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