1 #include <xen/xen.h>
2 #include <xen/events.h>
3 #include <xen/grant_table.h>
4 #include <xen/hvm.h>
5 #include <xen/interface/vcpu.h>
6 #include <xen/interface/xen.h>
7 #include <xen/interface/memory.h>
8 #include <xen/interface/hvm/params.h>
9 #include <xen/features.h>
10 #include <xen/platform_pci.h>
11 #include <xen/xenbus.h>
12 #include <xen/page.h>
13 #include <xen/interface/sched.h>
14 #include <xen/xen-ops.h>
15 #include <asm/xen/hypervisor.h>
16 #include <asm/xen/hypercall.h>
17 #include <asm/xen/xen-ops.h>
18 #include <asm/system_misc.h>
19 #include <asm/efi.h>
20 #include <linux/interrupt.h>
21 #include <linux/irqreturn.h>
22 #include <linux/module.h>
23 #include <linux/of.h>
24 #include <linux/of_fdt.h>
25 #include <linux/of_irq.h>
26 #include <linux/of_address.h>
27 #include <linux/cpuidle.h>
28 #include <linux/cpufreq.h>
29 #include <linux/cpu.h>
30 #include <linux/console.h>
31 #include <linux/pvclock_gtod.h>
32 #include <linux/time64.h>
33 #include <linux/timekeeping.h>
34 #include <linux/timekeeper_internal.h>
35 #include <linux/acpi.h>
36
37 #include <linux/mm.h>
38
39 struct start_info _xen_start_info;
40 struct start_info *xen_start_info = &_xen_start_info;
41 EXPORT_SYMBOL(xen_start_info);
42
43 enum xen_domain_type xen_domain_type = XEN_NATIVE;
44 EXPORT_SYMBOL(xen_domain_type);
45
46 struct shared_info xen_dummy_shared_info;
47 struct shared_info *HYPERVISOR_shared_info = (void *)&xen_dummy_shared_info;
48
49 DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
50 static struct vcpu_info __percpu *xen_vcpu_info;
51
52 /* Linux <-> Xen vCPU id mapping */
53 DEFINE_PER_CPU(uint32_t, xen_vcpu_id);
54 EXPORT_PER_CPU_SYMBOL(xen_vcpu_id);
55
56 /* These are unused until we support booting "pre-ballooned" */
57 unsigned long xen_released_pages;
58 struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata;
59
60 static __read_mostly unsigned int xen_events_irq;
61
62 uint32_t xen_start_flags;
63 EXPORT_SYMBOL(xen_start_flags);
64
xen_remap_domain_gfn_array(struct vm_area_struct * vma,unsigned long addr,xen_pfn_t * gfn,int nr,int * err_ptr,pgprot_t prot,unsigned domid,struct page ** pages)65 int xen_remap_domain_gfn_array(struct vm_area_struct *vma,
66 unsigned long addr,
67 xen_pfn_t *gfn, int nr,
68 int *err_ptr, pgprot_t prot,
69 unsigned domid,
70 struct page **pages)
71 {
72 return xen_xlate_remap_gfn_array(vma, addr, gfn, nr, err_ptr,
73 prot, domid, pages);
74 }
75 EXPORT_SYMBOL_GPL(xen_remap_domain_gfn_array);
76
77 /* Not used by XENFEAT_auto_translated guests. */
xen_remap_domain_gfn_range(struct vm_area_struct * vma,unsigned long addr,xen_pfn_t gfn,int nr,pgprot_t prot,unsigned domid,struct page ** pages)78 int xen_remap_domain_gfn_range(struct vm_area_struct *vma,
79 unsigned long addr,
80 xen_pfn_t gfn, int nr,
81 pgprot_t prot, unsigned domid,
82 struct page **pages)
83 {
84 return -ENOSYS;
85 }
86 EXPORT_SYMBOL_GPL(xen_remap_domain_gfn_range);
87
xen_unmap_domain_gfn_range(struct vm_area_struct * vma,int nr,struct page ** pages)88 int xen_unmap_domain_gfn_range(struct vm_area_struct *vma,
89 int nr, struct page **pages)
90 {
91 return xen_xlate_unmap_gfn_range(vma, nr, pages);
92 }
93 EXPORT_SYMBOL_GPL(xen_unmap_domain_gfn_range);
94
95 /* Not used by XENFEAT_auto_translated guests. */
xen_remap_domain_mfn_array(struct vm_area_struct * vma,unsigned long addr,xen_pfn_t * mfn,int nr,int * err_ptr,pgprot_t prot,unsigned int domid,struct page ** pages)96 int xen_remap_domain_mfn_array(struct vm_area_struct *vma,
97 unsigned long addr,
98 xen_pfn_t *mfn, int nr,
99 int *err_ptr, pgprot_t prot,
100 unsigned int domid, struct page **pages)
101 {
102 return -ENOSYS;
103 }
104 EXPORT_SYMBOL_GPL(xen_remap_domain_mfn_array);
105
xen_read_wallclock(struct timespec64 * ts)106 static void xen_read_wallclock(struct timespec64 *ts)
107 {
108 u32 version;
109 struct timespec64 now, ts_monotonic;
110 struct shared_info *s = HYPERVISOR_shared_info;
111 struct pvclock_wall_clock *wall_clock = &(s->wc);
112
113 /* get wallclock at system boot */
114 do {
115 version = wall_clock->version;
116 rmb(); /* fetch version before time */
117 now.tv_sec = ((uint64_t)wall_clock->sec_hi << 32) | wall_clock->sec;
118 now.tv_nsec = wall_clock->nsec;
119 rmb(); /* fetch time before checking version */
120 } while ((wall_clock->version & 1) || (version != wall_clock->version));
121
122 /* time since system boot */
123 ktime_get_ts64(&ts_monotonic);
124 *ts = timespec64_add(now, ts_monotonic);
125 }
126
xen_pvclock_gtod_notify(struct notifier_block * nb,unsigned long was_set,void * priv)127 static int xen_pvclock_gtod_notify(struct notifier_block *nb,
128 unsigned long was_set, void *priv)
129 {
130 /* Protected by the calling core code serialization */
131 static struct timespec64 next_sync;
132
133 struct xen_platform_op op;
134 struct timespec64 now, system_time;
135 struct timekeeper *tk = priv;
136
137 now.tv_sec = tk->xtime_sec;
138 now.tv_nsec = (long)(tk->tkr_mono.xtime_nsec >> tk->tkr_mono.shift);
139 system_time = timespec64_add(now, tk->wall_to_monotonic);
140
141 /*
142 * We only take the expensive HV call when the clock was set
143 * or when the 11 minutes RTC synchronization time elapsed.
144 */
145 if (!was_set && timespec64_compare(&now, &next_sync) < 0)
146 return NOTIFY_OK;
147
148 op.cmd = XENPF_settime64;
149 op.u.settime64.mbz = 0;
150 op.u.settime64.secs = now.tv_sec;
151 op.u.settime64.nsecs = now.tv_nsec;
152 op.u.settime64.system_time = timespec64_to_ns(&system_time);
153 (void)HYPERVISOR_platform_op(&op);
154
155 /*
156 * Move the next drift compensation time 11 minutes
157 * ahead. That's emulating the sync_cmos_clock() update for
158 * the hardware RTC.
159 */
160 next_sync = now;
161 next_sync.tv_sec += 11 * 60;
162
163 return NOTIFY_OK;
164 }
165
166 static struct notifier_block xen_pvclock_gtod_notifier = {
167 .notifier_call = xen_pvclock_gtod_notify,
168 };
169
xen_starting_cpu(unsigned int cpu)170 static int xen_starting_cpu(unsigned int cpu)
171 {
172 struct vcpu_register_vcpu_info info;
173 struct vcpu_info *vcpup;
174 int err;
175
176 /*
177 * VCPUOP_register_vcpu_info cannot be called twice for the same
178 * vcpu, so if vcpu_info is already registered, just get out. This
179 * can happen with cpu-hotplug.
180 */
181 if (per_cpu(xen_vcpu, cpu) != NULL)
182 goto after_register_vcpu_info;
183
184 pr_info("Xen: initializing cpu%d\n", cpu);
185 vcpup = per_cpu_ptr(xen_vcpu_info, cpu);
186
187 info.mfn = virt_to_gfn(vcpup);
188 info.offset = xen_offset_in_page(vcpup);
189
190 err = HYPERVISOR_vcpu_op(VCPUOP_register_vcpu_info, xen_vcpu_nr(cpu),
191 &info);
192 BUG_ON(err);
193 per_cpu(xen_vcpu, cpu) = vcpup;
194
195 xen_setup_runstate_info(cpu);
196
197 after_register_vcpu_info:
198 enable_percpu_irq(xen_events_irq, 0);
199 return 0;
200 }
201
xen_dying_cpu(unsigned int cpu)202 static int xen_dying_cpu(unsigned int cpu)
203 {
204 disable_percpu_irq(xen_events_irq);
205 return 0;
206 }
207
xen_reboot(int reason)208 void xen_reboot(int reason)
209 {
210 struct sched_shutdown r = { .reason = reason };
211 int rc;
212
213 rc = HYPERVISOR_sched_op(SCHEDOP_shutdown, &r);
214 BUG_ON(rc);
215 }
216
xen_restart(enum reboot_mode reboot_mode,const char * cmd)217 static void xen_restart(enum reboot_mode reboot_mode, const char *cmd)
218 {
219 xen_reboot(SHUTDOWN_reboot);
220 }
221
222
xen_power_off(void)223 static void xen_power_off(void)
224 {
225 xen_reboot(SHUTDOWN_poweroff);
226 }
227
xen_arm_callback(int irq,void * arg)228 static irqreturn_t xen_arm_callback(int irq, void *arg)
229 {
230 xen_hvm_evtchn_do_upcall();
231 return IRQ_HANDLED;
232 }
233
234 static __initdata struct {
235 const char *compat;
236 const char *prefix;
237 const char *version;
238 bool found;
239 } hyper_node = {"xen,xen", "xen,xen-", NULL, false};
240
fdt_find_hyper_node(unsigned long node,const char * uname,int depth,void * data)241 static int __init fdt_find_hyper_node(unsigned long node, const char *uname,
242 int depth, void *data)
243 {
244 const void *s = NULL;
245 int len;
246
247 if (depth != 1 || strcmp(uname, "hypervisor") != 0)
248 return 0;
249
250 if (of_flat_dt_is_compatible(node, hyper_node.compat))
251 hyper_node.found = true;
252
253 s = of_get_flat_dt_prop(node, "compatible", &len);
254 if (strlen(hyper_node.prefix) + 3 < len &&
255 !strncmp(hyper_node.prefix, s, strlen(hyper_node.prefix)))
256 hyper_node.version = s + strlen(hyper_node.prefix);
257
258 /*
259 * Check if Xen supports EFI by checking whether there is the
260 * "/hypervisor/uefi" node in DT. If so, runtime services are available
261 * through proxy functions (e.g. in case of Xen dom0 EFI implementation
262 * they call special hypercall which executes relevant EFI functions)
263 * and that is why they are always enabled.
264 */
265 if (IS_ENABLED(CONFIG_XEN_EFI)) {
266 if ((of_get_flat_dt_subnode_by_name(node, "uefi") > 0) &&
267 !efi_runtime_disabled())
268 set_bit(EFI_RUNTIME_SERVICES, &efi.flags);
269 }
270
271 return 0;
272 }
273
274 /*
275 * see Documentation/devicetree/bindings/arm/xen.txt for the
276 * documentation of the Xen Device Tree format.
277 */
278 #define GRANT_TABLE_PHYSADDR 0
xen_early_init(void)279 void __init xen_early_init(void)
280 {
281 of_scan_flat_dt(fdt_find_hyper_node, NULL);
282 if (!hyper_node.found) {
283 pr_debug("No Xen support\n");
284 return;
285 }
286
287 if (hyper_node.version == NULL) {
288 pr_debug("Xen version not found\n");
289 return;
290 }
291
292 pr_info("Xen %s support found\n", hyper_node.version);
293
294 xen_domain_type = XEN_HVM_DOMAIN;
295
296 xen_setup_features();
297
298 if (xen_feature(XENFEAT_dom0))
299 xen_start_flags |= SIF_INITDOMAIN|SIF_PRIVILEGED;
300
301 if (!console_set_on_cmdline && !xen_initial_domain())
302 add_preferred_console("hvc", 0, NULL);
303 }
304
xen_acpi_guest_init(void)305 static void __init xen_acpi_guest_init(void)
306 {
307 #ifdef CONFIG_ACPI
308 struct xen_hvm_param a;
309 int interrupt, trigger, polarity;
310
311 a.domid = DOMID_SELF;
312 a.index = HVM_PARAM_CALLBACK_IRQ;
313
314 if (HYPERVISOR_hvm_op(HVMOP_get_param, &a)
315 || (a.value >> 56) != HVM_PARAM_CALLBACK_TYPE_PPI) {
316 xen_events_irq = 0;
317 return;
318 }
319
320 interrupt = a.value & 0xff;
321 trigger = ((a.value >> 8) & 0x1) ? ACPI_EDGE_SENSITIVE
322 : ACPI_LEVEL_SENSITIVE;
323 polarity = ((a.value >> 8) & 0x2) ? ACPI_ACTIVE_LOW
324 : ACPI_ACTIVE_HIGH;
325 xen_events_irq = acpi_register_gsi(NULL, interrupt, trigger, polarity);
326 #endif
327 }
328
xen_dt_guest_init(void)329 static void __init xen_dt_guest_init(void)
330 {
331 struct device_node *xen_node;
332
333 xen_node = of_find_compatible_node(NULL, NULL, "xen,xen");
334 if (!xen_node) {
335 pr_err("Xen support was detected before, but it has disappeared\n");
336 return;
337 }
338
339 xen_events_irq = irq_of_parse_and_map(xen_node, 0);
340 }
341
xen_guest_init(void)342 static int __init xen_guest_init(void)
343 {
344 struct xen_add_to_physmap xatp;
345 struct shared_info *shared_info_page = NULL;
346 int cpu;
347
348 if (!xen_domain())
349 return 0;
350
351 if (!acpi_disabled)
352 xen_acpi_guest_init();
353 else
354 xen_dt_guest_init();
355
356 if (!xen_events_irq) {
357 pr_err("Xen event channel interrupt not found\n");
358 return -ENODEV;
359 }
360
361 /*
362 * The fdt parsing codes have set EFI_RUNTIME_SERVICES if Xen EFI
363 * parameters are found. Force enable runtime services.
364 */
365 if (efi_enabled(EFI_RUNTIME_SERVICES))
366 xen_efi_runtime_setup();
367
368 shared_info_page = (struct shared_info *)get_zeroed_page(GFP_KERNEL);
369
370 if (!shared_info_page) {
371 pr_err("not enough memory\n");
372 return -ENOMEM;
373 }
374 xatp.domid = DOMID_SELF;
375 xatp.idx = 0;
376 xatp.space = XENMAPSPACE_shared_info;
377 xatp.gpfn = virt_to_gfn(shared_info_page);
378 if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp))
379 BUG();
380
381 HYPERVISOR_shared_info = (struct shared_info *)shared_info_page;
382
383 /* xen_vcpu is a pointer to the vcpu_info struct in the shared_info
384 * page, we use it in the event channel upcall and in some pvclock
385 * related functions.
386 * The shared info contains exactly 1 CPU (the boot CPU). The guest
387 * is required to use VCPUOP_register_vcpu_info to place vcpu info
388 * for secondary CPUs as they are brought up.
389 * For uniformity we use VCPUOP_register_vcpu_info even on cpu0.
390 */
391 xen_vcpu_info = alloc_percpu(struct vcpu_info);
392 if (xen_vcpu_info == NULL)
393 return -ENOMEM;
394
395 /* Direct vCPU id mapping for ARM guests. */
396 for_each_possible_cpu(cpu)
397 per_cpu(xen_vcpu_id, cpu) = cpu;
398
399 xen_auto_xlat_grant_frames.count = gnttab_max_grant_frames();
400 if (xen_xlate_map_ballooned_pages(&xen_auto_xlat_grant_frames.pfn,
401 &xen_auto_xlat_grant_frames.vaddr,
402 xen_auto_xlat_grant_frames.count)) {
403 free_percpu(xen_vcpu_info);
404 return -ENOMEM;
405 }
406 gnttab_init();
407 if (!xen_initial_domain())
408 xenbus_probe(NULL);
409
410 /*
411 * Making sure board specific code will not set up ops for
412 * cpu idle and cpu freq.
413 */
414 disable_cpuidle();
415 disable_cpufreq();
416
417 xen_init_IRQ();
418
419 if (request_percpu_irq(xen_events_irq, xen_arm_callback,
420 "events", &xen_vcpu)) {
421 pr_err("Error request IRQ %d\n", xen_events_irq);
422 return -EINVAL;
423 }
424
425 xen_time_setup_guest();
426
427 if (xen_initial_domain())
428 pvclock_gtod_register_notifier(&xen_pvclock_gtod_notifier);
429
430 return cpuhp_setup_state(CPUHP_AP_ARM_XEN_STARTING,
431 "arm/xen:starting", xen_starting_cpu,
432 xen_dying_cpu);
433 }
434 early_initcall(xen_guest_init);
435
xen_pm_init(void)436 static int __init xen_pm_init(void)
437 {
438 if (!xen_domain())
439 return -ENODEV;
440
441 pm_power_off = xen_power_off;
442 arm_pm_restart = xen_restart;
443 if (!xen_initial_domain()) {
444 struct timespec64 ts;
445 xen_read_wallclock(&ts);
446 do_settimeofday64(&ts);
447 }
448
449 return 0;
450 }
451 late_initcall(xen_pm_init);
452
453
454 /* empty stubs */
xen_arch_pre_suspend(void)455 void xen_arch_pre_suspend(void) { }
xen_arch_post_suspend(int suspend_cancelled)456 void xen_arch_post_suspend(int suspend_cancelled) { }
xen_timer_resume(void)457 void xen_timer_resume(void) { }
xen_arch_resume(void)458 void xen_arch_resume(void) { }
xen_arch_suspend(void)459 void xen_arch_suspend(void) { }
460
461
462 /* In the hypercall.S file. */
463 EXPORT_SYMBOL_GPL(HYPERVISOR_event_channel_op);
464 EXPORT_SYMBOL_GPL(HYPERVISOR_grant_table_op);
465 EXPORT_SYMBOL_GPL(HYPERVISOR_xen_version);
466 EXPORT_SYMBOL_GPL(HYPERVISOR_console_io);
467 EXPORT_SYMBOL_GPL(HYPERVISOR_sched_op);
468 EXPORT_SYMBOL_GPL(HYPERVISOR_hvm_op);
469 EXPORT_SYMBOL_GPL(HYPERVISOR_memory_op);
470 EXPORT_SYMBOL_GPL(HYPERVISOR_physdev_op);
471 EXPORT_SYMBOL_GPL(HYPERVISOR_vcpu_op);
472 EXPORT_SYMBOL_GPL(HYPERVISOR_tmem_op);
473 EXPORT_SYMBOL_GPL(HYPERVISOR_platform_op);
474 EXPORT_SYMBOL_GPL(HYPERVISOR_multicall);
475 EXPORT_SYMBOL_GPL(HYPERVISOR_vm_assist);
476 EXPORT_SYMBOL_GPL(HYPERVISOR_dm_op);
477 EXPORT_SYMBOL_GPL(privcmd_call);
478