Commit | Line | Data |
---|---|---|
4c071ee5 | 1 | #include <xen/xen.h> |
0ec53ecf | 2 | #include <xen/events.h> |
b3b52fd8 SS |
3 | #include <xen/grant_table.h> |
4 | #include <xen/hvm.h> | |
9a9ab3cc | 5 | #include <xen/interface/vcpu.h> |
4c071ee5 SS |
6 | #include <xen/interface/xen.h> |
7 | #include <xen/interface/memory.h> | |
b3b52fd8 | 8 | #include <xen/interface/hvm/params.h> |
ef61ee0d | 9 | #include <xen/features.h> |
4c071ee5 | 10 | #include <xen/platform_pci.h> |
b3b52fd8 | 11 | #include <xen/xenbus.h> |
c61ba729 | 12 | #include <xen/page.h> |
6abb749e | 13 | #include <xen/interface/sched.h> |
f832da06 | 14 | #include <xen/xen-ops.h> |
4c071ee5 SS |
15 | #include <asm/xen/hypervisor.h> |
16 | #include <asm/xen/hypercall.h> | |
0cac5c30 | 17 | #include <asm/xen/xen-ops.h> |
6abb749e | 18 | #include <asm/system_misc.h> |
0cac5c30 | 19 | #include <asm/efi.h> |
0ec53ecf SS |
20 | #include <linux/interrupt.h> |
21 | #include <linux/irqreturn.h> | |
4c071ee5 | 22 | #include <linux/module.h> |
2e01f166 | 23 | #include <linux/of.h> |
9b08aaa3 | 24 | #include <linux/of_fdt.h> |
2e01f166 SS |
25 | #include <linux/of_irq.h> |
26 | #include <linux/of_address.h> | |
e1a9c16b JG |
27 | #include <linux/cpuidle.h> |
28 | #include <linux/cpufreq.h> | |
8b271d57 | 29 | #include <linux/cpu.h> |
f1dddd11 | 30 | #include <linux/console.h> |
7d5f6f81 SS |
31 | #include <linux/pvclock_gtod.h> |
32 | #include <linux/time64.h> | |
e709fba1 | 33 | #include <linux/timekeeping.h> |
7d5f6f81 | 34 | #include <linux/timekeeper_internal.h> |
d22cbe65 | 35 | #include <linux/acpi.h> |
4c071ee5 | 36 | |
f832da06 IC |
37 | #include <linux/mm.h> |
38 | ||
4c071ee5 SS |
39 | struct start_info _xen_start_info; |
40 | struct start_info *xen_start_info = &_xen_start_info; | |
35c8ab4c | 41 | EXPORT_SYMBOL(xen_start_info); |
4c071ee5 SS |
42 | |
43 | enum xen_domain_type xen_domain_type = XEN_NATIVE; | |
35c8ab4c | 44 | EXPORT_SYMBOL(xen_domain_type); |
4c071ee5 SS |
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); | |
9a9ab3cc | 50 | static struct vcpu_info __percpu *xen_vcpu_info; |
4c071ee5 | 51 | |
88e957d6 | 52 | /* Linux <-> Xen vCPU id mapping */ |
55467dea | 53 | DEFINE_PER_CPU(uint32_t, xen_vcpu_id); |
88e957d6 VK |
54 | EXPORT_PER_CPU_SYMBOL(xen_vcpu_id); |
55 | ||
c61ba729 IC |
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 | ||
81e863c3 | 60 | static __read_mostly unsigned int xen_events_irq; |
0ec53ecf | 61 | |
a13d7201 | 62 | int xen_remap_domain_gfn_array(struct vm_area_struct *vma, |
4c071ee5 | 63 | unsigned long addr, |
a13d7201 | 64 | xen_pfn_t *gfn, int nr, |
4e8c0c8c DV |
65 | int *err_ptr, pgprot_t prot, |
66 | unsigned domid, | |
f832da06 | 67 | struct page **pages) |
4c071ee5 | 68 | { |
a13d7201 | 69 | return xen_xlate_remap_gfn_array(vma, addr, gfn, nr, err_ptr, |
628c28ee | 70 | prot, domid, pages); |
4c071ee5 | 71 | } |
a13d7201 | 72 | EXPORT_SYMBOL_GPL(xen_remap_domain_gfn_array); |
4e8c0c8c DV |
73 | |
74 | /* Not used by XENFEAT_auto_translated guests. */ | |
a13d7201 | 75 | int xen_remap_domain_gfn_range(struct vm_area_struct *vma, |
4e8c0c8c | 76 | unsigned long addr, |
a13d7201 | 77 | xen_pfn_t gfn, int nr, |
4e8c0c8c DV |
78 | pgprot_t prot, unsigned domid, |
79 | struct page **pages) | |
80 | { | |
81 | return -ENOSYS; | |
82 | } | |
a13d7201 | 83 | EXPORT_SYMBOL_GPL(xen_remap_domain_gfn_range); |
2e01f166 | 84 | |
a13d7201 | 85 | int xen_unmap_domain_gfn_range(struct vm_area_struct *vma, |
f832da06 IC |
86 | int nr, struct page **pages) |
87 | { | |
628c28ee | 88 | return xen_xlate_unmap_gfn_range(vma, nr, pages); |
f832da06 | 89 | } |
a13d7201 | 90 | EXPORT_SYMBOL_GPL(xen_unmap_domain_gfn_range); |
f832da06 | 91 | |
e709fba1 SS |
92 | static void xen_read_wallclock(struct timespec64 *ts) |
93 | { | |
94 | u32 version; | |
95 | struct timespec64 now, ts_monotonic; | |
96 | struct shared_info *s = HYPERVISOR_shared_info; | |
97 | struct pvclock_wall_clock *wall_clock = &(s->wc); | |
98 | ||
99 | /* get wallclock at system boot */ | |
100 | do { | |
101 | version = wall_clock->version; | |
102 | rmb(); /* fetch version before time */ | |
103 | now.tv_sec = ((uint64_t)wall_clock->sec_hi << 32) | wall_clock->sec; | |
104 | now.tv_nsec = wall_clock->nsec; | |
105 | rmb(); /* fetch time before checking version */ | |
106 | } while ((wall_clock->version & 1) || (version != wall_clock->version)); | |
107 | ||
108 | /* time since system boot */ | |
109 | ktime_get_ts64(&ts_monotonic); | |
110 | *ts = timespec64_add(now, ts_monotonic); | |
111 | } | |
112 | ||
7d5f6f81 SS |
113 | static int xen_pvclock_gtod_notify(struct notifier_block *nb, |
114 | unsigned long was_set, void *priv) | |
115 | { | |
116 | /* Protected by the calling core code serialization */ | |
117 | static struct timespec64 next_sync; | |
118 | ||
119 | struct xen_platform_op op; | |
120 | struct timespec64 now, system_time; | |
121 | struct timekeeper *tk = priv; | |
122 | ||
123 | now.tv_sec = tk->xtime_sec; | |
124 | now.tv_nsec = (long)(tk->tkr_mono.xtime_nsec >> tk->tkr_mono.shift); | |
125 | system_time = timespec64_add(now, tk->wall_to_monotonic); | |
126 | ||
127 | /* | |
128 | * We only take the expensive HV call when the clock was set | |
129 | * or when the 11 minutes RTC synchronization time elapsed. | |
130 | */ | |
131 | if (!was_set && timespec64_compare(&now, &next_sync) < 0) | |
132 | return NOTIFY_OK; | |
133 | ||
134 | op.cmd = XENPF_settime64; | |
135 | op.u.settime64.mbz = 0; | |
136 | op.u.settime64.secs = now.tv_sec; | |
137 | op.u.settime64.nsecs = now.tv_nsec; | |
138 | op.u.settime64.system_time = timespec64_to_ns(&system_time); | |
139 | (void)HYPERVISOR_platform_op(&op); | |
140 | ||
141 | /* | |
142 | * Move the next drift compensation time 11 minutes | |
143 | * ahead. That's emulating the sync_cmos_clock() update for | |
144 | * the hardware RTC. | |
145 | */ | |
146 | next_sync = now; | |
147 | next_sync.tv_sec += 11 * 60; | |
148 | ||
149 | return NOTIFY_OK; | |
150 | } | |
151 | ||
152 | static struct notifier_block xen_pvclock_gtod_notifier = { | |
153 | .notifier_call = xen_pvclock_gtod_notify, | |
154 | }; | |
155 | ||
4761adb6 | 156 | static int xen_starting_cpu(unsigned int cpu) |
9a9ab3cc SS |
157 | { |
158 | struct vcpu_register_vcpu_info info; | |
159 | struct vcpu_info *vcpup; | |
160 | int err; | |
161 | ||
cb9644bf SS |
162 | /* |
163 | * VCPUOP_register_vcpu_info cannot be called twice for the same | |
164 | * vcpu, so if vcpu_info is already registered, just get out. This | |
165 | * can happen with cpu-hotplug. | |
166 | */ | |
167 | if (per_cpu(xen_vcpu, cpu) != NULL) | |
168 | goto after_register_vcpu_info; | |
169 | ||
9a9ab3cc SS |
170 | pr_info("Xen: initializing cpu%d\n", cpu); |
171 | vcpup = per_cpu_ptr(xen_vcpu_info, cpu); | |
172 | ||
88e957d6 VK |
173 | /* Direct vCPU id mapping for ARM guests. */ |
174 | per_cpu(xen_vcpu_id, cpu) = cpu; | |
175 | ||
250c9af3 JG |
176 | info.mfn = virt_to_gfn(vcpup); |
177 | info.offset = xen_offset_in_page(vcpup); | |
9a9ab3cc | 178 | |
ad5475f9 VK |
179 | err = HYPERVISOR_vcpu_op(VCPUOP_register_vcpu_info, xen_vcpu_nr(cpu), |
180 | &info); | |
d7266d78 SS |
181 | BUG_ON(err); |
182 | per_cpu(xen_vcpu, cpu) = vcpup; | |
183 | ||
34e38523 SS |
184 | xen_setup_runstate_info(cpu); |
185 | ||
cb9644bf | 186 | after_register_vcpu_info: |
3cc8e40e | 187 | enable_percpu_irq(xen_events_irq, 0); |
4761adb6 RC |
188 | return 0; |
189 | } | |
190 | ||
191 | static int xen_dying_cpu(unsigned int cpu) | |
192 | { | |
193 | disable_percpu_irq(xen_events_irq); | |
194 | return 0; | |
9a9ab3cc SS |
195 | } |
196 | ||
2451ade0 | 197 | static void xen_restart(enum reboot_mode reboot_mode, const char *cmd) |
6abb749e SS |
198 | { |
199 | struct sched_shutdown r = { .reason = SHUTDOWN_reboot }; | |
200 | int rc; | |
201 | rc = HYPERVISOR_sched_op(SCHEDOP_shutdown, &r); | |
a91c7775 | 202 | BUG_ON(rc); |
6abb749e SS |
203 | } |
204 | ||
205 | static void xen_power_off(void) | |
206 | { | |
207 | struct sched_shutdown r = { .reason = SHUTDOWN_poweroff }; | |
208 | int rc; | |
209 | rc = HYPERVISOR_sched_op(SCHEDOP_shutdown, &r); | |
a91c7775 | 210 | BUG_ON(rc); |
6abb749e SS |
211 | } |
212 | ||
8b271d57 JG |
213 | static irqreturn_t xen_arm_callback(int irq, void *arg) |
214 | { | |
215 | xen_hvm_evtchn_do_upcall(); | |
216 | return IRQ_HANDLED; | |
217 | } | |
218 | ||
9b08aaa3 SZ |
219 | static __initdata struct { |
220 | const char *compat; | |
221 | const char *prefix; | |
222 | const char *version; | |
223 | bool found; | |
224 | } hyper_node = {"xen,xen", "xen,xen-", NULL, false}; | |
225 | ||
226 | static int __init fdt_find_hyper_node(unsigned long node, const char *uname, | |
227 | int depth, void *data) | |
228 | { | |
229 | const void *s = NULL; | |
230 | int len; | |
231 | ||
232 | if (depth != 1 || strcmp(uname, "hypervisor") != 0) | |
233 | return 0; | |
234 | ||
235 | if (of_flat_dt_is_compatible(node, hyper_node.compat)) | |
236 | hyper_node.found = true; | |
237 | ||
238 | s = of_get_flat_dt_prop(node, "compatible", &len); | |
239 | if (strlen(hyper_node.prefix) + 3 < len && | |
240 | !strncmp(hyper_node.prefix, s, strlen(hyper_node.prefix))) | |
241 | hyper_node.version = s + strlen(hyper_node.prefix); | |
242 | ||
0cac5c30 SZ |
243 | /* |
244 | * Check if Xen supports EFI by checking whether there is the | |
245 | * "/hypervisor/uefi" node in DT. If so, runtime services are available | |
246 | * through proxy functions (e.g. in case of Xen dom0 EFI implementation | |
247 | * they call special hypercall which executes relevant EFI functions) | |
248 | * and that is why they are always enabled. | |
249 | */ | |
250 | if (IS_ENABLED(CONFIG_XEN_EFI)) { | |
251 | if ((of_get_flat_dt_subnode_by_name(node, "uefi") > 0) && | |
252 | !efi_runtime_disabled()) | |
253 | set_bit(EFI_RUNTIME_SERVICES, &efi.flags); | |
254 | } | |
255 | ||
9b08aaa3 SZ |
256 | return 0; |
257 | } | |
258 | ||
2e01f166 SS |
259 | /* |
260 | * see Documentation/devicetree/bindings/arm/xen.txt for the | |
261 | * documentation of the Xen Device Tree format. | |
262 | */ | |
b3b52fd8 | 263 | #define GRANT_TABLE_PHYSADDR 0 |
5882bfef | 264 | void __init xen_early_init(void) |
2e01f166 | 265 | { |
9b08aaa3 SZ |
266 | of_scan_flat_dt(fdt_find_hyper_node, NULL); |
267 | if (!hyper_node.found) { | |
2e01f166 | 268 | pr_debug("No Xen support\n"); |
5882bfef | 269 | return; |
2e01f166 | 270 | } |
9b08aaa3 SZ |
271 | |
272 | if (hyper_node.version == NULL) { | |
2e01f166 | 273 | pr_debug("Xen version not found\n"); |
5882bfef | 274 | return; |
81e863c3 JG |
275 | } |
276 | ||
9b08aaa3 | 277 | pr_info("Xen %s support found\n", hyper_node.version); |
8b271d57 | 278 | |
2e01f166 SS |
279 | xen_domain_type = XEN_HVM_DOMAIN; |
280 | ||
ef61ee0d | 281 | xen_setup_features(); |
5ebc77de | 282 | |
ef61ee0d SS |
283 | if (xen_feature(XENFEAT_dom0)) |
284 | xen_start_info->flags |= SIF_INITDOMAIN|SIF_PRIVILEGED; | |
285 | else | |
286 | xen_start_info->flags &= ~(SIF_INITDOMAIN|SIF_PRIVILEGED); | |
f1dddd11 AB |
287 | |
288 | if (!console_set_on_cmdline && !xen_initial_domain()) | |
289 | add_preferred_console("hvc", 0, NULL); | |
5882bfef SS |
290 | } |
291 | ||
d22cbe65 SZ |
292 | static void __init xen_acpi_guest_init(void) |
293 | { | |
294 | #ifdef CONFIG_ACPI | |
295 | struct xen_hvm_param a; | |
296 | int interrupt, trigger, polarity; | |
297 | ||
298 | a.domid = DOMID_SELF; | |
299 | a.index = HVM_PARAM_CALLBACK_IRQ; | |
300 | ||
301 | if (HYPERVISOR_hvm_op(HVMOP_get_param, &a) | |
302 | || (a.value >> 56) != HVM_PARAM_CALLBACK_TYPE_PPI) { | |
303 | xen_events_irq = 0; | |
304 | return; | |
305 | } | |
306 | ||
307 | interrupt = a.value & 0xff; | |
308 | trigger = ((a.value >> 8) & 0x1) ? ACPI_EDGE_SENSITIVE | |
309 | : ACPI_LEVEL_SENSITIVE; | |
310 | polarity = ((a.value >> 8) & 0x2) ? ACPI_ACTIVE_LOW | |
311 | : ACPI_ACTIVE_HIGH; | |
312 | xen_events_irq = acpi_register_gsi(NULL, interrupt, trigger, polarity); | |
313 | #endif | |
314 | } | |
315 | ||
316 | static void __init xen_dt_guest_init(void) | |
317 | { | |
9b08aaa3 SZ |
318 | struct device_node *xen_node; |
319 | ||
320 | xen_node = of_find_compatible_node(NULL, NULL, "xen,xen"); | |
321 | if (!xen_node) { | |
322 | pr_err("Xen support was detected before, but it has disappeared\n"); | |
323 | return; | |
324 | } | |
325 | ||
d22cbe65 SZ |
326 | xen_events_irq = irq_of_parse_and_map(xen_node, 0); |
327 | } | |
328 | ||
5882bfef SS |
329 | static int __init xen_guest_init(void) |
330 | { | |
331 | struct xen_add_to_physmap xatp; | |
332 | struct shared_info *shared_info_page = NULL; | |
5882bfef SS |
333 | |
334 | if (!xen_domain()) | |
335 | return 0; | |
336 | ||
d22cbe65 SZ |
337 | if (!acpi_disabled) |
338 | xen_acpi_guest_init(); | |
339 | else | |
340 | xen_dt_guest_init(); | |
341 | ||
5882bfef SS |
342 | if (!xen_events_irq) { |
343 | pr_err("Xen event channel interrupt not found\n"); | |
344 | return -ENODEV; | |
345 | } | |
346 | ||
0cac5c30 SZ |
347 | /* |
348 | * The fdt parsing codes have set EFI_RUNTIME_SERVICES if Xen EFI | |
349 | * parameters are found. Force enable runtime services. | |
350 | */ | |
351 | if (efi_enabled(EFI_RUNTIME_SERVICES)) | |
352 | xen_efi_runtime_setup(); | |
353 | ||
5882bfef | 354 | shared_info_page = (struct shared_info *)get_zeroed_page(GFP_KERNEL); |
ef61ee0d | 355 | |
2e01f166 SS |
356 | if (!shared_info_page) { |
357 | pr_err("not enough memory\n"); | |
358 | return -ENOMEM; | |
359 | } | |
360 | xatp.domid = DOMID_SELF; | |
361 | xatp.idx = 0; | |
362 | xatp.space = XENMAPSPACE_shared_info; | |
250c9af3 | 363 | xatp.gpfn = virt_to_gfn(shared_info_page); |
2e01f166 SS |
364 | if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp)) |
365 | BUG(); | |
366 | ||
367 | HYPERVISOR_shared_info = (struct shared_info *)shared_info_page; | |
368 | ||
369 | /* xen_vcpu is a pointer to the vcpu_info struct in the shared_info | |
370 | * page, we use it in the event channel upcall and in some pvclock | |
9a9ab3cc | 371 | * related functions. |
2e01f166 SS |
372 | * The shared info contains exactly 1 CPU (the boot CPU). The guest |
373 | * is required to use VCPUOP_register_vcpu_info to place vcpu info | |
9a9ab3cc SS |
374 | * for secondary CPUs as they are brought up. |
375 | * For uniformity we use VCPUOP_register_vcpu_info even on cpu0. | |
376 | */ | |
377 | xen_vcpu_info = __alloc_percpu(sizeof(struct vcpu_info), | |
378 | sizeof(struct vcpu_info)); | |
379 | if (xen_vcpu_info == NULL) | |
380 | return -ENOMEM; | |
b3b52fd8 | 381 | |
88e957d6 VK |
382 | /* Direct vCPU id mapping for ARM guests. */ |
383 | per_cpu(xen_vcpu_id, 0) = 0; | |
384 | ||
3cf4095d SZ |
385 | xen_auto_xlat_grant_frames.count = gnttab_max_grant_frames(); |
386 | if (xen_xlate_map_ballooned_pages(&xen_auto_xlat_grant_frames.pfn, | |
387 | &xen_auto_xlat_grant_frames.vaddr, | |
388 | xen_auto_xlat_grant_frames.count)) { | |
efaf30a3 KRW |
389 | free_percpu(xen_vcpu_info); |
390 | return -ENOMEM; | |
391 | } | |
b3b52fd8 SS |
392 | gnttab_init(); |
393 | if (!xen_initial_domain()) | |
394 | xenbus_probe(NULL); | |
395 | ||
e1a9c16b JG |
396 | /* |
397 | * Making sure board specific code will not set up ops for | |
398 | * cpu idle and cpu freq. | |
399 | */ | |
400 | disable_cpuidle(); | |
401 | disable_cpufreq(); | |
402 | ||
8b271d57 JG |
403 | xen_init_IRQ(); |
404 | ||
405 | if (request_percpu_irq(xen_events_irq, xen_arm_callback, | |
406 | "events", &xen_vcpu)) { | |
407 | pr_err("Error request IRQ %d\n", xen_events_irq); | |
408 | return -EINVAL; | |
409 | } | |
410 | ||
ecb23dc6 JG |
411 | xen_time_setup_guest(); |
412 | ||
7d5f6f81 SS |
413 | if (xen_initial_domain()) |
414 | pvclock_gtod_register_notifier(&xen_pvclock_gtod_notifier); | |
34e38523 | 415 | |
4761adb6 RC |
416 | return cpuhp_setup_state(CPUHP_AP_ARM_XEN_STARTING, |
417 | "AP_ARM_XEN_STARTING", xen_starting_cpu, | |
418 | xen_dying_cpu); | |
1aa3d8d9 | 419 | } |
8b271d57 | 420 | early_initcall(xen_guest_init); |
1aa3d8d9 SS |
421 | |
422 | static int __init xen_pm_init(void) | |
423 | { | |
9dd4b294 RH |
424 | if (!xen_domain()) |
425 | return -ENODEV; | |
426 | ||
6abb749e SS |
427 | pm_power_off = xen_power_off; |
428 | arm_pm_restart = xen_restart; | |
e709fba1 SS |
429 | if (!xen_initial_domain()) { |
430 | struct timespec64 ts; | |
431 | xen_read_wallclock(&ts); | |
432 | do_settimeofday64(&ts); | |
433 | } | |
6abb749e | 434 | |
2e01f166 SS |
435 | return 0; |
436 | } | |
9dd4b294 | 437 | late_initcall(xen_pm_init); |
0ec53ecf | 438 | |
79390289 SS |
439 | |
440 | /* empty stubs */ | |
441 | void xen_arch_pre_suspend(void) { } | |
442 | void xen_arch_post_suspend(int suspend_cancelled) { } | |
443 | void xen_timer_resume(void) { } | |
444 | void xen_arch_resume(void) { } | |
ffb7dbed | 445 | void xen_arch_suspend(void) { } |
79390289 SS |
446 | |
447 | ||
d5f985c8 | 448 | /* In the hypercall.S file. */ |
911dec0d KRW |
449 | EXPORT_SYMBOL_GPL(HYPERVISOR_event_channel_op); |
450 | EXPORT_SYMBOL_GPL(HYPERVISOR_grant_table_op); | |
ab277bbf SS |
451 | EXPORT_SYMBOL_GPL(HYPERVISOR_xen_version); |
452 | EXPORT_SYMBOL_GPL(HYPERVISOR_console_io); | |
453 | EXPORT_SYMBOL_GPL(HYPERVISOR_sched_op); | |
454 | EXPORT_SYMBOL_GPL(HYPERVISOR_hvm_op); | |
455 | EXPORT_SYMBOL_GPL(HYPERVISOR_memory_op); | |
456 | EXPORT_SYMBOL_GPL(HYPERVISOR_physdev_op); | |
ea0af613 | 457 | EXPORT_SYMBOL_GPL(HYPERVISOR_vcpu_op); |
176455e9 | 458 | EXPORT_SYMBOL_GPL(HYPERVISOR_tmem_op); |
72d39c69 | 459 | EXPORT_SYMBOL_GPL(HYPERVISOR_platform_op); |
9f1d3414 | 460 | EXPORT_SYMBOL_GPL(HYPERVISOR_multicall); |
4b5ae015 | 461 | EXPORT_SYMBOL_GPL(HYPERVISOR_vm_assist); |
911dec0d | 462 | EXPORT_SYMBOL_GPL(privcmd_call); |