]>
Commit | Line | Data |
---|---|---|
f65c9bb3 LJ |
1 | /****************************************************************************** |
2 | * pcpu.c | |
3 | * Management physical cpu in dom0, get pcpu info and provide sys interface | |
4 | * | |
5 | * Copyright (c) 2012 Intel Corporation | |
6 | * Author: Liu, Jinsong <[email protected]> | |
7 | * Author: Jiang, Yunhong <[email protected]> | |
8 | * | |
9 | * This program is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU General Public License version 2 | |
11 | * as published by the Free Software Foundation; or, when distributed | |
12 | * separately from the Linux kernel or incorporated into other | |
13 | * software packages, subject to the following license: | |
14 | * | |
15 | * Permission is hereby granted, free of charge, to any person obtaining a copy | |
16 | * of this source file (the "Software"), to deal in the Software without | |
17 | * restriction, including without limitation the rights to use, copy, modify, | |
18 | * merge, publish, distribute, sublicense, and/or sell copies of the Software, | |
19 | * and to permit persons to whom the Software is furnished to do so, subject to | |
20 | * the following conditions: | |
21 | * | |
22 | * The above copyright notice and this permission notice shall be included in | |
23 | * all copies or substantial portions of the Software. | |
24 | * | |
25 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR | |
26 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, | |
27 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE | |
28 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER | |
29 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING | |
30 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS | |
31 | * IN THE SOFTWARE. | |
32 | */ | |
33 | ||
283c0972 JP |
34 | #define pr_fmt(fmt) "xen_cpu: " fmt |
35 | ||
f65c9bb3 LJ |
36 | #include <linux/interrupt.h> |
37 | #include <linux/spinlock.h> | |
38 | #include <linux/cpu.h> | |
39 | #include <linux/stat.h> | |
40 | #include <linux/capability.h> | |
41 | ||
42 | #include <xen/xen.h> | |
6cb606f1 | 43 | #include <xen/acpi.h> |
f65c9bb3 LJ |
44 | #include <xen/xenbus.h> |
45 | #include <xen/events.h> | |
46 | #include <xen/interface/platform.h> | |
47 | #include <asm/xen/hypervisor.h> | |
48 | #include <asm/xen/hypercall.h> | |
49 | ||
f65c9bb3 LJ |
50 | |
51 | /* | |
52 | * @cpu_id: Xen physical cpu logic number | |
53 | * @flags: Xen physical cpu status flag | |
54 | * - XEN_PCPU_FLAGS_ONLINE: cpu is online | |
55 | * - XEN_PCPU_FLAGS_INVALID: cpu is not present | |
56 | */ | |
57 | struct pcpu { | |
58 | struct list_head list; | |
59 | struct device dev; | |
60 | uint32_t cpu_id; | |
61 | uint32_t flags; | |
62 | }; | |
63 | ||
64 | static struct bus_type xen_pcpu_subsys = { | |
65 | .name = "xen_cpu", | |
66 | .dev_name = "xen_cpu", | |
67 | }; | |
68 | ||
69 | static DEFINE_MUTEX(xen_pcpu_lock); | |
70 | ||
71 | static LIST_HEAD(xen_pcpus); | |
72 | ||
73 | static int xen_pcpu_down(uint32_t cpu_id) | |
74 | { | |
75 | struct xen_platform_op op = { | |
76 | .cmd = XENPF_cpu_offline, | |
77 | .interface_version = XENPF_INTERFACE_VERSION, | |
78 | .u.cpu_ol.cpuid = cpu_id, | |
79 | }; | |
80 | ||
cfafae94 | 81 | return HYPERVISOR_platform_op(&op); |
f65c9bb3 LJ |
82 | } |
83 | ||
84 | static int xen_pcpu_up(uint32_t cpu_id) | |
85 | { | |
86 | struct xen_platform_op op = { | |
87 | .cmd = XENPF_cpu_online, | |
88 | .interface_version = XENPF_INTERFACE_VERSION, | |
89 | .u.cpu_ol.cpuid = cpu_id, | |
90 | }; | |
91 | ||
cfafae94 | 92 | return HYPERVISOR_platform_op(&op); |
f65c9bb3 LJ |
93 | } |
94 | ||
95 | static ssize_t show_online(struct device *dev, | |
96 | struct device_attribute *attr, | |
97 | char *buf) | |
98 | { | |
99 | struct pcpu *cpu = container_of(dev, struct pcpu, dev); | |
100 | ||
101 | return sprintf(buf, "%u\n", !!(cpu->flags & XEN_PCPU_FLAGS_ONLINE)); | |
102 | } | |
103 | ||
104 | static ssize_t __ref store_online(struct device *dev, | |
105 | struct device_attribute *attr, | |
106 | const char *buf, size_t count) | |
107 | { | |
108 | struct pcpu *pcpu = container_of(dev, struct pcpu, dev); | |
109 | unsigned long long val; | |
110 | ssize_t ret; | |
111 | ||
112 | if (!capable(CAP_SYS_ADMIN)) | |
113 | return -EPERM; | |
114 | ||
115 | if (kstrtoull(buf, 0, &val) < 0) | |
116 | return -EINVAL; | |
117 | ||
118 | switch (val) { | |
119 | case 0: | |
120 | ret = xen_pcpu_down(pcpu->cpu_id); | |
121 | break; | |
122 | case 1: | |
123 | ret = xen_pcpu_up(pcpu->cpu_id); | |
124 | break; | |
125 | default: | |
126 | ret = -EINVAL; | |
127 | } | |
128 | ||
129 | if (ret >= 0) | |
130 | ret = count; | |
131 | return ret; | |
132 | } | |
133 | static DEVICE_ATTR(online, S_IRUGO | S_IWUSR, show_online, store_online); | |
134 | ||
4644e5ab TI |
135 | static struct attribute *pcpu_dev_attrs[] = { |
136 | &dev_attr_online.attr, | |
137 | NULL | |
138 | }; | |
139 | ||
140 | static umode_t pcpu_dev_is_visible(struct kobject *kobj, | |
141 | struct attribute *attr, int idx) | |
142 | { | |
143 | struct device *dev = kobj_to_dev(kobj); | |
144 | /* | |
145 | * Xen never offline cpu0 due to several restrictions | |
146 | * and assumptions. This basically doesn't add a sys control | |
147 | * to user, one cannot attempt to offline BSP. | |
148 | */ | |
149 | return dev->id ? attr->mode : 0; | |
150 | } | |
151 | ||
152 | static const struct attribute_group pcpu_dev_group = { | |
153 | .attrs = pcpu_dev_attrs, | |
154 | .is_visible = pcpu_dev_is_visible, | |
155 | }; | |
156 | ||
157 | static const struct attribute_group *pcpu_dev_groups[] = { | |
158 | &pcpu_dev_group, | |
159 | NULL | |
160 | }; | |
161 | ||
f65c9bb3 LJ |
162 | static bool xen_pcpu_online(uint32_t flags) |
163 | { | |
164 | return !!(flags & XEN_PCPU_FLAGS_ONLINE); | |
165 | } | |
166 | ||
167 | static void pcpu_online_status(struct xenpf_pcpuinfo *info, | |
168 | struct pcpu *pcpu) | |
169 | { | |
170 | if (xen_pcpu_online(info->flags) && | |
171 | !xen_pcpu_online(pcpu->flags)) { | |
172 | /* the pcpu is onlined */ | |
173 | pcpu->flags |= XEN_PCPU_FLAGS_ONLINE; | |
174 | kobject_uevent(&pcpu->dev.kobj, KOBJ_ONLINE); | |
175 | } else if (!xen_pcpu_online(info->flags) && | |
176 | xen_pcpu_online(pcpu->flags)) { | |
177 | /* The pcpu is offlined */ | |
178 | pcpu->flags &= ~XEN_PCPU_FLAGS_ONLINE; | |
179 | kobject_uevent(&pcpu->dev.kobj, KOBJ_OFFLINE); | |
180 | } | |
181 | } | |
182 | ||
183 | static struct pcpu *get_pcpu(uint32_t cpu_id) | |
184 | { | |
185 | struct pcpu *pcpu; | |
186 | ||
187 | list_for_each_entry(pcpu, &xen_pcpus, list) { | |
188 | if (pcpu->cpu_id == cpu_id) | |
189 | return pcpu; | |
190 | } | |
191 | ||
192 | return NULL; | |
193 | } | |
194 | ||
195 | static void pcpu_release(struct device *dev) | |
196 | { | |
197 | struct pcpu *pcpu = container_of(dev, struct pcpu, dev); | |
198 | ||
199 | list_del(&pcpu->list); | |
200 | kfree(pcpu); | |
201 | } | |
202 | ||
203 | static void unregister_and_remove_pcpu(struct pcpu *pcpu) | |
204 | { | |
205 | struct device *dev; | |
206 | ||
207 | if (!pcpu) | |
208 | return; | |
209 | ||
210 | dev = &pcpu->dev; | |
f65c9bb3 LJ |
211 | /* pcpu remove would be implicitly done */ |
212 | device_unregister(dev); | |
213 | } | |
214 | ||
215 | static int register_pcpu(struct pcpu *pcpu) | |
216 | { | |
217 | struct device *dev; | |
218 | int err = -EINVAL; | |
219 | ||
220 | if (!pcpu) | |
221 | return err; | |
222 | ||
223 | dev = &pcpu->dev; | |
224 | dev->bus = &xen_pcpu_subsys; | |
225 | dev->id = pcpu->cpu_id; | |
226 | dev->release = pcpu_release; | |
4644e5ab | 227 | dev->groups = pcpu_dev_groups; |
f65c9bb3 LJ |
228 | |
229 | err = device_register(dev); | |
230 | if (err) { | |
231 | pcpu_release(dev); | |
232 | return err; | |
233 | } | |
234 | ||
f65c9bb3 LJ |
235 | return 0; |
236 | } | |
237 | ||
238 | static struct pcpu *create_and_register_pcpu(struct xenpf_pcpuinfo *info) | |
239 | { | |
240 | struct pcpu *pcpu; | |
241 | int err; | |
242 | ||
243 | if (info->flags & XEN_PCPU_FLAGS_INVALID) | |
244 | return ERR_PTR(-ENODEV); | |
245 | ||
246 | pcpu = kzalloc(sizeof(struct pcpu), GFP_KERNEL); | |
247 | if (!pcpu) | |
248 | return ERR_PTR(-ENOMEM); | |
249 | ||
250 | INIT_LIST_HEAD(&pcpu->list); | |
251 | pcpu->cpu_id = info->xen_cpuid; | |
252 | pcpu->flags = info->flags; | |
253 | ||
254 | /* Need hold on xen_pcpu_lock before pcpu list manipulations */ | |
255 | list_add_tail(&pcpu->list, &xen_pcpus); | |
256 | ||
257 | err = register_pcpu(pcpu); | |
258 | if (err) { | |
283c0972 | 259 | pr_warn("Failed to register pcpu%u\n", info->xen_cpuid); |
f65c9bb3 LJ |
260 | return ERR_PTR(-ENOENT); |
261 | } | |
262 | ||
263 | return pcpu; | |
264 | } | |
265 | ||
266 | /* | |
267 | * Caller should hold the xen_pcpu_lock | |
268 | */ | |
269 | static int sync_pcpu(uint32_t cpu, uint32_t *max_cpu) | |
270 | { | |
271 | int ret; | |
272 | struct pcpu *pcpu = NULL; | |
273 | struct xenpf_pcpuinfo *info; | |
274 | struct xen_platform_op op = { | |
275 | .cmd = XENPF_get_cpuinfo, | |
276 | .interface_version = XENPF_INTERFACE_VERSION, | |
277 | .u.pcpu_info.xen_cpuid = cpu, | |
278 | }; | |
279 | ||
cfafae94 | 280 | ret = HYPERVISOR_platform_op(&op); |
f65c9bb3 LJ |
281 | if (ret) |
282 | return ret; | |
283 | ||
284 | info = &op.u.pcpu_info; | |
285 | if (max_cpu) | |
286 | *max_cpu = info->max_present; | |
287 | ||
288 | pcpu = get_pcpu(cpu); | |
289 | ||
290 | /* | |
291 | * Only those at cpu present map has its sys interface. | |
292 | */ | |
293 | if (info->flags & XEN_PCPU_FLAGS_INVALID) { | |
4f8c8527 | 294 | unregister_and_remove_pcpu(pcpu); |
f65c9bb3 LJ |
295 | return 0; |
296 | } | |
297 | ||
298 | if (!pcpu) { | |
299 | pcpu = create_and_register_pcpu(info); | |
300 | if (IS_ERR_OR_NULL(pcpu)) | |
301 | return -ENODEV; | |
302 | } else | |
303 | pcpu_online_status(info, pcpu); | |
304 | ||
305 | return 0; | |
306 | } | |
307 | ||
308 | /* | |
309 | * Sync dom0's pcpu information with xen hypervisor's | |
310 | */ | |
311 | static int xen_sync_pcpus(void) | |
312 | { | |
313 | /* | |
314 | * Boot cpu always have cpu_id 0 in xen | |
315 | */ | |
316 | uint32_t cpu = 0, max_cpu = 0; | |
317 | int err = 0; | |
318 | struct pcpu *pcpu, *tmp; | |
319 | ||
320 | mutex_lock(&xen_pcpu_lock); | |
321 | ||
322 | while (!err && (cpu <= max_cpu)) { | |
323 | err = sync_pcpu(cpu, &max_cpu); | |
324 | cpu++; | |
325 | } | |
326 | ||
327 | if (err) | |
328 | list_for_each_entry_safe(pcpu, tmp, &xen_pcpus, list) | |
329 | unregister_and_remove_pcpu(pcpu); | |
330 | ||
331 | mutex_unlock(&xen_pcpu_lock); | |
332 | ||
333 | return err; | |
334 | } | |
335 | ||
336 | static void xen_pcpu_work_fn(struct work_struct *work) | |
337 | { | |
338 | xen_sync_pcpus(); | |
339 | } | |
340 | static DECLARE_WORK(xen_pcpu_work, xen_pcpu_work_fn); | |
341 | ||
342 | static irqreturn_t xen_pcpu_interrupt(int irq, void *dev_id) | |
343 | { | |
344 | schedule_work(&xen_pcpu_work); | |
345 | return IRQ_HANDLED; | |
346 | } | |
347 | ||
39adc483 LJ |
348 | /* Sync with Xen hypervisor after cpu hotadded */ |
349 | void xen_pcpu_hotplug_sync(void) | |
350 | { | |
351 | schedule_work(&xen_pcpu_work); | |
352 | } | |
353 | EXPORT_SYMBOL_GPL(xen_pcpu_hotplug_sync); | |
354 | ||
355 | /* | |
356 | * For hypervisor presented cpu, return logic cpu id; | |
357 | * For hypervisor non-presented cpu, return -ENODEV. | |
358 | */ | |
359 | int xen_pcpu_id(uint32_t acpi_id) | |
360 | { | |
361 | int cpu_id = 0, max_id = 0; | |
362 | struct xen_platform_op op; | |
363 | ||
364 | op.cmd = XENPF_get_cpuinfo; | |
365 | while (cpu_id <= max_id) { | |
366 | op.u.pcpu_info.xen_cpuid = cpu_id; | |
cfafae94 | 367 | if (HYPERVISOR_platform_op(&op)) { |
39adc483 LJ |
368 | cpu_id++; |
369 | continue; | |
370 | } | |
371 | ||
372 | if (acpi_id == op.u.pcpu_info.acpi_id) | |
373 | return cpu_id; | |
374 | if (op.u.pcpu_info.max_present > max_id) | |
375 | max_id = op.u.pcpu_info.max_present; | |
376 | cpu_id++; | |
377 | } | |
378 | ||
379 | return -ENODEV; | |
380 | } | |
381 | EXPORT_SYMBOL_GPL(xen_pcpu_id); | |
382 | ||
f65c9bb3 LJ |
383 | static int __init xen_pcpu_init(void) |
384 | { | |
385 | int irq, ret; | |
386 | ||
387 | if (!xen_initial_domain()) | |
388 | return -ENODEV; | |
389 | ||
390 | irq = bind_virq_to_irqhandler(VIRQ_PCPU_STATE, 0, | |
391 | xen_pcpu_interrupt, 0, | |
392 | "xen-pcpu", NULL); | |
393 | if (irq < 0) { | |
283c0972 | 394 | pr_warn("Failed to bind pcpu virq\n"); |
f65c9bb3 LJ |
395 | return irq; |
396 | } | |
397 | ||
398 | ret = subsys_system_register(&xen_pcpu_subsys, NULL); | |
399 | if (ret) { | |
283c0972 | 400 | pr_warn("Failed to register pcpu subsys\n"); |
f65c9bb3 LJ |
401 | goto err1; |
402 | } | |
403 | ||
404 | ret = xen_sync_pcpus(); | |
405 | if (ret) { | |
283c0972 | 406 | pr_warn("Failed to sync pcpu info\n"); |
f65c9bb3 LJ |
407 | goto err2; |
408 | } | |
409 | ||
410 | return 0; | |
411 | ||
412 | err2: | |
413 | bus_unregister(&xen_pcpu_subsys); | |
414 | err1: | |
415 | unbind_from_irqhandler(irq, NULL); | |
416 | return ret; | |
417 | } | |
418 | arch_initcall(xen_pcpu_init); |