]>
Commit | Line | Data |
---|---|---|
6aa8b732 AK |
1 | /* |
2 | * Kernel-based Virtual Machine driver for Linux | |
3 | * | |
4 | * This module enables machines with Intel VT-x extensions to run virtual | |
5 | * machines without emulation or binary translation. | |
6 | * | |
7 | * Copyright (C) 2006 Qumranet, Inc. | |
221d059d | 8 | * Copyright 2010 Red Hat, Inc. and/or its affilates. |
6aa8b732 AK |
9 | * |
10 | * Authors: | |
11 | * Avi Kivity <avi@qumranet.com> | |
12 | * Yaniv Kamay <yaniv@qumranet.com> | |
13 | * | |
14 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
15 | * the COPYING file in the top-level directory. | |
16 | * | |
17 | */ | |
18 | ||
85f455f7 | 19 | #include "irq.h" |
1d737c8a | 20 | #include "mmu.h" |
e495606d | 21 | |
edf88417 | 22 | #include <linux/kvm_host.h> |
6aa8b732 | 23 | #include <linux/module.h> |
9d8f549d | 24 | #include <linux/kernel.h> |
6aa8b732 AK |
25 | #include <linux/mm.h> |
26 | #include <linux/highmem.h> | |
e8edc6e0 | 27 | #include <linux/sched.h> |
c7addb90 | 28 | #include <linux/moduleparam.h> |
229456fc | 29 | #include <linux/ftrace_event.h> |
5a0e3ad6 | 30 | #include <linux/slab.h> |
cafd6659 | 31 | #include <linux/tboot.h> |
5fdbf976 | 32 | #include "kvm_cache_regs.h" |
35920a35 | 33 | #include "x86.h" |
e495606d | 34 | |
6aa8b732 | 35 | #include <asm/io.h> |
3b3be0d1 | 36 | #include <asm/desc.h> |
13673a90 | 37 | #include <asm/vmx.h> |
6210e37b | 38 | #include <asm/virtext.h> |
a0861c02 | 39 | #include <asm/mce.h> |
2acf923e DC |
40 | #include <asm/i387.h> |
41 | #include <asm/xcr.h> | |
6aa8b732 | 42 | |
229456fc MT |
43 | #include "trace.h" |
44 | ||
4ecac3fd AK |
45 | #define __ex(x) __kvm_handle_fault_on_reboot(x) |
46 | ||
6aa8b732 AK |
47 | MODULE_AUTHOR("Qumranet"); |
48 | MODULE_LICENSE("GPL"); | |
49 | ||
4462d21a | 50 | static int __read_mostly bypass_guest_pf = 1; |
c1f8bc04 | 51 | module_param(bypass_guest_pf, bool, S_IRUGO); |
c7addb90 | 52 | |
4462d21a | 53 | static int __read_mostly enable_vpid = 1; |
736caefe | 54 | module_param_named(vpid, enable_vpid, bool, 0444); |
2384d2b3 | 55 | |
4462d21a | 56 | static int __read_mostly flexpriority_enabled = 1; |
736caefe | 57 | module_param_named(flexpriority, flexpriority_enabled, bool, S_IRUGO); |
4c9fc8ef | 58 | |
4462d21a | 59 | static int __read_mostly enable_ept = 1; |
736caefe | 60 | module_param_named(ept, enable_ept, bool, S_IRUGO); |
d56f546d | 61 | |
3a624e29 NK |
62 | static int __read_mostly enable_unrestricted_guest = 1; |
63 | module_param_named(unrestricted_guest, | |
64 | enable_unrestricted_guest, bool, S_IRUGO); | |
65 | ||
4462d21a | 66 | static int __read_mostly emulate_invalid_guest_state = 0; |
c1f8bc04 | 67 | module_param(emulate_invalid_guest_state, bool, S_IRUGO); |
04fa4d32 | 68 | |
b923e62e DX |
69 | static int __read_mostly vmm_exclusive = 1; |
70 | module_param(vmm_exclusive, bool, S_IRUGO); | |
71 | ||
cdc0e244 AK |
72 | #define KVM_GUEST_CR0_MASK_UNRESTRICTED_GUEST \ |
73 | (X86_CR0_WP | X86_CR0_NE | X86_CR0_NW | X86_CR0_CD) | |
74 | #define KVM_GUEST_CR0_MASK \ | |
75 | (KVM_GUEST_CR0_MASK_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE) | |
76 | #define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST \ | |
81231c69 | 77 | (X86_CR0_WP | X86_CR0_NE) |
cdc0e244 AK |
78 | #define KVM_VM_CR0_ALWAYS_ON \ |
79 | (KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE) | |
4c38609a AK |
80 | #define KVM_CR4_GUEST_OWNED_BITS \ |
81 | (X86_CR4_PVI | X86_CR4_DE | X86_CR4_PCE | X86_CR4_OSFXSR \ | |
82 | | X86_CR4_OSXMMEXCPT) | |
83 | ||
cdc0e244 AK |
84 | #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE) |
85 | #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE) | |
86 | ||
78ac8b47 AK |
87 | #define RMODE_GUEST_OWNED_EFLAGS_BITS (~(X86_EFLAGS_IOPL | X86_EFLAGS_VM)) |
88 | ||
4b8d54f9 ZE |
89 | /* |
90 | * These 2 parameters are used to config the controls for Pause-Loop Exiting: | |
91 | * ple_gap: upper bound on the amount of time between two successive | |
92 | * executions of PAUSE in a loop. Also indicate if ple enabled. | |
93 | * According to test, this time is usually small than 41 cycles. | |
94 | * ple_window: upper bound on the amount of time a guest is allowed to execute | |
95 | * in a PAUSE loop. Tests indicate that most spinlocks are held for | |
96 | * less than 2^12 cycles | |
97 | * Time is measured based on a counter that runs at the same rate as the TSC, | |
98 | * refer SDM volume 3b section 21.6.13 & 22.1.3. | |
99 | */ | |
100 | #define KVM_VMX_DEFAULT_PLE_GAP 41 | |
101 | #define KVM_VMX_DEFAULT_PLE_WINDOW 4096 | |
102 | static int ple_gap = KVM_VMX_DEFAULT_PLE_GAP; | |
103 | module_param(ple_gap, int, S_IRUGO); | |
104 | ||
105 | static int ple_window = KVM_VMX_DEFAULT_PLE_WINDOW; | |
106 | module_param(ple_window, int, S_IRUGO); | |
107 | ||
61d2ef2c AK |
108 | #define NR_AUTOLOAD_MSRS 1 |
109 | ||
a2fa3e9f GH |
110 | struct vmcs { |
111 | u32 revision_id; | |
112 | u32 abort; | |
113 | char data[0]; | |
114 | }; | |
115 | ||
26bb0981 AK |
116 | struct shared_msr_entry { |
117 | unsigned index; | |
118 | u64 data; | |
d5696725 | 119 | u64 mask; |
26bb0981 AK |
120 | }; |
121 | ||
a2fa3e9f | 122 | struct vcpu_vmx { |
fb3f0f51 | 123 | struct kvm_vcpu vcpu; |
543e4243 | 124 | struct list_head local_vcpus_link; |
313dbd49 | 125 | unsigned long host_rsp; |
a2fa3e9f | 126 | int launched; |
29bd8a78 | 127 | u8 fail; |
51aa01d1 | 128 | u32 exit_intr_info; |
1155f76a | 129 | u32 idt_vectoring_info; |
26bb0981 | 130 | struct shared_msr_entry *guest_msrs; |
a2fa3e9f GH |
131 | int nmsrs; |
132 | int save_nmsrs; | |
a2fa3e9f | 133 | #ifdef CONFIG_X86_64 |
44ea2b17 AK |
134 | u64 msr_host_kernel_gs_base; |
135 | u64 msr_guest_kernel_gs_base; | |
a2fa3e9f GH |
136 | #endif |
137 | struct vmcs *vmcs; | |
61d2ef2c AK |
138 | struct msr_autoload { |
139 | unsigned nr; | |
140 | struct vmx_msr_entry guest[NR_AUTOLOAD_MSRS]; | |
141 | struct vmx_msr_entry host[NR_AUTOLOAD_MSRS]; | |
142 | } msr_autoload; | |
a2fa3e9f GH |
143 | struct { |
144 | int loaded; | |
145 | u16 fs_sel, gs_sel, ldt_sel; | |
152d3f2f LV |
146 | int gs_ldt_reload_needed; |
147 | int fs_reload_needed; | |
d77c26fc | 148 | } host_state; |
9c8cba37 | 149 | struct { |
7ffd92c5 | 150 | int vm86_active; |
78ac8b47 | 151 | ulong save_rflags; |
7ffd92c5 AK |
152 | struct kvm_save_segment { |
153 | u16 selector; | |
154 | unsigned long base; | |
155 | u32 limit; | |
156 | u32 ar; | |
157 | } tr, es, ds, fs, gs; | |
9c8cba37 AK |
158 | struct { |
159 | bool pending; | |
160 | u8 vector; | |
161 | unsigned rip; | |
162 | } irq; | |
163 | } rmode; | |
2384d2b3 | 164 | int vpid; |
04fa4d32 | 165 | bool emulation_required; |
3b86cd99 JK |
166 | |
167 | /* Support for vnmi-less CPUs */ | |
168 | int soft_vnmi_blocked; | |
169 | ktime_t entry_time; | |
170 | s64 vnmi_blocked_time; | |
a0861c02 | 171 | u32 exit_reason; |
4e47c7a6 SY |
172 | |
173 | bool rdtscp_enabled; | |
a2fa3e9f GH |
174 | }; |
175 | ||
176 | static inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu) | |
177 | { | |
fb3f0f51 | 178 | return container_of(vcpu, struct vcpu_vmx, vcpu); |
a2fa3e9f GH |
179 | } |
180 | ||
b7ebfb05 | 181 | static int init_rmode(struct kvm *kvm); |
4e1096d2 | 182 | static u64 construct_eptp(unsigned long root_hpa); |
4610c9cc DX |
183 | static void kvm_cpu_vmxon(u64 addr); |
184 | static void kvm_cpu_vmxoff(void); | |
75880a01 | 185 | |
6aa8b732 AK |
186 | static DEFINE_PER_CPU(struct vmcs *, vmxarea); |
187 | static DEFINE_PER_CPU(struct vmcs *, current_vmcs); | |
543e4243 | 188 | static DEFINE_PER_CPU(struct list_head, vcpus_on_cpu); |
3444d7da | 189 | static DEFINE_PER_CPU(struct desc_ptr, host_gdt); |
6aa8b732 | 190 | |
3e7c73e9 AK |
191 | static unsigned long *vmx_io_bitmap_a; |
192 | static unsigned long *vmx_io_bitmap_b; | |
5897297b AK |
193 | static unsigned long *vmx_msr_bitmap_legacy; |
194 | static unsigned long *vmx_msr_bitmap_longmode; | |
fdef3ad1 | 195 | |
2384d2b3 SY |
196 | static DECLARE_BITMAP(vmx_vpid_bitmap, VMX_NR_VPIDS); |
197 | static DEFINE_SPINLOCK(vmx_vpid_lock); | |
198 | ||
1c3d14fe | 199 | static struct vmcs_config { |
6aa8b732 AK |
200 | int size; |
201 | int order; | |
202 | u32 revision_id; | |
1c3d14fe YS |
203 | u32 pin_based_exec_ctrl; |
204 | u32 cpu_based_exec_ctrl; | |
f78e0e2e | 205 | u32 cpu_based_2nd_exec_ctrl; |
1c3d14fe YS |
206 | u32 vmexit_ctrl; |
207 | u32 vmentry_ctrl; | |
208 | } vmcs_config; | |
6aa8b732 | 209 | |
efff9e53 | 210 | static struct vmx_capability { |
d56f546d SY |
211 | u32 ept; |
212 | u32 vpid; | |
213 | } vmx_capability; | |
214 | ||
6aa8b732 AK |
215 | #define VMX_SEGMENT_FIELD(seg) \ |
216 | [VCPU_SREG_##seg] = { \ | |
217 | .selector = GUEST_##seg##_SELECTOR, \ | |
218 | .base = GUEST_##seg##_BASE, \ | |
219 | .limit = GUEST_##seg##_LIMIT, \ | |
220 | .ar_bytes = GUEST_##seg##_AR_BYTES, \ | |
221 | } | |
222 | ||
223 | static struct kvm_vmx_segment_field { | |
224 | unsigned selector; | |
225 | unsigned base; | |
226 | unsigned limit; | |
227 | unsigned ar_bytes; | |
228 | } kvm_vmx_segment_fields[] = { | |
229 | VMX_SEGMENT_FIELD(CS), | |
230 | VMX_SEGMENT_FIELD(DS), | |
231 | VMX_SEGMENT_FIELD(ES), | |
232 | VMX_SEGMENT_FIELD(FS), | |
233 | VMX_SEGMENT_FIELD(GS), | |
234 | VMX_SEGMENT_FIELD(SS), | |
235 | VMX_SEGMENT_FIELD(TR), | |
236 | VMX_SEGMENT_FIELD(LDTR), | |
237 | }; | |
238 | ||
26bb0981 AK |
239 | static u64 host_efer; |
240 | ||
6de4f3ad AK |
241 | static void ept_save_pdptrs(struct kvm_vcpu *vcpu); |
242 | ||
4d56c8a7 | 243 | /* |
8c06585d | 244 | * Keep MSR_STAR at the end, as setup_msrs() will try to optimize it |
4d56c8a7 AK |
245 | * away by decrementing the array size. |
246 | */ | |
6aa8b732 | 247 | static const u32 vmx_msr_index[] = { |
05b3e0c2 | 248 | #ifdef CONFIG_X86_64 |
44ea2b17 | 249 | MSR_SYSCALL_MASK, MSR_LSTAR, MSR_CSTAR, |
6aa8b732 | 250 | #endif |
8c06585d | 251 | MSR_EFER, MSR_TSC_AUX, MSR_STAR, |
6aa8b732 | 252 | }; |
9d8f549d | 253 | #define NR_VMX_MSR ARRAY_SIZE(vmx_msr_index) |
6aa8b732 | 254 | |
31299944 | 255 | static inline bool is_page_fault(u32 intr_info) |
6aa8b732 AK |
256 | { |
257 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
258 | INTR_INFO_VALID_MASK)) == | |
8ab2d2e2 | 259 | (INTR_TYPE_HARD_EXCEPTION | PF_VECTOR | INTR_INFO_VALID_MASK); |
6aa8b732 AK |
260 | } |
261 | ||
31299944 | 262 | static inline bool is_no_device(u32 intr_info) |
2ab455cc AL |
263 | { |
264 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
265 | INTR_INFO_VALID_MASK)) == | |
8ab2d2e2 | 266 | (INTR_TYPE_HARD_EXCEPTION | NM_VECTOR | INTR_INFO_VALID_MASK); |
2ab455cc AL |
267 | } |
268 | ||
31299944 | 269 | static inline bool is_invalid_opcode(u32 intr_info) |
7aa81cc0 AL |
270 | { |
271 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
272 | INTR_INFO_VALID_MASK)) == | |
8ab2d2e2 | 273 | (INTR_TYPE_HARD_EXCEPTION | UD_VECTOR | INTR_INFO_VALID_MASK); |
7aa81cc0 AL |
274 | } |
275 | ||
31299944 | 276 | static inline bool is_external_interrupt(u32 intr_info) |
6aa8b732 AK |
277 | { |
278 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK)) | |
279 | == (INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK); | |
280 | } | |
281 | ||
31299944 | 282 | static inline bool is_machine_check(u32 intr_info) |
a0861c02 AK |
283 | { |
284 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
285 | INTR_INFO_VALID_MASK)) == | |
286 | (INTR_TYPE_HARD_EXCEPTION | MC_VECTOR | INTR_INFO_VALID_MASK); | |
287 | } | |
288 | ||
31299944 | 289 | static inline bool cpu_has_vmx_msr_bitmap(void) |
25c5f225 | 290 | { |
04547156 | 291 | return vmcs_config.cpu_based_exec_ctrl & CPU_BASED_USE_MSR_BITMAPS; |
25c5f225 SY |
292 | } |
293 | ||
31299944 | 294 | static inline bool cpu_has_vmx_tpr_shadow(void) |
6e5d865c | 295 | { |
04547156 | 296 | return vmcs_config.cpu_based_exec_ctrl & CPU_BASED_TPR_SHADOW; |
6e5d865c YS |
297 | } |
298 | ||
31299944 | 299 | static inline bool vm_need_tpr_shadow(struct kvm *kvm) |
6e5d865c | 300 | { |
04547156 | 301 | return (cpu_has_vmx_tpr_shadow()) && (irqchip_in_kernel(kvm)); |
6e5d865c YS |
302 | } |
303 | ||
31299944 | 304 | static inline bool cpu_has_secondary_exec_ctrls(void) |
f78e0e2e | 305 | { |
04547156 SY |
306 | return vmcs_config.cpu_based_exec_ctrl & |
307 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; | |
f78e0e2e SY |
308 | } |
309 | ||
774ead3a | 310 | static inline bool cpu_has_vmx_virtualize_apic_accesses(void) |
f78e0e2e | 311 | { |
04547156 SY |
312 | return vmcs_config.cpu_based_2nd_exec_ctrl & |
313 | SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
314 | } | |
315 | ||
316 | static inline bool cpu_has_vmx_flexpriority(void) | |
317 | { | |
318 | return cpu_has_vmx_tpr_shadow() && | |
319 | cpu_has_vmx_virtualize_apic_accesses(); | |
f78e0e2e SY |
320 | } |
321 | ||
e799794e MT |
322 | static inline bool cpu_has_vmx_ept_execute_only(void) |
323 | { | |
31299944 | 324 | return vmx_capability.ept & VMX_EPT_EXECUTE_ONLY_BIT; |
e799794e MT |
325 | } |
326 | ||
327 | static inline bool cpu_has_vmx_eptp_uncacheable(void) | |
328 | { | |
31299944 | 329 | return vmx_capability.ept & VMX_EPTP_UC_BIT; |
e799794e MT |
330 | } |
331 | ||
332 | static inline bool cpu_has_vmx_eptp_writeback(void) | |
333 | { | |
31299944 | 334 | return vmx_capability.ept & VMX_EPTP_WB_BIT; |
e799794e MT |
335 | } |
336 | ||
337 | static inline bool cpu_has_vmx_ept_2m_page(void) | |
338 | { | |
31299944 | 339 | return vmx_capability.ept & VMX_EPT_2MB_PAGE_BIT; |
e799794e MT |
340 | } |
341 | ||
878403b7 SY |
342 | static inline bool cpu_has_vmx_ept_1g_page(void) |
343 | { | |
31299944 | 344 | return vmx_capability.ept & VMX_EPT_1GB_PAGE_BIT; |
878403b7 SY |
345 | } |
346 | ||
4bc9b982 SY |
347 | static inline bool cpu_has_vmx_ept_4levels(void) |
348 | { | |
349 | return vmx_capability.ept & VMX_EPT_PAGE_WALK_4_BIT; | |
350 | } | |
351 | ||
31299944 | 352 | static inline bool cpu_has_vmx_invept_individual_addr(void) |
d56f546d | 353 | { |
31299944 | 354 | return vmx_capability.ept & VMX_EPT_EXTENT_INDIVIDUAL_BIT; |
d56f546d SY |
355 | } |
356 | ||
31299944 | 357 | static inline bool cpu_has_vmx_invept_context(void) |
d56f546d | 358 | { |
31299944 | 359 | return vmx_capability.ept & VMX_EPT_EXTENT_CONTEXT_BIT; |
d56f546d SY |
360 | } |
361 | ||
31299944 | 362 | static inline bool cpu_has_vmx_invept_global(void) |
d56f546d | 363 | { |
31299944 | 364 | return vmx_capability.ept & VMX_EPT_EXTENT_GLOBAL_BIT; |
d56f546d SY |
365 | } |
366 | ||
518c8aee GJ |
367 | static inline bool cpu_has_vmx_invvpid_single(void) |
368 | { | |
369 | return vmx_capability.vpid & VMX_VPID_EXTENT_SINGLE_CONTEXT_BIT; | |
370 | } | |
371 | ||
b9d762fa GJ |
372 | static inline bool cpu_has_vmx_invvpid_global(void) |
373 | { | |
374 | return vmx_capability.vpid & VMX_VPID_EXTENT_GLOBAL_CONTEXT_BIT; | |
375 | } | |
376 | ||
31299944 | 377 | static inline bool cpu_has_vmx_ept(void) |
d56f546d | 378 | { |
04547156 SY |
379 | return vmcs_config.cpu_based_2nd_exec_ctrl & |
380 | SECONDARY_EXEC_ENABLE_EPT; | |
d56f546d SY |
381 | } |
382 | ||
31299944 | 383 | static inline bool cpu_has_vmx_unrestricted_guest(void) |
3a624e29 NK |
384 | { |
385 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
386 | SECONDARY_EXEC_UNRESTRICTED_GUEST; | |
387 | } | |
388 | ||
31299944 | 389 | static inline bool cpu_has_vmx_ple(void) |
4b8d54f9 ZE |
390 | { |
391 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
392 | SECONDARY_EXEC_PAUSE_LOOP_EXITING; | |
393 | } | |
394 | ||
31299944 | 395 | static inline bool vm_need_virtualize_apic_accesses(struct kvm *kvm) |
f78e0e2e | 396 | { |
6d3e435e | 397 | return flexpriority_enabled && irqchip_in_kernel(kvm); |
f78e0e2e SY |
398 | } |
399 | ||
31299944 | 400 | static inline bool cpu_has_vmx_vpid(void) |
2384d2b3 | 401 | { |
04547156 SY |
402 | return vmcs_config.cpu_based_2nd_exec_ctrl & |
403 | SECONDARY_EXEC_ENABLE_VPID; | |
2384d2b3 SY |
404 | } |
405 | ||
31299944 | 406 | static inline bool cpu_has_vmx_rdtscp(void) |
4e47c7a6 SY |
407 | { |
408 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
409 | SECONDARY_EXEC_RDTSCP; | |
410 | } | |
411 | ||
31299944 | 412 | static inline bool cpu_has_virtual_nmis(void) |
f08864b4 SY |
413 | { |
414 | return vmcs_config.pin_based_exec_ctrl & PIN_BASED_VIRTUAL_NMIS; | |
415 | } | |
416 | ||
f5f48ee1 SY |
417 | static inline bool cpu_has_vmx_wbinvd_exit(void) |
418 | { | |
419 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
420 | SECONDARY_EXEC_WBINVD_EXITING; | |
421 | } | |
422 | ||
04547156 SY |
423 | static inline bool report_flexpriority(void) |
424 | { | |
425 | return flexpriority_enabled; | |
426 | } | |
427 | ||
8b9cf98c | 428 | static int __find_msr_index(struct vcpu_vmx *vmx, u32 msr) |
7725f0ba AK |
429 | { |
430 | int i; | |
431 | ||
a2fa3e9f | 432 | for (i = 0; i < vmx->nmsrs; ++i) |
26bb0981 | 433 | if (vmx_msr_index[vmx->guest_msrs[i].index] == msr) |
a75beee6 ED |
434 | return i; |
435 | return -1; | |
436 | } | |
437 | ||
2384d2b3 SY |
438 | static inline void __invvpid(int ext, u16 vpid, gva_t gva) |
439 | { | |
440 | struct { | |
441 | u64 vpid : 16; | |
442 | u64 rsvd : 48; | |
443 | u64 gva; | |
444 | } operand = { vpid, 0, gva }; | |
445 | ||
4ecac3fd | 446 | asm volatile (__ex(ASM_VMX_INVVPID) |
2384d2b3 SY |
447 | /* CF==1 or ZF==1 --> rc = -1 */ |
448 | "; ja 1f ; ud2 ; 1:" | |
449 | : : "a"(&operand), "c"(ext) : "cc", "memory"); | |
450 | } | |
451 | ||
1439442c SY |
452 | static inline void __invept(int ext, u64 eptp, gpa_t gpa) |
453 | { | |
454 | struct { | |
455 | u64 eptp, gpa; | |
456 | } operand = {eptp, gpa}; | |
457 | ||
4ecac3fd | 458 | asm volatile (__ex(ASM_VMX_INVEPT) |
1439442c SY |
459 | /* CF==1 or ZF==1 --> rc = -1 */ |
460 | "; ja 1f ; ud2 ; 1:\n" | |
461 | : : "a" (&operand), "c" (ext) : "cc", "memory"); | |
462 | } | |
463 | ||
26bb0981 | 464 | static struct shared_msr_entry *find_msr_entry(struct vcpu_vmx *vmx, u32 msr) |
a75beee6 ED |
465 | { |
466 | int i; | |
467 | ||
8b9cf98c | 468 | i = __find_msr_index(vmx, msr); |
a75beee6 | 469 | if (i >= 0) |
a2fa3e9f | 470 | return &vmx->guest_msrs[i]; |
8b6d44c7 | 471 | return NULL; |
7725f0ba AK |
472 | } |
473 | ||
6aa8b732 AK |
474 | static void vmcs_clear(struct vmcs *vmcs) |
475 | { | |
476 | u64 phys_addr = __pa(vmcs); | |
477 | u8 error; | |
478 | ||
4ecac3fd | 479 | asm volatile (__ex(ASM_VMX_VMCLEAR_RAX) "; setna %0" |
6aa8b732 AK |
480 | : "=g"(error) : "a"(&phys_addr), "m"(phys_addr) |
481 | : "cc", "memory"); | |
482 | if (error) | |
483 | printk(KERN_ERR "kvm: vmclear fail: %p/%llx\n", | |
484 | vmcs, phys_addr); | |
485 | } | |
486 | ||
7725b894 DX |
487 | static void vmcs_load(struct vmcs *vmcs) |
488 | { | |
489 | u64 phys_addr = __pa(vmcs); | |
490 | u8 error; | |
491 | ||
492 | asm volatile (__ex(ASM_VMX_VMPTRLD_RAX) "; setna %0" | |
493 | : "=g"(error) : "a"(&phys_addr), "m"(phys_addr) | |
494 | : "cc", "memory"); | |
495 | if (error) | |
496 | printk(KERN_ERR "kvm: vmptrld %p/%llx fail\n", | |
497 | vmcs, phys_addr); | |
498 | } | |
499 | ||
6aa8b732 AK |
500 | static void __vcpu_clear(void *arg) |
501 | { | |
8b9cf98c | 502 | struct vcpu_vmx *vmx = arg; |
d3b2c338 | 503 | int cpu = raw_smp_processor_id(); |
6aa8b732 | 504 | |
8b9cf98c | 505 | if (vmx->vcpu.cpu == cpu) |
a2fa3e9f GH |
506 | vmcs_clear(vmx->vmcs); |
507 | if (per_cpu(current_vmcs, cpu) == vmx->vmcs) | |
6aa8b732 | 508 | per_cpu(current_vmcs, cpu) = NULL; |
543e4243 AK |
509 | list_del(&vmx->local_vcpus_link); |
510 | vmx->vcpu.cpu = -1; | |
511 | vmx->launched = 0; | |
6aa8b732 AK |
512 | } |
513 | ||
8b9cf98c | 514 | static void vcpu_clear(struct vcpu_vmx *vmx) |
8d0be2b3 | 515 | { |
eae5ecb5 AK |
516 | if (vmx->vcpu.cpu == -1) |
517 | return; | |
8691e5a8 | 518 | smp_call_function_single(vmx->vcpu.cpu, __vcpu_clear, vmx, 1); |
8d0be2b3 AK |
519 | } |
520 | ||
1760dd49 | 521 | static inline void vpid_sync_vcpu_single(struct vcpu_vmx *vmx) |
2384d2b3 SY |
522 | { |
523 | if (vmx->vpid == 0) | |
524 | return; | |
525 | ||
518c8aee GJ |
526 | if (cpu_has_vmx_invvpid_single()) |
527 | __invvpid(VMX_VPID_EXTENT_SINGLE_CONTEXT, vmx->vpid, 0); | |
2384d2b3 SY |
528 | } |
529 | ||
b9d762fa GJ |
530 | static inline void vpid_sync_vcpu_global(void) |
531 | { | |
532 | if (cpu_has_vmx_invvpid_global()) | |
533 | __invvpid(VMX_VPID_EXTENT_ALL_CONTEXT, 0, 0); | |
534 | } | |
535 | ||
536 | static inline void vpid_sync_context(struct vcpu_vmx *vmx) | |
537 | { | |
538 | if (cpu_has_vmx_invvpid_single()) | |
1760dd49 | 539 | vpid_sync_vcpu_single(vmx); |
b9d762fa GJ |
540 | else |
541 | vpid_sync_vcpu_global(); | |
542 | } | |
543 | ||
1439442c SY |
544 | static inline void ept_sync_global(void) |
545 | { | |
546 | if (cpu_has_vmx_invept_global()) | |
547 | __invept(VMX_EPT_EXTENT_GLOBAL, 0, 0); | |
548 | } | |
549 | ||
550 | static inline void ept_sync_context(u64 eptp) | |
551 | { | |
089d034e | 552 | if (enable_ept) { |
1439442c SY |
553 | if (cpu_has_vmx_invept_context()) |
554 | __invept(VMX_EPT_EXTENT_CONTEXT, eptp, 0); | |
555 | else | |
556 | ept_sync_global(); | |
557 | } | |
558 | } | |
559 | ||
560 | static inline void ept_sync_individual_addr(u64 eptp, gpa_t gpa) | |
561 | { | |
089d034e | 562 | if (enable_ept) { |
1439442c SY |
563 | if (cpu_has_vmx_invept_individual_addr()) |
564 | __invept(VMX_EPT_EXTENT_INDIVIDUAL_ADDR, | |
565 | eptp, gpa); | |
566 | else | |
567 | ept_sync_context(eptp); | |
568 | } | |
569 | } | |
570 | ||
6aa8b732 AK |
571 | static unsigned long vmcs_readl(unsigned long field) |
572 | { | |
573 | unsigned long value; | |
574 | ||
4ecac3fd | 575 | asm volatile (__ex(ASM_VMX_VMREAD_RDX_RAX) |
6aa8b732 AK |
576 | : "=a"(value) : "d"(field) : "cc"); |
577 | return value; | |
578 | } | |
579 | ||
580 | static u16 vmcs_read16(unsigned long field) | |
581 | { | |
582 | return vmcs_readl(field); | |
583 | } | |
584 | ||
585 | static u32 vmcs_read32(unsigned long field) | |
586 | { | |
587 | return vmcs_readl(field); | |
588 | } | |
589 | ||
590 | static u64 vmcs_read64(unsigned long field) | |
591 | { | |
05b3e0c2 | 592 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
593 | return vmcs_readl(field); |
594 | #else | |
595 | return vmcs_readl(field) | ((u64)vmcs_readl(field+1) << 32); | |
596 | #endif | |
597 | } | |
598 | ||
e52de1b8 AK |
599 | static noinline void vmwrite_error(unsigned long field, unsigned long value) |
600 | { | |
601 | printk(KERN_ERR "vmwrite error: reg %lx value %lx (err %d)\n", | |
602 | field, value, vmcs_read32(VM_INSTRUCTION_ERROR)); | |
603 | dump_stack(); | |
604 | } | |
605 | ||
6aa8b732 AK |
606 | static void vmcs_writel(unsigned long field, unsigned long value) |
607 | { | |
608 | u8 error; | |
609 | ||
4ecac3fd | 610 | asm volatile (__ex(ASM_VMX_VMWRITE_RAX_RDX) "; setna %0" |
d77c26fc | 611 | : "=q"(error) : "a"(value), "d"(field) : "cc"); |
e52de1b8 AK |
612 | if (unlikely(error)) |
613 | vmwrite_error(field, value); | |
6aa8b732 AK |
614 | } |
615 | ||
616 | static void vmcs_write16(unsigned long field, u16 value) | |
617 | { | |
618 | vmcs_writel(field, value); | |
619 | } | |
620 | ||
621 | static void vmcs_write32(unsigned long field, u32 value) | |
622 | { | |
623 | vmcs_writel(field, value); | |
624 | } | |
625 | ||
626 | static void vmcs_write64(unsigned long field, u64 value) | |
627 | { | |
6aa8b732 | 628 | vmcs_writel(field, value); |
7682f2d0 | 629 | #ifndef CONFIG_X86_64 |
6aa8b732 AK |
630 | asm volatile (""); |
631 | vmcs_writel(field+1, value >> 32); | |
632 | #endif | |
633 | } | |
634 | ||
2ab455cc AL |
635 | static void vmcs_clear_bits(unsigned long field, u32 mask) |
636 | { | |
637 | vmcs_writel(field, vmcs_readl(field) & ~mask); | |
638 | } | |
639 | ||
640 | static void vmcs_set_bits(unsigned long field, u32 mask) | |
641 | { | |
642 | vmcs_writel(field, vmcs_readl(field) | mask); | |
643 | } | |
644 | ||
abd3f2d6 AK |
645 | static void update_exception_bitmap(struct kvm_vcpu *vcpu) |
646 | { | |
647 | u32 eb; | |
648 | ||
fd7373cc JK |
649 | eb = (1u << PF_VECTOR) | (1u << UD_VECTOR) | (1u << MC_VECTOR) | |
650 | (1u << NM_VECTOR) | (1u << DB_VECTOR); | |
651 | if ((vcpu->guest_debug & | |
652 | (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) == | |
653 | (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) | |
654 | eb |= 1u << BP_VECTOR; | |
7ffd92c5 | 655 | if (to_vmx(vcpu)->rmode.vm86_active) |
abd3f2d6 | 656 | eb = ~0; |
089d034e | 657 | if (enable_ept) |
1439442c | 658 | eb &= ~(1u << PF_VECTOR); /* bypass_guest_pf = 0 */ |
02daab21 AK |
659 | if (vcpu->fpu_active) |
660 | eb &= ~(1u << NM_VECTOR); | |
abd3f2d6 AK |
661 | vmcs_write32(EXCEPTION_BITMAP, eb); |
662 | } | |
663 | ||
61d2ef2c AK |
664 | static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr) |
665 | { | |
666 | unsigned i; | |
667 | struct msr_autoload *m = &vmx->msr_autoload; | |
668 | ||
669 | for (i = 0; i < m->nr; ++i) | |
670 | if (m->guest[i].index == msr) | |
671 | break; | |
672 | ||
673 | if (i == m->nr) | |
674 | return; | |
675 | --m->nr; | |
676 | m->guest[i] = m->guest[m->nr]; | |
677 | m->host[i] = m->host[m->nr]; | |
678 | vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->nr); | |
679 | vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr); | |
680 | } | |
681 | ||
682 | static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr, | |
683 | u64 guest_val, u64 host_val) | |
684 | { | |
685 | unsigned i; | |
686 | struct msr_autoload *m = &vmx->msr_autoload; | |
687 | ||
688 | for (i = 0; i < m->nr; ++i) | |
689 | if (m->guest[i].index == msr) | |
690 | break; | |
691 | ||
692 | if (i == m->nr) { | |
693 | ++m->nr; | |
694 | vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->nr); | |
695 | vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr); | |
696 | } | |
697 | ||
698 | m->guest[i].index = msr; | |
699 | m->guest[i].value = guest_val; | |
700 | m->host[i].index = msr; | |
701 | m->host[i].value = host_val; | |
702 | } | |
703 | ||
33ed6329 AK |
704 | static void reload_tss(void) |
705 | { | |
33ed6329 AK |
706 | /* |
707 | * VT restores TR but not its size. Useless. | |
708 | */ | |
d359192f | 709 | struct desc_ptr *gdt = &__get_cpu_var(host_gdt); |
a5f61300 | 710 | struct desc_struct *descs; |
33ed6329 | 711 | |
d359192f | 712 | descs = (void *)gdt->address; |
33ed6329 AK |
713 | descs[GDT_ENTRY_TSS].type = 9; /* available TSS */ |
714 | load_TR_desc(); | |
33ed6329 AK |
715 | } |
716 | ||
92c0d900 | 717 | static bool update_transition_efer(struct vcpu_vmx *vmx, int efer_offset) |
2cc51560 | 718 | { |
3a34a881 | 719 | u64 guest_efer; |
51c6cf66 AK |
720 | u64 ignore_bits; |
721 | ||
f6801dff | 722 | guest_efer = vmx->vcpu.arch.efer; |
3a34a881 | 723 | |
51c6cf66 AK |
724 | /* |
725 | * NX is emulated; LMA and LME handled by hardware; SCE meaninless | |
726 | * outside long mode | |
727 | */ | |
728 | ignore_bits = EFER_NX | EFER_SCE; | |
729 | #ifdef CONFIG_X86_64 | |
730 | ignore_bits |= EFER_LMA | EFER_LME; | |
731 | /* SCE is meaningful only in long mode on Intel */ | |
732 | if (guest_efer & EFER_LMA) | |
733 | ignore_bits &= ~(u64)EFER_SCE; | |
734 | #endif | |
51c6cf66 AK |
735 | guest_efer &= ~ignore_bits; |
736 | guest_efer |= host_efer & ignore_bits; | |
26bb0981 | 737 | vmx->guest_msrs[efer_offset].data = guest_efer; |
d5696725 | 738 | vmx->guest_msrs[efer_offset].mask = ~ignore_bits; |
84ad33ef AK |
739 | |
740 | clear_atomic_switch_msr(vmx, MSR_EFER); | |
741 | /* On ept, can't emulate nx, and must switch nx atomically */ | |
742 | if (enable_ept && ((vmx->vcpu.arch.efer ^ host_efer) & EFER_NX)) { | |
743 | guest_efer = vmx->vcpu.arch.efer; | |
744 | if (!(guest_efer & EFER_LMA)) | |
745 | guest_efer &= ~EFER_LME; | |
746 | add_atomic_switch_msr(vmx, MSR_EFER, guest_efer, host_efer); | |
747 | return false; | |
748 | } | |
749 | ||
26bb0981 | 750 | return true; |
51c6cf66 AK |
751 | } |
752 | ||
2d49ec72 GN |
753 | static unsigned long segment_base(u16 selector) |
754 | { | |
d359192f | 755 | struct desc_ptr *gdt = &__get_cpu_var(host_gdt); |
2d49ec72 GN |
756 | struct desc_struct *d; |
757 | unsigned long table_base; | |
758 | unsigned long v; | |
759 | ||
760 | if (!(selector & ~3)) | |
761 | return 0; | |
762 | ||
d359192f | 763 | table_base = gdt->address; |
2d49ec72 GN |
764 | |
765 | if (selector & 4) { /* from ldt */ | |
766 | u16 ldt_selector = kvm_read_ldt(); | |
767 | ||
768 | if (!(ldt_selector & ~3)) | |
769 | return 0; | |
770 | ||
771 | table_base = segment_base(ldt_selector); | |
772 | } | |
773 | d = (struct desc_struct *)(table_base + (selector & ~7)); | |
774 | v = get_desc_base(d); | |
775 | #ifdef CONFIG_X86_64 | |
776 | if (d->s == 0 && (d->type == 2 || d->type == 9 || d->type == 11)) | |
777 | v |= ((unsigned long)((struct ldttss_desc64 *)d)->base3) << 32; | |
778 | #endif | |
779 | return v; | |
780 | } | |
781 | ||
782 | static inline unsigned long kvm_read_tr_base(void) | |
783 | { | |
784 | u16 tr; | |
785 | asm("str %0" : "=g"(tr)); | |
786 | return segment_base(tr); | |
787 | } | |
788 | ||
04d2cc77 | 789 | static void vmx_save_host_state(struct kvm_vcpu *vcpu) |
33ed6329 | 790 | { |
04d2cc77 | 791 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
26bb0981 | 792 | int i; |
04d2cc77 | 793 | |
a2fa3e9f | 794 | if (vmx->host_state.loaded) |
33ed6329 AK |
795 | return; |
796 | ||
a2fa3e9f | 797 | vmx->host_state.loaded = 1; |
33ed6329 AK |
798 | /* |
799 | * Set host fs and gs selectors. Unfortunately, 22.2.3 does not | |
800 | * allow segment selectors with cpl > 0 or ti == 1. | |
801 | */ | |
d6e88aec | 802 | vmx->host_state.ldt_sel = kvm_read_ldt(); |
152d3f2f | 803 | vmx->host_state.gs_ldt_reload_needed = vmx->host_state.ldt_sel; |
9581d442 | 804 | savesegment(fs, vmx->host_state.fs_sel); |
152d3f2f | 805 | if (!(vmx->host_state.fs_sel & 7)) { |
a2fa3e9f | 806 | vmcs_write16(HOST_FS_SELECTOR, vmx->host_state.fs_sel); |
152d3f2f LV |
807 | vmx->host_state.fs_reload_needed = 0; |
808 | } else { | |
33ed6329 | 809 | vmcs_write16(HOST_FS_SELECTOR, 0); |
152d3f2f | 810 | vmx->host_state.fs_reload_needed = 1; |
33ed6329 | 811 | } |
9581d442 | 812 | savesegment(gs, vmx->host_state.gs_sel); |
a2fa3e9f GH |
813 | if (!(vmx->host_state.gs_sel & 7)) |
814 | vmcs_write16(HOST_GS_SELECTOR, vmx->host_state.gs_sel); | |
33ed6329 AK |
815 | else { |
816 | vmcs_write16(HOST_GS_SELECTOR, 0); | |
152d3f2f | 817 | vmx->host_state.gs_ldt_reload_needed = 1; |
33ed6329 AK |
818 | } |
819 | ||
820 | #ifdef CONFIG_X86_64 | |
821 | vmcs_writel(HOST_FS_BASE, read_msr(MSR_FS_BASE)); | |
822 | vmcs_writel(HOST_GS_BASE, read_msr(MSR_GS_BASE)); | |
823 | #else | |
a2fa3e9f GH |
824 | vmcs_writel(HOST_FS_BASE, segment_base(vmx->host_state.fs_sel)); |
825 | vmcs_writel(HOST_GS_BASE, segment_base(vmx->host_state.gs_sel)); | |
33ed6329 | 826 | #endif |
707c0874 AK |
827 | |
828 | #ifdef CONFIG_X86_64 | |
44ea2b17 AK |
829 | if (is_long_mode(&vmx->vcpu)) { |
830 | rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); | |
831 | wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); | |
832 | } | |
707c0874 | 833 | #endif |
26bb0981 AK |
834 | for (i = 0; i < vmx->save_nmsrs; ++i) |
835 | kvm_set_shared_msr(vmx->guest_msrs[i].index, | |
d5696725 AK |
836 | vmx->guest_msrs[i].data, |
837 | vmx->guest_msrs[i].mask); | |
33ed6329 AK |
838 | } |
839 | ||
a9b21b62 | 840 | static void __vmx_load_host_state(struct vcpu_vmx *vmx) |
33ed6329 | 841 | { |
a2fa3e9f | 842 | if (!vmx->host_state.loaded) |
33ed6329 AK |
843 | return; |
844 | ||
e1beb1d3 | 845 | ++vmx->vcpu.stat.host_state_reload; |
a2fa3e9f | 846 | vmx->host_state.loaded = 0; |
152d3f2f | 847 | if (vmx->host_state.fs_reload_needed) |
9581d442 | 848 | loadsegment(fs, vmx->host_state.fs_sel); |
152d3f2f | 849 | if (vmx->host_state.gs_ldt_reload_needed) { |
d6e88aec | 850 | kvm_load_ldt(vmx->host_state.ldt_sel); |
33ed6329 | 851 | #ifdef CONFIG_X86_64 |
9581d442 AK |
852 | load_gs_index(vmx->host_state.gs_sel); |
853 | wrmsrl(MSR_KERNEL_GS_BASE, current->thread.gs); | |
854 | #else | |
855 | loadsegment(gs, vmx->host_state.gs_sel); | |
33ed6329 | 856 | #endif |
33ed6329 | 857 | } |
152d3f2f | 858 | reload_tss(); |
44ea2b17 AK |
859 | #ifdef CONFIG_X86_64 |
860 | if (is_long_mode(&vmx->vcpu)) { | |
861 | rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); | |
862 | wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); | |
863 | } | |
864 | #endif | |
1c11e713 AK |
865 | if (current_thread_info()->status & TS_USEDFPU) |
866 | clts(); | |
3444d7da | 867 | load_gdt(&__get_cpu_var(host_gdt)); |
33ed6329 AK |
868 | } |
869 | ||
a9b21b62 AK |
870 | static void vmx_load_host_state(struct vcpu_vmx *vmx) |
871 | { | |
872 | preempt_disable(); | |
873 | __vmx_load_host_state(vmx); | |
874 | preempt_enable(); | |
875 | } | |
876 | ||
6aa8b732 AK |
877 | /* |
878 | * Switches to specified vcpu, until a matching vcpu_put(), but assumes | |
879 | * vcpu mutex is already taken. | |
880 | */ | |
15ad7146 | 881 | static void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
6aa8b732 | 882 | { |
a2fa3e9f | 883 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
4610c9cc | 884 | u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); |
6aa8b732 | 885 | |
4610c9cc DX |
886 | if (!vmm_exclusive) |
887 | kvm_cpu_vmxon(phys_addr); | |
888 | else if (vcpu->cpu != cpu) | |
8b9cf98c | 889 | vcpu_clear(vmx); |
6aa8b732 | 890 | |
a2fa3e9f | 891 | if (per_cpu(current_vmcs, cpu) != vmx->vmcs) { |
a2fa3e9f | 892 | per_cpu(current_vmcs, cpu) = vmx->vmcs; |
7725b894 | 893 | vmcs_load(vmx->vmcs); |
6aa8b732 AK |
894 | } |
895 | ||
896 | if (vcpu->cpu != cpu) { | |
d359192f | 897 | struct desc_ptr *gdt = &__get_cpu_var(host_gdt); |
6aa8b732 AK |
898 | unsigned long sysenter_esp; |
899 | ||
a8eeb04a | 900 | kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu); |
92fe13be DX |
901 | local_irq_disable(); |
902 | list_add(&vmx->local_vcpus_link, | |
903 | &per_cpu(vcpus_on_cpu, cpu)); | |
904 | local_irq_enable(); | |
905 | ||
6aa8b732 AK |
906 | /* |
907 | * Linux uses per-cpu TSS and GDT, so set these when switching | |
908 | * processors. | |
909 | */ | |
d6e88aec | 910 | vmcs_writel(HOST_TR_BASE, kvm_read_tr_base()); /* 22.2.4 */ |
d359192f | 911 | vmcs_writel(HOST_GDTR_BASE, gdt->address); /* 22.2.4 */ |
6aa8b732 AK |
912 | |
913 | rdmsrl(MSR_IA32_SYSENTER_ESP, sysenter_esp); | |
914 | vmcs_writel(HOST_IA32_SYSENTER_ESP, sysenter_esp); /* 22.2.3 */ | |
915 | } | |
6aa8b732 AK |
916 | } |
917 | ||
918 | static void vmx_vcpu_put(struct kvm_vcpu *vcpu) | |
919 | { | |
a9b21b62 | 920 | __vmx_load_host_state(to_vmx(vcpu)); |
4610c9cc | 921 | if (!vmm_exclusive) { |
b923e62e | 922 | __vcpu_clear(to_vmx(vcpu)); |
4610c9cc DX |
923 | kvm_cpu_vmxoff(); |
924 | } | |
6aa8b732 AK |
925 | } |
926 | ||
5fd86fcf AK |
927 | static void vmx_fpu_activate(struct kvm_vcpu *vcpu) |
928 | { | |
81231c69 AK |
929 | ulong cr0; |
930 | ||
5fd86fcf AK |
931 | if (vcpu->fpu_active) |
932 | return; | |
933 | vcpu->fpu_active = 1; | |
81231c69 AK |
934 | cr0 = vmcs_readl(GUEST_CR0); |
935 | cr0 &= ~(X86_CR0_TS | X86_CR0_MP); | |
936 | cr0 |= kvm_read_cr0_bits(vcpu, X86_CR0_TS | X86_CR0_MP); | |
937 | vmcs_writel(GUEST_CR0, cr0); | |
5fd86fcf | 938 | update_exception_bitmap(vcpu); |
edcafe3c AK |
939 | vcpu->arch.cr0_guest_owned_bits = X86_CR0_TS; |
940 | vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits); | |
5fd86fcf AK |
941 | } |
942 | ||
edcafe3c AK |
943 | static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu); |
944 | ||
5fd86fcf AK |
945 | static void vmx_fpu_deactivate(struct kvm_vcpu *vcpu) |
946 | { | |
edcafe3c | 947 | vmx_decache_cr0_guest_bits(vcpu); |
81231c69 | 948 | vmcs_set_bits(GUEST_CR0, X86_CR0_TS | X86_CR0_MP); |
5fd86fcf | 949 | update_exception_bitmap(vcpu); |
edcafe3c AK |
950 | vcpu->arch.cr0_guest_owned_bits = 0; |
951 | vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits); | |
952 | vmcs_writel(CR0_READ_SHADOW, vcpu->arch.cr0); | |
5fd86fcf AK |
953 | } |
954 | ||
6aa8b732 AK |
955 | static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu) |
956 | { | |
78ac8b47 | 957 | unsigned long rflags, save_rflags; |
345dcaa8 AK |
958 | |
959 | rflags = vmcs_readl(GUEST_RFLAGS); | |
78ac8b47 AK |
960 | if (to_vmx(vcpu)->rmode.vm86_active) { |
961 | rflags &= RMODE_GUEST_OWNED_EFLAGS_BITS; | |
962 | save_rflags = to_vmx(vcpu)->rmode.save_rflags; | |
963 | rflags |= save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; | |
964 | } | |
345dcaa8 | 965 | return rflags; |
6aa8b732 AK |
966 | } |
967 | ||
968 | static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) | |
969 | { | |
78ac8b47 AK |
970 | if (to_vmx(vcpu)->rmode.vm86_active) { |
971 | to_vmx(vcpu)->rmode.save_rflags = rflags; | |
053de044 | 972 | rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
78ac8b47 | 973 | } |
6aa8b732 AK |
974 | vmcs_writel(GUEST_RFLAGS, rflags); |
975 | } | |
976 | ||
2809f5d2 GC |
977 | static u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu, int mask) |
978 | { | |
979 | u32 interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); | |
980 | int ret = 0; | |
981 | ||
982 | if (interruptibility & GUEST_INTR_STATE_STI) | |
48005f64 | 983 | ret |= KVM_X86_SHADOW_INT_STI; |
2809f5d2 | 984 | if (interruptibility & GUEST_INTR_STATE_MOV_SS) |
48005f64 | 985 | ret |= KVM_X86_SHADOW_INT_MOV_SS; |
2809f5d2 GC |
986 | |
987 | return ret & mask; | |
988 | } | |
989 | ||
990 | static void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask) | |
991 | { | |
992 | u32 interruptibility_old = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); | |
993 | u32 interruptibility = interruptibility_old; | |
994 | ||
995 | interruptibility &= ~(GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS); | |
996 | ||
48005f64 | 997 | if (mask & KVM_X86_SHADOW_INT_MOV_SS) |
2809f5d2 | 998 | interruptibility |= GUEST_INTR_STATE_MOV_SS; |
48005f64 | 999 | else if (mask & KVM_X86_SHADOW_INT_STI) |
2809f5d2 GC |
1000 | interruptibility |= GUEST_INTR_STATE_STI; |
1001 | ||
1002 | if ((interruptibility != interruptibility_old)) | |
1003 | vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, interruptibility); | |
1004 | } | |
1005 | ||
6aa8b732 AK |
1006 | static void skip_emulated_instruction(struct kvm_vcpu *vcpu) |
1007 | { | |
1008 | unsigned long rip; | |
6aa8b732 | 1009 | |
5fdbf976 | 1010 | rip = kvm_rip_read(vcpu); |
6aa8b732 | 1011 | rip += vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
5fdbf976 | 1012 | kvm_rip_write(vcpu, rip); |
6aa8b732 | 1013 | |
2809f5d2 GC |
1014 | /* skipping an emulated instruction also counts */ |
1015 | vmx_set_interrupt_shadow(vcpu, 0); | |
6aa8b732 AK |
1016 | } |
1017 | ||
298101da | 1018 | static void vmx_queue_exception(struct kvm_vcpu *vcpu, unsigned nr, |
ce7ddec4 JR |
1019 | bool has_error_code, u32 error_code, |
1020 | bool reinject) | |
298101da | 1021 | { |
77ab6db0 | 1022 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
8ab2d2e2 | 1023 | u32 intr_info = nr | INTR_INFO_VALID_MASK; |
77ab6db0 | 1024 | |
8ab2d2e2 | 1025 | if (has_error_code) { |
77ab6db0 | 1026 | vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code); |
8ab2d2e2 JK |
1027 | intr_info |= INTR_INFO_DELIVER_CODE_MASK; |
1028 | } | |
77ab6db0 | 1029 | |
7ffd92c5 | 1030 | if (vmx->rmode.vm86_active) { |
77ab6db0 JK |
1031 | vmx->rmode.irq.pending = true; |
1032 | vmx->rmode.irq.vector = nr; | |
1033 | vmx->rmode.irq.rip = kvm_rip_read(vcpu); | |
ae0bb3e0 GN |
1034 | if (kvm_exception_is_soft(nr)) |
1035 | vmx->rmode.irq.rip += | |
1036 | vmx->vcpu.arch.event_exit_inst_len; | |
8ab2d2e2 JK |
1037 | intr_info |= INTR_TYPE_SOFT_INTR; |
1038 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr_info); | |
77ab6db0 JK |
1039 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, 1); |
1040 | kvm_rip_write(vcpu, vmx->rmode.irq.rip - 1); | |
1041 | return; | |
1042 | } | |
1043 | ||
66fd3f7f GN |
1044 | if (kvm_exception_is_soft(nr)) { |
1045 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, | |
1046 | vmx->vcpu.arch.event_exit_inst_len); | |
8ab2d2e2 JK |
1047 | intr_info |= INTR_TYPE_SOFT_EXCEPTION; |
1048 | } else | |
1049 | intr_info |= INTR_TYPE_HARD_EXCEPTION; | |
1050 | ||
1051 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr_info); | |
298101da AK |
1052 | } |
1053 | ||
4e47c7a6 SY |
1054 | static bool vmx_rdtscp_supported(void) |
1055 | { | |
1056 | return cpu_has_vmx_rdtscp(); | |
1057 | } | |
1058 | ||
a75beee6 ED |
1059 | /* |
1060 | * Swap MSR entry in host/guest MSR entry array. | |
1061 | */ | |
8b9cf98c | 1062 | static void move_msr_up(struct vcpu_vmx *vmx, int from, int to) |
a75beee6 | 1063 | { |
26bb0981 | 1064 | struct shared_msr_entry tmp; |
a2fa3e9f GH |
1065 | |
1066 | tmp = vmx->guest_msrs[to]; | |
1067 | vmx->guest_msrs[to] = vmx->guest_msrs[from]; | |
1068 | vmx->guest_msrs[from] = tmp; | |
a75beee6 ED |
1069 | } |
1070 | ||
e38aea3e AK |
1071 | /* |
1072 | * Set up the vmcs to automatically save and restore system | |
1073 | * msrs. Don't touch the 64-bit msrs if the guest is in legacy | |
1074 | * mode, as fiddling with msrs is very expensive. | |
1075 | */ | |
8b9cf98c | 1076 | static void setup_msrs(struct vcpu_vmx *vmx) |
e38aea3e | 1077 | { |
26bb0981 | 1078 | int save_nmsrs, index; |
5897297b | 1079 | unsigned long *msr_bitmap; |
e38aea3e | 1080 | |
33f9c505 | 1081 | vmx_load_host_state(vmx); |
a75beee6 ED |
1082 | save_nmsrs = 0; |
1083 | #ifdef CONFIG_X86_64 | |
8b9cf98c | 1084 | if (is_long_mode(&vmx->vcpu)) { |
8b9cf98c | 1085 | index = __find_msr_index(vmx, MSR_SYSCALL_MASK); |
a75beee6 | 1086 | if (index >= 0) |
8b9cf98c RR |
1087 | move_msr_up(vmx, index, save_nmsrs++); |
1088 | index = __find_msr_index(vmx, MSR_LSTAR); | |
a75beee6 | 1089 | if (index >= 0) |
8b9cf98c RR |
1090 | move_msr_up(vmx, index, save_nmsrs++); |
1091 | index = __find_msr_index(vmx, MSR_CSTAR); | |
a75beee6 | 1092 | if (index >= 0) |
8b9cf98c | 1093 | move_msr_up(vmx, index, save_nmsrs++); |
4e47c7a6 SY |
1094 | index = __find_msr_index(vmx, MSR_TSC_AUX); |
1095 | if (index >= 0 && vmx->rdtscp_enabled) | |
1096 | move_msr_up(vmx, index, save_nmsrs++); | |
a75beee6 | 1097 | /* |
8c06585d | 1098 | * MSR_STAR is only needed on long mode guests, and only |
a75beee6 ED |
1099 | * if efer.sce is enabled. |
1100 | */ | |
8c06585d | 1101 | index = __find_msr_index(vmx, MSR_STAR); |
f6801dff | 1102 | if ((index >= 0) && (vmx->vcpu.arch.efer & EFER_SCE)) |
8b9cf98c | 1103 | move_msr_up(vmx, index, save_nmsrs++); |
a75beee6 ED |
1104 | } |
1105 | #endif | |
92c0d900 AK |
1106 | index = __find_msr_index(vmx, MSR_EFER); |
1107 | if (index >= 0 && update_transition_efer(vmx, index)) | |
26bb0981 | 1108 | move_msr_up(vmx, index, save_nmsrs++); |
e38aea3e | 1109 | |
26bb0981 | 1110 | vmx->save_nmsrs = save_nmsrs; |
5897297b AK |
1111 | |
1112 | if (cpu_has_vmx_msr_bitmap()) { | |
1113 | if (is_long_mode(&vmx->vcpu)) | |
1114 | msr_bitmap = vmx_msr_bitmap_longmode; | |
1115 | else | |
1116 | msr_bitmap = vmx_msr_bitmap_legacy; | |
1117 | ||
1118 | vmcs_write64(MSR_BITMAP, __pa(msr_bitmap)); | |
1119 | } | |
e38aea3e AK |
1120 | } |
1121 | ||
6aa8b732 AK |
1122 | /* |
1123 | * reads and returns guest's timestamp counter "register" | |
1124 | * guest_tsc = host_tsc + tsc_offset -- 21.3 | |
1125 | */ | |
1126 | static u64 guest_read_tsc(void) | |
1127 | { | |
1128 | u64 host_tsc, tsc_offset; | |
1129 | ||
1130 | rdtscll(host_tsc); | |
1131 | tsc_offset = vmcs_read64(TSC_OFFSET); | |
1132 | return host_tsc + tsc_offset; | |
1133 | } | |
1134 | ||
1135 | /* | |
99e3e30a | 1136 | * writes 'offset' into guest's timestamp counter offset register |
6aa8b732 | 1137 | */ |
99e3e30a | 1138 | static void vmx_write_tsc_offset(struct kvm_vcpu *vcpu, u64 offset) |
6aa8b732 | 1139 | { |
f4e1b3c8 | 1140 | vmcs_write64(TSC_OFFSET, offset); |
6aa8b732 AK |
1141 | } |
1142 | ||
e48672fa ZA |
1143 | static void vmx_adjust_tsc_offset(struct kvm_vcpu *vcpu, s64 adjustment) |
1144 | { | |
1145 | u64 offset = vmcs_read64(TSC_OFFSET); | |
1146 | vmcs_write64(TSC_OFFSET, offset + adjustment); | |
1147 | } | |
1148 | ||
6aa8b732 AK |
1149 | /* |
1150 | * Reads an msr value (of 'msr_index') into 'pdata'. | |
1151 | * Returns 0 on success, non-0 otherwise. | |
1152 | * Assumes vcpu_load() was already called. | |
1153 | */ | |
1154 | static int vmx_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata) | |
1155 | { | |
1156 | u64 data; | |
26bb0981 | 1157 | struct shared_msr_entry *msr; |
6aa8b732 AK |
1158 | |
1159 | if (!pdata) { | |
1160 | printk(KERN_ERR "BUG: get_msr called with NULL pdata\n"); | |
1161 | return -EINVAL; | |
1162 | } | |
1163 | ||
1164 | switch (msr_index) { | |
05b3e0c2 | 1165 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
1166 | case MSR_FS_BASE: |
1167 | data = vmcs_readl(GUEST_FS_BASE); | |
1168 | break; | |
1169 | case MSR_GS_BASE: | |
1170 | data = vmcs_readl(GUEST_GS_BASE); | |
1171 | break; | |
44ea2b17 AK |
1172 | case MSR_KERNEL_GS_BASE: |
1173 | vmx_load_host_state(to_vmx(vcpu)); | |
1174 | data = to_vmx(vcpu)->msr_guest_kernel_gs_base; | |
1175 | break; | |
26bb0981 | 1176 | #endif |
6aa8b732 | 1177 | case MSR_EFER: |
3bab1f5d | 1178 | return kvm_get_msr_common(vcpu, msr_index, pdata); |
af24a4e4 | 1179 | case MSR_IA32_TSC: |
6aa8b732 AK |
1180 | data = guest_read_tsc(); |
1181 | break; | |
1182 | case MSR_IA32_SYSENTER_CS: | |
1183 | data = vmcs_read32(GUEST_SYSENTER_CS); | |
1184 | break; | |
1185 | case MSR_IA32_SYSENTER_EIP: | |
f5b42c33 | 1186 | data = vmcs_readl(GUEST_SYSENTER_EIP); |
6aa8b732 AK |
1187 | break; |
1188 | case MSR_IA32_SYSENTER_ESP: | |
f5b42c33 | 1189 | data = vmcs_readl(GUEST_SYSENTER_ESP); |
6aa8b732 | 1190 | break; |
4e47c7a6 SY |
1191 | case MSR_TSC_AUX: |
1192 | if (!to_vmx(vcpu)->rdtscp_enabled) | |
1193 | return 1; | |
1194 | /* Otherwise falls through */ | |
6aa8b732 | 1195 | default: |
26bb0981 | 1196 | vmx_load_host_state(to_vmx(vcpu)); |
8b9cf98c | 1197 | msr = find_msr_entry(to_vmx(vcpu), msr_index); |
3bab1f5d | 1198 | if (msr) { |
542423b0 | 1199 | vmx_load_host_state(to_vmx(vcpu)); |
3bab1f5d AK |
1200 | data = msr->data; |
1201 | break; | |
6aa8b732 | 1202 | } |
3bab1f5d | 1203 | return kvm_get_msr_common(vcpu, msr_index, pdata); |
6aa8b732 AK |
1204 | } |
1205 | ||
1206 | *pdata = data; | |
1207 | return 0; | |
1208 | } | |
1209 | ||
1210 | /* | |
1211 | * Writes msr value into into the appropriate "register". | |
1212 | * Returns 0 on success, non-0 otherwise. | |
1213 | * Assumes vcpu_load() was already called. | |
1214 | */ | |
1215 | static int vmx_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data) | |
1216 | { | |
a2fa3e9f | 1217 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
26bb0981 | 1218 | struct shared_msr_entry *msr; |
2cc51560 ED |
1219 | int ret = 0; |
1220 | ||
6aa8b732 | 1221 | switch (msr_index) { |
3bab1f5d | 1222 | case MSR_EFER: |
a9b21b62 | 1223 | vmx_load_host_state(vmx); |
2cc51560 | 1224 | ret = kvm_set_msr_common(vcpu, msr_index, data); |
2cc51560 | 1225 | break; |
16175a79 | 1226 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
1227 | case MSR_FS_BASE: |
1228 | vmcs_writel(GUEST_FS_BASE, data); | |
1229 | break; | |
1230 | case MSR_GS_BASE: | |
1231 | vmcs_writel(GUEST_GS_BASE, data); | |
1232 | break; | |
44ea2b17 AK |
1233 | case MSR_KERNEL_GS_BASE: |
1234 | vmx_load_host_state(vmx); | |
1235 | vmx->msr_guest_kernel_gs_base = data; | |
1236 | break; | |
6aa8b732 AK |
1237 | #endif |
1238 | case MSR_IA32_SYSENTER_CS: | |
1239 | vmcs_write32(GUEST_SYSENTER_CS, data); | |
1240 | break; | |
1241 | case MSR_IA32_SYSENTER_EIP: | |
f5b42c33 | 1242 | vmcs_writel(GUEST_SYSENTER_EIP, data); |
6aa8b732 AK |
1243 | break; |
1244 | case MSR_IA32_SYSENTER_ESP: | |
f5b42c33 | 1245 | vmcs_writel(GUEST_SYSENTER_ESP, data); |
6aa8b732 | 1246 | break; |
af24a4e4 | 1247 | case MSR_IA32_TSC: |
99e3e30a | 1248 | kvm_write_tsc(vcpu, data); |
6aa8b732 | 1249 | break; |
468d472f SY |
1250 | case MSR_IA32_CR_PAT: |
1251 | if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) { | |
1252 | vmcs_write64(GUEST_IA32_PAT, data); | |
1253 | vcpu->arch.pat = data; | |
1254 | break; | |
1255 | } | |
4e47c7a6 SY |
1256 | ret = kvm_set_msr_common(vcpu, msr_index, data); |
1257 | break; | |
1258 | case MSR_TSC_AUX: | |
1259 | if (!vmx->rdtscp_enabled) | |
1260 | return 1; | |
1261 | /* Check reserved bit, higher 32 bits should be zero */ | |
1262 | if ((data >> 32) != 0) | |
1263 | return 1; | |
1264 | /* Otherwise falls through */ | |
6aa8b732 | 1265 | default: |
8b9cf98c | 1266 | msr = find_msr_entry(vmx, msr_index); |
3bab1f5d | 1267 | if (msr) { |
542423b0 | 1268 | vmx_load_host_state(vmx); |
3bab1f5d AK |
1269 | msr->data = data; |
1270 | break; | |
6aa8b732 | 1271 | } |
2cc51560 | 1272 | ret = kvm_set_msr_common(vcpu, msr_index, data); |
6aa8b732 AK |
1273 | } |
1274 | ||
2cc51560 | 1275 | return ret; |
6aa8b732 AK |
1276 | } |
1277 | ||
5fdbf976 | 1278 | static void vmx_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg) |
6aa8b732 | 1279 | { |
5fdbf976 MT |
1280 | __set_bit(reg, (unsigned long *)&vcpu->arch.regs_avail); |
1281 | switch (reg) { | |
1282 | case VCPU_REGS_RSP: | |
1283 | vcpu->arch.regs[VCPU_REGS_RSP] = vmcs_readl(GUEST_RSP); | |
1284 | break; | |
1285 | case VCPU_REGS_RIP: | |
1286 | vcpu->arch.regs[VCPU_REGS_RIP] = vmcs_readl(GUEST_RIP); | |
1287 | break; | |
6de4f3ad AK |
1288 | case VCPU_EXREG_PDPTR: |
1289 | if (enable_ept) | |
1290 | ept_save_pdptrs(vcpu); | |
1291 | break; | |
5fdbf976 MT |
1292 | default: |
1293 | break; | |
1294 | } | |
6aa8b732 AK |
1295 | } |
1296 | ||
355be0b9 | 1297 | static void set_guest_debug(struct kvm_vcpu *vcpu, struct kvm_guest_debug *dbg) |
6aa8b732 | 1298 | { |
ae675ef0 JK |
1299 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) |
1300 | vmcs_writel(GUEST_DR7, dbg->arch.debugreg[7]); | |
1301 | else | |
1302 | vmcs_writel(GUEST_DR7, vcpu->arch.dr7); | |
1303 | ||
abd3f2d6 | 1304 | update_exception_bitmap(vcpu); |
6aa8b732 AK |
1305 | } |
1306 | ||
1307 | static __init int cpu_has_kvm_support(void) | |
1308 | { | |
6210e37b | 1309 | return cpu_has_vmx(); |
6aa8b732 AK |
1310 | } |
1311 | ||
1312 | static __init int vmx_disabled_by_bios(void) | |
1313 | { | |
1314 | u64 msr; | |
1315 | ||
1316 | rdmsrl(MSR_IA32_FEATURE_CONTROL, msr); | |
cafd6659 SW |
1317 | if (msr & FEATURE_CONTROL_LOCKED) { |
1318 | if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) | |
1319 | && tboot_enabled()) | |
1320 | return 1; | |
1321 | if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) | |
1322 | && !tboot_enabled()) | |
1323 | return 1; | |
1324 | } | |
1325 | ||
1326 | return 0; | |
62b3ffb8 | 1327 | /* locked but not enabled */ |
6aa8b732 AK |
1328 | } |
1329 | ||
7725b894 DX |
1330 | static void kvm_cpu_vmxon(u64 addr) |
1331 | { | |
1332 | asm volatile (ASM_VMX_VMXON_RAX | |
1333 | : : "a"(&addr), "m"(addr) | |
1334 | : "memory", "cc"); | |
1335 | } | |
1336 | ||
10474ae8 | 1337 | static int hardware_enable(void *garbage) |
6aa8b732 AK |
1338 | { |
1339 | int cpu = raw_smp_processor_id(); | |
1340 | u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); | |
cafd6659 | 1341 | u64 old, test_bits; |
6aa8b732 | 1342 | |
10474ae8 AG |
1343 | if (read_cr4() & X86_CR4_VMXE) |
1344 | return -EBUSY; | |
1345 | ||
543e4243 | 1346 | INIT_LIST_HEAD(&per_cpu(vcpus_on_cpu, cpu)); |
6aa8b732 | 1347 | rdmsrl(MSR_IA32_FEATURE_CONTROL, old); |
cafd6659 SW |
1348 | |
1349 | test_bits = FEATURE_CONTROL_LOCKED; | |
1350 | test_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; | |
1351 | if (tboot_enabled()) | |
1352 | test_bits |= FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX; | |
1353 | ||
1354 | if ((old & test_bits) != test_bits) { | |
6aa8b732 | 1355 | /* enable and lock */ |
cafd6659 SW |
1356 | wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits); |
1357 | } | |
66aee91a | 1358 | write_cr4(read_cr4() | X86_CR4_VMXE); /* FIXME: not cpu hotplug safe */ |
10474ae8 | 1359 | |
4610c9cc DX |
1360 | if (vmm_exclusive) { |
1361 | kvm_cpu_vmxon(phys_addr); | |
1362 | ept_sync_global(); | |
1363 | } | |
10474ae8 | 1364 | |
3444d7da AK |
1365 | store_gdt(&__get_cpu_var(host_gdt)); |
1366 | ||
10474ae8 | 1367 | return 0; |
6aa8b732 AK |
1368 | } |
1369 | ||
543e4243 AK |
1370 | static void vmclear_local_vcpus(void) |
1371 | { | |
1372 | int cpu = raw_smp_processor_id(); | |
1373 | struct vcpu_vmx *vmx, *n; | |
1374 | ||
1375 | list_for_each_entry_safe(vmx, n, &per_cpu(vcpus_on_cpu, cpu), | |
1376 | local_vcpus_link) | |
1377 | __vcpu_clear(vmx); | |
1378 | } | |
1379 | ||
710ff4a8 EH |
1380 | |
1381 | /* Just like cpu_vmxoff(), but with the __kvm_handle_fault_on_reboot() | |
1382 | * tricks. | |
1383 | */ | |
1384 | static void kvm_cpu_vmxoff(void) | |
6aa8b732 | 1385 | { |
4ecac3fd | 1386 | asm volatile (__ex(ASM_VMX_VMXOFF) : : : "cc"); |
6aa8b732 AK |
1387 | } |
1388 | ||
710ff4a8 EH |
1389 | static void hardware_disable(void *garbage) |
1390 | { | |
4610c9cc DX |
1391 | if (vmm_exclusive) { |
1392 | vmclear_local_vcpus(); | |
1393 | kvm_cpu_vmxoff(); | |
1394 | } | |
7725b894 | 1395 | write_cr4(read_cr4() & ~X86_CR4_VMXE); |
710ff4a8 EH |
1396 | } |
1397 | ||
1c3d14fe | 1398 | static __init int adjust_vmx_controls(u32 ctl_min, u32 ctl_opt, |
d77c26fc | 1399 | u32 msr, u32 *result) |
1c3d14fe YS |
1400 | { |
1401 | u32 vmx_msr_low, vmx_msr_high; | |
1402 | u32 ctl = ctl_min | ctl_opt; | |
1403 | ||
1404 | rdmsr(msr, vmx_msr_low, vmx_msr_high); | |
1405 | ||
1406 | ctl &= vmx_msr_high; /* bit == 0 in high word ==> must be zero */ | |
1407 | ctl |= vmx_msr_low; /* bit == 1 in low word ==> must be one */ | |
1408 | ||
1409 | /* Ensure minimum (required) set of control bits are supported. */ | |
1410 | if (ctl_min & ~ctl) | |
002c7f7c | 1411 | return -EIO; |
1c3d14fe YS |
1412 | |
1413 | *result = ctl; | |
1414 | return 0; | |
1415 | } | |
1416 | ||
002c7f7c | 1417 | static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf) |
6aa8b732 AK |
1418 | { |
1419 | u32 vmx_msr_low, vmx_msr_high; | |
d56f546d | 1420 | u32 min, opt, min2, opt2; |
1c3d14fe YS |
1421 | u32 _pin_based_exec_control = 0; |
1422 | u32 _cpu_based_exec_control = 0; | |
f78e0e2e | 1423 | u32 _cpu_based_2nd_exec_control = 0; |
1c3d14fe YS |
1424 | u32 _vmexit_control = 0; |
1425 | u32 _vmentry_control = 0; | |
1426 | ||
1427 | min = PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING; | |
f08864b4 | 1428 | opt = PIN_BASED_VIRTUAL_NMIS; |
1c3d14fe YS |
1429 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PINBASED_CTLS, |
1430 | &_pin_based_exec_control) < 0) | |
002c7f7c | 1431 | return -EIO; |
1c3d14fe YS |
1432 | |
1433 | min = CPU_BASED_HLT_EXITING | | |
1434 | #ifdef CONFIG_X86_64 | |
1435 | CPU_BASED_CR8_LOAD_EXITING | | |
1436 | CPU_BASED_CR8_STORE_EXITING | | |
1437 | #endif | |
d56f546d SY |
1438 | CPU_BASED_CR3_LOAD_EXITING | |
1439 | CPU_BASED_CR3_STORE_EXITING | | |
1c3d14fe YS |
1440 | CPU_BASED_USE_IO_BITMAPS | |
1441 | CPU_BASED_MOV_DR_EXITING | | |
a7052897 | 1442 | CPU_BASED_USE_TSC_OFFSETING | |
59708670 SY |
1443 | CPU_BASED_MWAIT_EXITING | |
1444 | CPU_BASED_MONITOR_EXITING | | |
a7052897 | 1445 | CPU_BASED_INVLPG_EXITING; |
f78e0e2e | 1446 | opt = CPU_BASED_TPR_SHADOW | |
25c5f225 | 1447 | CPU_BASED_USE_MSR_BITMAPS | |
f78e0e2e | 1448 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; |
1c3d14fe YS |
1449 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PROCBASED_CTLS, |
1450 | &_cpu_based_exec_control) < 0) | |
002c7f7c | 1451 | return -EIO; |
6e5d865c YS |
1452 | #ifdef CONFIG_X86_64 |
1453 | if ((_cpu_based_exec_control & CPU_BASED_TPR_SHADOW)) | |
1454 | _cpu_based_exec_control &= ~CPU_BASED_CR8_LOAD_EXITING & | |
1455 | ~CPU_BASED_CR8_STORE_EXITING; | |
1456 | #endif | |
f78e0e2e | 1457 | if (_cpu_based_exec_control & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) { |
d56f546d SY |
1458 | min2 = 0; |
1459 | opt2 = SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | | |
2384d2b3 | 1460 | SECONDARY_EXEC_WBINVD_EXITING | |
d56f546d | 1461 | SECONDARY_EXEC_ENABLE_VPID | |
3a624e29 | 1462 | SECONDARY_EXEC_ENABLE_EPT | |
4b8d54f9 | 1463 | SECONDARY_EXEC_UNRESTRICTED_GUEST | |
4e47c7a6 SY |
1464 | SECONDARY_EXEC_PAUSE_LOOP_EXITING | |
1465 | SECONDARY_EXEC_RDTSCP; | |
d56f546d SY |
1466 | if (adjust_vmx_controls(min2, opt2, |
1467 | MSR_IA32_VMX_PROCBASED_CTLS2, | |
f78e0e2e SY |
1468 | &_cpu_based_2nd_exec_control) < 0) |
1469 | return -EIO; | |
1470 | } | |
1471 | #ifndef CONFIG_X86_64 | |
1472 | if (!(_cpu_based_2nd_exec_control & | |
1473 | SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)) | |
1474 | _cpu_based_exec_control &= ~CPU_BASED_TPR_SHADOW; | |
1475 | #endif | |
d56f546d | 1476 | if (_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_EPT) { |
a7052897 MT |
1477 | /* CR3 accesses and invlpg don't need to cause VM Exits when EPT |
1478 | enabled */ | |
5fff7d27 GN |
1479 | _cpu_based_exec_control &= ~(CPU_BASED_CR3_LOAD_EXITING | |
1480 | CPU_BASED_CR3_STORE_EXITING | | |
1481 | CPU_BASED_INVLPG_EXITING); | |
d56f546d SY |
1482 | rdmsr(MSR_IA32_VMX_EPT_VPID_CAP, |
1483 | vmx_capability.ept, vmx_capability.vpid); | |
1484 | } | |
1c3d14fe YS |
1485 | |
1486 | min = 0; | |
1487 | #ifdef CONFIG_X86_64 | |
1488 | min |= VM_EXIT_HOST_ADDR_SPACE_SIZE; | |
1489 | #endif | |
468d472f | 1490 | opt = VM_EXIT_SAVE_IA32_PAT | VM_EXIT_LOAD_IA32_PAT; |
1c3d14fe YS |
1491 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_EXIT_CTLS, |
1492 | &_vmexit_control) < 0) | |
002c7f7c | 1493 | return -EIO; |
1c3d14fe | 1494 | |
468d472f SY |
1495 | min = 0; |
1496 | opt = VM_ENTRY_LOAD_IA32_PAT; | |
1c3d14fe YS |
1497 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_ENTRY_CTLS, |
1498 | &_vmentry_control) < 0) | |
002c7f7c | 1499 | return -EIO; |
6aa8b732 | 1500 | |
c68876fd | 1501 | rdmsr(MSR_IA32_VMX_BASIC, vmx_msr_low, vmx_msr_high); |
1c3d14fe YS |
1502 | |
1503 | /* IA-32 SDM Vol 3B: VMCS size is never greater than 4kB. */ | |
1504 | if ((vmx_msr_high & 0x1fff) > PAGE_SIZE) | |
002c7f7c | 1505 | return -EIO; |
1c3d14fe YS |
1506 | |
1507 | #ifdef CONFIG_X86_64 | |
1508 | /* IA-32 SDM Vol 3B: 64-bit CPUs always have VMX_BASIC_MSR[48]==0. */ | |
1509 | if (vmx_msr_high & (1u<<16)) | |
002c7f7c | 1510 | return -EIO; |
1c3d14fe YS |
1511 | #endif |
1512 | ||
1513 | /* Require Write-Back (WB) memory type for VMCS accesses. */ | |
1514 | if (((vmx_msr_high >> 18) & 15) != 6) | |
002c7f7c | 1515 | return -EIO; |
1c3d14fe | 1516 | |
002c7f7c YS |
1517 | vmcs_conf->size = vmx_msr_high & 0x1fff; |
1518 | vmcs_conf->order = get_order(vmcs_config.size); | |
1519 | vmcs_conf->revision_id = vmx_msr_low; | |
1c3d14fe | 1520 | |
002c7f7c YS |
1521 | vmcs_conf->pin_based_exec_ctrl = _pin_based_exec_control; |
1522 | vmcs_conf->cpu_based_exec_ctrl = _cpu_based_exec_control; | |
f78e0e2e | 1523 | vmcs_conf->cpu_based_2nd_exec_ctrl = _cpu_based_2nd_exec_control; |
002c7f7c YS |
1524 | vmcs_conf->vmexit_ctrl = _vmexit_control; |
1525 | vmcs_conf->vmentry_ctrl = _vmentry_control; | |
1c3d14fe YS |
1526 | |
1527 | return 0; | |
c68876fd | 1528 | } |
6aa8b732 AK |
1529 | |
1530 | static struct vmcs *alloc_vmcs_cpu(int cpu) | |
1531 | { | |
1532 | int node = cpu_to_node(cpu); | |
1533 | struct page *pages; | |
1534 | struct vmcs *vmcs; | |
1535 | ||
6484eb3e | 1536 | pages = alloc_pages_exact_node(node, GFP_KERNEL, vmcs_config.order); |
6aa8b732 AK |
1537 | if (!pages) |
1538 | return NULL; | |
1539 | vmcs = page_address(pages); | |
1c3d14fe YS |
1540 | memset(vmcs, 0, vmcs_config.size); |
1541 | vmcs->revision_id = vmcs_config.revision_id; /* vmcs revision id */ | |
6aa8b732 AK |
1542 | return vmcs; |
1543 | } | |
1544 | ||
1545 | static struct vmcs *alloc_vmcs(void) | |
1546 | { | |
d3b2c338 | 1547 | return alloc_vmcs_cpu(raw_smp_processor_id()); |
6aa8b732 AK |
1548 | } |
1549 | ||
1550 | static void free_vmcs(struct vmcs *vmcs) | |
1551 | { | |
1c3d14fe | 1552 | free_pages((unsigned long)vmcs, vmcs_config.order); |
6aa8b732 AK |
1553 | } |
1554 | ||
39959588 | 1555 | static void free_kvm_area(void) |
6aa8b732 AK |
1556 | { |
1557 | int cpu; | |
1558 | ||
3230bb47 | 1559 | for_each_possible_cpu(cpu) { |
6aa8b732 | 1560 | free_vmcs(per_cpu(vmxarea, cpu)); |
3230bb47 ZA |
1561 | per_cpu(vmxarea, cpu) = NULL; |
1562 | } | |
6aa8b732 AK |
1563 | } |
1564 | ||
6aa8b732 AK |
1565 | static __init int alloc_kvm_area(void) |
1566 | { | |
1567 | int cpu; | |
1568 | ||
3230bb47 | 1569 | for_each_possible_cpu(cpu) { |
6aa8b732 AK |
1570 | struct vmcs *vmcs; |
1571 | ||
1572 | vmcs = alloc_vmcs_cpu(cpu); | |
1573 | if (!vmcs) { | |
1574 | free_kvm_area(); | |
1575 | return -ENOMEM; | |
1576 | } | |
1577 | ||
1578 | per_cpu(vmxarea, cpu) = vmcs; | |
1579 | } | |
1580 | return 0; | |
1581 | } | |
1582 | ||
1583 | static __init int hardware_setup(void) | |
1584 | { | |
002c7f7c YS |
1585 | if (setup_vmcs_config(&vmcs_config) < 0) |
1586 | return -EIO; | |
50a37eb4 JR |
1587 | |
1588 | if (boot_cpu_has(X86_FEATURE_NX)) | |
1589 | kvm_enable_efer_bits(EFER_NX); | |
1590 | ||
93ba03c2 SY |
1591 | if (!cpu_has_vmx_vpid()) |
1592 | enable_vpid = 0; | |
1593 | ||
4bc9b982 SY |
1594 | if (!cpu_has_vmx_ept() || |
1595 | !cpu_has_vmx_ept_4levels()) { | |
93ba03c2 | 1596 | enable_ept = 0; |
3a624e29 NK |
1597 | enable_unrestricted_guest = 0; |
1598 | } | |
1599 | ||
1600 | if (!cpu_has_vmx_unrestricted_guest()) | |
1601 | enable_unrestricted_guest = 0; | |
93ba03c2 SY |
1602 | |
1603 | if (!cpu_has_vmx_flexpriority()) | |
1604 | flexpriority_enabled = 0; | |
1605 | ||
95ba8273 GN |
1606 | if (!cpu_has_vmx_tpr_shadow()) |
1607 | kvm_x86_ops->update_cr8_intercept = NULL; | |
1608 | ||
54dee993 MT |
1609 | if (enable_ept && !cpu_has_vmx_ept_2m_page()) |
1610 | kvm_disable_largepages(); | |
1611 | ||
4b8d54f9 ZE |
1612 | if (!cpu_has_vmx_ple()) |
1613 | ple_gap = 0; | |
1614 | ||
6aa8b732 AK |
1615 | return alloc_kvm_area(); |
1616 | } | |
1617 | ||
1618 | static __exit void hardware_unsetup(void) | |
1619 | { | |
1620 | free_kvm_area(); | |
1621 | } | |
1622 | ||
6aa8b732 AK |
1623 | static void fix_pmode_dataseg(int seg, struct kvm_save_segment *save) |
1624 | { | |
1625 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
1626 | ||
6af11b9e | 1627 | if (vmcs_readl(sf->base) == save->base && (save->base & AR_S_MASK)) { |
6aa8b732 AK |
1628 | vmcs_write16(sf->selector, save->selector); |
1629 | vmcs_writel(sf->base, save->base); | |
1630 | vmcs_write32(sf->limit, save->limit); | |
1631 | vmcs_write32(sf->ar_bytes, save->ar); | |
1632 | } else { | |
1633 | u32 dpl = (vmcs_read16(sf->selector) & SELECTOR_RPL_MASK) | |
1634 | << AR_DPL_SHIFT; | |
1635 | vmcs_write32(sf->ar_bytes, 0x93 | dpl); | |
1636 | } | |
1637 | } | |
1638 | ||
1639 | static void enter_pmode(struct kvm_vcpu *vcpu) | |
1640 | { | |
1641 | unsigned long flags; | |
a89a8fb9 | 1642 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
6aa8b732 | 1643 | |
a89a8fb9 | 1644 | vmx->emulation_required = 1; |
7ffd92c5 | 1645 | vmx->rmode.vm86_active = 0; |
6aa8b732 | 1646 | |
7ffd92c5 AK |
1647 | vmcs_writel(GUEST_TR_BASE, vmx->rmode.tr.base); |
1648 | vmcs_write32(GUEST_TR_LIMIT, vmx->rmode.tr.limit); | |
1649 | vmcs_write32(GUEST_TR_AR_BYTES, vmx->rmode.tr.ar); | |
6aa8b732 AK |
1650 | |
1651 | flags = vmcs_readl(GUEST_RFLAGS); | |
78ac8b47 AK |
1652 | flags &= RMODE_GUEST_OWNED_EFLAGS_BITS; |
1653 | flags |= vmx->rmode.save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; | |
6aa8b732 AK |
1654 | vmcs_writel(GUEST_RFLAGS, flags); |
1655 | ||
66aee91a RR |
1656 | vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~X86_CR4_VME) | |
1657 | (vmcs_readl(CR4_READ_SHADOW) & X86_CR4_VME)); | |
6aa8b732 AK |
1658 | |
1659 | update_exception_bitmap(vcpu); | |
1660 | ||
a89a8fb9 MG |
1661 | if (emulate_invalid_guest_state) |
1662 | return; | |
1663 | ||
7ffd92c5 AK |
1664 | fix_pmode_dataseg(VCPU_SREG_ES, &vmx->rmode.es); |
1665 | fix_pmode_dataseg(VCPU_SREG_DS, &vmx->rmode.ds); | |
1666 | fix_pmode_dataseg(VCPU_SREG_GS, &vmx->rmode.gs); | |
1667 | fix_pmode_dataseg(VCPU_SREG_FS, &vmx->rmode.fs); | |
6aa8b732 AK |
1668 | |
1669 | vmcs_write16(GUEST_SS_SELECTOR, 0); | |
1670 | vmcs_write32(GUEST_SS_AR_BYTES, 0x93); | |
1671 | ||
1672 | vmcs_write16(GUEST_CS_SELECTOR, | |
1673 | vmcs_read16(GUEST_CS_SELECTOR) & ~SELECTOR_RPL_MASK); | |
1674 | vmcs_write32(GUEST_CS_AR_BYTES, 0x9b); | |
1675 | } | |
1676 | ||
d77c26fc | 1677 | static gva_t rmode_tss_base(struct kvm *kvm) |
6aa8b732 | 1678 | { |
bfc6d222 | 1679 | if (!kvm->arch.tss_addr) { |
bc6678a3 MT |
1680 | struct kvm_memslots *slots; |
1681 | gfn_t base_gfn; | |
1682 | ||
90d83dc3 | 1683 | slots = kvm_memslots(kvm); |
f495c6e5 | 1684 | base_gfn = slots->memslots[0].base_gfn + |
46a26bf5 | 1685 | kvm->memslots->memslots[0].npages - 3; |
cbc94022 IE |
1686 | return base_gfn << PAGE_SHIFT; |
1687 | } | |
bfc6d222 | 1688 | return kvm->arch.tss_addr; |
6aa8b732 AK |
1689 | } |
1690 | ||
1691 | static void fix_rmode_seg(int seg, struct kvm_save_segment *save) | |
1692 | { | |
1693 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
1694 | ||
1695 | save->selector = vmcs_read16(sf->selector); | |
1696 | save->base = vmcs_readl(sf->base); | |
1697 | save->limit = vmcs_read32(sf->limit); | |
1698 | save->ar = vmcs_read32(sf->ar_bytes); | |
15b00f32 JK |
1699 | vmcs_write16(sf->selector, save->base >> 4); |
1700 | vmcs_write32(sf->base, save->base & 0xfffff); | |
6aa8b732 AK |
1701 | vmcs_write32(sf->limit, 0xffff); |
1702 | vmcs_write32(sf->ar_bytes, 0xf3); | |
1703 | } | |
1704 | ||
1705 | static void enter_rmode(struct kvm_vcpu *vcpu) | |
1706 | { | |
1707 | unsigned long flags; | |
a89a8fb9 | 1708 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
6aa8b732 | 1709 | |
3a624e29 NK |
1710 | if (enable_unrestricted_guest) |
1711 | return; | |
1712 | ||
a89a8fb9 | 1713 | vmx->emulation_required = 1; |
7ffd92c5 | 1714 | vmx->rmode.vm86_active = 1; |
6aa8b732 | 1715 | |
7ffd92c5 | 1716 | vmx->rmode.tr.base = vmcs_readl(GUEST_TR_BASE); |
6aa8b732 AK |
1717 | vmcs_writel(GUEST_TR_BASE, rmode_tss_base(vcpu->kvm)); |
1718 | ||
7ffd92c5 | 1719 | vmx->rmode.tr.limit = vmcs_read32(GUEST_TR_LIMIT); |
6aa8b732 AK |
1720 | vmcs_write32(GUEST_TR_LIMIT, RMODE_TSS_SIZE - 1); |
1721 | ||
7ffd92c5 | 1722 | vmx->rmode.tr.ar = vmcs_read32(GUEST_TR_AR_BYTES); |
6aa8b732 AK |
1723 | vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); |
1724 | ||
1725 | flags = vmcs_readl(GUEST_RFLAGS); | |
78ac8b47 | 1726 | vmx->rmode.save_rflags = flags; |
6aa8b732 | 1727 | |
053de044 | 1728 | flags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
6aa8b732 AK |
1729 | |
1730 | vmcs_writel(GUEST_RFLAGS, flags); | |
66aee91a | 1731 | vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | X86_CR4_VME); |
6aa8b732 AK |
1732 | update_exception_bitmap(vcpu); |
1733 | ||
a89a8fb9 MG |
1734 | if (emulate_invalid_guest_state) |
1735 | goto continue_rmode; | |
1736 | ||
6aa8b732 AK |
1737 | vmcs_write16(GUEST_SS_SELECTOR, vmcs_readl(GUEST_SS_BASE) >> 4); |
1738 | vmcs_write32(GUEST_SS_LIMIT, 0xffff); | |
1739 | vmcs_write32(GUEST_SS_AR_BYTES, 0xf3); | |
1740 | ||
1741 | vmcs_write32(GUEST_CS_AR_BYTES, 0xf3); | |
abacf8df | 1742 | vmcs_write32(GUEST_CS_LIMIT, 0xffff); |
8cb5b033 AK |
1743 | if (vmcs_readl(GUEST_CS_BASE) == 0xffff0000) |
1744 | vmcs_writel(GUEST_CS_BASE, 0xf0000); | |
6aa8b732 AK |
1745 | vmcs_write16(GUEST_CS_SELECTOR, vmcs_readl(GUEST_CS_BASE) >> 4); |
1746 | ||
7ffd92c5 AK |
1747 | fix_rmode_seg(VCPU_SREG_ES, &vmx->rmode.es); |
1748 | fix_rmode_seg(VCPU_SREG_DS, &vmx->rmode.ds); | |
1749 | fix_rmode_seg(VCPU_SREG_GS, &vmx->rmode.gs); | |
1750 | fix_rmode_seg(VCPU_SREG_FS, &vmx->rmode.fs); | |
75880a01 | 1751 | |
a89a8fb9 | 1752 | continue_rmode: |
8668a3c4 | 1753 | kvm_mmu_reset_context(vcpu); |
b7ebfb05 | 1754 | init_rmode(vcpu->kvm); |
6aa8b732 AK |
1755 | } |
1756 | ||
401d10de AS |
1757 | static void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer) |
1758 | { | |
1759 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
26bb0981 AK |
1760 | struct shared_msr_entry *msr = find_msr_entry(vmx, MSR_EFER); |
1761 | ||
1762 | if (!msr) | |
1763 | return; | |
401d10de | 1764 | |
44ea2b17 AK |
1765 | /* |
1766 | * Force kernel_gs_base reloading before EFER changes, as control | |
1767 | * of this msr depends on is_long_mode(). | |
1768 | */ | |
1769 | vmx_load_host_state(to_vmx(vcpu)); | |
f6801dff | 1770 | vcpu->arch.efer = efer; |
401d10de AS |
1771 | if (efer & EFER_LMA) { |
1772 | vmcs_write32(VM_ENTRY_CONTROLS, | |
1773 | vmcs_read32(VM_ENTRY_CONTROLS) | | |
1774 | VM_ENTRY_IA32E_MODE); | |
1775 | msr->data = efer; | |
1776 | } else { | |
1777 | vmcs_write32(VM_ENTRY_CONTROLS, | |
1778 | vmcs_read32(VM_ENTRY_CONTROLS) & | |
1779 | ~VM_ENTRY_IA32E_MODE); | |
1780 | ||
1781 | msr->data = efer & ~EFER_LME; | |
1782 | } | |
1783 | setup_msrs(vmx); | |
1784 | } | |
1785 | ||
05b3e0c2 | 1786 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
1787 | |
1788 | static void enter_lmode(struct kvm_vcpu *vcpu) | |
1789 | { | |
1790 | u32 guest_tr_ar; | |
1791 | ||
1792 | guest_tr_ar = vmcs_read32(GUEST_TR_AR_BYTES); | |
1793 | if ((guest_tr_ar & AR_TYPE_MASK) != AR_TYPE_BUSY_64_TSS) { | |
1794 | printk(KERN_DEBUG "%s: tss fixup for long mode. \n", | |
b8688d51 | 1795 | __func__); |
6aa8b732 AK |
1796 | vmcs_write32(GUEST_TR_AR_BYTES, |
1797 | (guest_tr_ar & ~AR_TYPE_MASK) | |
1798 | | AR_TYPE_BUSY_64_TSS); | |
1799 | } | |
da38f438 | 1800 | vmx_set_efer(vcpu, vcpu->arch.efer | EFER_LMA); |
6aa8b732 AK |
1801 | } |
1802 | ||
1803 | static void exit_lmode(struct kvm_vcpu *vcpu) | |
1804 | { | |
6aa8b732 AK |
1805 | vmcs_write32(VM_ENTRY_CONTROLS, |
1806 | vmcs_read32(VM_ENTRY_CONTROLS) | |
1e4e6e00 | 1807 | & ~VM_ENTRY_IA32E_MODE); |
da38f438 | 1808 | vmx_set_efer(vcpu, vcpu->arch.efer & ~EFER_LMA); |
6aa8b732 AK |
1809 | } |
1810 | ||
1811 | #endif | |
1812 | ||
2384d2b3 SY |
1813 | static void vmx_flush_tlb(struct kvm_vcpu *vcpu) |
1814 | { | |
b9d762fa | 1815 | vpid_sync_context(to_vmx(vcpu)); |
dd180b3e XG |
1816 | if (enable_ept) { |
1817 | if (!VALID_PAGE(vcpu->arch.mmu.root_hpa)) | |
1818 | return; | |
4e1096d2 | 1819 | ept_sync_context(construct_eptp(vcpu->arch.mmu.root_hpa)); |
dd180b3e | 1820 | } |
2384d2b3 SY |
1821 | } |
1822 | ||
e8467fda AK |
1823 | static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu) |
1824 | { | |
1825 | ulong cr0_guest_owned_bits = vcpu->arch.cr0_guest_owned_bits; | |
1826 | ||
1827 | vcpu->arch.cr0 &= ~cr0_guest_owned_bits; | |
1828 | vcpu->arch.cr0 |= vmcs_readl(GUEST_CR0) & cr0_guest_owned_bits; | |
1829 | } | |
1830 | ||
25c4c276 | 1831 | static void vmx_decache_cr4_guest_bits(struct kvm_vcpu *vcpu) |
399badf3 | 1832 | { |
fc78f519 AK |
1833 | ulong cr4_guest_owned_bits = vcpu->arch.cr4_guest_owned_bits; |
1834 | ||
1835 | vcpu->arch.cr4 &= ~cr4_guest_owned_bits; | |
1836 | vcpu->arch.cr4 |= vmcs_readl(GUEST_CR4) & cr4_guest_owned_bits; | |
399badf3 AK |
1837 | } |
1838 | ||
1439442c SY |
1839 | static void ept_load_pdptrs(struct kvm_vcpu *vcpu) |
1840 | { | |
6de4f3ad AK |
1841 | if (!test_bit(VCPU_EXREG_PDPTR, |
1842 | (unsigned long *)&vcpu->arch.regs_dirty)) | |
1843 | return; | |
1844 | ||
1439442c | 1845 | if (is_paging(vcpu) && is_pae(vcpu) && !is_long_mode(vcpu)) { |
ff03a073 JR |
1846 | vmcs_write64(GUEST_PDPTR0, vcpu->arch.mmu.pdptrs[0]); |
1847 | vmcs_write64(GUEST_PDPTR1, vcpu->arch.mmu.pdptrs[1]); | |
1848 | vmcs_write64(GUEST_PDPTR2, vcpu->arch.mmu.pdptrs[2]); | |
1849 | vmcs_write64(GUEST_PDPTR3, vcpu->arch.mmu.pdptrs[3]); | |
1439442c SY |
1850 | } |
1851 | } | |
1852 | ||
8f5d549f AK |
1853 | static void ept_save_pdptrs(struct kvm_vcpu *vcpu) |
1854 | { | |
1855 | if (is_paging(vcpu) && is_pae(vcpu) && !is_long_mode(vcpu)) { | |
ff03a073 JR |
1856 | vcpu->arch.mmu.pdptrs[0] = vmcs_read64(GUEST_PDPTR0); |
1857 | vcpu->arch.mmu.pdptrs[1] = vmcs_read64(GUEST_PDPTR1); | |
1858 | vcpu->arch.mmu.pdptrs[2] = vmcs_read64(GUEST_PDPTR2); | |
1859 | vcpu->arch.mmu.pdptrs[3] = vmcs_read64(GUEST_PDPTR3); | |
8f5d549f | 1860 | } |
6de4f3ad AK |
1861 | |
1862 | __set_bit(VCPU_EXREG_PDPTR, | |
1863 | (unsigned long *)&vcpu->arch.regs_avail); | |
1864 | __set_bit(VCPU_EXREG_PDPTR, | |
1865 | (unsigned long *)&vcpu->arch.regs_dirty); | |
8f5d549f AK |
1866 | } |
1867 | ||
1439442c SY |
1868 | static void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4); |
1869 | ||
1870 | static void ept_update_paging_mode_cr0(unsigned long *hw_cr0, | |
1871 | unsigned long cr0, | |
1872 | struct kvm_vcpu *vcpu) | |
1873 | { | |
1874 | if (!(cr0 & X86_CR0_PG)) { | |
1875 | /* From paging/starting to nonpaging */ | |
1876 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, | |
65267ea1 | 1877 | vmcs_read32(CPU_BASED_VM_EXEC_CONTROL) | |
1439442c SY |
1878 | (CPU_BASED_CR3_LOAD_EXITING | |
1879 | CPU_BASED_CR3_STORE_EXITING)); | |
1880 | vcpu->arch.cr0 = cr0; | |
fc78f519 | 1881 | vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
1439442c SY |
1882 | } else if (!is_paging(vcpu)) { |
1883 | /* From nonpaging to paging */ | |
1884 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, | |
65267ea1 | 1885 | vmcs_read32(CPU_BASED_VM_EXEC_CONTROL) & |
1439442c SY |
1886 | ~(CPU_BASED_CR3_LOAD_EXITING | |
1887 | CPU_BASED_CR3_STORE_EXITING)); | |
1888 | vcpu->arch.cr0 = cr0; | |
fc78f519 | 1889 | vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
1439442c | 1890 | } |
95eb84a7 SY |
1891 | |
1892 | if (!(cr0 & X86_CR0_WP)) | |
1893 | *hw_cr0 &= ~X86_CR0_WP; | |
1439442c SY |
1894 | } |
1895 | ||
6aa8b732 AK |
1896 | static void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) |
1897 | { | |
7ffd92c5 | 1898 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
3a624e29 NK |
1899 | unsigned long hw_cr0; |
1900 | ||
1901 | if (enable_unrestricted_guest) | |
1902 | hw_cr0 = (cr0 & ~KVM_GUEST_CR0_MASK_UNRESTRICTED_GUEST) | |
1903 | | KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST; | |
1904 | else | |
1905 | hw_cr0 = (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON; | |
1439442c | 1906 | |
7ffd92c5 | 1907 | if (vmx->rmode.vm86_active && (cr0 & X86_CR0_PE)) |
6aa8b732 AK |
1908 | enter_pmode(vcpu); |
1909 | ||
7ffd92c5 | 1910 | if (!vmx->rmode.vm86_active && !(cr0 & X86_CR0_PE)) |
6aa8b732 AK |
1911 | enter_rmode(vcpu); |
1912 | ||
05b3e0c2 | 1913 | #ifdef CONFIG_X86_64 |
f6801dff | 1914 | if (vcpu->arch.efer & EFER_LME) { |
707d92fa | 1915 | if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) |
6aa8b732 | 1916 | enter_lmode(vcpu); |
707d92fa | 1917 | if (is_paging(vcpu) && !(cr0 & X86_CR0_PG)) |
6aa8b732 AK |
1918 | exit_lmode(vcpu); |
1919 | } | |
1920 | #endif | |
1921 | ||
089d034e | 1922 | if (enable_ept) |
1439442c SY |
1923 | ept_update_paging_mode_cr0(&hw_cr0, cr0, vcpu); |
1924 | ||
02daab21 | 1925 | if (!vcpu->fpu_active) |
81231c69 | 1926 | hw_cr0 |= X86_CR0_TS | X86_CR0_MP; |
02daab21 | 1927 | |
6aa8b732 | 1928 | vmcs_writel(CR0_READ_SHADOW, cr0); |
1439442c | 1929 | vmcs_writel(GUEST_CR0, hw_cr0); |
ad312c7c | 1930 | vcpu->arch.cr0 = cr0; |
6aa8b732 AK |
1931 | } |
1932 | ||
1439442c SY |
1933 | static u64 construct_eptp(unsigned long root_hpa) |
1934 | { | |
1935 | u64 eptp; | |
1936 | ||
1937 | /* TODO write the value reading from MSR */ | |
1938 | eptp = VMX_EPT_DEFAULT_MT | | |
1939 | VMX_EPT_DEFAULT_GAW << VMX_EPT_GAW_EPTP_SHIFT; | |
1940 | eptp |= (root_hpa & PAGE_MASK); | |
1941 | ||
1942 | return eptp; | |
1943 | } | |
1944 | ||
6aa8b732 AK |
1945 | static void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) |
1946 | { | |
1439442c SY |
1947 | unsigned long guest_cr3; |
1948 | u64 eptp; | |
1949 | ||
1950 | guest_cr3 = cr3; | |
089d034e | 1951 | if (enable_ept) { |
1439442c SY |
1952 | eptp = construct_eptp(cr3); |
1953 | vmcs_write64(EPT_POINTER, eptp); | |
1439442c | 1954 | guest_cr3 = is_paging(vcpu) ? vcpu->arch.cr3 : |
b927a3ce | 1955 | vcpu->kvm->arch.ept_identity_map_addr; |
7c93be44 | 1956 | ept_load_pdptrs(vcpu); |
1439442c SY |
1957 | } |
1958 | ||
2384d2b3 | 1959 | vmx_flush_tlb(vcpu); |
1439442c | 1960 | vmcs_writel(GUEST_CR3, guest_cr3); |
6aa8b732 AK |
1961 | } |
1962 | ||
1963 | static void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) | |
1964 | { | |
7ffd92c5 | 1965 | unsigned long hw_cr4 = cr4 | (to_vmx(vcpu)->rmode.vm86_active ? |
1439442c SY |
1966 | KVM_RMODE_VM_CR4_ALWAYS_ON : KVM_PMODE_VM_CR4_ALWAYS_ON); |
1967 | ||
ad312c7c | 1968 | vcpu->arch.cr4 = cr4; |
bc23008b AK |
1969 | if (enable_ept) { |
1970 | if (!is_paging(vcpu)) { | |
1971 | hw_cr4 &= ~X86_CR4_PAE; | |
1972 | hw_cr4 |= X86_CR4_PSE; | |
1973 | } else if (!(cr4 & X86_CR4_PAE)) { | |
1974 | hw_cr4 &= ~X86_CR4_PAE; | |
1975 | } | |
1976 | } | |
1439442c SY |
1977 | |
1978 | vmcs_writel(CR4_READ_SHADOW, cr4); | |
1979 | vmcs_writel(GUEST_CR4, hw_cr4); | |
6aa8b732 AK |
1980 | } |
1981 | ||
6aa8b732 AK |
1982 | static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg) |
1983 | { | |
1984 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
1985 | ||
1986 | return vmcs_readl(sf->base); | |
1987 | } | |
1988 | ||
1989 | static void vmx_get_segment(struct kvm_vcpu *vcpu, | |
1990 | struct kvm_segment *var, int seg) | |
1991 | { | |
1992 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
1993 | u32 ar; | |
1994 | ||
1995 | var->base = vmcs_readl(sf->base); | |
1996 | var->limit = vmcs_read32(sf->limit); | |
1997 | var->selector = vmcs_read16(sf->selector); | |
1998 | ar = vmcs_read32(sf->ar_bytes); | |
9fd4a3b7 | 1999 | if ((ar & AR_UNUSABLE_MASK) && !emulate_invalid_guest_state) |
6aa8b732 AK |
2000 | ar = 0; |
2001 | var->type = ar & 15; | |
2002 | var->s = (ar >> 4) & 1; | |
2003 | var->dpl = (ar >> 5) & 3; | |
2004 | var->present = (ar >> 7) & 1; | |
2005 | var->avl = (ar >> 12) & 1; | |
2006 | var->l = (ar >> 13) & 1; | |
2007 | var->db = (ar >> 14) & 1; | |
2008 | var->g = (ar >> 15) & 1; | |
2009 | var->unusable = (ar >> 16) & 1; | |
2010 | } | |
2011 | ||
2e4d2653 IE |
2012 | static int vmx_get_cpl(struct kvm_vcpu *vcpu) |
2013 | { | |
3eeb3288 | 2014 | if (!is_protmode(vcpu)) |
2e4d2653 IE |
2015 | return 0; |
2016 | ||
2017 | if (vmx_get_rflags(vcpu) & X86_EFLAGS_VM) /* if virtual 8086 */ | |
2018 | return 3; | |
2019 | ||
eab4b8aa | 2020 | return vmcs_read16(GUEST_CS_SELECTOR) & 3; |
2e4d2653 IE |
2021 | } |
2022 | ||
653e3108 | 2023 | static u32 vmx_segment_access_rights(struct kvm_segment *var) |
6aa8b732 | 2024 | { |
6aa8b732 AK |
2025 | u32 ar; |
2026 | ||
653e3108 | 2027 | if (var->unusable) |
6aa8b732 AK |
2028 | ar = 1 << 16; |
2029 | else { | |
2030 | ar = var->type & 15; | |
2031 | ar |= (var->s & 1) << 4; | |
2032 | ar |= (var->dpl & 3) << 5; | |
2033 | ar |= (var->present & 1) << 7; | |
2034 | ar |= (var->avl & 1) << 12; | |
2035 | ar |= (var->l & 1) << 13; | |
2036 | ar |= (var->db & 1) << 14; | |
2037 | ar |= (var->g & 1) << 15; | |
2038 | } | |
f7fbf1fd UL |
2039 | if (ar == 0) /* a 0 value means unusable */ |
2040 | ar = AR_UNUSABLE_MASK; | |
653e3108 AK |
2041 | |
2042 | return ar; | |
2043 | } | |
2044 | ||
2045 | static void vmx_set_segment(struct kvm_vcpu *vcpu, | |
2046 | struct kvm_segment *var, int seg) | |
2047 | { | |
7ffd92c5 | 2048 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
653e3108 AK |
2049 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
2050 | u32 ar; | |
2051 | ||
7ffd92c5 AK |
2052 | if (vmx->rmode.vm86_active && seg == VCPU_SREG_TR) { |
2053 | vmx->rmode.tr.selector = var->selector; | |
2054 | vmx->rmode.tr.base = var->base; | |
2055 | vmx->rmode.tr.limit = var->limit; | |
2056 | vmx->rmode.tr.ar = vmx_segment_access_rights(var); | |
653e3108 AK |
2057 | return; |
2058 | } | |
2059 | vmcs_writel(sf->base, var->base); | |
2060 | vmcs_write32(sf->limit, var->limit); | |
2061 | vmcs_write16(sf->selector, var->selector); | |
7ffd92c5 | 2062 | if (vmx->rmode.vm86_active && var->s) { |
653e3108 AK |
2063 | /* |
2064 | * Hack real-mode segments into vm86 compatibility. | |
2065 | */ | |
2066 | if (var->base == 0xffff0000 && var->selector == 0xf000) | |
2067 | vmcs_writel(sf->base, 0xf0000); | |
2068 | ar = 0xf3; | |
2069 | } else | |
2070 | ar = vmx_segment_access_rights(var); | |
3a624e29 NK |
2071 | |
2072 | /* | |
2073 | * Fix the "Accessed" bit in AR field of segment registers for older | |
2074 | * qemu binaries. | |
2075 | * IA32 arch specifies that at the time of processor reset the | |
2076 | * "Accessed" bit in the AR field of segment registers is 1. And qemu | |
2077 | * is setting it to 0 in the usedland code. This causes invalid guest | |
2078 | * state vmexit when "unrestricted guest" mode is turned on. | |
2079 | * Fix for this setup issue in cpu_reset is being pushed in the qemu | |
2080 | * tree. Newer qemu binaries with that qemu fix would not need this | |
2081 | * kvm hack. | |
2082 | */ | |
2083 | if (enable_unrestricted_guest && (seg != VCPU_SREG_LDTR)) | |
2084 | ar |= 0x1; /* Accessed */ | |
2085 | ||
6aa8b732 AK |
2086 | vmcs_write32(sf->ar_bytes, ar); |
2087 | } | |
2088 | ||
6aa8b732 AK |
2089 | static void vmx_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l) |
2090 | { | |
2091 | u32 ar = vmcs_read32(GUEST_CS_AR_BYTES); | |
2092 | ||
2093 | *db = (ar >> 14) & 1; | |
2094 | *l = (ar >> 13) & 1; | |
2095 | } | |
2096 | ||
89a27f4d | 2097 | static void vmx_get_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 2098 | { |
89a27f4d GN |
2099 | dt->size = vmcs_read32(GUEST_IDTR_LIMIT); |
2100 | dt->address = vmcs_readl(GUEST_IDTR_BASE); | |
6aa8b732 AK |
2101 | } |
2102 | ||
89a27f4d | 2103 | static void vmx_set_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 2104 | { |
89a27f4d GN |
2105 | vmcs_write32(GUEST_IDTR_LIMIT, dt->size); |
2106 | vmcs_writel(GUEST_IDTR_BASE, dt->address); | |
6aa8b732 AK |
2107 | } |
2108 | ||
89a27f4d | 2109 | static void vmx_get_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 2110 | { |
89a27f4d GN |
2111 | dt->size = vmcs_read32(GUEST_GDTR_LIMIT); |
2112 | dt->address = vmcs_readl(GUEST_GDTR_BASE); | |
6aa8b732 AK |
2113 | } |
2114 | ||
89a27f4d | 2115 | static void vmx_set_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 2116 | { |
89a27f4d GN |
2117 | vmcs_write32(GUEST_GDTR_LIMIT, dt->size); |
2118 | vmcs_writel(GUEST_GDTR_BASE, dt->address); | |
6aa8b732 AK |
2119 | } |
2120 | ||
648dfaa7 MG |
2121 | static bool rmode_segment_valid(struct kvm_vcpu *vcpu, int seg) |
2122 | { | |
2123 | struct kvm_segment var; | |
2124 | u32 ar; | |
2125 | ||
2126 | vmx_get_segment(vcpu, &var, seg); | |
2127 | ar = vmx_segment_access_rights(&var); | |
2128 | ||
2129 | if (var.base != (var.selector << 4)) | |
2130 | return false; | |
2131 | if (var.limit != 0xffff) | |
2132 | return false; | |
2133 | if (ar != 0xf3) | |
2134 | return false; | |
2135 | ||
2136 | return true; | |
2137 | } | |
2138 | ||
2139 | static bool code_segment_valid(struct kvm_vcpu *vcpu) | |
2140 | { | |
2141 | struct kvm_segment cs; | |
2142 | unsigned int cs_rpl; | |
2143 | ||
2144 | vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); | |
2145 | cs_rpl = cs.selector & SELECTOR_RPL_MASK; | |
2146 | ||
1872a3f4 AK |
2147 | if (cs.unusable) |
2148 | return false; | |
648dfaa7 MG |
2149 | if (~cs.type & (AR_TYPE_CODE_MASK|AR_TYPE_ACCESSES_MASK)) |
2150 | return false; | |
2151 | if (!cs.s) | |
2152 | return false; | |
1872a3f4 | 2153 | if (cs.type & AR_TYPE_WRITEABLE_MASK) { |
648dfaa7 MG |
2154 | if (cs.dpl > cs_rpl) |
2155 | return false; | |
1872a3f4 | 2156 | } else { |
648dfaa7 MG |
2157 | if (cs.dpl != cs_rpl) |
2158 | return false; | |
2159 | } | |
2160 | if (!cs.present) | |
2161 | return false; | |
2162 | ||
2163 | /* TODO: Add Reserved field check, this'll require a new member in the kvm_segment_field structure */ | |
2164 | return true; | |
2165 | } | |
2166 | ||
2167 | static bool stack_segment_valid(struct kvm_vcpu *vcpu) | |
2168 | { | |
2169 | struct kvm_segment ss; | |
2170 | unsigned int ss_rpl; | |
2171 | ||
2172 | vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); | |
2173 | ss_rpl = ss.selector & SELECTOR_RPL_MASK; | |
2174 | ||
1872a3f4 AK |
2175 | if (ss.unusable) |
2176 | return true; | |
2177 | if (ss.type != 3 && ss.type != 7) | |
648dfaa7 MG |
2178 | return false; |
2179 | if (!ss.s) | |
2180 | return false; | |
2181 | if (ss.dpl != ss_rpl) /* DPL != RPL */ | |
2182 | return false; | |
2183 | if (!ss.present) | |
2184 | return false; | |
2185 | ||
2186 | return true; | |
2187 | } | |
2188 | ||
2189 | static bool data_segment_valid(struct kvm_vcpu *vcpu, int seg) | |
2190 | { | |
2191 | struct kvm_segment var; | |
2192 | unsigned int rpl; | |
2193 | ||
2194 | vmx_get_segment(vcpu, &var, seg); | |
2195 | rpl = var.selector & SELECTOR_RPL_MASK; | |
2196 | ||
1872a3f4 AK |
2197 | if (var.unusable) |
2198 | return true; | |
648dfaa7 MG |
2199 | if (!var.s) |
2200 | return false; | |
2201 | if (!var.present) | |
2202 | return false; | |
2203 | if (~var.type & (AR_TYPE_CODE_MASK|AR_TYPE_WRITEABLE_MASK)) { | |
2204 | if (var.dpl < rpl) /* DPL < RPL */ | |
2205 | return false; | |
2206 | } | |
2207 | ||
2208 | /* TODO: Add other members to kvm_segment_field to allow checking for other access | |
2209 | * rights flags | |
2210 | */ | |
2211 | return true; | |
2212 | } | |
2213 | ||
2214 | static bool tr_valid(struct kvm_vcpu *vcpu) | |
2215 | { | |
2216 | struct kvm_segment tr; | |
2217 | ||
2218 | vmx_get_segment(vcpu, &tr, VCPU_SREG_TR); | |
2219 | ||
1872a3f4 AK |
2220 | if (tr.unusable) |
2221 | return false; | |
648dfaa7 MG |
2222 | if (tr.selector & SELECTOR_TI_MASK) /* TI = 1 */ |
2223 | return false; | |
1872a3f4 | 2224 | if (tr.type != 3 && tr.type != 11) /* TODO: Check if guest is in IA32e mode */ |
648dfaa7 MG |
2225 | return false; |
2226 | if (!tr.present) | |
2227 | return false; | |
2228 | ||
2229 | return true; | |
2230 | } | |
2231 | ||
2232 | static bool ldtr_valid(struct kvm_vcpu *vcpu) | |
2233 | { | |
2234 | struct kvm_segment ldtr; | |
2235 | ||
2236 | vmx_get_segment(vcpu, &ldtr, VCPU_SREG_LDTR); | |
2237 | ||
1872a3f4 AK |
2238 | if (ldtr.unusable) |
2239 | return true; | |
648dfaa7 MG |
2240 | if (ldtr.selector & SELECTOR_TI_MASK) /* TI = 1 */ |
2241 | return false; | |
2242 | if (ldtr.type != 2) | |
2243 | return false; | |
2244 | if (!ldtr.present) | |
2245 | return false; | |
2246 | ||
2247 | return true; | |
2248 | } | |
2249 | ||
2250 | static bool cs_ss_rpl_check(struct kvm_vcpu *vcpu) | |
2251 | { | |
2252 | struct kvm_segment cs, ss; | |
2253 | ||
2254 | vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); | |
2255 | vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); | |
2256 | ||
2257 | return ((cs.selector & SELECTOR_RPL_MASK) == | |
2258 | (ss.selector & SELECTOR_RPL_MASK)); | |
2259 | } | |
2260 | ||
2261 | /* | |
2262 | * Check if guest state is valid. Returns true if valid, false if | |
2263 | * not. | |
2264 | * We assume that registers are always usable | |
2265 | */ | |
2266 | static bool guest_state_valid(struct kvm_vcpu *vcpu) | |
2267 | { | |
2268 | /* real mode guest state checks */ | |
3eeb3288 | 2269 | if (!is_protmode(vcpu)) { |
648dfaa7 MG |
2270 | if (!rmode_segment_valid(vcpu, VCPU_SREG_CS)) |
2271 | return false; | |
2272 | if (!rmode_segment_valid(vcpu, VCPU_SREG_SS)) | |
2273 | return false; | |
2274 | if (!rmode_segment_valid(vcpu, VCPU_SREG_DS)) | |
2275 | return false; | |
2276 | if (!rmode_segment_valid(vcpu, VCPU_SREG_ES)) | |
2277 | return false; | |
2278 | if (!rmode_segment_valid(vcpu, VCPU_SREG_FS)) | |
2279 | return false; | |
2280 | if (!rmode_segment_valid(vcpu, VCPU_SREG_GS)) | |
2281 | return false; | |
2282 | } else { | |
2283 | /* protected mode guest state checks */ | |
2284 | if (!cs_ss_rpl_check(vcpu)) | |
2285 | return false; | |
2286 | if (!code_segment_valid(vcpu)) | |
2287 | return false; | |
2288 | if (!stack_segment_valid(vcpu)) | |
2289 | return false; | |
2290 | if (!data_segment_valid(vcpu, VCPU_SREG_DS)) | |
2291 | return false; | |
2292 | if (!data_segment_valid(vcpu, VCPU_SREG_ES)) | |
2293 | return false; | |
2294 | if (!data_segment_valid(vcpu, VCPU_SREG_FS)) | |
2295 | return false; | |
2296 | if (!data_segment_valid(vcpu, VCPU_SREG_GS)) | |
2297 | return false; | |
2298 | if (!tr_valid(vcpu)) | |
2299 | return false; | |
2300 | if (!ldtr_valid(vcpu)) | |
2301 | return false; | |
2302 | } | |
2303 | /* TODO: | |
2304 | * - Add checks on RIP | |
2305 | * - Add checks on RFLAGS | |
2306 | */ | |
2307 | ||
2308 | return true; | |
2309 | } | |
2310 | ||
d77c26fc | 2311 | static int init_rmode_tss(struct kvm *kvm) |
6aa8b732 | 2312 | { |
6aa8b732 | 2313 | gfn_t fn = rmode_tss_base(kvm) >> PAGE_SHIFT; |
195aefde | 2314 | u16 data = 0; |
10589a46 | 2315 | int ret = 0; |
195aefde | 2316 | int r; |
6aa8b732 | 2317 | |
195aefde IE |
2318 | r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
2319 | if (r < 0) | |
10589a46 | 2320 | goto out; |
195aefde | 2321 | data = TSS_BASE_SIZE + TSS_REDIRECTION_SIZE; |
464d17c8 SY |
2322 | r = kvm_write_guest_page(kvm, fn++, &data, |
2323 | TSS_IOPB_BASE_OFFSET, sizeof(u16)); | |
195aefde | 2324 | if (r < 0) |
10589a46 | 2325 | goto out; |
195aefde IE |
2326 | r = kvm_clear_guest_page(kvm, fn++, 0, PAGE_SIZE); |
2327 | if (r < 0) | |
10589a46 | 2328 | goto out; |
195aefde IE |
2329 | r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
2330 | if (r < 0) | |
10589a46 | 2331 | goto out; |
195aefde | 2332 | data = ~0; |
10589a46 MT |
2333 | r = kvm_write_guest_page(kvm, fn, &data, |
2334 | RMODE_TSS_SIZE - 2 * PAGE_SIZE - 1, | |
2335 | sizeof(u8)); | |
195aefde | 2336 | if (r < 0) |
10589a46 MT |
2337 | goto out; |
2338 | ||
2339 | ret = 1; | |
2340 | out: | |
10589a46 | 2341 | return ret; |
6aa8b732 AK |
2342 | } |
2343 | ||
b7ebfb05 SY |
2344 | static int init_rmode_identity_map(struct kvm *kvm) |
2345 | { | |
2346 | int i, r, ret; | |
2347 | pfn_t identity_map_pfn; | |
2348 | u32 tmp; | |
2349 | ||
089d034e | 2350 | if (!enable_ept) |
b7ebfb05 SY |
2351 | return 1; |
2352 | if (unlikely(!kvm->arch.ept_identity_pagetable)) { | |
2353 | printk(KERN_ERR "EPT: identity-mapping pagetable " | |
2354 | "haven't been allocated!\n"); | |
2355 | return 0; | |
2356 | } | |
2357 | if (likely(kvm->arch.ept_identity_pagetable_done)) | |
2358 | return 1; | |
2359 | ret = 0; | |
b927a3ce | 2360 | identity_map_pfn = kvm->arch.ept_identity_map_addr >> PAGE_SHIFT; |
b7ebfb05 SY |
2361 | r = kvm_clear_guest_page(kvm, identity_map_pfn, 0, PAGE_SIZE); |
2362 | if (r < 0) | |
2363 | goto out; | |
2364 | /* Set up identity-mapping pagetable for EPT in real mode */ | |
2365 | for (i = 0; i < PT32_ENT_PER_PAGE; i++) { | |
2366 | tmp = (i << 22) + (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | | |
2367 | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_PSE); | |
2368 | r = kvm_write_guest_page(kvm, identity_map_pfn, | |
2369 | &tmp, i * sizeof(tmp), sizeof(tmp)); | |
2370 | if (r < 0) | |
2371 | goto out; | |
2372 | } | |
2373 | kvm->arch.ept_identity_pagetable_done = true; | |
2374 | ret = 1; | |
2375 | out: | |
2376 | return ret; | |
2377 | } | |
2378 | ||
6aa8b732 AK |
2379 | static void seg_setup(int seg) |
2380 | { | |
2381 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
3a624e29 | 2382 | unsigned int ar; |
6aa8b732 AK |
2383 | |
2384 | vmcs_write16(sf->selector, 0); | |
2385 | vmcs_writel(sf->base, 0); | |
2386 | vmcs_write32(sf->limit, 0xffff); | |
3a624e29 NK |
2387 | if (enable_unrestricted_guest) { |
2388 | ar = 0x93; | |
2389 | if (seg == VCPU_SREG_CS) | |
2390 | ar |= 0x08; /* code segment */ | |
2391 | } else | |
2392 | ar = 0xf3; | |
2393 | ||
2394 | vmcs_write32(sf->ar_bytes, ar); | |
6aa8b732 AK |
2395 | } |
2396 | ||
f78e0e2e SY |
2397 | static int alloc_apic_access_page(struct kvm *kvm) |
2398 | { | |
2399 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2400 | int r = 0; | |
2401 | ||
79fac95e | 2402 | mutex_lock(&kvm->slots_lock); |
bfc6d222 | 2403 | if (kvm->arch.apic_access_page) |
f78e0e2e SY |
2404 | goto out; |
2405 | kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT; | |
2406 | kvm_userspace_mem.flags = 0; | |
2407 | kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL; | |
2408 | kvm_userspace_mem.memory_size = PAGE_SIZE; | |
2409 | r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0); | |
2410 | if (r) | |
2411 | goto out; | |
72dc67a6 | 2412 | |
bfc6d222 | 2413 | kvm->arch.apic_access_page = gfn_to_page(kvm, 0xfee00); |
f78e0e2e | 2414 | out: |
79fac95e | 2415 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
2416 | return r; |
2417 | } | |
2418 | ||
b7ebfb05 SY |
2419 | static int alloc_identity_pagetable(struct kvm *kvm) |
2420 | { | |
2421 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2422 | int r = 0; | |
2423 | ||
79fac95e | 2424 | mutex_lock(&kvm->slots_lock); |
b7ebfb05 SY |
2425 | if (kvm->arch.ept_identity_pagetable) |
2426 | goto out; | |
2427 | kvm_userspace_mem.slot = IDENTITY_PAGETABLE_PRIVATE_MEMSLOT; | |
2428 | kvm_userspace_mem.flags = 0; | |
b927a3ce SY |
2429 | kvm_userspace_mem.guest_phys_addr = |
2430 | kvm->arch.ept_identity_map_addr; | |
b7ebfb05 SY |
2431 | kvm_userspace_mem.memory_size = PAGE_SIZE; |
2432 | r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0); | |
2433 | if (r) | |
2434 | goto out; | |
2435 | ||
b7ebfb05 | 2436 | kvm->arch.ept_identity_pagetable = gfn_to_page(kvm, |
b927a3ce | 2437 | kvm->arch.ept_identity_map_addr >> PAGE_SHIFT); |
b7ebfb05 | 2438 | out: |
79fac95e | 2439 | mutex_unlock(&kvm->slots_lock); |
b7ebfb05 SY |
2440 | return r; |
2441 | } | |
2442 | ||
2384d2b3 SY |
2443 | static void allocate_vpid(struct vcpu_vmx *vmx) |
2444 | { | |
2445 | int vpid; | |
2446 | ||
2447 | vmx->vpid = 0; | |
919818ab | 2448 | if (!enable_vpid) |
2384d2b3 SY |
2449 | return; |
2450 | spin_lock(&vmx_vpid_lock); | |
2451 | vpid = find_first_zero_bit(vmx_vpid_bitmap, VMX_NR_VPIDS); | |
2452 | if (vpid < VMX_NR_VPIDS) { | |
2453 | vmx->vpid = vpid; | |
2454 | __set_bit(vpid, vmx_vpid_bitmap); | |
2455 | } | |
2456 | spin_unlock(&vmx_vpid_lock); | |
2457 | } | |
2458 | ||
cdbecfc3 LJ |
2459 | static void free_vpid(struct vcpu_vmx *vmx) |
2460 | { | |
2461 | if (!enable_vpid) | |
2462 | return; | |
2463 | spin_lock(&vmx_vpid_lock); | |
2464 | if (vmx->vpid != 0) | |
2465 | __clear_bit(vmx->vpid, vmx_vpid_bitmap); | |
2466 | spin_unlock(&vmx_vpid_lock); | |
2467 | } | |
2468 | ||
5897297b | 2469 | static void __vmx_disable_intercept_for_msr(unsigned long *msr_bitmap, u32 msr) |
25c5f225 | 2470 | { |
3e7c73e9 | 2471 | int f = sizeof(unsigned long); |
25c5f225 SY |
2472 | |
2473 | if (!cpu_has_vmx_msr_bitmap()) | |
2474 | return; | |
2475 | ||
2476 | /* | |
2477 | * See Intel PRM Vol. 3, 20.6.9 (MSR-Bitmap Address). Early manuals | |
2478 | * have the write-low and read-high bitmap offsets the wrong way round. | |
2479 | * We can control MSRs 0x00000000-0x00001fff and 0xc0000000-0xc0001fff. | |
2480 | */ | |
25c5f225 | 2481 | if (msr <= 0x1fff) { |
3e7c73e9 AK |
2482 | __clear_bit(msr, msr_bitmap + 0x000 / f); /* read-low */ |
2483 | __clear_bit(msr, msr_bitmap + 0x800 / f); /* write-low */ | |
25c5f225 SY |
2484 | } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
2485 | msr &= 0x1fff; | |
3e7c73e9 AK |
2486 | __clear_bit(msr, msr_bitmap + 0x400 / f); /* read-high */ |
2487 | __clear_bit(msr, msr_bitmap + 0xc00 / f); /* write-high */ | |
25c5f225 | 2488 | } |
25c5f225 SY |
2489 | } |
2490 | ||
5897297b AK |
2491 | static void vmx_disable_intercept_for_msr(u32 msr, bool longmode_only) |
2492 | { | |
2493 | if (!longmode_only) | |
2494 | __vmx_disable_intercept_for_msr(vmx_msr_bitmap_legacy, msr); | |
2495 | __vmx_disable_intercept_for_msr(vmx_msr_bitmap_longmode, msr); | |
2496 | } | |
2497 | ||
6aa8b732 AK |
2498 | /* |
2499 | * Sets up the vmcs for emulated real mode. | |
2500 | */ | |
8b9cf98c | 2501 | static int vmx_vcpu_setup(struct vcpu_vmx *vmx) |
6aa8b732 | 2502 | { |
468d472f | 2503 | u32 host_sysenter_cs, msr_low, msr_high; |
6aa8b732 | 2504 | u32 junk; |
f4e1b3c8 | 2505 | u64 host_pat; |
6aa8b732 | 2506 | unsigned long a; |
89a27f4d | 2507 | struct desc_ptr dt; |
6aa8b732 | 2508 | int i; |
cd2276a7 | 2509 | unsigned long kvm_vmx_return; |
6e5d865c | 2510 | u32 exec_control; |
6aa8b732 | 2511 | |
6aa8b732 | 2512 | /* I/O */ |
3e7c73e9 AK |
2513 | vmcs_write64(IO_BITMAP_A, __pa(vmx_io_bitmap_a)); |
2514 | vmcs_write64(IO_BITMAP_B, __pa(vmx_io_bitmap_b)); | |
6aa8b732 | 2515 | |
25c5f225 | 2516 | if (cpu_has_vmx_msr_bitmap()) |
5897297b | 2517 | vmcs_write64(MSR_BITMAP, __pa(vmx_msr_bitmap_legacy)); |
25c5f225 | 2518 | |
6aa8b732 AK |
2519 | vmcs_write64(VMCS_LINK_POINTER, -1ull); /* 22.3.1.5 */ |
2520 | ||
6aa8b732 | 2521 | /* Control */ |
1c3d14fe YS |
2522 | vmcs_write32(PIN_BASED_VM_EXEC_CONTROL, |
2523 | vmcs_config.pin_based_exec_ctrl); | |
6e5d865c YS |
2524 | |
2525 | exec_control = vmcs_config.cpu_based_exec_ctrl; | |
2526 | if (!vm_need_tpr_shadow(vmx->vcpu.kvm)) { | |
2527 | exec_control &= ~CPU_BASED_TPR_SHADOW; | |
2528 | #ifdef CONFIG_X86_64 | |
2529 | exec_control |= CPU_BASED_CR8_STORE_EXITING | | |
2530 | CPU_BASED_CR8_LOAD_EXITING; | |
2531 | #endif | |
2532 | } | |
089d034e | 2533 | if (!enable_ept) |
d56f546d | 2534 | exec_control |= CPU_BASED_CR3_STORE_EXITING | |
83dbc83a MT |
2535 | CPU_BASED_CR3_LOAD_EXITING | |
2536 | CPU_BASED_INVLPG_EXITING; | |
6e5d865c | 2537 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, exec_control); |
6aa8b732 | 2538 | |
83ff3b9d SY |
2539 | if (cpu_has_secondary_exec_ctrls()) { |
2540 | exec_control = vmcs_config.cpu_based_2nd_exec_ctrl; | |
2541 | if (!vm_need_virtualize_apic_accesses(vmx->vcpu.kvm)) | |
2542 | exec_control &= | |
2543 | ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
2384d2b3 SY |
2544 | if (vmx->vpid == 0) |
2545 | exec_control &= ~SECONDARY_EXEC_ENABLE_VPID; | |
046d8710 | 2546 | if (!enable_ept) { |
d56f546d | 2547 | exec_control &= ~SECONDARY_EXEC_ENABLE_EPT; |
046d8710 SY |
2548 | enable_unrestricted_guest = 0; |
2549 | } | |
3a624e29 NK |
2550 | if (!enable_unrestricted_guest) |
2551 | exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST; | |
4b8d54f9 ZE |
2552 | if (!ple_gap) |
2553 | exec_control &= ~SECONDARY_EXEC_PAUSE_LOOP_EXITING; | |
83ff3b9d SY |
2554 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, exec_control); |
2555 | } | |
f78e0e2e | 2556 | |
4b8d54f9 ZE |
2557 | if (ple_gap) { |
2558 | vmcs_write32(PLE_GAP, ple_gap); | |
2559 | vmcs_write32(PLE_WINDOW, ple_window); | |
2560 | } | |
2561 | ||
c7addb90 AK |
2562 | vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, !!bypass_guest_pf); |
2563 | vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, !!bypass_guest_pf); | |
6aa8b732 AK |
2564 | vmcs_write32(CR3_TARGET_COUNT, 0); /* 22.2.1 */ |
2565 | ||
1c11e713 | 2566 | vmcs_writel(HOST_CR0, read_cr0() | X86_CR0_TS); /* 22.2.3 */ |
6aa8b732 AK |
2567 | vmcs_writel(HOST_CR4, read_cr4()); /* 22.2.3, 22.2.5 */ |
2568 | vmcs_writel(HOST_CR3, read_cr3()); /* 22.2.3 FIXME: shadow tables */ | |
2569 | ||
2570 | vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS); /* 22.2.4 */ | |
2571 | vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ | |
2572 | vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS); /* 22.2.4 */ | |
9581d442 AK |
2573 | vmcs_write16(HOST_FS_SELECTOR, 0); /* 22.2.4 */ |
2574 | vmcs_write16(HOST_GS_SELECTOR, 0); /* 22.2.4 */ | |
6aa8b732 | 2575 | vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
05b3e0c2 | 2576 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2577 | rdmsrl(MSR_FS_BASE, a); |
2578 | vmcs_writel(HOST_FS_BASE, a); /* 22.2.4 */ | |
2579 | rdmsrl(MSR_GS_BASE, a); | |
2580 | vmcs_writel(HOST_GS_BASE, a); /* 22.2.4 */ | |
2581 | #else | |
2582 | vmcs_writel(HOST_FS_BASE, 0); /* 22.2.4 */ | |
2583 | vmcs_writel(HOST_GS_BASE, 0); /* 22.2.4 */ | |
2584 | #endif | |
2585 | ||
2586 | vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8); /* 22.2.4 */ | |
2587 | ||
ec68798c | 2588 | native_store_idt(&dt); |
89a27f4d | 2589 | vmcs_writel(HOST_IDTR_BASE, dt.address); /* 22.2.4 */ |
6aa8b732 | 2590 | |
d77c26fc | 2591 | asm("mov $.Lkvm_vmx_return, %0" : "=r"(kvm_vmx_return)); |
cd2276a7 | 2592 | vmcs_writel(HOST_RIP, kvm_vmx_return); /* 22.2.5 */ |
2cc51560 ED |
2593 | vmcs_write32(VM_EXIT_MSR_STORE_COUNT, 0); |
2594 | vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, 0); | |
61d2ef2c | 2595 | vmcs_write64(VM_EXIT_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.host)); |
2cc51560 | 2596 | vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, 0); |
61d2ef2c | 2597 | vmcs_write64(VM_ENTRY_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.guest)); |
6aa8b732 AK |
2598 | |
2599 | rdmsr(MSR_IA32_SYSENTER_CS, host_sysenter_cs, junk); | |
2600 | vmcs_write32(HOST_IA32_SYSENTER_CS, host_sysenter_cs); | |
2601 | rdmsrl(MSR_IA32_SYSENTER_ESP, a); | |
2602 | vmcs_writel(HOST_IA32_SYSENTER_ESP, a); /* 22.2.3 */ | |
2603 | rdmsrl(MSR_IA32_SYSENTER_EIP, a); | |
2604 | vmcs_writel(HOST_IA32_SYSENTER_EIP, a); /* 22.2.3 */ | |
2605 | ||
468d472f SY |
2606 | if (vmcs_config.vmexit_ctrl & VM_EXIT_LOAD_IA32_PAT) { |
2607 | rdmsr(MSR_IA32_CR_PAT, msr_low, msr_high); | |
2608 | host_pat = msr_low | ((u64) msr_high << 32); | |
2609 | vmcs_write64(HOST_IA32_PAT, host_pat); | |
2610 | } | |
2611 | if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) { | |
2612 | rdmsr(MSR_IA32_CR_PAT, msr_low, msr_high); | |
2613 | host_pat = msr_low | ((u64) msr_high << 32); | |
2614 | /* Write the default value follow host pat */ | |
2615 | vmcs_write64(GUEST_IA32_PAT, host_pat); | |
2616 | /* Keep arch.pat sync with GUEST_IA32_PAT */ | |
2617 | vmx->vcpu.arch.pat = host_pat; | |
2618 | } | |
2619 | ||
6aa8b732 AK |
2620 | for (i = 0; i < NR_VMX_MSR; ++i) { |
2621 | u32 index = vmx_msr_index[i]; | |
2622 | u32 data_low, data_high; | |
a2fa3e9f | 2623 | int j = vmx->nmsrs; |
6aa8b732 AK |
2624 | |
2625 | if (rdmsr_safe(index, &data_low, &data_high) < 0) | |
2626 | continue; | |
432bd6cb AK |
2627 | if (wrmsr_safe(index, data_low, data_high) < 0) |
2628 | continue; | |
26bb0981 AK |
2629 | vmx->guest_msrs[j].index = i; |
2630 | vmx->guest_msrs[j].data = 0; | |
d5696725 | 2631 | vmx->guest_msrs[j].mask = -1ull; |
a2fa3e9f | 2632 | ++vmx->nmsrs; |
6aa8b732 | 2633 | } |
6aa8b732 | 2634 | |
1c3d14fe | 2635 | vmcs_write32(VM_EXIT_CONTROLS, vmcs_config.vmexit_ctrl); |
6aa8b732 AK |
2636 | |
2637 | /* 22.2.1, 20.8.1 */ | |
1c3d14fe YS |
2638 | vmcs_write32(VM_ENTRY_CONTROLS, vmcs_config.vmentry_ctrl); |
2639 | ||
e00c8cf2 | 2640 | vmcs_writel(CR0_GUEST_HOST_MASK, ~0UL); |
4c38609a | 2641 | vmx->vcpu.arch.cr4_guest_owned_bits = KVM_CR4_GUEST_OWNED_BITS; |
ce03e4f2 AK |
2642 | if (enable_ept) |
2643 | vmx->vcpu.arch.cr4_guest_owned_bits |= X86_CR4_PGE; | |
4c38609a | 2644 | vmcs_writel(CR4_GUEST_HOST_MASK, ~vmx->vcpu.arch.cr4_guest_owned_bits); |
e00c8cf2 | 2645 | |
99e3e30a | 2646 | kvm_write_tsc(&vmx->vcpu, 0); |
f78e0e2e | 2647 | |
e00c8cf2 AK |
2648 | return 0; |
2649 | } | |
2650 | ||
b7ebfb05 SY |
2651 | static int init_rmode(struct kvm *kvm) |
2652 | { | |
4b9d3a04 XG |
2653 | int idx, ret = 0; |
2654 | ||
2655 | idx = srcu_read_lock(&kvm->srcu); | |
b7ebfb05 | 2656 | if (!init_rmode_tss(kvm)) |
4b9d3a04 | 2657 | goto exit; |
b7ebfb05 | 2658 | if (!init_rmode_identity_map(kvm)) |
4b9d3a04 XG |
2659 | goto exit; |
2660 | ||
2661 | ret = 1; | |
2662 | exit: | |
2663 | srcu_read_unlock(&kvm->srcu, idx); | |
2664 | return ret; | |
b7ebfb05 SY |
2665 | } |
2666 | ||
e00c8cf2 AK |
2667 | static int vmx_vcpu_reset(struct kvm_vcpu *vcpu) |
2668 | { | |
2669 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
2670 | u64 msr; | |
4b9d3a04 | 2671 | int ret; |
e00c8cf2 | 2672 | |
5fdbf976 | 2673 | vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP)); |
b7ebfb05 | 2674 | if (!init_rmode(vmx->vcpu.kvm)) { |
e00c8cf2 AK |
2675 | ret = -ENOMEM; |
2676 | goto out; | |
2677 | } | |
2678 | ||
7ffd92c5 | 2679 | vmx->rmode.vm86_active = 0; |
e00c8cf2 | 2680 | |
3b86cd99 JK |
2681 | vmx->soft_vnmi_blocked = 0; |
2682 | ||
ad312c7c | 2683 | vmx->vcpu.arch.regs[VCPU_REGS_RDX] = get_rdx_init_val(); |
2d3ad1f4 | 2684 | kvm_set_cr8(&vmx->vcpu, 0); |
e00c8cf2 | 2685 | msr = 0xfee00000 | MSR_IA32_APICBASE_ENABLE; |
c5af89b6 | 2686 | if (kvm_vcpu_is_bsp(&vmx->vcpu)) |
e00c8cf2 AK |
2687 | msr |= MSR_IA32_APICBASE_BSP; |
2688 | kvm_set_apic_base(&vmx->vcpu, msr); | |
2689 | ||
10ab25cd JK |
2690 | ret = fx_init(&vmx->vcpu); |
2691 | if (ret != 0) | |
2692 | goto out; | |
e00c8cf2 | 2693 | |
5706be0d | 2694 | seg_setup(VCPU_SREG_CS); |
e00c8cf2 AK |
2695 | /* |
2696 | * GUEST_CS_BASE should really be 0xffff0000, but VT vm86 mode | |
2697 | * insists on having GUEST_CS_BASE == GUEST_CS_SELECTOR << 4. Sigh. | |
2698 | */ | |
c5af89b6 | 2699 | if (kvm_vcpu_is_bsp(&vmx->vcpu)) { |
e00c8cf2 AK |
2700 | vmcs_write16(GUEST_CS_SELECTOR, 0xf000); |
2701 | vmcs_writel(GUEST_CS_BASE, 0x000f0000); | |
2702 | } else { | |
ad312c7c ZX |
2703 | vmcs_write16(GUEST_CS_SELECTOR, vmx->vcpu.arch.sipi_vector << 8); |
2704 | vmcs_writel(GUEST_CS_BASE, vmx->vcpu.arch.sipi_vector << 12); | |
e00c8cf2 | 2705 | } |
e00c8cf2 AK |
2706 | |
2707 | seg_setup(VCPU_SREG_DS); | |
2708 | seg_setup(VCPU_SREG_ES); | |
2709 | seg_setup(VCPU_SREG_FS); | |
2710 | seg_setup(VCPU_SREG_GS); | |
2711 | seg_setup(VCPU_SREG_SS); | |
2712 | ||
2713 | vmcs_write16(GUEST_TR_SELECTOR, 0); | |
2714 | vmcs_writel(GUEST_TR_BASE, 0); | |
2715 | vmcs_write32(GUEST_TR_LIMIT, 0xffff); | |
2716 | vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); | |
2717 | ||
2718 | vmcs_write16(GUEST_LDTR_SELECTOR, 0); | |
2719 | vmcs_writel(GUEST_LDTR_BASE, 0); | |
2720 | vmcs_write32(GUEST_LDTR_LIMIT, 0xffff); | |
2721 | vmcs_write32(GUEST_LDTR_AR_BYTES, 0x00082); | |
2722 | ||
2723 | vmcs_write32(GUEST_SYSENTER_CS, 0); | |
2724 | vmcs_writel(GUEST_SYSENTER_ESP, 0); | |
2725 | vmcs_writel(GUEST_SYSENTER_EIP, 0); | |
2726 | ||
2727 | vmcs_writel(GUEST_RFLAGS, 0x02); | |
c5af89b6 | 2728 | if (kvm_vcpu_is_bsp(&vmx->vcpu)) |
5fdbf976 | 2729 | kvm_rip_write(vcpu, 0xfff0); |
e00c8cf2 | 2730 | else |
5fdbf976 MT |
2731 | kvm_rip_write(vcpu, 0); |
2732 | kvm_register_write(vcpu, VCPU_REGS_RSP, 0); | |
e00c8cf2 | 2733 | |
e00c8cf2 AK |
2734 | vmcs_writel(GUEST_DR7, 0x400); |
2735 | ||
2736 | vmcs_writel(GUEST_GDTR_BASE, 0); | |
2737 | vmcs_write32(GUEST_GDTR_LIMIT, 0xffff); | |
2738 | ||
2739 | vmcs_writel(GUEST_IDTR_BASE, 0); | |
2740 | vmcs_write32(GUEST_IDTR_LIMIT, 0xffff); | |
2741 | ||
2742 | vmcs_write32(GUEST_ACTIVITY_STATE, 0); | |
2743 | vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, 0); | |
2744 | vmcs_write32(GUEST_PENDING_DBG_EXCEPTIONS, 0); | |
2745 | ||
e00c8cf2 AK |
2746 | /* Special registers */ |
2747 | vmcs_write64(GUEST_IA32_DEBUGCTL, 0); | |
2748 | ||
2749 | setup_msrs(vmx); | |
2750 | ||
6aa8b732 AK |
2751 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); /* 22.2.1 */ |
2752 | ||
f78e0e2e SY |
2753 | if (cpu_has_vmx_tpr_shadow()) { |
2754 | vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, 0); | |
2755 | if (vm_need_tpr_shadow(vmx->vcpu.kvm)) | |
2756 | vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, | |
ad312c7c | 2757 | page_to_phys(vmx->vcpu.arch.apic->regs_page)); |
f78e0e2e SY |
2758 | vmcs_write32(TPR_THRESHOLD, 0); |
2759 | } | |
2760 | ||
2761 | if (vm_need_virtualize_apic_accesses(vmx->vcpu.kvm)) | |
2762 | vmcs_write64(APIC_ACCESS_ADDR, | |
bfc6d222 | 2763 | page_to_phys(vmx->vcpu.kvm->arch.apic_access_page)); |
6aa8b732 | 2764 | |
2384d2b3 SY |
2765 | if (vmx->vpid != 0) |
2766 | vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->vpid); | |
2767 | ||
fa40052c | 2768 | vmx->vcpu.arch.cr0 = X86_CR0_NW | X86_CR0_CD | X86_CR0_ET; |
4d4ec087 | 2769 | vmx_set_cr0(&vmx->vcpu, kvm_read_cr0(vcpu)); /* enter rmode */ |
8b9cf98c | 2770 | vmx_set_cr4(&vmx->vcpu, 0); |
8b9cf98c | 2771 | vmx_set_efer(&vmx->vcpu, 0); |
8b9cf98c RR |
2772 | vmx_fpu_activate(&vmx->vcpu); |
2773 | update_exception_bitmap(&vmx->vcpu); | |
6aa8b732 | 2774 | |
b9d762fa | 2775 | vpid_sync_context(vmx); |
2384d2b3 | 2776 | |
3200f405 | 2777 | ret = 0; |
6aa8b732 | 2778 | |
a89a8fb9 MG |
2779 | /* HACK: Don't enable emulation on guest boot/reset */ |
2780 | vmx->emulation_required = 0; | |
2781 | ||
6aa8b732 AK |
2782 | out: |
2783 | return ret; | |
2784 | } | |
2785 | ||
3b86cd99 JK |
2786 | static void enable_irq_window(struct kvm_vcpu *vcpu) |
2787 | { | |
2788 | u32 cpu_based_vm_exec_control; | |
2789 | ||
2790 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
2791 | cpu_based_vm_exec_control |= CPU_BASED_VIRTUAL_INTR_PENDING; | |
2792 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
2793 | } | |
2794 | ||
2795 | static void enable_nmi_window(struct kvm_vcpu *vcpu) | |
2796 | { | |
2797 | u32 cpu_based_vm_exec_control; | |
2798 | ||
2799 | if (!cpu_has_virtual_nmis()) { | |
2800 | enable_irq_window(vcpu); | |
2801 | return; | |
2802 | } | |
2803 | ||
2804 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
2805 | cpu_based_vm_exec_control |= CPU_BASED_VIRTUAL_NMI_PENDING; | |
2806 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
2807 | } | |
2808 | ||
66fd3f7f | 2809 | static void vmx_inject_irq(struct kvm_vcpu *vcpu) |
85f455f7 | 2810 | { |
9c8cba37 | 2811 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
66fd3f7f GN |
2812 | uint32_t intr; |
2813 | int irq = vcpu->arch.interrupt.nr; | |
9c8cba37 | 2814 | |
229456fc | 2815 | trace_kvm_inj_virq(irq); |
2714d1d3 | 2816 | |
fa89a817 | 2817 | ++vcpu->stat.irq_injections; |
7ffd92c5 | 2818 | if (vmx->rmode.vm86_active) { |
9c8cba37 AK |
2819 | vmx->rmode.irq.pending = true; |
2820 | vmx->rmode.irq.vector = irq; | |
5fdbf976 | 2821 | vmx->rmode.irq.rip = kvm_rip_read(vcpu); |
ae0bb3e0 GN |
2822 | if (vcpu->arch.interrupt.soft) |
2823 | vmx->rmode.irq.rip += | |
2824 | vmx->vcpu.arch.event_exit_inst_len; | |
9c5623e3 AK |
2825 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, |
2826 | irq | INTR_TYPE_SOFT_INTR | INTR_INFO_VALID_MASK); | |
2827 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, 1); | |
5fdbf976 | 2828 | kvm_rip_write(vcpu, vmx->rmode.irq.rip - 1); |
85f455f7 ED |
2829 | return; |
2830 | } | |
66fd3f7f GN |
2831 | intr = irq | INTR_INFO_VALID_MASK; |
2832 | if (vcpu->arch.interrupt.soft) { | |
2833 | intr |= INTR_TYPE_SOFT_INTR; | |
2834 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, | |
2835 | vmx->vcpu.arch.event_exit_inst_len); | |
2836 | } else | |
2837 | intr |= INTR_TYPE_EXT_INTR; | |
2838 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr); | |
85f455f7 ED |
2839 | } |
2840 | ||
f08864b4 SY |
2841 | static void vmx_inject_nmi(struct kvm_vcpu *vcpu) |
2842 | { | |
66a5a347 JK |
2843 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
2844 | ||
3b86cd99 JK |
2845 | if (!cpu_has_virtual_nmis()) { |
2846 | /* | |
2847 | * Tracking the NMI-blocked state in software is built upon | |
2848 | * finding the next open IRQ window. This, in turn, depends on | |
2849 | * well-behaving guests: They have to keep IRQs disabled at | |
2850 | * least as long as the NMI handler runs. Otherwise we may | |
2851 | * cause NMI nesting, maybe breaking the guest. But as this is | |
2852 | * highly unlikely, we can live with the residual risk. | |
2853 | */ | |
2854 | vmx->soft_vnmi_blocked = 1; | |
2855 | vmx->vnmi_blocked_time = 0; | |
2856 | } | |
2857 | ||
487b391d | 2858 | ++vcpu->stat.nmi_injections; |
7ffd92c5 | 2859 | if (vmx->rmode.vm86_active) { |
66a5a347 JK |
2860 | vmx->rmode.irq.pending = true; |
2861 | vmx->rmode.irq.vector = NMI_VECTOR; | |
2862 | vmx->rmode.irq.rip = kvm_rip_read(vcpu); | |
2863 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, | |
2864 | NMI_VECTOR | INTR_TYPE_SOFT_INTR | | |
2865 | INTR_INFO_VALID_MASK); | |
2866 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, 1); | |
2867 | kvm_rip_write(vcpu, vmx->rmode.irq.rip - 1); | |
2868 | return; | |
2869 | } | |
f08864b4 SY |
2870 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, |
2871 | INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK | NMI_VECTOR); | |
f08864b4 SY |
2872 | } |
2873 | ||
c4282df9 | 2874 | static int vmx_nmi_allowed(struct kvm_vcpu *vcpu) |
33f089ca | 2875 | { |
3b86cd99 | 2876 | if (!cpu_has_virtual_nmis() && to_vmx(vcpu)->soft_vnmi_blocked) |
c4282df9 | 2877 | return 0; |
33f089ca | 2878 | |
c4282df9 | 2879 | return !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & |
f8c5fae1 | 2880 | (GUEST_INTR_STATE_MOV_SS | GUEST_INTR_STATE_NMI)); |
33f089ca JK |
2881 | } |
2882 | ||
3cfc3092 JK |
2883 | static bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu) |
2884 | { | |
2885 | if (!cpu_has_virtual_nmis()) | |
2886 | return to_vmx(vcpu)->soft_vnmi_blocked; | |
c332c83a | 2887 | return vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_NMI; |
3cfc3092 JK |
2888 | } |
2889 | ||
2890 | static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked) | |
2891 | { | |
2892 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
2893 | ||
2894 | if (!cpu_has_virtual_nmis()) { | |
2895 | if (vmx->soft_vnmi_blocked != masked) { | |
2896 | vmx->soft_vnmi_blocked = masked; | |
2897 | vmx->vnmi_blocked_time = 0; | |
2898 | } | |
2899 | } else { | |
2900 | if (masked) | |
2901 | vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, | |
2902 | GUEST_INTR_STATE_NMI); | |
2903 | else | |
2904 | vmcs_clear_bits(GUEST_INTERRUPTIBILITY_INFO, | |
2905 | GUEST_INTR_STATE_NMI); | |
2906 | } | |
2907 | } | |
2908 | ||
78646121 GN |
2909 | static int vmx_interrupt_allowed(struct kvm_vcpu *vcpu) |
2910 | { | |
c4282df9 GN |
2911 | return (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF) && |
2912 | !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & | |
2913 | (GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS)); | |
78646121 GN |
2914 | } |
2915 | ||
cbc94022 IE |
2916 | static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr) |
2917 | { | |
2918 | int ret; | |
2919 | struct kvm_userspace_memory_region tss_mem = { | |
6fe63979 | 2920 | .slot = TSS_PRIVATE_MEMSLOT, |
cbc94022 IE |
2921 | .guest_phys_addr = addr, |
2922 | .memory_size = PAGE_SIZE * 3, | |
2923 | .flags = 0, | |
2924 | }; | |
2925 | ||
2926 | ret = kvm_set_memory_region(kvm, &tss_mem, 0); | |
2927 | if (ret) | |
2928 | return ret; | |
bfc6d222 | 2929 | kvm->arch.tss_addr = addr; |
cbc94022 IE |
2930 | return 0; |
2931 | } | |
2932 | ||
6aa8b732 AK |
2933 | static int handle_rmode_exception(struct kvm_vcpu *vcpu, |
2934 | int vec, u32 err_code) | |
2935 | { | |
b3f37707 NK |
2936 | /* |
2937 | * Instruction with address size override prefix opcode 0x67 | |
2938 | * Cause the #SS fault with 0 error code in VM86 mode. | |
2939 | */ | |
2940 | if (((vec == GP_VECTOR) || (vec == SS_VECTOR)) && err_code == 0) | |
851ba692 | 2941 | if (emulate_instruction(vcpu, 0, 0, 0) == EMULATE_DONE) |
6aa8b732 | 2942 | return 1; |
77ab6db0 JK |
2943 | /* |
2944 | * Forward all other exceptions that are valid in real mode. | |
2945 | * FIXME: Breaks guest debugging in real mode, needs to be fixed with | |
2946 | * the required debugging infrastructure rework. | |
2947 | */ | |
2948 | switch (vec) { | |
77ab6db0 | 2949 | case DB_VECTOR: |
d0bfb940 JK |
2950 | if (vcpu->guest_debug & |
2951 | (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP)) | |
2952 | return 0; | |
2953 | kvm_queue_exception(vcpu, vec); | |
2954 | return 1; | |
77ab6db0 | 2955 | case BP_VECTOR: |
c573cd22 JK |
2956 | /* |
2957 | * Update instruction length as we may reinject the exception | |
2958 | * from user space while in guest debugging mode. | |
2959 | */ | |
2960 | to_vmx(vcpu)->vcpu.arch.event_exit_inst_len = | |
2961 | vmcs_read32(VM_EXIT_INSTRUCTION_LEN); | |
d0bfb940 JK |
2962 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) |
2963 | return 0; | |
2964 | /* fall through */ | |
2965 | case DE_VECTOR: | |
77ab6db0 JK |
2966 | case OF_VECTOR: |
2967 | case BR_VECTOR: | |
2968 | case UD_VECTOR: | |
2969 | case DF_VECTOR: | |
2970 | case SS_VECTOR: | |
2971 | case GP_VECTOR: | |
2972 | case MF_VECTOR: | |
2973 | kvm_queue_exception(vcpu, vec); | |
2974 | return 1; | |
2975 | } | |
6aa8b732 AK |
2976 | return 0; |
2977 | } | |
2978 | ||
a0861c02 AK |
2979 | /* |
2980 | * Trigger machine check on the host. We assume all the MSRs are already set up | |
2981 | * by the CPU and that we still run on the same CPU as the MCE occurred on. | |
2982 | * We pass a fake environment to the machine check handler because we want | |
2983 | * the guest to be always treated like user space, no matter what context | |
2984 | * it used internally. | |
2985 | */ | |
2986 | static void kvm_machine_check(void) | |
2987 | { | |
2988 | #if defined(CONFIG_X86_MCE) && defined(CONFIG_X86_64) | |
2989 | struct pt_regs regs = { | |
2990 | .cs = 3, /* Fake ring 3 no matter what the guest ran on */ | |
2991 | .flags = X86_EFLAGS_IF, | |
2992 | }; | |
2993 | ||
2994 | do_machine_check(®s, 0); | |
2995 | #endif | |
2996 | } | |
2997 | ||
851ba692 | 2998 | static int handle_machine_check(struct kvm_vcpu *vcpu) |
a0861c02 AK |
2999 | { |
3000 | /* already handled by vcpu_run */ | |
3001 | return 1; | |
3002 | } | |
3003 | ||
851ba692 | 3004 | static int handle_exception(struct kvm_vcpu *vcpu) |
6aa8b732 | 3005 | { |
1155f76a | 3006 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
851ba692 | 3007 | struct kvm_run *kvm_run = vcpu->run; |
d0bfb940 | 3008 | u32 intr_info, ex_no, error_code; |
42dbaa5a | 3009 | unsigned long cr2, rip, dr6; |
6aa8b732 AK |
3010 | u32 vect_info; |
3011 | enum emulation_result er; | |
3012 | ||
1155f76a | 3013 | vect_info = vmx->idt_vectoring_info; |
6aa8b732 AK |
3014 | intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
3015 | ||
a0861c02 | 3016 | if (is_machine_check(intr_info)) |
851ba692 | 3017 | return handle_machine_check(vcpu); |
a0861c02 | 3018 | |
6aa8b732 | 3019 | if ((vect_info & VECTORING_INFO_VALID_MASK) && |
65ac7264 AK |
3020 | !is_page_fault(intr_info)) { |
3021 | vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; | |
3022 | vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_SIMUL_EX; | |
3023 | vcpu->run->internal.ndata = 2; | |
3024 | vcpu->run->internal.data[0] = vect_info; | |
3025 | vcpu->run->internal.data[1] = intr_info; | |
3026 | return 0; | |
3027 | } | |
6aa8b732 | 3028 | |
e4a41889 | 3029 | if ((intr_info & INTR_INFO_INTR_TYPE_MASK) == INTR_TYPE_NMI_INTR) |
1b6269db | 3030 | return 1; /* already handled by vmx_vcpu_run() */ |
2ab455cc AL |
3031 | |
3032 | if (is_no_device(intr_info)) { | |
5fd86fcf | 3033 | vmx_fpu_activate(vcpu); |
2ab455cc AL |
3034 | return 1; |
3035 | } | |
3036 | ||
7aa81cc0 | 3037 | if (is_invalid_opcode(intr_info)) { |
851ba692 | 3038 | er = emulate_instruction(vcpu, 0, 0, EMULTYPE_TRAP_UD); |
7aa81cc0 | 3039 | if (er != EMULATE_DONE) |
7ee5d940 | 3040 | kvm_queue_exception(vcpu, UD_VECTOR); |
7aa81cc0 AL |
3041 | return 1; |
3042 | } | |
3043 | ||
6aa8b732 | 3044 | error_code = 0; |
5fdbf976 | 3045 | rip = kvm_rip_read(vcpu); |
2e11384c | 3046 | if (intr_info & INTR_INFO_DELIVER_CODE_MASK) |
6aa8b732 AK |
3047 | error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE); |
3048 | if (is_page_fault(intr_info)) { | |
1439442c | 3049 | /* EPT won't cause page fault directly */ |
089d034e | 3050 | if (enable_ept) |
1439442c | 3051 | BUG(); |
6aa8b732 | 3052 | cr2 = vmcs_readl(EXIT_QUALIFICATION); |
229456fc MT |
3053 | trace_kvm_page_fault(cr2, error_code); |
3054 | ||
3298b75c | 3055 | if (kvm_event_needs_reinjection(vcpu)) |
577bdc49 | 3056 | kvm_mmu_unprotect_page_virt(vcpu, cr2); |
3067714c | 3057 | return kvm_mmu_page_fault(vcpu, cr2, error_code); |
6aa8b732 AK |
3058 | } |
3059 | ||
7ffd92c5 | 3060 | if (vmx->rmode.vm86_active && |
6aa8b732 | 3061 | handle_rmode_exception(vcpu, intr_info & INTR_INFO_VECTOR_MASK, |
72d6e5a0 | 3062 | error_code)) { |
ad312c7c ZX |
3063 | if (vcpu->arch.halt_request) { |
3064 | vcpu->arch.halt_request = 0; | |
72d6e5a0 AK |
3065 | return kvm_emulate_halt(vcpu); |
3066 | } | |
6aa8b732 | 3067 | return 1; |
72d6e5a0 | 3068 | } |
6aa8b732 | 3069 | |
d0bfb940 | 3070 | ex_no = intr_info & INTR_INFO_VECTOR_MASK; |
42dbaa5a JK |
3071 | switch (ex_no) { |
3072 | case DB_VECTOR: | |
3073 | dr6 = vmcs_readl(EXIT_QUALIFICATION); | |
3074 | if (!(vcpu->guest_debug & | |
3075 | (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP))) { | |
3076 | vcpu->arch.dr6 = dr6 | DR6_FIXED_1; | |
3077 | kvm_queue_exception(vcpu, DB_VECTOR); | |
3078 | return 1; | |
3079 | } | |
3080 | kvm_run->debug.arch.dr6 = dr6 | DR6_FIXED_1; | |
3081 | kvm_run->debug.arch.dr7 = vmcs_readl(GUEST_DR7); | |
3082 | /* fall through */ | |
3083 | case BP_VECTOR: | |
c573cd22 JK |
3084 | /* |
3085 | * Update instruction length as we may reinject #BP from | |
3086 | * user space while in guest debugging mode. Reading it for | |
3087 | * #DB as well causes no harm, it is not used in that case. | |
3088 | */ | |
3089 | vmx->vcpu.arch.event_exit_inst_len = | |
3090 | vmcs_read32(VM_EXIT_INSTRUCTION_LEN); | |
6aa8b732 | 3091 | kvm_run->exit_reason = KVM_EXIT_DEBUG; |
d0bfb940 JK |
3092 | kvm_run->debug.arch.pc = vmcs_readl(GUEST_CS_BASE) + rip; |
3093 | kvm_run->debug.arch.exception = ex_no; | |
42dbaa5a JK |
3094 | break; |
3095 | default: | |
d0bfb940 JK |
3096 | kvm_run->exit_reason = KVM_EXIT_EXCEPTION; |
3097 | kvm_run->ex.exception = ex_no; | |
3098 | kvm_run->ex.error_code = error_code; | |
42dbaa5a | 3099 | break; |
6aa8b732 | 3100 | } |
6aa8b732 AK |
3101 | return 0; |
3102 | } | |
3103 | ||
851ba692 | 3104 | static int handle_external_interrupt(struct kvm_vcpu *vcpu) |
6aa8b732 | 3105 | { |
1165f5fe | 3106 | ++vcpu->stat.irq_exits; |
6aa8b732 AK |
3107 | return 1; |
3108 | } | |
3109 | ||
851ba692 | 3110 | static int handle_triple_fault(struct kvm_vcpu *vcpu) |
988ad74f | 3111 | { |
851ba692 | 3112 | vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; |
988ad74f AK |
3113 | return 0; |
3114 | } | |
6aa8b732 | 3115 | |
851ba692 | 3116 | static int handle_io(struct kvm_vcpu *vcpu) |
6aa8b732 | 3117 | { |
bfdaab09 | 3118 | unsigned long exit_qualification; |
34c33d16 | 3119 | int size, in, string; |
039576c0 | 3120 | unsigned port; |
6aa8b732 | 3121 | |
bfdaab09 | 3122 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
039576c0 | 3123 | string = (exit_qualification & 16) != 0; |
cf8f70bf | 3124 | in = (exit_qualification & 8) != 0; |
e70669ab | 3125 | |
cf8f70bf | 3126 | ++vcpu->stat.io_exits; |
e70669ab | 3127 | |
cf8f70bf | 3128 | if (string || in) |
6d77dbfc | 3129 | return emulate_instruction(vcpu, 0, 0, 0) == EMULATE_DONE; |
e70669ab | 3130 | |
cf8f70bf GN |
3131 | port = exit_qualification >> 16; |
3132 | size = (exit_qualification & 7) + 1; | |
e93f36bc | 3133 | skip_emulated_instruction(vcpu); |
cf8f70bf GN |
3134 | |
3135 | return kvm_fast_pio_out(vcpu, size, port); | |
6aa8b732 AK |
3136 | } |
3137 | ||
102d8325 IM |
3138 | static void |
3139 | vmx_patch_hypercall(struct kvm_vcpu *vcpu, unsigned char *hypercall) | |
3140 | { | |
3141 | /* | |
3142 | * Patch in the VMCALL instruction: | |
3143 | */ | |
3144 | hypercall[0] = 0x0f; | |
3145 | hypercall[1] = 0x01; | |
3146 | hypercall[2] = 0xc1; | |
102d8325 IM |
3147 | } |
3148 | ||
49a9b07e AK |
3149 | static void complete_insn_gp(struct kvm_vcpu *vcpu, int err) |
3150 | { | |
3151 | if (err) | |
3152 | kvm_inject_gp(vcpu, 0); | |
3153 | else | |
3154 | skip_emulated_instruction(vcpu); | |
3155 | } | |
3156 | ||
851ba692 | 3157 | static int handle_cr(struct kvm_vcpu *vcpu) |
6aa8b732 | 3158 | { |
229456fc | 3159 | unsigned long exit_qualification, val; |
6aa8b732 AK |
3160 | int cr; |
3161 | int reg; | |
49a9b07e | 3162 | int err; |
6aa8b732 | 3163 | |
bfdaab09 | 3164 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6aa8b732 AK |
3165 | cr = exit_qualification & 15; |
3166 | reg = (exit_qualification >> 8) & 15; | |
3167 | switch ((exit_qualification >> 4) & 3) { | |
3168 | case 0: /* mov to cr */ | |
229456fc MT |
3169 | val = kvm_register_read(vcpu, reg); |
3170 | trace_kvm_cr_write(cr, val); | |
6aa8b732 AK |
3171 | switch (cr) { |
3172 | case 0: | |
49a9b07e AK |
3173 | err = kvm_set_cr0(vcpu, val); |
3174 | complete_insn_gp(vcpu, err); | |
6aa8b732 AK |
3175 | return 1; |
3176 | case 3: | |
2390218b AK |
3177 | err = kvm_set_cr3(vcpu, val); |
3178 | complete_insn_gp(vcpu, err); | |
6aa8b732 AK |
3179 | return 1; |
3180 | case 4: | |
a83b29c6 AK |
3181 | err = kvm_set_cr4(vcpu, val); |
3182 | complete_insn_gp(vcpu, err); | |
6aa8b732 | 3183 | return 1; |
0a5fff19 GN |
3184 | case 8: { |
3185 | u8 cr8_prev = kvm_get_cr8(vcpu); | |
3186 | u8 cr8 = kvm_register_read(vcpu, reg); | |
3187 | kvm_set_cr8(vcpu, cr8); | |
3188 | skip_emulated_instruction(vcpu); | |
3189 | if (irqchip_in_kernel(vcpu->kvm)) | |
3190 | return 1; | |
3191 | if (cr8_prev <= cr8) | |
3192 | return 1; | |
851ba692 | 3193 | vcpu->run->exit_reason = KVM_EXIT_SET_TPR; |
0a5fff19 GN |
3194 | return 0; |
3195 | } | |
6aa8b732 AK |
3196 | }; |
3197 | break; | |
25c4c276 | 3198 | case 2: /* clts */ |
edcafe3c | 3199 | vmx_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS)); |
4d4ec087 | 3200 | trace_kvm_cr_write(0, kvm_read_cr0(vcpu)); |
25c4c276 | 3201 | skip_emulated_instruction(vcpu); |
6b52d186 | 3202 | vmx_fpu_activate(vcpu); |
25c4c276 | 3203 | return 1; |
6aa8b732 AK |
3204 | case 1: /*mov from cr*/ |
3205 | switch (cr) { | |
3206 | case 3: | |
5fdbf976 | 3207 | kvm_register_write(vcpu, reg, vcpu->arch.cr3); |
229456fc | 3208 | trace_kvm_cr_read(cr, vcpu->arch.cr3); |
6aa8b732 AK |
3209 | skip_emulated_instruction(vcpu); |
3210 | return 1; | |
3211 | case 8: | |
229456fc MT |
3212 | val = kvm_get_cr8(vcpu); |
3213 | kvm_register_write(vcpu, reg, val); | |
3214 | trace_kvm_cr_read(cr, val); | |
6aa8b732 AK |
3215 | skip_emulated_instruction(vcpu); |
3216 | return 1; | |
3217 | } | |
3218 | break; | |
3219 | case 3: /* lmsw */ | |
a1f83a74 | 3220 | val = (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f; |
4d4ec087 | 3221 | trace_kvm_cr_write(0, (kvm_read_cr0(vcpu) & ~0xful) | val); |
a1f83a74 | 3222 | kvm_lmsw(vcpu, val); |
6aa8b732 AK |
3223 | |
3224 | skip_emulated_instruction(vcpu); | |
3225 | return 1; | |
3226 | default: | |
3227 | break; | |
3228 | } | |
851ba692 | 3229 | vcpu->run->exit_reason = 0; |
f0242478 | 3230 | pr_unimpl(vcpu, "unhandled control register: op %d cr %d\n", |
6aa8b732 AK |
3231 | (int)(exit_qualification >> 4) & 3, cr); |
3232 | return 0; | |
3233 | } | |
3234 | ||
851ba692 | 3235 | static int handle_dr(struct kvm_vcpu *vcpu) |
6aa8b732 | 3236 | { |
bfdaab09 | 3237 | unsigned long exit_qualification; |
6aa8b732 AK |
3238 | int dr, reg; |
3239 | ||
f2483415 | 3240 | /* Do not handle if the CPL > 0, will trigger GP on re-entry */ |
0a79b009 AK |
3241 | if (!kvm_require_cpl(vcpu, 0)) |
3242 | return 1; | |
42dbaa5a JK |
3243 | dr = vmcs_readl(GUEST_DR7); |
3244 | if (dr & DR7_GD) { | |
3245 | /* | |
3246 | * As the vm-exit takes precedence over the debug trap, we | |
3247 | * need to emulate the latter, either for the host or the | |
3248 | * guest debugging itself. | |
3249 | */ | |
3250 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { | |
851ba692 AK |
3251 | vcpu->run->debug.arch.dr6 = vcpu->arch.dr6; |
3252 | vcpu->run->debug.arch.dr7 = dr; | |
3253 | vcpu->run->debug.arch.pc = | |
42dbaa5a JK |
3254 | vmcs_readl(GUEST_CS_BASE) + |
3255 | vmcs_readl(GUEST_RIP); | |
851ba692 AK |
3256 | vcpu->run->debug.arch.exception = DB_VECTOR; |
3257 | vcpu->run->exit_reason = KVM_EXIT_DEBUG; | |
42dbaa5a JK |
3258 | return 0; |
3259 | } else { | |
3260 | vcpu->arch.dr7 &= ~DR7_GD; | |
3261 | vcpu->arch.dr6 |= DR6_BD; | |
3262 | vmcs_writel(GUEST_DR7, vcpu->arch.dr7); | |
3263 | kvm_queue_exception(vcpu, DB_VECTOR); | |
3264 | return 1; | |
3265 | } | |
3266 | } | |
3267 | ||
bfdaab09 | 3268 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
42dbaa5a JK |
3269 | dr = exit_qualification & DEBUG_REG_ACCESS_NUM; |
3270 | reg = DEBUG_REG_ACCESS_REG(exit_qualification); | |
3271 | if (exit_qualification & TYPE_MOV_FROM_DR) { | |
020df079 GN |
3272 | unsigned long val; |
3273 | if (!kvm_get_dr(vcpu, dr, &val)) | |
3274 | kvm_register_write(vcpu, reg, val); | |
3275 | } else | |
3276 | kvm_set_dr(vcpu, dr, vcpu->arch.regs[reg]); | |
6aa8b732 AK |
3277 | skip_emulated_instruction(vcpu); |
3278 | return 1; | |
3279 | } | |
3280 | ||
020df079 GN |
3281 | static void vmx_set_dr7(struct kvm_vcpu *vcpu, unsigned long val) |
3282 | { | |
3283 | vmcs_writel(GUEST_DR7, val); | |
3284 | } | |
3285 | ||
851ba692 | 3286 | static int handle_cpuid(struct kvm_vcpu *vcpu) |
6aa8b732 | 3287 | { |
06465c5a AK |
3288 | kvm_emulate_cpuid(vcpu); |
3289 | return 1; | |
6aa8b732 AK |
3290 | } |
3291 | ||
851ba692 | 3292 | static int handle_rdmsr(struct kvm_vcpu *vcpu) |
6aa8b732 | 3293 | { |
ad312c7c | 3294 | u32 ecx = vcpu->arch.regs[VCPU_REGS_RCX]; |
6aa8b732 AK |
3295 | u64 data; |
3296 | ||
3297 | if (vmx_get_msr(vcpu, ecx, &data)) { | |
59200273 | 3298 | trace_kvm_msr_read_ex(ecx); |
c1a5d4f9 | 3299 | kvm_inject_gp(vcpu, 0); |
6aa8b732 AK |
3300 | return 1; |
3301 | } | |
3302 | ||
229456fc | 3303 | trace_kvm_msr_read(ecx, data); |
2714d1d3 | 3304 | |
6aa8b732 | 3305 | /* FIXME: handling of bits 32:63 of rax, rdx */ |
ad312c7c ZX |
3306 | vcpu->arch.regs[VCPU_REGS_RAX] = data & -1u; |
3307 | vcpu->arch.regs[VCPU_REGS_RDX] = (data >> 32) & -1u; | |
6aa8b732 AK |
3308 | skip_emulated_instruction(vcpu); |
3309 | return 1; | |
3310 | } | |
3311 | ||
851ba692 | 3312 | static int handle_wrmsr(struct kvm_vcpu *vcpu) |
6aa8b732 | 3313 | { |
ad312c7c ZX |
3314 | u32 ecx = vcpu->arch.regs[VCPU_REGS_RCX]; |
3315 | u64 data = (vcpu->arch.regs[VCPU_REGS_RAX] & -1u) | |
3316 | | ((u64)(vcpu->arch.regs[VCPU_REGS_RDX] & -1u) << 32); | |
6aa8b732 AK |
3317 | |
3318 | if (vmx_set_msr(vcpu, ecx, data) != 0) { | |
59200273 | 3319 | trace_kvm_msr_write_ex(ecx, data); |
c1a5d4f9 | 3320 | kvm_inject_gp(vcpu, 0); |
6aa8b732 AK |
3321 | return 1; |
3322 | } | |
3323 | ||
59200273 | 3324 | trace_kvm_msr_write(ecx, data); |
6aa8b732 AK |
3325 | skip_emulated_instruction(vcpu); |
3326 | return 1; | |
3327 | } | |
3328 | ||
851ba692 | 3329 | static int handle_tpr_below_threshold(struct kvm_vcpu *vcpu) |
6e5d865c | 3330 | { |
3842d135 | 3331 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
6e5d865c YS |
3332 | return 1; |
3333 | } | |
3334 | ||
851ba692 | 3335 | static int handle_interrupt_window(struct kvm_vcpu *vcpu) |
6aa8b732 | 3336 | { |
85f455f7 ED |
3337 | u32 cpu_based_vm_exec_control; |
3338 | ||
3339 | /* clear pending irq */ | |
3340 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
3341 | cpu_based_vm_exec_control &= ~CPU_BASED_VIRTUAL_INTR_PENDING; | |
3342 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
2714d1d3 | 3343 | |
3842d135 AK |
3344 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
3345 | ||
a26bf12a | 3346 | ++vcpu->stat.irq_window_exits; |
2714d1d3 | 3347 | |
c1150d8c DL |
3348 | /* |
3349 | * If the user space waits to inject interrupts, exit as soon as | |
3350 | * possible | |
3351 | */ | |
8061823a | 3352 | if (!irqchip_in_kernel(vcpu->kvm) && |
851ba692 | 3353 | vcpu->run->request_interrupt_window && |
8061823a | 3354 | !kvm_cpu_has_interrupt(vcpu)) { |
851ba692 | 3355 | vcpu->run->exit_reason = KVM_EXIT_IRQ_WINDOW_OPEN; |
c1150d8c DL |
3356 | return 0; |
3357 | } | |
6aa8b732 AK |
3358 | return 1; |
3359 | } | |
3360 | ||
851ba692 | 3361 | static int handle_halt(struct kvm_vcpu *vcpu) |
6aa8b732 AK |
3362 | { |
3363 | skip_emulated_instruction(vcpu); | |
d3bef15f | 3364 | return kvm_emulate_halt(vcpu); |
6aa8b732 AK |
3365 | } |
3366 | ||
851ba692 | 3367 | static int handle_vmcall(struct kvm_vcpu *vcpu) |
c21415e8 | 3368 | { |
510043da | 3369 | skip_emulated_instruction(vcpu); |
7aa81cc0 AL |
3370 | kvm_emulate_hypercall(vcpu); |
3371 | return 1; | |
c21415e8 IM |
3372 | } |
3373 | ||
851ba692 | 3374 | static int handle_vmx_insn(struct kvm_vcpu *vcpu) |
e3c7cb6a AK |
3375 | { |
3376 | kvm_queue_exception(vcpu, UD_VECTOR); | |
3377 | return 1; | |
3378 | } | |
3379 | ||
851ba692 | 3380 | static int handle_invlpg(struct kvm_vcpu *vcpu) |
a7052897 | 3381 | { |
f9c617f6 | 3382 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
a7052897 MT |
3383 | |
3384 | kvm_mmu_invlpg(vcpu, exit_qualification); | |
3385 | skip_emulated_instruction(vcpu); | |
3386 | return 1; | |
3387 | } | |
3388 | ||
851ba692 | 3389 | static int handle_wbinvd(struct kvm_vcpu *vcpu) |
e5edaa01 ED |
3390 | { |
3391 | skip_emulated_instruction(vcpu); | |
f5f48ee1 | 3392 | kvm_emulate_wbinvd(vcpu); |
e5edaa01 ED |
3393 | return 1; |
3394 | } | |
3395 | ||
2acf923e DC |
3396 | static int handle_xsetbv(struct kvm_vcpu *vcpu) |
3397 | { | |
3398 | u64 new_bv = kvm_read_edx_eax(vcpu); | |
3399 | u32 index = kvm_register_read(vcpu, VCPU_REGS_RCX); | |
3400 | ||
3401 | if (kvm_set_xcr(vcpu, index, new_bv) == 0) | |
3402 | skip_emulated_instruction(vcpu); | |
3403 | return 1; | |
3404 | } | |
3405 | ||
851ba692 | 3406 | static int handle_apic_access(struct kvm_vcpu *vcpu) |
f78e0e2e | 3407 | { |
6d77dbfc | 3408 | return emulate_instruction(vcpu, 0, 0, 0) == EMULATE_DONE; |
f78e0e2e SY |
3409 | } |
3410 | ||
851ba692 | 3411 | static int handle_task_switch(struct kvm_vcpu *vcpu) |
37817f29 | 3412 | { |
60637aac | 3413 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
37817f29 | 3414 | unsigned long exit_qualification; |
e269fb21 JK |
3415 | bool has_error_code = false; |
3416 | u32 error_code = 0; | |
37817f29 | 3417 | u16 tss_selector; |
64a7ec06 GN |
3418 | int reason, type, idt_v; |
3419 | ||
3420 | idt_v = (vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK); | |
3421 | type = (vmx->idt_vectoring_info & VECTORING_INFO_TYPE_MASK); | |
37817f29 IE |
3422 | |
3423 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
3424 | ||
3425 | reason = (u32)exit_qualification >> 30; | |
64a7ec06 GN |
3426 | if (reason == TASK_SWITCH_GATE && idt_v) { |
3427 | switch (type) { | |
3428 | case INTR_TYPE_NMI_INTR: | |
3429 | vcpu->arch.nmi_injected = false; | |
3430 | if (cpu_has_virtual_nmis()) | |
3431 | vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, | |
3432 | GUEST_INTR_STATE_NMI); | |
3433 | break; | |
3434 | case INTR_TYPE_EXT_INTR: | |
66fd3f7f | 3435 | case INTR_TYPE_SOFT_INTR: |
64a7ec06 GN |
3436 | kvm_clear_interrupt_queue(vcpu); |
3437 | break; | |
3438 | case INTR_TYPE_HARD_EXCEPTION: | |
e269fb21 JK |
3439 | if (vmx->idt_vectoring_info & |
3440 | VECTORING_INFO_DELIVER_CODE_MASK) { | |
3441 | has_error_code = true; | |
3442 | error_code = | |
3443 | vmcs_read32(IDT_VECTORING_ERROR_CODE); | |
3444 | } | |
3445 | /* fall through */ | |
64a7ec06 GN |
3446 | case INTR_TYPE_SOFT_EXCEPTION: |
3447 | kvm_clear_exception_queue(vcpu); | |
3448 | break; | |
3449 | default: | |
3450 | break; | |
3451 | } | |
60637aac | 3452 | } |
37817f29 IE |
3453 | tss_selector = exit_qualification; |
3454 | ||
64a7ec06 GN |
3455 | if (!idt_v || (type != INTR_TYPE_HARD_EXCEPTION && |
3456 | type != INTR_TYPE_EXT_INTR && | |
3457 | type != INTR_TYPE_NMI_INTR)) | |
3458 | skip_emulated_instruction(vcpu); | |
3459 | ||
acb54517 GN |
3460 | if (kvm_task_switch(vcpu, tss_selector, reason, |
3461 | has_error_code, error_code) == EMULATE_FAIL) { | |
3462 | vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; | |
3463 | vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION; | |
3464 | vcpu->run->internal.ndata = 0; | |
42dbaa5a | 3465 | return 0; |
acb54517 | 3466 | } |
42dbaa5a JK |
3467 | |
3468 | /* clear all local breakpoint enable flags */ | |
3469 | vmcs_writel(GUEST_DR7, vmcs_readl(GUEST_DR7) & ~55); | |
3470 | ||
3471 | /* | |
3472 | * TODO: What about debug traps on tss switch? | |
3473 | * Are we supposed to inject them and update dr6? | |
3474 | */ | |
3475 | ||
3476 | return 1; | |
37817f29 IE |
3477 | } |
3478 | ||
851ba692 | 3479 | static int handle_ept_violation(struct kvm_vcpu *vcpu) |
1439442c | 3480 | { |
f9c617f6 | 3481 | unsigned long exit_qualification; |
1439442c | 3482 | gpa_t gpa; |
1439442c | 3483 | int gla_validity; |
1439442c | 3484 | |
f9c617f6 | 3485 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
1439442c SY |
3486 | |
3487 | if (exit_qualification & (1 << 6)) { | |
3488 | printk(KERN_ERR "EPT: GPA exceeds GAW!\n"); | |
7f582ab6 | 3489 | return -EINVAL; |
1439442c SY |
3490 | } |
3491 | ||
3492 | gla_validity = (exit_qualification >> 7) & 0x3; | |
3493 | if (gla_validity != 0x3 && gla_validity != 0x1 && gla_validity != 0) { | |
3494 | printk(KERN_ERR "EPT: Handling EPT violation failed!\n"); | |
3495 | printk(KERN_ERR "EPT: GPA: 0x%lx, GVA: 0x%lx\n", | |
3496 | (long unsigned int)vmcs_read64(GUEST_PHYSICAL_ADDRESS), | |
f9c617f6 | 3497 | vmcs_readl(GUEST_LINEAR_ADDRESS)); |
1439442c SY |
3498 | printk(KERN_ERR "EPT: Exit qualification is 0x%lx\n", |
3499 | (long unsigned int)exit_qualification); | |
851ba692 AK |
3500 | vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; |
3501 | vcpu->run->hw.hardware_exit_reason = EXIT_REASON_EPT_VIOLATION; | |
596ae895 | 3502 | return 0; |
1439442c SY |
3503 | } |
3504 | ||
3505 | gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); | |
229456fc | 3506 | trace_kvm_page_fault(gpa, exit_qualification); |
49cd7d22 | 3507 | return kvm_mmu_page_fault(vcpu, gpa & PAGE_MASK, 0); |
1439442c SY |
3508 | } |
3509 | ||
68f89400 MT |
3510 | static u64 ept_rsvd_mask(u64 spte, int level) |
3511 | { | |
3512 | int i; | |
3513 | u64 mask = 0; | |
3514 | ||
3515 | for (i = 51; i > boot_cpu_data.x86_phys_bits; i--) | |
3516 | mask |= (1ULL << i); | |
3517 | ||
3518 | if (level > 2) | |
3519 | /* bits 7:3 reserved */ | |
3520 | mask |= 0xf8; | |
3521 | else if (level == 2) { | |
3522 | if (spte & (1ULL << 7)) | |
3523 | /* 2MB ref, bits 20:12 reserved */ | |
3524 | mask |= 0x1ff000; | |
3525 | else | |
3526 | /* bits 6:3 reserved */ | |
3527 | mask |= 0x78; | |
3528 | } | |
3529 | ||
3530 | return mask; | |
3531 | } | |
3532 | ||
3533 | static void ept_misconfig_inspect_spte(struct kvm_vcpu *vcpu, u64 spte, | |
3534 | int level) | |
3535 | { | |
3536 | printk(KERN_ERR "%s: spte 0x%llx level %d\n", __func__, spte, level); | |
3537 | ||
3538 | /* 010b (write-only) */ | |
3539 | WARN_ON((spte & 0x7) == 0x2); | |
3540 | ||
3541 | /* 110b (write/execute) */ | |
3542 | WARN_ON((spte & 0x7) == 0x6); | |
3543 | ||
3544 | /* 100b (execute-only) and value not supported by logical processor */ | |
3545 | if (!cpu_has_vmx_ept_execute_only()) | |
3546 | WARN_ON((spte & 0x7) == 0x4); | |
3547 | ||
3548 | /* not 000b */ | |
3549 | if ((spte & 0x7)) { | |
3550 | u64 rsvd_bits = spte & ept_rsvd_mask(spte, level); | |
3551 | ||
3552 | if (rsvd_bits != 0) { | |
3553 | printk(KERN_ERR "%s: rsvd_bits = 0x%llx\n", | |
3554 | __func__, rsvd_bits); | |
3555 | WARN_ON(1); | |
3556 | } | |
3557 | ||
3558 | if (level == 1 || (level == 2 && (spte & (1ULL << 7)))) { | |
3559 | u64 ept_mem_type = (spte & 0x38) >> 3; | |
3560 | ||
3561 | if (ept_mem_type == 2 || ept_mem_type == 3 || | |
3562 | ept_mem_type == 7) { | |
3563 | printk(KERN_ERR "%s: ept_mem_type=0x%llx\n", | |
3564 | __func__, ept_mem_type); | |
3565 | WARN_ON(1); | |
3566 | } | |
3567 | } | |
3568 | } | |
3569 | } | |
3570 | ||
851ba692 | 3571 | static int handle_ept_misconfig(struct kvm_vcpu *vcpu) |
68f89400 MT |
3572 | { |
3573 | u64 sptes[4]; | |
3574 | int nr_sptes, i; | |
3575 | gpa_t gpa; | |
3576 | ||
3577 | gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); | |
3578 | ||
3579 | printk(KERN_ERR "EPT: Misconfiguration.\n"); | |
3580 | printk(KERN_ERR "EPT: GPA: 0x%llx\n", gpa); | |
3581 | ||
3582 | nr_sptes = kvm_mmu_get_spte_hierarchy(vcpu, gpa, sptes); | |
3583 | ||
3584 | for (i = PT64_ROOT_LEVEL; i > PT64_ROOT_LEVEL - nr_sptes; --i) | |
3585 | ept_misconfig_inspect_spte(vcpu, sptes[i-1], i); | |
3586 | ||
851ba692 AK |
3587 | vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; |
3588 | vcpu->run->hw.hardware_exit_reason = EXIT_REASON_EPT_MISCONFIG; | |
68f89400 MT |
3589 | |
3590 | return 0; | |
3591 | } | |
3592 | ||
851ba692 | 3593 | static int handle_nmi_window(struct kvm_vcpu *vcpu) |
f08864b4 SY |
3594 | { |
3595 | u32 cpu_based_vm_exec_control; | |
3596 | ||
3597 | /* clear pending NMI */ | |
3598 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
3599 | cpu_based_vm_exec_control &= ~CPU_BASED_VIRTUAL_NMI_PENDING; | |
3600 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
3601 | ++vcpu->stat.nmi_window_exits; | |
3842d135 | 3602 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
f08864b4 SY |
3603 | |
3604 | return 1; | |
3605 | } | |
3606 | ||
80ced186 | 3607 | static int handle_invalid_guest_state(struct kvm_vcpu *vcpu) |
ea953ef0 | 3608 | { |
8b3079a5 AK |
3609 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
3610 | enum emulation_result err = EMULATE_DONE; | |
80ced186 | 3611 | int ret = 1; |
ea953ef0 MG |
3612 | |
3613 | while (!guest_state_valid(vcpu)) { | |
851ba692 | 3614 | err = emulate_instruction(vcpu, 0, 0, 0); |
ea953ef0 | 3615 | |
80ced186 MG |
3616 | if (err == EMULATE_DO_MMIO) { |
3617 | ret = 0; | |
3618 | goto out; | |
3619 | } | |
1d5a4d9b | 3620 | |
6d77dbfc GN |
3621 | if (err != EMULATE_DONE) |
3622 | return 0; | |
ea953ef0 MG |
3623 | |
3624 | if (signal_pending(current)) | |
80ced186 | 3625 | goto out; |
ea953ef0 MG |
3626 | if (need_resched()) |
3627 | schedule(); | |
3628 | } | |
3629 | ||
80ced186 MG |
3630 | vmx->emulation_required = 0; |
3631 | out: | |
3632 | return ret; | |
ea953ef0 MG |
3633 | } |
3634 | ||
4b8d54f9 ZE |
3635 | /* |
3636 | * Indicate a busy-waiting vcpu in spinlock. We do not enable the PAUSE | |
3637 | * exiting, so only get here on cpu with PAUSE-Loop-Exiting. | |
3638 | */ | |
9fb41ba8 | 3639 | static int handle_pause(struct kvm_vcpu *vcpu) |
4b8d54f9 ZE |
3640 | { |
3641 | skip_emulated_instruction(vcpu); | |
3642 | kvm_vcpu_on_spin(vcpu); | |
3643 | ||
3644 | return 1; | |
3645 | } | |
3646 | ||
59708670 SY |
3647 | static int handle_invalid_op(struct kvm_vcpu *vcpu) |
3648 | { | |
3649 | kvm_queue_exception(vcpu, UD_VECTOR); | |
3650 | return 1; | |
3651 | } | |
3652 | ||
6aa8b732 AK |
3653 | /* |
3654 | * The exit handlers return 1 if the exit was handled fully and guest execution | |
3655 | * may resume. Otherwise they set the kvm_run parameter to indicate what needs | |
3656 | * to be done to userspace and return 0. | |
3657 | */ | |
851ba692 | 3658 | static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = { |
6aa8b732 AK |
3659 | [EXIT_REASON_EXCEPTION_NMI] = handle_exception, |
3660 | [EXIT_REASON_EXTERNAL_INTERRUPT] = handle_external_interrupt, | |
988ad74f | 3661 | [EXIT_REASON_TRIPLE_FAULT] = handle_triple_fault, |
f08864b4 | 3662 | [EXIT_REASON_NMI_WINDOW] = handle_nmi_window, |
6aa8b732 | 3663 | [EXIT_REASON_IO_INSTRUCTION] = handle_io, |
6aa8b732 AK |
3664 | [EXIT_REASON_CR_ACCESS] = handle_cr, |
3665 | [EXIT_REASON_DR_ACCESS] = handle_dr, | |
3666 | [EXIT_REASON_CPUID] = handle_cpuid, | |
3667 | [EXIT_REASON_MSR_READ] = handle_rdmsr, | |
3668 | [EXIT_REASON_MSR_WRITE] = handle_wrmsr, | |
3669 | [EXIT_REASON_PENDING_INTERRUPT] = handle_interrupt_window, | |
3670 | [EXIT_REASON_HLT] = handle_halt, | |
a7052897 | 3671 | [EXIT_REASON_INVLPG] = handle_invlpg, |
c21415e8 | 3672 | [EXIT_REASON_VMCALL] = handle_vmcall, |
e3c7cb6a AK |
3673 | [EXIT_REASON_VMCLEAR] = handle_vmx_insn, |
3674 | [EXIT_REASON_VMLAUNCH] = handle_vmx_insn, | |
3675 | [EXIT_REASON_VMPTRLD] = handle_vmx_insn, | |
3676 | [EXIT_REASON_VMPTRST] = handle_vmx_insn, | |
3677 | [EXIT_REASON_VMREAD] = handle_vmx_insn, | |
3678 | [EXIT_REASON_VMRESUME] = handle_vmx_insn, | |
3679 | [EXIT_REASON_VMWRITE] = handle_vmx_insn, | |
3680 | [EXIT_REASON_VMOFF] = handle_vmx_insn, | |
3681 | [EXIT_REASON_VMON] = handle_vmx_insn, | |
f78e0e2e SY |
3682 | [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, |
3683 | [EXIT_REASON_APIC_ACCESS] = handle_apic_access, | |
e5edaa01 | 3684 | [EXIT_REASON_WBINVD] = handle_wbinvd, |
2acf923e | 3685 | [EXIT_REASON_XSETBV] = handle_xsetbv, |
37817f29 | 3686 | [EXIT_REASON_TASK_SWITCH] = handle_task_switch, |
a0861c02 | 3687 | [EXIT_REASON_MCE_DURING_VMENTRY] = handle_machine_check, |
68f89400 MT |
3688 | [EXIT_REASON_EPT_VIOLATION] = handle_ept_violation, |
3689 | [EXIT_REASON_EPT_MISCONFIG] = handle_ept_misconfig, | |
4b8d54f9 | 3690 | [EXIT_REASON_PAUSE_INSTRUCTION] = handle_pause, |
59708670 SY |
3691 | [EXIT_REASON_MWAIT_INSTRUCTION] = handle_invalid_op, |
3692 | [EXIT_REASON_MONITOR_INSTRUCTION] = handle_invalid_op, | |
6aa8b732 AK |
3693 | }; |
3694 | ||
3695 | static const int kvm_vmx_max_exit_handlers = | |
50a3485c | 3696 | ARRAY_SIZE(kvm_vmx_exit_handlers); |
6aa8b732 AK |
3697 | |
3698 | /* | |
3699 | * The guest has exited. See if we can fix it or if we need userspace | |
3700 | * assistance. | |
3701 | */ | |
851ba692 | 3702 | static int vmx_handle_exit(struct kvm_vcpu *vcpu) |
6aa8b732 | 3703 | { |
29bd8a78 | 3704 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
a0861c02 | 3705 | u32 exit_reason = vmx->exit_reason; |
1155f76a | 3706 | u32 vectoring_info = vmx->idt_vectoring_info; |
29bd8a78 | 3707 | |
5bfd8b54 | 3708 | trace_kvm_exit(exit_reason, vcpu); |
2714d1d3 | 3709 | |
80ced186 MG |
3710 | /* If guest state is invalid, start emulating */ |
3711 | if (vmx->emulation_required && emulate_invalid_guest_state) | |
3712 | return handle_invalid_guest_state(vcpu); | |
1d5a4d9b | 3713 | |
1439442c SY |
3714 | /* Access CR3 don't cause VMExit in paging mode, so we need |
3715 | * to sync with guest real CR3. */ | |
6de4f3ad | 3716 | if (enable_ept && is_paging(vcpu)) |
1439442c | 3717 | vcpu->arch.cr3 = vmcs_readl(GUEST_CR3); |
1439442c | 3718 | |
5120702e MG |
3719 | if (exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY) { |
3720 | vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; | |
3721 | vcpu->run->fail_entry.hardware_entry_failure_reason | |
3722 | = exit_reason; | |
3723 | return 0; | |
3724 | } | |
3725 | ||
29bd8a78 | 3726 | if (unlikely(vmx->fail)) { |
851ba692 AK |
3727 | vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; |
3728 | vcpu->run->fail_entry.hardware_entry_failure_reason | |
29bd8a78 AK |
3729 | = vmcs_read32(VM_INSTRUCTION_ERROR); |
3730 | return 0; | |
3731 | } | |
6aa8b732 | 3732 | |
d77c26fc | 3733 | if ((vectoring_info & VECTORING_INFO_VALID_MASK) && |
1439442c | 3734 | (exit_reason != EXIT_REASON_EXCEPTION_NMI && |
60637aac JK |
3735 | exit_reason != EXIT_REASON_EPT_VIOLATION && |
3736 | exit_reason != EXIT_REASON_TASK_SWITCH)) | |
3737 | printk(KERN_WARNING "%s: unexpected, valid vectoring info " | |
3738 | "(0x%x) and exit reason is 0x%x\n", | |
3739 | __func__, vectoring_info, exit_reason); | |
3b86cd99 JK |
3740 | |
3741 | if (unlikely(!cpu_has_virtual_nmis() && vmx->soft_vnmi_blocked)) { | |
c4282df9 | 3742 | if (vmx_interrupt_allowed(vcpu)) { |
3b86cd99 | 3743 | vmx->soft_vnmi_blocked = 0; |
3b86cd99 | 3744 | } else if (vmx->vnmi_blocked_time > 1000000000LL && |
4531220b | 3745 | vcpu->arch.nmi_pending) { |
3b86cd99 JK |
3746 | /* |
3747 | * This CPU don't support us in finding the end of an | |
3748 | * NMI-blocked window if the guest runs with IRQs | |
3749 | * disabled. So we pull the trigger after 1 s of | |
3750 | * futile waiting, but inform the user about this. | |
3751 | */ | |
3752 | printk(KERN_WARNING "%s: Breaking out of NMI-blocked " | |
3753 | "state on VCPU %d after 1 s timeout\n", | |
3754 | __func__, vcpu->vcpu_id); | |
3755 | vmx->soft_vnmi_blocked = 0; | |
3b86cd99 | 3756 | } |
3b86cd99 JK |
3757 | } |
3758 | ||
6aa8b732 AK |
3759 | if (exit_reason < kvm_vmx_max_exit_handlers |
3760 | && kvm_vmx_exit_handlers[exit_reason]) | |
851ba692 | 3761 | return kvm_vmx_exit_handlers[exit_reason](vcpu); |
6aa8b732 | 3762 | else { |
851ba692 AK |
3763 | vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; |
3764 | vcpu->run->hw.hardware_exit_reason = exit_reason; | |
6aa8b732 AK |
3765 | } |
3766 | return 0; | |
3767 | } | |
3768 | ||
95ba8273 | 3769 | static void update_cr8_intercept(struct kvm_vcpu *vcpu, int tpr, int irr) |
6e5d865c | 3770 | { |
95ba8273 | 3771 | if (irr == -1 || tpr < irr) { |
6e5d865c YS |
3772 | vmcs_write32(TPR_THRESHOLD, 0); |
3773 | return; | |
3774 | } | |
3775 | ||
95ba8273 | 3776 | vmcs_write32(TPR_THRESHOLD, irr); |
6e5d865c YS |
3777 | } |
3778 | ||
51aa01d1 | 3779 | static void vmx_complete_atomic_exit(struct vcpu_vmx *vmx) |
cf393f75 | 3780 | { |
51aa01d1 | 3781 | u32 exit_intr_info = vmx->exit_intr_info; |
a0861c02 AK |
3782 | |
3783 | /* Handle machine checks before interrupts are enabled */ | |
3784 | if ((vmx->exit_reason == EXIT_REASON_MCE_DURING_VMENTRY) | |
3785 | || (vmx->exit_reason == EXIT_REASON_EXCEPTION_NMI | |
3786 | && is_machine_check(exit_intr_info))) | |
3787 | kvm_machine_check(); | |
3788 | ||
20f65983 GN |
3789 | /* We need to handle NMIs before interrupts are enabled */ |
3790 | if ((exit_intr_info & INTR_INFO_INTR_TYPE_MASK) == INTR_TYPE_NMI_INTR && | |
ff9d07a0 ZY |
3791 | (exit_intr_info & INTR_INFO_VALID_MASK)) { |
3792 | kvm_before_handle_nmi(&vmx->vcpu); | |
20f65983 | 3793 | asm("int $2"); |
ff9d07a0 ZY |
3794 | kvm_after_handle_nmi(&vmx->vcpu); |
3795 | } | |
51aa01d1 | 3796 | } |
20f65983 | 3797 | |
51aa01d1 AK |
3798 | static void vmx_recover_nmi_blocking(struct vcpu_vmx *vmx) |
3799 | { | |
3800 | u32 exit_intr_info = vmx->exit_intr_info; | |
3801 | bool unblock_nmi; | |
3802 | u8 vector; | |
3803 | bool idtv_info_valid; | |
3804 | ||
3805 | idtv_info_valid = vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK; | |
20f65983 | 3806 | |
cf393f75 AK |
3807 | if (cpu_has_virtual_nmis()) { |
3808 | unblock_nmi = (exit_intr_info & INTR_INFO_UNBLOCK_NMI) != 0; | |
3809 | vector = exit_intr_info & INTR_INFO_VECTOR_MASK; | |
3810 | /* | |
7b4a25cb | 3811 | * SDM 3: 27.7.1.2 (September 2008) |
cf393f75 AK |
3812 | * Re-set bit "block by NMI" before VM entry if vmexit caused by |
3813 | * a guest IRET fault. | |
7b4a25cb GN |
3814 | * SDM 3: 23.2.2 (September 2008) |
3815 | * Bit 12 is undefined in any of the following cases: | |
3816 | * If the VM exit sets the valid bit in the IDT-vectoring | |
3817 | * information field. | |
3818 | * If the VM exit is due to a double fault. | |
cf393f75 | 3819 | */ |
7b4a25cb GN |
3820 | if ((exit_intr_info & INTR_INFO_VALID_MASK) && unblock_nmi && |
3821 | vector != DF_VECTOR && !idtv_info_valid) | |
cf393f75 AK |
3822 | vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
3823 | GUEST_INTR_STATE_NMI); | |
3b86cd99 JK |
3824 | } else if (unlikely(vmx->soft_vnmi_blocked)) |
3825 | vmx->vnmi_blocked_time += | |
3826 | ktime_to_ns(ktime_sub(ktime_get(), vmx->entry_time)); | |
51aa01d1 AK |
3827 | } |
3828 | ||
625831a3 AK |
3829 | /* |
3830 | * Failure to inject an interrupt should give us the information | |
3831 | * in IDT_VECTORING_INFO_FIELD. However, if the failure occurs | |
3832 | * when fetching the interrupt redirection bitmap in the real-mode | |
3833 | * tss, this doesn't happen. So we do it ourselves. | |
3834 | */ | |
3835 | static void fixup_rmode_irq(struct vcpu_vmx *vmx, u32 *idt_vectoring_info) | |
3836 | { | |
3837 | vmx->rmode.irq.pending = 0; | |
3838 | if (kvm_rip_read(&vmx->vcpu) + 1 != vmx->rmode.irq.rip) | |
3839 | return; | |
3840 | kvm_rip_write(&vmx->vcpu, vmx->rmode.irq.rip); | |
3841 | if (*idt_vectoring_info & VECTORING_INFO_VALID_MASK) { | |
3842 | *idt_vectoring_info &= ~VECTORING_INFO_TYPE_MASK; | |
3843 | *idt_vectoring_info |= INTR_TYPE_EXT_INTR; | |
3844 | return; | |
3845 | } | |
3846 | *idt_vectoring_info = | |
3847 | VECTORING_INFO_VALID_MASK | |
3848 | | INTR_TYPE_EXT_INTR | |
3849 | | vmx->rmode.irq.vector; | |
3850 | } | |
3851 | ||
83422e17 AK |
3852 | static void __vmx_complete_interrupts(struct vcpu_vmx *vmx, |
3853 | u32 idt_vectoring_info, | |
3854 | int instr_len_field, | |
3855 | int error_code_field) | |
51aa01d1 | 3856 | { |
51aa01d1 AK |
3857 | u8 vector; |
3858 | int type; | |
3859 | bool idtv_info_valid; | |
3860 | ||
537b37e2 | 3861 | if (vmx->rmode.irq.pending) |
83422e17 | 3862 | fixup_rmode_irq(vmx, &idt_vectoring_info); |
537b37e2 | 3863 | |
51aa01d1 | 3864 | idtv_info_valid = idt_vectoring_info & VECTORING_INFO_VALID_MASK; |
668f612f | 3865 | |
37b96e98 GN |
3866 | vmx->vcpu.arch.nmi_injected = false; |
3867 | kvm_clear_exception_queue(&vmx->vcpu); | |
3868 | kvm_clear_interrupt_queue(&vmx->vcpu); | |
3869 | ||
3870 | if (!idtv_info_valid) | |
3871 | return; | |
3872 | ||
3842d135 AK |
3873 | kvm_make_request(KVM_REQ_EVENT, &vmx->vcpu); |
3874 | ||
668f612f AK |
3875 | vector = idt_vectoring_info & VECTORING_INFO_VECTOR_MASK; |
3876 | type = idt_vectoring_info & VECTORING_INFO_TYPE_MASK; | |
37b96e98 | 3877 | |
64a7ec06 | 3878 | switch (type) { |
37b96e98 GN |
3879 | case INTR_TYPE_NMI_INTR: |
3880 | vmx->vcpu.arch.nmi_injected = true; | |
668f612f | 3881 | /* |
7b4a25cb | 3882 | * SDM 3: 27.7.1.2 (September 2008) |
37b96e98 GN |
3883 | * Clear bit "block by NMI" before VM entry if a NMI |
3884 | * delivery faulted. | |
668f612f | 3885 | */ |
37b96e98 GN |
3886 | vmcs_clear_bits(GUEST_INTERRUPTIBILITY_INFO, |
3887 | GUEST_INTR_STATE_NMI); | |
3888 | break; | |
37b96e98 | 3889 | case INTR_TYPE_SOFT_EXCEPTION: |
66fd3f7f | 3890 | vmx->vcpu.arch.event_exit_inst_len = |
83422e17 | 3891 | vmcs_read32(instr_len_field); |
66fd3f7f GN |
3892 | /* fall through */ |
3893 | case INTR_TYPE_HARD_EXCEPTION: | |
35920a35 | 3894 | if (idt_vectoring_info & VECTORING_INFO_DELIVER_CODE_MASK) { |
83422e17 | 3895 | u32 err = vmcs_read32(error_code_field); |
37b96e98 | 3896 | kvm_queue_exception_e(&vmx->vcpu, vector, err); |
35920a35 AK |
3897 | } else |
3898 | kvm_queue_exception(&vmx->vcpu, vector); | |
37b96e98 | 3899 | break; |
66fd3f7f GN |
3900 | case INTR_TYPE_SOFT_INTR: |
3901 | vmx->vcpu.arch.event_exit_inst_len = | |
83422e17 | 3902 | vmcs_read32(instr_len_field); |
66fd3f7f | 3903 | /* fall through */ |
37b96e98 | 3904 | case INTR_TYPE_EXT_INTR: |
66fd3f7f GN |
3905 | kvm_queue_interrupt(&vmx->vcpu, vector, |
3906 | type == INTR_TYPE_SOFT_INTR); | |
37b96e98 GN |
3907 | break; |
3908 | default: | |
3909 | break; | |
f7d9238f | 3910 | } |
cf393f75 AK |
3911 | } |
3912 | ||
83422e17 AK |
3913 | static void vmx_complete_interrupts(struct vcpu_vmx *vmx) |
3914 | { | |
3915 | __vmx_complete_interrupts(vmx, vmx->idt_vectoring_info, | |
3916 | VM_EXIT_INSTRUCTION_LEN, | |
3917 | IDT_VECTORING_ERROR_CODE); | |
3918 | } | |
3919 | ||
b463a6f7 AK |
3920 | static void vmx_cancel_injection(struct kvm_vcpu *vcpu) |
3921 | { | |
3922 | __vmx_complete_interrupts(to_vmx(vcpu), | |
3923 | vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), | |
3924 | VM_ENTRY_INSTRUCTION_LEN, | |
3925 | VM_ENTRY_EXCEPTION_ERROR_CODE); | |
3926 | ||
3927 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); | |
3928 | } | |
3929 | ||
c801949d AK |
3930 | #ifdef CONFIG_X86_64 |
3931 | #define R "r" | |
3932 | #define Q "q" | |
3933 | #else | |
3934 | #define R "e" | |
3935 | #define Q "l" | |
3936 | #endif | |
3937 | ||
851ba692 | 3938 | static void vmx_vcpu_run(struct kvm_vcpu *vcpu) |
6aa8b732 | 3939 | { |
a2fa3e9f | 3940 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
e6adf283 | 3941 | |
3b86cd99 JK |
3942 | /* Record the guest's net vcpu time for enforced NMI injections. */ |
3943 | if (unlikely(!cpu_has_virtual_nmis() && vmx->soft_vnmi_blocked)) | |
3944 | vmx->entry_time = ktime_get(); | |
3945 | ||
80ced186 MG |
3946 | /* Don't enter VMX if guest state is invalid, let the exit handler |
3947 | start emulation until we arrive back to a valid state */ | |
3948 | if (vmx->emulation_required && emulate_invalid_guest_state) | |
a89a8fb9 | 3949 | return; |
a89a8fb9 | 3950 | |
5fdbf976 MT |
3951 | if (test_bit(VCPU_REGS_RSP, (unsigned long *)&vcpu->arch.regs_dirty)) |
3952 | vmcs_writel(GUEST_RSP, vcpu->arch.regs[VCPU_REGS_RSP]); | |
3953 | if (test_bit(VCPU_REGS_RIP, (unsigned long *)&vcpu->arch.regs_dirty)) | |
3954 | vmcs_writel(GUEST_RIP, vcpu->arch.regs[VCPU_REGS_RIP]); | |
3955 | ||
787ff736 GN |
3956 | /* When single-stepping over STI and MOV SS, we must clear the |
3957 | * corresponding interruptibility bits in the guest state. Otherwise | |
3958 | * vmentry fails as it then expects bit 14 (BS) in pending debug | |
3959 | * exceptions being set, but that's not correct for the guest debugging | |
3960 | * case. */ | |
3961 | if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) | |
3962 | vmx_set_interrupt_shadow(vcpu, 0); | |
3963 | ||
d77c26fc | 3964 | asm( |
6aa8b732 | 3965 | /* Store host registers */ |
c801949d AK |
3966 | "push %%"R"dx; push %%"R"bp;" |
3967 | "push %%"R"cx \n\t" | |
313dbd49 AK |
3968 | "cmp %%"R"sp, %c[host_rsp](%0) \n\t" |
3969 | "je 1f \n\t" | |
3970 | "mov %%"R"sp, %c[host_rsp](%0) \n\t" | |
4ecac3fd | 3971 | __ex(ASM_VMX_VMWRITE_RSP_RDX) "\n\t" |
313dbd49 | 3972 | "1: \n\t" |
d3edefc0 AK |
3973 | /* Reload cr2 if changed */ |
3974 | "mov %c[cr2](%0), %%"R"ax \n\t" | |
3975 | "mov %%cr2, %%"R"dx \n\t" | |
3976 | "cmp %%"R"ax, %%"R"dx \n\t" | |
3977 | "je 2f \n\t" | |
3978 | "mov %%"R"ax, %%cr2 \n\t" | |
3979 | "2: \n\t" | |
6aa8b732 | 3980 | /* Check if vmlaunch of vmresume is needed */ |
e08aa78a | 3981 | "cmpl $0, %c[launched](%0) \n\t" |
6aa8b732 | 3982 | /* Load guest registers. Don't clobber flags. */ |
c801949d AK |
3983 | "mov %c[rax](%0), %%"R"ax \n\t" |
3984 | "mov %c[rbx](%0), %%"R"bx \n\t" | |
3985 | "mov %c[rdx](%0), %%"R"dx \n\t" | |
3986 | "mov %c[rsi](%0), %%"R"si \n\t" | |
3987 | "mov %c[rdi](%0), %%"R"di \n\t" | |
3988 | "mov %c[rbp](%0), %%"R"bp \n\t" | |
05b3e0c2 | 3989 | #ifdef CONFIG_X86_64 |
e08aa78a AK |
3990 | "mov %c[r8](%0), %%r8 \n\t" |
3991 | "mov %c[r9](%0), %%r9 \n\t" | |
3992 | "mov %c[r10](%0), %%r10 \n\t" | |
3993 | "mov %c[r11](%0), %%r11 \n\t" | |
3994 | "mov %c[r12](%0), %%r12 \n\t" | |
3995 | "mov %c[r13](%0), %%r13 \n\t" | |
3996 | "mov %c[r14](%0), %%r14 \n\t" | |
3997 | "mov %c[r15](%0), %%r15 \n\t" | |
6aa8b732 | 3998 | #endif |
c801949d AK |
3999 | "mov %c[rcx](%0), %%"R"cx \n\t" /* kills %0 (ecx) */ |
4000 | ||
6aa8b732 | 4001 | /* Enter guest mode */ |
cd2276a7 | 4002 | "jne .Llaunched \n\t" |
4ecac3fd | 4003 | __ex(ASM_VMX_VMLAUNCH) "\n\t" |
cd2276a7 | 4004 | "jmp .Lkvm_vmx_return \n\t" |
4ecac3fd | 4005 | ".Llaunched: " __ex(ASM_VMX_VMRESUME) "\n\t" |
cd2276a7 | 4006 | ".Lkvm_vmx_return: " |
6aa8b732 | 4007 | /* Save guest registers, load host registers, keep flags */ |
c801949d AK |
4008 | "xchg %0, (%%"R"sp) \n\t" |
4009 | "mov %%"R"ax, %c[rax](%0) \n\t" | |
4010 | "mov %%"R"bx, %c[rbx](%0) \n\t" | |
4011 | "push"Q" (%%"R"sp); pop"Q" %c[rcx](%0) \n\t" | |
4012 | "mov %%"R"dx, %c[rdx](%0) \n\t" | |
4013 | "mov %%"R"si, %c[rsi](%0) \n\t" | |
4014 | "mov %%"R"di, %c[rdi](%0) \n\t" | |
4015 | "mov %%"R"bp, %c[rbp](%0) \n\t" | |
05b3e0c2 | 4016 | #ifdef CONFIG_X86_64 |
e08aa78a AK |
4017 | "mov %%r8, %c[r8](%0) \n\t" |
4018 | "mov %%r9, %c[r9](%0) \n\t" | |
4019 | "mov %%r10, %c[r10](%0) \n\t" | |
4020 | "mov %%r11, %c[r11](%0) \n\t" | |
4021 | "mov %%r12, %c[r12](%0) \n\t" | |
4022 | "mov %%r13, %c[r13](%0) \n\t" | |
4023 | "mov %%r14, %c[r14](%0) \n\t" | |
4024 | "mov %%r15, %c[r15](%0) \n\t" | |
6aa8b732 | 4025 | #endif |
c801949d AK |
4026 | "mov %%cr2, %%"R"ax \n\t" |
4027 | "mov %%"R"ax, %c[cr2](%0) \n\t" | |
4028 | ||
4029 | "pop %%"R"bp; pop %%"R"bp; pop %%"R"dx \n\t" | |
e08aa78a AK |
4030 | "setbe %c[fail](%0) \n\t" |
4031 | : : "c"(vmx), "d"((unsigned long)HOST_RSP), | |
4032 | [launched]"i"(offsetof(struct vcpu_vmx, launched)), | |
4033 | [fail]"i"(offsetof(struct vcpu_vmx, fail)), | |
313dbd49 | 4034 | [host_rsp]"i"(offsetof(struct vcpu_vmx, host_rsp)), |
ad312c7c ZX |
4035 | [rax]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RAX])), |
4036 | [rbx]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RBX])), | |
4037 | [rcx]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RCX])), | |
4038 | [rdx]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RDX])), | |
4039 | [rsi]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RSI])), | |
4040 | [rdi]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RDI])), | |
4041 | [rbp]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RBP])), | |
05b3e0c2 | 4042 | #ifdef CONFIG_X86_64 |
ad312c7c ZX |
4043 | [r8]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R8])), |
4044 | [r9]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R9])), | |
4045 | [r10]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R10])), | |
4046 | [r11]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R11])), | |
4047 | [r12]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R12])), | |
4048 | [r13]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R13])), | |
4049 | [r14]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R14])), | |
4050 | [r15]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R15])), | |
6aa8b732 | 4051 | #endif |
ad312c7c | 4052 | [cr2]"i"(offsetof(struct vcpu_vmx, vcpu.arch.cr2)) |
c2036300 | 4053 | : "cc", "memory" |
c801949d | 4054 | , R"bx", R"di", R"si" |
c2036300 | 4055 | #ifdef CONFIG_X86_64 |
c2036300 LV |
4056 | , "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" |
4057 | #endif | |
4058 | ); | |
6aa8b732 | 4059 | |
6de4f3ad AK |
4060 | vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP) |
4061 | | (1 << VCPU_EXREG_PDPTR)); | |
5fdbf976 MT |
4062 | vcpu->arch.regs_dirty = 0; |
4063 | ||
1155f76a AK |
4064 | vmx->idt_vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD); |
4065 | ||
d77c26fc | 4066 | asm("mov %0, %%ds; mov %0, %%es" : : "r"(__USER_DS)); |
15ad7146 | 4067 | vmx->launched = 1; |
1b6269db | 4068 | |
51aa01d1 AK |
4069 | vmx->exit_reason = vmcs_read32(VM_EXIT_REASON); |
4070 | vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); | |
4071 | ||
4072 | vmx_complete_atomic_exit(vmx); | |
4073 | vmx_recover_nmi_blocking(vmx); | |
cf393f75 | 4074 | vmx_complete_interrupts(vmx); |
6aa8b732 AK |
4075 | } |
4076 | ||
c801949d AK |
4077 | #undef R |
4078 | #undef Q | |
4079 | ||
6aa8b732 AK |
4080 | static void vmx_free_vmcs(struct kvm_vcpu *vcpu) |
4081 | { | |
a2fa3e9f GH |
4082 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
4083 | ||
4084 | if (vmx->vmcs) { | |
543e4243 | 4085 | vcpu_clear(vmx); |
a2fa3e9f GH |
4086 | free_vmcs(vmx->vmcs); |
4087 | vmx->vmcs = NULL; | |
6aa8b732 AK |
4088 | } |
4089 | } | |
4090 | ||
4091 | static void vmx_free_vcpu(struct kvm_vcpu *vcpu) | |
4092 | { | |
fb3f0f51 RR |
4093 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
4094 | ||
cdbecfc3 | 4095 | free_vpid(vmx); |
6aa8b732 | 4096 | vmx_free_vmcs(vcpu); |
fb3f0f51 RR |
4097 | kfree(vmx->guest_msrs); |
4098 | kvm_vcpu_uninit(vcpu); | |
a4770347 | 4099 | kmem_cache_free(kvm_vcpu_cache, vmx); |
6aa8b732 AK |
4100 | } |
4101 | ||
4610c9cc DX |
4102 | static inline void vmcs_init(struct vmcs *vmcs) |
4103 | { | |
4104 | u64 phys_addr = __pa(per_cpu(vmxarea, raw_smp_processor_id())); | |
4105 | ||
4106 | if (!vmm_exclusive) | |
4107 | kvm_cpu_vmxon(phys_addr); | |
4108 | ||
4109 | vmcs_clear(vmcs); | |
4110 | ||
4111 | if (!vmm_exclusive) | |
4112 | kvm_cpu_vmxoff(); | |
4113 | } | |
4114 | ||
fb3f0f51 | 4115 | static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id) |
6aa8b732 | 4116 | { |
fb3f0f51 | 4117 | int err; |
c16f862d | 4118 | struct vcpu_vmx *vmx = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL); |
15ad7146 | 4119 | int cpu; |
6aa8b732 | 4120 | |
a2fa3e9f | 4121 | if (!vmx) |
fb3f0f51 RR |
4122 | return ERR_PTR(-ENOMEM); |
4123 | ||
2384d2b3 SY |
4124 | allocate_vpid(vmx); |
4125 | ||
fb3f0f51 RR |
4126 | err = kvm_vcpu_init(&vmx->vcpu, kvm, id); |
4127 | if (err) | |
4128 | goto free_vcpu; | |
965b58a5 | 4129 | |
a2fa3e9f | 4130 | vmx->guest_msrs = kmalloc(PAGE_SIZE, GFP_KERNEL); |
fb3f0f51 RR |
4131 | if (!vmx->guest_msrs) { |
4132 | err = -ENOMEM; | |
4133 | goto uninit_vcpu; | |
4134 | } | |
965b58a5 | 4135 | |
a2fa3e9f GH |
4136 | vmx->vmcs = alloc_vmcs(); |
4137 | if (!vmx->vmcs) | |
fb3f0f51 | 4138 | goto free_msrs; |
a2fa3e9f | 4139 | |
4610c9cc | 4140 | vmcs_init(vmx->vmcs); |
a2fa3e9f | 4141 | |
15ad7146 AK |
4142 | cpu = get_cpu(); |
4143 | vmx_vcpu_load(&vmx->vcpu, cpu); | |
e48672fa | 4144 | vmx->vcpu.cpu = cpu; |
8b9cf98c | 4145 | err = vmx_vcpu_setup(vmx); |
fb3f0f51 | 4146 | vmx_vcpu_put(&vmx->vcpu); |
15ad7146 | 4147 | put_cpu(); |
fb3f0f51 RR |
4148 | if (err) |
4149 | goto free_vmcs; | |
5e4a0b3c MT |
4150 | if (vm_need_virtualize_apic_accesses(kvm)) |
4151 | if (alloc_apic_access_page(kvm) != 0) | |
4152 | goto free_vmcs; | |
fb3f0f51 | 4153 | |
b927a3ce SY |
4154 | if (enable_ept) { |
4155 | if (!kvm->arch.ept_identity_map_addr) | |
4156 | kvm->arch.ept_identity_map_addr = | |
4157 | VMX_EPT_IDENTITY_PAGETABLE_ADDR; | |
b7ebfb05 SY |
4158 | if (alloc_identity_pagetable(kvm) != 0) |
4159 | goto free_vmcs; | |
b927a3ce | 4160 | } |
b7ebfb05 | 4161 | |
fb3f0f51 RR |
4162 | return &vmx->vcpu; |
4163 | ||
4164 | free_vmcs: | |
4165 | free_vmcs(vmx->vmcs); | |
4166 | free_msrs: | |
fb3f0f51 RR |
4167 | kfree(vmx->guest_msrs); |
4168 | uninit_vcpu: | |
4169 | kvm_vcpu_uninit(&vmx->vcpu); | |
4170 | free_vcpu: | |
cdbecfc3 | 4171 | free_vpid(vmx); |
a4770347 | 4172 | kmem_cache_free(kvm_vcpu_cache, vmx); |
fb3f0f51 | 4173 | return ERR_PTR(err); |
6aa8b732 AK |
4174 | } |
4175 | ||
002c7f7c YS |
4176 | static void __init vmx_check_processor_compat(void *rtn) |
4177 | { | |
4178 | struct vmcs_config vmcs_conf; | |
4179 | ||
4180 | *(int *)rtn = 0; | |
4181 | if (setup_vmcs_config(&vmcs_conf) < 0) | |
4182 | *(int *)rtn = -EIO; | |
4183 | if (memcmp(&vmcs_config, &vmcs_conf, sizeof(struct vmcs_config)) != 0) { | |
4184 | printk(KERN_ERR "kvm: CPU %d feature inconsistency!\n", | |
4185 | smp_processor_id()); | |
4186 | *(int *)rtn = -EIO; | |
4187 | } | |
4188 | } | |
4189 | ||
67253af5 SY |
4190 | static int get_ept_level(void) |
4191 | { | |
4192 | return VMX_EPT_DEFAULT_GAW + 1; | |
4193 | } | |
4194 | ||
4b12f0de | 4195 | static u64 vmx_get_mt_mask(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio) |
64d4d521 | 4196 | { |
4b12f0de SY |
4197 | u64 ret; |
4198 | ||
522c68c4 SY |
4199 | /* For VT-d and EPT combination |
4200 | * 1. MMIO: always map as UC | |
4201 | * 2. EPT with VT-d: | |
4202 | * a. VT-d without snooping control feature: can't guarantee the | |
4203 | * result, try to trust guest. | |
4204 | * b. VT-d with snooping control feature: snooping control feature of | |
4205 | * VT-d engine can guarantee the cache correctness. Just set it | |
4206 | * to WB to keep consistent with host. So the same as item 3. | |
a19a6d11 | 4207 | * 3. EPT without VT-d: always map as WB and set IPAT=1 to keep |
522c68c4 SY |
4208 | * consistent with host MTRR |
4209 | */ | |
4b12f0de SY |
4210 | if (is_mmio) |
4211 | ret = MTRR_TYPE_UNCACHABLE << VMX_EPT_MT_EPTE_SHIFT; | |
522c68c4 SY |
4212 | else if (vcpu->kvm->arch.iommu_domain && |
4213 | !(vcpu->kvm->arch.iommu_flags & KVM_IOMMU_CACHE_COHERENCY)) | |
4214 | ret = kvm_get_guest_memory_type(vcpu, gfn) << | |
4215 | VMX_EPT_MT_EPTE_SHIFT; | |
4b12f0de | 4216 | else |
522c68c4 | 4217 | ret = (MTRR_TYPE_WRBACK << VMX_EPT_MT_EPTE_SHIFT) |
a19a6d11 | 4218 | | VMX_EPT_IPAT_BIT; |
4b12f0de SY |
4219 | |
4220 | return ret; | |
64d4d521 SY |
4221 | } |
4222 | ||
f4c9e87c AK |
4223 | #define _ER(x) { EXIT_REASON_##x, #x } |
4224 | ||
229456fc | 4225 | static const struct trace_print_flags vmx_exit_reasons_str[] = { |
f4c9e87c AK |
4226 | _ER(EXCEPTION_NMI), |
4227 | _ER(EXTERNAL_INTERRUPT), | |
4228 | _ER(TRIPLE_FAULT), | |
4229 | _ER(PENDING_INTERRUPT), | |
4230 | _ER(NMI_WINDOW), | |
4231 | _ER(TASK_SWITCH), | |
4232 | _ER(CPUID), | |
4233 | _ER(HLT), | |
4234 | _ER(INVLPG), | |
4235 | _ER(RDPMC), | |
4236 | _ER(RDTSC), | |
4237 | _ER(VMCALL), | |
4238 | _ER(VMCLEAR), | |
4239 | _ER(VMLAUNCH), | |
4240 | _ER(VMPTRLD), | |
4241 | _ER(VMPTRST), | |
4242 | _ER(VMREAD), | |
4243 | _ER(VMRESUME), | |
4244 | _ER(VMWRITE), | |
4245 | _ER(VMOFF), | |
4246 | _ER(VMON), | |
4247 | _ER(CR_ACCESS), | |
4248 | _ER(DR_ACCESS), | |
4249 | _ER(IO_INSTRUCTION), | |
4250 | _ER(MSR_READ), | |
4251 | _ER(MSR_WRITE), | |
4252 | _ER(MWAIT_INSTRUCTION), | |
4253 | _ER(MONITOR_INSTRUCTION), | |
4254 | _ER(PAUSE_INSTRUCTION), | |
4255 | _ER(MCE_DURING_VMENTRY), | |
4256 | _ER(TPR_BELOW_THRESHOLD), | |
4257 | _ER(APIC_ACCESS), | |
4258 | _ER(EPT_VIOLATION), | |
4259 | _ER(EPT_MISCONFIG), | |
4260 | _ER(WBINVD), | |
229456fc MT |
4261 | { -1, NULL } |
4262 | }; | |
4263 | ||
f4c9e87c AK |
4264 | #undef _ER |
4265 | ||
17cc3935 | 4266 | static int vmx_get_lpage_level(void) |
344f414f | 4267 | { |
878403b7 SY |
4268 | if (enable_ept && !cpu_has_vmx_ept_1g_page()) |
4269 | return PT_DIRECTORY_LEVEL; | |
4270 | else | |
4271 | /* For shadow and EPT supported 1GB page */ | |
4272 | return PT_PDPE_LEVEL; | |
344f414f JR |
4273 | } |
4274 | ||
4e47c7a6 SY |
4275 | static inline u32 bit(int bitno) |
4276 | { | |
4277 | return 1 << (bitno & 31); | |
4278 | } | |
4279 | ||
0e851880 SY |
4280 | static void vmx_cpuid_update(struct kvm_vcpu *vcpu) |
4281 | { | |
4e47c7a6 SY |
4282 | struct kvm_cpuid_entry2 *best; |
4283 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
4284 | u32 exec_control; | |
4285 | ||
4286 | vmx->rdtscp_enabled = false; | |
4287 | if (vmx_rdtscp_supported()) { | |
4288 | exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL); | |
4289 | if (exec_control & SECONDARY_EXEC_RDTSCP) { | |
4290 | best = kvm_find_cpuid_entry(vcpu, 0x80000001, 0); | |
4291 | if (best && (best->edx & bit(X86_FEATURE_RDTSCP))) | |
4292 | vmx->rdtscp_enabled = true; | |
4293 | else { | |
4294 | exec_control &= ~SECONDARY_EXEC_RDTSCP; | |
4295 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, | |
4296 | exec_control); | |
4297 | } | |
4298 | } | |
4299 | } | |
0e851880 SY |
4300 | } |
4301 | ||
d4330ef2 JR |
4302 | static void vmx_set_supported_cpuid(u32 func, struct kvm_cpuid_entry2 *entry) |
4303 | { | |
4304 | } | |
4305 | ||
cbdd1bea | 4306 | static struct kvm_x86_ops vmx_x86_ops = { |
6aa8b732 AK |
4307 | .cpu_has_kvm_support = cpu_has_kvm_support, |
4308 | .disabled_by_bios = vmx_disabled_by_bios, | |
4309 | .hardware_setup = hardware_setup, | |
4310 | .hardware_unsetup = hardware_unsetup, | |
002c7f7c | 4311 | .check_processor_compatibility = vmx_check_processor_compat, |
6aa8b732 AK |
4312 | .hardware_enable = hardware_enable, |
4313 | .hardware_disable = hardware_disable, | |
04547156 | 4314 | .cpu_has_accelerated_tpr = report_flexpriority, |
6aa8b732 AK |
4315 | |
4316 | .vcpu_create = vmx_create_vcpu, | |
4317 | .vcpu_free = vmx_free_vcpu, | |
04d2cc77 | 4318 | .vcpu_reset = vmx_vcpu_reset, |
6aa8b732 | 4319 | |
04d2cc77 | 4320 | .prepare_guest_switch = vmx_save_host_state, |
6aa8b732 AK |
4321 | .vcpu_load = vmx_vcpu_load, |
4322 | .vcpu_put = vmx_vcpu_put, | |
4323 | ||
4324 | .set_guest_debug = set_guest_debug, | |
4325 | .get_msr = vmx_get_msr, | |
4326 | .set_msr = vmx_set_msr, | |
4327 | .get_segment_base = vmx_get_segment_base, | |
4328 | .get_segment = vmx_get_segment, | |
4329 | .set_segment = vmx_set_segment, | |
2e4d2653 | 4330 | .get_cpl = vmx_get_cpl, |
6aa8b732 | 4331 | .get_cs_db_l_bits = vmx_get_cs_db_l_bits, |
e8467fda | 4332 | .decache_cr0_guest_bits = vmx_decache_cr0_guest_bits, |
25c4c276 | 4333 | .decache_cr4_guest_bits = vmx_decache_cr4_guest_bits, |
6aa8b732 | 4334 | .set_cr0 = vmx_set_cr0, |
6aa8b732 AK |
4335 | .set_cr3 = vmx_set_cr3, |
4336 | .set_cr4 = vmx_set_cr4, | |
6aa8b732 | 4337 | .set_efer = vmx_set_efer, |
6aa8b732 AK |
4338 | .get_idt = vmx_get_idt, |
4339 | .set_idt = vmx_set_idt, | |
4340 | .get_gdt = vmx_get_gdt, | |
4341 | .set_gdt = vmx_set_gdt, | |
020df079 | 4342 | .set_dr7 = vmx_set_dr7, |
5fdbf976 | 4343 | .cache_reg = vmx_cache_reg, |
6aa8b732 AK |
4344 | .get_rflags = vmx_get_rflags, |
4345 | .set_rflags = vmx_set_rflags, | |
ebcbab4c | 4346 | .fpu_activate = vmx_fpu_activate, |
02daab21 | 4347 | .fpu_deactivate = vmx_fpu_deactivate, |
6aa8b732 AK |
4348 | |
4349 | .tlb_flush = vmx_flush_tlb, | |
6aa8b732 | 4350 | |
6aa8b732 | 4351 | .run = vmx_vcpu_run, |
6062d012 | 4352 | .handle_exit = vmx_handle_exit, |
6aa8b732 | 4353 | .skip_emulated_instruction = skip_emulated_instruction, |
2809f5d2 GC |
4354 | .set_interrupt_shadow = vmx_set_interrupt_shadow, |
4355 | .get_interrupt_shadow = vmx_get_interrupt_shadow, | |
102d8325 | 4356 | .patch_hypercall = vmx_patch_hypercall, |
2a8067f1 | 4357 | .set_irq = vmx_inject_irq, |
95ba8273 | 4358 | .set_nmi = vmx_inject_nmi, |
298101da | 4359 | .queue_exception = vmx_queue_exception, |
b463a6f7 | 4360 | .cancel_injection = vmx_cancel_injection, |
78646121 | 4361 | .interrupt_allowed = vmx_interrupt_allowed, |
95ba8273 | 4362 | .nmi_allowed = vmx_nmi_allowed, |
3cfc3092 JK |
4363 | .get_nmi_mask = vmx_get_nmi_mask, |
4364 | .set_nmi_mask = vmx_set_nmi_mask, | |
95ba8273 GN |
4365 | .enable_nmi_window = enable_nmi_window, |
4366 | .enable_irq_window = enable_irq_window, | |
4367 | .update_cr8_intercept = update_cr8_intercept, | |
95ba8273 | 4368 | |
cbc94022 | 4369 | .set_tss_addr = vmx_set_tss_addr, |
67253af5 | 4370 | .get_tdp_level = get_ept_level, |
4b12f0de | 4371 | .get_mt_mask = vmx_get_mt_mask, |
229456fc MT |
4372 | |
4373 | .exit_reasons_str = vmx_exit_reasons_str, | |
17cc3935 | 4374 | .get_lpage_level = vmx_get_lpage_level, |
0e851880 SY |
4375 | |
4376 | .cpuid_update = vmx_cpuid_update, | |
4e47c7a6 SY |
4377 | |
4378 | .rdtscp_supported = vmx_rdtscp_supported, | |
d4330ef2 JR |
4379 | |
4380 | .set_supported_cpuid = vmx_set_supported_cpuid, | |
f5f48ee1 SY |
4381 | |
4382 | .has_wbinvd_exit = cpu_has_vmx_wbinvd_exit, | |
99e3e30a ZA |
4383 | |
4384 | .write_tsc_offset = vmx_write_tsc_offset, | |
e48672fa | 4385 | .adjust_tsc_offset = vmx_adjust_tsc_offset, |
1c97f0a0 JR |
4386 | |
4387 | .set_tdp_cr3 = vmx_set_cr3, | |
6aa8b732 AK |
4388 | }; |
4389 | ||
4390 | static int __init vmx_init(void) | |
4391 | { | |
26bb0981 AK |
4392 | int r, i; |
4393 | ||
4394 | rdmsrl_safe(MSR_EFER, &host_efer); | |
4395 | ||
4396 | for (i = 0; i < NR_VMX_MSR; ++i) | |
4397 | kvm_define_shared_msr(i, vmx_msr_index[i]); | |
fdef3ad1 | 4398 | |
3e7c73e9 | 4399 | vmx_io_bitmap_a = (unsigned long *)__get_free_page(GFP_KERNEL); |
fdef3ad1 HQ |
4400 | if (!vmx_io_bitmap_a) |
4401 | return -ENOMEM; | |
4402 | ||
3e7c73e9 | 4403 | vmx_io_bitmap_b = (unsigned long *)__get_free_page(GFP_KERNEL); |
fdef3ad1 HQ |
4404 | if (!vmx_io_bitmap_b) { |
4405 | r = -ENOMEM; | |
4406 | goto out; | |
4407 | } | |
4408 | ||
5897297b AK |
4409 | vmx_msr_bitmap_legacy = (unsigned long *)__get_free_page(GFP_KERNEL); |
4410 | if (!vmx_msr_bitmap_legacy) { | |
25c5f225 SY |
4411 | r = -ENOMEM; |
4412 | goto out1; | |
4413 | } | |
4414 | ||
5897297b AK |
4415 | vmx_msr_bitmap_longmode = (unsigned long *)__get_free_page(GFP_KERNEL); |
4416 | if (!vmx_msr_bitmap_longmode) { | |
4417 | r = -ENOMEM; | |
4418 | goto out2; | |
4419 | } | |
4420 | ||
fdef3ad1 HQ |
4421 | /* |
4422 | * Allow direct access to the PC debug port (it is often used for I/O | |
4423 | * delays, but the vmexits simply slow things down). | |
4424 | */ | |
3e7c73e9 AK |
4425 | memset(vmx_io_bitmap_a, 0xff, PAGE_SIZE); |
4426 | clear_bit(0x80, vmx_io_bitmap_a); | |
fdef3ad1 | 4427 | |
3e7c73e9 | 4428 | memset(vmx_io_bitmap_b, 0xff, PAGE_SIZE); |
fdef3ad1 | 4429 | |
5897297b AK |
4430 | memset(vmx_msr_bitmap_legacy, 0xff, PAGE_SIZE); |
4431 | memset(vmx_msr_bitmap_longmode, 0xff, PAGE_SIZE); | |
25c5f225 | 4432 | |
2384d2b3 SY |
4433 | set_bit(0, vmx_vpid_bitmap); /* 0 is reserved for host */ |
4434 | ||
0ee75bea AK |
4435 | r = kvm_init(&vmx_x86_ops, sizeof(struct vcpu_vmx), |
4436 | __alignof__(struct vcpu_vmx), THIS_MODULE); | |
fdef3ad1 | 4437 | if (r) |
5897297b | 4438 | goto out3; |
25c5f225 | 4439 | |
5897297b AK |
4440 | vmx_disable_intercept_for_msr(MSR_FS_BASE, false); |
4441 | vmx_disable_intercept_for_msr(MSR_GS_BASE, false); | |
4442 | vmx_disable_intercept_for_msr(MSR_KERNEL_GS_BASE, true); | |
4443 | vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_CS, false); | |
4444 | vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_ESP, false); | |
4445 | vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_EIP, false); | |
fdef3ad1 | 4446 | |
089d034e | 4447 | if (enable_ept) { |
1439442c | 4448 | bypass_guest_pf = 0; |
5fdbcb9d | 4449 | kvm_mmu_set_base_ptes(VMX_EPT_READABLE_MASK | |
2aaf69dc | 4450 | VMX_EPT_WRITABLE_MASK); |
534e38b4 | 4451 | kvm_mmu_set_mask_ptes(0ull, 0ull, 0ull, 0ull, |
4b12f0de | 4452 | VMX_EPT_EXECUTABLE_MASK); |
5fdbcb9d SY |
4453 | kvm_enable_tdp(); |
4454 | } else | |
4455 | kvm_disable_tdp(); | |
1439442c | 4456 | |
c7addb90 AK |
4457 | if (bypass_guest_pf) |
4458 | kvm_mmu_set_nonpresent_ptes(~0xffeull, 0ull); | |
4459 | ||
fdef3ad1 HQ |
4460 | return 0; |
4461 | ||
5897297b AK |
4462 | out3: |
4463 | free_page((unsigned long)vmx_msr_bitmap_longmode); | |
25c5f225 | 4464 | out2: |
5897297b | 4465 | free_page((unsigned long)vmx_msr_bitmap_legacy); |
fdef3ad1 | 4466 | out1: |
3e7c73e9 | 4467 | free_page((unsigned long)vmx_io_bitmap_b); |
fdef3ad1 | 4468 | out: |
3e7c73e9 | 4469 | free_page((unsigned long)vmx_io_bitmap_a); |
fdef3ad1 | 4470 | return r; |
6aa8b732 AK |
4471 | } |
4472 | ||
4473 | static void __exit vmx_exit(void) | |
4474 | { | |
5897297b AK |
4475 | free_page((unsigned long)vmx_msr_bitmap_legacy); |
4476 | free_page((unsigned long)vmx_msr_bitmap_longmode); | |
3e7c73e9 AK |
4477 | free_page((unsigned long)vmx_io_bitmap_b); |
4478 | free_page((unsigned long)vmx_io_bitmap_a); | |
fdef3ad1 | 4479 | |
cb498ea2 | 4480 | kvm_exit(); |
6aa8b732 AK |
4481 | } |
4482 | ||
4483 | module_init(vmx_init) | |
4484 | module_exit(vmx_exit) |