xref: /linux/include/asm-generic/mshyperv.h (revision db912b8954c23a55dbc6dc683e0e06ffcb433848)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 
3 /*
4  * Linux-specific definitions for managing interactions with Microsoft's
5  * Hyper-V hypervisor. The definitions in this file are architecture
6  * independent. See arch/<arch>/include/asm/mshyperv.h for definitions
7  * that are specific to architecture <arch>.
8  *
9  * Definitions that are derived from Hyper-V code or headers should not go in
10  * this file, but should instead go in the relevant files in include/hyperv.
11  *
12  * Copyright (C) 2019, Microsoft, Inc.
13  *
14  * Author : Michael Kelley <mikelley@microsoft.com>
15  */
16 
17 #ifndef _ASM_GENERIC_MSHYPERV_H
18 #define _ASM_GENERIC_MSHYPERV_H
19 
20 #include <linux/types.h>
21 #include <linux/atomic.h>
22 #include <linux/bitops.h>
23 #include <acpi/acpi_numa.h>
24 #include <linux/cpumask.h>
25 #include <linux/nmi.h>
26 #include <asm/ptrace.h>
27 #include <hyperv/hvhdk.h>
28 
29 #define VTPM_BASE_ADDRESS 0xfed40000
30 
31 enum hv_partition_type {
32 	HV_PARTITION_TYPE_GUEST,
33 	HV_PARTITION_TYPE_ROOT,
34 };
35 
36 struct ms_hyperv_info {
37 	u32 features;
38 	u32 priv_high;
39 	u32 misc_features;
40 	u32 hints;
41 	u32 nested_features;
42 	u32 max_vp_index;
43 	u32 max_lp_index;
44 	u8 vtl;
45 	union {
46 		u32 isolation_config_a;
47 		struct {
48 			u32 paravisor_present : 1;
49 			u32 reserved_a1 : 31;
50 		};
51 	};
52 	union {
53 		u32 isolation_config_b;
54 		struct {
55 			u32 cvm_type : 4;
56 			u32 reserved_b1 : 1;
57 			u32 shared_gpa_boundary_active : 1;
58 			u32 shared_gpa_boundary_bits : 6;
59 			u32 reserved_b2 : 20;
60 		};
61 	};
62 	u64 shared_gpa_boundary;
63 };
64 extern struct ms_hyperv_info ms_hyperv;
65 extern bool hv_nested;
66 extern u64 hv_current_partition_id;
67 extern enum hv_partition_type hv_curr_partition_type;
68 
69 extern void * __percpu *hyperv_pcpu_input_arg;
70 extern void * __percpu *hyperv_pcpu_output_arg;
71 
72 extern u64 hv_do_hypercall(u64 control, void *inputaddr, void *outputaddr);
73 extern u64 hv_do_fast_hypercall8(u16 control, u64 input8);
74 bool hv_isolation_type_snp(void);
75 bool hv_isolation_type_tdx(void);
76 
77 static inline struct hv_proximity_domain_info hv_numa_node_to_pxm_info(int node)
78 {
79 	struct hv_proximity_domain_info pxm_info = {};
80 
81 	if (node != NUMA_NO_NODE) {
82 		pxm_info.domain_id = node_to_pxm(node);
83 		pxm_info.flags.proximity_info_valid = 1;
84 		pxm_info.flags.proximity_preferred = 1;
85 	}
86 
87 	return pxm_info;
88 }
89 
90 /* Helper functions that provide a consistent pattern for checking Hyper-V hypercall status. */
91 static inline int hv_result(u64 status)
92 {
93 	return status & HV_HYPERCALL_RESULT_MASK;
94 }
95 
96 static inline bool hv_result_success(u64 status)
97 {
98 	return hv_result(status) == HV_STATUS_SUCCESS;
99 }
100 
101 static inline unsigned int hv_repcomp(u64 status)
102 {
103 	/* Bits [43:32] of status have 'Reps completed' data. */
104 	return (status & HV_HYPERCALL_REP_COMP_MASK) >>
105 			 HV_HYPERCALL_REP_COMP_OFFSET;
106 }
107 
108 /*
109  * Rep hypercalls. Callers of this functions are supposed to ensure that
110  * rep_count and varhead_size comply with Hyper-V hypercall definition.
111  */
112 static inline u64 hv_do_rep_hypercall(u16 code, u16 rep_count, u16 varhead_size,
113 				      void *input, void *output)
114 {
115 	u64 control = code;
116 	u64 status;
117 	u16 rep_comp;
118 
119 	control |= (u64)varhead_size << HV_HYPERCALL_VARHEAD_OFFSET;
120 	control |= (u64)rep_count << HV_HYPERCALL_REP_COMP_OFFSET;
121 
122 	do {
123 		status = hv_do_hypercall(control, input, output);
124 		if (!hv_result_success(status))
125 			return status;
126 
127 		rep_comp = hv_repcomp(status);
128 
129 		control &= ~HV_HYPERCALL_REP_START_MASK;
130 		control |= (u64)rep_comp << HV_HYPERCALL_REP_START_OFFSET;
131 
132 		touch_nmi_watchdog();
133 	} while (rep_comp < rep_count);
134 
135 	return status;
136 }
137 
138 /* Generate the guest OS identifier as described in the Hyper-V TLFS */
139 static inline u64 hv_generate_guest_id(u64 kernel_version)
140 {
141 	u64 guest_id;
142 
143 	guest_id = (((u64)HV_LINUX_VENDOR_ID) << 48);
144 	guest_id |= (kernel_version << 16);
145 
146 	return guest_id;
147 }
148 
149 /* Free the message slot and signal end-of-message if required */
150 static inline void vmbus_signal_eom(struct hv_message *msg, u32 old_msg_type)
151 {
152 	/*
153 	 * On crash we're reading some other CPU's message page and we need
154 	 * to be careful: this other CPU may already had cleared the header
155 	 * and the host may already had delivered some other message there.
156 	 * In case we blindly write msg->header.message_type we're going
157 	 * to lose it. We can still lose a message of the same type but
158 	 * we count on the fact that there can only be one
159 	 * CHANNELMSG_UNLOAD_RESPONSE and we don't care about other messages
160 	 * on crash.
161 	 */
162 	if (cmpxchg(&msg->header.message_type, old_msg_type,
163 		    HVMSG_NONE) != old_msg_type)
164 		return;
165 
166 	/*
167 	 * The cmxchg() above does an implicit memory barrier to
168 	 * ensure the write to MessageType (ie set to
169 	 * HVMSG_NONE) happens before we read the
170 	 * MessagePending and EOMing. Otherwise, the EOMing
171 	 * will not deliver any more messages since there is
172 	 * no empty slot
173 	 */
174 	if (msg->header.message_flags.msg_pending) {
175 		/*
176 		 * This will cause message queue rescan to
177 		 * possibly deliver another msg from the
178 		 * hypervisor
179 		 */
180 		hv_set_msr(HV_MSR_EOM, 0);
181 	}
182 }
183 
184 int hv_get_hypervisor_version(union hv_hypervisor_version_info *info);
185 
186 void hv_setup_vmbus_handler(void (*handler)(void));
187 void hv_remove_vmbus_handler(void);
188 void hv_setup_stimer0_handler(void (*handler)(void));
189 void hv_remove_stimer0_handler(void);
190 
191 void hv_setup_kexec_handler(void (*handler)(void));
192 void hv_remove_kexec_handler(void);
193 void hv_setup_crash_handler(void (*handler)(struct pt_regs *regs));
194 void hv_remove_crash_handler(void);
195 
196 extern int vmbus_interrupt;
197 extern int vmbus_irq;
198 
199 #if IS_ENABLED(CONFIG_HYPERV)
200 /*
201  * Hypervisor's notion of virtual processor ID is different from
202  * Linux' notion of CPU ID. This information can only be retrieved
203  * in the context of the calling CPU. Setup a map for easy access
204  * to this information.
205  */
206 extern u32 *hv_vp_index;
207 extern u32 hv_max_vp_index;
208 
209 extern u64 (*hv_read_reference_counter)(void);
210 
211 /* Sentinel value for an uninitialized entry in hv_vp_index array */
212 #define VP_INVAL	U32_MAX
213 
214 int __init hv_common_init(void);
215 void __init hv_get_partition_id(void);
216 void __init hv_common_free(void);
217 void __init ms_hyperv_late_init(void);
218 int hv_common_cpu_init(unsigned int cpu);
219 int hv_common_cpu_die(unsigned int cpu);
220 void hv_identify_partition_type(void);
221 
222 void *hv_alloc_hyperv_page(void);
223 void *hv_alloc_hyperv_zeroed_page(void);
224 void hv_free_hyperv_page(void *addr);
225 
226 int hv_call_deposit_pages(int node, u64 partition_id, u32 num_pages);
227 int hv_call_add_logical_proc(int node, u32 lp_index, u32 acpi_id);
228 int hv_call_create_vp(int node, u64 partition_id, u32 vp_index, u32 flags);
229 
230 /**
231  * hv_cpu_number_to_vp_number() - Map CPU to VP.
232  * @cpu_number: CPU number in Linux terms
233  *
234  * This function returns the mapping between the Linux processor
235  * number and the hypervisor's virtual processor number, useful
236  * in making hypercalls and such that talk about specific
237  * processors.
238  *
239  * Return: Virtual processor number in Hyper-V terms
240  */
241 static inline int hv_cpu_number_to_vp_number(int cpu_number)
242 {
243 	return hv_vp_index[cpu_number];
244 }
245 
246 static inline int __cpumask_to_vpset(struct hv_vpset *vpset,
247 				    const struct cpumask *cpus,
248 				    bool (*func)(int cpu))
249 {
250 	int cpu, vcpu, vcpu_bank, vcpu_offset, nr_bank = 1;
251 	int max_vcpu_bank = hv_max_vp_index / HV_VCPUS_PER_SPARSE_BANK;
252 
253 	/* vpset.valid_bank_mask can represent up to HV_MAX_SPARSE_VCPU_BANKS banks */
254 	if (max_vcpu_bank >= HV_MAX_SPARSE_VCPU_BANKS)
255 		return 0;
256 
257 	/*
258 	 * Clear all banks up to the maximum possible bank as hv_tlb_flush_ex
259 	 * structs are not cleared between calls, we risk flushing unneeded
260 	 * vCPUs otherwise.
261 	 */
262 	for (vcpu_bank = 0; vcpu_bank <= max_vcpu_bank; vcpu_bank++)
263 		vpset->bank_contents[vcpu_bank] = 0;
264 
265 	/*
266 	 * Some banks may end up being empty but this is acceptable.
267 	 */
268 	for_each_cpu(cpu, cpus) {
269 		if (func && func(cpu))
270 			continue;
271 		vcpu = hv_cpu_number_to_vp_number(cpu);
272 		if (vcpu == VP_INVAL)
273 			return -1;
274 		vcpu_bank = vcpu / HV_VCPUS_PER_SPARSE_BANK;
275 		vcpu_offset = vcpu % HV_VCPUS_PER_SPARSE_BANK;
276 		__set_bit(vcpu_offset, (unsigned long *)
277 			  &vpset->bank_contents[vcpu_bank]);
278 		if (vcpu_bank >= nr_bank)
279 			nr_bank = vcpu_bank + 1;
280 	}
281 	vpset->valid_bank_mask = GENMASK_ULL(nr_bank - 1, 0);
282 	return nr_bank;
283 }
284 
285 /*
286  * Convert a Linux cpumask into a Hyper-V VPset. In the _skip variant,
287  * 'func' is called for each CPU present in cpumask.  If 'func' returns
288  * true, that CPU is skipped -- i.e., that CPU from cpumask is *not*
289  * added to the Hyper-V VPset. If 'func' is NULL, no CPUs are
290  * skipped.
291  */
292 static inline int cpumask_to_vpset(struct hv_vpset *vpset,
293 				    const struct cpumask *cpus)
294 {
295 	return __cpumask_to_vpset(vpset, cpus, NULL);
296 }
297 
298 static inline int cpumask_to_vpset_skip(struct hv_vpset *vpset,
299 				    const struct cpumask *cpus,
300 				    bool (*func)(int cpu))
301 {
302 	return __cpumask_to_vpset(vpset, cpus, func);
303 }
304 
305 int hv_result_to_errno(u64 status);
306 void hyperv_report_panic(struct pt_regs *regs, long err, bool in_die);
307 bool hv_is_hyperv_initialized(void);
308 bool hv_is_hibernation_supported(void);
309 enum hv_isolation_type hv_get_isolation_type(void);
310 bool hv_is_isolation_supported(void);
311 bool hv_isolation_type_snp(void);
312 u64 hv_ghcb_hypercall(u64 control, void *input, void *output, u32 input_size);
313 u64 hv_tdx_hypercall(u64 control, u64 param1, u64 param2);
314 void hyperv_cleanup(void);
315 bool hv_query_ext_cap(u64 cap_query);
316 void hv_setup_dma_ops(struct device *dev, bool coherent);
317 #else /* CONFIG_HYPERV */
318 static inline void hv_identify_partition_type(void) {}
319 static inline bool hv_is_hyperv_initialized(void) { return false; }
320 static inline bool hv_is_hibernation_supported(void) { return false; }
321 static inline void hyperv_cleanup(void) {}
322 static inline void ms_hyperv_late_init(void) {}
323 static inline bool hv_is_isolation_supported(void) { return false; }
324 static inline enum hv_isolation_type hv_get_isolation_type(void)
325 {
326 	return HV_ISOLATION_TYPE_NONE;
327 }
328 #endif /* CONFIG_HYPERV */
329 
330 static inline bool hv_root_partition(void)
331 {
332 	return hv_curr_partition_type == HV_PARTITION_TYPE_ROOT;
333 }
334 
335 #endif
336