, NULL); 127static DEVICE_ATTR(modalias, 0444, acpi_device_modalias_show, NULL);
122 128
129static acpi_status acpi_bus_offline_companions(acpi_handle handle, u32 lvl,
130 void *data, void **ret_p)
131{
132 struct acpi_device *device = NULL;
133 struct acpi_device_physical_node *pn;
134 bool second_pass = (bool)data;
135 acpi_status status = AE_OK;
136
137 if (acpi_bus_get_device(handle, &device))
138 return AE_OK;
139
140 mutex_lock(&device->physical_node_lock);
141
142 list_for_each_entry(pn, &device->physical_node_list, node) {
143 int ret;
144
145 if (second_pass) {
146 /* Skip devices offlined by the first pass. */
147 if (pn->put_online)
148 continue;
149 } else {
150 pn->put_online = false;
151 }
152 ret = device_offline(pn->dev);
153 if (acpi_force_hot_remove)
154 continue;
155
156 if (ret >= 0) {
157 pn->put_online = !ret;
158 } else {
159 *ret_p = pn->dev;
160 if (second_pass) {
161 status = AE_ERROR;
162 break;
163 }
164 }
165 }
166
167 mutex_unlock(&device->physical_node_lock);
168
169 return status;
170}
171
172static acpi_status acpi_bus_online_companions(acpi_handle handle, u32 lvl,
173 void *data, void **ret_p)
174{
175 struct acpi_device *device = NULL;
176 struct acpi_device_physical_node *pn;
177
178 if (acpi_bus_get_device(handle, &device))
179 return AE_OK;
180
181 mutex_lock(&device->physical_node_lock);
182
183 list_for_each_entry(pn, &device->physical_node_list, node)
184 if (pn->put_online) {
185 device_online(pn->dev);
186 pn->put_online = false;
187 }
188
189 mutex_unlock(&device->physical_node_lock);
190
191 return AE_OK;
192}
193
123static int acpi_scan_hot_remove(struct acpi_device *device) 194static int acpi_scan_hot_remove(struct acpi_device *device)
124{ 195{
125 acpi_handle handle = device->handle; 196 acpi_handle handle = device->handle;
126 acpi_handle not_used; 197 acpi_handle not_used;
127 struct acpi_object_list arg_list; 198 struct acpi_object_list arg_list;
128 union acpi_object arg; 199 union acpi_object arg;
200 struct device *errdev;
129 acpi_status status; 201 acpi_status status;
130 unsigned long long sta; 202 unsigned long long sta;
131 203
@@ -136,10 +208,53 @@ static int acpi_scan_hot_remove(struct acpi_device *device)
136 return -EINVAL; 208 return -EINVAL;
137 } 209 }
138 210
211 lock_device_hotplug();
212
213 /*
214 * Carry out two passes here and ignore errors in the first pass,
215 * because if the devices in question are memory blocks and
216 * CONFIG_MEMCG is set, one of the blocks may hold data structures
217 * that the other blocks depend on, but it is not known in advance which
218 * block holds them.
219 *
220 * If the first pass is successful, the second one isn't needed, though.
221 */
222 errdev = NULL;
223 acpi_walk_namespace(ACPI_TYPE_ANY, handle, ACPI_UINT32_MAX,
224 NULL, acpi_bus_offline_companions,
225 (void *)false, (void **)&errdev);
226 acpi_bus_offline_companions(handle, 0, (void *)false, (void **)&errdev);
227 if (errdev) {
228 errdev = NULL;
229 acpi_walk_namespace(ACPI_TYPE_ANY, handle, ACPI_UINT32_MAX,
230 NULL, acpi_bus_offline_companions,
231 (void *)true , (void **)&errdev);
232 if (!errdev || acpi_force_hot_remove)
233 acpi_bus_offline_companions(handle, 0, (void *)true,
234 (void **)&errdev);
235
236 if (errdev && !acpi_force_hot_remove) {
237 dev_warn(errdev, "Offline failed.\n");
238 acpi_bus_online_companions(handle, 0, NULL, NULL);
239 acpi_walk_namespace(ACPI_TYPE_ANY, handle,
240 ACPI_UINT32_MAX,
241 acpi_bus_online_companions, NULL,
242 NULL, NULL);
243
244 unlock_device_hotplug();
245
246 put_device(&device->dev);
247 return -EBUSY;
248 }
249 }
250
139 ACPI_DEBUG_PRINT((ACPI_DB_INFO, 251 ACPI_DEBUG_PRINT((ACPI_DB_INFO,
140 "Hot-removing device %s...\n", dev_name(&device->dev))); 252 "Hot-removing device %s...\n", dev_name(&device->dev)));
141 253
142 acpi_bus_trim(device); 254 acpi_bus_trim(device);
255
256 unlock_device_hotplug();
257
143 /* Device node has been unregistered. */ 258 /* Device node has been unregistered. */
144 put_device(&device->dev); 259 put_device(&device->dev);
145 device = NULL; 260 device = NULL;
@@ -236,6 +351,7 @@ static void acpi_scan_bus_device_check(acpi_handle handle, u32 ost_source)
236 int error; 351 int error;
237 352
238 mutex_lock(&acpi_scan_lock); 353 mutex_lock(&acpi_scan_lock);
354 lock_device_hotplug();
239 355
240 acpi_bus_get_device(handle, &device); 356 acpi_bus_get_device(handle, &device);
241 if (device) { 357 if (device) {
@@ -259,6 +375,7 @@ static void acpi_scan_bus_device_check(acpi_handle handle, u32 ost_source)
259 kobject_uevent(&device->dev.kobj, KOBJ_ONLINE); 375 kobject_uevent(&device->dev.kobj, KOBJ_ONLINE);
260 376
261 out: 377 out:
378 unlock_device_hotplug();
262 acpi_evaluate_hotplug_ost(handle, ost_source, ost_code, NULL); 379 acpi_evaluate_hotplug_ost(handle, ost_source, ost_code, NULL);
263 mutex_unlock(&acpi_scan_lock); 380 mutex_unlock(&acpi_scan_lock);
264} 381}
@@ -952,7 +1069,6 @@ int acpi_device_add(struct acpi_device *device,
952 printk(KERN_ERR PREFIX "Error creating sysfs interface for device %s\n", 1069 printk(KERN_ERR PREFIX "Error creating sysfs interface for device %s\n",
953 dev_name(&device->dev)); 1070 dev_name(&device->dev));
954 1071
955 device->removal_type = ACPI_BUS_REMOVAL_NORMAL;
956 return 0; 1072 return 0;
957 1073
958 err: 1074 err:
@@ -1939,7 +2055,6 @@ static acpi_status acpi_bus_device_detach(acpi_handle handle, u32 lvl_not_used,
1939 if (!acpi_bus_get_device(handle, &device)) { 2055 if (!acpi_bus_get_device(handle, &device)) {
1940 struct acpi_scan_handler *dev_handler = device->handler; 2056 struct acpi_scan_handler *dev_handler = device->handler;
1941 2057
1942 device->removal_type = ACPI_BUS_REMOVAL_EJECT;
1943 if (dev_handler) { 2058 if (dev_handler) {
1944 if (dev_handler->detach) 2059 if (dev_handler->detach)
1945 dev_handler->detach(device); 2060 dev_handler->detach(device);
@@ -2038,6 +2153,7 @@ int __init acpi_scan_init(void)
2038 2153
2039 acpi_pci_root_init(); 2154 acpi_pci_root_init();
2040 acpi_pci_link_init(); 2155 acpi_pci_link_init();
2156 acpi_processor_init();
2041 acpi_platform_init(); 2157 acpi_platform_init();
2042 acpi_lpss_init(); 2158 acpi_lpss_init();
2043 acpi_container_init(); 2159 acpi_container_init();
diff --git a/drivers/acpi/sysfs.c b/drivers/acpi/sysfs.c
index fcae5fa2e1b3..5c5d1624fa2c 100644
--- a/drivers/acpi/sysfs.c
+++ b/drivers/acpi/sysfs.c
@@ -780,6 +780,33 @@ void acpi_sysfs_add_hotplug_profile(struct acpi_hotplug_profile *hotplug,
780 pr_err(PREFIX "Unable to add hotplug profile '%s'\n", name); 780 pr_err(PREFIX "Unable to add hotplug profile '%s'\n", name);
781} 781}
782 782
783static ssize_t force_remove_show(struct kobject *kobj,
784 struct kobj_attribute *attr, char *buf)
785{
786 return sprintf(buf, "%d\n", !!acpi_force_hot_remove);
787}
788
789static ssize_t force_remove_store(struct kobject *kobj,
790 struct kobj_attribute *attr,
791 const char *buf, size_t size)
792{
793 bool val;
794 int ret;
795
796 ret = strtobool(buf, &val);
797 if (ret < 0)
798 return ret;
799
800 lock_device_hotplug();
801 acpi_force_hot_remove = val;
802 unlock_device_hotplug();
803 return size;
804}
805
806static const struct kobj_attribute force_remove_attr =
807 __ATTR(force_remove, S_IRUGO | S_IWUSR, force_remove_show,
808 force_remove_store);
809
783int __init acpi_sysfs_init(void) 810int __init acpi_sysfs_init(void)
784{ 811{
785 int result; 812 int result;
@@ -789,6 +816,10 @@ int __init acpi_sysfs_init(void)
789 return result; 816 return result;
790 817
791 hotplug_kobj = kobject_create_and_add("hotplug", acpi_kobj); 818 hotplug_kobj = kobject_create_and_add("hotplug", acpi_kobj);
819 result = sysfs_create_file(hotplug_kobj, &force_remove_attr.attr);
820 if (result)
821 return result;
822
792 result = sysfs_create_file(acpi_kobj, &pm_profile_attr.attr); 823 result = sysfs_create_file(acpi_kobj, &pm_profile_attr.attr);
793 return result; 824 return result;
794} 825}
diff --git a/drivers/base/core.c b/drivers/base/core.c
index 2499cefdcdf2..2166f34b7d84 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -403,6 +403,36 @@ static ssize_t store_uevent(struct device *dev, struct device_attribute *attr,
403static struct device_attribute uevent_attr = 403static struct device_attribute uevent_attr =
404 __ATTR(uevent, S_IRUGO | S_IWUSR, show_uevent, store_uevent); 404 __ATTR(uevent, S_IRUGO | S_IWUSR, show_uevent, store_uevent);
405 405
406static ssize_t show_online(struct device *dev, struct device_attribute *attr,
407 char *buf)
408{
409 bool val;
410
411 lock_device_hotplug();
412 val = !dev->offline;
413 unlock_device_hotplug();
414 return sprintf(buf, "%u\n", val);
415}
416
417static ssize_t store_online(struct device *dev, struct device_attribute *attr,
418 const char *buf, size_t count)
419{
420 bool val;
421 int ret;
422
423 ret = strtobool(buf, &val);
424 if (ret < 0)
425 return ret;
426
427 lock_device_hotplug();
428 ret = val ? device_online(dev) : device_offline(dev);
429 unlock_device_hotplug();
430 return ret < 0 ? ret : count;
431}
432
433static struct device_attribute online_attr =
434 __ATTR(online, S_IRUGO | S_IWUSR, show_online, store_online);
435
406static int device_add_attributes(struct device *dev, 436static int device_add_attributes(struct device *dev,
407 struct device_attribute *attrs) 437 struct device_attribute *attrs)
408{ 438{
@@ -516,6 +546,12 @@ static int device_add_attrs(struct device *dev)
516 if (error) 546 if (error)
517 goto err_remove_type_groups; 547 goto err_remove_type_groups;
518 548
549 if (device_supports_offline(dev) && !dev->offline_disabled) {
550 error = device_create_file(dev, &online_attr);
551 if (error)
552 goto err_remove_type_groups;
553 }
554
519 return 0; 555 return 0;
520 556
521 err_remove_type_groups: 557 err_remove_type_groups:
@@ -536,6 +572,7 @@ static void device_remove_attrs(struct device *dev)
536 struct class *class = dev->class; 572 struct class *class = dev->class;
537 const struct device_type *type = dev->type; 573 const struct device_type *type = dev->type;
538 574
575 device_remove_file(dev, &online_attr);
539 device_remove_groups(dev, dev->groups); 576 device_remove_groups(dev, dev->groups);
540 577
541 if (type) 578 if (type)
@@ -1433,6 +1470,99 @@ EXPORT_SYMBOL_GPL(put_device);
1433EXPORT_SYMBOL_GPL(device_create_file); 1470EXPORT_SYMBOL_GPL(device_create_file);
1434EXPORT_SYMBOL_GPL(device_remove_file); 1471EXPORT_SYMBOL_GPL(device_remove_file);
1435 1472
1473static DEFINE_MUTEX(device_hotplug_lock);
1474
1475void lock_device_hotplug(void)
1476{
1477 mutex_lock(&device_hotplug_lock);
1478}
1479
1480void unlock_device_hotplug(void)
1481{
1482 mutex_unlock(&device_hotplug_lock);
1483}
1484
1485static int device_check_offline(struct device *dev, void *not_used)
1486{
1487 int ret;
1488
1489 ret = device_for_each_child(dev, NULL, device_check_offline);
1490 if (ret)
1491 return ret;
1492
1493 return device_supports_offline(dev) && !dev->offline ? -EBUSY : 0;
1494}
1495
1496/**
1497 * device_offline - Prepare the device for hot-removal.
1498 * @dev: Device to be put offline.
1499 *
1500 * Execute the device bus type's .offline() callback, if present, to prepare
1501 * the device for a subsequent hot-removal. If that succeeds, the device must
1502 * not be used until either it is removed or its bus type's .online() callback
1503 * is executed.
1504 *
1505 * Call under device_hotplug_lock.
1506 */
1507int device_offline(struct device *dev)
1508{
1509 int ret;
1510
1511 if (dev->offline_disabled)
1512 return -EPERM;
1513
1514 ret = device_for_each_child(dev, NULL, device_check_offline);
1515 if (ret)
1516 return ret;
1517
1518 device_lock(dev);
1519 if (device_supports_offline(dev)) {
1520 if (dev->offline) {
1521 ret = 1;
1522 } else {
1523 ret = dev->bus->offline(dev);
1524 if (!ret) {
1525 kobject_uevent(&dev->kobj, KOBJ_OFFLINE);
1526 dev->offline = true;
1527 }
1528 }
1529 }
1530 device_unlock(dev);
1531
1532 return ret;
1533}
1534
1535/**
1536 * device_online - Put the device back online after successful device_offline().
1537 * @dev: Device to be put back online.
1538 *
1539 * If device_offline() has been successfully executed for @dev, but the device
1540 * has not been removed subsequently, execute its bus type's .online() callback
1541 * to indicate that the device can be used again.
1542 *
1543 * Call under device_hotplug_lock.
1544 */
1545int device_online(struct device *dev)
1546{
1547 int ret = 0;
1548
1549 device_lock(dev);
1550 if (device_supports_offline(dev)) {
1551 if (dev->offline) {
1552 ret = dev->bus->online(dev);
1553 if (!ret) {
1554 kobject_uevent(&dev->kobj, KOBJ_ONLINE);
1555 dev->offline = false;
1556 }
1557 } else {
1558 ret = 1;
1559 }
1560 }
1561 device_unlock(dev);
1562
1563 return ret;
1564}
1565
1436struct root_device { 1566struct root_device {
1437 struct device dev; 1567 struct device dev;
1438 struct module *owner; 1568 struct module *owner;
diff --git a/drivers/base/cpu.c b/drivers/base/cpu.c
index 3d48fc887ef4..1d110dc6f0c1 100644
--- a/drivers/base/cpu.c
+++ b/drivers/base/cpu.c
@@ -13,17 +13,21 @@
13#include <linux/gfp.h> 13#include <linux/gfp.h>
14#include <linux/slab.h> 14#include <linux/slab.h>
15#include <linux/percpu.h> 15#include <linux/percpu.h>
16#include <linux/acpi.h>
16 17
17#include "base.h" 18#include "base.h"
18 19
19struct bus_type cpu_subsys = {
20 .name = "cpu",
21 .dev_name = "cpu",
22};
23EXPORT_SYMBOL_GPL(cpu_subsys);
24
25static DEFINE_PER_CPU(struct device *, cpu_sys_devices); 20static DEFINE_PER_CPU(struct device *, cpu_sys_devices);
26 21
22static int cpu_subsys_match(struct device *dev, struct device_driver *drv)
23{
24 /* ACPI style match is the only one that may succeed. */
25 if (acpi_driver_match_device(dev, drv))
26 return 1;
27
28 return 0;
29}
30
27#ifdef CONFIG_HOTPLUG_CPU 31#ifdef CONFIG_HOTPLUG_CPU
28static void change_cpu_under_node(struct cpu *cpu, 32static void change_cpu_under_node(struct cpu *cpu,
29 unsigned int from_nid, unsigned int to_nid) 33 unsigned int from_nid, unsigned int to_nid)
@@ -34,69 +38,45 @@ static void change_cpu_under_node(struct cpu *cpu,
34 cpu->node_id = to_nid; 38 cpu->node_id = to_nid;
35} 39}
36 40
37static ssize_t show_online(struct device *dev, 41static int __ref cpu_subsys_online(struct device *dev)
38 struct device_attribute *attr,
39 char *buf)
40{ 42{
41 struct cpu *cpu = container_of(dev, struct cpu, dev); 43 struct cpu *cpu = container_of(dev, struct cpu, dev);
44 int cpuid = dev->id;
45 int from_nid, to_nid;
46 int ret;
42 47
43 return sprintf(buf, "%u\n", !!cpu_online(cpu->dev.id)); 48 cpu_hotplug_driver_lock();
49
50 from_nid = cpu_to_node(cpuid);
51 ret = cpu_up(cpuid);
52 /*
53 * When hot adding memory to memoryless node and enabling a cpu
54 * on the node, node number of the cpu may internally change.
55 */
56 to_nid = cpu_to_node(cpuid);
57 if (from_nid != to_nid)
58 change_cpu_under_node(cpu, from_nid, to_nid);
59
60 cpu_hotplug_driver_unlock();
61 return ret;
44} 62}
45 63
46static ssize_t __ref store_online(struct device *dev, 64static int cpu_subsys_offline(struct device *dev)
47 struct device_attribute *attr,
48 const char *buf, size_t count)
49{ 65{
50 struct cpu *cpu = container_of(dev, struct cpu, dev); 66 int ret;
51 int cpuid = cpu->dev.id;
52 int from_nid, to_nid;
53 ssize_t ret;
54 67
55 cpu_hotplug_driver_lock(); 68 cpu_hotplug_driver_lock();
56 switch (buf[0]) { 69 ret = cpu_down(dev->id);
57 case '0':
58 ret = cpu_down(cpuid);
59 if (!ret)
60 kobject_uevent(&dev->kobj, KOBJ_OFFLINE);
61 break;
62 case '1':
63 from_nid = cpu_to_node(cpuid);
64 ret = cpu_up(cpuid);
65
66 /*
67 * When hot adding memory to memoryless node and enabling a cpu
68 * on the node, node number of the cpu may internally change.
69 */
70 to_nid = cpu_to_node(cpuid);
71 if (from_nid != to_nid)
72 change_cpu_under_node(cpu, from_nid, to_nid);
73
74 if (!ret)
75 kobject_uevent(&dev->kobj, KOBJ_ONLINE);
76 break;
77 default:
78 ret = -EINVAL;
79 }
80 cpu_hotplug_driver_unlock(); 70 cpu_hotplug_driver_unlock();
81
82 if (ret >= 0)
83 ret = count;
84 return ret; 71 return ret;
85} 72}
86static DEVICE_ATTR(online, 0644, show_online, store_online);
87 73
88static void __cpuinit register_cpu_control(struct cpu *cpu)
89{
90 device_create_file(&cpu->dev, &dev_attr_online);
91}
92void unregister_cpu(struct cpu *cpu) 74void unregister_cpu(struct cpu *cpu)
93{ 75{
94 int logical_cpu = cpu->dev.id; 76 int logical_cpu = cpu->dev.id;
95 77
96 unregister_cpu_under_node(logical_cpu, cpu_to_node(logical_cpu)); 78 unregister_cpu_under_node(logical_cpu, cpu_to_node(logical_cpu));
97 79
98 device_remove_file(&cpu->dev, &dev_attr_online);
99
100 device_unregister(&cpu->dev); 80 device_unregister(&cpu->dev);
101 per_cpu(cpu_sys_devices, logical_cpu) = NULL; 81 per_cpu(cpu_sys_devices, logical_cpu) = NULL;
102 return; 82 return;
@@ -123,12 +103,19 @@ static DEVICE_ATTR(probe, S_IWUSR, NULL, cpu_probe_store);
123static DEVICE_ATTR(release, S_IWUSR, NULL, cpu_release_store); 103static DEVICE_ATTR(release, S_IWUSR, NULL, cpu_release_store);
124#endif /* CONFIG_ARCH_CPU_PROBE_RELEASE */ 104#endif /* CONFIG_ARCH_CPU_PROBE_RELEASE */
125 105
126#else /* ... !CONFIG_HOTPLUG_CPU */
127static inline void register_cpu_control(struct cpu *cpu)
128{
129}
130#endif /* CONFIG_HOTPLUG_CPU */ 106#endif /* CONFIG_HOTPLUG_CPU */
131 107
108struct bus_type cpu_subsys = {
109 .name = "cpu",
110 .dev_name = "cpu",
111 .match = cpu_subsys_match,
112#ifdef CONFIG_HOTPLUG_CPU
113 .online = cpu_subsys_online,
114 .offline = cpu_subsys_offline,
115#endif
116};
117EXPORT_SYMBOL_GPL(cpu_subsys);
118
132#ifdef CONFIG_KEXEC 119#ifdef CONFIG_KEXEC
133#include <linux/kexec.h> 120#include <linux/kexec.h>
134 121
@@ -277,12 +264,12 @@ int __cpuinit register_cpu(struct cpu *cpu, int num)
277 cpu->dev.id = num; 264 cpu->dev.id = num;
278 cpu->dev.bus = &cpu_subsys; 265 cpu->dev.bus = &cpu_subsys;
279 cpu->dev.release = cpu_device_release; 266 cpu->dev.release = cpu_device_release;
267 cpu->dev.offline_disabled = !cpu->hotpluggable;
268 cpu->dev.offline = !cpu_online(num);
280#ifdef CONFIG_ARCH_HAS_CPU_AUTOPROBE 269#ifdef CONFIG_ARCH_HAS_CPU_AUTOPROBE
281 cpu->dev.bus->uevent = arch_cpu_uevent; 270 cpu->dev.bus->uevent = arch_cpu_uevent;
282#endif 271#endif
283 error = device_register(&cpu->dev); 272 error = device_register(&cpu->dev);
284 if (!error && cpu->hotpluggable)
285 register_cpu_control(cpu);
286 if (!error) 273 if (!error)
287 per_cpu(cpu_sys_devices, num) = &cpu->dev; 274 per_cpu(cpu_sys_devices, num) = &cpu->dev;
288 if (!error) 275 if (!error)
diff --git a/drivers/base/memory.c b/drivers/base/memory.c
index 14f8a6954da0..4ebf97f99fae 100644
--- a/drivers/base/memory.c
+++ b/drivers/base/memory.c
@@ -37,9 +37,14 @@ static inline int base_memory_block_id(int section_nr)
37 return section_nr / sections_per_block; 37 return section_nr / sections_per_block;
38} 38}
39 39
40static int memory_subsys_online(struct device *dev);
41static int memory_subsys_offline(struct device *dev);
42
40static struct bus_type memory_subsys = { 43static struct bus_type memory_subsys = {
41 .name = MEMORY_CLASS_NAME, 44 .name = MEMORY_CLASS_NAME,
42 .dev_name = MEMORY_CLASS_NAME, 45 .dev_name = MEMORY_CLASS_NAME,
46 .online = memory_subsys_online,
47 .offline = memory_subsys_offline,
43}; 48};
44 49
45static BLOCKING_NOTIFIER_HEAD(memory_chain); 50static BLOCKING_NOTIFIER_HEAD(memory_chain);
@@ -88,6 +93,7 @@ int register_memory(struct memory_block *memory)
88 memory->dev.bus = &memory_subsys; 93 memory->dev.bus = &memory_subsys;
89 memory->dev.id = memory->start_section_nr / sections_per_block; 94 memory->dev.id = memory->start_section_nr / sections_per_block;
90 memory->dev.release = memory_block_release; 95 memory->dev.release = memory_block_release;
96 memory->dev.offline = memory->state == MEM_OFFLINE;
91 97
92 error = device_register(&memory->dev); 98 error = device_register(&memory->dev);
93 return error; 99 return error;
@@ -278,33 +284,64 @@ static int __memory_block_change_state(struct memory_block *mem,
278{ 284{
279 int ret = 0; 285 int ret = 0;
280 286
281 if (mem->state != from_state_req) { 287 if (mem->state != from_state_req)
282 ret = -EINVAL; 288 return -EINVAL;
283 goto out;
284 }
285 289
286 if (to_state == MEM_OFFLINE) 290 if (to_state == MEM_OFFLINE)
287 mem->state = MEM_GOING_OFFLINE; 291 mem->state = MEM_GOING_OFFLINE;
288 292
289 ret = memory_block_action(mem->start_section_nr, to_state, online_type); 293 ret = memory_block_action(mem->start_section_nr, to_state, online_type);
294 mem->state = ret ? from_state_req : to_state;
295 return ret;
296}
290 297
291 if (ret) { 298static int memory_subsys_online(struct device *dev)
292 mem->state = from_state_req; 299{
293 goto out; 300 struct memory_block *mem = container_of(dev, struct memory_block, dev);
294 } 301 int ret;
295 302
296 mem->state = to_state; 303 mutex_lock(&mem->state_mutex);
297 switch (mem->state) { 304
298 case MEM_OFFLINE: 305 ret = mem->state == MEM_ONLINE ? 0 :
299 kobject_uevent(&mem->dev.kobj, KOBJ_OFFLINE); 306 __memory_block_change_state(mem, MEM_ONLINE, MEM_OFFLINE,
300 break; 307 ONLINE_KEEP);
301 case MEM_ONLINE: 308
302 kobject_uevent(&mem->dev.kobj, KOBJ_ONLINE); 309 mutex_unlock(&mem->state_mutex);
303 break; 310 return ret;
304 default: 311}
305 break; 312
313static int memory_subsys_offline(struct device *dev)
314{
315 struct memory_block *mem = container_of(dev, struct memory_block, dev);
316 int ret;
317
318 mutex_lock(&mem->state_mutex);
319
320 ret = mem->state == MEM_OFFLINE ? 0 :
321 __memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE, -1);
322
323 mutex_unlock(&mem->state_mutex);
324 return ret;
325}
326
327static int __memory_block_change_state_uevent(struct memory_block *mem,
328 unsigned long to_state, unsigned long from_state_req,
329 int online_type)
330{
331 int ret = __memory_block_change_state(mem, to_state, from_state_req,
332 online_type);
333 if (!ret) {
334 switch (mem->state) {
335 case MEM_OFFLINE:
336 kobject_uevent(&mem->dev.kobj, KOBJ_OFFLINE);
337 break;
338 case MEM_ONLINE:
339 kobject_uevent(&mem->dev.kobj, KOBJ_ONLINE);
340 break;
341 default:
342 break;
343 }
306 } 344 }
307out:
308 return ret; 345 return ret;
309} 346}
310 347
@@ -315,8 +352,8 @@ static int memory_block_change_state(struct memory_block *mem,
315 int ret; 352 int ret;
316 353
317 mutex_lock(&mem->state_mutex); 354 mutex_lock(&mem->state_mutex);
318 ret = __memory_block_change_state(mem, to_state, from_state_req, 355 ret = __memory_block_change_state_uevent(mem, to_state, from_state_req,
319 online_type); 356 online_type);
320 mutex_unlock(&mem->state_mutex); 357 mutex_unlock(&mem->state_mutex);
321 358
322 return ret; 359 return ret;
@@ -326,22 +363,34 @@ store_mem_state(struct device *dev,
326 struct device_attribute *attr, const char *buf, size_t count) 363 struct device_attribute *attr, const char *buf, size_t count)
327{ 364{
328 struct memory_block *mem; 365 struct memory_block *mem;
366 bool offline;
329 int ret = -EINVAL; 367 int ret = -EINVAL;
330 368
331 mem = container_of(dev, struct memory_block, dev); 369 mem = container_of(dev, struct memory_block, dev);
332 370
333 if (!strncmp(buf, "online_kernel", min_t(int, count, 13))) 371 lock_device_hotplug();
372
373 if (!strncmp(buf, "online_kernel", min_t(int, count, 13))) {
374 offline = false;
334 ret = memory_block_change_state(mem, MEM_ONLINE, 375 ret = memory_block_change_state(mem, MEM_ONLINE,
335 MEM_OFFLINE, ONLINE_KERNEL); 376 MEM_OFFLINE, ONLINE_KERNEL);
336 else if (!strncmp(buf, "online_movable", min_t(int, count, 14))) 377 } else if (!strncmp(buf, "online_movable", min_t(int, count, 14))) {
378 offline = false;
337 ret = memory_block_change_state(mem, MEM_ONLINE, 379 ret = memory_block_change_state(mem, MEM_ONLINE,
338 MEM_OFFLINE, ONLINE_MOVABLE); 380 MEM_OFFLINE, ONLINE_MOVABLE);
339 else if (!strncmp(buf, "online", min_t(int, count, 6))) 381 } else if (!strncmp(buf, "online", min_t(int, count, 6))) {
382 offline = false;
340 ret = memory_block_change_state(mem, MEM_ONLINE, 383 ret = memory_block_change_state(mem, MEM_ONLINE,
341 MEM_OFFLINE, ONLINE_KEEP); 384 MEM_OFFLINE, ONLINE_KEEP);
342 else if(!strncmp(buf, "offline", min_t(int, count, 7))) 385 } else if(!strncmp(buf, "offline", min_t(int, count, 7))) {
386 offline = true;
343 ret = memory_block_change_state(mem, MEM_OFFLINE, 387 ret = memory_block_change_state(mem, MEM_OFFLINE,
344 MEM_ONLINE, -1); 388 MEM_ONLINE, -1);
389 }
390 if (!ret)
391 dev->offline = offline;
392
393 unlock_device_hotplug();
345 394
346 if (ret) 395 if (ret)
347 return ret; 396 return ret;
@@ -679,21 +728,6 @@ int unregister_memory_section(struct mem_section *section)
679} 728}
680#endif /* CONFIG_MEMORY_HOTREMOVE */ 729#endif /* CONFIG_MEMORY_HOTREMOVE */
681 730
682/*
683 * offline one memory block. If the memory block has been offlined, do nothing.
684 */
685int offline_memory_block(struct memory_block *mem)
686{
687 int ret = 0;
688
689 mutex_lock(&mem->state_mutex);
690 if (mem->state != MEM_OFFLINE)
691 ret = __memory_block_change_state(mem, MEM_OFFLINE, MEM_ONLINE, -1);
692 mutex_unlock(&mem->state_mutex);
693
694 return ret;
695}
696
697/* return true if the memory block is offlined, otherwise, return false */ 731/* return true if the memory block is offlined, otherwise, return false */
698bool is_memblock_offlined(struct memory_block *mem) 732bool is_memblock_offlined(struct memory_block *mem)
699{ 733{
diff --git a/drivers/cpufreq/acpi-cpufreq.c b/drivers/cpufreq/acpi-cpufreq.c
index edc089e9d0c4..b8efacf01d69 100644
--- a/drivers/cpufreq/acpi-cpufreq.c
+++ b/drivers/cpufreq/acpi-cpufreq.c
@@ -1034,4 +1034,11 @@ static const struct x86_cpu_id acpi_cpufreq_ids[] = {
1034}; 1034};
1035MODULE_DEVICE_TABLE(x86cpu, acpi_cpufreq_ids); 1035MODULE_DEVICE_TABLE(x86cpu, acpi_cpufreq_ids);
1036 1036
1037static const struct acpi_device_id processor_device_ids[] = {
1038 {ACPI_PROCESSOR_OBJECT_HID, },
1039 {ACPI_PROCESSOR_DEVICE_HID, },
1040 {},
1041};
1042MODULE_DEVICE_TABLE(acpi, processor_device_ids);
1043
1037MODULE_ALIAS("acpi"); 1044MODULE_ALIAS("acpi");
diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
index c13c919ab99e..cf0817a555e4 100644
--- a/include/acpi/acpi_bus.h
+++ b/include/acpi/acpi_bus.h
@@ -63,13 +63,6 @@ acpi_get_physical_device_location(acpi_handle handle, struct acpi_pld_info **pld
63#define ACPI_BUS_FILE_ROOT "acpi" 63#define ACPI_BUS_FILE_ROOT "acpi"
64extern struct proc_dir_entry *acpi_root_dir; 64extern struct proc_dir_entry *acpi_root_dir;
65 65
66enum acpi_bus_removal_type {
67 ACPI_BUS_REMOVAL_NORMAL = 0,
68 ACPI_BUS_REMOVAL_EJECT,
69 ACPI_BUS_REMOVAL_SUPRISE,
70 ACPI_BUS_REMOVAL_TYPE_COUNT
71};
72
73enum acpi_bus_device_type { 66enum acpi_bus_device_type {
74 ACPI_BUS_TYPE_DEVICE = 0, 67 ACPI_BUS_TYPE_DEVICE = 0,
75 ACPI_BUS_TYPE_POWER, 68 ACPI_BUS_TYPE_POWER,
@@ -286,6 +279,7 @@ struct acpi_device_physical_node {
286 u8 node_id; 279 u8 node_id;
287 struct list_head node; 280 struct list_head node;
288 struct device *dev; 281 struct device *dev;
282 bool put_online:1;
289}; 283};
290 284
291/* set maximum of physical nodes to 32 for expansibility */ 285/* set maximum of physical nodes to 32 for expansibility */
@@ -310,7 +304,6 @@ struct acpi_device {
310 struct acpi_driver *driver; 304 struct acpi_driver *driver;
311 void *driver_data; 305 void *driver_data;
312 struct device dev; 306 struct device dev;
313 enum acpi_bus_removal_type removal_type; /* indicate for different removal type */
314 u8 physical_node_count; 307 u8 physical_node_count;
315 struct list_head physical_node_list; 308 struct list_head physical_node_list;
316 struct mutex physical_node_lock; 309 struct mutex physical_node_lock;
diff --git a/include/acpi/processor.h b/include/acpi/processor.h
index ea69367fdd3b..66096d06925e 100644
--- a/include/acpi/processor.h
+++ b/include/acpi/processor.h
@@ -6,6 +6,10 @@
6#include <linux/thermal.h> 6#include <linux/thermal.h>
7#include <asm/acpi.h> 7#include <asm/acpi.h>
8 8
9#define ACPI_PROCESSOR_CLASS "processor"
10#define ACPI_PROCESSOR_DEVICE_NAME "Processor"
11#define ACPI_PROCESSOR_DEVICE_HID "ACPI0007"
12
9#define ACPI_PROCESSOR_BUSY_METRIC 10 13#define ACPI_PROCESSOR_BUSY_METRIC 10
10 14
11#define ACPI_PROCESSOR_MAX_POWER 8 15#define ACPI_PROCESSOR_MAX_POWER 8
@@ -207,6 +211,7 @@ struct acpi_processor {
207 struct acpi_processor_throttling throttling; 211 struct acpi_processor_throttling throttling;
208 struct acpi_processor_limit limit; 212 struct acpi_processor_limit limit;
209 struct thermal_cooling_device *cdev; 213 struct thermal_cooling_device *cdev;
214 struct device *dev; /* Processor device. */
210}; 215};
211 216
212struct acpi_processor_errata { 217struct acpi_processor_errata {
diff --git a/include/linux/acpi.h b/include/linux/acpi.h
index 17b5b5967641..353ba256f368 100644
--- a/include/linux/acpi.h
+++ b/include/linux/acpi.h
@@ -352,8 +352,7 @@ extern acpi_status acpi_pci_osc_control_set(acpi_handle handle,
352 352
353/* Enable _OST when all relevant hotplug operations are enabled */ 353/* Enable _OST when all relevant hotplug operations are enabled */
354#if defined(CONFIG_ACPI_HOTPLUG_CPU) && \ 354#if defined(CONFIG_ACPI_HOTPLUG_CPU) && \
355 (defined(CONFIG_ACPI_HOTPLUG_MEMORY) || \ 355 defined(CONFIG_ACPI_HOTPLUG_MEMORY) && \
356 defined(CONFIG_ACPI_HOTPLUG_MEMORY_MODULE)) && \
357 defined(CONFIG_ACPI_CONTAINER) 356 defined(CONFIG_ACPI_CONTAINER)
358#define ACPI_HOTPLUG_OST 357#define ACPI_HOTPLUG_OST
359#endif 358#endif
diff --git a/include/linux/device.h b/include/linux/device.h
index c0a126125325..eeb33315514c 100644
--- a/include/linux/device.h
+++ b/include/linux/device.h
@@ -71,6 +71,10 @@ extern void bus_remove_file(struct bus_type *, struct bus_attribute *);
71 * the specific driver's probe to initial the matched device. 71 * the specific driver's probe to initial the matched device.
72 * @remove: Called when a device removed from this bus. 72 * @remove: Called when a device removed from this bus.
73 * @shutdown: Called at shut-down time to quiesce the device. 73 * @shutdown: Called at shut-down time to quiesce the device.
74 *
75 * @online: Called to put the device back online (after offlining it).
76 * @offline: Called to put the device offline for hot-removal. May fail.
77 *
74 * @suspend: Called when a device on this bus wants to go to sleep mode. 78 * @suspend: Called when a device on this bus wants to go to sleep mode.
75 * @resume: Called to bring a device on this bus out of sleep mode. 79 * @resume: Called to bring a device on this bus out of sleep mode.
76 * @pm: Power management operations of this bus, callback the specific 80 * @pm: Power management operations of this bus, callback the specific
@@ -104,6 +108,9 @@ struct bus_type {
104 int (*remove)(struct device *dev); 108 int (*remove)(struct device *dev);
105 void (*shutdown)(struct device *dev); 109 void (*shutdown)(struct device *dev);
106 110
111 int (*online)(struct device *dev);
112 int (*offline)(struct device *dev);
113
107 int (*suspend)(struct device *dev, pm_message_t state); 114 int (*suspend)(struct device *dev, pm_message_t state);
108 int (*resume)(struct device *dev); 115 int (*resume)(struct device *dev);
109 116
@@ -648,6 +655,8 @@ struct acpi_dev_node {
648 * @release: Callback to free the device after all references have 655 * @release: Callback to free the device after all references have
649 * gone away. This should be set by the allocator of the 656 * gone away. This should be set by the allocator of the
650 * device (i.e. the bus driver that discovered the device). 657 * device (i.e. the bus driver that discovered the device).
658 * @offline_disabled: If set, the device is permanently online.
659 * @offline: Set after successful invocation of bus type's .offline().
651 * 660 *
652 * At the lowest level, every device in a Linux system is represented by an 661 * At the lowest level, every device in a Linux system is represented by an
653 * instance of struct device. The device structure contains the information 662 * instance of struct device. The device structure contains the information
@@ -720,6 +729,9 @@ struct device {
720 729
721 void (*release)(struct device *dev); 730 void (*release)(struct device *dev);
722 struct iommu_group *iommu_group; 731 struct iommu_group *iommu_group;
732
733 bool offline_disabled:1;
734 bool offline:1;
723}; 735};
724 736
725static inline struct device *kobj_to_dev(struct kobject *kobj) 737static inline struct device *kobj_to_dev(struct kobject *kobj)
@@ -856,6 +868,15 @@ extern const char *device_get_devnode(struct device *dev,
856extern void *dev_get_drvdata(const struct device *dev); 868extern void *dev_get_drvdata(const struct device *dev);
857extern int dev_set_drvdata(struct device *dev, void *data); 869extern int dev_set_drvdata(struct device *dev, void *data);
858 870
871static inline bool device_supports_offline(struct device *dev)
872{
873 return dev->bus && dev->bus->offline && dev->bus->online;
874}
875
876extern void lock_device_hotplug(void);
877extern void unlock_device_hotplug(void);
878extern int device_offline(struct device *dev);
879extern int device_online(struct device *dev);
859/* 880/*
860 * Root device objects for grouping under /sys/devices 881 * Root device objects for grouping under /sys/devices
861 */ 882 */
diff --git a/include/linux/memory_hotplug.h b/include/linux/memory_hotplug.h
index 3e622c610925..dd38e62b84d2 100644
--- a/include/linux/memory_hotplug.h
+++ b/include/linux/memory_hotplug.h
@@ -234,6 +234,8 @@ static inline void unlock_memory_hotplug(void) {}
234 234
235extern int is_mem_section_removable(unsigned long pfn, unsigned long nr_pages); 235extern int is_mem_section_removable(unsigned long pfn, unsigned long nr_pages);
236extern void try_offline_node(int nid); 236extern void try_offline_node(int nid);
237extern int offline_pages(unsigned long start_pfn, unsigned long nr_pages);
238extern void remove_memory(int nid, u64 start, u64 size);
237 239
238#else 240#else
239static inline int is_mem_section_removable(unsigned long pfn, 241static inline int is_mem_section_removable(unsigned long pfn,
@@ -243,15 +245,23 @@ static inline int is_mem_section_removable(unsigned long pfn,
243} 245}
244 246
245static inline void try_offline_node(int nid) {} 247static inline void try_offline_node(int nid) {}
248
249static inline int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
250{
251 return -EINVAL;
252}
253
254static inline void remove_memory(int nid, u64 start, u64 size) {}
246#endif /* CONFIG_MEMORY_HOTREMOVE */ 255#endif /* CONFIG_MEMORY_HOTREMOVE */
247 256
257extern int walk_memory_range(unsigned long start_pfn, unsigned long end_pfn,
258 void *arg, int (*func)(struct memory_block *, void *));
248extern int mem_online_node(int nid); 259extern int mem_online_node(int nid);
249extern int add_memory(int nid, u64 start, u64 size); 260extern int add_memory(int nid, u64 start, u64 size);
250extern int arch_add_memory(int nid, u64 start, u64 size); 261extern int arch_add_memory(int nid, u64 start, u64 size);
251extern int offline_pages(unsigned long start_pfn, unsigned long nr_pages); 262extern int offline_pages(unsigned long start_pfn, unsigned long nr_pages);
252extern int offline_memory_block(struct memory_block *mem);
253extern bool is_memblock_offlined(struct memory_block *mem); 263extern bool is_memblock_offlined(struct memory_block *mem);
254extern int remove_memory(int nid, u64 start, u64 size); 264extern void remove_memory(int nid, u64 start, u64 size);
255extern int sparse_add_one_section(struct zone *zone, unsigned long start_pfn, 265extern int sparse_add_one_section(struct zone *zone, unsigned long start_pfn,
256 int nr_pages); 266 int nr_pages);
257extern void sparse_remove_one_section(struct zone *zone, struct mem_section *ms); 267extern void sparse_remove_one_section(struct zone *zone, struct mem_section *ms);
diff --git a/mm/memory_hotplug.c b/mm/memory_hotplug.c
index 1ad92b46753e..081b4d654ed6 100644
--- a/mm/memory_hotplug.c
+++ b/mm/memory_hotplug.c
@@ -1621,6 +1621,7 @@ int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
1621{ 1621{
1622 return __offline_pages(start_pfn, start_pfn + nr_pages, 120 * HZ); 1622 return __offline_pages(start_pfn, start_pfn + nr_pages, 120 * HZ);
1623} 1623}
1624#endif /* CONFIG_MEMORY_HOTREMOVE */
1624 1625
1625/** 1626/**
1626 * walk_memory_range - walks through all mem sections in [start_pfn, end_pfn) 1627 * walk_memory_range - walks through all mem sections in [start_pfn, end_pfn)
@@ -1634,7 +1635,7 @@ int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
1634 * 1635 *
1635 * Returns the return value of func. 1636 * Returns the return value of func.
1636 */ 1637 */
1637static int walk_memory_range(unsigned long start_pfn, unsigned long end_pfn, 1638int walk_memory_range(unsigned long start_pfn, unsigned long end_pfn,
1638 void *arg, int (*func)(struct memory_block *, void *)) 1639 void *arg, int (*func)(struct memory_block *, void *))
1639{ 1640{
1640 struct memory_block *mem = NULL; 1641 struct memory_block *mem = NULL;
@@ -1671,24 +1672,7 @@ static int walk_memory_range(unsigned long start_pfn, unsigned long end_pfn,
1671 return 0; 1672 return 0;
1672} 1673}
1673 1674
1674/** 1675#ifdef CONFIG_MEMORY_HOTREMOVE
1675 * offline_memory_block_cb - callback function for offlining memory block
1676 * @mem: the memory block to be offlined
1677 * @arg: buffer to hold error msg
1678 *
1679 * Always return 0, and put the error msg in arg if any.
1680 */
1681static int offline_memory_block_cb(struct memory_block *mem, void *arg)
1682{
1683 int *ret = arg;
1684 int error = offline_memory_block(mem);
1685
1686 if (error != 0 && *ret == 0)
1687 *ret = error;
1688
1689 return 0;
1690}
1691
1692static int is_memblock_offlined_cb(struct memory_block *mem, void *arg) 1676static int is_memblock_offlined_cb(struct memory_block *mem, void *arg)
1693{ 1677{
1694 int ret = !is_memblock_offlined(mem); 1678 int ret = !is_memblock_offlined(mem);
@@ -1814,54 +1798,22 @@ void try_offline_node(int nid)
1814} 1798}
1815EXPORT_SYMBOL(try_offline_node); 1799EXPORT_SYMBOL(try_offline_node);
1816 1800
1817int __ref remove_memory(int nid, u64 start, u64 size) 1801void __ref remove_memory(int nid, u64 start, u64 size)
1818{ 1802{
1819 unsigned long start_pfn, end_pfn; 1803 int ret;
1820 int ret = 0;
1821 int retry = 1;
1822
1823 start_pfn = PFN_DOWN(start);
1824 end_pfn = PFN_UP(start + size - 1);
1825
1826 /*
1827 * When CONFIG_MEMCG is on, one memory block may be used by other
1828 * blocks to store page cgroup when onlining pages. But we don't know
1829 * in what order pages are onlined. So we iterate twice to offline
1830 * memory:
1831 * 1st iterate: offline every non primary memory block.
1832 * 2nd iterate: offline primary (i.e. first added) memory block.
1833 */
1834repeat:
1835 walk_memory_range(start_pfn, end_pfn, &ret,
1836 offline_memory_block_cb);
1837 if (ret) {
1838 if (!retry)
1839 return ret;
1840
1841 retry = 0;
1842 ret = 0;
1843 goto repeat;
1844 }
1845 1804
1846 lock_memory_hotplug(); 1805 lock_memory_hotplug();
1847 1806
1848 /* 1807 /*
1849 * we have offlined all memory blocks like this: 1808 * All memory blocks must be offlined before removing memory. Check
1850 * 1. lock memory hotplug 1809 * whether all memory blocks in question are offline and trigger a BUG()
1851 * 2. offline a memory block 1810 * if this is not the case.
1852 * 3. unlock memory hotplug
1853 *
1854 * repeat step1-3 to offline the memory block. All memory blocks
1855 * must be offlined before removing memory. But we don't hold the
1856 * lock in the whole operation. So we should check whether all
1857 * memory blocks are offlined.
1858 */ 1811 */
1859 1812 ret = walk_memory_range(PFN_DOWN(start), PFN_UP(start + size - 1), NULL,
1860 ret = walk_memory_range(start_pfn, end_pfn, NULL,
1861 is_memblock_offlined_cb); 1813 is_memblock_offlined_cb);
1862 if (ret) { 1814 if (ret) {
1863 unlock_memory_hotplug(); 1815 unlock_memory_hotplug();
1864 return ret; 1816 BUG();
1865 } 1817 }
1866 1818
1867 /* remove memmap entry */ 1819 /* remove memmap entry */
@@ -1872,17 +1824,6 @@ repeat:
1872 try_offline_node(nid); 1824 try_offline_node(nid);
1873 1825
1874 unlock_memory_hotplug(); 1826 unlock_memory_hotplug();
1875
1876 return 0;
1877} 1827}
1878#else
1879int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
1880{
1881 return -EINVAL;
1882}
1883int remove_memory(int nid, u64 start, u64 size)
1884{
1885 return -EINVAL;
1886}
1887#endif /* CONFIG_MEMORY_HOTREMOVE */
1888EXPORT_SYMBOL_GPL(remove_memory); 1828EXPORT_SYMBOL_GPL(remove_memory);
1829#endif /* CONFIG_MEMORY_HOTREMOVE */
generated by cgit v1.2.2 (git 2.25.0) at 2026-07-08 00:10:51 -0400