projects/linux·blob master · refs

425 lines · 9.3 KiB

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
34#define pr_fmt(fmt) "xen_cpu: " fmt
35
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>
43#include <xen/acpi.h>
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
50#ifdef CONFIG_ACPI
51#include <acpi/processor.h>
52#endif
53
54/*
55 * @cpu_id: Xen physical cpu logic number
56 * @flags: Xen physical cpu status flag
57 * - XEN_PCPU_FLAGS_ONLINE: cpu is online
58 * - XEN_PCPU_FLAGS_INVALID: cpu is not present
59 */
60struct pcpu {
61 struct list_head list;
62 struct device dev;
63 uint32_t cpu_id;
64 uint32_t acpi_id;
65 uint32_t flags;
66};
67
68static const struct bus_type xen_pcpu_subsys = {
69 .name = "xen_cpu",
70 .dev_name = "xen_cpu",
71};
72
73static DEFINE_MUTEX(xen_pcpu_lock);
74
75static LIST_HEAD(xen_pcpus);
76
77static int xen_pcpu_down(uint32_t cpu_id)
78{
79 struct xen_platform_op op = {
80 .cmd = XENPF_cpu_offline,
81 .interface_version = XENPF_INTERFACE_VERSION,
82 .u.cpu_ol.cpuid = cpu_id,
83 };
84
85 return HYPERVISOR_platform_op(&op);
86}
87
88static int xen_pcpu_up(uint32_t cpu_id)
89{
90 struct xen_platform_op op = {
91 .cmd = XENPF_cpu_online,
92 .interface_version = XENPF_INTERFACE_VERSION,
93 .u.cpu_ol.cpuid = cpu_id,
94 };
95
96 return HYPERVISOR_platform_op(&op);
97}
98
99static ssize_t online_show(struct device *dev,
100 struct device_attribute *attr,
101 char *buf)
102{
103 struct pcpu *cpu = container_of(dev, struct pcpu, dev);
104
105 return sprintf(buf, "%u\n", !!(cpu->flags & XEN_PCPU_FLAGS_ONLINE));
106}
107
108static ssize_t online_store(struct device *dev,
109 struct device_attribute *attr,
110 const char *buf, size_t count)
111{
112 struct pcpu *pcpu = container_of(dev, struct pcpu, dev);
113 unsigned long long val;
114 ssize_t ret;
115
116 if (!capable(CAP_SYS_ADMIN))
117 return -EPERM;
118
119 if (kstrtoull(buf, 0, &val) < 0)
120 return -EINVAL;
121
122 switch (val) {
123 case 0:
124 ret = xen_pcpu_down(pcpu->cpu_id);
125 break;
126 case 1:
127 ret = xen_pcpu_up(pcpu->cpu_id);
128 break;
129 default:
130 ret = -EINVAL;
131 }
132
133 if (ret >= 0)
134 ret = count;
135 return ret;
136}
137static DEVICE_ATTR_RW(online);
138
139static struct attribute *pcpu_dev_attrs[] = {
140 &dev_attr_online.attr,
141 NULL
142};
143
144static umode_t pcpu_dev_is_visible(struct kobject *kobj,
145 struct attribute *attr, int idx)
146{
147 struct device *dev = kobj_to_dev(kobj);
148 /*
149 * Xen never offline cpu0 due to several restrictions
150 * and assumptions. This basically doesn't add a sys control
151 * to user, one cannot attempt to offline BSP.
152 */
153 return dev->id ? attr->mode : 0;
154}
155
156static const struct attribute_group pcpu_dev_group = {
157 .attrs = pcpu_dev_attrs,
158 .is_visible = pcpu_dev_is_visible,
159};
160
161static const struct attribute_group *pcpu_dev_groups[] = {
162 &pcpu_dev_group,
163 NULL
164};
165
166static bool xen_pcpu_online(uint32_t flags)
167{
168 return !!(flags & XEN_PCPU_FLAGS_ONLINE);
169}
170
171static void pcpu_online_status(struct xenpf_pcpuinfo *info,
172 struct pcpu *pcpu)
173{
174 if (xen_pcpu_online(info->flags) &&
175 !xen_pcpu_online(pcpu->flags)) {
176 /* the pcpu is onlined */
177 pcpu->flags |= XEN_PCPU_FLAGS_ONLINE;
178 kobject_uevent(&pcpu->dev.kobj, KOBJ_ONLINE);
179 } else if (!xen_pcpu_online(info->flags) &&
180 xen_pcpu_online(pcpu->flags)) {
181 /* The pcpu is offlined */
182 pcpu->flags &= ~XEN_PCPU_FLAGS_ONLINE;
183 kobject_uevent(&pcpu->dev.kobj, KOBJ_OFFLINE);
184 }
185}
186
187static struct pcpu *get_pcpu(uint32_t cpu_id)
188{
189 struct pcpu *pcpu;
190
191 list_for_each_entry(pcpu, &xen_pcpus, list) {
192 if (pcpu->cpu_id == cpu_id)
193 return pcpu;
194 }
195
196 return NULL;
197}
198
199static void pcpu_release(struct device *dev)
200{
201 struct pcpu *pcpu = container_of(dev, struct pcpu, dev);
202
203 list_del(&pcpu->list);
204 kfree(pcpu);
205}
206
207static void unregister_and_remove_pcpu(struct pcpu *pcpu)
208{
209 struct device *dev;
210
211 if (!pcpu)
212 return;
213
214 dev = &pcpu->dev;
215 /* pcpu remove would be implicitly done */
216 device_unregister(dev);
217}
218
219static int register_pcpu(struct pcpu *pcpu)
220{
221 struct device *dev;
222 int err = -EINVAL;
223
224 if (!pcpu)
225 return err;
226
227 dev = &pcpu->dev;
228 dev->bus = &xen_pcpu_subsys;
229 dev->id = pcpu->cpu_id;
230 dev->release = pcpu_release;
231 dev->groups = pcpu_dev_groups;
232
233 err = device_register(dev);
234 if (err) {
235 put_device(dev);
236 return err;
237 }
238
239 return 0;
240}
241
242static struct pcpu *create_and_register_pcpu(struct xenpf_pcpuinfo *info)
243{
244 struct pcpu *pcpu;
245 int err;
246
247 if (info->flags & XEN_PCPU_FLAGS_INVALID)
248 return ERR_PTR(-ENODEV);
249
250 pcpu = kzalloc_obj(struct pcpu);
251 if (!pcpu)
252 return ERR_PTR(-ENOMEM);
253
254 INIT_LIST_HEAD(&pcpu->list);
255 pcpu->cpu_id = info->xen_cpuid;
256 pcpu->acpi_id = info->acpi_id;
257 pcpu->flags = info->flags;
258
259 /* Need hold on xen_pcpu_lock before pcpu list manipulations */
260 list_add_tail(&pcpu->list, &xen_pcpus);
261
262 err = register_pcpu(pcpu);
263 if (err) {
264 pr_warn("Failed to register pcpu%u\n", info->xen_cpuid);
265 return ERR_PTR(-ENOENT);
266 }
267
268 return pcpu;
269}
270
271/*
272 * Caller should hold the xen_pcpu_lock
273 */
274static int sync_pcpu(uint32_t cpu, uint32_t *max_cpu)
275{
276 int ret;
277 struct pcpu *pcpu = NULL;
278 struct xenpf_pcpuinfo *info;
279 struct xen_platform_op op = {
280 .cmd = XENPF_get_cpuinfo,
281 .interface_version = XENPF_INTERFACE_VERSION,
282 .u.pcpu_info.xen_cpuid = cpu,
283 };
284
285 ret = HYPERVISOR_platform_op(&op);
286 if (ret)
287 return ret;
288
289 info = &op.u.pcpu_info;
290 if (max_cpu)
291 *max_cpu = info->max_present;
292
293 pcpu = get_pcpu(cpu);
294
295 /*
296 * Only those at cpu present map has its sys interface.
297 */
298 if (info->flags & XEN_PCPU_FLAGS_INVALID) {
299 unregister_and_remove_pcpu(pcpu);
300 return 0;
301 }
302
303 if (!pcpu) {
304 pcpu = create_and_register_pcpu(info);
305 if (IS_ERR_OR_NULL(pcpu))
306 return -ENODEV;
307 } else
308 pcpu_online_status(info, pcpu);
309
310 return 0;
311}
312
313/*
314 * Sync dom0's pcpu information with xen hypervisor's
315 */
316static int xen_sync_pcpus(void)
317{
318 /*
319 * Boot cpu always have cpu_id 0 in xen
320 */
321 uint32_t cpu = 0, max_cpu = 0;
322 int err = 0;
323 struct pcpu *pcpu, *tmp;
324
325 mutex_lock(&xen_pcpu_lock);
326
327 while (!err && (cpu <= max_cpu)) {
328 err = sync_pcpu(cpu, &max_cpu);
329 cpu++;
330 }
331
332 if (err)
333 list_for_each_entry_safe(pcpu, tmp, &xen_pcpus, list)
334 unregister_and_remove_pcpu(pcpu);
335
336 mutex_unlock(&xen_pcpu_lock);
337
338 return err;
339}
340
341static void xen_pcpu_work_fn(struct work_struct *work)
342{
343 xen_sync_pcpus();
344}
345static DECLARE_WORK(xen_pcpu_work, xen_pcpu_work_fn);
346
347static irqreturn_t xen_pcpu_interrupt(int irq, void *dev_id)
348{
349 schedule_work(&xen_pcpu_work);
350 return IRQ_HANDLED;
351}
352
353static int __init xen_pcpu_init(void)
354{
355 int irq, ret;
356
357 if (!xen_initial_domain())
358 return -ENODEV;
359
360 irq = bind_virq_to_irqhandler(VIRQ_PCPU_STATE, 0,
361 xen_pcpu_interrupt, 0,
362 "xen-pcpu", NULL);
363 if (irq < 0) {
364 pr_warn("Failed to bind pcpu virq\n");
365 return irq;
366 }
367
368 ret = subsys_system_register(&xen_pcpu_subsys, NULL);
369 if (ret) {
370 pr_warn("Failed to register pcpu subsys\n");
371 goto err1;
372 }
373
374 ret = xen_sync_pcpus();
375 if (ret) {
376 pr_warn("Failed to sync pcpu info\n");
377 goto err2;
378 }
379
380 return 0;
381
382err2:
383 bus_unregister(&xen_pcpu_subsys);
384err1:
385 unbind_from_irqhandler(irq, NULL);
386 return ret;
387}
388arch_initcall(xen_pcpu_init);
389
390#ifdef CONFIG_ACPI
391bool __init xen_processor_present(uint32_t acpi_id)
392{
393 const struct pcpu *pcpu;
394 bool online = false;
395
396 mutex_lock(&xen_pcpu_lock);
397 list_for_each_entry(pcpu, &xen_pcpus, list)
398 if (pcpu->acpi_id == acpi_id) {
399 online = pcpu->flags & XEN_PCPU_FLAGS_ONLINE;
400 break;
401 }
402 mutex_unlock(&xen_pcpu_lock);
403
404 return online;
405}
406
407void xen_sanitize_proc_cap_bits(uint32_t *cap)
408{
409 struct xen_platform_op op = {
410 .cmd = XENPF_set_processor_pminfo,
411 .u.set_pminfo.id = -1,
412 .u.set_pminfo.type = XEN_PM_PDC,
413 };
414 u32 buf[3] = { ACPI_PDC_REVISION_ID, 1, *cap };
415 int ret;
416
417 set_xen_guest_handle(op.u.set_pminfo.pdc, buf);
418 ret = HYPERVISOR_platform_op(&op);
419 if (ret)
420 pr_err("sanitize of _PDC buffer bits from Xen failed: %d\n",
421 ret);
422 else
423 *cap = buf[2];
424}
425#endif