xref: /kernel/linux/linux-5.10/arch/arm/xen/enlighten.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0-only
2#include <xen/xen.h>
3#include <xen/events.h>
4#include <xen/grant_table.h>
5#include <xen/hvm.h>
6#include <xen/interface/vcpu.h>
7#include <xen/interface/xen.h>
8#include <xen/interface/memory.h>
9#include <xen/interface/hvm/params.h>
10#include <xen/features.h>
11#include <xen/platform_pci.h>
12#include <xen/xenbus.h>
13#include <xen/page.h>
14#include <xen/interface/sched.h>
15#include <xen/xen-ops.h>
16#include <asm/xen/hypervisor.h>
17#include <asm/xen/hypercall.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
39static struct start_info _xen_start_info;
40struct start_info *xen_start_info = &_xen_start_info;
41EXPORT_SYMBOL(xen_start_info);
42
43enum xen_domain_type xen_domain_type = XEN_NATIVE;
44EXPORT_SYMBOL(xen_domain_type);
45
46struct shared_info xen_dummy_shared_info;
47struct shared_info *HYPERVISOR_shared_info = (void *)&xen_dummy_shared_info;
48
49DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
50static struct vcpu_info __percpu *xen_vcpu_info;
51
52/* Linux <-> Xen vCPU id mapping */
53DEFINE_PER_CPU(uint32_t, xen_vcpu_id);
54EXPORT_PER_CPU_SYMBOL(xen_vcpu_id);
55
56/* These are unused until we support booting "pre-ballooned" */
57unsigned long xen_released_pages;
58struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata;
59
60static __read_mostly unsigned int xen_events_irq;
61
62uint32_t xen_start_flags;
63EXPORT_SYMBOL(xen_start_flags);
64
65int xen_unmap_domain_gfn_range(struct vm_area_struct *vma,
66			       int nr, struct page **pages)
67{
68	return xen_xlate_unmap_gfn_range(vma, nr, pages);
69}
70EXPORT_SYMBOL_GPL(xen_unmap_domain_gfn_range);
71
72static void xen_read_wallclock(struct timespec64 *ts)
73{
74	u32 version;
75	struct timespec64 now, ts_monotonic;
76	struct shared_info *s = HYPERVISOR_shared_info;
77	struct pvclock_wall_clock *wall_clock = &(s->wc);
78
79	/* get wallclock at system boot */
80	do {
81		version = wall_clock->version;
82		rmb();		/* fetch version before time */
83		now.tv_sec  = ((uint64_t)wall_clock->sec_hi << 32) | wall_clock->sec;
84		now.tv_nsec = wall_clock->nsec;
85		rmb();		/* fetch time before checking version */
86	} while ((wall_clock->version & 1) || (version != wall_clock->version));
87
88	/* time since system boot */
89	ktime_get_ts64(&ts_monotonic);
90	*ts = timespec64_add(now, ts_monotonic);
91}
92
93static int xen_pvclock_gtod_notify(struct notifier_block *nb,
94				   unsigned long was_set, void *priv)
95{
96	/* Protected by the calling core code serialization */
97	static struct timespec64 next_sync;
98
99	struct xen_platform_op op;
100	struct timespec64 now, system_time;
101	struct timekeeper *tk = priv;
102
103	now.tv_sec = tk->xtime_sec;
104	now.tv_nsec = (long)(tk->tkr_mono.xtime_nsec >> tk->tkr_mono.shift);
105	system_time = timespec64_add(now, tk->wall_to_monotonic);
106
107	/*
108	 * We only take the expensive HV call when the clock was set
109	 * or when the 11 minutes RTC synchronization time elapsed.
110	 */
111	if (!was_set && timespec64_compare(&now, &next_sync) < 0)
112		return NOTIFY_OK;
113
114	op.cmd = XENPF_settime64;
115	op.u.settime64.mbz = 0;
116	op.u.settime64.secs = now.tv_sec;
117	op.u.settime64.nsecs = now.tv_nsec;
118	op.u.settime64.system_time = timespec64_to_ns(&system_time);
119	(void)HYPERVISOR_platform_op(&op);
120
121	/*
122	 * Move the next drift compensation time 11 minutes
123	 * ahead. That's emulating the sync_cmos_clock() update for
124	 * the hardware RTC.
125	 */
126	next_sync = now;
127	next_sync.tv_sec += 11 * 60;
128
129	return NOTIFY_OK;
130}
131
132static struct notifier_block xen_pvclock_gtod_notifier = {
133	.notifier_call = xen_pvclock_gtod_notify,
134};
135
136static int xen_starting_cpu(unsigned int cpu)
137{
138	struct vcpu_register_vcpu_info info;
139	struct vcpu_info *vcpup;
140	int err;
141
142	/*
143	 * VCPUOP_register_vcpu_info cannot be called twice for the same
144	 * vcpu, so if vcpu_info is already registered, just get out. This
145	 * can happen with cpu-hotplug.
146	 */
147	if (per_cpu(xen_vcpu, cpu) != NULL)
148		goto after_register_vcpu_info;
149
150	pr_info("Xen: initializing cpu%d\n", cpu);
151	vcpup = per_cpu_ptr(xen_vcpu_info, cpu);
152
153	info.mfn = percpu_to_gfn(vcpup);
154	info.offset = xen_offset_in_page(vcpup);
155
156	err = HYPERVISOR_vcpu_op(VCPUOP_register_vcpu_info, xen_vcpu_nr(cpu),
157				 &info);
158	BUG_ON(err);
159	per_cpu(xen_vcpu, cpu) = vcpup;
160
161after_register_vcpu_info:
162	enable_percpu_irq(xen_events_irq, 0);
163	return 0;
164}
165
166static int xen_dying_cpu(unsigned int cpu)
167{
168	disable_percpu_irq(xen_events_irq);
169	return 0;
170}
171
172void xen_reboot(int reason)
173{
174	struct sched_shutdown r = { .reason = reason };
175	int rc;
176
177	rc = HYPERVISOR_sched_op(SCHEDOP_shutdown, &r);
178	BUG_ON(rc);
179}
180
181static void xen_restart(enum reboot_mode reboot_mode, const char *cmd)
182{
183	xen_reboot(SHUTDOWN_reboot);
184}
185
186
187static void xen_power_off(void)
188{
189	xen_reboot(SHUTDOWN_poweroff);
190}
191
192static irqreturn_t xen_arm_callback(int irq, void *arg)
193{
194	xen_hvm_evtchn_do_upcall();
195	return IRQ_HANDLED;
196}
197
198static __initdata struct {
199	const char *compat;
200	const char *prefix;
201	const char *version;
202	bool found;
203} hyper_node = {"xen,xen", "xen,xen-", NULL, false};
204
205static int __init fdt_find_hyper_node(unsigned long node, const char *uname,
206				      int depth, void *data)
207{
208	const void *s = NULL;
209	int len;
210
211	if (depth != 1 || strcmp(uname, "hypervisor") != 0)
212		return 0;
213
214	if (of_flat_dt_is_compatible(node, hyper_node.compat))
215		hyper_node.found = true;
216
217	s = of_get_flat_dt_prop(node, "compatible", &len);
218	if (strlen(hyper_node.prefix) + 3  < len &&
219	    !strncmp(hyper_node.prefix, s, strlen(hyper_node.prefix)))
220		hyper_node.version = s + strlen(hyper_node.prefix);
221
222	/*
223	 * Check if Xen supports EFI by checking whether there is the
224	 * "/hypervisor/uefi" node in DT. If so, runtime services are available
225	 * through proxy functions (e.g. in case of Xen dom0 EFI implementation
226	 * they call special hypercall which executes relevant EFI functions)
227	 * and that is why they are always enabled.
228	 */
229	if (IS_ENABLED(CONFIG_XEN_EFI)) {
230		if ((of_get_flat_dt_subnode_by_name(node, "uefi") > 0) &&
231		    !efi_runtime_disabled())
232			set_bit(EFI_RUNTIME_SERVICES, &efi.flags);
233	}
234
235	return 0;
236}
237
238/*
239 * see Documentation/devicetree/bindings/arm/xen.txt for the
240 * documentation of the Xen Device Tree format.
241 */
242void __init xen_early_init(void)
243{
244	of_scan_flat_dt(fdt_find_hyper_node, NULL);
245	if (!hyper_node.found) {
246		pr_debug("No Xen support\n");
247		return;
248	}
249
250	if (hyper_node.version == NULL) {
251		pr_debug("Xen version not found\n");
252		return;
253	}
254
255	pr_info("Xen %s support found\n", hyper_node.version);
256
257	xen_domain_type = XEN_HVM_DOMAIN;
258
259	xen_setup_features();
260
261	if (xen_feature(XENFEAT_dom0))
262		xen_start_flags |= SIF_INITDOMAIN|SIF_PRIVILEGED;
263
264	if (!console_set_on_cmdline && !xen_initial_domain())
265		add_preferred_console("hvc", 0, NULL);
266}
267
268static void __init xen_acpi_guest_init(void)
269{
270#ifdef CONFIG_ACPI
271	struct xen_hvm_param a;
272	int interrupt, trigger, polarity;
273
274	a.domid = DOMID_SELF;
275	a.index = HVM_PARAM_CALLBACK_IRQ;
276
277	if (HYPERVISOR_hvm_op(HVMOP_get_param, &a)
278	    || (a.value >> 56) != HVM_PARAM_CALLBACK_TYPE_PPI) {
279		xen_events_irq = 0;
280		return;
281	}
282
283	interrupt = a.value & 0xff;
284	trigger = ((a.value >> 8) & 0x1) ? ACPI_EDGE_SENSITIVE
285					 : ACPI_LEVEL_SENSITIVE;
286	polarity = ((a.value >> 8) & 0x2) ? ACPI_ACTIVE_LOW
287					  : ACPI_ACTIVE_HIGH;
288	xen_events_irq = acpi_register_gsi(NULL, interrupt, trigger, polarity);
289#endif
290}
291
292static void __init xen_dt_guest_init(void)
293{
294	struct device_node *xen_node;
295
296	xen_node = of_find_compatible_node(NULL, NULL, "xen,xen");
297	if (!xen_node) {
298		pr_err("Xen support was detected before, but it has disappeared\n");
299		return;
300	}
301
302	xen_events_irq = irq_of_parse_and_map(xen_node, 0);
303}
304
305static int __init xen_guest_init(void)
306{
307	struct xen_add_to_physmap xatp;
308	struct shared_info *shared_info_page = NULL;
309	int cpu;
310
311	if (!xen_domain())
312		return 0;
313
314	if (!acpi_disabled)
315		xen_acpi_guest_init();
316	else
317		xen_dt_guest_init();
318
319	if (!xen_events_irq) {
320		pr_err("Xen event channel interrupt not found\n");
321		return -ENODEV;
322	}
323
324	/*
325	 * The fdt parsing codes have set EFI_RUNTIME_SERVICES if Xen EFI
326	 * parameters are found. Force enable runtime services.
327	 */
328	if (efi_enabled(EFI_RUNTIME_SERVICES))
329		xen_efi_runtime_setup();
330
331	shared_info_page = (struct shared_info *)get_zeroed_page(GFP_KERNEL);
332
333	if (!shared_info_page) {
334		pr_err("not enough memory\n");
335		return -ENOMEM;
336	}
337	xatp.domid = DOMID_SELF;
338	xatp.idx = 0;
339	xatp.space = XENMAPSPACE_shared_info;
340	xatp.gpfn = virt_to_gfn(shared_info_page);
341	if (HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp))
342		BUG();
343
344	HYPERVISOR_shared_info = (struct shared_info *)shared_info_page;
345
346	/* xen_vcpu is a pointer to the vcpu_info struct in the shared_info
347	 * page, we use it in the event channel upcall and in some pvclock
348	 * related functions.
349	 * The shared info contains exactly 1 CPU (the boot CPU). The guest
350	 * is required to use VCPUOP_register_vcpu_info to place vcpu info
351	 * for secondary CPUs as they are brought up.
352	 * For uniformity we use VCPUOP_register_vcpu_info even on cpu0.
353	 */
354	xen_vcpu_info = __alloc_percpu(sizeof(struct vcpu_info),
355				       1 << fls(sizeof(struct vcpu_info) - 1));
356	if (xen_vcpu_info == NULL)
357		return -ENOMEM;
358
359	/* Direct vCPU id mapping for ARM guests. */
360	for_each_possible_cpu(cpu)
361		per_cpu(xen_vcpu_id, cpu) = cpu;
362
363	xen_auto_xlat_grant_frames.count = gnttab_max_grant_frames();
364	if (xen_xlate_map_ballooned_pages(&xen_auto_xlat_grant_frames.pfn,
365					  &xen_auto_xlat_grant_frames.vaddr,
366					  xen_auto_xlat_grant_frames.count)) {
367		free_percpu(xen_vcpu_info);
368		return -ENOMEM;
369	}
370	gnttab_init();
371
372	/*
373	 * Making sure board specific code will not set up ops for
374	 * cpu idle and cpu freq.
375	 */
376	disable_cpuidle();
377	disable_cpufreq();
378
379	xen_init_IRQ();
380
381	if (request_percpu_irq(xen_events_irq, xen_arm_callback,
382			       "events", &xen_vcpu)) {
383		pr_err("Error request IRQ %d\n", xen_events_irq);
384		return -EINVAL;
385	}
386
387	if (xen_initial_domain())
388		pvclock_gtod_register_notifier(&xen_pvclock_gtod_notifier);
389
390	return cpuhp_setup_state(CPUHP_AP_ARM_XEN_STARTING,
391				 "arm/xen:starting", xen_starting_cpu,
392				 xen_dying_cpu);
393}
394early_initcall(xen_guest_init);
395
396static int xen_starting_runstate_cpu(unsigned int cpu)
397{
398	xen_setup_runstate_info(cpu);
399	return 0;
400}
401
402static int __init xen_late_init(void)
403{
404	if (!xen_domain())
405		return -ENODEV;
406
407	pm_power_off = xen_power_off;
408	arm_pm_restart = xen_restart;
409	if (!xen_initial_domain()) {
410		struct timespec64 ts;
411		xen_read_wallclock(&ts);
412		do_settimeofday64(&ts);
413	}
414
415	if (xen_kernel_unmapped_at_usr())
416		return 0;
417
418	xen_time_setup_guest();
419
420	return cpuhp_setup_state(CPUHP_AP_ARM_XEN_RUNSTATE_STARTING,
421				 "arm/xen_runstate:starting",
422				 xen_starting_runstate_cpu, NULL);
423}
424late_initcall(xen_late_init);
425
426
427/* empty stubs */
428void xen_arch_pre_suspend(void) { }
429void xen_arch_post_suspend(int suspend_cancelled) { }
430void xen_timer_resume(void) { }
431void xen_arch_resume(void) { }
432void xen_arch_suspend(void) { }
433
434
435/* In the hypercall.S file. */
436EXPORT_SYMBOL_GPL(HYPERVISOR_event_channel_op);
437EXPORT_SYMBOL_GPL(HYPERVISOR_grant_table_op);
438EXPORT_SYMBOL_GPL(HYPERVISOR_xen_version);
439EXPORT_SYMBOL_GPL(HYPERVISOR_console_io);
440EXPORT_SYMBOL_GPL(HYPERVISOR_sched_op);
441EXPORT_SYMBOL_GPL(HYPERVISOR_hvm_op);
442EXPORT_SYMBOL_GPL(HYPERVISOR_memory_op);
443EXPORT_SYMBOL_GPL(HYPERVISOR_physdev_op);
444EXPORT_SYMBOL_GPL(HYPERVISOR_vcpu_op);
445EXPORT_SYMBOL_GPL(HYPERVISOR_tmem_op);
446EXPORT_SYMBOL_GPL(HYPERVISOR_platform_op_raw);
447EXPORT_SYMBOL_GPL(HYPERVISOR_multicall);
448EXPORT_SYMBOL_GPL(HYPERVISOR_vm_assist);
449EXPORT_SYMBOL_GPL(HYPERVISOR_dm_op);
450EXPORT_SYMBOL_GPL(privcmd_call);
451