]>
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. | |
9611c187 | 8 | * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
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" |
00b27a3e | 21 | #include "cpuid.h" |
e495606d | 22 | |
edf88417 | 23 | #include <linux/kvm_host.h> |
6aa8b732 | 24 | #include <linux/module.h> |
9d8f549d | 25 | #include <linux/kernel.h> |
6aa8b732 AK |
26 | #include <linux/mm.h> |
27 | #include <linux/highmem.h> | |
e8edc6e0 | 28 | #include <linux/sched.h> |
c7addb90 | 29 | #include <linux/moduleparam.h> |
e9bda3b3 | 30 | #include <linux/mod_devicetable.h> |
229456fc | 31 | #include <linux/ftrace_event.h> |
5a0e3ad6 | 32 | #include <linux/slab.h> |
cafd6659 | 33 | #include <linux/tboot.h> |
5fdbf976 | 34 | #include "kvm_cache_regs.h" |
35920a35 | 35 | #include "x86.h" |
e495606d | 36 | |
6aa8b732 | 37 | #include <asm/io.h> |
3b3be0d1 | 38 | #include <asm/desc.h> |
13673a90 | 39 | #include <asm/vmx.h> |
6210e37b | 40 | #include <asm/virtext.h> |
a0861c02 | 41 | #include <asm/mce.h> |
2acf923e DC |
42 | #include <asm/i387.h> |
43 | #include <asm/xcr.h> | |
d7cd9796 | 44 | #include <asm/perf_event.h> |
6aa8b732 | 45 | |
229456fc MT |
46 | #include "trace.h" |
47 | ||
4ecac3fd | 48 | #define __ex(x) __kvm_handle_fault_on_reboot(x) |
5e520e62 AK |
49 | #define __ex_clear(x, reg) \ |
50 | ____kvm_handle_fault_on_reboot(x, "xor " reg " , " reg) | |
4ecac3fd | 51 | |
6aa8b732 AK |
52 | MODULE_AUTHOR("Qumranet"); |
53 | MODULE_LICENSE("GPL"); | |
54 | ||
e9bda3b3 JT |
55 | static const struct x86_cpu_id vmx_cpu_id[] = { |
56 | X86_FEATURE_MATCH(X86_FEATURE_VMX), | |
57 | {} | |
58 | }; | |
59 | MODULE_DEVICE_TABLE(x86cpu, vmx_cpu_id); | |
60 | ||
476bc001 | 61 | static bool __read_mostly enable_vpid = 1; |
736caefe | 62 | module_param_named(vpid, enable_vpid, bool, 0444); |
2384d2b3 | 63 | |
476bc001 | 64 | static bool __read_mostly flexpriority_enabled = 1; |
736caefe | 65 | module_param_named(flexpriority, flexpriority_enabled, bool, S_IRUGO); |
4c9fc8ef | 66 | |
476bc001 | 67 | static bool __read_mostly enable_ept = 1; |
736caefe | 68 | module_param_named(ept, enable_ept, bool, S_IRUGO); |
d56f546d | 69 | |
476bc001 | 70 | static bool __read_mostly enable_unrestricted_guest = 1; |
3a624e29 NK |
71 | module_param_named(unrestricted_guest, |
72 | enable_unrestricted_guest, bool, S_IRUGO); | |
73 | ||
83c3a331 XH |
74 | static bool __read_mostly enable_ept_ad_bits = 1; |
75 | module_param_named(eptad, enable_ept_ad_bits, bool, S_IRUGO); | |
76 | ||
a27685c3 | 77 | static bool __read_mostly emulate_invalid_guest_state = true; |
c1f8bc04 | 78 | module_param(emulate_invalid_guest_state, bool, S_IRUGO); |
04fa4d32 | 79 | |
476bc001 | 80 | static bool __read_mostly vmm_exclusive = 1; |
b923e62e DX |
81 | module_param(vmm_exclusive, bool, S_IRUGO); |
82 | ||
476bc001 | 83 | static bool __read_mostly fasteoi = 1; |
58fbbf26 KT |
84 | module_param(fasteoi, bool, S_IRUGO); |
85 | ||
801d3424 NHE |
86 | /* |
87 | * If nested=1, nested virtualization is supported, i.e., guests may use | |
88 | * VMX and be a hypervisor for its own guests. If nested=0, guests may not | |
89 | * use VMX instructions. | |
90 | */ | |
476bc001 | 91 | static bool __read_mostly nested = 0; |
801d3424 NHE |
92 | module_param(nested, bool, S_IRUGO); |
93 | ||
cdc0e244 AK |
94 | #define KVM_GUEST_CR0_MASK_UNRESTRICTED_GUEST \ |
95 | (X86_CR0_WP | X86_CR0_NE | X86_CR0_NW | X86_CR0_CD) | |
96 | #define KVM_GUEST_CR0_MASK \ | |
97 | (KVM_GUEST_CR0_MASK_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE) | |
98 | #define KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST \ | |
81231c69 | 99 | (X86_CR0_WP | X86_CR0_NE) |
cdc0e244 AK |
100 | #define KVM_VM_CR0_ALWAYS_ON \ |
101 | (KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST | X86_CR0_PG | X86_CR0_PE) | |
4c38609a AK |
102 | #define KVM_CR4_GUEST_OWNED_BITS \ |
103 | (X86_CR4_PVI | X86_CR4_DE | X86_CR4_PCE | X86_CR4_OSFXSR \ | |
104 | | X86_CR4_OSXMMEXCPT) | |
105 | ||
cdc0e244 AK |
106 | #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE) |
107 | #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE) | |
108 | ||
78ac8b47 AK |
109 | #define RMODE_GUEST_OWNED_EFLAGS_BITS (~(X86_EFLAGS_IOPL | X86_EFLAGS_VM)) |
110 | ||
4b8d54f9 ZE |
111 | /* |
112 | * These 2 parameters are used to config the controls for Pause-Loop Exiting: | |
113 | * ple_gap: upper bound on the amount of time between two successive | |
114 | * executions of PAUSE in a loop. Also indicate if ple enabled. | |
00c25bce | 115 | * According to test, this time is usually smaller than 128 cycles. |
4b8d54f9 ZE |
116 | * ple_window: upper bound on the amount of time a guest is allowed to execute |
117 | * in a PAUSE loop. Tests indicate that most spinlocks are held for | |
118 | * less than 2^12 cycles | |
119 | * Time is measured based on a counter that runs at the same rate as the TSC, | |
120 | * refer SDM volume 3b section 21.6.13 & 22.1.3. | |
121 | */ | |
00c25bce | 122 | #define KVM_VMX_DEFAULT_PLE_GAP 128 |
4b8d54f9 ZE |
123 | #define KVM_VMX_DEFAULT_PLE_WINDOW 4096 |
124 | static int ple_gap = KVM_VMX_DEFAULT_PLE_GAP; | |
125 | module_param(ple_gap, int, S_IRUGO); | |
126 | ||
127 | static int ple_window = KVM_VMX_DEFAULT_PLE_WINDOW; | |
128 | module_param(ple_window, int, S_IRUGO); | |
129 | ||
8bf00a52 | 130 | #define NR_AUTOLOAD_MSRS 8 |
ff2f6fe9 | 131 | #define VMCS02_POOL_SIZE 1 |
61d2ef2c | 132 | |
a2fa3e9f GH |
133 | struct vmcs { |
134 | u32 revision_id; | |
135 | u32 abort; | |
136 | char data[0]; | |
137 | }; | |
138 | ||
d462b819 NHE |
139 | /* |
140 | * Track a VMCS that may be loaded on a certain CPU. If it is (cpu!=-1), also | |
141 | * remember whether it was VMLAUNCHed, and maintain a linked list of all VMCSs | |
142 | * loaded on this CPU (so we can clear them if the CPU goes down). | |
143 | */ | |
144 | struct loaded_vmcs { | |
145 | struct vmcs *vmcs; | |
146 | int cpu; | |
147 | int launched; | |
148 | struct list_head loaded_vmcss_on_cpu_link; | |
149 | }; | |
150 | ||
26bb0981 AK |
151 | struct shared_msr_entry { |
152 | unsigned index; | |
153 | u64 data; | |
d5696725 | 154 | u64 mask; |
26bb0981 AK |
155 | }; |
156 | ||
a9d30f33 NHE |
157 | /* |
158 | * struct vmcs12 describes the state that our guest hypervisor (L1) keeps for a | |
159 | * single nested guest (L2), hence the name vmcs12. Any VMX implementation has | |
160 | * a VMCS structure, and vmcs12 is our emulated VMX's VMCS. This structure is | |
161 | * stored in guest memory specified by VMPTRLD, but is opaque to the guest, | |
162 | * which must access it using VMREAD/VMWRITE/VMCLEAR instructions. | |
163 | * More than one of these structures may exist, if L1 runs multiple L2 guests. | |
164 | * nested_vmx_run() will use the data here to build a vmcs02: a VMCS for the | |
165 | * underlying hardware which will be used to run L2. | |
166 | * This structure is packed to ensure that its layout is identical across | |
167 | * machines (necessary for live migration). | |
168 | * If there are changes in this struct, VMCS12_REVISION must be changed. | |
169 | */ | |
22bd0358 | 170 | typedef u64 natural_width; |
a9d30f33 NHE |
171 | struct __packed vmcs12 { |
172 | /* According to the Intel spec, a VMCS region must start with the | |
173 | * following two fields. Then follow implementation-specific data. | |
174 | */ | |
175 | u32 revision_id; | |
176 | u32 abort; | |
22bd0358 | 177 | |
27d6c865 NHE |
178 | u32 launch_state; /* set to 0 by VMCLEAR, to 1 by VMLAUNCH */ |
179 | u32 padding[7]; /* room for future expansion */ | |
180 | ||
22bd0358 NHE |
181 | u64 io_bitmap_a; |
182 | u64 io_bitmap_b; | |
183 | u64 msr_bitmap; | |
184 | u64 vm_exit_msr_store_addr; | |
185 | u64 vm_exit_msr_load_addr; | |
186 | u64 vm_entry_msr_load_addr; | |
187 | u64 tsc_offset; | |
188 | u64 virtual_apic_page_addr; | |
189 | u64 apic_access_addr; | |
190 | u64 ept_pointer; | |
191 | u64 guest_physical_address; | |
192 | u64 vmcs_link_pointer; | |
193 | u64 guest_ia32_debugctl; | |
194 | u64 guest_ia32_pat; | |
195 | u64 guest_ia32_efer; | |
196 | u64 guest_ia32_perf_global_ctrl; | |
197 | u64 guest_pdptr0; | |
198 | u64 guest_pdptr1; | |
199 | u64 guest_pdptr2; | |
200 | u64 guest_pdptr3; | |
201 | u64 host_ia32_pat; | |
202 | u64 host_ia32_efer; | |
203 | u64 host_ia32_perf_global_ctrl; | |
204 | u64 padding64[8]; /* room for future expansion */ | |
205 | /* | |
206 | * To allow migration of L1 (complete with its L2 guests) between | |
207 | * machines of different natural widths (32 or 64 bit), we cannot have | |
208 | * unsigned long fields with no explict size. We use u64 (aliased | |
209 | * natural_width) instead. Luckily, x86 is little-endian. | |
210 | */ | |
211 | natural_width cr0_guest_host_mask; | |
212 | natural_width cr4_guest_host_mask; | |
213 | natural_width cr0_read_shadow; | |
214 | natural_width cr4_read_shadow; | |
215 | natural_width cr3_target_value0; | |
216 | natural_width cr3_target_value1; | |
217 | natural_width cr3_target_value2; | |
218 | natural_width cr3_target_value3; | |
219 | natural_width exit_qualification; | |
220 | natural_width guest_linear_address; | |
221 | natural_width guest_cr0; | |
222 | natural_width guest_cr3; | |
223 | natural_width guest_cr4; | |
224 | natural_width guest_es_base; | |
225 | natural_width guest_cs_base; | |
226 | natural_width guest_ss_base; | |
227 | natural_width guest_ds_base; | |
228 | natural_width guest_fs_base; | |
229 | natural_width guest_gs_base; | |
230 | natural_width guest_ldtr_base; | |
231 | natural_width guest_tr_base; | |
232 | natural_width guest_gdtr_base; | |
233 | natural_width guest_idtr_base; | |
234 | natural_width guest_dr7; | |
235 | natural_width guest_rsp; | |
236 | natural_width guest_rip; | |
237 | natural_width guest_rflags; | |
238 | natural_width guest_pending_dbg_exceptions; | |
239 | natural_width guest_sysenter_esp; | |
240 | natural_width guest_sysenter_eip; | |
241 | natural_width host_cr0; | |
242 | natural_width host_cr3; | |
243 | natural_width host_cr4; | |
244 | natural_width host_fs_base; | |
245 | natural_width host_gs_base; | |
246 | natural_width host_tr_base; | |
247 | natural_width host_gdtr_base; | |
248 | natural_width host_idtr_base; | |
249 | natural_width host_ia32_sysenter_esp; | |
250 | natural_width host_ia32_sysenter_eip; | |
251 | natural_width host_rsp; | |
252 | natural_width host_rip; | |
253 | natural_width paddingl[8]; /* room for future expansion */ | |
254 | u32 pin_based_vm_exec_control; | |
255 | u32 cpu_based_vm_exec_control; | |
256 | u32 exception_bitmap; | |
257 | u32 page_fault_error_code_mask; | |
258 | u32 page_fault_error_code_match; | |
259 | u32 cr3_target_count; | |
260 | u32 vm_exit_controls; | |
261 | u32 vm_exit_msr_store_count; | |
262 | u32 vm_exit_msr_load_count; | |
263 | u32 vm_entry_controls; | |
264 | u32 vm_entry_msr_load_count; | |
265 | u32 vm_entry_intr_info_field; | |
266 | u32 vm_entry_exception_error_code; | |
267 | u32 vm_entry_instruction_len; | |
268 | u32 tpr_threshold; | |
269 | u32 secondary_vm_exec_control; | |
270 | u32 vm_instruction_error; | |
271 | u32 vm_exit_reason; | |
272 | u32 vm_exit_intr_info; | |
273 | u32 vm_exit_intr_error_code; | |
274 | u32 idt_vectoring_info_field; | |
275 | u32 idt_vectoring_error_code; | |
276 | u32 vm_exit_instruction_len; | |
277 | u32 vmx_instruction_info; | |
278 | u32 guest_es_limit; | |
279 | u32 guest_cs_limit; | |
280 | u32 guest_ss_limit; | |
281 | u32 guest_ds_limit; | |
282 | u32 guest_fs_limit; | |
283 | u32 guest_gs_limit; | |
284 | u32 guest_ldtr_limit; | |
285 | u32 guest_tr_limit; | |
286 | u32 guest_gdtr_limit; | |
287 | u32 guest_idtr_limit; | |
288 | u32 guest_es_ar_bytes; | |
289 | u32 guest_cs_ar_bytes; | |
290 | u32 guest_ss_ar_bytes; | |
291 | u32 guest_ds_ar_bytes; | |
292 | u32 guest_fs_ar_bytes; | |
293 | u32 guest_gs_ar_bytes; | |
294 | u32 guest_ldtr_ar_bytes; | |
295 | u32 guest_tr_ar_bytes; | |
296 | u32 guest_interruptibility_info; | |
297 | u32 guest_activity_state; | |
298 | u32 guest_sysenter_cs; | |
299 | u32 host_ia32_sysenter_cs; | |
300 | u32 padding32[8]; /* room for future expansion */ | |
301 | u16 virtual_processor_id; | |
302 | u16 guest_es_selector; | |
303 | u16 guest_cs_selector; | |
304 | u16 guest_ss_selector; | |
305 | u16 guest_ds_selector; | |
306 | u16 guest_fs_selector; | |
307 | u16 guest_gs_selector; | |
308 | u16 guest_ldtr_selector; | |
309 | u16 guest_tr_selector; | |
310 | u16 host_es_selector; | |
311 | u16 host_cs_selector; | |
312 | u16 host_ss_selector; | |
313 | u16 host_ds_selector; | |
314 | u16 host_fs_selector; | |
315 | u16 host_gs_selector; | |
316 | u16 host_tr_selector; | |
a9d30f33 NHE |
317 | }; |
318 | ||
319 | /* | |
320 | * VMCS12_REVISION is an arbitrary id that should be changed if the content or | |
321 | * layout of struct vmcs12 is changed. MSR_IA32_VMX_BASIC returns this id, and | |
322 | * VMPTRLD verifies that the VMCS region that L1 is loading contains this id. | |
323 | */ | |
324 | #define VMCS12_REVISION 0x11e57ed0 | |
325 | ||
326 | /* | |
327 | * VMCS12_SIZE is the number of bytes L1 should allocate for the VMXON region | |
328 | * and any VMCS region. Although only sizeof(struct vmcs12) are used by the | |
329 | * current implementation, 4K are reserved to avoid future complications. | |
330 | */ | |
331 | #define VMCS12_SIZE 0x1000 | |
332 | ||
ff2f6fe9 NHE |
333 | /* Used to remember the last vmcs02 used for some recently used vmcs12s */ |
334 | struct vmcs02_list { | |
335 | struct list_head list; | |
336 | gpa_t vmptr; | |
337 | struct loaded_vmcs vmcs02; | |
338 | }; | |
339 | ||
ec378aee NHE |
340 | /* |
341 | * The nested_vmx structure is part of vcpu_vmx, and holds information we need | |
342 | * for correct emulation of VMX (i.e., nested VMX) on this vcpu. | |
343 | */ | |
344 | struct nested_vmx { | |
345 | /* Has the level1 guest done vmxon? */ | |
346 | bool vmxon; | |
a9d30f33 NHE |
347 | |
348 | /* The guest-physical address of the current VMCS L1 keeps for L2 */ | |
349 | gpa_t current_vmptr; | |
350 | /* The host-usable pointer to the above */ | |
351 | struct page *current_vmcs12_page; | |
352 | struct vmcs12 *current_vmcs12; | |
ff2f6fe9 NHE |
353 | |
354 | /* vmcs02_list cache of VMCSs recently used to run L2 guests */ | |
355 | struct list_head vmcs02_pool; | |
356 | int vmcs02_num; | |
fe3ef05c | 357 | u64 vmcs01_tsc_offset; |
644d711a NHE |
358 | /* L2 must run next, and mustn't decide to exit to L1. */ |
359 | bool nested_run_pending; | |
fe3ef05c NHE |
360 | /* |
361 | * Guest pages referred to in vmcs02 with host-physical pointers, so | |
362 | * we must keep them pinned while L2 runs. | |
363 | */ | |
364 | struct page *apic_access_page; | |
ec378aee NHE |
365 | }; |
366 | ||
a2fa3e9f | 367 | struct vcpu_vmx { |
fb3f0f51 | 368 | struct kvm_vcpu vcpu; |
313dbd49 | 369 | unsigned long host_rsp; |
29bd8a78 | 370 | u8 fail; |
69c73028 | 371 | u8 cpl; |
9d58b931 | 372 | bool nmi_known_unmasked; |
51aa01d1 | 373 | u32 exit_intr_info; |
1155f76a | 374 | u32 idt_vectoring_info; |
6de12732 | 375 | ulong rflags; |
26bb0981 | 376 | struct shared_msr_entry *guest_msrs; |
a2fa3e9f GH |
377 | int nmsrs; |
378 | int save_nmsrs; | |
a2fa3e9f | 379 | #ifdef CONFIG_X86_64 |
44ea2b17 AK |
380 | u64 msr_host_kernel_gs_base; |
381 | u64 msr_guest_kernel_gs_base; | |
a2fa3e9f | 382 | #endif |
d462b819 NHE |
383 | /* |
384 | * loaded_vmcs points to the VMCS currently used in this vcpu. For a | |
385 | * non-nested (L1) guest, it always points to vmcs01. For a nested | |
386 | * guest (L2), it points to a different VMCS. | |
387 | */ | |
388 | struct loaded_vmcs vmcs01; | |
389 | struct loaded_vmcs *loaded_vmcs; | |
390 | bool __launched; /* temporary, used in vmx_vcpu_run */ | |
61d2ef2c AK |
391 | struct msr_autoload { |
392 | unsigned nr; | |
393 | struct vmx_msr_entry guest[NR_AUTOLOAD_MSRS]; | |
394 | struct vmx_msr_entry host[NR_AUTOLOAD_MSRS]; | |
395 | } msr_autoload; | |
a2fa3e9f GH |
396 | struct { |
397 | int loaded; | |
398 | u16 fs_sel, gs_sel, ldt_sel; | |
b2da15ac AK |
399 | #ifdef CONFIG_X86_64 |
400 | u16 ds_sel, es_sel; | |
401 | #endif | |
152d3f2f LV |
402 | int gs_ldt_reload_needed; |
403 | int fs_reload_needed; | |
d77c26fc | 404 | } host_state; |
9c8cba37 | 405 | struct { |
7ffd92c5 | 406 | int vm86_active; |
78ac8b47 | 407 | ulong save_rflags; |
f5f7b2fe AK |
408 | struct kvm_segment segs[8]; |
409 | } rmode; | |
410 | struct { | |
411 | u32 bitmask; /* 4 bits per segment (1 bit per field) */ | |
7ffd92c5 AK |
412 | struct kvm_save_segment { |
413 | u16 selector; | |
414 | unsigned long base; | |
415 | u32 limit; | |
416 | u32 ar; | |
f5f7b2fe | 417 | } seg[8]; |
2fb92db1 | 418 | } segment_cache; |
2384d2b3 | 419 | int vpid; |
04fa4d32 | 420 | bool emulation_required; |
3b86cd99 JK |
421 | |
422 | /* Support for vnmi-less CPUs */ | |
423 | int soft_vnmi_blocked; | |
424 | ktime_t entry_time; | |
425 | s64 vnmi_blocked_time; | |
a0861c02 | 426 | u32 exit_reason; |
4e47c7a6 SY |
427 | |
428 | bool rdtscp_enabled; | |
ec378aee NHE |
429 | |
430 | /* Support for a guest hypervisor (nested VMX) */ | |
431 | struct nested_vmx nested; | |
a2fa3e9f GH |
432 | }; |
433 | ||
2fb92db1 AK |
434 | enum segment_cache_field { |
435 | SEG_FIELD_SEL = 0, | |
436 | SEG_FIELD_BASE = 1, | |
437 | SEG_FIELD_LIMIT = 2, | |
438 | SEG_FIELD_AR = 3, | |
439 | ||
440 | SEG_FIELD_NR = 4 | |
441 | }; | |
442 | ||
a2fa3e9f GH |
443 | static inline struct vcpu_vmx *to_vmx(struct kvm_vcpu *vcpu) |
444 | { | |
fb3f0f51 | 445 | return container_of(vcpu, struct vcpu_vmx, vcpu); |
a2fa3e9f GH |
446 | } |
447 | ||
22bd0358 NHE |
448 | #define VMCS12_OFFSET(x) offsetof(struct vmcs12, x) |
449 | #define FIELD(number, name) [number] = VMCS12_OFFSET(name) | |
450 | #define FIELD64(number, name) [number] = VMCS12_OFFSET(name), \ | |
451 | [number##_HIGH] = VMCS12_OFFSET(name)+4 | |
452 | ||
453 | static unsigned short vmcs_field_to_offset_table[] = { | |
454 | FIELD(VIRTUAL_PROCESSOR_ID, virtual_processor_id), | |
455 | FIELD(GUEST_ES_SELECTOR, guest_es_selector), | |
456 | FIELD(GUEST_CS_SELECTOR, guest_cs_selector), | |
457 | FIELD(GUEST_SS_SELECTOR, guest_ss_selector), | |
458 | FIELD(GUEST_DS_SELECTOR, guest_ds_selector), | |
459 | FIELD(GUEST_FS_SELECTOR, guest_fs_selector), | |
460 | FIELD(GUEST_GS_SELECTOR, guest_gs_selector), | |
461 | FIELD(GUEST_LDTR_SELECTOR, guest_ldtr_selector), | |
462 | FIELD(GUEST_TR_SELECTOR, guest_tr_selector), | |
463 | FIELD(HOST_ES_SELECTOR, host_es_selector), | |
464 | FIELD(HOST_CS_SELECTOR, host_cs_selector), | |
465 | FIELD(HOST_SS_SELECTOR, host_ss_selector), | |
466 | FIELD(HOST_DS_SELECTOR, host_ds_selector), | |
467 | FIELD(HOST_FS_SELECTOR, host_fs_selector), | |
468 | FIELD(HOST_GS_SELECTOR, host_gs_selector), | |
469 | FIELD(HOST_TR_SELECTOR, host_tr_selector), | |
470 | FIELD64(IO_BITMAP_A, io_bitmap_a), | |
471 | FIELD64(IO_BITMAP_B, io_bitmap_b), | |
472 | FIELD64(MSR_BITMAP, msr_bitmap), | |
473 | FIELD64(VM_EXIT_MSR_STORE_ADDR, vm_exit_msr_store_addr), | |
474 | FIELD64(VM_EXIT_MSR_LOAD_ADDR, vm_exit_msr_load_addr), | |
475 | FIELD64(VM_ENTRY_MSR_LOAD_ADDR, vm_entry_msr_load_addr), | |
476 | FIELD64(TSC_OFFSET, tsc_offset), | |
477 | FIELD64(VIRTUAL_APIC_PAGE_ADDR, virtual_apic_page_addr), | |
478 | FIELD64(APIC_ACCESS_ADDR, apic_access_addr), | |
479 | FIELD64(EPT_POINTER, ept_pointer), | |
480 | FIELD64(GUEST_PHYSICAL_ADDRESS, guest_physical_address), | |
481 | FIELD64(VMCS_LINK_POINTER, vmcs_link_pointer), | |
482 | FIELD64(GUEST_IA32_DEBUGCTL, guest_ia32_debugctl), | |
483 | FIELD64(GUEST_IA32_PAT, guest_ia32_pat), | |
484 | FIELD64(GUEST_IA32_EFER, guest_ia32_efer), | |
485 | FIELD64(GUEST_IA32_PERF_GLOBAL_CTRL, guest_ia32_perf_global_ctrl), | |
486 | FIELD64(GUEST_PDPTR0, guest_pdptr0), | |
487 | FIELD64(GUEST_PDPTR1, guest_pdptr1), | |
488 | FIELD64(GUEST_PDPTR2, guest_pdptr2), | |
489 | FIELD64(GUEST_PDPTR3, guest_pdptr3), | |
490 | FIELD64(HOST_IA32_PAT, host_ia32_pat), | |
491 | FIELD64(HOST_IA32_EFER, host_ia32_efer), | |
492 | FIELD64(HOST_IA32_PERF_GLOBAL_CTRL, host_ia32_perf_global_ctrl), | |
493 | FIELD(PIN_BASED_VM_EXEC_CONTROL, pin_based_vm_exec_control), | |
494 | FIELD(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control), | |
495 | FIELD(EXCEPTION_BITMAP, exception_bitmap), | |
496 | FIELD(PAGE_FAULT_ERROR_CODE_MASK, page_fault_error_code_mask), | |
497 | FIELD(PAGE_FAULT_ERROR_CODE_MATCH, page_fault_error_code_match), | |
498 | FIELD(CR3_TARGET_COUNT, cr3_target_count), | |
499 | FIELD(VM_EXIT_CONTROLS, vm_exit_controls), | |
500 | FIELD(VM_EXIT_MSR_STORE_COUNT, vm_exit_msr_store_count), | |
501 | FIELD(VM_EXIT_MSR_LOAD_COUNT, vm_exit_msr_load_count), | |
502 | FIELD(VM_ENTRY_CONTROLS, vm_entry_controls), | |
503 | FIELD(VM_ENTRY_MSR_LOAD_COUNT, vm_entry_msr_load_count), | |
504 | FIELD(VM_ENTRY_INTR_INFO_FIELD, vm_entry_intr_info_field), | |
505 | FIELD(VM_ENTRY_EXCEPTION_ERROR_CODE, vm_entry_exception_error_code), | |
506 | FIELD(VM_ENTRY_INSTRUCTION_LEN, vm_entry_instruction_len), | |
507 | FIELD(TPR_THRESHOLD, tpr_threshold), | |
508 | FIELD(SECONDARY_VM_EXEC_CONTROL, secondary_vm_exec_control), | |
509 | FIELD(VM_INSTRUCTION_ERROR, vm_instruction_error), | |
510 | FIELD(VM_EXIT_REASON, vm_exit_reason), | |
511 | FIELD(VM_EXIT_INTR_INFO, vm_exit_intr_info), | |
512 | FIELD(VM_EXIT_INTR_ERROR_CODE, vm_exit_intr_error_code), | |
513 | FIELD(IDT_VECTORING_INFO_FIELD, idt_vectoring_info_field), | |
514 | FIELD(IDT_VECTORING_ERROR_CODE, idt_vectoring_error_code), | |
515 | FIELD(VM_EXIT_INSTRUCTION_LEN, vm_exit_instruction_len), | |
516 | FIELD(VMX_INSTRUCTION_INFO, vmx_instruction_info), | |
517 | FIELD(GUEST_ES_LIMIT, guest_es_limit), | |
518 | FIELD(GUEST_CS_LIMIT, guest_cs_limit), | |
519 | FIELD(GUEST_SS_LIMIT, guest_ss_limit), | |
520 | FIELD(GUEST_DS_LIMIT, guest_ds_limit), | |
521 | FIELD(GUEST_FS_LIMIT, guest_fs_limit), | |
522 | FIELD(GUEST_GS_LIMIT, guest_gs_limit), | |
523 | FIELD(GUEST_LDTR_LIMIT, guest_ldtr_limit), | |
524 | FIELD(GUEST_TR_LIMIT, guest_tr_limit), | |
525 | FIELD(GUEST_GDTR_LIMIT, guest_gdtr_limit), | |
526 | FIELD(GUEST_IDTR_LIMIT, guest_idtr_limit), | |
527 | FIELD(GUEST_ES_AR_BYTES, guest_es_ar_bytes), | |
528 | FIELD(GUEST_CS_AR_BYTES, guest_cs_ar_bytes), | |
529 | FIELD(GUEST_SS_AR_BYTES, guest_ss_ar_bytes), | |
530 | FIELD(GUEST_DS_AR_BYTES, guest_ds_ar_bytes), | |
531 | FIELD(GUEST_FS_AR_BYTES, guest_fs_ar_bytes), | |
532 | FIELD(GUEST_GS_AR_BYTES, guest_gs_ar_bytes), | |
533 | FIELD(GUEST_LDTR_AR_BYTES, guest_ldtr_ar_bytes), | |
534 | FIELD(GUEST_TR_AR_BYTES, guest_tr_ar_bytes), | |
535 | FIELD(GUEST_INTERRUPTIBILITY_INFO, guest_interruptibility_info), | |
536 | FIELD(GUEST_ACTIVITY_STATE, guest_activity_state), | |
537 | FIELD(GUEST_SYSENTER_CS, guest_sysenter_cs), | |
538 | FIELD(HOST_IA32_SYSENTER_CS, host_ia32_sysenter_cs), | |
539 | FIELD(CR0_GUEST_HOST_MASK, cr0_guest_host_mask), | |
540 | FIELD(CR4_GUEST_HOST_MASK, cr4_guest_host_mask), | |
541 | FIELD(CR0_READ_SHADOW, cr0_read_shadow), | |
542 | FIELD(CR4_READ_SHADOW, cr4_read_shadow), | |
543 | FIELD(CR3_TARGET_VALUE0, cr3_target_value0), | |
544 | FIELD(CR3_TARGET_VALUE1, cr3_target_value1), | |
545 | FIELD(CR3_TARGET_VALUE2, cr3_target_value2), | |
546 | FIELD(CR3_TARGET_VALUE3, cr3_target_value3), | |
547 | FIELD(EXIT_QUALIFICATION, exit_qualification), | |
548 | FIELD(GUEST_LINEAR_ADDRESS, guest_linear_address), | |
549 | FIELD(GUEST_CR0, guest_cr0), | |
550 | FIELD(GUEST_CR3, guest_cr3), | |
551 | FIELD(GUEST_CR4, guest_cr4), | |
552 | FIELD(GUEST_ES_BASE, guest_es_base), | |
553 | FIELD(GUEST_CS_BASE, guest_cs_base), | |
554 | FIELD(GUEST_SS_BASE, guest_ss_base), | |
555 | FIELD(GUEST_DS_BASE, guest_ds_base), | |
556 | FIELD(GUEST_FS_BASE, guest_fs_base), | |
557 | FIELD(GUEST_GS_BASE, guest_gs_base), | |
558 | FIELD(GUEST_LDTR_BASE, guest_ldtr_base), | |
559 | FIELD(GUEST_TR_BASE, guest_tr_base), | |
560 | FIELD(GUEST_GDTR_BASE, guest_gdtr_base), | |
561 | FIELD(GUEST_IDTR_BASE, guest_idtr_base), | |
562 | FIELD(GUEST_DR7, guest_dr7), | |
563 | FIELD(GUEST_RSP, guest_rsp), | |
564 | FIELD(GUEST_RIP, guest_rip), | |
565 | FIELD(GUEST_RFLAGS, guest_rflags), | |
566 | FIELD(GUEST_PENDING_DBG_EXCEPTIONS, guest_pending_dbg_exceptions), | |
567 | FIELD(GUEST_SYSENTER_ESP, guest_sysenter_esp), | |
568 | FIELD(GUEST_SYSENTER_EIP, guest_sysenter_eip), | |
569 | FIELD(HOST_CR0, host_cr0), | |
570 | FIELD(HOST_CR3, host_cr3), | |
571 | FIELD(HOST_CR4, host_cr4), | |
572 | FIELD(HOST_FS_BASE, host_fs_base), | |
573 | FIELD(HOST_GS_BASE, host_gs_base), | |
574 | FIELD(HOST_TR_BASE, host_tr_base), | |
575 | FIELD(HOST_GDTR_BASE, host_gdtr_base), | |
576 | FIELD(HOST_IDTR_BASE, host_idtr_base), | |
577 | FIELD(HOST_IA32_SYSENTER_ESP, host_ia32_sysenter_esp), | |
578 | FIELD(HOST_IA32_SYSENTER_EIP, host_ia32_sysenter_eip), | |
579 | FIELD(HOST_RSP, host_rsp), | |
580 | FIELD(HOST_RIP, host_rip), | |
581 | }; | |
582 | static const int max_vmcs_field = ARRAY_SIZE(vmcs_field_to_offset_table); | |
583 | ||
584 | static inline short vmcs_field_to_offset(unsigned long field) | |
585 | { | |
586 | if (field >= max_vmcs_field || vmcs_field_to_offset_table[field] == 0) | |
587 | return -1; | |
588 | return vmcs_field_to_offset_table[field]; | |
589 | } | |
590 | ||
a9d30f33 NHE |
591 | static inline struct vmcs12 *get_vmcs12(struct kvm_vcpu *vcpu) |
592 | { | |
593 | return to_vmx(vcpu)->nested.current_vmcs12; | |
594 | } | |
595 | ||
596 | static struct page *nested_get_page(struct kvm_vcpu *vcpu, gpa_t addr) | |
597 | { | |
598 | struct page *page = gfn_to_page(vcpu->kvm, addr >> PAGE_SHIFT); | |
32cad84f | 599 | if (is_error_page(page)) |
a9d30f33 | 600 | return NULL; |
32cad84f | 601 | |
a9d30f33 NHE |
602 | return page; |
603 | } | |
604 | ||
605 | static void nested_release_page(struct page *page) | |
606 | { | |
607 | kvm_release_page_dirty(page); | |
608 | } | |
609 | ||
610 | static void nested_release_page_clean(struct page *page) | |
611 | { | |
612 | kvm_release_page_clean(page); | |
613 | } | |
614 | ||
4e1096d2 | 615 | static u64 construct_eptp(unsigned long root_hpa); |
4610c9cc DX |
616 | static void kvm_cpu_vmxon(u64 addr); |
617 | static void kvm_cpu_vmxoff(void); | |
aff48baa | 618 | static void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3); |
776e58ea | 619 | static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr); |
b246dd5d OW |
620 | static void vmx_set_segment(struct kvm_vcpu *vcpu, |
621 | struct kvm_segment *var, int seg); | |
622 | static void vmx_get_segment(struct kvm_vcpu *vcpu, | |
623 | struct kvm_segment *var, int seg); | |
75880a01 | 624 | |
6aa8b732 AK |
625 | static DEFINE_PER_CPU(struct vmcs *, vmxarea); |
626 | static DEFINE_PER_CPU(struct vmcs *, current_vmcs); | |
d462b819 NHE |
627 | /* |
628 | * We maintain a per-CPU linked-list of VMCS loaded on that CPU. This is needed | |
629 | * when a CPU is brought down, and we need to VMCLEAR all VMCSs loaded on it. | |
630 | */ | |
631 | static DEFINE_PER_CPU(struct list_head, loaded_vmcss_on_cpu); | |
3444d7da | 632 | static DEFINE_PER_CPU(struct desc_ptr, host_gdt); |
6aa8b732 | 633 | |
3e7c73e9 AK |
634 | static unsigned long *vmx_io_bitmap_a; |
635 | static unsigned long *vmx_io_bitmap_b; | |
5897297b AK |
636 | static unsigned long *vmx_msr_bitmap_legacy; |
637 | static unsigned long *vmx_msr_bitmap_longmode; | |
fdef3ad1 | 638 | |
110312c8 | 639 | static bool cpu_has_load_ia32_efer; |
8bf00a52 | 640 | static bool cpu_has_load_perf_global_ctrl; |
110312c8 | 641 | |
2384d2b3 SY |
642 | static DECLARE_BITMAP(vmx_vpid_bitmap, VMX_NR_VPIDS); |
643 | static DEFINE_SPINLOCK(vmx_vpid_lock); | |
644 | ||
1c3d14fe | 645 | static struct vmcs_config { |
6aa8b732 AK |
646 | int size; |
647 | int order; | |
648 | u32 revision_id; | |
1c3d14fe YS |
649 | u32 pin_based_exec_ctrl; |
650 | u32 cpu_based_exec_ctrl; | |
f78e0e2e | 651 | u32 cpu_based_2nd_exec_ctrl; |
1c3d14fe YS |
652 | u32 vmexit_ctrl; |
653 | u32 vmentry_ctrl; | |
654 | } vmcs_config; | |
6aa8b732 | 655 | |
efff9e53 | 656 | static struct vmx_capability { |
d56f546d SY |
657 | u32 ept; |
658 | u32 vpid; | |
659 | } vmx_capability; | |
660 | ||
6aa8b732 AK |
661 | #define VMX_SEGMENT_FIELD(seg) \ |
662 | [VCPU_SREG_##seg] = { \ | |
663 | .selector = GUEST_##seg##_SELECTOR, \ | |
664 | .base = GUEST_##seg##_BASE, \ | |
665 | .limit = GUEST_##seg##_LIMIT, \ | |
666 | .ar_bytes = GUEST_##seg##_AR_BYTES, \ | |
667 | } | |
668 | ||
669 | static struct kvm_vmx_segment_field { | |
670 | unsigned selector; | |
671 | unsigned base; | |
672 | unsigned limit; | |
673 | unsigned ar_bytes; | |
674 | } kvm_vmx_segment_fields[] = { | |
675 | VMX_SEGMENT_FIELD(CS), | |
676 | VMX_SEGMENT_FIELD(DS), | |
677 | VMX_SEGMENT_FIELD(ES), | |
678 | VMX_SEGMENT_FIELD(FS), | |
679 | VMX_SEGMENT_FIELD(GS), | |
680 | VMX_SEGMENT_FIELD(SS), | |
681 | VMX_SEGMENT_FIELD(TR), | |
682 | VMX_SEGMENT_FIELD(LDTR), | |
683 | }; | |
684 | ||
26bb0981 AK |
685 | static u64 host_efer; |
686 | ||
6de4f3ad AK |
687 | static void ept_save_pdptrs(struct kvm_vcpu *vcpu); |
688 | ||
4d56c8a7 | 689 | /* |
8c06585d | 690 | * Keep MSR_STAR at the end, as setup_msrs() will try to optimize it |
4d56c8a7 AK |
691 | * away by decrementing the array size. |
692 | */ | |
6aa8b732 | 693 | static const u32 vmx_msr_index[] = { |
05b3e0c2 | 694 | #ifdef CONFIG_X86_64 |
44ea2b17 | 695 | MSR_SYSCALL_MASK, MSR_LSTAR, MSR_CSTAR, |
6aa8b732 | 696 | #endif |
8c06585d | 697 | MSR_EFER, MSR_TSC_AUX, MSR_STAR, |
6aa8b732 | 698 | }; |
9d8f549d | 699 | #define NR_VMX_MSR ARRAY_SIZE(vmx_msr_index) |
6aa8b732 | 700 | |
31299944 | 701 | static inline bool is_page_fault(u32 intr_info) |
6aa8b732 AK |
702 | { |
703 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
704 | INTR_INFO_VALID_MASK)) == | |
8ab2d2e2 | 705 | (INTR_TYPE_HARD_EXCEPTION | PF_VECTOR | INTR_INFO_VALID_MASK); |
6aa8b732 AK |
706 | } |
707 | ||
31299944 | 708 | static inline bool is_no_device(u32 intr_info) |
2ab455cc AL |
709 | { |
710 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
711 | INTR_INFO_VALID_MASK)) == | |
8ab2d2e2 | 712 | (INTR_TYPE_HARD_EXCEPTION | NM_VECTOR | INTR_INFO_VALID_MASK); |
2ab455cc AL |
713 | } |
714 | ||
31299944 | 715 | static inline bool is_invalid_opcode(u32 intr_info) |
7aa81cc0 AL |
716 | { |
717 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
718 | INTR_INFO_VALID_MASK)) == | |
8ab2d2e2 | 719 | (INTR_TYPE_HARD_EXCEPTION | UD_VECTOR | INTR_INFO_VALID_MASK); |
7aa81cc0 AL |
720 | } |
721 | ||
31299944 | 722 | static inline bool is_external_interrupt(u32 intr_info) |
6aa8b732 AK |
723 | { |
724 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK)) | |
725 | == (INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK); | |
726 | } | |
727 | ||
31299944 | 728 | static inline bool is_machine_check(u32 intr_info) |
a0861c02 AK |
729 | { |
730 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK | | |
731 | INTR_INFO_VALID_MASK)) == | |
732 | (INTR_TYPE_HARD_EXCEPTION | MC_VECTOR | INTR_INFO_VALID_MASK); | |
733 | } | |
734 | ||
31299944 | 735 | static inline bool cpu_has_vmx_msr_bitmap(void) |
25c5f225 | 736 | { |
04547156 | 737 | return vmcs_config.cpu_based_exec_ctrl & CPU_BASED_USE_MSR_BITMAPS; |
25c5f225 SY |
738 | } |
739 | ||
31299944 | 740 | static inline bool cpu_has_vmx_tpr_shadow(void) |
6e5d865c | 741 | { |
04547156 | 742 | return vmcs_config.cpu_based_exec_ctrl & CPU_BASED_TPR_SHADOW; |
6e5d865c YS |
743 | } |
744 | ||
31299944 | 745 | static inline bool vm_need_tpr_shadow(struct kvm *kvm) |
6e5d865c | 746 | { |
04547156 | 747 | return (cpu_has_vmx_tpr_shadow()) && (irqchip_in_kernel(kvm)); |
6e5d865c YS |
748 | } |
749 | ||
31299944 | 750 | static inline bool cpu_has_secondary_exec_ctrls(void) |
f78e0e2e | 751 | { |
04547156 SY |
752 | return vmcs_config.cpu_based_exec_ctrl & |
753 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; | |
f78e0e2e SY |
754 | } |
755 | ||
774ead3a | 756 | static inline bool cpu_has_vmx_virtualize_apic_accesses(void) |
f78e0e2e | 757 | { |
04547156 SY |
758 | return vmcs_config.cpu_based_2nd_exec_ctrl & |
759 | SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
760 | } | |
761 | ||
762 | static inline bool cpu_has_vmx_flexpriority(void) | |
763 | { | |
764 | return cpu_has_vmx_tpr_shadow() && | |
765 | cpu_has_vmx_virtualize_apic_accesses(); | |
f78e0e2e SY |
766 | } |
767 | ||
e799794e MT |
768 | static inline bool cpu_has_vmx_ept_execute_only(void) |
769 | { | |
31299944 | 770 | return vmx_capability.ept & VMX_EPT_EXECUTE_ONLY_BIT; |
e799794e MT |
771 | } |
772 | ||
773 | static inline bool cpu_has_vmx_eptp_uncacheable(void) | |
774 | { | |
31299944 | 775 | return vmx_capability.ept & VMX_EPTP_UC_BIT; |
e799794e MT |
776 | } |
777 | ||
778 | static inline bool cpu_has_vmx_eptp_writeback(void) | |
779 | { | |
31299944 | 780 | return vmx_capability.ept & VMX_EPTP_WB_BIT; |
e799794e MT |
781 | } |
782 | ||
783 | static inline bool cpu_has_vmx_ept_2m_page(void) | |
784 | { | |
31299944 | 785 | return vmx_capability.ept & VMX_EPT_2MB_PAGE_BIT; |
e799794e MT |
786 | } |
787 | ||
878403b7 SY |
788 | static inline bool cpu_has_vmx_ept_1g_page(void) |
789 | { | |
31299944 | 790 | return vmx_capability.ept & VMX_EPT_1GB_PAGE_BIT; |
878403b7 SY |
791 | } |
792 | ||
4bc9b982 SY |
793 | static inline bool cpu_has_vmx_ept_4levels(void) |
794 | { | |
795 | return vmx_capability.ept & VMX_EPT_PAGE_WALK_4_BIT; | |
796 | } | |
797 | ||
83c3a331 XH |
798 | static inline bool cpu_has_vmx_ept_ad_bits(void) |
799 | { | |
800 | return vmx_capability.ept & VMX_EPT_AD_BIT; | |
801 | } | |
802 | ||
31299944 | 803 | static inline bool cpu_has_vmx_invept_individual_addr(void) |
d56f546d | 804 | { |
31299944 | 805 | return vmx_capability.ept & VMX_EPT_EXTENT_INDIVIDUAL_BIT; |
d56f546d SY |
806 | } |
807 | ||
31299944 | 808 | static inline bool cpu_has_vmx_invept_context(void) |
d56f546d | 809 | { |
31299944 | 810 | return vmx_capability.ept & VMX_EPT_EXTENT_CONTEXT_BIT; |
d56f546d SY |
811 | } |
812 | ||
31299944 | 813 | static inline bool cpu_has_vmx_invept_global(void) |
d56f546d | 814 | { |
31299944 | 815 | return vmx_capability.ept & VMX_EPT_EXTENT_GLOBAL_BIT; |
d56f546d SY |
816 | } |
817 | ||
518c8aee GJ |
818 | static inline bool cpu_has_vmx_invvpid_single(void) |
819 | { | |
820 | return vmx_capability.vpid & VMX_VPID_EXTENT_SINGLE_CONTEXT_BIT; | |
821 | } | |
822 | ||
b9d762fa GJ |
823 | static inline bool cpu_has_vmx_invvpid_global(void) |
824 | { | |
825 | return vmx_capability.vpid & VMX_VPID_EXTENT_GLOBAL_CONTEXT_BIT; | |
826 | } | |
827 | ||
31299944 | 828 | static inline bool cpu_has_vmx_ept(void) |
d56f546d | 829 | { |
04547156 SY |
830 | return vmcs_config.cpu_based_2nd_exec_ctrl & |
831 | SECONDARY_EXEC_ENABLE_EPT; | |
d56f546d SY |
832 | } |
833 | ||
31299944 | 834 | static inline bool cpu_has_vmx_unrestricted_guest(void) |
3a624e29 NK |
835 | { |
836 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
837 | SECONDARY_EXEC_UNRESTRICTED_GUEST; | |
838 | } | |
839 | ||
31299944 | 840 | static inline bool cpu_has_vmx_ple(void) |
4b8d54f9 ZE |
841 | { |
842 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
843 | SECONDARY_EXEC_PAUSE_LOOP_EXITING; | |
844 | } | |
845 | ||
31299944 | 846 | static inline bool vm_need_virtualize_apic_accesses(struct kvm *kvm) |
f78e0e2e | 847 | { |
6d3e435e | 848 | return flexpriority_enabled && irqchip_in_kernel(kvm); |
f78e0e2e SY |
849 | } |
850 | ||
31299944 | 851 | static inline bool cpu_has_vmx_vpid(void) |
2384d2b3 | 852 | { |
04547156 SY |
853 | return vmcs_config.cpu_based_2nd_exec_ctrl & |
854 | SECONDARY_EXEC_ENABLE_VPID; | |
2384d2b3 SY |
855 | } |
856 | ||
31299944 | 857 | static inline bool cpu_has_vmx_rdtscp(void) |
4e47c7a6 SY |
858 | { |
859 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
860 | SECONDARY_EXEC_RDTSCP; | |
861 | } | |
862 | ||
ad756a16 MJ |
863 | static inline bool cpu_has_vmx_invpcid(void) |
864 | { | |
865 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
866 | SECONDARY_EXEC_ENABLE_INVPCID; | |
867 | } | |
868 | ||
31299944 | 869 | static inline bool cpu_has_virtual_nmis(void) |
f08864b4 SY |
870 | { |
871 | return vmcs_config.pin_based_exec_ctrl & PIN_BASED_VIRTUAL_NMIS; | |
872 | } | |
873 | ||
f5f48ee1 SY |
874 | static inline bool cpu_has_vmx_wbinvd_exit(void) |
875 | { | |
876 | return vmcs_config.cpu_based_2nd_exec_ctrl & | |
877 | SECONDARY_EXEC_WBINVD_EXITING; | |
878 | } | |
879 | ||
04547156 SY |
880 | static inline bool report_flexpriority(void) |
881 | { | |
882 | return flexpriority_enabled; | |
883 | } | |
884 | ||
fe3ef05c NHE |
885 | static inline bool nested_cpu_has(struct vmcs12 *vmcs12, u32 bit) |
886 | { | |
887 | return vmcs12->cpu_based_vm_exec_control & bit; | |
888 | } | |
889 | ||
890 | static inline bool nested_cpu_has2(struct vmcs12 *vmcs12, u32 bit) | |
891 | { | |
892 | return (vmcs12->cpu_based_vm_exec_control & | |
893 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) && | |
894 | (vmcs12->secondary_vm_exec_control & bit); | |
895 | } | |
896 | ||
644d711a NHE |
897 | static inline bool nested_cpu_has_virtual_nmis(struct vmcs12 *vmcs12, |
898 | struct kvm_vcpu *vcpu) | |
899 | { | |
900 | return vmcs12->pin_based_vm_exec_control & PIN_BASED_VIRTUAL_NMIS; | |
901 | } | |
902 | ||
903 | static inline bool is_exception(u32 intr_info) | |
904 | { | |
905 | return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK)) | |
906 | == (INTR_TYPE_HARD_EXCEPTION | INTR_INFO_VALID_MASK); | |
907 | } | |
908 | ||
909 | static void nested_vmx_vmexit(struct kvm_vcpu *vcpu); | |
7c177938 NHE |
910 | static void nested_vmx_entry_failure(struct kvm_vcpu *vcpu, |
911 | struct vmcs12 *vmcs12, | |
912 | u32 reason, unsigned long qualification); | |
913 | ||
8b9cf98c | 914 | static int __find_msr_index(struct vcpu_vmx *vmx, u32 msr) |
7725f0ba AK |
915 | { |
916 | int i; | |
917 | ||
a2fa3e9f | 918 | for (i = 0; i < vmx->nmsrs; ++i) |
26bb0981 | 919 | if (vmx_msr_index[vmx->guest_msrs[i].index] == msr) |
a75beee6 ED |
920 | return i; |
921 | return -1; | |
922 | } | |
923 | ||
2384d2b3 SY |
924 | static inline void __invvpid(int ext, u16 vpid, gva_t gva) |
925 | { | |
926 | struct { | |
927 | u64 vpid : 16; | |
928 | u64 rsvd : 48; | |
929 | u64 gva; | |
930 | } operand = { vpid, 0, gva }; | |
931 | ||
4ecac3fd | 932 | asm volatile (__ex(ASM_VMX_INVVPID) |
2384d2b3 SY |
933 | /* CF==1 or ZF==1 --> rc = -1 */ |
934 | "; ja 1f ; ud2 ; 1:" | |
935 | : : "a"(&operand), "c"(ext) : "cc", "memory"); | |
936 | } | |
937 | ||
1439442c SY |
938 | static inline void __invept(int ext, u64 eptp, gpa_t gpa) |
939 | { | |
940 | struct { | |
941 | u64 eptp, gpa; | |
942 | } operand = {eptp, gpa}; | |
943 | ||
4ecac3fd | 944 | asm volatile (__ex(ASM_VMX_INVEPT) |
1439442c SY |
945 | /* CF==1 or ZF==1 --> rc = -1 */ |
946 | "; ja 1f ; ud2 ; 1:\n" | |
947 | : : "a" (&operand), "c" (ext) : "cc", "memory"); | |
948 | } | |
949 | ||
26bb0981 | 950 | static struct shared_msr_entry *find_msr_entry(struct vcpu_vmx *vmx, u32 msr) |
a75beee6 ED |
951 | { |
952 | int i; | |
953 | ||
8b9cf98c | 954 | i = __find_msr_index(vmx, msr); |
a75beee6 | 955 | if (i >= 0) |
a2fa3e9f | 956 | return &vmx->guest_msrs[i]; |
8b6d44c7 | 957 | return NULL; |
7725f0ba AK |
958 | } |
959 | ||
6aa8b732 AK |
960 | static void vmcs_clear(struct vmcs *vmcs) |
961 | { | |
962 | u64 phys_addr = __pa(vmcs); | |
963 | u8 error; | |
964 | ||
4ecac3fd | 965 | asm volatile (__ex(ASM_VMX_VMCLEAR_RAX) "; setna %0" |
16d8f72f | 966 | : "=qm"(error) : "a"(&phys_addr), "m"(phys_addr) |
6aa8b732 AK |
967 | : "cc", "memory"); |
968 | if (error) | |
969 | printk(KERN_ERR "kvm: vmclear fail: %p/%llx\n", | |
970 | vmcs, phys_addr); | |
971 | } | |
972 | ||
d462b819 NHE |
973 | static inline void loaded_vmcs_init(struct loaded_vmcs *loaded_vmcs) |
974 | { | |
975 | vmcs_clear(loaded_vmcs->vmcs); | |
976 | loaded_vmcs->cpu = -1; | |
977 | loaded_vmcs->launched = 0; | |
978 | } | |
979 | ||
7725b894 DX |
980 | static void vmcs_load(struct vmcs *vmcs) |
981 | { | |
982 | u64 phys_addr = __pa(vmcs); | |
983 | u8 error; | |
984 | ||
985 | asm volatile (__ex(ASM_VMX_VMPTRLD_RAX) "; setna %0" | |
16d8f72f | 986 | : "=qm"(error) : "a"(&phys_addr), "m"(phys_addr) |
7725b894 DX |
987 | : "cc", "memory"); |
988 | if (error) | |
2844d849 | 989 | printk(KERN_ERR "kvm: vmptrld %p/%llx failed\n", |
7725b894 DX |
990 | vmcs, phys_addr); |
991 | } | |
992 | ||
d462b819 | 993 | static void __loaded_vmcs_clear(void *arg) |
6aa8b732 | 994 | { |
d462b819 | 995 | struct loaded_vmcs *loaded_vmcs = arg; |
d3b2c338 | 996 | int cpu = raw_smp_processor_id(); |
6aa8b732 | 997 | |
d462b819 NHE |
998 | if (loaded_vmcs->cpu != cpu) |
999 | return; /* vcpu migration can race with cpu offline */ | |
1000 | if (per_cpu(current_vmcs, cpu) == loaded_vmcs->vmcs) | |
6aa8b732 | 1001 | per_cpu(current_vmcs, cpu) = NULL; |
d462b819 NHE |
1002 | list_del(&loaded_vmcs->loaded_vmcss_on_cpu_link); |
1003 | loaded_vmcs_init(loaded_vmcs); | |
6aa8b732 AK |
1004 | } |
1005 | ||
d462b819 | 1006 | static void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs) |
8d0be2b3 | 1007 | { |
d462b819 NHE |
1008 | if (loaded_vmcs->cpu != -1) |
1009 | smp_call_function_single( | |
1010 | loaded_vmcs->cpu, __loaded_vmcs_clear, loaded_vmcs, 1); | |
8d0be2b3 AK |
1011 | } |
1012 | ||
1760dd49 | 1013 | static inline void vpid_sync_vcpu_single(struct vcpu_vmx *vmx) |
2384d2b3 SY |
1014 | { |
1015 | if (vmx->vpid == 0) | |
1016 | return; | |
1017 | ||
518c8aee GJ |
1018 | if (cpu_has_vmx_invvpid_single()) |
1019 | __invvpid(VMX_VPID_EXTENT_SINGLE_CONTEXT, vmx->vpid, 0); | |
2384d2b3 SY |
1020 | } |
1021 | ||
b9d762fa GJ |
1022 | static inline void vpid_sync_vcpu_global(void) |
1023 | { | |
1024 | if (cpu_has_vmx_invvpid_global()) | |
1025 | __invvpid(VMX_VPID_EXTENT_ALL_CONTEXT, 0, 0); | |
1026 | } | |
1027 | ||
1028 | static inline void vpid_sync_context(struct vcpu_vmx *vmx) | |
1029 | { | |
1030 | if (cpu_has_vmx_invvpid_single()) | |
1760dd49 | 1031 | vpid_sync_vcpu_single(vmx); |
b9d762fa GJ |
1032 | else |
1033 | vpid_sync_vcpu_global(); | |
1034 | } | |
1035 | ||
1439442c SY |
1036 | static inline void ept_sync_global(void) |
1037 | { | |
1038 | if (cpu_has_vmx_invept_global()) | |
1039 | __invept(VMX_EPT_EXTENT_GLOBAL, 0, 0); | |
1040 | } | |
1041 | ||
1042 | static inline void ept_sync_context(u64 eptp) | |
1043 | { | |
089d034e | 1044 | if (enable_ept) { |
1439442c SY |
1045 | if (cpu_has_vmx_invept_context()) |
1046 | __invept(VMX_EPT_EXTENT_CONTEXT, eptp, 0); | |
1047 | else | |
1048 | ept_sync_global(); | |
1049 | } | |
1050 | } | |
1051 | ||
1052 | static inline void ept_sync_individual_addr(u64 eptp, gpa_t gpa) | |
1053 | { | |
089d034e | 1054 | if (enable_ept) { |
1439442c SY |
1055 | if (cpu_has_vmx_invept_individual_addr()) |
1056 | __invept(VMX_EPT_EXTENT_INDIVIDUAL_ADDR, | |
1057 | eptp, gpa); | |
1058 | else | |
1059 | ept_sync_context(eptp); | |
1060 | } | |
1061 | } | |
1062 | ||
96304217 | 1063 | static __always_inline unsigned long vmcs_readl(unsigned long field) |
6aa8b732 | 1064 | { |
5e520e62 | 1065 | unsigned long value; |
6aa8b732 | 1066 | |
5e520e62 AK |
1067 | asm volatile (__ex_clear(ASM_VMX_VMREAD_RDX_RAX, "%0") |
1068 | : "=a"(value) : "d"(field) : "cc"); | |
6aa8b732 AK |
1069 | return value; |
1070 | } | |
1071 | ||
96304217 | 1072 | static __always_inline u16 vmcs_read16(unsigned long field) |
6aa8b732 AK |
1073 | { |
1074 | return vmcs_readl(field); | |
1075 | } | |
1076 | ||
96304217 | 1077 | static __always_inline u32 vmcs_read32(unsigned long field) |
6aa8b732 AK |
1078 | { |
1079 | return vmcs_readl(field); | |
1080 | } | |
1081 | ||
96304217 | 1082 | static __always_inline u64 vmcs_read64(unsigned long field) |
6aa8b732 | 1083 | { |
05b3e0c2 | 1084 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
1085 | return vmcs_readl(field); |
1086 | #else | |
1087 | return vmcs_readl(field) | ((u64)vmcs_readl(field+1) << 32); | |
1088 | #endif | |
1089 | } | |
1090 | ||
e52de1b8 AK |
1091 | static noinline void vmwrite_error(unsigned long field, unsigned long value) |
1092 | { | |
1093 | printk(KERN_ERR "vmwrite error: reg %lx value %lx (err %d)\n", | |
1094 | field, value, vmcs_read32(VM_INSTRUCTION_ERROR)); | |
1095 | dump_stack(); | |
1096 | } | |
1097 | ||
6aa8b732 AK |
1098 | static void vmcs_writel(unsigned long field, unsigned long value) |
1099 | { | |
1100 | u8 error; | |
1101 | ||
4ecac3fd | 1102 | asm volatile (__ex(ASM_VMX_VMWRITE_RAX_RDX) "; setna %0" |
d77c26fc | 1103 | : "=q"(error) : "a"(value), "d"(field) : "cc"); |
e52de1b8 AK |
1104 | if (unlikely(error)) |
1105 | vmwrite_error(field, value); | |
6aa8b732 AK |
1106 | } |
1107 | ||
1108 | static void vmcs_write16(unsigned long field, u16 value) | |
1109 | { | |
1110 | vmcs_writel(field, value); | |
1111 | } | |
1112 | ||
1113 | static void vmcs_write32(unsigned long field, u32 value) | |
1114 | { | |
1115 | vmcs_writel(field, value); | |
1116 | } | |
1117 | ||
1118 | static void vmcs_write64(unsigned long field, u64 value) | |
1119 | { | |
6aa8b732 | 1120 | vmcs_writel(field, value); |
7682f2d0 | 1121 | #ifndef CONFIG_X86_64 |
6aa8b732 AK |
1122 | asm volatile (""); |
1123 | vmcs_writel(field+1, value >> 32); | |
1124 | #endif | |
1125 | } | |
1126 | ||
2ab455cc AL |
1127 | static void vmcs_clear_bits(unsigned long field, u32 mask) |
1128 | { | |
1129 | vmcs_writel(field, vmcs_readl(field) & ~mask); | |
1130 | } | |
1131 | ||
1132 | static void vmcs_set_bits(unsigned long field, u32 mask) | |
1133 | { | |
1134 | vmcs_writel(field, vmcs_readl(field) | mask); | |
1135 | } | |
1136 | ||
2fb92db1 AK |
1137 | static void vmx_segment_cache_clear(struct vcpu_vmx *vmx) |
1138 | { | |
1139 | vmx->segment_cache.bitmask = 0; | |
1140 | } | |
1141 | ||
1142 | static bool vmx_segment_cache_test_set(struct vcpu_vmx *vmx, unsigned seg, | |
1143 | unsigned field) | |
1144 | { | |
1145 | bool ret; | |
1146 | u32 mask = 1 << (seg * SEG_FIELD_NR + field); | |
1147 | ||
1148 | if (!(vmx->vcpu.arch.regs_avail & (1 << VCPU_EXREG_SEGMENTS))) { | |
1149 | vmx->vcpu.arch.regs_avail |= (1 << VCPU_EXREG_SEGMENTS); | |
1150 | vmx->segment_cache.bitmask = 0; | |
1151 | } | |
1152 | ret = vmx->segment_cache.bitmask & mask; | |
1153 | vmx->segment_cache.bitmask |= mask; | |
1154 | return ret; | |
1155 | } | |
1156 | ||
1157 | static u16 vmx_read_guest_seg_selector(struct vcpu_vmx *vmx, unsigned seg) | |
1158 | { | |
1159 | u16 *p = &vmx->segment_cache.seg[seg].selector; | |
1160 | ||
1161 | if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_SEL)) | |
1162 | *p = vmcs_read16(kvm_vmx_segment_fields[seg].selector); | |
1163 | return *p; | |
1164 | } | |
1165 | ||
1166 | static ulong vmx_read_guest_seg_base(struct vcpu_vmx *vmx, unsigned seg) | |
1167 | { | |
1168 | ulong *p = &vmx->segment_cache.seg[seg].base; | |
1169 | ||
1170 | if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_BASE)) | |
1171 | *p = vmcs_readl(kvm_vmx_segment_fields[seg].base); | |
1172 | return *p; | |
1173 | } | |
1174 | ||
1175 | static u32 vmx_read_guest_seg_limit(struct vcpu_vmx *vmx, unsigned seg) | |
1176 | { | |
1177 | u32 *p = &vmx->segment_cache.seg[seg].limit; | |
1178 | ||
1179 | if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_LIMIT)) | |
1180 | *p = vmcs_read32(kvm_vmx_segment_fields[seg].limit); | |
1181 | return *p; | |
1182 | } | |
1183 | ||
1184 | static u32 vmx_read_guest_seg_ar(struct vcpu_vmx *vmx, unsigned seg) | |
1185 | { | |
1186 | u32 *p = &vmx->segment_cache.seg[seg].ar; | |
1187 | ||
1188 | if (!vmx_segment_cache_test_set(vmx, seg, SEG_FIELD_AR)) | |
1189 | *p = vmcs_read32(kvm_vmx_segment_fields[seg].ar_bytes); | |
1190 | return *p; | |
1191 | } | |
1192 | ||
abd3f2d6 AK |
1193 | static void update_exception_bitmap(struct kvm_vcpu *vcpu) |
1194 | { | |
1195 | u32 eb; | |
1196 | ||
fd7373cc JK |
1197 | eb = (1u << PF_VECTOR) | (1u << UD_VECTOR) | (1u << MC_VECTOR) | |
1198 | (1u << NM_VECTOR) | (1u << DB_VECTOR); | |
1199 | if ((vcpu->guest_debug & | |
1200 | (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) == | |
1201 | (KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_USE_SW_BP)) | |
1202 | eb |= 1u << BP_VECTOR; | |
7ffd92c5 | 1203 | if (to_vmx(vcpu)->rmode.vm86_active) |
abd3f2d6 | 1204 | eb = ~0; |
089d034e | 1205 | if (enable_ept) |
1439442c | 1206 | eb &= ~(1u << PF_VECTOR); /* bypass_guest_pf = 0 */ |
02daab21 AK |
1207 | if (vcpu->fpu_active) |
1208 | eb &= ~(1u << NM_VECTOR); | |
36cf24e0 NHE |
1209 | |
1210 | /* When we are running a nested L2 guest and L1 specified for it a | |
1211 | * certain exception bitmap, we must trap the same exceptions and pass | |
1212 | * them to L1. When running L2, we will only handle the exceptions | |
1213 | * specified above if L1 did not want them. | |
1214 | */ | |
1215 | if (is_guest_mode(vcpu)) | |
1216 | eb |= get_vmcs12(vcpu)->exception_bitmap; | |
1217 | ||
abd3f2d6 AK |
1218 | vmcs_write32(EXCEPTION_BITMAP, eb); |
1219 | } | |
1220 | ||
8bf00a52 GN |
1221 | static void clear_atomic_switch_msr_special(unsigned long entry, |
1222 | unsigned long exit) | |
1223 | { | |
1224 | vmcs_clear_bits(VM_ENTRY_CONTROLS, entry); | |
1225 | vmcs_clear_bits(VM_EXIT_CONTROLS, exit); | |
1226 | } | |
1227 | ||
61d2ef2c AK |
1228 | static void clear_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr) |
1229 | { | |
1230 | unsigned i; | |
1231 | struct msr_autoload *m = &vmx->msr_autoload; | |
1232 | ||
8bf00a52 GN |
1233 | switch (msr) { |
1234 | case MSR_EFER: | |
1235 | if (cpu_has_load_ia32_efer) { | |
1236 | clear_atomic_switch_msr_special(VM_ENTRY_LOAD_IA32_EFER, | |
1237 | VM_EXIT_LOAD_IA32_EFER); | |
1238 | return; | |
1239 | } | |
1240 | break; | |
1241 | case MSR_CORE_PERF_GLOBAL_CTRL: | |
1242 | if (cpu_has_load_perf_global_ctrl) { | |
1243 | clear_atomic_switch_msr_special( | |
1244 | VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, | |
1245 | VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL); | |
1246 | return; | |
1247 | } | |
1248 | break; | |
110312c8 AK |
1249 | } |
1250 | ||
61d2ef2c AK |
1251 | for (i = 0; i < m->nr; ++i) |
1252 | if (m->guest[i].index == msr) | |
1253 | break; | |
1254 | ||
1255 | if (i == m->nr) | |
1256 | return; | |
1257 | --m->nr; | |
1258 | m->guest[i] = m->guest[m->nr]; | |
1259 | m->host[i] = m->host[m->nr]; | |
1260 | vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->nr); | |
1261 | vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr); | |
1262 | } | |
1263 | ||
8bf00a52 GN |
1264 | static void add_atomic_switch_msr_special(unsigned long entry, |
1265 | unsigned long exit, unsigned long guest_val_vmcs, | |
1266 | unsigned long host_val_vmcs, u64 guest_val, u64 host_val) | |
1267 | { | |
1268 | vmcs_write64(guest_val_vmcs, guest_val); | |
1269 | vmcs_write64(host_val_vmcs, host_val); | |
1270 | vmcs_set_bits(VM_ENTRY_CONTROLS, entry); | |
1271 | vmcs_set_bits(VM_EXIT_CONTROLS, exit); | |
1272 | } | |
1273 | ||
61d2ef2c AK |
1274 | static void add_atomic_switch_msr(struct vcpu_vmx *vmx, unsigned msr, |
1275 | u64 guest_val, u64 host_val) | |
1276 | { | |
1277 | unsigned i; | |
1278 | struct msr_autoload *m = &vmx->msr_autoload; | |
1279 | ||
8bf00a52 GN |
1280 | switch (msr) { |
1281 | case MSR_EFER: | |
1282 | if (cpu_has_load_ia32_efer) { | |
1283 | add_atomic_switch_msr_special(VM_ENTRY_LOAD_IA32_EFER, | |
1284 | VM_EXIT_LOAD_IA32_EFER, | |
1285 | GUEST_IA32_EFER, | |
1286 | HOST_IA32_EFER, | |
1287 | guest_val, host_val); | |
1288 | return; | |
1289 | } | |
1290 | break; | |
1291 | case MSR_CORE_PERF_GLOBAL_CTRL: | |
1292 | if (cpu_has_load_perf_global_ctrl) { | |
1293 | add_atomic_switch_msr_special( | |
1294 | VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL, | |
1295 | VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL, | |
1296 | GUEST_IA32_PERF_GLOBAL_CTRL, | |
1297 | HOST_IA32_PERF_GLOBAL_CTRL, | |
1298 | guest_val, host_val); | |
1299 | return; | |
1300 | } | |
1301 | break; | |
110312c8 AK |
1302 | } |
1303 | ||
61d2ef2c AK |
1304 | for (i = 0; i < m->nr; ++i) |
1305 | if (m->guest[i].index == msr) | |
1306 | break; | |
1307 | ||
e7fc6f93 GN |
1308 | if (i == NR_AUTOLOAD_MSRS) { |
1309 | printk_once(KERN_WARNING"Not enough mst switch entries. " | |
1310 | "Can't add msr %x\n", msr); | |
1311 | return; | |
1312 | } else if (i == m->nr) { | |
61d2ef2c AK |
1313 | ++m->nr; |
1314 | vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, m->nr); | |
1315 | vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, m->nr); | |
1316 | } | |
1317 | ||
1318 | m->guest[i].index = msr; | |
1319 | m->guest[i].value = guest_val; | |
1320 | m->host[i].index = msr; | |
1321 | m->host[i].value = host_val; | |
1322 | } | |
1323 | ||
33ed6329 AK |
1324 | static void reload_tss(void) |
1325 | { | |
33ed6329 AK |
1326 | /* |
1327 | * VT restores TR but not its size. Useless. | |
1328 | */ | |
d359192f | 1329 | struct desc_ptr *gdt = &__get_cpu_var(host_gdt); |
a5f61300 | 1330 | struct desc_struct *descs; |
33ed6329 | 1331 | |
d359192f | 1332 | descs = (void *)gdt->address; |
33ed6329 AK |
1333 | descs[GDT_ENTRY_TSS].type = 9; /* available TSS */ |
1334 | load_TR_desc(); | |
33ed6329 AK |
1335 | } |
1336 | ||
92c0d900 | 1337 | static bool update_transition_efer(struct vcpu_vmx *vmx, int efer_offset) |
2cc51560 | 1338 | { |
3a34a881 | 1339 | u64 guest_efer; |
51c6cf66 AK |
1340 | u64 ignore_bits; |
1341 | ||
f6801dff | 1342 | guest_efer = vmx->vcpu.arch.efer; |
3a34a881 | 1343 | |
51c6cf66 | 1344 | /* |
0fa06071 | 1345 | * NX is emulated; LMA and LME handled by hardware; SCE meaningless |
51c6cf66 AK |
1346 | * outside long mode |
1347 | */ | |
1348 | ignore_bits = EFER_NX | EFER_SCE; | |
1349 | #ifdef CONFIG_X86_64 | |
1350 | ignore_bits |= EFER_LMA | EFER_LME; | |
1351 | /* SCE is meaningful only in long mode on Intel */ | |
1352 | if (guest_efer & EFER_LMA) | |
1353 | ignore_bits &= ~(u64)EFER_SCE; | |
1354 | #endif | |
51c6cf66 AK |
1355 | guest_efer &= ~ignore_bits; |
1356 | guest_efer |= host_efer & ignore_bits; | |
26bb0981 | 1357 | vmx->guest_msrs[efer_offset].data = guest_efer; |
d5696725 | 1358 | vmx->guest_msrs[efer_offset].mask = ~ignore_bits; |
84ad33ef AK |
1359 | |
1360 | clear_atomic_switch_msr(vmx, MSR_EFER); | |
1361 | /* On ept, can't emulate nx, and must switch nx atomically */ | |
1362 | if (enable_ept && ((vmx->vcpu.arch.efer ^ host_efer) & EFER_NX)) { | |
1363 | guest_efer = vmx->vcpu.arch.efer; | |
1364 | if (!(guest_efer & EFER_LMA)) | |
1365 | guest_efer &= ~EFER_LME; | |
1366 | add_atomic_switch_msr(vmx, MSR_EFER, guest_efer, host_efer); | |
1367 | return false; | |
1368 | } | |
1369 | ||
26bb0981 | 1370 | return true; |
51c6cf66 AK |
1371 | } |
1372 | ||
2d49ec72 GN |
1373 | static unsigned long segment_base(u16 selector) |
1374 | { | |
d359192f | 1375 | struct desc_ptr *gdt = &__get_cpu_var(host_gdt); |
2d49ec72 GN |
1376 | struct desc_struct *d; |
1377 | unsigned long table_base; | |
1378 | unsigned long v; | |
1379 | ||
1380 | if (!(selector & ~3)) | |
1381 | return 0; | |
1382 | ||
d359192f | 1383 | table_base = gdt->address; |
2d49ec72 GN |
1384 | |
1385 | if (selector & 4) { /* from ldt */ | |
1386 | u16 ldt_selector = kvm_read_ldt(); | |
1387 | ||
1388 | if (!(ldt_selector & ~3)) | |
1389 | return 0; | |
1390 | ||
1391 | table_base = segment_base(ldt_selector); | |
1392 | } | |
1393 | d = (struct desc_struct *)(table_base + (selector & ~7)); | |
1394 | v = get_desc_base(d); | |
1395 | #ifdef CONFIG_X86_64 | |
1396 | if (d->s == 0 && (d->type == 2 || d->type == 9 || d->type == 11)) | |
1397 | v |= ((unsigned long)((struct ldttss_desc64 *)d)->base3) << 32; | |
1398 | #endif | |
1399 | return v; | |
1400 | } | |
1401 | ||
1402 | static inline unsigned long kvm_read_tr_base(void) | |
1403 | { | |
1404 | u16 tr; | |
1405 | asm("str %0" : "=g"(tr)); | |
1406 | return segment_base(tr); | |
1407 | } | |
1408 | ||
04d2cc77 | 1409 | static void vmx_save_host_state(struct kvm_vcpu *vcpu) |
33ed6329 | 1410 | { |
04d2cc77 | 1411 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
26bb0981 | 1412 | int i; |
04d2cc77 | 1413 | |
a2fa3e9f | 1414 | if (vmx->host_state.loaded) |
33ed6329 AK |
1415 | return; |
1416 | ||
a2fa3e9f | 1417 | vmx->host_state.loaded = 1; |
33ed6329 AK |
1418 | /* |
1419 | * Set host fs and gs selectors. Unfortunately, 22.2.3 does not | |
1420 | * allow segment selectors with cpl > 0 or ti == 1. | |
1421 | */ | |
d6e88aec | 1422 | vmx->host_state.ldt_sel = kvm_read_ldt(); |
152d3f2f | 1423 | vmx->host_state.gs_ldt_reload_needed = vmx->host_state.ldt_sel; |
9581d442 | 1424 | savesegment(fs, vmx->host_state.fs_sel); |
152d3f2f | 1425 | if (!(vmx->host_state.fs_sel & 7)) { |
a2fa3e9f | 1426 | vmcs_write16(HOST_FS_SELECTOR, vmx->host_state.fs_sel); |
152d3f2f LV |
1427 | vmx->host_state.fs_reload_needed = 0; |
1428 | } else { | |
33ed6329 | 1429 | vmcs_write16(HOST_FS_SELECTOR, 0); |
152d3f2f | 1430 | vmx->host_state.fs_reload_needed = 1; |
33ed6329 | 1431 | } |
9581d442 | 1432 | savesegment(gs, vmx->host_state.gs_sel); |
a2fa3e9f GH |
1433 | if (!(vmx->host_state.gs_sel & 7)) |
1434 | vmcs_write16(HOST_GS_SELECTOR, vmx->host_state.gs_sel); | |
33ed6329 AK |
1435 | else { |
1436 | vmcs_write16(HOST_GS_SELECTOR, 0); | |
152d3f2f | 1437 | vmx->host_state.gs_ldt_reload_needed = 1; |
33ed6329 AK |
1438 | } |
1439 | ||
b2da15ac AK |
1440 | #ifdef CONFIG_X86_64 |
1441 | savesegment(ds, vmx->host_state.ds_sel); | |
1442 | savesegment(es, vmx->host_state.es_sel); | |
1443 | #endif | |
1444 | ||
33ed6329 AK |
1445 | #ifdef CONFIG_X86_64 |
1446 | vmcs_writel(HOST_FS_BASE, read_msr(MSR_FS_BASE)); | |
1447 | vmcs_writel(HOST_GS_BASE, read_msr(MSR_GS_BASE)); | |
1448 | #else | |
a2fa3e9f GH |
1449 | vmcs_writel(HOST_FS_BASE, segment_base(vmx->host_state.fs_sel)); |
1450 | vmcs_writel(HOST_GS_BASE, segment_base(vmx->host_state.gs_sel)); | |
33ed6329 | 1451 | #endif |
707c0874 AK |
1452 | |
1453 | #ifdef CONFIG_X86_64 | |
c8770e7b AK |
1454 | rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); |
1455 | if (is_long_mode(&vmx->vcpu)) | |
44ea2b17 | 1456 | wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); |
707c0874 | 1457 | #endif |
26bb0981 AK |
1458 | for (i = 0; i < vmx->save_nmsrs; ++i) |
1459 | kvm_set_shared_msr(vmx->guest_msrs[i].index, | |
d5696725 AK |
1460 | vmx->guest_msrs[i].data, |
1461 | vmx->guest_msrs[i].mask); | |
33ed6329 AK |
1462 | } |
1463 | ||
a9b21b62 | 1464 | static void __vmx_load_host_state(struct vcpu_vmx *vmx) |
33ed6329 | 1465 | { |
a2fa3e9f | 1466 | if (!vmx->host_state.loaded) |
33ed6329 AK |
1467 | return; |
1468 | ||
e1beb1d3 | 1469 | ++vmx->vcpu.stat.host_state_reload; |
a2fa3e9f | 1470 | vmx->host_state.loaded = 0; |
c8770e7b AK |
1471 | #ifdef CONFIG_X86_64 |
1472 | if (is_long_mode(&vmx->vcpu)) | |
1473 | rdmsrl(MSR_KERNEL_GS_BASE, vmx->msr_guest_kernel_gs_base); | |
1474 | #endif | |
152d3f2f | 1475 | if (vmx->host_state.gs_ldt_reload_needed) { |
d6e88aec | 1476 | kvm_load_ldt(vmx->host_state.ldt_sel); |
33ed6329 | 1477 | #ifdef CONFIG_X86_64 |
9581d442 | 1478 | load_gs_index(vmx->host_state.gs_sel); |
9581d442 AK |
1479 | #else |
1480 | loadsegment(gs, vmx->host_state.gs_sel); | |
33ed6329 | 1481 | #endif |
33ed6329 | 1482 | } |
0a77fe4c AK |
1483 | if (vmx->host_state.fs_reload_needed) |
1484 | loadsegment(fs, vmx->host_state.fs_sel); | |
b2da15ac AK |
1485 | #ifdef CONFIG_X86_64 |
1486 | if (unlikely(vmx->host_state.ds_sel | vmx->host_state.es_sel)) { | |
1487 | loadsegment(ds, vmx->host_state.ds_sel); | |
1488 | loadsegment(es, vmx->host_state.es_sel); | |
1489 | } | |
b2da15ac | 1490 | #endif |
152d3f2f | 1491 | reload_tss(); |
44ea2b17 | 1492 | #ifdef CONFIG_X86_64 |
c8770e7b | 1493 | wrmsrl(MSR_KERNEL_GS_BASE, vmx->msr_host_kernel_gs_base); |
44ea2b17 | 1494 | #endif |
1361b83a | 1495 | if (user_has_fpu()) |
1c11e713 | 1496 | clts(); |
3444d7da | 1497 | load_gdt(&__get_cpu_var(host_gdt)); |
33ed6329 AK |
1498 | } |
1499 | ||
a9b21b62 AK |
1500 | static void vmx_load_host_state(struct vcpu_vmx *vmx) |
1501 | { | |
1502 | preempt_disable(); | |
1503 | __vmx_load_host_state(vmx); | |
1504 | preempt_enable(); | |
1505 | } | |
1506 | ||
6aa8b732 AK |
1507 | /* |
1508 | * Switches to specified vcpu, until a matching vcpu_put(), but assumes | |
1509 | * vcpu mutex is already taken. | |
1510 | */ | |
15ad7146 | 1511 | static void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
6aa8b732 | 1512 | { |
a2fa3e9f | 1513 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
4610c9cc | 1514 | u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); |
6aa8b732 | 1515 | |
4610c9cc DX |
1516 | if (!vmm_exclusive) |
1517 | kvm_cpu_vmxon(phys_addr); | |
d462b819 NHE |
1518 | else if (vmx->loaded_vmcs->cpu != cpu) |
1519 | loaded_vmcs_clear(vmx->loaded_vmcs); | |
6aa8b732 | 1520 | |
d462b819 NHE |
1521 | if (per_cpu(current_vmcs, cpu) != vmx->loaded_vmcs->vmcs) { |
1522 | per_cpu(current_vmcs, cpu) = vmx->loaded_vmcs->vmcs; | |
1523 | vmcs_load(vmx->loaded_vmcs->vmcs); | |
6aa8b732 AK |
1524 | } |
1525 | ||
d462b819 | 1526 | if (vmx->loaded_vmcs->cpu != cpu) { |
d359192f | 1527 | struct desc_ptr *gdt = &__get_cpu_var(host_gdt); |
6aa8b732 AK |
1528 | unsigned long sysenter_esp; |
1529 | ||
a8eeb04a | 1530 | kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu); |
92fe13be | 1531 | local_irq_disable(); |
d462b819 NHE |
1532 | list_add(&vmx->loaded_vmcs->loaded_vmcss_on_cpu_link, |
1533 | &per_cpu(loaded_vmcss_on_cpu, cpu)); | |
92fe13be DX |
1534 | local_irq_enable(); |
1535 | ||
6aa8b732 AK |
1536 | /* |
1537 | * Linux uses per-cpu TSS and GDT, so set these when switching | |
1538 | * processors. | |
1539 | */ | |
d6e88aec | 1540 | vmcs_writel(HOST_TR_BASE, kvm_read_tr_base()); /* 22.2.4 */ |
d359192f | 1541 | vmcs_writel(HOST_GDTR_BASE, gdt->address); /* 22.2.4 */ |
6aa8b732 AK |
1542 | |
1543 | rdmsrl(MSR_IA32_SYSENTER_ESP, sysenter_esp); | |
1544 | vmcs_writel(HOST_IA32_SYSENTER_ESP, sysenter_esp); /* 22.2.3 */ | |
d462b819 | 1545 | vmx->loaded_vmcs->cpu = cpu; |
6aa8b732 | 1546 | } |
6aa8b732 AK |
1547 | } |
1548 | ||
1549 | static void vmx_vcpu_put(struct kvm_vcpu *vcpu) | |
1550 | { | |
a9b21b62 | 1551 | __vmx_load_host_state(to_vmx(vcpu)); |
4610c9cc | 1552 | if (!vmm_exclusive) { |
d462b819 NHE |
1553 | __loaded_vmcs_clear(to_vmx(vcpu)->loaded_vmcs); |
1554 | vcpu->cpu = -1; | |
4610c9cc DX |
1555 | kvm_cpu_vmxoff(); |
1556 | } | |
6aa8b732 AK |
1557 | } |
1558 | ||
5fd86fcf AK |
1559 | static void vmx_fpu_activate(struct kvm_vcpu *vcpu) |
1560 | { | |
81231c69 AK |
1561 | ulong cr0; |
1562 | ||
5fd86fcf AK |
1563 | if (vcpu->fpu_active) |
1564 | return; | |
1565 | vcpu->fpu_active = 1; | |
81231c69 AK |
1566 | cr0 = vmcs_readl(GUEST_CR0); |
1567 | cr0 &= ~(X86_CR0_TS | X86_CR0_MP); | |
1568 | cr0 |= kvm_read_cr0_bits(vcpu, X86_CR0_TS | X86_CR0_MP); | |
1569 | vmcs_writel(GUEST_CR0, cr0); | |
5fd86fcf | 1570 | update_exception_bitmap(vcpu); |
edcafe3c | 1571 | vcpu->arch.cr0_guest_owned_bits = X86_CR0_TS; |
36cf24e0 NHE |
1572 | if (is_guest_mode(vcpu)) |
1573 | vcpu->arch.cr0_guest_owned_bits &= | |
1574 | ~get_vmcs12(vcpu)->cr0_guest_host_mask; | |
edcafe3c | 1575 | vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits); |
5fd86fcf AK |
1576 | } |
1577 | ||
edcafe3c AK |
1578 | static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu); |
1579 | ||
fe3ef05c NHE |
1580 | /* |
1581 | * Return the cr0 value that a nested guest would read. This is a combination | |
1582 | * of the real cr0 used to run the guest (guest_cr0), and the bits shadowed by | |
1583 | * its hypervisor (cr0_read_shadow). | |
1584 | */ | |
1585 | static inline unsigned long nested_read_cr0(struct vmcs12 *fields) | |
1586 | { | |
1587 | return (fields->guest_cr0 & ~fields->cr0_guest_host_mask) | | |
1588 | (fields->cr0_read_shadow & fields->cr0_guest_host_mask); | |
1589 | } | |
1590 | static inline unsigned long nested_read_cr4(struct vmcs12 *fields) | |
1591 | { | |
1592 | return (fields->guest_cr4 & ~fields->cr4_guest_host_mask) | | |
1593 | (fields->cr4_read_shadow & fields->cr4_guest_host_mask); | |
1594 | } | |
1595 | ||
5fd86fcf AK |
1596 | static void vmx_fpu_deactivate(struct kvm_vcpu *vcpu) |
1597 | { | |
36cf24e0 NHE |
1598 | /* Note that there is no vcpu->fpu_active = 0 here. The caller must |
1599 | * set this *before* calling this function. | |
1600 | */ | |
edcafe3c | 1601 | vmx_decache_cr0_guest_bits(vcpu); |
81231c69 | 1602 | vmcs_set_bits(GUEST_CR0, X86_CR0_TS | X86_CR0_MP); |
5fd86fcf | 1603 | update_exception_bitmap(vcpu); |
edcafe3c AK |
1604 | vcpu->arch.cr0_guest_owned_bits = 0; |
1605 | vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits); | |
36cf24e0 NHE |
1606 | if (is_guest_mode(vcpu)) { |
1607 | /* | |
1608 | * L1's specified read shadow might not contain the TS bit, | |
1609 | * so now that we turned on shadowing of this bit, we need to | |
1610 | * set this bit of the shadow. Like in nested_vmx_run we need | |
1611 | * nested_read_cr0(vmcs12), but vmcs12->guest_cr0 is not yet | |
1612 | * up-to-date here because we just decached cr0.TS (and we'll | |
1613 | * only update vmcs12->guest_cr0 on nested exit). | |
1614 | */ | |
1615 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); | |
1616 | vmcs12->guest_cr0 = (vmcs12->guest_cr0 & ~X86_CR0_TS) | | |
1617 | (vcpu->arch.cr0 & X86_CR0_TS); | |
1618 | vmcs_writel(CR0_READ_SHADOW, nested_read_cr0(vmcs12)); | |
1619 | } else | |
1620 | vmcs_writel(CR0_READ_SHADOW, vcpu->arch.cr0); | |
5fd86fcf AK |
1621 | } |
1622 | ||
6aa8b732 AK |
1623 | static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu) |
1624 | { | |
78ac8b47 | 1625 | unsigned long rflags, save_rflags; |
345dcaa8 | 1626 | |
6de12732 AK |
1627 | if (!test_bit(VCPU_EXREG_RFLAGS, (ulong *)&vcpu->arch.regs_avail)) { |
1628 | __set_bit(VCPU_EXREG_RFLAGS, (ulong *)&vcpu->arch.regs_avail); | |
1629 | rflags = vmcs_readl(GUEST_RFLAGS); | |
1630 | if (to_vmx(vcpu)->rmode.vm86_active) { | |
1631 | rflags &= RMODE_GUEST_OWNED_EFLAGS_BITS; | |
1632 | save_rflags = to_vmx(vcpu)->rmode.save_rflags; | |
1633 | rflags |= save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; | |
1634 | } | |
1635 | to_vmx(vcpu)->rflags = rflags; | |
78ac8b47 | 1636 | } |
6de12732 | 1637 | return to_vmx(vcpu)->rflags; |
6aa8b732 AK |
1638 | } |
1639 | ||
1640 | static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags) | |
1641 | { | |
6de12732 | 1642 | __set_bit(VCPU_EXREG_RFLAGS, (ulong *)&vcpu->arch.regs_avail); |
69c73028 | 1643 | __clear_bit(VCPU_EXREG_CPL, (ulong *)&vcpu->arch.regs_avail); |
6de12732 | 1644 | to_vmx(vcpu)->rflags = rflags; |
78ac8b47 AK |
1645 | if (to_vmx(vcpu)->rmode.vm86_active) { |
1646 | to_vmx(vcpu)->rmode.save_rflags = rflags; | |
053de044 | 1647 | rflags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
78ac8b47 | 1648 | } |
6aa8b732 AK |
1649 | vmcs_writel(GUEST_RFLAGS, rflags); |
1650 | } | |
1651 | ||
2809f5d2 GC |
1652 | static u32 vmx_get_interrupt_shadow(struct kvm_vcpu *vcpu, int mask) |
1653 | { | |
1654 | u32 interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); | |
1655 | int ret = 0; | |
1656 | ||
1657 | if (interruptibility & GUEST_INTR_STATE_STI) | |
48005f64 | 1658 | ret |= KVM_X86_SHADOW_INT_STI; |
2809f5d2 | 1659 | if (interruptibility & GUEST_INTR_STATE_MOV_SS) |
48005f64 | 1660 | ret |= KVM_X86_SHADOW_INT_MOV_SS; |
2809f5d2 GC |
1661 | |
1662 | return ret & mask; | |
1663 | } | |
1664 | ||
1665 | static void vmx_set_interrupt_shadow(struct kvm_vcpu *vcpu, int mask) | |
1666 | { | |
1667 | u32 interruptibility_old = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); | |
1668 | u32 interruptibility = interruptibility_old; | |
1669 | ||
1670 | interruptibility &= ~(GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS); | |
1671 | ||
48005f64 | 1672 | if (mask & KVM_X86_SHADOW_INT_MOV_SS) |
2809f5d2 | 1673 | interruptibility |= GUEST_INTR_STATE_MOV_SS; |
48005f64 | 1674 | else if (mask & KVM_X86_SHADOW_INT_STI) |
2809f5d2 GC |
1675 | interruptibility |= GUEST_INTR_STATE_STI; |
1676 | ||
1677 | if ((interruptibility != interruptibility_old)) | |
1678 | vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, interruptibility); | |
1679 | } | |
1680 | ||
6aa8b732 AK |
1681 | static void skip_emulated_instruction(struct kvm_vcpu *vcpu) |
1682 | { | |
1683 | unsigned long rip; | |
6aa8b732 | 1684 | |
5fdbf976 | 1685 | rip = kvm_rip_read(vcpu); |
6aa8b732 | 1686 | rip += vmcs_read32(VM_EXIT_INSTRUCTION_LEN); |
5fdbf976 | 1687 | kvm_rip_write(vcpu, rip); |
6aa8b732 | 1688 | |
2809f5d2 GC |
1689 | /* skipping an emulated instruction also counts */ |
1690 | vmx_set_interrupt_shadow(vcpu, 0); | |
6aa8b732 AK |
1691 | } |
1692 | ||
0b6ac343 NHE |
1693 | /* |
1694 | * KVM wants to inject page-faults which it got to the guest. This function | |
1695 | * checks whether in a nested guest, we need to inject them to L1 or L2. | |
1696 | * This function assumes it is called with the exit reason in vmcs02 being | |
1697 | * a #PF exception (this is the only case in which KVM injects a #PF when L2 | |
1698 | * is running). | |
1699 | */ | |
1700 | static int nested_pf_handled(struct kvm_vcpu *vcpu) | |
1701 | { | |
1702 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); | |
1703 | ||
1704 | /* TODO: also check PFEC_MATCH/MASK, not just EB.PF. */ | |
95871901 | 1705 | if (!(vmcs12->exception_bitmap & (1u << PF_VECTOR))) |
0b6ac343 NHE |
1706 | return 0; |
1707 | ||
1708 | nested_vmx_vmexit(vcpu); | |
1709 | return 1; | |
1710 | } | |
1711 | ||
298101da | 1712 | static void vmx_queue_exception(struct kvm_vcpu *vcpu, unsigned nr, |
ce7ddec4 JR |
1713 | bool has_error_code, u32 error_code, |
1714 | bool reinject) | |
298101da | 1715 | { |
77ab6db0 | 1716 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
8ab2d2e2 | 1717 | u32 intr_info = nr | INTR_INFO_VALID_MASK; |
77ab6db0 | 1718 | |
0b6ac343 NHE |
1719 | if (nr == PF_VECTOR && is_guest_mode(vcpu) && |
1720 | nested_pf_handled(vcpu)) | |
1721 | return; | |
1722 | ||
8ab2d2e2 | 1723 | if (has_error_code) { |
77ab6db0 | 1724 | vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code); |
8ab2d2e2 JK |
1725 | intr_info |= INTR_INFO_DELIVER_CODE_MASK; |
1726 | } | |
77ab6db0 | 1727 | |
7ffd92c5 | 1728 | if (vmx->rmode.vm86_active) { |
71f9833b SH |
1729 | int inc_eip = 0; |
1730 | if (kvm_exception_is_soft(nr)) | |
1731 | inc_eip = vcpu->arch.event_exit_inst_len; | |
1732 | if (kvm_inject_realmode_interrupt(vcpu, nr, inc_eip) != EMULATE_DONE) | |
a92601bb | 1733 | kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu); |
77ab6db0 JK |
1734 | return; |
1735 | } | |
1736 | ||
66fd3f7f GN |
1737 | if (kvm_exception_is_soft(nr)) { |
1738 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, | |
1739 | vmx->vcpu.arch.event_exit_inst_len); | |
8ab2d2e2 JK |
1740 | intr_info |= INTR_TYPE_SOFT_EXCEPTION; |
1741 | } else | |
1742 | intr_info |= INTR_TYPE_HARD_EXCEPTION; | |
1743 | ||
1744 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr_info); | |
298101da AK |
1745 | } |
1746 | ||
4e47c7a6 SY |
1747 | static bool vmx_rdtscp_supported(void) |
1748 | { | |
1749 | return cpu_has_vmx_rdtscp(); | |
1750 | } | |
1751 | ||
ad756a16 MJ |
1752 | static bool vmx_invpcid_supported(void) |
1753 | { | |
1754 | return cpu_has_vmx_invpcid() && enable_ept; | |
1755 | } | |
1756 | ||
a75beee6 ED |
1757 | /* |
1758 | * Swap MSR entry in host/guest MSR entry array. | |
1759 | */ | |
8b9cf98c | 1760 | static void move_msr_up(struct vcpu_vmx *vmx, int from, int to) |
a75beee6 | 1761 | { |
26bb0981 | 1762 | struct shared_msr_entry tmp; |
a2fa3e9f GH |
1763 | |
1764 | tmp = vmx->guest_msrs[to]; | |
1765 | vmx->guest_msrs[to] = vmx->guest_msrs[from]; | |
1766 | vmx->guest_msrs[from] = tmp; | |
a75beee6 ED |
1767 | } |
1768 | ||
e38aea3e AK |
1769 | /* |
1770 | * Set up the vmcs to automatically save and restore system | |
1771 | * msrs. Don't touch the 64-bit msrs if the guest is in legacy | |
1772 | * mode, as fiddling with msrs is very expensive. | |
1773 | */ | |
8b9cf98c | 1774 | static void setup_msrs(struct vcpu_vmx *vmx) |
e38aea3e | 1775 | { |
26bb0981 | 1776 | int save_nmsrs, index; |
5897297b | 1777 | unsigned long *msr_bitmap; |
e38aea3e | 1778 | |
a75beee6 ED |
1779 | save_nmsrs = 0; |
1780 | #ifdef CONFIG_X86_64 | |
8b9cf98c | 1781 | if (is_long_mode(&vmx->vcpu)) { |
8b9cf98c | 1782 | index = __find_msr_index(vmx, MSR_SYSCALL_MASK); |
a75beee6 | 1783 | if (index >= 0) |
8b9cf98c RR |
1784 | move_msr_up(vmx, index, save_nmsrs++); |
1785 | index = __find_msr_index(vmx, MSR_LSTAR); | |
a75beee6 | 1786 | if (index >= 0) |
8b9cf98c RR |
1787 | move_msr_up(vmx, index, save_nmsrs++); |
1788 | index = __find_msr_index(vmx, MSR_CSTAR); | |
a75beee6 | 1789 | if (index >= 0) |
8b9cf98c | 1790 | move_msr_up(vmx, index, save_nmsrs++); |
4e47c7a6 SY |
1791 | index = __find_msr_index(vmx, MSR_TSC_AUX); |
1792 | if (index >= 0 && vmx->rdtscp_enabled) | |
1793 | move_msr_up(vmx, index, save_nmsrs++); | |
a75beee6 | 1794 | /* |
8c06585d | 1795 | * MSR_STAR is only needed on long mode guests, and only |
a75beee6 ED |
1796 | * if efer.sce is enabled. |
1797 | */ | |
8c06585d | 1798 | index = __find_msr_index(vmx, MSR_STAR); |
f6801dff | 1799 | if ((index >= 0) && (vmx->vcpu.arch.efer & EFER_SCE)) |
8b9cf98c | 1800 | move_msr_up(vmx, index, save_nmsrs++); |
a75beee6 ED |
1801 | } |
1802 | #endif | |
92c0d900 AK |
1803 | index = __find_msr_index(vmx, MSR_EFER); |
1804 | if (index >= 0 && update_transition_efer(vmx, index)) | |
26bb0981 | 1805 | move_msr_up(vmx, index, save_nmsrs++); |
e38aea3e | 1806 | |
26bb0981 | 1807 | vmx->save_nmsrs = save_nmsrs; |
5897297b AK |
1808 | |
1809 | if (cpu_has_vmx_msr_bitmap()) { | |
1810 | if (is_long_mode(&vmx->vcpu)) | |
1811 | msr_bitmap = vmx_msr_bitmap_longmode; | |
1812 | else | |
1813 | msr_bitmap = vmx_msr_bitmap_legacy; | |
1814 | ||
1815 | vmcs_write64(MSR_BITMAP, __pa(msr_bitmap)); | |
1816 | } | |
e38aea3e AK |
1817 | } |
1818 | ||
6aa8b732 AK |
1819 | /* |
1820 | * reads and returns guest's timestamp counter "register" | |
1821 | * guest_tsc = host_tsc + tsc_offset -- 21.3 | |
1822 | */ | |
1823 | static u64 guest_read_tsc(void) | |
1824 | { | |
1825 | u64 host_tsc, tsc_offset; | |
1826 | ||
1827 | rdtscll(host_tsc); | |
1828 | tsc_offset = vmcs_read64(TSC_OFFSET); | |
1829 | return host_tsc + tsc_offset; | |
1830 | } | |
1831 | ||
d5c1785d NHE |
1832 | /* |
1833 | * Like guest_read_tsc, but always returns L1's notion of the timestamp | |
1834 | * counter, even if a nested guest (L2) is currently running. | |
1835 | */ | |
1836 | u64 vmx_read_l1_tsc(struct kvm_vcpu *vcpu) | |
1837 | { | |
1838 | u64 host_tsc, tsc_offset; | |
1839 | ||
1840 | rdtscll(host_tsc); | |
1841 | tsc_offset = is_guest_mode(vcpu) ? | |
1842 | to_vmx(vcpu)->nested.vmcs01_tsc_offset : | |
1843 | vmcs_read64(TSC_OFFSET); | |
1844 | return host_tsc + tsc_offset; | |
1845 | } | |
1846 | ||
4051b188 | 1847 | /* |
cc578287 ZA |
1848 | * Engage any workarounds for mis-matched TSC rates. Currently limited to |
1849 | * software catchup for faster rates on slower CPUs. | |
4051b188 | 1850 | */ |
cc578287 | 1851 | static void vmx_set_tsc_khz(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale) |
4051b188 | 1852 | { |
cc578287 ZA |
1853 | if (!scale) |
1854 | return; | |
1855 | ||
1856 | if (user_tsc_khz > tsc_khz) { | |
1857 | vcpu->arch.tsc_catchup = 1; | |
1858 | vcpu->arch.tsc_always_catchup = 1; | |
1859 | } else | |
1860 | WARN(1, "user requested TSC rate below hardware speed\n"); | |
4051b188 JR |
1861 | } |
1862 | ||
6aa8b732 | 1863 | /* |
99e3e30a | 1864 | * writes 'offset' into guest's timestamp counter offset register |
6aa8b732 | 1865 | */ |
99e3e30a | 1866 | static void vmx_write_tsc_offset(struct kvm_vcpu *vcpu, u64 offset) |
6aa8b732 | 1867 | { |
27fc51b2 | 1868 | if (is_guest_mode(vcpu)) { |
7991825b | 1869 | /* |
27fc51b2 NHE |
1870 | * We're here if L1 chose not to trap WRMSR to TSC. According |
1871 | * to the spec, this should set L1's TSC; The offset that L1 | |
1872 | * set for L2 remains unchanged, and still needs to be added | |
1873 | * to the newly set TSC to get L2's TSC. | |
7991825b | 1874 | */ |
27fc51b2 NHE |
1875 | struct vmcs12 *vmcs12; |
1876 | to_vmx(vcpu)->nested.vmcs01_tsc_offset = offset; | |
1877 | /* recalculate vmcs02.TSC_OFFSET: */ | |
1878 | vmcs12 = get_vmcs12(vcpu); | |
1879 | vmcs_write64(TSC_OFFSET, offset + | |
1880 | (nested_cpu_has(vmcs12, CPU_BASED_USE_TSC_OFFSETING) ? | |
1881 | vmcs12->tsc_offset : 0)); | |
1882 | } else { | |
1883 | vmcs_write64(TSC_OFFSET, offset); | |
1884 | } | |
6aa8b732 AK |
1885 | } |
1886 | ||
f1e2b260 | 1887 | static void vmx_adjust_tsc_offset(struct kvm_vcpu *vcpu, s64 adjustment, bool host) |
e48672fa ZA |
1888 | { |
1889 | u64 offset = vmcs_read64(TSC_OFFSET); | |
1890 | vmcs_write64(TSC_OFFSET, offset + adjustment); | |
7991825b NHE |
1891 | if (is_guest_mode(vcpu)) { |
1892 | /* Even when running L2, the adjustment needs to apply to L1 */ | |
1893 | to_vmx(vcpu)->nested.vmcs01_tsc_offset += adjustment; | |
1894 | } | |
e48672fa ZA |
1895 | } |
1896 | ||
857e4099 JR |
1897 | static u64 vmx_compute_tsc_offset(struct kvm_vcpu *vcpu, u64 target_tsc) |
1898 | { | |
1899 | return target_tsc - native_read_tsc(); | |
1900 | } | |
1901 | ||
801d3424 NHE |
1902 | static bool guest_cpuid_has_vmx(struct kvm_vcpu *vcpu) |
1903 | { | |
1904 | struct kvm_cpuid_entry2 *best = kvm_find_cpuid_entry(vcpu, 1, 0); | |
1905 | return best && (best->ecx & (1 << (X86_FEATURE_VMX & 31))); | |
1906 | } | |
1907 | ||
1908 | /* | |
1909 | * nested_vmx_allowed() checks whether a guest should be allowed to use VMX | |
1910 | * instructions and MSRs (i.e., nested VMX). Nested VMX is disabled for | |
1911 | * all guests if the "nested" module option is off, and can also be disabled | |
1912 | * for a single guest by disabling its VMX cpuid bit. | |
1913 | */ | |
1914 | static inline bool nested_vmx_allowed(struct kvm_vcpu *vcpu) | |
1915 | { | |
1916 | return nested && guest_cpuid_has_vmx(vcpu); | |
1917 | } | |
1918 | ||
b87a51ae NHE |
1919 | /* |
1920 | * nested_vmx_setup_ctls_msrs() sets up variables containing the values to be | |
1921 | * returned for the various VMX controls MSRs when nested VMX is enabled. | |
1922 | * The same values should also be used to verify that vmcs12 control fields are | |
1923 | * valid during nested entry from L1 to L2. | |
1924 | * Each of these control msrs has a low and high 32-bit half: A low bit is on | |
1925 | * if the corresponding bit in the (32-bit) control field *must* be on, and a | |
1926 | * bit in the high half is on if the corresponding bit in the control field | |
1927 | * may be on. See also vmx_control_verify(). | |
1928 | * TODO: allow these variables to be modified (downgraded) by module options | |
1929 | * or other means. | |
1930 | */ | |
1931 | static u32 nested_vmx_procbased_ctls_low, nested_vmx_procbased_ctls_high; | |
1932 | static u32 nested_vmx_secondary_ctls_low, nested_vmx_secondary_ctls_high; | |
1933 | static u32 nested_vmx_pinbased_ctls_low, nested_vmx_pinbased_ctls_high; | |
1934 | static u32 nested_vmx_exit_ctls_low, nested_vmx_exit_ctls_high; | |
1935 | static u32 nested_vmx_entry_ctls_low, nested_vmx_entry_ctls_high; | |
1936 | static __init void nested_vmx_setup_ctls_msrs(void) | |
1937 | { | |
1938 | /* | |
1939 | * Note that as a general rule, the high half of the MSRs (bits in | |
1940 | * the control fields which may be 1) should be initialized by the | |
1941 | * intersection of the underlying hardware's MSR (i.e., features which | |
1942 | * can be supported) and the list of features we want to expose - | |
1943 | * because they are known to be properly supported in our code. | |
1944 | * Also, usually, the low half of the MSRs (bits which must be 1) can | |
1945 | * be set to 0, meaning that L1 may turn off any of these bits. The | |
1946 | * reason is that if one of these bits is necessary, it will appear | |
1947 | * in vmcs01 and prepare_vmcs02, when it bitwise-or's the control | |
1948 | * fields of vmcs01 and vmcs02, will turn these bits off - and | |
1949 | * nested_vmx_exit_handled() will not pass related exits to L1. | |
1950 | * These rules have exceptions below. | |
1951 | */ | |
1952 | ||
1953 | /* pin-based controls */ | |
1954 | /* | |
1955 | * According to the Intel spec, if bit 55 of VMX_BASIC is off (as it is | |
1956 | * in our case), bits 1, 2 and 4 (i.e., 0x16) must be 1 in this MSR. | |
1957 | */ | |
1958 | nested_vmx_pinbased_ctls_low = 0x16 ; | |
1959 | nested_vmx_pinbased_ctls_high = 0x16 | | |
1960 | PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING | | |
1961 | PIN_BASED_VIRTUAL_NMIS; | |
1962 | ||
1963 | /* exit controls */ | |
1964 | nested_vmx_exit_ctls_low = 0; | |
b6f1250e | 1965 | /* Note that guest use of VM_EXIT_ACK_INTR_ON_EXIT is not supported. */ |
b87a51ae NHE |
1966 | #ifdef CONFIG_X86_64 |
1967 | nested_vmx_exit_ctls_high = VM_EXIT_HOST_ADDR_SPACE_SIZE; | |
1968 | #else | |
1969 | nested_vmx_exit_ctls_high = 0; | |
1970 | #endif | |
1971 | ||
1972 | /* entry controls */ | |
1973 | rdmsr(MSR_IA32_VMX_ENTRY_CTLS, | |
1974 | nested_vmx_entry_ctls_low, nested_vmx_entry_ctls_high); | |
1975 | nested_vmx_entry_ctls_low = 0; | |
1976 | nested_vmx_entry_ctls_high &= | |
1977 | VM_ENTRY_LOAD_IA32_PAT | VM_ENTRY_IA32E_MODE; | |
1978 | ||
1979 | /* cpu-based controls */ | |
1980 | rdmsr(MSR_IA32_VMX_PROCBASED_CTLS, | |
1981 | nested_vmx_procbased_ctls_low, nested_vmx_procbased_ctls_high); | |
1982 | nested_vmx_procbased_ctls_low = 0; | |
1983 | nested_vmx_procbased_ctls_high &= | |
1984 | CPU_BASED_VIRTUAL_INTR_PENDING | CPU_BASED_USE_TSC_OFFSETING | | |
1985 | CPU_BASED_HLT_EXITING | CPU_BASED_INVLPG_EXITING | | |
1986 | CPU_BASED_MWAIT_EXITING | CPU_BASED_CR3_LOAD_EXITING | | |
1987 | CPU_BASED_CR3_STORE_EXITING | | |
1988 | #ifdef CONFIG_X86_64 | |
1989 | CPU_BASED_CR8_LOAD_EXITING | CPU_BASED_CR8_STORE_EXITING | | |
1990 | #endif | |
1991 | CPU_BASED_MOV_DR_EXITING | CPU_BASED_UNCOND_IO_EXITING | | |
1992 | CPU_BASED_USE_IO_BITMAPS | CPU_BASED_MONITOR_EXITING | | |
dbcb4e79 | 1993 | CPU_BASED_RDPMC_EXITING | CPU_BASED_RDTSC_EXITING | |
b87a51ae NHE |
1994 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; |
1995 | /* | |
1996 | * We can allow some features even when not supported by the | |
1997 | * hardware. For example, L1 can specify an MSR bitmap - and we | |
1998 | * can use it to avoid exits to L1 - even when L0 runs L2 | |
1999 | * without MSR bitmaps. | |
2000 | */ | |
2001 | nested_vmx_procbased_ctls_high |= CPU_BASED_USE_MSR_BITMAPS; | |
2002 | ||
2003 | /* secondary cpu-based controls */ | |
2004 | rdmsr(MSR_IA32_VMX_PROCBASED_CTLS2, | |
2005 | nested_vmx_secondary_ctls_low, nested_vmx_secondary_ctls_high); | |
2006 | nested_vmx_secondary_ctls_low = 0; | |
2007 | nested_vmx_secondary_ctls_high &= | |
2008 | SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
2009 | } | |
2010 | ||
2011 | static inline bool vmx_control_verify(u32 control, u32 low, u32 high) | |
2012 | { | |
2013 | /* | |
2014 | * Bits 0 in high must be 0, and bits 1 in low must be 1. | |
2015 | */ | |
2016 | return ((control & high) | low) == control; | |
2017 | } | |
2018 | ||
2019 | static inline u64 vmx_control_msr(u32 low, u32 high) | |
2020 | { | |
2021 | return low | ((u64)high << 32); | |
2022 | } | |
2023 | ||
2024 | /* | |
2025 | * If we allow our guest to use VMX instructions (i.e., nested VMX), we should | |
2026 | * also let it use VMX-specific MSRs. | |
2027 | * vmx_get_vmx_msr() and vmx_set_vmx_msr() return 1 when we handled a | |
2028 | * VMX-specific MSR, or 0 when we haven't (and the caller should handle it | |
2029 | * like all other MSRs). | |
2030 | */ | |
2031 | static int vmx_get_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata) | |
2032 | { | |
2033 | if (!nested_vmx_allowed(vcpu) && msr_index >= MSR_IA32_VMX_BASIC && | |
2034 | msr_index <= MSR_IA32_VMX_TRUE_ENTRY_CTLS) { | |
2035 | /* | |
2036 | * According to the spec, processors which do not support VMX | |
2037 | * should throw a #GP(0) when VMX capability MSRs are read. | |
2038 | */ | |
2039 | kvm_queue_exception_e(vcpu, GP_VECTOR, 0); | |
2040 | return 1; | |
2041 | } | |
2042 | ||
2043 | switch (msr_index) { | |
2044 | case MSR_IA32_FEATURE_CONTROL: | |
2045 | *pdata = 0; | |
2046 | break; | |
2047 | case MSR_IA32_VMX_BASIC: | |
2048 | /* | |
2049 | * This MSR reports some information about VMX support. We | |
2050 | * should return information about the VMX we emulate for the | |
2051 | * guest, and the VMCS structure we give it - not about the | |
2052 | * VMX support of the underlying hardware. | |
2053 | */ | |
2054 | *pdata = VMCS12_REVISION | | |
2055 | ((u64)VMCS12_SIZE << VMX_BASIC_VMCS_SIZE_SHIFT) | | |
2056 | (VMX_BASIC_MEM_TYPE_WB << VMX_BASIC_MEM_TYPE_SHIFT); | |
2057 | break; | |
2058 | case MSR_IA32_VMX_TRUE_PINBASED_CTLS: | |
2059 | case MSR_IA32_VMX_PINBASED_CTLS: | |
2060 | *pdata = vmx_control_msr(nested_vmx_pinbased_ctls_low, | |
2061 | nested_vmx_pinbased_ctls_high); | |
2062 | break; | |
2063 | case MSR_IA32_VMX_TRUE_PROCBASED_CTLS: | |
2064 | case MSR_IA32_VMX_PROCBASED_CTLS: | |
2065 | *pdata = vmx_control_msr(nested_vmx_procbased_ctls_low, | |
2066 | nested_vmx_procbased_ctls_high); | |
2067 | break; | |
2068 | case MSR_IA32_VMX_TRUE_EXIT_CTLS: | |
2069 | case MSR_IA32_VMX_EXIT_CTLS: | |
2070 | *pdata = vmx_control_msr(nested_vmx_exit_ctls_low, | |
2071 | nested_vmx_exit_ctls_high); | |
2072 | break; | |
2073 | case MSR_IA32_VMX_TRUE_ENTRY_CTLS: | |
2074 | case MSR_IA32_VMX_ENTRY_CTLS: | |
2075 | *pdata = vmx_control_msr(nested_vmx_entry_ctls_low, | |
2076 | nested_vmx_entry_ctls_high); | |
2077 | break; | |
2078 | case MSR_IA32_VMX_MISC: | |
2079 | *pdata = 0; | |
2080 | break; | |
2081 | /* | |
2082 | * These MSRs specify bits which the guest must keep fixed (on or off) | |
2083 | * while L1 is in VMXON mode (in L1's root mode, or running an L2). | |
2084 | * We picked the standard core2 setting. | |
2085 | */ | |
2086 | #define VMXON_CR0_ALWAYSON (X86_CR0_PE | X86_CR0_PG | X86_CR0_NE) | |
2087 | #define VMXON_CR4_ALWAYSON X86_CR4_VMXE | |
2088 | case MSR_IA32_VMX_CR0_FIXED0: | |
2089 | *pdata = VMXON_CR0_ALWAYSON; | |
2090 | break; | |
2091 | case MSR_IA32_VMX_CR0_FIXED1: | |
2092 | *pdata = -1ULL; | |
2093 | break; | |
2094 | case MSR_IA32_VMX_CR4_FIXED0: | |
2095 | *pdata = VMXON_CR4_ALWAYSON; | |
2096 | break; | |
2097 | case MSR_IA32_VMX_CR4_FIXED1: | |
2098 | *pdata = -1ULL; | |
2099 | break; | |
2100 | case MSR_IA32_VMX_VMCS_ENUM: | |
2101 | *pdata = 0x1f; | |
2102 | break; | |
2103 | case MSR_IA32_VMX_PROCBASED_CTLS2: | |
2104 | *pdata = vmx_control_msr(nested_vmx_secondary_ctls_low, | |
2105 | nested_vmx_secondary_ctls_high); | |
2106 | break; | |
2107 | case MSR_IA32_VMX_EPT_VPID_CAP: | |
2108 | /* Currently, no nested ept or nested vpid */ | |
2109 | *pdata = 0; | |
2110 | break; | |
2111 | default: | |
2112 | return 0; | |
2113 | } | |
2114 | ||
2115 | return 1; | |
2116 | } | |
2117 | ||
2118 | static int vmx_set_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data) | |
2119 | { | |
2120 | if (!nested_vmx_allowed(vcpu)) | |
2121 | return 0; | |
2122 | ||
2123 | if (msr_index == MSR_IA32_FEATURE_CONTROL) | |
2124 | /* TODO: the right thing. */ | |
2125 | return 1; | |
2126 | /* | |
2127 | * No need to treat VMX capability MSRs specially: If we don't handle | |
2128 | * them, handle_wrmsr will #GP(0), which is correct (they are readonly) | |
2129 | */ | |
2130 | return 0; | |
2131 | } | |
2132 | ||
6aa8b732 AK |
2133 | /* |
2134 | * Reads an msr value (of 'msr_index') into 'pdata'. | |
2135 | * Returns 0 on success, non-0 otherwise. | |
2136 | * Assumes vcpu_load() was already called. | |
2137 | */ | |
2138 | static int vmx_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata) | |
2139 | { | |
2140 | u64 data; | |
26bb0981 | 2141 | struct shared_msr_entry *msr; |
6aa8b732 AK |
2142 | |
2143 | if (!pdata) { | |
2144 | printk(KERN_ERR "BUG: get_msr called with NULL pdata\n"); | |
2145 | return -EINVAL; | |
2146 | } | |
2147 | ||
2148 | switch (msr_index) { | |
05b3e0c2 | 2149 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2150 | case MSR_FS_BASE: |
2151 | data = vmcs_readl(GUEST_FS_BASE); | |
2152 | break; | |
2153 | case MSR_GS_BASE: | |
2154 | data = vmcs_readl(GUEST_GS_BASE); | |
2155 | break; | |
44ea2b17 AK |
2156 | case MSR_KERNEL_GS_BASE: |
2157 | vmx_load_host_state(to_vmx(vcpu)); | |
2158 | data = to_vmx(vcpu)->msr_guest_kernel_gs_base; | |
2159 | break; | |
26bb0981 | 2160 | #endif |
6aa8b732 | 2161 | case MSR_EFER: |
3bab1f5d | 2162 | return kvm_get_msr_common(vcpu, msr_index, pdata); |
af24a4e4 | 2163 | case MSR_IA32_TSC: |
6aa8b732 AK |
2164 | data = guest_read_tsc(); |
2165 | break; | |
2166 | case MSR_IA32_SYSENTER_CS: | |
2167 | data = vmcs_read32(GUEST_SYSENTER_CS); | |
2168 | break; | |
2169 | case MSR_IA32_SYSENTER_EIP: | |
f5b42c33 | 2170 | data = vmcs_readl(GUEST_SYSENTER_EIP); |
6aa8b732 AK |
2171 | break; |
2172 | case MSR_IA32_SYSENTER_ESP: | |
f5b42c33 | 2173 | data = vmcs_readl(GUEST_SYSENTER_ESP); |
6aa8b732 | 2174 | break; |
4e47c7a6 SY |
2175 | case MSR_TSC_AUX: |
2176 | if (!to_vmx(vcpu)->rdtscp_enabled) | |
2177 | return 1; | |
2178 | /* Otherwise falls through */ | |
6aa8b732 | 2179 | default: |
b87a51ae NHE |
2180 | if (vmx_get_vmx_msr(vcpu, msr_index, pdata)) |
2181 | return 0; | |
8b9cf98c | 2182 | msr = find_msr_entry(to_vmx(vcpu), msr_index); |
3bab1f5d AK |
2183 | if (msr) { |
2184 | data = msr->data; | |
2185 | break; | |
6aa8b732 | 2186 | } |
3bab1f5d | 2187 | return kvm_get_msr_common(vcpu, msr_index, pdata); |
6aa8b732 AK |
2188 | } |
2189 | ||
2190 | *pdata = data; | |
2191 | return 0; | |
2192 | } | |
2193 | ||
2194 | /* | |
2195 | * Writes msr value into into the appropriate "register". | |
2196 | * Returns 0 on success, non-0 otherwise. | |
2197 | * Assumes vcpu_load() was already called. | |
2198 | */ | |
2199 | static int vmx_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data) | |
2200 | { | |
a2fa3e9f | 2201 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
26bb0981 | 2202 | struct shared_msr_entry *msr; |
2cc51560 ED |
2203 | int ret = 0; |
2204 | ||
6aa8b732 | 2205 | switch (msr_index) { |
3bab1f5d | 2206 | case MSR_EFER: |
2cc51560 | 2207 | ret = kvm_set_msr_common(vcpu, msr_index, data); |
2cc51560 | 2208 | break; |
16175a79 | 2209 | #ifdef CONFIG_X86_64 |
6aa8b732 | 2210 | case MSR_FS_BASE: |
2fb92db1 | 2211 | vmx_segment_cache_clear(vmx); |
6aa8b732 AK |
2212 | vmcs_writel(GUEST_FS_BASE, data); |
2213 | break; | |
2214 | case MSR_GS_BASE: | |
2fb92db1 | 2215 | vmx_segment_cache_clear(vmx); |
6aa8b732 AK |
2216 | vmcs_writel(GUEST_GS_BASE, data); |
2217 | break; | |
44ea2b17 AK |
2218 | case MSR_KERNEL_GS_BASE: |
2219 | vmx_load_host_state(vmx); | |
2220 | vmx->msr_guest_kernel_gs_base = data; | |
2221 | break; | |
6aa8b732 AK |
2222 | #endif |
2223 | case MSR_IA32_SYSENTER_CS: | |
2224 | vmcs_write32(GUEST_SYSENTER_CS, data); | |
2225 | break; | |
2226 | case MSR_IA32_SYSENTER_EIP: | |
f5b42c33 | 2227 | vmcs_writel(GUEST_SYSENTER_EIP, data); |
6aa8b732 AK |
2228 | break; |
2229 | case MSR_IA32_SYSENTER_ESP: | |
f5b42c33 | 2230 | vmcs_writel(GUEST_SYSENTER_ESP, data); |
6aa8b732 | 2231 | break; |
af24a4e4 | 2232 | case MSR_IA32_TSC: |
99e3e30a | 2233 | kvm_write_tsc(vcpu, data); |
6aa8b732 | 2234 | break; |
468d472f SY |
2235 | case MSR_IA32_CR_PAT: |
2236 | if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) { | |
2237 | vmcs_write64(GUEST_IA32_PAT, data); | |
2238 | vcpu->arch.pat = data; | |
2239 | break; | |
2240 | } | |
4e47c7a6 SY |
2241 | ret = kvm_set_msr_common(vcpu, msr_index, data); |
2242 | break; | |
2243 | case MSR_TSC_AUX: | |
2244 | if (!vmx->rdtscp_enabled) | |
2245 | return 1; | |
2246 | /* Check reserved bit, higher 32 bits should be zero */ | |
2247 | if ((data >> 32) != 0) | |
2248 | return 1; | |
2249 | /* Otherwise falls through */ | |
6aa8b732 | 2250 | default: |
b87a51ae NHE |
2251 | if (vmx_set_vmx_msr(vcpu, msr_index, data)) |
2252 | break; | |
8b9cf98c | 2253 | msr = find_msr_entry(vmx, msr_index); |
3bab1f5d AK |
2254 | if (msr) { |
2255 | msr->data = data; | |
2225fd56 AK |
2256 | if (msr - vmx->guest_msrs < vmx->save_nmsrs) { |
2257 | preempt_disable(); | |
9ee73970 AK |
2258 | kvm_set_shared_msr(msr->index, msr->data, |
2259 | msr->mask); | |
2225fd56 AK |
2260 | preempt_enable(); |
2261 | } | |
3bab1f5d | 2262 | break; |
6aa8b732 | 2263 | } |
2cc51560 | 2264 | ret = kvm_set_msr_common(vcpu, msr_index, data); |
6aa8b732 AK |
2265 | } |
2266 | ||
2cc51560 | 2267 | return ret; |
6aa8b732 AK |
2268 | } |
2269 | ||
5fdbf976 | 2270 | static void vmx_cache_reg(struct kvm_vcpu *vcpu, enum kvm_reg reg) |
6aa8b732 | 2271 | { |
5fdbf976 MT |
2272 | __set_bit(reg, (unsigned long *)&vcpu->arch.regs_avail); |
2273 | switch (reg) { | |
2274 | case VCPU_REGS_RSP: | |
2275 | vcpu->arch.regs[VCPU_REGS_RSP] = vmcs_readl(GUEST_RSP); | |
2276 | break; | |
2277 | case VCPU_REGS_RIP: | |
2278 | vcpu->arch.regs[VCPU_REGS_RIP] = vmcs_readl(GUEST_RIP); | |
2279 | break; | |
6de4f3ad AK |
2280 | case VCPU_EXREG_PDPTR: |
2281 | if (enable_ept) | |
2282 | ept_save_pdptrs(vcpu); | |
2283 | break; | |
5fdbf976 MT |
2284 | default: |
2285 | break; | |
2286 | } | |
6aa8b732 AK |
2287 | } |
2288 | ||
355be0b9 | 2289 | static void set_guest_debug(struct kvm_vcpu *vcpu, struct kvm_guest_debug *dbg) |
6aa8b732 | 2290 | { |
ae675ef0 JK |
2291 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) |
2292 | vmcs_writel(GUEST_DR7, dbg->arch.debugreg[7]); | |
2293 | else | |
2294 | vmcs_writel(GUEST_DR7, vcpu->arch.dr7); | |
2295 | ||
abd3f2d6 | 2296 | update_exception_bitmap(vcpu); |
6aa8b732 AK |
2297 | } |
2298 | ||
2299 | static __init int cpu_has_kvm_support(void) | |
2300 | { | |
6210e37b | 2301 | return cpu_has_vmx(); |
6aa8b732 AK |
2302 | } |
2303 | ||
2304 | static __init int vmx_disabled_by_bios(void) | |
2305 | { | |
2306 | u64 msr; | |
2307 | ||
2308 | rdmsrl(MSR_IA32_FEATURE_CONTROL, msr); | |
cafd6659 | 2309 | if (msr & FEATURE_CONTROL_LOCKED) { |
23f3e991 | 2310 | /* launched w/ TXT and VMX disabled */ |
cafd6659 SW |
2311 | if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) |
2312 | && tboot_enabled()) | |
2313 | return 1; | |
23f3e991 | 2314 | /* launched w/o TXT and VMX only enabled w/ TXT */ |
cafd6659 | 2315 | if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) |
23f3e991 | 2316 | && (msr & FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX) |
f9335afe SW |
2317 | && !tboot_enabled()) { |
2318 | printk(KERN_WARNING "kvm: disable TXT in the BIOS or " | |
23f3e991 | 2319 | "activate TXT before enabling KVM\n"); |
cafd6659 | 2320 | return 1; |
f9335afe | 2321 | } |
23f3e991 JC |
2322 | /* launched w/o TXT and VMX disabled */ |
2323 | if (!(msr & FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX) | |
2324 | && !tboot_enabled()) | |
2325 | return 1; | |
cafd6659 SW |
2326 | } |
2327 | ||
2328 | return 0; | |
6aa8b732 AK |
2329 | } |
2330 | ||
7725b894 DX |
2331 | static void kvm_cpu_vmxon(u64 addr) |
2332 | { | |
2333 | asm volatile (ASM_VMX_VMXON_RAX | |
2334 | : : "a"(&addr), "m"(addr) | |
2335 | : "memory", "cc"); | |
2336 | } | |
2337 | ||
10474ae8 | 2338 | static int hardware_enable(void *garbage) |
6aa8b732 AK |
2339 | { |
2340 | int cpu = raw_smp_processor_id(); | |
2341 | u64 phys_addr = __pa(per_cpu(vmxarea, cpu)); | |
cafd6659 | 2342 | u64 old, test_bits; |
6aa8b732 | 2343 | |
10474ae8 AG |
2344 | if (read_cr4() & X86_CR4_VMXE) |
2345 | return -EBUSY; | |
2346 | ||
d462b819 | 2347 | INIT_LIST_HEAD(&per_cpu(loaded_vmcss_on_cpu, cpu)); |
6aa8b732 | 2348 | rdmsrl(MSR_IA32_FEATURE_CONTROL, old); |
cafd6659 SW |
2349 | |
2350 | test_bits = FEATURE_CONTROL_LOCKED; | |
2351 | test_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; | |
2352 | if (tboot_enabled()) | |
2353 | test_bits |= FEATURE_CONTROL_VMXON_ENABLED_INSIDE_SMX; | |
2354 | ||
2355 | if ((old & test_bits) != test_bits) { | |
6aa8b732 | 2356 | /* enable and lock */ |
cafd6659 SW |
2357 | wrmsrl(MSR_IA32_FEATURE_CONTROL, old | test_bits); |
2358 | } | |
66aee91a | 2359 | write_cr4(read_cr4() | X86_CR4_VMXE); /* FIXME: not cpu hotplug safe */ |
10474ae8 | 2360 | |
4610c9cc DX |
2361 | if (vmm_exclusive) { |
2362 | kvm_cpu_vmxon(phys_addr); | |
2363 | ept_sync_global(); | |
2364 | } | |
10474ae8 | 2365 | |
3444d7da AK |
2366 | store_gdt(&__get_cpu_var(host_gdt)); |
2367 | ||
10474ae8 | 2368 | return 0; |
6aa8b732 AK |
2369 | } |
2370 | ||
d462b819 | 2371 | static void vmclear_local_loaded_vmcss(void) |
543e4243 AK |
2372 | { |
2373 | int cpu = raw_smp_processor_id(); | |
d462b819 | 2374 | struct loaded_vmcs *v, *n; |
543e4243 | 2375 | |
d462b819 NHE |
2376 | list_for_each_entry_safe(v, n, &per_cpu(loaded_vmcss_on_cpu, cpu), |
2377 | loaded_vmcss_on_cpu_link) | |
2378 | __loaded_vmcs_clear(v); | |
543e4243 AK |
2379 | } |
2380 | ||
710ff4a8 EH |
2381 | |
2382 | /* Just like cpu_vmxoff(), but with the __kvm_handle_fault_on_reboot() | |
2383 | * tricks. | |
2384 | */ | |
2385 | static void kvm_cpu_vmxoff(void) | |
6aa8b732 | 2386 | { |
4ecac3fd | 2387 | asm volatile (__ex(ASM_VMX_VMXOFF) : : : "cc"); |
6aa8b732 AK |
2388 | } |
2389 | ||
710ff4a8 EH |
2390 | static void hardware_disable(void *garbage) |
2391 | { | |
4610c9cc | 2392 | if (vmm_exclusive) { |
d462b819 | 2393 | vmclear_local_loaded_vmcss(); |
4610c9cc DX |
2394 | kvm_cpu_vmxoff(); |
2395 | } | |
7725b894 | 2396 | write_cr4(read_cr4() & ~X86_CR4_VMXE); |
710ff4a8 EH |
2397 | } |
2398 | ||
1c3d14fe | 2399 | static __init int adjust_vmx_controls(u32 ctl_min, u32 ctl_opt, |
d77c26fc | 2400 | u32 msr, u32 *result) |
1c3d14fe YS |
2401 | { |
2402 | u32 vmx_msr_low, vmx_msr_high; | |
2403 | u32 ctl = ctl_min | ctl_opt; | |
2404 | ||
2405 | rdmsr(msr, vmx_msr_low, vmx_msr_high); | |
2406 | ||
2407 | ctl &= vmx_msr_high; /* bit == 0 in high word ==> must be zero */ | |
2408 | ctl |= vmx_msr_low; /* bit == 1 in low word ==> must be one */ | |
2409 | ||
2410 | /* Ensure minimum (required) set of control bits are supported. */ | |
2411 | if (ctl_min & ~ctl) | |
002c7f7c | 2412 | return -EIO; |
1c3d14fe YS |
2413 | |
2414 | *result = ctl; | |
2415 | return 0; | |
2416 | } | |
2417 | ||
110312c8 AK |
2418 | static __init bool allow_1_setting(u32 msr, u32 ctl) |
2419 | { | |
2420 | u32 vmx_msr_low, vmx_msr_high; | |
2421 | ||
2422 | rdmsr(msr, vmx_msr_low, vmx_msr_high); | |
2423 | return vmx_msr_high & ctl; | |
2424 | } | |
2425 | ||
002c7f7c | 2426 | static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf) |
6aa8b732 AK |
2427 | { |
2428 | u32 vmx_msr_low, vmx_msr_high; | |
d56f546d | 2429 | u32 min, opt, min2, opt2; |
1c3d14fe YS |
2430 | u32 _pin_based_exec_control = 0; |
2431 | u32 _cpu_based_exec_control = 0; | |
f78e0e2e | 2432 | u32 _cpu_based_2nd_exec_control = 0; |
1c3d14fe YS |
2433 | u32 _vmexit_control = 0; |
2434 | u32 _vmentry_control = 0; | |
2435 | ||
2436 | min = PIN_BASED_EXT_INTR_MASK | PIN_BASED_NMI_EXITING; | |
f08864b4 | 2437 | opt = PIN_BASED_VIRTUAL_NMIS; |
1c3d14fe YS |
2438 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PINBASED_CTLS, |
2439 | &_pin_based_exec_control) < 0) | |
002c7f7c | 2440 | return -EIO; |
1c3d14fe | 2441 | |
10166744 | 2442 | min = CPU_BASED_HLT_EXITING | |
1c3d14fe YS |
2443 | #ifdef CONFIG_X86_64 |
2444 | CPU_BASED_CR8_LOAD_EXITING | | |
2445 | CPU_BASED_CR8_STORE_EXITING | | |
2446 | #endif | |
d56f546d SY |
2447 | CPU_BASED_CR3_LOAD_EXITING | |
2448 | CPU_BASED_CR3_STORE_EXITING | | |
1c3d14fe YS |
2449 | CPU_BASED_USE_IO_BITMAPS | |
2450 | CPU_BASED_MOV_DR_EXITING | | |
a7052897 | 2451 | CPU_BASED_USE_TSC_OFFSETING | |
59708670 SY |
2452 | CPU_BASED_MWAIT_EXITING | |
2453 | CPU_BASED_MONITOR_EXITING | | |
fee84b07 AK |
2454 | CPU_BASED_INVLPG_EXITING | |
2455 | CPU_BASED_RDPMC_EXITING; | |
443381a8 | 2456 | |
f78e0e2e | 2457 | opt = CPU_BASED_TPR_SHADOW | |
25c5f225 | 2458 | CPU_BASED_USE_MSR_BITMAPS | |
f78e0e2e | 2459 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS; |
1c3d14fe YS |
2460 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_PROCBASED_CTLS, |
2461 | &_cpu_based_exec_control) < 0) | |
002c7f7c | 2462 | return -EIO; |
6e5d865c YS |
2463 | #ifdef CONFIG_X86_64 |
2464 | if ((_cpu_based_exec_control & CPU_BASED_TPR_SHADOW)) | |
2465 | _cpu_based_exec_control &= ~CPU_BASED_CR8_LOAD_EXITING & | |
2466 | ~CPU_BASED_CR8_STORE_EXITING; | |
2467 | #endif | |
f78e0e2e | 2468 | if (_cpu_based_exec_control & CPU_BASED_ACTIVATE_SECONDARY_CONTROLS) { |
d56f546d SY |
2469 | min2 = 0; |
2470 | opt2 = SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES | | |
2384d2b3 | 2471 | SECONDARY_EXEC_WBINVD_EXITING | |
d56f546d | 2472 | SECONDARY_EXEC_ENABLE_VPID | |
3a624e29 | 2473 | SECONDARY_EXEC_ENABLE_EPT | |
4b8d54f9 | 2474 | SECONDARY_EXEC_UNRESTRICTED_GUEST | |
4e47c7a6 | 2475 | SECONDARY_EXEC_PAUSE_LOOP_EXITING | |
ad756a16 MJ |
2476 | SECONDARY_EXEC_RDTSCP | |
2477 | SECONDARY_EXEC_ENABLE_INVPCID; | |
d56f546d SY |
2478 | if (adjust_vmx_controls(min2, opt2, |
2479 | MSR_IA32_VMX_PROCBASED_CTLS2, | |
f78e0e2e SY |
2480 | &_cpu_based_2nd_exec_control) < 0) |
2481 | return -EIO; | |
2482 | } | |
2483 | #ifndef CONFIG_X86_64 | |
2484 | if (!(_cpu_based_2nd_exec_control & | |
2485 | SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES)) | |
2486 | _cpu_based_exec_control &= ~CPU_BASED_TPR_SHADOW; | |
2487 | #endif | |
d56f546d | 2488 | if (_cpu_based_2nd_exec_control & SECONDARY_EXEC_ENABLE_EPT) { |
a7052897 MT |
2489 | /* CR3 accesses and invlpg don't need to cause VM Exits when EPT |
2490 | enabled */ | |
5fff7d27 GN |
2491 | _cpu_based_exec_control &= ~(CPU_BASED_CR3_LOAD_EXITING | |
2492 | CPU_BASED_CR3_STORE_EXITING | | |
2493 | CPU_BASED_INVLPG_EXITING); | |
d56f546d SY |
2494 | rdmsr(MSR_IA32_VMX_EPT_VPID_CAP, |
2495 | vmx_capability.ept, vmx_capability.vpid); | |
2496 | } | |
1c3d14fe YS |
2497 | |
2498 | min = 0; | |
2499 | #ifdef CONFIG_X86_64 | |
2500 | min |= VM_EXIT_HOST_ADDR_SPACE_SIZE; | |
2501 | #endif | |
468d472f | 2502 | opt = VM_EXIT_SAVE_IA32_PAT | VM_EXIT_LOAD_IA32_PAT; |
1c3d14fe YS |
2503 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_EXIT_CTLS, |
2504 | &_vmexit_control) < 0) | |
002c7f7c | 2505 | return -EIO; |
1c3d14fe | 2506 | |
468d472f SY |
2507 | min = 0; |
2508 | opt = VM_ENTRY_LOAD_IA32_PAT; | |
1c3d14fe YS |
2509 | if (adjust_vmx_controls(min, opt, MSR_IA32_VMX_ENTRY_CTLS, |
2510 | &_vmentry_control) < 0) | |
002c7f7c | 2511 | return -EIO; |
6aa8b732 | 2512 | |
c68876fd | 2513 | rdmsr(MSR_IA32_VMX_BASIC, vmx_msr_low, vmx_msr_high); |
1c3d14fe YS |
2514 | |
2515 | /* IA-32 SDM Vol 3B: VMCS size is never greater than 4kB. */ | |
2516 | if ((vmx_msr_high & 0x1fff) > PAGE_SIZE) | |
002c7f7c | 2517 | return -EIO; |
1c3d14fe YS |
2518 | |
2519 | #ifdef CONFIG_X86_64 | |
2520 | /* IA-32 SDM Vol 3B: 64-bit CPUs always have VMX_BASIC_MSR[48]==0. */ | |
2521 | if (vmx_msr_high & (1u<<16)) | |
002c7f7c | 2522 | return -EIO; |
1c3d14fe YS |
2523 | #endif |
2524 | ||
2525 | /* Require Write-Back (WB) memory type for VMCS accesses. */ | |
2526 | if (((vmx_msr_high >> 18) & 15) != 6) | |
002c7f7c | 2527 | return -EIO; |
1c3d14fe | 2528 | |
002c7f7c YS |
2529 | vmcs_conf->size = vmx_msr_high & 0x1fff; |
2530 | vmcs_conf->order = get_order(vmcs_config.size); | |
2531 | vmcs_conf->revision_id = vmx_msr_low; | |
1c3d14fe | 2532 | |
002c7f7c YS |
2533 | vmcs_conf->pin_based_exec_ctrl = _pin_based_exec_control; |
2534 | vmcs_conf->cpu_based_exec_ctrl = _cpu_based_exec_control; | |
f78e0e2e | 2535 | vmcs_conf->cpu_based_2nd_exec_ctrl = _cpu_based_2nd_exec_control; |
002c7f7c YS |
2536 | vmcs_conf->vmexit_ctrl = _vmexit_control; |
2537 | vmcs_conf->vmentry_ctrl = _vmentry_control; | |
1c3d14fe | 2538 | |
110312c8 AK |
2539 | cpu_has_load_ia32_efer = |
2540 | allow_1_setting(MSR_IA32_VMX_ENTRY_CTLS, | |
2541 | VM_ENTRY_LOAD_IA32_EFER) | |
2542 | && allow_1_setting(MSR_IA32_VMX_EXIT_CTLS, | |
2543 | VM_EXIT_LOAD_IA32_EFER); | |
2544 | ||
8bf00a52 GN |
2545 | cpu_has_load_perf_global_ctrl = |
2546 | allow_1_setting(MSR_IA32_VMX_ENTRY_CTLS, | |
2547 | VM_ENTRY_LOAD_IA32_PERF_GLOBAL_CTRL) | |
2548 | && allow_1_setting(MSR_IA32_VMX_EXIT_CTLS, | |
2549 | VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL); | |
2550 | ||
2551 | /* | |
2552 | * Some cpus support VM_ENTRY_(LOAD|SAVE)_IA32_PERF_GLOBAL_CTRL | |
2553 | * but due to arrata below it can't be used. Workaround is to use | |
2554 | * msr load mechanism to switch IA32_PERF_GLOBAL_CTRL. | |
2555 | * | |
2556 | * VM Exit May Incorrectly Clear IA32_PERF_GLOBAL_CTRL [34:32] | |
2557 | * | |
2558 | * AAK155 (model 26) | |
2559 | * AAP115 (model 30) | |
2560 | * AAT100 (model 37) | |
2561 | * BC86,AAY89,BD102 (model 44) | |
2562 | * BA97 (model 46) | |
2563 | * | |
2564 | */ | |
2565 | if (cpu_has_load_perf_global_ctrl && boot_cpu_data.x86 == 0x6) { | |
2566 | switch (boot_cpu_data.x86_model) { | |
2567 | case 26: | |
2568 | case 30: | |
2569 | case 37: | |
2570 | case 44: | |
2571 | case 46: | |
2572 | cpu_has_load_perf_global_ctrl = false; | |
2573 | printk_once(KERN_WARNING"kvm: VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL " | |
2574 | "does not work properly. Using workaround\n"); | |
2575 | break; | |
2576 | default: | |
2577 | break; | |
2578 | } | |
2579 | } | |
2580 | ||
1c3d14fe | 2581 | return 0; |
c68876fd | 2582 | } |
6aa8b732 AK |
2583 | |
2584 | static struct vmcs *alloc_vmcs_cpu(int cpu) | |
2585 | { | |
2586 | int node = cpu_to_node(cpu); | |
2587 | struct page *pages; | |
2588 | struct vmcs *vmcs; | |
2589 | ||
6484eb3e | 2590 | pages = alloc_pages_exact_node(node, GFP_KERNEL, vmcs_config.order); |
6aa8b732 AK |
2591 | if (!pages) |
2592 | return NULL; | |
2593 | vmcs = page_address(pages); | |
1c3d14fe YS |
2594 | memset(vmcs, 0, vmcs_config.size); |
2595 | vmcs->revision_id = vmcs_config.revision_id; /* vmcs revision id */ | |
6aa8b732 AK |
2596 | return vmcs; |
2597 | } | |
2598 | ||
2599 | static struct vmcs *alloc_vmcs(void) | |
2600 | { | |
d3b2c338 | 2601 | return alloc_vmcs_cpu(raw_smp_processor_id()); |
6aa8b732 AK |
2602 | } |
2603 | ||
2604 | static void free_vmcs(struct vmcs *vmcs) | |
2605 | { | |
1c3d14fe | 2606 | free_pages((unsigned long)vmcs, vmcs_config.order); |
6aa8b732 AK |
2607 | } |
2608 | ||
d462b819 NHE |
2609 | /* |
2610 | * Free a VMCS, but before that VMCLEAR it on the CPU where it was last loaded | |
2611 | */ | |
2612 | static void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs) | |
2613 | { | |
2614 | if (!loaded_vmcs->vmcs) | |
2615 | return; | |
2616 | loaded_vmcs_clear(loaded_vmcs); | |
2617 | free_vmcs(loaded_vmcs->vmcs); | |
2618 | loaded_vmcs->vmcs = NULL; | |
2619 | } | |
2620 | ||
39959588 | 2621 | static void free_kvm_area(void) |
6aa8b732 AK |
2622 | { |
2623 | int cpu; | |
2624 | ||
3230bb47 | 2625 | for_each_possible_cpu(cpu) { |
6aa8b732 | 2626 | free_vmcs(per_cpu(vmxarea, cpu)); |
3230bb47 ZA |
2627 | per_cpu(vmxarea, cpu) = NULL; |
2628 | } | |
6aa8b732 AK |
2629 | } |
2630 | ||
6aa8b732 AK |
2631 | static __init int alloc_kvm_area(void) |
2632 | { | |
2633 | int cpu; | |
2634 | ||
3230bb47 | 2635 | for_each_possible_cpu(cpu) { |
6aa8b732 AK |
2636 | struct vmcs *vmcs; |
2637 | ||
2638 | vmcs = alloc_vmcs_cpu(cpu); | |
2639 | if (!vmcs) { | |
2640 | free_kvm_area(); | |
2641 | return -ENOMEM; | |
2642 | } | |
2643 | ||
2644 | per_cpu(vmxarea, cpu) = vmcs; | |
2645 | } | |
2646 | return 0; | |
2647 | } | |
2648 | ||
2649 | static __init int hardware_setup(void) | |
2650 | { | |
002c7f7c YS |
2651 | if (setup_vmcs_config(&vmcs_config) < 0) |
2652 | return -EIO; | |
50a37eb4 JR |
2653 | |
2654 | if (boot_cpu_has(X86_FEATURE_NX)) | |
2655 | kvm_enable_efer_bits(EFER_NX); | |
2656 | ||
93ba03c2 SY |
2657 | if (!cpu_has_vmx_vpid()) |
2658 | enable_vpid = 0; | |
2659 | ||
4bc9b982 SY |
2660 | if (!cpu_has_vmx_ept() || |
2661 | !cpu_has_vmx_ept_4levels()) { | |
93ba03c2 | 2662 | enable_ept = 0; |
3a624e29 | 2663 | enable_unrestricted_guest = 0; |
83c3a331 | 2664 | enable_ept_ad_bits = 0; |
3a624e29 NK |
2665 | } |
2666 | ||
83c3a331 XH |
2667 | if (!cpu_has_vmx_ept_ad_bits()) |
2668 | enable_ept_ad_bits = 0; | |
2669 | ||
3a624e29 NK |
2670 | if (!cpu_has_vmx_unrestricted_guest()) |
2671 | enable_unrestricted_guest = 0; | |
93ba03c2 SY |
2672 | |
2673 | if (!cpu_has_vmx_flexpriority()) | |
2674 | flexpriority_enabled = 0; | |
2675 | ||
95ba8273 GN |
2676 | if (!cpu_has_vmx_tpr_shadow()) |
2677 | kvm_x86_ops->update_cr8_intercept = NULL; | |
2678 | ||
54dee993 MT |
2679 | if (enable_ept && !cpu_has_vmx_ept_2m_page()) |
2680 | kvm_disable_largepages(); | |
2681 | ||
4b8d54f9 ZE |
2682 | if (!cpu_has_vmx_ple()) |
2683 | ple_gap = 0; | |
2684 | ||
b87a51ae NHE |
2685 | if (nested) |
2686 | nested_vmx_setup_ctls_msrs(); | |
2687 | ||
6aa8b732 AK |
2688 | return alloc_kvm_area(); |
2689 | } | |
2690 | ||
2691 | static __exit void hardware_unsetup(void) | |
2692 | { | |
2693 | free_kvm_area(); | |
2694 | } | |
2695 | ||
f5f7b2fe | 2696 | static void fix_pmode_dataseg(struct kvm_vcpu *vcpu, int seg, struct kvm_segment *save) |
6aa8b732 AK |
2697 | { |
2698 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
c865c43d | 2699 | struct kvm_segment tmp = *save; |
6aa8b732 | 2700 | |
c865c43d AK |
2701 | if (!(vmcs_readl(sf->base) == tmp.base && tmp.s)) { |
2702 | tmp.base = vmcs_readl(sf->base); | |
2703 | tmp.selector = vmcs_read16(sf->selector); | |
2704 | tmp.s = 1; | |
6aa8b732 | 2705 | } |
c865c43d | 2706 | vmx_set_segment(vcpu, &tmp, seg); |
6aa8b732 AK |
2707 | } |
2708 | ||
2709 | static void enter_pmode(struct kvm_vcpu *vcpu) | |
2710 | { | |
2711 | unsigned long flags; | |
a89a8fb9 | 2712 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
6aa8b732 | 2713 | |
a89a8fb9 | 2714 | vmx->emulation_required = 1; |
7ffd92c5 | 2715 | vmx->rmode.vm86_active = 0; |
6aa8b732 | 2716 | |
2fb92db1 AK |
2717 | vmx_segment_cache_clear(vmx); |
2718 | ||
f5f7b2fe | 2719 | vmx_set_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR); |
6aa8b732 AK |
2720 | |
2721 | flags = vmcs_readl(GUEST_RFLAGS); | |
78ac8b47 AK |
2722 | flags &= RMODE_GUEST_OWNED_EFLAGS_BITS; |
2723 | flags |= vmx->rmode.save_rflags & ~RMODE_GUEST_OWNED_EFLAGS_BITS; | |
6aa8b732 AK |
2724 | vmcs_writel(GUEST_RFLAGS, flags); |
2725 | ||
66aee91a RR |
2726 | vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~X86_CR4_VME) | |
2727 | (vmcs_readl(CR4_READ_SHADOW) & X86_CR4_VME)); | |
6aa8b732 AK |
2728 | |
2729 | update_exception_bitmap(vcpu); | |
2730 | ||
a89a8fb9 MG |
2731 | if (emulate_invalid_guest_state) |
2732 | return; | |
2733 | ||
f5f7b2fe AK |
2734 | fix_pmode_dataseg(vcpu, VCPU_SREG_ES, &vmx->rmode.segs[VCPU_SREG_ES]); |
2735 | fix_pmode_dataseg(vcpu, VCPU_SREG_DS, &vmx->rmode.segs[VCPU_SREG_DS]); | |
2736 | fix_pmode_dataseg(vcpu, VCPU_SREG_FS, &vmx->rmode.segs[VCPU_SREG_FS]); | |
2737 | fix_pmode_dataseg(vcpu, VCPU_SREG_GS, &vmx->rmode.segs[VCPU_SREG_GS]); | |
6aa8b732 | 2738 | |
2fb92db1 AK |
2739 | vmx_segment_cache_clear(vmx); |
2740 | ||
6aa8b732 AK |
2741 | vmcs_write16(GUEST_SS_SELECTOR, 0); |
2742 | vmcs_write32(GUEST_SS_AR_BYTES, 0x93); | |
2743 | ||
2744 | vmcs_write16(GUEST_CS_SELECTOR, | |
2745 | vmcs_read16(GUEST_CS_SELECTOR) & ~SELECTOR_RPL_MASK); | |
2746 | vmcs_write32(GUEST_CS_AR_BYTES, 0x9b); | |
2747 | } | |
2748 | ||
d77c26fc | 2749 | static gva_t rmode_tss_base(struct kvm *kvm) |
6aa8b732 | 2750 | { |
bfc6d222 | 2751 | if (!kvm->arch.tss_addr) { |
bc6678a3 | 2752 | struct kvm_memslots *slots; |
28a37544 | 2753 | struct kvm_memory_slot *slot; |
bc6678a3 MT |
2754 | gfn_t base_gfn; |
2755 | ||
90d83dc3 | 2756 | slots = kvm_memslots(kvm); |
28a37544 XG |
2757 | slot = id_to_memslot(slots, 0); |
2758 | base_gfn = slot->base_gfn + slot->npages - 3; | |
2759 | ||
cbc94022 IE |
2760 | return base_gfn << PAGE_SHIFT; |
2761 | } | |
bfc6d222 | 2762 | return kvm->arch.tss_addr; |
6aa8b732 AK |
2763 | } |
2764 | ||
f5f7b2fe | 2765 | static void fix_rmode_seg(int seg, struct kvm_segment *save) |
baa7e81e AK |
2766 | { |
2767 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
2768 | ||
15b00f32 | 2769 | vmcs_write16(sf->selector, save->base >> 4); |
444e863d | 2770 | vmcs_write32(sf->base, save->base & 0xffff0); |
6aa8b732 AK |
2771 | vmcs_write32(sf->limit, 0xffff); |
2772 | vmcs_write32(sf->ar_bytes, 0xf3); | |
444e863d GN |
2773 | if (save->base & 0xf) |
2774 | printk_once(KERN_WARNING "kvm: segment base is not paragraph" | |
2775 | " aligned when entering protected mode (seg=%d)", | |
2776 | seg); | |
6aa8b732 AK |
2777 | } |
2778 | ||
2779 | static void enter_rmode(struct kvm_vcpu *vcpu) | |
2780 | { | |
2781 | unsigned long flags; | |
a89a8fb9 | 2782 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
b246dd5d | 2783 | struct kvm_segment var; |
6aa8b732 | 2784 | |
3a624e29 NK |
2785 | if (enable_unrestricted_guest) |
2786 | return; | |
2787 | ||
f5f7b2fe AK |
2788 | vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_TR], VCPU_SREG_TR); |
2789 | vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_ES], VCPU_SREG_ES); | |
2790 | vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_DS], VCPU_SREG_DS); | |
2791 | vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_FS], VCPU_SREG_FS); | |
2792 | vmx_get_segment(vcpu, &vmx->rmode.segs[VCPU_SREG_GS], VCPU_SREG_GS); | |
2793 | ||
a89a8fb9 | 2794 | vmx->emulation_required = 1; |
7ffd92c5 | 2795 | vmx->rmode.vm86_active = 1; |
6aa8b732 | 2796 | |
baa7e81e | 2797 | |
776e58ea GN |
2798 | /* |
2799 | * Very old userspace does not call KVM_SET_TSS_ADDR before entering | |
2800 | * vcpu. Call it here with phys address pointing 16M below 4G. | |
2801 | */ | |
2802 | if (!vcpu->kvm->arch.tss_addr) { | |
2803 | printk_once(KERN_WARNING "kvm: KVM_SET_TSS_ADDR need to be " | |
2804 | "called before entering vcpu\n"); | |
2805 | srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); | |
2806 | vmx_set_tss_addr(vcpu->kvm, 0xfeffd000); | |
2807 | vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); | |
2808 | } | |
2809 | ||
2fb92db1 AK |
2810 | vmx_segment_cache_clear(vmx); |
2811 | ||
6aa8b732 | 2812 | vmcs_writel(GUEST_TR_BASE, rmode_tss_base(vcpu->kvm)); |
6aa8b732 | 2813 | vmcs_write32(GUEST_TR_LIMIT, RMODE_TSS_SIZE - 1); |
6aa8b732 AK |
2814 | vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); |
2815 | ||
2816 | flags = vmcs_readl(GUEST_RFLAGS); | |
78ac8b47 | 2817 | vmx->rmode.save_rflags = flags; |
6aa8b732 | 2818 | |
053de044 | 2819 | flags |= X86_EFLAGS_IOPL | X86_EFLAGS_VM; |
6aa8b732 AK |
2820 | |
2821 | vmcs_writel(GUEST_RFLAGS, flags); | |
66aee91a | 2822 | vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | X86_CR4_VME); |
6aa8b732 AK |
2823 | update_exception_bitmap(vcpu); |
2824 | ||
a89a8fb9 MG |
2825 | if (emulate_invalid_guest_state) |
2826 | goto continue_rmode; | |
2827 | ||
b246dd5d OW |
2828 | vmx_get_segment(vcpu, &var, VCPU_SREG_SS); |
2829 | vmx_set_segment(vcpu, &var, VCPU_SREG_SS); | |
2830 | ||
2831 | vmx_get_segment(vcpu, &var, VCPU_SREG_CS); | |
2832 | vmx_set_segment(vcpu, &var, VCPU_SREG_CS); | |
2833 | ||
2834 | vmx_get_segment(vcpu, &var, VCPU_SREG_ES); | |
2835 | vmx_set_segment(vcpu, &var, VCPU_SREG_ES); | |
2836 | ||
2837 | vmx_get_segment(vcpu, &var, VCPU_SREG_DS); | |
2838 | vmx_set_segment(vcpu, &var, VCPU_SREG_DS); | |
6aa8b732 | 2839 | |
b246dd5d OW |
2840 | vmx_get_segment(vcpu, &var, VCPU_SREG_GS); |
2841 | vmx_set_segment(vcpu, &var, VCPU_SREG_GS); | |
6aa8b732 | 2842 | |
b246dd5d OW |
2843 | vmx_get_segment(vcpu, &var, VCPU_SREG_FS); |
2844 | vmx_set_segment(vcpu, &var, VCPU_SREG_FS); | |
75880a01 | 2845 | |
a89a8fb9 | 2846 | continue_rmode: |
8668a3c4 | 2847 | kvm_mmu_reset_context(vcpu); |
6aa8b732 AK |
2848 | } |
2849 | ||
401d10de AS |
2850 | static void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer) |
2851 | { | |
2852 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
26bb0981 AK |
2853 | struct shared_msr_entry *msr = find_msr_entry(vmx, MSR_EFER); |
2854 | ||
2855 | if (!msr) | |
2856 | return; | |
401d10de | 2857 | |
44ea2b17 AK |
2858 | /* |
2859 | * Force kernel_gs_base reloading before EFER changes, as control | |
2860 | * of this msr depends on is_long_mode(). | |
2861 | */ | |
2862 | vmx_load_host_state(to_vmx(vcpu)); | |
f6801dff | 2863 | vcpu->arch.efer = efer; |
401d10de AS |
2864 | if (efer & EFER_LMA) { |
2865 | vmcs_write32(VM_ENTRY_CONTROLS, | |
2866 | vmcs_read32(VM_ENTRY_CONTROLS) | | |
2867 | VM_ENTRY_IA32E_MODE); | |
2868 | msr->data = efer; | |
2869 | } else { | |
2870 | vmcs_write32(VM_ENTRY_CONTROLS, | |
2871 | vmcs_read32(VM_ENTRY_CONTROLS) & | |
2872 | ~VM_ENTRY_IA32E_MODE); | |
2873 | ||
2874 | msr->data = efer & ~EFER_LME; | |
2875 | } | |
2876 | setup_msrs(vmx); | |
2877 | } | |
2878 | ||
05b3e0c2 | 2879 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2880 | |
2881 | static void enter_lmode(struct kvm_vcpu *vcpu) | |
2882 | { | |
2883 | u32 guest_tr_ar; | |
2884 | ||
2fb92db1 AK |
2885 | vmx_segment_cache_clear(to_vmx(vcpu)); |
2886 | ||
6aa8b732 AK |
2887 | guest_tr_ar = vmcs_read32(GUEST_TR_AR_BYTES); |
2888 | if ((guest_tr_ar & AR_TYPE_MASK) != AR_TYPE_BUSY_64_TSS) { | |
bd80158a JK |
2889 | pr_debug_ratelimited("%s: tss fixup for long mode. \n", |
2890 | __func__); | |
6aa8b732 AK |
2891 | vmcs_write32(GUEST_TR_AR_BYTES, |
2892 | (guest_tr_ar & ~AR_TYPE_MASK) | |
2893 | | AR_TYPE_BUSY_64_TSS); | |
2894 | } | |
da38f438 | 2895 | vmx_set_efer(vcpu, vcpu->arch.efer | EFER_LMA); |
6aa8b732 AK |
2896 | } |
2897 | ||
2898 | static void exit_lmode(struct kvm_vcpu *vcpu) | |
2899 | { | |
6aa8b732 AK |
2900 | vmcs_write32(VM_ENTRY_CONTROLS, |
2901 | vmcs_read32(VM_ENTRY_CONTROLS) | |
1e4e6e00 | 2902 | & ~VM_ENTRY_IA32E_MODE); |
da38f438 | 2903 | vmx_set_efer(vcpu, vcpu->arch.efer & ~EFER_LMA); |
6aa8b732 AK |
2904 | } |
2905 | ||
2906 | #endif | |
2907 | ||
2384d2b3 SY |
2908 | static void vmx_flush_tlb(struct kvm_vcpu *vcpu) |
2909 | { | |
b9d762fa | 2910 | vpid_sync_context(to_vmx(vcpu)); |
dd180b3e XG |
2911 | if (enable_ept) { |
2912 | if (!VALID_PAGE(vcpu->arch.mmu.root_hpa)) | |
2913 | return; | |
4e1096d2 | 2914 | ept_sync_context(construct_eptp(vcpu->arch.mmu.root_hpa)); |
dd180b3e | 2915 | } |
2384d2b3 SY |
2916 | } |
2917 | ||
e8467fda AK |
2918 | static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu) |
2919 | { | |
2920 | ulong cr0_guest_owned_bits = vcpu->arch.cr0_guest_owned_bits; | |
2921 | ||
2922 | vcpu->arch.cr0 &= ~cr0_guest_owned_bits; | |
2923 | vcpu->arch.cr0 |= vmcs_readl(GUEST_CR0) & cr0_guest_owned_bits; | |
2924 | } | |
2925 | ||
aff48baa AK |
2926 | static void vmx_decache_cr3(struct kvm_vcpu *vcpu) |
2927 | { | |
2928 | if (enable_ept && is_paging(vcpu)) | |
2929 | vcpu->arch.cr3 = vmcs_readl(GUEST_CR3); | |
2930 | __set_bit(VCPU_EXREG_CR3, (ulong *)&vcpu->arch.regs_avail); | |
2931 | } | |
2932 | ||
25c4c276 | 2933 | static void vmx_decache_cr4_guest_bits(struct kvm_vcpu *vcpu) |
399badf3 | 2934 | { |
fc78f519 AK |
2935 | ulong cr4_guest_owned_bits = vcpu->arch.cr4_guest_owned_bits; |
2936 | ||
2937 | vcpu->arch.cr4 &= ~cr4_guest_owned_bits; | |
2938 | vcpu->arch.cr4 |= vmcs_readl(GUEST_CR4) & cr4_guest_owned_bits; | |
399badf3 AK |
2939 | } |
2940 | ||
1439442c SY |
2941 | static void ept_load_pdptrs(struct kvm_vcpu *vcpu) |
2942 | { | |
6de4f3ad AK |
2943 | if (!test_bit(VCPU_EXREG_PDPTR, |
2944 | (unsigned long *)&vcpu->arch.regs_dirty)) | |
2945 | return; | |
2946 | ||
1439442c | 2947 | if (is_paging(vcpu) && is_pae(vcpu) && !is_long_mode(vcpu)) { |
ff03a073 JR |
2948 | vmcs_write64(GUEST_PDPTR0, vcpu->arch.mmu.pdptrs[0]); |
2949 | vmcs_write64(GUEST_PDPTR1, vcpu->arch.mmu.pdptrs[1]); | |
2950 | vmcs_write64(GUEST_PDPTR2, vcpu->arch.mmu.pdptrs[2]); | |
2951 | vmcs_write64(GUEST_PDPTR3, vcpu->arch.mmu.pdptrs[3]); | |
1439442c SY |
2952 | } |
2953 | } | |
2954 | ||
8f5d549f AK |
2955 | static void ept_save_pdptrs(struct kvm_vcpu *vcpu) |
2956 | { | |
2957 | if (is_paging(vcpu) && is_pae(vcpu) && !is_long_mode(vcpu)) { | |
ff03a073 JR |
2958 | vcpu->arch.mmu.pdptrs[0] = vmcs_read64(GUEST_PDPTR0); |
2959 | vcpu->arch.mmu.pdptrs[1] = vmcs_read64(GUEST_PDPTR1); | |
2960 | vcpu->arch.mmu.pdptrs[2] = vmcs_read64(GUEST_PDPTR2); | |
2961 | vcpu->arch.mmu.pdptrs[3] = vmcs_read64(GUEST_PDPTR3); | |
8f5d549f | 2962 | } |
6de4f3ad AK |
2963 | |
2964 | __set_bit(VCPU_EXREG_PDPTR, | |
2965 | (unsigned long *)&vcpu->arch.regs_avail); | |
2966 | __set_bit(VCPU_EXREG_PDPTR, | |
2967 | (unsigned long *)&vcpu->arch.regs_dirty); | |
8f5d549f AK |
2968 | } |
2969 | ||
5e1746d6 | 2970 | static int vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4); |
1439442c SY |
2971 | |
2972 | static void ept_update_paging_mode_cr0(unsigned long *hw_cr0, | |
2973 | unsigned long cr0, | |
2974 | struct kvm_vcpu *vcpu) | |
2975 | { | |
5233dd51 MT |
2976 | if (!test_bit(VCPU_EXREG_CR3, (ulong *)&vcpu->arch.regs_avail)) |
2977 | vmx_decache_cr3(vcpu); | |
1439442c SY |
2978 | if (!(cr0 & X86_CR0_PG)) { |
2979 | /* From paging/starting to nonpaging */ | |
2980 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, | |
65267ea1 | 2981 | vmcs_read32(CPU_BASED_VM_EXEC_CONTROL) | |
1439442c SY |
2982 | (CPU_BASED_CR3_LOAD_EXITING | |
2983 | CPU_BASED_CR3_STORE_EXITING)); | |
2984 | vcpu->arch.cr0 = cr0; | |
fc78f519 | 2985 | vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
1439442c SY |
2986 | } else if (!is_paging(vcpu)) { |
2987 | /* From nonpaging to paging */ | |
2988 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, | |
65267ea1 | 2989 | vmcs_read32(CPU_BASED_VM_EXEC_CONTROL) & |
1439442c SY |
2990 | ~(CPU_BASED_CR3_LOAD_EXITING | |
2991 | CPU_BASED_CR3_STORE_EXITING)); | |
2992 | vcpu->arch.cr0 = cr0; | |
fc78f519 | 2993 | vmx_set_cr4(vcpu, kvm_read_cr4(vcpu)); |
1439442c | 2994 | } |
95eb84a7 SY |
2995 | |
2996 | if (!(cr0 & X86_CR0_WP)) | |
2997 | *hw_cr0 &= ~X86_CR0_WP; | |
1439442c SY |
2998 | } |
2999 | ||
6aa8b732 AK |
3000 | static void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) |
3001 | { | |
7ffd92c5 | 3002 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
3a624e29 NK |
3003 | unsigned long hw_cr0; |
3004 | ||
3005 | if (enable_unrestricted_guest) | |
3006 | hw_cr0 = (cr0 & ~KVM_GUEST_CR0_MASK_UNRESTRICTED_GUEST) | |
3007 | | KVM_VM_CR0_ALWAYS_ON_UNRESTRICTED_GUEST; | |
3008 | else | |
3009 | hw_cr0 = (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON; | |
1439442c | 3010 | |
7ffd92c5 | 3011 | if (vmx->rmode.vm86_active && (cr0 & X86_CR0_PE)) |
6aa8b732 AK |
3012 | enter_pmode(vcpu); |
3013 | ||
7ffd92c5 | 3014 | if (!vmx->rmode.vm86_active && !(cr0 & X86_CR0_PE)) |
6aa8b732 AK |
3015 | enter_rmode(vcpu); |
3016 | ||
05b3e0c2 | 3017 | #ifdef CONFIG_X86_64 |
f6801dff | 3018 | if (vcpu->arch.efer & EFER_LME) { |
707d92fa | 3019 | if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) |
6aa8b732 | 3020 | enter_lmode(vcpu); |
707d92fa | 3021 | if (is_paging(vcpu) && !(cr0 & X86_CR0_PG)) |
6aa8b732 AK |
3022 | exit_lmode(vcpu); |
3023 | } | |
3024 | #endif | |
3025 | ||
089d034e | 3026 | if (enable_ept) |
1439442c SY |
3027 | ept_update_paging_mode_cr0(&hw_cr0, cr0, vcpu); |
3028 | ||
02daab21 | 3029 | if (!vcpu->fpu_active) |
81231c69 | 3030 | hw_cr0 |= X86_CR0_TS | X86_CR0_MP; |
02daab21 | 3031 | |
6aa8b732 | 3032 | vmcs_writel(CR0_READ_SHADOW, cr0); |
1439442c | 3033 | vmcs_writel(GUEST_CR0, hw_cr0); |
ad312c7c | 3034 | vcpu->arch.cr0 = cr0; |
69c73028 | 3035 | __clear_bit(VCPU_EXREG_CPL, (ulong *)&vcpu->arch.regs_avail); |
6aa8b732 AK |
3036 | } |
3037 | ||
1439442c SY |
3038 | static u64 construct_eptp(unsigned long root_hpa) |
3039 | { | |
3040 | u64 eptp; | |
3041 | ||
3042 | /* TODO write the value reading from MSR */ | |
3043 | eptp = VMX_EPT_DEFAULT_MT | | |
3044 | VMX_EPT_DEFAULT_GAW << VMX_EPT_GAW_EPTP_SHIFT; | |
b38f9934 XH |
3045 | if (enable_ept_ad_bits) |
3046 | eptp |= VMX_EPT_AD_ENABLE_BIT; | |
1439442c SY |
3047 | eptp |= (root_hpa & PAGE_MASK); |
3048 | ||
3049 | return eptp; | |
3050 | } | |
3051 | ||
6aa8b732 AK |
3052 | static void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) |
3053 | { | |
1439442c SY |
3054 | unsigned long guest_cr3; |
3055 | u64 eptp; | |
3056 | ||
3057 | guest_cr3 = cr3; | |
089d034e | 3058 | if (enable_ept) { |
1439442c SY |
3059 | eptp = construct_eptp(cr3); |
3060 | vmcs_write64(EPT_POINTER, eptp); | |
9f8fe504 | 3061 | guest_cr3 = is_paging(vcpu) ? kvm_read_cr3(vcpu) : |
b927a3ce | 3062 | vcpu->kvm->arch.ept_identity_map_addr; |
7c93be44 | 3063 | ept_load_pdptrs(vcpu); |
1439442c SY |
3064 | } |
3065 | ||
2384d2b3 | 3066 | vmx_flush_tlb(vcpu); |
1439442c | 3067 | vmcs_writel(GUEST_CR3, guest_cr3); |
6aa8b732 AK |
3068 | } |
3069 | ||
5e1746d6 | 3070 | static int vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) |
6aa8b732 | 3071 | { |
7ffd92c5 | 3072 | unsigned long hw_cr4 = cr4 | (to_vmx(vcpu)->rmode.vm86_active ? |
1439442c SY |
3073 | KVM_RMODE_VM_CR4_ALWAYS_ON : KVM_PMODE_VM_CR4_ALWAYS_ON); |
3074 | ||
5e1746d6 NHE |
3075 | if (cr4 & X86_CR4_VMXE) { |
3076 | /* | |
3077 | * To use VMXON (and later other VMX instructions), a guest | |
3078 | * must first be able to turn on cr4.VMXE (see handle_vmon()). | |
3079 | * So basically the check on whether to allow nested VMX | |
3080 | * is here. | |
3081 | */ | |
3082 | if (!nested_vmx_allowed(vcpu)) | |
3083 | return 1; | |
3084 | } else if (to_vmx(vcpu)->nested.vmxon) | |
3085 | return 1; | |
3086 | ||
ad312c7c | 3087 | vcpu->arch.cr4 = cr4; |
bc23008b AK |
3088 | if (enable_ept) { |
3089 | if (!is_paging(vcpu)) { | |
3090 | hw_cr4 &= ~X86_CR4_PAE; | |
3091 | hw_cr4 |= X86_CR4_PSE; | |
3092 | } else if (!(cr4 & X86_CR4_PAE)) { | |
3093 | hw_cr4 &= ~X86_CR4_PAE; | |
3094 | } | |
3095 | } | |
1439442c SY |
3096 | |
3097 | vmcs_writel(CR4_READ_SHADOW, cr4); | |
3098 | vmcs_writel(GUEST_CR4, hw_cr4); | |
5e1746d6 | 3099 | return 0; |
6aa8b732 AK |
3100 | } |
3101 | ||
6aa8b732 AK |
3102 | static void vmx_get_segment(struct kvm_vcpu *vcpu, |
3103 | struct kvm_segment *var, int seg) | |
3104 | { | |
a9179499 | 3105 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
6aa8b732 AK |
3106 | u32 ar; |
3107 | ||
a9179499 AK |
3108 | if (vmx->rmode.vm86_active |
3109 | && (seg == VCPU_SREG_TR || seg == VCPU_SREG_ES | |
3110 | || seg == VCPU_SREG_DS || seg == VCPU_SREG_FS | |
72636420 | 3111 | || seg == VCPU_SREG_GS)) { |
f5f7b2fe | 3112 | *var = vmx->rmode.segs[seg]; |
a9179499 | 3113 | if (seg == VCPU_SREG_TR |
2fb92db1 | 3114 | || var->selector == vmx_read_guest_seg_selector(vmx, seg)) |
f5f7b2fe | 3115 | return; |
1390a28b AK |
3116 | var->base = vmx_read_guest_seg_base(vmx, seg); |
3117 | var->selector = vmx_read_guest_seg_selector(vmx, seg); | |
3118 | return; | |
a9179499 | 3119 | } |
2fb92db1 AK |
3120 | var->base = vmx_read_guest_seg_base(vmx, seg); |
3121 | var->limit = vmx_read_guest_seg_limit(vmx, seg); | |
3122 | var->selector = vmx_read_guest_seg_selector(vmx, seg); | |
3123 | ar = vmx_read_guest_seg_ar(vmx, seg); | |
9fd4a3b7 | 3124 | if ((ar & AR_UNUSABLE_MASK) && !emulate_invalid_guest_state) |
6aa8b732 AK |
3125 | ar = 0; |
3126 | var->type = ar & 15; | |
3127 | var->s = (ar >> 4) & 1; | |
3128 | var->dpl = (ar >> 5) & 3; | |
3129 | var->present = (ar >> 7) & 1; | |
3130 | var->avl = (ar >> 12) & 1; | |
3131 | var->l = (ar >> 13) & 1; | |
3132 | var->db = (ar >> 14) & 1; | |
3133 | var->g = (ar >> 15) & 1; | |
3134 | var->unusable = (ar >> 16) & 1; | |
3135 | } | |
3136 | ||
a9179499 AK |
3137 | static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg) |
3138 | { | |
a9179499 AK |
3139 | struct kvm_segment s; |
3140 | ||
3141 | if (to_vmx(vcpu)->rmode.vm86_active) { | |
3142 | vmx_get_segment(vcpu, &s, seg); | |
3143 | return s.base; | |
3144 | } | |
2fb92db1 | 3145 | return vmx_read_guest_seg_base(to_vmx(vcpu), seg); |
a9179499 AK |
3146 | } |
3147 | ||
69c73028 | 3148 | static int __vmx_get_cpl(struct kvm_vcpu *vcpu) |
2e4d2653 | 3149 | { |
3eeb3288 | 3150 | if (!is_protmode(vcpu)) |
2e4d2653 IE |
3151 | return 0; |
3152 | ||
f4c63e5d AK |
3153 | if (!is_long_mode(vcpu) |
3154 | && (kvm_get_rflags(vcpu) & X86_EFLAGS_VM)) /* if virtual 8086 */ | |
2e4d2653 IE |
3155 | return 3; |
3156 | ||
2fb92db1 | 3157 | return vmx_read_guest_seg_selector(to_vmx(vcpu), VCPU_SREG_CS) & 3; |
2e4d2653 IE |
3158 | } |
3159 | ||
69c73028 AK |
3160 | static int vmx_get_cpl(struct kvm_vcpu *vcpu) |
3161 | { | |
d881e6f6 AK |
3162 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
3163 | ||
3164 | /* | |
3165 | * If we enter real mode with cs.sel & 3 != 0, the normal CPL calculations | |
3166 | * fail; use the cache instead. | |
3167 | */ | |
3168 | if (unlikely(vmx->emulation_required && emulate_invalid_guest_state)) { | |
3169 | return vmx->cpl; | |
3170 | } | |
3171 | ||
69c73028 AK |
3172 | if (!test_bit(VCPU_EXREG_CPL, (ulong *)&vcpu->arch.regs_avail)) { |
3173 | __set_bit(VCPU_EXREG_CPL, (ulong *)&vcpu->arch.regs_avail); | |
d881e6f6 | 3174 | vmx->cpl = __vmx_get_cpl(vcpu); |
69c73028 | 3175 | } |
d881e6f6 AK |
3176 | |
3177 | return vmx->cpl; | |
69c73028 AK |
3178 | } |
3179 | ||
3180 | ||
653e3108 | 3181 | static u32 vmx_segment_access_rights(struct kvm_segment *var) |
6aa8b732 | 3182 | { |
6aa8b732 AK |
3183 | u32 ar; |
3184 | ||
f0495f9b | 3185 | if (var->unusable || !var->present) |
6aa8b732 AK |
3186 | ar = 1 << 16; |
3187 | else { | |
3188 | ar = var->type & 15; | |
3189 | ar |= (var->s & 1) << 4; | |
3190 | ar |= (var->dpl & 3) << 5; | |
3191 | ar |= (var->present & 1) << 7; | |
3192 | ar |= (var->avl & 1) << 12; | |
3193 | ar |= (var->l & 1) << 13; | |
3194 | ar |= (var->db & 1) << 14; | |
3195 | ar |= (var->g & 1) << 15; | |
3196 | } | |
653e3108 AK |
3197 | |
3198 | return ar; | |
3199 | } | |
3200 | ||
3201 | static void vmx_set_segment(struct kvm_vcpu *vcpu, | |
3202 | struct kvm_segment *var, int seg) | |
3203 | { | |
7ffd92c5 | 3204 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
653e3108 AK |
3205 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; |
3206 | u32 ar; | |
3207 | ||
2fb92db1 AK |
3208 | vmx_segment_cache_clear(vmx); |
3209 | ||
7ffd92c5 | 3210 | if (vmx->rmode.vm86_active && seg == VCPU_SREG_TR) { |
a8ba6c26 | 3211 | vmcs_write16(sf->selector, var->selector); |
f5f7b2fe | 3212 | vmx->rmode.segs[VCPU_SREG_TR] = *var; |
653e3108 AK |
3213 | return; |
3214 | } | |
3215 | vmcs_writel(sf->base, var->base); | |
3216 | vmcs_write32(sf->limit, var->limit); | |
3217 | vmcs_write16(sf->selector, var->selector); | |
7ffd92c5 | 3218 | if (vmx->rmode.vm86_active && var->s) { |
653e3108 AK |
3219 | /* |
3220 | * Hack real-mode segments into vm86 compatibility. | |
3221 | */ | |
3222 | if (var->base == 0xffff0000 && var->selector == 0xf000) | |
3223 | vmcs_writel(sf->base, 0xf0000); | |
3224 | ar = 0xf3; | |
3225 | } else | |
3226 | ar = vmx_segment_access_rights(var); | |
3a624e29 NK |
3227 | |
3228 | /* | |
3229 | * Fix the "Accessed" bit in AR field of segment registers for older | |
3230 | * qemu binaries. | |
3231 | * IA32 arch specifies that at the time of processor reset the | |
3232 | * "Accessed" bit in the AR field of segment registers is 1. And qemu | |
0fa06071 | 3233 | * is setting it to 0 in the userland code. This causes invalid guest |
3a624e29 NK |
3234 | * state vmexit when "unrestricted guest" mode is turned on. |
3235 | * Fix for this setup issue in cpu_reset is being pushed in the qemu | |
3236 | * tree. Newer qemu binaries with that qemu fix would not need this | |
3237 | * kvm hack. | |
3238 | */ | |
3239 | if (enable_unrestricted_guest && (seg != VCPU_SREG_LDTR)) | |
3240 | ar |= 0x1; /* Accessed */ | |
3241 | ||
6aa8b732 | 3242 | vmcs_write32(sf->ar_bytes, ar); |
69c73028 | 3243 | __clear_bit(VCPU_EXREG_CPL, (ulong *)&vcpu->arch.regs_avail); |
b246dd5d OW |
3244 | |
3245 | /* | |
3246 | * Fix segments for real mode guest in hosts that don't have | |
3247 | * "unrestricted_mode" or it was disabled. | |
3248 | * This is done to allow migration of the guests from hosts with | |
3249 | * unrestricted guest like Westmere to older host that don't have | |
3250 | * unrestricted guest like Nehelem. | |
3251 | */ | |
3252 | if (!enable_unrestricted_guest && vmx->rmode.vm86_active) { | |
3253 | switch (seg) { | |
3254 | case VCPU_SREG_CS: | |
3255 | vmcs_write32(GUEST_CS_AR_BYTES, 0xf3); | |
3256 | vmcs_write32(GUEST_CS_LIMIT, 0xffff); | |
3257 | if (vmcs_readl(GUEST_CS_BASE) == 0xffff0000) | |
3258 | vmcs_writel(GUEST_CS_BASE, 0xf0000); | |
3259 | vmcs_write16(GUEST_CS_SELECTOR, | |
3260 | vmcs_readl(GUEST_CS_BASE) >> 4); | |
3261 | break; | |
3262 | case VCPU_SREG_ES: | |
b246dd5d | 3263 | case VCPU_SREG_DS: |
b246dd5d | 3264 | case VCPU_SREG_GS: |
b246dd5d | 3265 | case VCPU_SREG_FS: |
f5f7b2fe | 3266 | fix_rmode_seg(seg, &vmx->rmode.segs[seg]); |
b246dd5d OW |
3267 | break; |
3268 | case VCPU_SREG_SS: | |
3269 | vmcs_write16(GUEST_SS_SELECTOR, | |
3270 | vmcs_readl(GUEST_SS_BASE) >> 4); | |
3271 | vmcs_write32(GUEST_SS_LIMIT, 0xffff); | |
3272 | vmcs_write32(GUEST_SS_AR_BYTES, 0xf3); | |
3273 | break; | |
3274 | } | |
3275 | } | |
6aa8b732 AK |
3276 | } |
3277 | ||
6aa8b732 AK |
3278 | static void vmx_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l) |
3279 | { | |
2fb92db1 | 3280 | u32 ar = vmx_read_guest_seg_ar(to_vmx(vcpu), VCPU_SREG_CS); |
6aa8b732 AK |
3281 | |
3282 | *db = (ar >> 14) & 1; | |
3283 | *l = (ar >> 13) & 1; | |
3284 | } | |
3285 | ||
89a27f4d | 3286 | static void vmx_get_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 3287 | { |
89a27f4d GN |
3288 | dt->size = vmcs_read32(GUEST_IDTR_LIMIT); |
3289 | dt->address = vmcs_readl(GUEST_IDTR_BASE); | |
6aa8b732 AK |
3290 | } |
3291 | ||
89a27f4d | 3292 | static void vmx_set_idt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 3293 | { |
89a27f4d GN |
3294 | vmcs_write32(GUEST_IDTR_LIMIT, dt->size); |
3295 | vmcs_writel(GUEST_IDTR_BASE, dt->address); | |
6aa8b732 AK |
3296 | } |
3297 | ||
89a27f4d | 3298 | static void vmx_get_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 3299 | { |
89a27f4d GN |
3300 | dt->size = vmcs_read32(GUEST_GDTR_LIMIT); |
3301 | dt->address = vmcs_readl(GUEST_GDTR_BASE); | |
6aa8b732 AK |
3302 | } |
3303 | ||
89a27f4d | 3304 | static void vmx_set_gdt(struct kvm_vcpu *vcpu, struct desc_ptr *dt) |
6aa8b732 | 3305 | { |
89a27f4d GN |
3306 | vmcs_write32(GUEST_GDTR_LIMIT, dt->size); |
3307 | vmcs_writel(GUEST_GDTR_BASE, dt->address); | |
6aa8b732 AK |
3308 | } |
3309 | ||
648dfaa7 MG |
3310 | static bool rmode_segment_valid(struct kvm_vcpu *vcpu, int seg) |
3311 | { | |
3312 | struct kvm_segment var; | |
3313 | u32 ar; | |
3314 | ||
3315 | vmx_get_segment(vcpu, &var, seg); | |
3316 | ar = vmx_segment_access_rights(&var); | |
3317 | ||
3318 | if (var.base != (var.selector << 4)) | |
3319 | return false; | |
e2a610d7 | 3320 | if (var.limit < 0xffff) |
648dfaa7 | 3321 | return false; |
495e1166 | 3322 | if ((ar | (3 << AR_DPL_SHIFT)) != 0xf3) |
648dfaa7 MG |
3323 | return false; |
3324 | ||
3325 | return true; | |
3326 | } | |
3327 | ||
3328 | static bool code_segment_valid(struct kvm_vcpu *vcpu) | |
3329 | { | |
3330 | struct kvm_segment cs; | |
3331 | unsigned int cs_rpl; | |
3332 | ||
3333 | vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); | |
3334 | cs_rpl = cs.selector & SELECTOR_RPL_MASK; | |
3335 | ||
1872a3f4 AK |
3336 | if (cs.unusable) |
3337 | return false; | |
648dfaa7 MG |
3338 | if (~cs.type & (AR_TYPE_CODE_MASK|AR_TYPE_ACCESSES_MASK)) |
3339 | return false; | |
3340 | if (!cs.s) | |
3341 | return false; | |
1872a3f4 | 3342 | if (cs.type & AR_TYPE_WRITEABLE_MASK) { |
648dfaa7 MG |
3343 | if (cs.dpl > cs_rpl) |
3344 | return false; | |
1872a3f4 | 3345 | } else { |
648dfaa7 MG |
3346 | if (cs.dpl != cs_rpl) |
3347 | return false; | |
3348 | } | |
3349 | if (!cs.present) | |
3350 | return false; | |
3351 | ||
3352 | /* TODO: Add Reserved field check, this'll require a new member in the kvm_segment_field structure */ | |
3353 | return true; | |
3354 | } | |
3355 | ||
3356 | static bool stack_segment_valid(struct kvm_vcpu *vcpu) | |
3357 | { | |
3358 | struct kvm_segment ss; | |
3359 | unsigned int ss_rpl; | |
3360 | ||
3361 | vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); | |
3362 | ss_rpl = ss.selector & SELECTOR_RPL_MASK; | |
3363 | ||
1872a3f4 AK |
3364 | if (ss.unusable) |
3365 | return true; | |
3366 | if (ss.type != 3 && ss.type != 7) | |
648dfaa7 MG |
3367 | return false; |
3368 | if (!ss.s) | |
3369 | return false; | |
3370 | if (ss.dpl != ss_rpl) /* DPL != RPL */ | |
3371 | return false; | |
3372 | if (!ss.present) | |
3373 | return false; | |
3374 | ||
3375 | return true; | |
3376 | } | |
3377 | ||
3378 | static bool data_segment_valid(struct kvm_vcpu *vcpu, int seg) | |
3379 | { | |
3380 | struct kvm_segment var; | |
3381 | unsigned int rpl; | |
3382 | ||
3383 | vmx_get_segment(vcpu, &var, seg); | |
3384 | rpl = var.selector & SELECTOR_RPL_MASK; | |
3385 | ||
1872a3f4 AK |
3386 | if (var.unusable) |
3387 | return true; | |
648dfaa7 MG |
3388 | if (!var.s) |
3389 | return false; | |
3390 | if (!var.present) | |
3391 | return false; | |
3392 | if (~var.type & (AR_TYPE_CODE_MASK|AR_TYPE_WRITEABLE_MASK)) { | |
3393 | if (var.dpl < rpl) /* DPL < RPL */ | |
3394 | return false; | |
3395 | } | |
3396 | ||
3397 | /* TODO: Add other members to kvm_segment_field to allow checking for other access | |
3398 | * rights flags | |
3399 | */ | |
3400 | return true; | |
3401 | } | |
3402 | ||
3403 | static bool tr_valid(struct kvm_vcpu *vcpu) | |
3404 | { | |
3405 | struct kvm_segment tr; | |
3406 | ||
3407 | vmx_get_segment(vcpu, &tr, VCPU_SREG_TR); | |
3408 | ||
1872a3f4 AK |
3409 | if (tr.unusable) |
3410 | return false; | |
648dfaa7 MG |
3411 | if (tr.selector & SELECTOR_TI_MASK) /* TI = 1 */ |
3412 | return false; | |
1872a3f4 | 3413 | if (tr.type != 3 && tr.type != 11) /* TODO: Check if guest is in IA32e mode */ |
648dfaa7 MG |
3414 | return false; |
3415 | if (!tr.present) | |
3416 | return false; | |
3417 | ||
3418 | return true; | |
3419 | } | |
3420 | ||
3421 | static bool ldtr_valid(struct kvm_vcpu *vcpu) | |
3422 | { | |
3423 | struct kvm_segment ldtr; | |
3424 | ||
3425 | vmx_get_segment(vcpu, &ldtr, VCPU_SREG_LDTR); | |
3426 | ||
1872a3f4 AK |
3427 | if (ldtr.unusable) |
3428 | return true; | |
648dfaa7 MG |
3429 | if (ldtr.selector & SELECTOR_TI_MASK) /* TI = 1 */ |
3430 | return false; | |
3431 | if (ldtr.type != 2) | |
3432 | return false; | |
3433 | if (!ldtr.present) | |
3434 | return false; | |
3435 | ||
3436 | return true; | |
3437 | } | |
3438 | ||
3439 | static bool cs_ss_rpl_check(struct kvm_vcpu *vcpu) | |
3440 | { | |
3441 | struct kvm_segment cs, ss; | |
3442 | ||
3443 | vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); | |
3444 | vmx_get_segment(vcpu, &ss, VCPU_SREG_SS); | |
3445 | ||
3446 | return ((cs.selector & SELECTOR_RPL_MASK) == | |
3447 | (ss.selector & SELECTOR_RPL_MASK)); | |
3448 | } | |
3449 | ||
3450 | /* | |
3451 | * Check if guest state is valid. Returns true if valid, false if | |
3452 | * not. | |
3453 | * We assume that registers are always usable | |
3454 | */ | |
3455 | static bool guest_state_valid(struct kvm_vcpu *vcpu) | |
3456 | { | |
3457 | /* real mode guest state checks */ | |
3eeb3288 | 3458 | if (!is_protmode(vcpu)) { |
648dfaa7 MG |
3459 | if (!rmode_segment_valid(vcpu, VCPU_SREG_CS)) |
3460 | return false; | |
3461 | if (!rmode_segment_valid(vcpu, VCPU_SREG_SS)) | |
3462 | return false; | |
3463 | if (!rmode_segment_valid(vcpu, VCPU_SREG_DS)) | |
3464 | return false; | |
3465 | if (!rmode_segment_valid(vcpu, VCPU_SREG_ES)) | |
3466 | return false; | |
3467 | if (!rmode_segment_valid(vcpu, VCPU_SREG_FS)) | |
3468 | return false; | |
3469 | if (!rmode_segment_valid(vcpu, VCPU_SREG_GS)) | |
3470 | return false; | |
3471 | } else { | |
3472 | /* protected mode guest state checks */ | |
3473 | if (!cs_ss_rpl_check(vcpu)) | |
3474 | return false; | |
3475 | if (!code_segment_valid(vcpu)) | |
3476 | return false; | |
3477 | if (!stack_segment_valid(vcpu)) | |
3478 | return false; | |
3479 | if (!data_segment_valid(vcpu, VCPU_SREG_DS)) | |
3480 | return false; | |
3481 | if (!data_segment_valid(vcpu, VCPU_SREG_ES)) | |
3482 | return false; | |
3483 | if (!data_segment_valid(vcpu, VCPU_SREG_FS)) | |
3484 | return false; | |
3485 | if (!data_segment_valid(vcpu, VCPU_SREG_GS)) | |
3486 | return false; | |
3487 | if (!tr_valid(vcpu)) | |
3488 | return false; | |
3489 | if (!ldtr_valid(vcpu)) | |
3490 | return false; | |
3491 | } | |
3492 | /* TODO: | |
3493 | * - Add checks on RIP | |
3494 | * - Add checks on RFLAGS | |
3495 | */ | |
3496 | ||
3497 | return true; | |
3498 | } | |
3499 | ||
d77c26fc | 3500 | static int init_rmode_tss(struct kvm *kvm) |
6aa8b732 | 3501 | { |
40dcaa9f | 3502 | gfn_t fn; |
195aefde | 3503 | u16 data = 0; |
40dcaa9f | 3504 | int r, idx, ret = 0; |
6aa8b732 | 3505 | |
40dcaa9f XG |
3506 | idx = srcu_read_lock(&kvm->srcu); |
3507 | fn = rmode_tss_base(kvm) >> PAGE_SHIFT; | |
195aefde IE |
3508 | r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
3509 | if (r < 0) | |
10589a46 | 3510 | goto out; |
195aefde | 3511 | data = TSS_BASE_SIZE + TSS_REDIRECTION_SIZE; |
464d17c8 SY |
3512 | r = kvm_write_guest_page(kvm, fn++, &data, |
3513 | TSS_IOPB_BASE_OFFSET, sizeof(u16)); | |
195aefde | 3514 | if (r < 0) |
10589a46 | 3515 | goto out; |
195aefde IE |
3516 | r = kvm_clear_guest_page(kvm, fn++, 0, PAGE_SIZE); |
3517 | if (r < 0) | |
10589a46 | 3518 | goto out; |
195aefde IE |
3519 | r = kvm_clear_guest_page(kvm, fn, 0, PAGE_SIZE); |
3520 | if (r < 0) | |
10589a46 | 3521 | goto out; |
195aefde | 3522 | data = ~0; |
10589a46 MT |
3523 | r = kvm_write_guest_page(kvm, fn, &data, |
3524 | RMODE_TSS_SIZE - 2 * PAGE_SIZE - 1, | |
3525 | sizeof(u8)); | |
195aefde | 3526 | if (r < 0) |
10589a46 MT |
3527 | goto out; |
3528 | ||
3529 | ret = 1; | |
3530 | out: | |
40dcaa9f | 3531 | srcu_read_unlock(&kvm->srcu, idx); |
10589a46 | 3532 | return ret; |
6aa8b732 AK |
3533 | } |
3534 | ||
b7ebfb05 SY |
3535 | static int init_rmode_identity_map(struct kvm *kvm) |
3536 | { | |
40dcaa9f | 3537 | int i, idx, r, ret; |
b7ebfb05 SY |
3538 | pfn_t identity_map_pfn; |
3539 | u32 tmp; | |
3540 | ||
089d034e | 3541 | if (!enable_ept) |
b7ebfb05 SY |
3542 | return 1; |
3543 | if (unlikely(!kvm->arch.ept_identity_pagetable)) { | |
3544 | printk(KERN_ERR "EPT: identity-mapping pagetable " | |
3545 | "haven't been allocated!\n"); | |
3546 | return 0; | |
3547 | } | |
3548 | if (likely(kvm->arch.ept_identity_pagetable_done)) | |
3549 | return 1; | |
3550 | ret = 0; | |
b927a3ce | 3551 | identity_map_pfn = kvm->arch.ept_identity_map_addr >> PAGE_SHIFT; |
40dcaa9f | 3552 | idx = srcu_read_lock(&kvm->srcu); |
b7ebfb05 SY |
3553 | r = kvm_clear_guest_page(kvm, identity_map_pfn, 0, PAGE_SIZE); |
3554 | if (r < 0) | |
3555 | goto out; | |
3556 | /* Set up identity-mapping pagetable for EPT in real mode */ | |
3557 | for (i = 0; i < PT32_ENT_PER_PAGE; i++) { | |
3558 | tmp = (i << 22) + (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER | | |
3559 | _PAGE_ACCESSED | _PAGE_DIRTY | _PAGE_PSE); | |
3560 | r = kvm_write_guest_page(kvm, identity_map_pfn, | |
3561 | &tmp, i * sizeof(tmp), sizeof(tmp)); | |
3562 | if (r < 0) | |
3563 | goto out; | |
3564 | } | |
3565 | kvm->arch.ept_identity_pagetable_done = true; | |
3566 | ret = 1; | |
3567 | out: | |
40dcaa9f | 3568 | srcu_read_unlock(&kvm->srcu, idx); |
b7ebfb05 SY |
3569 | return ret; |
3570 | } | |
3571 | ||
6aa8b732 AK |
3572 | static void seg_setup(int seg) |
3573 | { | |
3574 | struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg]; | |
3a624e29 | 3575 | unsigned int ar; |
6aa8b732 AK |
3576 | |
3577 | vmcs_write16(sf->selector, 0); | |
3578 | vmcs_writel(sf->base, 0); | |
3579 | vmcs_write32(sf->limit, 0xffff); | |
3a624e29 NK |
3580 | if (enable_unrestricted_guest) { |
3581 | ar = 0x93; | |
3582 | if (seg == VCPU_SREG_CS) | |
3583 | ar |= 0x08; /* code segment */ | |
3584 | } else | |
3585 | ar = 0xf3; | |
3586 | ||
3587 | vmcs_write32(sf->ar_bytes, ar); | |
6aa8b732 AK |
3588 | } |
3589 | ||
f78e0e2e SY |
3590 | static int alloc_apic_access_page(struct kvm *kvm) |
3591 | { | |
3592 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
3593 | int r = 0; | |
3594 | ||
79fac95e | 3595 | mutex_lock(&kvm->slots_lock); |
bfc6d222 | 3596 | if (kvm->arch.apic_access_page) |
f78e0e2e SY |
3597 | goto out; |
3598 | kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT; | |
3599 | kvm_userspace_mem.flags = 0; | |
3600 | kvm_userspace_mem.guest_phys_addr = 0xfee00000ULL; | |
3601 | kvm_userspace_mem.memory_size = PAGE_SIZE; | |
3602 | r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0); | |
3603 | if (r) | |
3604 | goto out; | |
72dc67a6 | 3605 | |
bfc6d222 | 3606 | kvm->arch.apic_access_page = gfn_to_page(kvm, 0xfee00); |
f78e0e2e | 3607 | out: |
79fac95e | 3608 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
3609 | return r; |
3610 | } | |
3611 | ||
b7ebfb05 SY |
3612 | static int alloc_identity_pagetable(struct kvm *kvm) |
3613 | { | |
3614 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
3615 | int r = 0; | |
3616 | ||
79fac95e | 3617 | mutex_lock(&kvm->slots_lock); |
b7ebfb05 SY |
3618 | if (kvm->arch.ept_identity_pagetable) |
3619 | goto out; | |
3620 | kvm_userspace_mem.slot = IDENTITY_PAGETABLE_PRIVATE_MEMSLOT; | |
3621 | kvm_userspace_mem.flags = 0; | |
b927a3ce SY |
3622 | kvm_userspace_mem.guest_phys_addr = |
3623 | kvm->arch.ept_identity_map_addr; | |
b7ebfb05 SY |
3624 | kvm_userspace_mem.memory_size = PAGE_SIZE; |
3625 | r = __kvm_set_memory_region(kvm, &kvm_userspace_mem, 0); | |
3626 | if (r) | |
3627 | goto out; | |
3628 | ||
b7ebfb05 | 3629 | kvm->arch.ept_identity_pagetable = gfn_to_page(kvm, |
b927a3ce | 3630 | kvm->arch.ept_identity_map_addr >> PAGE_SHIFT); |
b7ebfb05 | 3631 | out: |
79fac95e | 3632 | mutex_unlock(&kvm->slots_lock); |
b7ebfb05 SY |
3633 | return r; |
3634 | } | |
3635 | ||
2384d2b3 SY |
3636 | static void allocate_vpid(struct vcpu_vmx *vmx) |
3637 | { | |
3638 | int vpid; | |
3639 | ||
3640 | vmx->vpid = 0; | |
919818ab | 3641 | if (!enable_vpid) |
2384d2b3 SY |
3642 | return; |
3643 | spin_lock(&vmx_vpid_lock); | |
3644 | vpid = find_first_zero_bit(vmx_vpid_bitmap, VMX_NR_VPIDS); | |
3645 | if (vpid < VMX_NR_VPIDS) { | |
3646 | vmx->vpid = vpid; | |
3647 | __set_bit(vpid, vmx_vpid_bitmap); | |
3648 | } | |
3649 | spin_unlock(&vmx_vpid_lock); | |
3650 | } | |
3651 | ||
cdbecfc3 LJ |
3652 | static void free_vpid(struct vcpu_vmx *vmx) |
3653 | { | |
3654 | if (!enable_vpid) | |
3655 | return; | |
3656 | spin_lock(&vmx_vpid_lock); | |
3657 | if (vmx->vpid != 0) | |
3658 | __clear_bit(vmx->vpid, vmx_vpid_bitmap); | |
3659 | spin_unlock(&vmx_vpid_lock); | |
3660 | } | |
3661 | ||
5897297b | 3662 | static void __vmx_disable_intercept_for_msr(unsigned long *msr_bitmap, u32 msr) |
25c5f225 | 3663 | { |
3e7c73e9 | 3664 | int f = sizeof(unsigned long); |
25c5f225 SY |
3665 | |
3666 | if (!cpu_has_vmx_msr_bitmap()) | |
3667 | return; | |
3668 | ||
3669 | /* | |
3670 | * See Intel PRM Vol. 3, 20.6.9 (MSR-Bitmap Address). Early manuals | |
3671 | * have the write-low and read-high bitmap offsets the wrong way round. | |
3672 | * We can control MSRs 0x00000000-0x00001fff and 0xc0000000-0xc0001fff. | |
3673 | */ | |
25c5f225 | 3674 | if (msr <= 0x1fff) { |
3e7c73e9 AK |
3675 | __clear_bit(msr, msr_bitmap + 0x000 / f); /* read-low */ |
3676 | __clear_bit(msr, msr_bitmap + 0x800 / f); /* write-low */ | |
25c5f225 SY |
3677 | } else if ((msr >= 0xc0000000) && (msr <= 0xc0001fff)) { |
3678 | msr &= 0x1fff; | |
3e7c73e9 AK |
3679 | __clear_bit(msr, msr_bitmap + 0x400 / f); /* read-high */ |
3680 | __clear_bit(msr, msr_bitmap + 0xc00 / f); /* write-high */ | |
25c5f225 | 3681 | } |
25c5f225 SY |
3682 | } |
3683 | ||
5897297b AK |
3684 | static void vmx_disable_intercept_for_msr(u32 msr, bool longmode_only) |
3685 | { | |
3686 | if (!longmode_only) | |
3687 | __vmx_disable_intercept_for_msr(vmx_msr_bitmap_legacy, msr); | |
3688 | __vmx_disable_intercept_for_msr(vmx_msr_bitmap_longmode, msr); | |
3689 | } | |
3690 | ||
a3a8ff8e NHE |
3691 | /* |
3692 | * Set up the vmcs's constant host-state fields, i.e., host-state fields that | |
3693 | * will not change in the lifetime of the guest. | |
3694 | * Note that host-state that does change is set elsewhere. E.g., host-state | |
3695 | * that is set differently for each CPU is set in vmx_vcpu_load(), not here. | |
3696 | */ | |
3697 | static void vmx_set_constant_host_state(void) | |
3698 | { | |
3699 | u32 low32, high32; | |
3700 | unsigned long tmpl; | |
3701 | struct desc_ptr dt; | |
3702 | ||
3703 | vmcs_writel(HOST_CR0, read_cr0() | X86_CR0_TS); /* 22.2.3 */ | |
3704 | vmcs_writel(HOST_CR4, read_cr4()); /* 22.2.3, 22.2.5 */ | |
3705 | vmcs_writel(HOST_CR3, read_cr3()); /* 22.2.3 FIXME: shadow tables */ | |
3706 | ||
3707 | vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS); /* 22.2.4 */ | |
b2da15ac AK |
3708 | #ifdef CONFIG_X86_64 |
3709 | /* | |
3710 | * Load null selectors, so we can avoid reloading them in | |
3711 | * __vmx_load_host_state(), in case userspace uses the null selectors | |
3712 | * too (the expected case). | |
3713 | */ | |
3714 | vmcs_write16(HOST_DS_SELECTOR, 0); | |
3715 | vmcs_write16(HOST_ES_SELECTOR, 0); | |
3716 | #else | |
a3a8ff8e NHE |
3717 | vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
3718 | vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS); /* 22.2.4 */ | |
b2da15ac | 3719 | #endif |
a3a8ff8e NHE |
3720 | vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS); /* 22.2.4 */ |
3721 | vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8); /* 22.2.4 */ | |
3722 | ||
3723 | native_store_idt(&dt); | |
3724 | vmcs_writel(HOST_IDTR_BASE, dt.address); /* 22.2.4 */ | |
3725 | ||
3726 | asm("mov $.Lkvm_vmx_return, %0" : "=r"(tmpl)); | |
3727 | vmcs_writel(HOST_RIP, tmpl); /* 22.2.5 */ | |
3728 | ||
3729 | rdmsr(MSR_IA32_SYSENTER_CS, low32, high32); | |
3730 | vmcs_write32(HOST_IA32_SYSENTER_CS, low32); | |
3731 | rdmsrl(MSR_IA32_SYSENTER_EIP, tmpl); | |
3732 | vmcs_writel(HOST_IA32_SYSENTER_EIP, tmpl); /* 22.2.3 */ | |
3733 | ||
3734 | if (vmcs_config.vmexit_ctrl & VM_EXIT_LOAD_IA32_PAT) { | |
3735 | rdmsr(MSR_IA32_CR_PAT, low32, high32); | |
3736 | vmcs_write64(HOST_IA32_PAT, low32 | ((u64) high32 << 32)); | |
3737 | } | |
3738 | } | |
3739 | ||
bf8179a0 NHE |
3740 | static void set_cr4_guest_host_mask(struct vcpu_vmx *vmx) |
3741 | { | |
3742 | vmx->vcpu.arch.cr4_guest_owned_bits = KVM_CR4_GUEST_OWNED_BITS; | |
3743 | if (enable_ept) | |
3744 | vmx->vcpu.arch.cr4_guest_owned_bits |= X86_CR4_PGE; | |
fe3ef05c NHE |
3745 | if (is_guest_mode(&vmx->vcpu)) |
3746 | vmx->vcpu.arch.cr4_guest_owned_bits &= | |
3747 | ~get_vmcs12(&vmx->vcpu)->cr4_guest_host_mask; | |
bf8179a0 NHE |
3748 | vmcs_writel(CR4_GUEST_HOST_MASK, ~vmx->vcpu.arch.cr4_guest_owned_bits); |
3749 | } | |
3750 | ||
3751 | static u32 vmx_exec_control(struct vcpu_vmx *vmx) | |
3752 | { | |
3753 | u32 exec_control = vmcs_config.cpu_based_exec_ctrl; | |
3754 | if (!vm_need_tpr_shadow(vmx->vcpu.kvm)) { | |
3755 | exec_control &= ~CPU_BASED_TPR_SHADOW; | |
3756 | #ifdef CONFIG_X86_64 | |
3757 | exec_control |= CPU_BASED_CR8_STORE_EXITING | | |
3758 | CPU_BASED_CR8_LOAD_EXITING; | |
3759 | #endif | |
3760 | } | |
3761 | if (!enable_ept) | |
3762 | exec_control |= CPU_BASED_CR3_STORE_EXITING | | |
3763 | CPU_BASED_CR3_LOAD_EXITING | | |
3764 | CPU_BASED_INVLPG_EXITING; | |
3765 | return exec_control; | |
3766 | } | |
3767 | ||
3768 | static u32 vmx_secondary_exec_control(struct vcpu_vmx *vmx) | |
3769 | { | |
3770 | u32 exec_control = vmcs_config.cpu_based_2nd_exec_ctrl; | |
3771 | if (!vm_need_virtualize_apic_accesses(vmx->vcpu.kvm)) | |
3772 | exec_control &= ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
3773 | if (vmx->vpid == 0) | |
3774 | exec_control &= ~SECONDARY_EXEC_ENABLE_VPID; | |
3775 | if (!enable_ept) { | |
3776 | exec_control &= ~SECONDARY_EXEC_ENABLE_EPT; | |
3777 | enable_unrestricted_guest = 0; | |
ad756a16 MJ |
3778 | /* Enable INVPCID for non-ept guests may cause performance regression. */ |
3779 | exec_control &= ~SECONDARY_EXEC_ENABLE_INVPCID; | |
bf8179a0 NHE |
3780 | } |
3781 | if (!enable_unrestricted_guest) | |
3782 | exec_control &= ~SECONDARY_EXEC_UNRESTRICTED_GUEST; | |
3783 | if (!ple_gap) | |
3784 | exec_control &= ~SECONDARY_EXEC_PAUSE_LOOP_EXITING; | |
3785 | return exec_control; | |
3786 | } | |
3787 | ||
ce88decf XG |
3788 | static void ept_set_mmio_spte_mask(void) |
3789 | { | |
3790 | /* | |
3791 | * EPT Misconfigurations can be generated if the value of bits 2:0 | |
3792 | * of an EPT paging-structure entry is 110b (write/execute). | |
3793 | * Also, magic bits (0xffull << 49) is set to quickly identify mmio | |
3794 | * spte. | |
3795 | */ | |
3796 | kvm_mmu_set_mmio_spte_mask(0xffull << 49 | 0x6ull); | |
3797 | } | |
3798 | ||
6aa8b732 AK |
3799 | /* |
3800 | * Sets up the vmcs for emulated real mode. | |
3801 | */ | |
8b9cf98c | 3802 | static int vmx_vcpu_setup(struct vcpu_vmx *vmx) |
6aa8b732 | 3803 | { |
2e4ce7f5 | 3804 | #ifdef CONFIG_X86_64 |
6aa8b732 | 3805 | unsigned long a; |
2e4ce7f5 | 3806 | #endif |
6aa8b732 | 3807 | int i; |
6aa8b732 | 3808 | |
6aa8b732 | 3809 | /* I/O */ |
3e7c73e9 AK |
3810 | vmcs_write64(IO_BITMAP_A, __pa(vmx_io_bitmap_a)); |
3811 | vmcs_write64(IO_BITMAP_B, __pa(vmx_io_bitmap_b)); | |
6aa8b732 | 3812 | |
25c5f225 | 3813 | if (cpu_has_vmx_msr_bitmap()) |
5897297b | 3814 | vmcs_write64(MSR_BITMAP, __pa(vmx_msr_bitmap_legacy)); |
25c5f225 | 3815 | |
6aa8b732 AK |
3816 | vmcs_write64(VMCS_LINK_POINTER, -1ull); /* 22.3.1.5 */ |
3817 | ||
6aa8b732 | 3818 | /* Control */ |
1c3d14fe YS |
3819 | vmcs_write32(PIN_BASED_VM_EXEC_CONTROL, |
3820 | vmcs_config.pin_based_exec_ctrl); | |
6e5d865c | 3821 | |
bf8179a0 | 3822 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, vmx_exec_control(vmx)); |
6aa8b732 | 3823 | |
83ff3b9d | 3824 | if (cpu_has_secondary_exec_ctrls()) { |
bf8179a0 NHE |
3825 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, |
3826 | vmx_secondary_exec_control(vmx)); | |
83ff3b9d | 3827 | } |
f78e0e2e | 3828 | |
4b8d54f9 ZE |
3829 | if (ple_gap) { |
3830 | vmcs_write32(PLE_GAP, ple_gap); | |
3831 | vmcs_write32(PLE_WINDOW, ple_window); | |
3832 | } | |
3833 | ||
c3707958 XG |
3834 | vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0); |
3835 | vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0); | |
6aa8b732 AK |
3836 | vmcs_write32(CR3_TARGET_COUNT, 0); /* 22.2.1 */ |
3837 | ||
9581d442 AK |
3838 | vmcs_write16(HOST_FS_SELECTOR, 0); /* 22.2.4 */ |
3839 | vmcs_write16(HOST_GS_SELECTOR, 0); /* 22.2.4 */ | |
a3a8ff8e | 3840 | vmx_set_constant_host_state(); |
05b3e0c2 | 3841 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
3842 | rdmsrl(MSR_FS_BASE, a); |
3843 | vmcs_writel(HOST_FS_BASE, a); /* 22.2.4 */ | |
3844 | rdmsrl(MSR_GS_BASE, a); | |
3845 | vmcs_writel(HOST_GS_BASE, a); /* 22.2.4 */ | |
3846 | #else | |
3847 | vmcs_writel(HOST_FS_BASE, 0); /* 22.2.4 */ | |
3848 | vmcs_writel(HOST_GS_BASE, 0); /* 22.2.4 */ | |
3849 | #endif | |
3850 | ||
2cc51560 ED |
3851 | vmcs_write32(VM_EXIT_MSR_STORE_COUNT, 0); |
3852 | vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, 0); | |
61d2ef2c | 3853 | vmcs_write64(VM_EXIT_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.host)); |
2cc51560 | 3854 | vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, 0); |
61d2ef2c | 3855 | vmcs_write64(VM_ENTRY_MSR_LOAD_ADDR, __pa(vmx->msr_autoload.guest)); |
6aa8b732 | 3856 | |
468d472f | 3857 | if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) { |
a3a8ff8e NHE |
3858 | u32 msr_low, msr_high; |
3859 | u64 host_pat; | |
468d472f SY |
3860 | rdmsr(MSR_IA32_CR_PAT, msr_low, msr_high); |
3861 | host_pat = msr_low | ((u64) msr_high << 32); | |
3862 | /* Write the default value follow host pat */ | |
3863 | vmcs_write64(GUEST_IA32_PAT, host_pat); | |
3864 | /* Keep arch.pat sync with GUEST_IA32_PAT */ | |
3865 | vmx->vcpu.arch.pat = host_pat; | |
3866 | } | |
3867 | ||
6aa8b732 AK |
3868 | for (i = 0; i < NR_VMX_MSR; ++i) { |
3869 | u32 index = vmx_msr_index[i]; | |
3870 | u32 data_low, data_high; | |
a2fa3e9f | 3871 | int j = vmx->nmsrs; |
6aa8b732 AK |
3872 | |
3873 | if (rdmsr_safe(index, &data_low, &data_high) < 0) | |
3874 | continue; | |
432bd6cb AK |
3875 | if (wrmsr_safe(index, data_low, data_high) < 0) |
3876 | continue; | |
26bb0981 AK |
3877 | vmx->guest_msrs[j].index = i; |
3878 | vmx->guest_msrs[j].data = 0; | |
d5696725 | 3879 | vmx->guest_msrs[j].mask = -1ull; |
a2fa3e9f | 3880 | ++vmx->nmsrs; |
6aa8b732 | 3881 | } |
6aa8b732 | 3882 | |
1c3d14fe | 3883 | vmcs_write32(VM_EXIT_CONTROLS, vmcs_config.vmexit_ctrl); |
6aa8b732 AK |
3884 | |
3885 | /* 22.2.1, 20.8.1 */ | |
1c3d14fe YS |
3886 | vmcs_write32(VM_ENTRY_CONTROLS, vmcs_config.vmentry_ctrl); |
3887 | ||
e00c8cf2 | 3888 | vmcs_writel(CR0_GUEST_HOST_MASK, ~0UL); |
bf8179a0 | 3889 | set_cr4_guest_host_mask(vmx); |
e00c8cf2 | 3890 | |
99e3e30a | 3891 | kvm_write_tsc(&vmx->vcpu, 0); |
f78e0e2e | 3892 | |
e00c8cf2 AK |
3893 | return 0; |
3894 | } | |
3895 | ||
3896 | static int vmx_vcpu_reset(struct kvm_vcpu *vcpu) | |
3897 | { | |
3898 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
3899 | u64 msr; | |
4b9d3a04 | 3900 | int ret; |
e00c8cf2 | 3901 | |
5fdbf976 | 3902 | vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP)); |
e00c8cf2 | 3903 | |
7ffd92c5 | 3904 | vmx->rmode.vm86_active = 0; |
e00c8cf2 | 3905 | |
3b86cd99 JK |
3906 | vmx->soft_vnmi_blocked = 0; |
3907 | ||
ad312c7c | 3908 | vmx->vcpu.arch.regs[VCPU_REGS_RDX] = get_rdx_init_val(); |
2d3ad1f4 | 3909 | kvm_set_cr8(&vmx->vcpu, 0); |
e00c8cf2 | 3910 | msr = 0xfee00000 | MSR_IA32_APICBASE_ENABLE; |
c5af89b6 | 3911 | if (kvm_vcpu_is_bsp(&vmx->vcpu)) |
e00c8cf2 AK |
3912 | msr |= MSR_IA32_APICBASE_BSP; |
3913 | kvm_set_apic_base(&vmx->vcpu, msr); | |
3914 | ||
10ab25cd JK |
3915 | ret = fx_init(&vmx->vcpu); |
3916 | if (ret != 0) | |
3917 | goto out; | |
e00c8cf2 | 3918 | |
2fb92db1 AK |
3919 | vmx_segment_cache_clear(vmx); |
3920 | ||
5706be0d | 3921 | seg_setup(VCPU_SREG_CS); |
e00c8cf2 AK |
3922 | /* |
3923 | * GUEST_CS_BASE should really be 0xffff0000, but VT vm86 mode | |
3924 | * insists on having GUEST_CS_BASE == GUEST_CS_SELECTOR << 4. Sigh. | |
3925 | */ | |
c5af89b6 | 3926 | if (kvm_vcpu_is_bsp(&vmx->vcpu)) { |
e00c8cf2 AK |
3927 | vmcs_write16(GUEST_CS_SELECTOR, 0xf000); |
3928 | vmcs_writel(GUEST_CS_BASE, 0x000f0000); | |
3929 | } else { | |
ad312c7c ZX |
3930 | vmcs_write16(GUEST_CS_SELECTOR, vmx->vcpu.arch.sipi_vector << 8); |
3931 | vmcs_writel(GUEST_CS_BASE, vmx->vcpu.arch.sipi_vector << 12); | |
e00c8cf2 | 3932 | } |
e00c8cf2 AK |
3933 | |
3934 | seg_setup(VCPU_SREG_DS); | |
3935 | seg_setup(VCPU_SREG_ES); | |
3936 | seg_setup(VCPU_SREG_FS); | |
3937 | seg_setup(VCPU_SREG_GS); | |
3938 | seg_setup(VCPU_SREG_SS); | |
3939 | ||
3940 | vmcs_write16(GUEST_TR_SELECTOR, 0); | |
3941 | vmcs_writel(GUEST_TR_BASE, 0); | |
3942 | vmcs_write32(GUEST_TR_LIMIT, 0xffff); | |
3943 | vmcs_write32(GUEST_TR_AR_BYTES, 0x008b); | |
3944 | ||
3945 | vmcs_write16(GUEST_LDTR_SELECTOR, 0); | |
3946 | vmcs_writel(GUEST_LDTR_BASE, 0); | |
3947 | vmcs_write32(GUEST_LDTR_LIMIT, 0xffff); | |
3948 | vmcs_write32(GUEST_LDTR_AR_BYTES, 0x00082); | |
3949 | ||
3950 | vmcs_write32(GUEST_SYSENTER_CS, 0); | |
3951 | vmcs_writel(GUEST_SYSENTER_ESP, 0); | |
3952 | vmcs_writel(GUEST_SYSENTER_EIP, 0); | |
3953 | ||
3954 | vmcs_writel(GUEST_RFLAGS, 0x02); | |
c5af89b6 | 3955 | if (kvm_vcpu_is_bsp(&vmx->vcpu)) |
5fdbf976 | 3956 | kvm_rip_write(vcpu, 0xfff0); |
e00c8cf2 | 3957 | else |
5fdbf976 MT |
3958 | kvm_rip_write(vcpu, 0); |
3959 | kvm_register_write(vcpu, VCPU_REGS_RSP, 0); | |
e00c8cf2 | 3960 | |
e00c8cf2 AK |
3961 | vmcs_writel(GUEST_DR7, 0x400); |
3962 | ||
3963 | vmcs_writel(GUEST_GDTR_BASE, 0); | |
3964 | vmcs_write32(GUEST_GDTR_LIMIT, 0xffff); | |
3965 | ||
3966 | vmcs_writel(GUEST_IDTR_BASE, 0); | |
3967 | vmcs_write32(GUEST_IDTR_LIMIT, 0xffff); | |
3968 | ||
443381a8 | 3969 | vmcs_write32(GUEST_ACTIVITY_STATE, GUEST_ACTIVITY_ACTIVE); |
e00c8cf2 AK |
3970 | vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, 0); |
3971 | vmcs_write32(GUEST_PENDING_DBG_EXCEPTIONS, 0); | |
3972 | ||
e00c8cf2 AK |
3973 | /* Special registers */ |
3974 | vmcs_write64(GUEST_IA32_DEBUGCTL, 0); | |
3975 | ||
3976 | setup_msrs(vmx); | |
3977 | ||
6aa8b732 AK |
3978 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); /* 22.2.1 */ |
3979 | ||
f78e0e2e SY |
3980 | if (cpu_has_vmx_tpr_shadow()) { |
3981 | vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, 0); | |
3982 | if (vm_need_tpr_shadow(vmx->vcpu.kvm)) | |
3983 | vmcs_write64(VIRTUAL_APIC_PAGE_ADDR, | |
afc20184 | 3984 | __pa(vmx->vcpu.arch.apic->regs)); |
f78e0e2e SY |
3985 | vmcs_write32(TPR_THRESHOLD, 0); |
3986 | } | |
3987 | ||
3988 | if (vm_need_virtualize_apic_accesses(vmx->vcpu.kvm)) | |
3989 | vmcs_write64(APIC_ACCESS_ADDR, | |
bfc6d222 | 3990 | page_to_phys(vmx->vcpu.kvm->arch.apic_access_page)); |
6aa8b732 | 3991 | |
2384d2b3 SY |
3992 | if (vmx->vpid != 0) |
3993 | vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->vpid); | |
3994 | ||
fa40052c | 3995 | vmx->vcpu.arch.cr0 = X86_CR0_NW | X86_CR0_CD | X86_CR0_ET; |
7a4f5ad0 | 3996 | vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu); |
4d4ec087 | 3997 | vmx_set_cr0(&vmx->vcpu, kvm_read_cr0(vcpu)); /* enter rmode */ |
7a4f5ad0 | 3998 | srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx); |
8b9cf98c | 3999 | vmx_set_cr4(&vmx->vcpu, 0); |
8b9cf98c | 4000 | vmx_set_efer(&vmx->vcpu, 0); |
8b9cf98c RR |
4001 | vmx_fpu_activate(&vmx->vcpu); |
4002 | update_exception_bitmap(&vmx->vcpu); | |
6aa8b732 | 4003 | |
b9d762fa | 4004 | vpid_sync_context(vmx); |
2384d2b3 | 4005 | |
3200f405 | 4006 | ret = 0; |
6aa8b732 | 4007 | |
a89a8fb9 MG |
4008 | /* HACK: Don't enable emulation on guest boot/reset */ |
4009 | vmx->emulation_required = 0; | |
4010 | ||
6aa8b732 AK |
4011 | out: |
4012 | return ret; | |
4013 | } | |
4014 | ||
b6f1250e NHE |
4015 | /* |
4016 | * In nested virtualization, check if L1 asked to exit on external interrupts. | |
4017 | * For most existing hypervisors, this will always return true. | |
4018 | */ | |
4019 | static bool nested_exit_on_intr(struct kvm_vcpu *vcpu) | |
4020 | { | |
4021 | return get_vmcs12(vcpu)->pin_based_vm_exec_control & | |
4022 | PIN_BASED_EXT_INTR_MASK; | |
4023 | } | |
4024 | ||
3b86cd99 JK |
4025 | static void enable_irq_window(struct kvm_vcpu *vcpu) |
4026 | { | |
4027 | u32 cpu_based_vm_exec_control; | |
d6185f20 NHE |
4028 | if (is_guest_mode(vcpu) && nested_exit_on_intr(vcpu)) { |
4029 | /* | |
4030 | * We get here if vmx_interrupt_allowed() said we can't | |
4031 | * inject to L1 now because L2 must run. Ask L2 to exit | |
4032 | * right after entry, so we can inject to L1 more promptly. | |
b6f1250e | 4033 | */ |
d6185f20 | 4034 | kvm_make_request(KVM_REQ_IMMEDIATE_EXIT, vcpu); |
b6f1250e | 4035 | return; |
d6185f20 | 4036 | } |
3b86cd99 JK |
4037 | |
4038 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
4039 | cpu_based_vm_exec_control |= CPU_BASED_VIRTUAL_INTR_PENDING; | |
4040 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
4041 | } | |
4042 | ||
4043 | static void enable_nmi_window(struct kvm_vcpu *vcpu) | |
4044 | { | |
4045 | u32 cpu_based_vm_exec_control; | |
4046 | ||
4047 | if (!cpu_has_virtual_nmis()) { | |
4048 | enable_irq_window(vcpu); | |
4049 | return; | |
4050 | } | |
4051 | ||
30bd0c4c AK |
4052 | if (vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_STI) { |
4053 | enable_irq_window(vcpu); | |
4054 | return; | |
4055 | } | |
3b86cd99 JK |
4056 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); |
4057 | cpu_based_vm_exec_control |= CPU_BASED_VIRTUAL_NMI_PENDING; | |
4058 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
4059 | } | |
4060 | ||
66fd3f7f | 4061 | static void vmx_inject_irq(struct kvm_vcpu *vcpu) |
85f455f7 | 4062 | { |
9c8cba37 | 4063 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
66fd3f7f GN |
4064 | uint32_t intr; |
4065 | int irq = vcpu->arch.interrupt.nr; | |
9c8cba37 | 4066 | |
229456fc | 4067 | trace_kvm_inj_virq(irq); |
2714d1d3 | 4068 | |
fa89a817 | 4069 | ++vcpu->stat.irq_injections; |
7ffd92c5 | 4070 | if (vmx->rmode.vm86_active) { |
71f9833b SH |
4071 | int inc_eip = 0; |
4072 | if (vcpu->arch.interrupt.soft) | |
4073 | inc_eip = vcpu->arch.event_exit_inst_len; | |
4074 | if (kvm_inject_realmode_interrupt(vcpu, irq, inc_eip) != EMULATE_DONE) | |
a92601bb | 4075 | kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu); |
85f455f7 ED |
4076 | return; |
4077 | } | |
66fd3f7f GN |
4078 | intr = irq | INTR_INFO_VALID_MASK; |
4079 | if (vcpu->arch.interrupt.soft) { | |
4080 | intr |= INTR_TYPE_SOFT_INTR; | |
4081 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, | |
4082 | vmx->vcpu.arch.event_exit_inst_len); | |
4083 | } else | |
4084 | intr |= INTR_TYPE_EXT_INTR; | |
4085 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, intr); | |
85f455f7 ED |
4086 | } |
4087 | ||
f08864b4 SY |
4088 | static void vmx_inject_nmi(struct kvm_vcpu *vcpu) |
4089 | { | |
66a5a347 JK |
4090 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
4091 | ||
0b6ac343 NHE |
4092 | if (is_guest_mode(vcpu)) |
4093 | return; | |
4094 | ||
3b86cd99 JK |
4095 | if (!cpu_has_virtual_nmis()) { |
4096 | /* | |
4097 | * Tracking the NMI-blocked state in software is built upon | |
4098 | * finding the next open IRQ window. This, in turn, depends on | |
4099 | * well-behaving guests: They have to keep IRQs disabled at | |
4100 | * least as long as the NMI handler runs. Otherwise we may | |
4101 | * cause NMI nesting, maybe breaking the guest. But as this is | |
4102 | * highly unlikely, we can live with the residual risk. | |
4103 | */ | |
4104 | vmx->soft_vnmi_blocked = 1; | |
4105 | vmx->vnmi_blocked_time = 0; | |
4106 | } | |
4107 | ||
487b391d | 4108 | ++vcpu->stat.nmi_injections; |
9d58b931 | 4109 | vmx->nmi_known_unmasked = false; |
7ffd92c5 | 4110 | if (vmx->rmode.vm86_active) { |
71f9833b | 4111 | if (kvm_inject_realmode_interrupt(vcpu, NMI_VECTOR, 0) != EMULATE_DONE) |
a92601bb | 4112 | kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu); |
66a5a347 JK |
4113 | return; |
4114 | } | |
f08864b4 SY |
4115 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, |
4116 | INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK | NMI_VECTOR); | |
f08864b4 SY |
4117 | } |
4118 | ||
c4282df9 | 4119 | static int vmx_nmi_allowed(struct kvm_vcpu *vcpu) |
33f089ca | 4120 | { |
3b86cd99 | 4121 | if (!cpu_has_virtual_nmis() && to_vmx(vcpu)->soft_vnmi_blocked) |
c4282df9 | 4122 | return 0; |
33f089ca | 4123 | |
c4282df9 | 4124 | return !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & |
30bd0c4c AK |
4125 | (GUEST_INTR_STATE_MOV_SS | GUEST_INTR_STATE_STI |
4126 | | GUEST_INTR_STATE_NMI)); | |
33f089ca JK |
4127 | } |
4128 | ||
3cfc3092 JK |
4129 | static bool vmx_get_nmi_mask(struct kvm_vcpu *vcpu) |
4130 | { | |
4131 | if (!cpu_has_virtual_nmis()) | |
4132 | return to_vmx(vcpu)->soft_vnmi_blocked; | |
9d58b931 AK |
4133 | if (to_vmx(vcpu)->nmi_known_unmasked) |
4134 | return false; | |
c332c83a | 4135 | return vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & GUEST_INTR_STATE_NMI; |
3cfc3092 JK |
4136 | } |
4137 | ||
4138 | static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked) | |
4139 | { | |
4140 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
4141 | ||
4142 | if (!cpu_has_virtual_nmis()) { | |
4143 | if (vmx->soft_vnmi_blocked != masked) { | |
4144 | vmx->soft_vnmi_blocked = masked; | |
4145 | vmx->vnmi_blocked_time = 0; | |
4146 | } | |
4147 | } else { | |
9d58b931 | 4148 | vmx->nmi_known_unmasked = !masked; |
3cfc3092 JK |
4149 | if (masked) |
4150 | vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, | |
4151 | GUEST_INTR_STATE_NMI); | |
4152 | else | |
4153 | vmcs_clear_bits(GUEST_INTERRUPTIBILITY_INFO, | |
4154 | GUEST_INTR_STATE_NMI); | |
4155 | } | |
4156 | } | |
4157 | ||
78646121 GN |
4158 | static int vmx_interrupt_allowed(struct kvm_vcpu *vcpu) |
4159 | { | |
b6f1250e | 4160 | if (is_guest_mode(vcpu) && nested_exit_on_intr(vcpu)) { |
51cfe38e NHE |
4161 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); |
4162 | if (to_vmx(vcpu)->nested.nested_run_pending || | |
4163 | (vmcs12->idt_vectoring_info_field & | |
4164 | VECTORING_INFO_VALID_MASK)) | |
b6f1250e NHE |
4165 | return 0; |
4166 | nested_vmx_vmexit(vcpu); | |
b6f1250e NHE |
4167 | vmcs12->vm_exit_reason = EXIT_REASON_EXTERNAL_INTERRUPT; |
4168 | vmcs12->vm_exit_intr_info = 0; | |
4169 | /* fall through to normal code, but now in L1, not L2 */ | |
4170 | } | |
4171 | ||
c4282df9 GN |
4172 | return (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF) && |
4173 | !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & | |
4174 | (GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS)); | |
78646121 GN |
4175 | } |
4176 | ||
cbc94022 IE |
4177 | static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr) |
4178 | { | |
4179 | int ret; | |
4180 | struct kvm_userspace_memory_region tss_mem = { | |
6fe63979 | 4181 | .slot = TSS_PRIVATE_MEMSLOT, |
cbc94022 IE |
4182 | .guest_phys_addr = addr, |
4183 | .memory_size = PAGE_SIZE * 3, | |
4184 | .flags = 0, | |
4185 | }; | |
4186 | ||
4187 | ret = kvm_set_memory_region(kvm, &tss_mem, 0); | |
4188 | if (ret) | |
4189 | return ret; | |
bfc6d222 | 4190 | kvm->arch.tss_addr = addr; |
93ea5388 GN |
4191 | if (!init_rmode_tss(kvm)) |
4192 | return -ENOMEM; | |
4193 | ||
cbc94022 IE |
4194 | return 0; |
4195 | } | |
4196 | ||
6aa8b732 AK |
4197 | static int handle_rmode_exception(struct kvm_vcpu *vcpu, |
4198 | int vec, u32 err_code) | |
4199 | { | |
b3f37707 NK |
4200 | /* |
4201 | * Instruction with address size override prefix opcode 0x67 | |
4202 | * Cause the #SS fault with 0 error code in VM86 mode. | |
4203 | */ | |
4204 | if (((vec == GP_VECTOR) || (vec == SS_VECTOR)) && err_code == 0) | |
51d8b661 | 4205 | if (emulate_instruction(vcpu, 0) == EMULATE_DONE) |
6aa8b732 | 4206 | return 1; |
77ab6db0 JK |
4207 | /* |
4208 | * Forward all other exceptions that are valid in real mode. | |
4209 | * FIXME: Breaks guest debugging in real mode, needs to be fixed with | |
4210 | * the required debugging infrastructure rework. | |
4211 | */ | |
4212 | switch (vec) { | |
77ab6db0 | 4213 | case DB_VECTOR: |
d0bfb940 JK |
4214 | if (vcpu->guest_debug & |
4215 | (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP)) | |
4216 | return 0; | |
4217 | kvm_queue_exception(vcpu, vec); | |
4218 | return 1; | |
77ab6db0 | 4219 | case BP_VECTOR: |
c573cd22 JK |
4220 | /* |
4221 | * Update instruction length as we may reinject the exception | |
4222 | * from user space while in guest debugging mode. | |
4223 | */ | |
4224 | to_vmx(vcpu)->vcpu.arch.event_exit_inst_len = | |
4225 | vmcs_read32(VM_EXIT_INSTRUCTION_LEN); | |
d0bfb940 JK |
4226 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_SW_BP) |
4227 | return 0; | |
4228 | /* fall through */ | |
4229 | case DE_VECTOR: | |
77ab6db0 JK |
4230 | case OF_VECTOR: |
4231 | case BR_VECTOR: | |
4232 | case UD_VECTOR: | |
4233 | case DF_VECTOR: | |
4234 | case SS_VECTOR: | |
4235 | case GP_VECTOR: | |
4236 | case MF_VECTOR: | |
4237 | kvm_queue_exception(vcpu, vec); | |
4238 | return 1; | |
4239 | } | |
6aa8b732 AK |
4240 | return 0; |
4241 | } | |
4242 | ||
a0861c02 AK |
4243 | /* |
4244 | * Trigger machine check on the host. We assume all the MSRs are already set up | |
4245 | * by the CPU and that we still run on the same CPU as the MCE occurred on. | |
4246 | * We pass a fake environment to the machine check handler because we want | |
4247 | * the guest to be always treated like user space, no matter what context | |
4248 | * it used internally. | |
4249 | */ | |
4250 | static void kvm_machine_check(void) | |
4251 | { | |
4252 | #if defined(CONFIG_X86_MCE) && defined(CONFIG_X86_64) | |
4253 | struct pt_regs regs = { | |
4254 | .cs = 3, /* Fake ring 3 no matter what the guest ran on */ | |
4255 | .flags = X86_EFLAGS_IF, | |
4256 | }; | |
4257 | ||
4258 | do_machine_check(®s, 0); | |
4259 | #endif | |
4260 | } | |
4261 | ||
851ba692 | 4262 | static int handle_machine_check(struct kvm_vcpu *vcpu) |
a0861c02 AK |
4263 | { |
4264 | /* already handled by vcpu_run */ | |
4265 | return 1; | |
4266 | } | |
4267 | ||
851ba692 | 4268 | static int handle_exception(struct kvm_vcpu *vcpu) |
6aa8b732 | 4269 | { |
1155f76a | 4270 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
851ba692 | 4271 | struct kvm_run *kvm_run = vcpu->run; |
d0bfb940 | 4272 | u32 intr_info, ex_no, error_code; |
42dbaa5a | 4273 | unsigned long cr2, rip, dr6; |
6aa8b732 AK |
4274 | u32 vect_info; |
4275 | enum emulation_result er; | |
4276 | ||
1155f76a | 4277 | vect_info = vmx->idt_vectoring_info; |
88786475 | 4278 | intr_info = vmx->exit_intr_info; |
6aa8b732 | 4279 | |
a0861c02 | 4280 | if (is_machine_check(intr_info)) |
851ba692 | 4281 | return handle_machine_check(vcpu); |
a0861c02 | 4282 | |
6aa8b732 | 4283 | if ((vect_info & VECTORING_INFO_VALID_MASK) && |
65ac7264 AK |
4284 | !is_page_fault(intr_info)) { |
4285 | vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; | |
4286 | vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_SIMUL_EX; | |
4287 | vcpu->run->internal.ndata = 2; | |
4288 | vcpu->run->internal.data[0] = vect_info; | |
4289 | vcpu->run->internal.data[1] = intr_info; | |
4290 | return 0; | |
4291 | } | |
6aa8b732 | 4292 | |
e4a41889 | 4293 | if ((intr_info & INTR_INFO_INTR_TYPE_MASK) == INTR_TYPE_NMI_INTR) |
1b6269db | 4294 | return 1; /* already handled by vmx_vcpu_run() */ |
2ab455cc AL |
4295 | |
4296 | if (is_no_device(intr_info)) { | |
5fd86fcf | 4297 | vmx_fpu_activate(vcpu); |
2ab455cc AL |
4298 | return 1; |
4299 | } | |
4300 | ||
7aa81cc0 | 4301 | if (is_invalid_opcode(intr_info)) { |
51d8b661 | 4302 | er = emulate_instruction(vcpu, EMULTYPE_TRAP_UD); |
7aa81cc0 | 4303 | if (er != EMULATE_DONE) |
7ee5d940 | 4304 | kvm_queue_exception(vcpu, UD_VECTOR); |
7aa81cc0 AL |
4305 | return 1; |
4306 | } | |
4307 | ||
6aa8b732 | 4308 | error_code = 0; |
2e11384c | 4309 | if (intr_info & INTR_INFO_DELIVER_CODE_MASK) |
6aa8b732 AK |
4310 | error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE); |
4311 | if (is_page_fault(intr_info)) { | |
1439442c | 4312 | /* EPT won't cause page fault directly */ |
cf3ace79 | 4313 | BUG_ON(enable_ept); |
6aa8b732 | 4314 | cr2 = vmcs_readl(EXIT_QUALIFICATION); |
229456fc MT |
4315 | trace_kvm_page_fault(cr2, error_code); |
4316 | ||
3298b75c | 4317 | if (kvm_event_needs_reinjection(vcpu)) |
577bdc49 | 4318 | kvm_mmu_unprotect_page_virt(vcpu, cr2); |
dc25e89e | 4319 | return kvm_mmu_page_fault(vcpu, cr2, error_code, NULL, 0); |
6aa8b732 AK |
4320 | } |
4321 | ||
7ffd92c5 | 4322 | if (vmx->rmode.vm86_active && |
6aa8b732 | 4323 | handle_rmode_exception(vcpu, intr_info & INTR_INFO_VECTOR_MASK, |
72d6e5a0 | 4324 | error_code)) { |
ad312c7c ZX |
4325 | if (vcpu->arch.halt_request) { |
4326 | vcpu->arch.halt_request = 0; | |
72d6e5a0 AK |
4327 | return kvm_emulate_halt(vcpu); |
4328 | } | |
6aa8b732 | 4329 | return 1; |
72d6e5a0 | 4330 | } |
6aa8b732 | 4331 | |
d0bfb940 | 4332 | ex_no = intr_info & INTR_INFO_VECTOR_MASK; |
42dbaa5a JK |
4333 | switch (ex_no) { |
4334 | case DB_VECTOR: | |
4335 | dr6 = vmcs_readl(EXIT_QUALIFICATION); | |
4336 | if (!(vcpu->guest_debug & | |
4337 | (KVM_GUESTDBG_SINGLESTEP | KVM_GUESTDBG_USE_HW_BP))) { | |
4338 | vcpu->arch.dr6 = dr6 | DR6_FIXED_1; | |
4339 | kvm_queue_exception(vcpu, DB_VECTOR); | |
4340 | return 1; | |
4341 | } | |
4342 | kvm_run->debug.arch.dr6 = dr6 | DR6_FIXED_1; | |
4343 | kvm_run->debug.arch.dr7 = vmcs_readl(GUEST_DR7); | |
4344 | /* fall through */ | |
4345 | case BP_VECTOR: | |
c573cd22 JK |
4346 | /* |
4347 | * Update instruction length as we may reinject #BP from | |
4348 | * user space while in guest debugging mode. Reading it for | |
4349 | * #DB as well causes no harm, it is not used in that case. | |
4350 | */ | |
4351 | vmx->vcpu.arch.event_exit_inst_len = | |
4352 | vmcs_read32(VM_EXIT_INSTRUCTION_LEN); | |
6aa8b732 | 4353 | kvm_run->exit_reason = KVM_EXIT_DEBUG; |
0a434bb2 | 4354 | rip = kvm_rip_read(vcpu); |
d0bfb940 JK |
4355 | kvm_run->debug.arch.pc = vmcs_readl(GUEST_CS_BASE) + rip; |
4356 | kvm_run->debug.arch.exception = ex_no; | |
42dbaa5a JK |
4357 | break; |
4358 | default: | |
d0bfb940 JK |
4359 | kvm_run->exit_reason = KVM_EXIT_EXCEPTION; |
4360 | kvm_run->ex.exception = ex_no; | |
4361 | kvm_run->ex.error_code = error_code; | |
42dbaa5a | 4362 | break; |
6aa8b732 | 4363 | } |
6aa8b732 AK |
4364 | return 0; |
4365 | } | |
4366 | ||
851ba692 | 4367 | static int handle_external_interrupt(struct kvm_vcpu *vcpu) |
6aa8b732 | 4368 | { |
1165f5fe | 4369 | ++vcpu->stat.irq_exits; |
6aa8b732 AK |
4370 | return 1; |
4371 | } | |
4372 | ||
851ba692 | 4373 | static int handle_triple_fault(struct kvm_vcpu *vcpu) |
988ad74f | 4374 | { |
851ba692 | 4375 | vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; |
988ad74f AK |
4376 | return 0; |
4377 | } | |
6aa8b732 | 4378 | |
851ba692 | 4379 | static int handle_io(struct kvm_vcpu *vcpu) |
6aa8b732 | 4380 | { |
bfdaab09 | 4381 | unsigned long exit_qualification; |
34c33d16 | 4382 | int size, in, string; |
039576c0 | 4383 | unsigned port; |
6aa8b732 | 4384 | |
bfdaab09 | 4385 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
039576c0 | 4386 | string = (exit_qualification & 16) != 0; |
cf8f70bf | 4387 | in = (exit_qualification & 8) != 0; |
e70669ab | 4388 | |
cf8f70bf | 4389 | ++vcpu->stat.io_exits; |
e70669ab | 4390 | |
cf8f70bf | 4391 | if (string || in) |
51d8b661 | 4392 | return emulate_instruction(vcpu, 0) == EMULATE_DONE; |
e70669ab | 4393 | |
cf8f70bf GN |
4394 | port = exit_qualification >> 16; |
4395 | size = (exit_qualification & 7) + 1; | |
e93f36bc | 4396 | skip_emulated_instruction(vcpu); |
cf8f70bf GN |
4397 | |
4398 | return kvm_fast_pio_out(vcpu, size, port); | |
6aa8b732 AK |
4399 | } |
4400 | ||
102d8325 IM |
4401 | static void |
4402 | vmx_patch_hypercall(struct kvm_vcpu *vcpu, unsigned char *hypercall) | |
4403 | { | |
4404 | /* | |
4405 | * Patch in the VMCALL instruction: | |
4406 | */ | |
4407 | hypercall[0] = 0x0f; | |
4408 | hypercall[1] = 0x01; | |
4409 | hypercall[2] = 0xc1; | |
102d8325 IM |
4410 | } |
4411 | ||
0fa06071 | 4412 | /* called to set cr0 as appropriate for a mov-to-cr0 exit. */ |
eeadf9e7 NHE |
4413 | static int handle_set_cr0(struct kvm_vcpu *vcpu, unsigned long val) |
4414 | { | |
4415 | if (to_vmx(vcpu)->nested.vmxon && | |
4416 | ((val & VMXON_CR0_ALWAYSON) != VMXON_CR0_ALWAYSON)) | |
4417 | return 1; | |
4418 | ||
4419 | if (is_guest_mode(vcpu)) { | |
4420 | /* | |
4421 | * We get here when L2 changed cr0 in a way that did not change | |
4422 | * any of L1's shadowed bits (see nested_vmx_exit_handled_cr), | |
4423 | * but did change L0 shadowed bits. This can currently happen | |
4424 | * with the TS bit: L0 may want to leave TS on (for lazy fpu | |
4425 | * loading) while pretending to allow the guest to change it. | |
4426 | */ | |
4427 | if (kvm_set_cr0(vcpu, (val & vcpu->arch.cr0_guest_owned_bits) | | |
4428 | (vcpu->arch.cr0 & ~vcpu->arch.cr0_guest_owned_bits))) | |
4429 | return 1; | |
4430 | vmcs_writel(CR0_READ_SHADOW, val); | |
4431 | return 0; | |
4432 | } else | |
4433 | return kvm_set_cr0(vcpu, val); | |
4434 | } | |
4435 | ||
4436 | static int handle_set_cr4(struct kvm_vcpu *vcpu, unsigned long val) | |
4437 | { | |
4438 | if (is_guest_mode(vcpu)) { | |
4439 | if (kvm_set_cr4(vcpu, (val & vcpu->arch.cr4_guest_owned_bits) | | |
4440 | (vcpu->arch.cr4 & ~vcpu->arch.cr4_guest_owned_bits))) | |
4441 | return 1; | |
4442 | vmcs_writel(CR4_READ_SHADOW, val); | |
4443 | return 0; | |
4444 | } else | |
4445 | return kvm_set_cr4(vcpu, val); | |
4446 | } | |
4447 | ||
4448 | /* called to set cr0 as approriate for clts instruction exit. */ | |
4449 | static void handle_clts(struct kvm_vcpu *vcpu) | |
4450 | { | |
4451 | if (is_guest_mode(vcpu)) { | |
4452 | /* | |
4453 | * We get here when L2 did CLTS, and L1 didn't shadow CR0.TS | |
4454 | * but we did (!fpu_active). We need to keep GUEST_CR0.TS on, | |
4455 | * just pretend it's off (also in arch.cr0 for fpu_activate). | |
4456 | */ | |
4457 | vmcs_writel(CR0_READ_SHADOW, | |
4458 | vmcs_readl(CR0_READ_SHADOW) & ~X86_CR0_TS); | |
4459 | vcpu->arch.cr0 &= ~X86_CR0_TS; | |
4460 | } else | |
4461 | vmx_set_cr0(vcpu, kvm_read_cr0_bits(vcpu, ~X86_CR0_TS)); | |
4462 | } | |
4463 | ||
851ba692 | 4464 | static int handle_cr(struct kvm_vcpu *vcpu) |
6aa8b732 | 4465 | { |
229456fc | 4466 | unsigned long exit_qualification, val; |
6aa8b732 AK |
4467 | int cr; |
4468 | int reg; | |
49a9b07e | 4469 | int err; |
6aa8b732 | 4470 | |
bfdaab09 | 4471 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
6aa8b732 AK |
4472 | cr = exit_qualification & 15; |
4473 | reg = (exit_qualification >> 8) & 15; | |
4474 | switch ((exit_qualification >> 4) & 3) { | |
4475 | case 0: /* mov to cr */ | |
229456fc MT |
4476 | val = kvm_register_read(vcpu, reg); |
4477 | trace_kvm_cr_write(cr, val); | |
6aa8b732 AK |
4478 | switch (cr) { |
4479 | case 0: | |
eeadf9e7 | 4480 | err = handle_set_cr0(vcpu, val); |
db8fcefa | 4481 | kvm_complete_insn_gp(vcpu, err); |
6aa8b732 AK |
4482 | return 1; |
4483 | case 3: | |
2390218b | 4484 | err = kvm_set_cr3(vcpu, val); |
db8fcefa | 4485 | kvm_complete_insn_gp(vcpu, err); |
6aa8b732 AK |
4486 | return 1; |
4487 | case 4: | |
eeadf9e7 | 4488 | err = handle_set_cr4(vcpu, val); |
db8fcefa | 4489 | kvm_complete_insn_gp(vcpu, err); |
6aa8b732 | 4490 | return 1; |
0a5fff19 GN |
4491 | case 8: { |
4492 | u8 cr8_prev = kvm_get_cr8(vcpu); | |
4493 | u8 cr8 = kvm_register_read(vcpu, reg); | |
eea1cff9 | 4494 | err = kvm_set_cr8(vcpu, cr8); |
db8fcefa | 4495 | kvm_complete_insn_gp(vcpu, err); |
0a5fff19 GN |
4496 | if (irqchip_in_kernel(vcpu->kvm)) |
4497 | return 1; | |
4498 | if (cr8_prev <= cr8) | |
4499 | return 1; | |
851ba692 | 4500 | vcpu->run->exit_reason = KVM_EXIT_SET_TPR; |
0a5fff19 GN |
4501 | return 0; |
4502 | } | |
6aa8b732 AK |
4503 | }; |
4504 | break; | |
25c4c276 | 4505 | case 2: /* clts */ |
eeadf9e7 | 4506 | handle_clts(vcpu); |
4d4ec087 | 4507 | trace_kvm_cr_write(0, kvm_read_cr0(vcpu)); |
25c4c276 | 4508 | skip_emulated_instruction(vcpu); |
6b52d186 | 4509 | vmx_fpu_activate(vcpu); |
25c4c276 | 4510 | return 1; |
6aa8b732 AK |
4511 | case 1: /*mov from cr*/ |
4512 | switch (cr) { | |
4513 | case 3: | |
9f8fe504 AK |
4514 | val = kvm_read_cr3(vcpu); |
4515 | kvm_register_write(vcpu, reg, val); | |
4516 | trace_kvm_cr_read(cr, val); | |
6aa8b732 AK |
4517 | skip_emulated_instruction(vcpu); |
4518 | return 1; | |
4519 | case 8: | |
229456fc MT |
4520 | val = kvm_get_cr8(vcpu); |
4521 | kvm_register_write(vcpu, reg, val); | |
4522 | trace_kvm_cr_read(cr, val); | |
6aa8b732 AK |
4523 | skip_emulated_instruction(vcpu); |
4524 | return 1; | |
4525 | } | |
4526 | break; | |
4527 | case 3: /* lmsw */ | |
a1f83a74 | 4528 | val = (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f; |
4d4ec087 | 4529 | trace_kvm_cr_write(0, (kvm_read_cr0(vcpu) & ~0xful) | val); |
a1f83a74 | 4530 | kvm_lmsw(vcpu, val); |
6aa8b732 AK |
4531 | |
4532 | skip_emulated_instruction(vcpu); | |
4533 | return 1; | |
4534 | default: | |
4535 | break; | |
4536 | } | |
851ba692 | 4537 | vcpu->run->exit_reason = 0; |
a737f256 | 4538 | vcpu_unimpl(vcpu, "unhandled control register: op %d cr %d\n", |
6aa8b732 AK |
4539 | (int)(exit_qualification >> 4) & 3, cr); |
4540 | return 0; | |
4541 | } | |
4542 | ||
851ba692 | 4543 | static int handle_dr(struct kvm_vcpu *vcpu) |
6aa8b732 | 4544 | { |
bfdaab09 | 4545 | unsigned long exit_qualification; |
6aa8b732 AK |
4546 | int dr, reg; |
4547 | ||
f2483415 | 4548 | /* Do not handle if the CPL > 0, will trigger GP on re-entry */ |
0a79b009 AK |
4549 | if (!kvm_require_cpl(vcpu, 0)) |
4550 | return 1; | |
42dbaa5a JK |
4551 | dr = vmcs_readl(GUEST_DR7); |
4552 | if (dr & DR7_GD) { | |
4553 | /* | |
4554 | * As the vm-exit takes precedence over the debug trap, we | |
4555 | * need to emulate the latter, either for the host or the | |
4556 | * guest debugging itself. | |
4557 | */ | |
4558 | if (vcpu->guest_debug & KVM_GUESTDBG_USE_HW_BP) { | |
851ba692 AK |
4559 | vcpu->run->debug.arch.dr6 = vcpu->arch.dr6; |
4560 | vcpu->run->debug.arch.dr7 = dr; | |
4561 | vcpu->run->debug.arch.pc = | |
42dbaa5a JK |
4562 | vmcs_readl(GUEST_CS_BASE) + |
4563 | vmcs_readl(GUEST_RIP); | |
851ba692 AK |
4564 | vcpu->run->debug.arch.exception = DB_VECTOR; |
4565 | vcpu->run->exit_reason = KVM_EXIT_DEBUG; | |
42dbaa5a JK |
4566 | return 0; |
4567 | } else { | |
4568 | vcpu->arch.dr7 &= ~DR7_GD; | |
4569 | vcpu->arch.dr6 |= DR6_BD; | |
4570 | vmcs_writel(GUEST_DR7, vcpu->arch.dr7); | |
4571 | kvm_queue_exception(vcpu, DB_VECTOR); | |
4572 | return 1; | |
4573 | } | |
4574 | } | |
4575 | ||
bfdaab09 | 4576 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
42dbaa5a JK |
4577 | dr = exit_qualification & DEBUG_REG_ACCESS_NUM; |
4578 | reg = DEBUG_REG_ACCESS_REG(exit_qualification); | |
4579 | if (exit_qualification & TYPE_MOV_FROM_DR) { | |
020df079 GN |
4580 | unsigned long val; |
4581 | if (!kvm_get_dr(vcpu, dr, &val)) | |
4582 | kvm_register_write(vcpu, reg, val); | |
4583 | } else | |
4584 | kvm_set_dr(vcpu, dr, vcpu->arch.regs[reg]); | |
6aa8b732 AK |
4585 | skip_emulated_instruction(vcpu); |
4586 | return 1; | |
4587 | } | |
4588 | ||
020df079 GN |
4589 | static void vmx_set_dr7(struct kvm_vcpu *vcpu, unsigned long val) |
4590 | { | |
4591 | vmcs_writel(GUEST_DR7, val); | |
4592 | } | |
4593 | ||
851ba692 | 4594 | static int handle_cpuid(struct kvm_vcpu *vcpu) |
6aa8b732 | 4595 | { |
06465c5a AK |
4596 | kvm_emulate_cpuid(vcpu); |
4597 | return 1; | |
6aa8b732 AK |
4598 | } |
4599 | ||
851ba692 | 4600 | static int handle_rdmsr(struct kvm_vcpu *vcpu) |
6aa8b732 | 4601 | { |
ad312c7c | 4602 | u32 ecx = vcpu->arch.regs[VCPU_REGS_RCX]; |
6aa8b732 AK |
4603 | u64 data; |
4604 | ||
4605 | if (vmx_get_msr(vcpu, ecx, &data)) { | |
59200273 | 4606 | trace_kvm_msr_read_ex(ecx); |
c1a5d4f9 | 4607 | kvm_inject_gp(vcpu, 0); |
6aa8b732 AK |
4608 | return 1; |
4609 | } | |
4610 | ||
229456fc | 4611 | trace_kvm_msr_read(ecx, data); |
2714d1d3 | 4612 | |
6aa8b732 | 4613 | /* FIXME: handling of bits 32:63 of rax, rdx */ |
ad312c7c ZX |
4614 | vcpu->arch.regs[VCPU_REGS_RAX] = data & -1u; |
4615 | vcpu->arch.regs[VCPU_REGS_RDX] = (data >> 32) & -1u; | |
6aa8b732 AK |
4616 | skip_emulated_instruction(vcpu); |
4617 | return 1; | |
4618 | } | |
4619 | ||
851ba692 | 4620 | static int handle_wrmsr(struct kvm_vcpu *vcpu) |
6aa8b732 | 4621 | { |
ad312c7c ZX |
4622 | u32 ecx = vcpu->arch.regs[VCPU_REGS_RCX]; |
4623 | u64 data = (vcpu->arch.regs[VCPU_REGS_RAX] & -1u) | |
4624 | | ((u64)(vcpu->arch.regs[VCPU_REGS_RDX] & -1u) << 32); | |
6aa8b732 AK |
4625 | |
4626 | if (vmx_set_msr(vcpu, ecx, data) != 0) { | |
59200273 | 4627 | trace_kvm_msr_write_ex(ecx, data); |
c1a5d4f9 | 4628 | kvm_inject_gp(vcpu, 0); |
6aa8b732 AK |
4629 | return 1; |
4630 | } | |
4631 | ||
59200273 | 4632 | trace_kvm_msr_write(ecx, data); |
6aa8b732 AK |
4633 | skip_emulated_instruction(vcpu); |
4634 | return 1; | |
4635 | } | |
4636 | ||
851ba692 | 4637 | static int handle_tpr_below_threshold(struct kvm_vcpu *vcpu) |
6e5d865c | 4638 | { |
3842d135 | 4639 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
6e5d865c YS |
4640 | return 1; |
4641 | } | |
4642 | ||
851ba692 | 4643 | static int handle_interrupt_window(struct kvm_vcpu *vcpu) |
6aa8b732 | 4644 | { |
85f455f7 ED |
4645 | u32 cpu_based_vm_exec_control; |
4646 | ||
4647 | /* clear pending irq */ | |
4648 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
4649 | cpu_based_vm_exec_control &= ~CPU_BASED_VIRTUAL_INTR_PENDING; | |
4650 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
2714d1d3 | 4651 | |
3842d135 AK |
4652 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
4653 | ||
a26bf12a | 4654 | ++vcpu->stat.irq_window_exits; |
2714d1d3 | 4655 | |
c1150d8c DL |
4656 | /* |
4657 | * If the user space waits to inject interrupts, exit as soon as | |
4658 | * possible | |
4659 | */ | |
8061823a | 4660 | if (!irqchip_in_kernel(vcpu->kvm) && |
851ba692 | 4661 | vcpu->run->request_interrupt_window && |
8061823a | 4662 | !kvm_cpu_has_interrupt(vcpu)) { |
851ba692 | 4663 | vcpu->run->exit_reason = KVM_EXIT_IRQ_WINDOW_OPEN; |
c1150d8c DL |
4664 | return 0; |
4665 | } | |
6aa8b732 AK |
4666 | return 1; |
4667 | } | |
4668 | ||
851ba692 | 4669 | static int handle_halt(struct kvm_vcpu *vcpu) |
6aa8b732 AK |
4670 | { |
4671 | skip_emulated_instruction(vcpu); | |
d3bef15f | 4672 | return kvm_emulate_halt(vcpu); |
6aa8b732 AK |
4673 | } |
4674 | ||
851ba692 | 4675 | static int handle_vmcall(struct kvm_vcpu *vcpu) |
c21415e8 | 4676 | { |
510043da | 4677 | skip_emulated_instruction(vcpu); |
7aa81cc0 AL |
4678 | kvm_emulate_hypercall(vcpu); |
4679 | return 1; | |
c21415e8 IM |
4680 | } |
4681 | ||
ec25d5e6 GN |
4682 | static int handle_invd(struct kvm_vcpu *vcpu) |
4683 | { | |
51d8b661 | 4684 | return emulate_instruction(vcpu, 0) == EMULATE_DONE; |
ec25d5e6 GN |
4685 | } |
4686 | ||
851ba692 | 4687 | static int handle_invlpg(struct kvm_vcpu *vcpu) |
a7052897 | 4688 | { |
f9c617f6 | 4689 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
a7052897 MT |
4690 | |
4691 | kvm_mmu_invlpg(vcpu, exit_qualification); | |
4692 | skip_emulated_instruction(vcpu); | |
4693 | return 1; | |
4694 | } | |
4695 | ||
fee84b07 AK |
4696 | static int handle_rdpmc(struct kvm_vcpu *vcpu) |
4697 | { | |
4698 | int err; | |
4699 | ||
4700 | err = kvm_rdpmc(vcpu); | |
4701 | kvm_complete_insn_gp(vcpu, err); | |
4702 | ||
4703 | return 1; | |
4704 | } | |
4705 | ||
851ba692 | 4706 | static int handle_wbinvd(struct kvm_vcpu *vcpu) |
e5edaa01 ED |
4707 | { |
4708 | skip_emulated_instruction(vcpu); | |
f5f48ee1 | 4709 | kvm_emulate_wbinvd(vcpu); |
e5edaa01 ED |
4710 | return 1; |
4711 | } | |
4712 | ||
2acf923e DC |
4713 | static int handle_xsetbv(struct kvm_vcpu *vcpu) |
4714 | { | |
4715 | u64 new_bv = kvm_read_edx_eax(vcpu); | |
4716 | u32 index = kvm_register_read(vcpu, VCPU_REGS_RCX); | |
4717 | ||
4718 | if (kvm_set_xcr(vcpu, index, new_bv) == 0) | |
4719 | skip_emulated_instruction(vcpu); | |
4720 | return 1; | |
4721 | } | |
4722 | ||
851ba692 | 4723 | static int handle_apic_access(struct kvm_vcpu *vcpu) |
f78e0e2e | 4724 | { |
58fbbf26 KT |
4725 | if (likely(fasteoi)) { |
4726 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
4727 | int access_type, offset; | |
4728 | ||
4729 | access_type = exit_qualification & APIC_ACCESS_TYPE; | |
4730 | offset = exit_qualification & APIC_ACCESS_OFFSET; | |
4731 | /* | |
4732 | * Sane guest uses MOV to write EOI, with written value | |
4733 | * not cared. So make a short-circuit here by avoiding | |
4734 | * heavy instruction emulation. | |
4735 | */ | |
4736 | if ((access_type == TYPE_LINEAR_APIC_INST_WRITE) && | |
4737 | (offset == APIC_EOI)) { | |
4738 | kvm_lapic_set_eoi(vcpu); | |
4739 | skip_emulated_instruction(vcpu); | |
4740 | return 1; | |
4741 | } | |
4742 | } | |
51d8b661 | 4743 | return emulate_instruction(vcpu, 0) == EMULATE_DONE; |
f78e0e2e SY |
4744 | } |
4745 | ||
851ba692 | 4746 | static int handle_task_switch(struct kvm_vcpu *vcpu) |
37817f29 | 4747 | { |
60637aac | 4748 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
37817f29 | 4749 | unsigned long exit_qualification; |
e269fb21 JK |
4750 | bool has_error_code = false; |
4751 | u32 error_code = 0; | |
37817f29 | 4752 | u16 tss_selector; |
7f3d35fd | 4753 | int reason, type, idt_v, idt_index; |
64a7ec06 GN |
4754 | |
4755 | idt_v = (vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK); | |
7f3d35fd | 4756 | idt_index = (vmx->idt_vectoring_info & VECTORING_INFO_VECTOR_MASK); |
64a7ec06 | 4757 | type = (vmx->idt_vectoring_info & VECTORING_INFO_TYPE_MASK); |
37817f29 IE |
4758 | |
4759 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
4760 | ||
4761 | reason = (u32)exit_qualification >> 30; | |
64a7ec06 GN |
4762 | if (reason == TASK_SWITCH_GATE && idt_v) { |
4763 | switch (type) { | |
4764 | case INTR_TYPE_NMI_INTR: | |
4765 | vcpu->arch.nmi_injected = false; | |
654f06fc | 4766 | vmx_set_nmi_mask(vcpu, true); |
64a7ec06 GN |
4767 | break; |
4768 | case INTR_TYPE_EXT_INTR: | |
66fd3f7f | 4769 | case INTR_TYPE_SOFT_INTR: |
64a7ec06 GN |
4770 | kvm_clear_interrupt_queue(vcpu); |
4771 | break; | |
4772 | case INTR_TYPE_HARD_EXCEPTION: | |
e269fb21 JK |
4773 | if (vmx->idt_vectoring_info & |
4774 | VECTORING_INFO_DELIVER_CODE_MASK) { | |
4775 | has_error_code = true; | |
4776 | error_code = | |
4777 | vmcs_read32(IDT_VECTORING_ERROR_CODE); | |
4778 | } | |
4779 | /* fall through */ | |
64a7ec06 GN |
4780 | case INTR_TYPE_SOFT_EXCEPTION: |
4781 | kvm_clear_exception_queue(vcpu); | |
4782 | break; | |
4783 | default: | |
4784 | break; | |
4785 | } | |
60637aac | 4786 | } |
37817f29 IE |
4787 | tss_selector = exit_qualification; |
4788 | ||
64a7ec06 GN |
4789 | if (!idt_v || (type != INTR_TYPE_HARD_EXCEPTION && |
4790 | type != INTR_TYPE_EXT_INTR && | |
4791 | type != INTR_TYPE_NMI_INTR)) | |
4792 | skip_emulated_instruction(vcpu); | |
4793 | ||
7f3d35fd KW |
4794 | if (kvm_task_switch(vcpu, tss_selector, |
4795 | type == INTR_TYPE_SOFT_INTR ? idt_index : -1, reason, | |
4796 | has_error_code, error_code) == EMULATE_FAIL) { | |
acb54517 GN |
4797 | vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; |
4798 | vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION; | |
4799 | vcpu->run->internal.ndata = 0; | |
42dbaa5a | 4800 | return 0; |
acb54517 | 4801 | } |
42dbaa5a JK |
4802 | |
4803 | /* clear all local breakpoint enable flags */ | |
4804 | vmcs_writel(GUEST_DR7, vmcs_readl(GUEST_DR7) & ~55); | |
4805 | ||
4806 | /* | |
4807 | * TODO: What about debug traps on tss switch? | |
4808 | * Are we supposed to inject them and update dr6? | |
4809 | */ | |
4810 | ||
4811 | return 1; | |
37817f29 IE |
4812 | } |
4813 | ||
851ba692 | 4814 | static int handle_ept_violation(struct kvm_vcpu *vcpu) |
1439442c | 4815 | { |
f9c617f6 | 4816 | unsigned long exit_qualification; |
1439442c | 4817 | gpa_t gpa; |
4f5982a5 | 4818 | u32 error_code; |
1439442c | 4819 | int gla_validity; |
1439442c | 4820 | |
f9c617f6 | 4821 | exit_qualification = vmcs_readl(EXIT_QUALIFICATION); |
1439442c SY |
4822 | |
4823 | if (exit_qualification & (1 << 6)) { | |
4824 | printk(KERN_ERR "EPT: GPA exceeds GAW!\n"); | |
7f582ab6 | 4825 | return -EINVAL; |
1439442c SY |
4826 | } |
4827 | ||
4828 | gla_validity = (exit_qualification >> 7) & 0x3; | |
4829 | if (gla_validity != 0x3 && gla_validity != 0x1 && gla_validity != 0) { | |
4830 | printk(KERN_ERR "EPT: Handling EPT violation failed!\n"); | |
4831 | printk(KERN_ERR "EPT: GPA: 0x%lx, GVA: 0x%lx\n", | |
4832 | (long unsigned int)vmcs_read64(GUEST_PHYSICAL_ADDRESS), | |
f9c617f6 | 4833 | vmcs_readl(GUEST_LINEAR_ADDRESS)); |
1439442c SY |
4834 | printk(KERN_ERR "EPT: Exit qualification is 0x%lx\n", |
4835 | (long unsigned int)exit_qualification); | |
851ba692 AK |
4836 | vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; |
4837 | vcpu->run->hw.hardware_exit_reason = EXIT_REASON_EPT_VIOLATION; | |
596ae895 | 4838 | return 0; |
1439442c SY |
4839 | } |
4840 | ||
4841 | gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); | |
229456fc | 4842 | trace_kvm_page_fault(gpa, exit_qualification); |
4f5982a5 XG |
4843 | |
4844 | /* It is a write fault? */ | |
4845 | error_code = exit_qualification & (1U << 1); | |
4846 | /* ept page table is present? */ | |
4847 | error_code |= (exit_qualification >> 3) & 0x1; | |
4848 | ||
4849 | return kvm_mmu_page_fault(vcpu, gpa, error_code, NULL, 0); | |
1439442c SY |
4850 | } |
4851 | ||
68f89400 MT |
4852 | static u64 ept_rsvd_mask(u64 spte, int level) |
4853 | { | |
4854 | int i; | |
4855 | u64 mask = 0; | |
4856 | ||
4857 | for (i = 51; i > boot_cpu_data.x86_phys_bits; i--) | |
4858 | mask |= (1ULL << i); | |
4859 | ||
4860 | if (level > 2) | |
4861 | /* bits 7:3 reserved */ | |
4862 | mask |= 0xf8; | |
4863 | else if (level == 2) { | |
4864 | if (spte & (1ULL << 7)) | |
4865 | /* 2MB ref, bits 20:12 reserved */ | |
4866 | mask |= 0x1ff000; | |
4867 | else | |
4868 | /* bits 6:3 reserved */ | |
4869 | mask |= 0x78; | |
4870 | } | |
4871 | ||
4872 | return mask; | |
4873 | } | |
4874 | ||
4875 | static void ept_misconfig_inspect_spte(struct kvm_vcpu *vcpu, u64 spte, | |
4876 | int level) | |
4877 | { | |
4878 | printk(KERN_ERR "%s: spte 0x%llx level %d\n", __func__, spte, level); | |
4879 | ||
4880 | /* 010b (write-only) */ | |
4881 | WARN_ON((spte & 0x7) == 0x2); | |
4882 | ||
4883 | /* 110b (write/execute) */ | |
4884 | WARN_ON((spte & 0x7) == 0x6); | |
4885 | ||
4886 | /* 100b (execute-only) and value not supported by logical processor */ | |
4887 | if (!cpu_has_vmx_ept_execute_only()) | |
4888 | WARN_ON((spte & 0x7) == 0x4); | |
4889 | ||
4890 | /* not 000b */ | |
4891 | if ((spte & 0x7)) { | |
4892 | u64 rsvd_bits = spte & ept_rsvd_mask(spte, level); | |
4893 | ||
4894 | if (rsvd_bits != 0) { | |
4895 | printk(KERN_ERR "%s: rsvd_bits = 0x%llx\n", | |
4896 | __func__, rsvd_bits); | |
4897 | WARN_ON(1); | |
4898 | } | |
4899 | ||
4900 | if (level == 1 || (level == 2 && (spte & (1ULL << 7)))) { | |
4901 | u64 ept_mem_type = (spte & 0x38) >> 3; | |
4902 | ||
4903 | if (ept_mem_type == 2 || ept_mem_type == 3 || | |
4904 | ept_mem_type == 7) { | |
4905 | printk(KERN_ERR "%s: ept_mem_type=0x%llx\n", | |
4906 | __func__, ept_mem_type); | |
4907 | WARN_ON(1); | |
4908 | } | |
4909 | } | |
4910 | } | |
4911 | } | |
4912 | ||
851ba692 | 4913 | static int handle_ept_misconfig(struct kvm_vcpu *vcpu) |
68f89400 MT |
4914 | { |
4915 | u64 sptes[4]; | |
ce88decf | 4916 | int nr_sptes, i, ret; |
68f89400 MT |
4917 | gpa_t gpa; |
4918 | ||
4919 | gpa = vmcs_read64(GUEST_PHYSICAL_ADDRESS); | |
4920 | ||
ce88decf XG |
4921 | ret = handle_mmio_page_fault_common(vcpu, gpa, true); |
4922 | if (likely(ret == 1)) | |
4923 | return x86_emulate_instruction(vcpu, gpa, 0, NULL, 0) == | |
4924 | EMULATE_DONE; | |
4925 | if (unlikely(!ret)) | |
4926 | return 1; | |
4927 | ||
4928 | /* It is the real ept misconfig */ | |
68f89400 MT |
4929 | printk(KERN_ERR "EPT: Misconfiguration.\n"); |
4930 | printk(KERN_ERR "EPT: GPA: 0x%llx\n", gpa); | |
4931 | ||
4932 | nr_sptes = kvm_mmu_get_spte_hierarchy(vcpu, gpa, sptes); | |
4933 | ||
4934 | for (i = PT64_ROOT_LEVEL; i > PT64_ROOT_LEVEL - nr_sptes; --i) | |
4935 | ept_misconfig_inspect_spte(vcpu, sptes[i-1], i); | |
4936 | ||
851ba692 AK |
4937 | vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; |
4938 | vcpu->run->hw.hardware_exit_reason = EXIT_REASON_EPT_MISCONFIG; | |
68f89400 MT |
4939 | |
4940 | return 0; | |
4941 | } | |
4942 | ||
851ba692 | 4943 | static int handle_nmi_window(struct kvm_vcpu *vcpu) |
f08864b4 SY |
4944 | { |
4945 | u32 cpu_based_vm_exec_control; | |
4946 | ||
4947 | /* clear pending NMI */ | |
4948 | cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
4949 | cpu_based_vm_exec_control &= ~CPU_BASED_VIRTUAL_NMI_PENDING; | |
4950 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control); | |
4951 | ++vcpu->stat.nmi_window_exits; | |
3842d135 | 4952 | kvm_make_request(KVM_REQ_EVENT, vcpu); |
f08864b4 SY |
4953 | |
4954 | return 1; | |
4955 | } | |
4956 | ||
80ced186 | 4957 | static int handle_invalid_guest_state(struct kvm_vcpu *vcpu) |
ea953ef0 | 4958 | { |
8b3079a5 AK |
4959 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
4960 | enum emulation_result err = EMULATE_DONE; | |
80ced186 | 4961 | int ret = 1; |
49e9d557 AK |
4962 | u32 cpu_exec_ctrl; |
4963 | bool intr_window_requested; | |
b8405c18 | 4964 | unsigned count = 130; |
49e9d557 AK |
4965 | |
4966 | cpu_exec_ctrl = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL); | |
4967 | intr_window_requested = cpu_exec_ctrl & CPU_BASED_VIRTUAL_INTR_PENDING; | |
ea953ef0 | 4968 | |
b8405c18 | 4969 | while (!guest_state_valid(vcpu) && count-- != 0) { |
bdea48e3 | 4970 | if (intr_window_requested && vmx_interrupt_allowed(vcpu)) |
49e9d557 AK |
4971 | return handle_interrupt_window(&vmx->vcpu); |
4972 | ||
de87dcdd AK |
4973 | if (test_bit(KVM_REQ_EVENT, &vcpu->requests)) |
4974 | return 1; | |
4975 | ||
51d8b661 | 4976 | err = emulate_instruction(vcpu, 0); |
ea953ef0 | 4977 | |
80ced186 MG |
4978 | if (err == EMULATE_DO_MMIO) { |
4979 | ret = 0; | |
4980 | goto out; | |
4981 | } | |
1d5a4d9b | 4982 | |
de5f70e0 AK |
4983 | if (err != EMULATE_DONE) { |
4984 | vcpu->run->exit_reason = KVM_EXIT_INTERNAL_ERROR; | |
4985 | vcpu->run->internal.suberror = KVM_INTERNAL_ERROR_EMULATION; | |
4986 | vcpu->run->internal.ndata = 0; | |
6d77dbfc | 4987 | return 0; |
de5f70e0 | 4988 | } |
ea953ef0 MG |
4989 | |
4990 | if (signal_pending(current)) | |
80ced186 | 4991 | goto out; |
ea953ef0 MG |
4992 | if (need_resched()) |
4993 | schedule(); | |
4994 | } | |
4995 | ||
7c068e45 | 4996 | vmx->emulation_required = !guest_state_valid(vcpu); |
80ced186 MG |
4997 | out: |
4998 | return ret; | |
ea953ef0 MG |
4999 | } |
5000 | ||
4b8d54f9 ZE |
5001 | /* |
5002 | * Indicate a busy-waiting vcpu in spinlock. We do not enable the PAUSE | |
5003 | * exiting, so only get here on cpu with PAUSE-Loop-Exiting. | |
5004 | */ | |
9fb41ba8 | 5005 | static int handle_pause(struct kvm_vcpu *vcpu) |
4b8d54f9 ZE |
5006 | { |
5007 | skip_emulated_instruction(vcpu); | |
5008 | kvm_vcpu_on_spin(vcpu); | |
5009 | ||
5010 | return 1; | |
5011 | } | |
5012 | ||
59708670 SY |
5013 | static int handle_invalid_op(struct kvm_vcpu *vcpu) |
5014 | { | |
5015 | kvm_queue_exception(vcpu, UD_VECTOR); | |
5016 | return 1; | |
5017 | } | |
5018 | ||
ff2f6fe9 NHE |
5019 | /* |
5020 | * To run an L2 guest, we need a vmcs02 based on the L1-specified vmcs12. | |
5021 | * We could reuse a single VMCS for all the L2 guests, but we also want the | |
5022 | * option to allocate a separate vmcs02 for each separate loaded vmcs12 - this | |
5023 | * allows keeping them loaded on the processor, and in the future will allow | |
5024 | * optimizations where prepare_vmcs02 doesn't need to set all the fields on | |
5025 | * every entry if they never change. | |
5026 | * So we keep, in vmx->nested.vmcs02_pool, a cache of size VMCS02_POOL_SIZE | |
5027 | * (>=0) with a vmcs02 for each recently loaded vmcs12s, most recent first. | |
5028 | * | |
5029 | * The following functions allocate and free a vmcs02 in this pool. | |
5030 | */ | |
5031 | ||
5032 | /* Get a VMCS from the pool to use as vmcs02 for the current vmcs12. */ | |
5033 | static struct loaded_vmcs *nested_get_current_vmcs02(struct vcpu_vmx *vmx) | |
5034 | { | |
5035 | struct vmcs02_list *item; | |
5036 | list_for_each_entry(item, &vmx->nested.vmcs02_pool, list) | |
5037 | if (item->vmptr == vmx->nested.current_vmptr) { | |
5038 | list_move(&item->list, &vmx->nested.vmcs02_pool); | |
5039 | return &item->vmcs02; | |
5040 | } | |
5041 | ||
5042 | if (vmx->nested.vmcs02_num >= max(VMCS02_POOL_SIZE, 1)) { | |
5043 | /* Recycle the least recently used VMCS. */ | |
5044 | item = list_entry(vmx->nested.vmcs02_pool.prev, | |
5045 | struct vmcs02_list, list); | |
5046 | item->vmptr = vmx->nested.current_vmptr; | |
5047 | list_move(&item->list, &vmx->nested.vmcs02_pool); | |
5048 | return &item->vmcs02; | |
5049 | } | |
5050 | ||
5051 | /* Create a new VMCS */ | |
5052 | item = (struct vmcs02_list *) | |
5053 | kmalloc(sizeof(struct vmcs02_list), GFP_KERNEL); | |
5054 | if (!item) | |
5055 | return NULL; | |
5056 | item->vmcs02.vmcs = alloc_vmcs(); | |
5057 | if (!item->vmcs02.vmcs) { | |
5058 | kfree(item); | |
5059 | return NULL; | |
5060 | } | |
5061 | loaded_vmcs_init(&item->vmcs02); | |
5062 | item->vmptr = vmx->nested.current_vmptr; | |
5063 | list_add(&(item->list), &(vmx->nested.vmcs02_pool)); | |
5064 | vmx->nested.vmcs02_num++; | |
5065 | return &item->vmcs02; | |
5066 | } | |
5067 | ||
5068 | /* Free and remove from pool a vmcs02 saved for a vmcs12 (if there is one) */ | |
5069 | static void nested_free_vmcs02(struct vcpu_vmx *vmx, gpa_t vmptr) | |
5070 | { | |
5071 | struct vmcs02_list *item; | |
5072 | list_for_each_entry(item, &vmx->nested.vmcs02_pool, list) | |
5073 | if (item->vmptr == vmptr) { | |
5074 | free_loaded_vmcs(&item->vmcs02); | |
5075 | list_del(&item->list); | |
5076 | kfree(item); | |
5077 | vmx->nested.vmcs02_num--; | |
5078 | return; | |
5079 | } | |
5080 | } | |
5081 | ||
5082 | /* | |
5083 | * Free all VMCSs saved for this vcpu, except the one pointed by | |
5084 | * vmx->loaded_vmcs. These include the VMCSs in vmcs02_pool (except the one | |
5085 | * currently used, if running L2), and vmcs01 when running L2. | |
5086 | */ | |
5087 | static void nested_free_all_saved_vmcss(struct vcpu_vmx *vmx) | |
5088 | { | |
5089 | struct vmcs02_list *item, *n; | |
5090 | list_for_each_entry_safe(item, n, &vmx->nested.vmcs02_pool, list) { | |
5091 | if (vmx->loaded_vmcs != &item->vmcs02) | |
5092 | free_loaded_vmcs(&item->vmcs02); | |
5093 | list_del(&item->list); | |
5094 | kfree(item); | |
5095 | } | |
5096 | vmx->nested.vmcs02_num = 0; | |
5097 | ||
5098 | if (vmx->loaded_vmcs != &vmx->vmcs01) | |
5099 | free_loaded_vmcs(&vmx->vmcs01); | |
5100 | } | |
5101 | ||
ec378aee NHE |
5102 | /* |
5103 | * Emulate the VMXON instruction. | |
5104 | * Currently, we just remember that VMX is active, and do not save or even | |
5105 | * inspect the argument to VMXON (the so-called "VMXON pointer") because we | |
5106 | * do not currently need to store anything in that guest-allocated memory | |
5107 | * region. Consequently, VMCLEAR and VMPTRLD also do not verify that the their | |
5108 | * argument is different from the VMXON pointer (which the spec says they do). | |
5109 | */ | |
5110 | static int handle_vmon(struct kvm_vcpu *vcpu) | |
5111 | { | |
5112 | struct kvm_segment cs; | |
5113 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
5114 | ||
5115 | /* The Intel VMX Instruction Reference lists a bunch of bits that | |
5116 | * are prerequisite to running VMXON, most notably cr4.VMXE must be | |
5117 | * set to 1 (see vmx_set_cr4() for when we allow the guest to set this). | |
5118 | * Otherwise, we should fail with #UD. We test these now: | |
5119 | */ | |
5120 | if (!kvm_read_cr4_bits(vcpu, X86_CR4_VMXE) || | |
5121 | !kvm_read_cr0_bits(vcpu, X86_CR0_PE) || | |
5122 | (vmx_get_rflags(vcpu) & X86_EFLAGS_VM)) { | |
5123 | kvm_queue_exception(vcpu, UD_VECTOR); | |
5124 | return 1; | |
5125 | } | |
5126 | ||
5127 | vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); | |
5128 | if (is_long_mode(vcpu) && !cs.l) { | |
5129 | kvm_queue_exception(vcpu, UD_VECTOR); | |
5130 | return 1; | |
5131 | } | |
5132 | ||
5133 | if (vmx_get_cpl(vcpu)) { | |
5134 | kvm_inject_gp(vcpu, 0); | |
5135 | return 1; | |
5136 | } | |
5137 | ||
ff2f6fe9 NHE |
5138 | INIT_LIST_HEAD(&(vmx->nested.vmcs02_pool)); |
5139 | vmx->nested.vmcs02_num = 0; | |
5140 | ||
ec378aee NHE |
5141 | vmx->nested.vmxon = true; |
5142 | ||
5143 | skip_emulated_instruction(vcpu); | |
5144 | return 1; | |
5145 | } | |
5146 | ||
5147 | /* | |
5148 | * Intel's VMX Instruction Reference specifies a common set of prerequisites | |
5149 | * for running VMX instructions (except VMXON, whose prerequisites are | |
5150 | * slightly different). It also specifies what exception to inject otherwise. | |
5151 | */ | |
5152 | static int nested_vmx_check_permission(struct kvm_vcpu *vcpu) | |
5153 | { | |
5154 | struct kvm_segment cs; | |
5155 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
5156 | ||
5157 | if (!vmx->nested.vmxon) { | |
5158 | kvm_queue_exception(vcpu, UD_VECTOR); | |
5159 | return 0; | |
5160 | } | |
5161 | ||
5162 | vmx_get_segment(vcpu, &cs, VCPU_SREG_CS); | |
5163 | if ((vmx_get_rflags(vcpu) & X86_EFLAGS_VM) || | |
5164 | (is_long_mode(vcpu) && !cs.l)) { | |
5165 | kvm_queue_exception(vcpu, UD_VECTOR); | |
5166 | return 0; | |
5167 | } | |
5168 | ||
5169 | if (vmx_get_cpl(vcpu)) { | |
5170 | kvm_inject_gp(vcpu, 0); | |
5171 | return 0; | |
5172 | } | |
5173 | ||
5174 | return 1; | |
5175 | } | |
5176 | ||
5177 | /* | |
5178 | * Free whatever needs to be freed from vmx->nested when L1 goes down, or | |
5179 | * just stops using VMX. | |
5180 | */ | |
5181 | static void free_nested(struct vcpu_vmx *vmx) | |
5182 | { | |
5183 | if (!vmx->nested.vmxon) | |
5184 | return; | |
5185 | vmx->nested.vmxon = false; | |
a9d30f33 NHE |
5186 | if (vmx->nested.current_vmptr != -1ull) { |
5187 | kunmap(vmx->nested.current_vmcs12_page); | |
5188 | nested_release_page(vmx->nested.current_vmcs12_page); | |
5189 | vmx->nested.current_vmptr = -1ull; | |
5190 | vmx->nested.current_vmcs12 = NULL; | |
5191 | } | |
fe3ef05c NHE |
5192 | /* Unpin physical memory we referred to in current vmcs02 */ |
5193 | if (vmx->nested.apic_access_page) { | |
5194 | nested_release_page(vmx->nested.apic_access_page); | |
5195 | vmx->nested.apic_access_page = 0; | |
5196 | } | |
ff2f6fe9 NHE |
5197 | |
5198 | nested_free_all_saved_vmcss(vmx); | |
ec378aee NHE |
5199 | } |
5200 | ||
5201 | /* Emulate the VMXOFF instruction */ | |
5202 | static int handle_vmoff(struct kvm_vcpu *vcpu) | |
5203 | { | |
5204 | if (!nested_vmx_check_permission(vcpu)) | |
5205 | return 1; | |
5206 | free_nested(to_vmx(vcpu)); | |
5207 | skip_emulated_instruction(vcpu); | |
5208 | return 1; | |
5209 | } | |
5210 | ||
064aea77 NHE |
5211 | /* |
5212 | * Decode the memory-address operand of a vmx instruction, as recorded on an | |
5213 | * exit caused by such an instruction (run by a guest hypervisor). | |
5214 | * On success, returns 0. When the operand is invalid, returns 1 and throws | |
5215 | * #UD or #GP. | |
5216 | */ | |
5217 | static int get_vmx_mem_address(struct kvm_vcpu *vcpu, | |
5218 | unsigned long exit_qualification, | |
5219 | u32 vmx_instruction_info, gva_t *ret) | |
5220 | { | |
5221 | /* | |
5222 | * According to Vol. 3B, "Information for VM Exits Due to Instruction | |
5223 | * Execution", on an exit, vmx_instruction_info holds most of the | |
5224 | * addressing components of the operand. Only the displacement part | |
5225 | * is put in exit_qualification (see 3B, "Basic VM-Exit Information"). | |
5226 | * For how an actual address is calculated from all these components, | |
5227 | * refer to Vol. 1, "Operand Addressing". | |
5228 | */ | |
5229 | int scaling = vmx_instruction_info & 3; | |
5230 | int addr_size = (vmx_instruction_info >> 7) & 7; | |
5231 | bool is_reg = vmx_instruction_info & (1u << 10); | |
5232 | int seg_reg = (vmx_instruction_info >> 15) & 7; | |
5233 | int index_reg = (vmx_instruction_info >> 18) & 0xf; | |
5234 | bool index_is_valid = !(vmx_instruction_info & (1u << 22)); | |
5235 | int base_reg = (vmx_instruction_info >> 23) & 0xf; | |
5236 | bool base_is_valid = !(vmx_instruction_info & (1u << 27)); | |
5237 | ||
5238 | if (is_reg) { | |
5239 | kvm_queue_exception(vcpu, UD_VECTOR); | |
5240 | return 1; | |
5241 | } | |
5242 | ||
5243 | /* Addr = segment_base + offset */ | |
5244 | /* offset = base + [index * scale] + displacement */ | |
5245 | *ret = vmx_get_segment_base(vcpu, seg_reg); | |
5246 | if (base_is_valid) | |
5247 | *ret += kvm_register_read(vcpu, base_reg); | |
5248 | if (index_is_valid) | |
5249 | *ret += kvm_register_read(vcpu, index_reg)<<scaling; | |
5250 | *ret += exit_qualification; /* holds the displacement */ | |
5251 | ||
5252 | if (addr_size == 1) /* 32 bit */ | |
5253 | *ret &= 0xffffffff; | |
5254 | ||
5255 | /* | |
5256 | * TODO: throw #GP (and return 1) in various cases that the VM* | |
5257 | * instructions require it - e.g., offset beyond segment limit, | |
5258 | * unusable or unreadable/unwritable segment, non-canonical 64-bit | |
5259 | * address, and so on. Currently these are not checked. | |
5260 | */ | |
5261 | return 0; | |
5262 | } | |
5263 | ||
0140caea NHE |
5264 | /* |
5265 | * The following 3 functions, nested_vmx_succeed()/failValid()/failInvalid(), | |
5266 | * set the success or error code of an emulated VMX instruction, as specified | |
5267 | * by Vol 2B, VMX Instruction Reference, "Conventions". | |
5268 | */ | |
5269 | static void nested_vmx_succeed(struct kvm_vcpu *vcpu) | |
5270 | { | |
5271 | vmx_set_rflags(vcpu, vmx_get_rflags(vcpu) | |
5272 | & ~(X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF | | |
5273 | X86_EFLAGS_ZF | X86_EFLAGS_SF | X86_EFLAGS_OF)); | |
5274 | } | |
5275 | ||
5276 | static void nested_vmx_failInvalid(struct kvm_vcpu *vcpu) | |
5277 | { | |
5278 | vmx_set_rflags(vcpu, (vmx_get_rflags(vcpu) | |
5279 | & ~(X86_EFLAGS_PF | X86_EFLAGS_AF | X86_EFLAGS_ZF | | |
5280 | X86_EFLAGS_SF | X86_EFLAGS_OF)) | |
5281 | | X86_EFLAGS_CF); | |
5282 | } | |
5283 | ||
5284 | static void nested_vmx_failValid(struct kvm_vcpu *vcpu, | |
5285 | u32 vm_instruction_error) | |
5286 | { | |
5287 | if (to_vmx(vcpu)->nested.current_vmptr == -1ull) { | |
5288 | /* | |
5289 | * failValid writes the error number to the current VMCS, which | |
5290 | * can't be done there isn't a current VMCS. | |
5291 | */ | |
5292 | nested_vmx_failInvalid(vcpu); | |
5293 | return; | |
5294 | } | |
5295 | vmx_set_rflags(vcpu, (vmx_get_rflags(vcpu) | |
5296 | & ~(X86_EFLAGS_CF | X86_EFLAGS_PF | X86_EFLAGS_AF | | |
5297 | X86_EFLAGS_SF | X86_EFLAGS_OF)) | |
5298 | | X86_EFLAGS_ZF); | |
5299 | get_vmcs12(vcpu)->vm_instruction_error = vm_instruction_error; | |
5300 | } | |
5301 | ||
27d6c865 NHE |
5302 | /* Emulate the VMCLEAR instruction */ |
5303 | static int handle_vmclear(struct kvm_vcpu *vcpu) | |
5304 | { | |
5305 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
5306 | gva_t gva; | |
5307 | gpa_t vmptr; | |
5308 | struct vmcs12 *vmcs12; | |
5309 | struct page *page; | |
5310 | struct x86_exception e; | |
5311 | ||
5312 | if (!nested_vmx_check_permission(vcpu)) | |
5313 | return 1; | |
5314 | ||
5315 | if (get_vmx_mem_address(vcpu, vmcs_readl(EXIT_QUALIFICATION), | |
5316 | vmcs_read32(VMX_INSTRUCTION_INFO), &gva)) | |
5317 | return 1; | |
5318 | ||
5319 | if (kvm_read_guest_virt(&vcpu->arch.emulate_ctxt, gva, &vmptr, | |
5320 | sizeof(vmptr), &e)) { | |
5321 | kvm_inject_page_fault(vcpu, &e); | |
5322 | return 1; | |
5323 | } | |
5324 | ||
5325 | if (!IS_ALIGNED(vmptr, PAGE_SIZE)) { | |
5326 | nested_vmx_failValid(vcpu, VMXERR_VMCLEAR_INVALID_ADDRESS); | |
5327 | skip_emulated_instruction(vcpu); | |
5328 | return 1; | |
5329 | } | |
5330 | ||
5331 | if (vmptr == vmx->nested.current_vmptr) { | |
5332 | kunmap(vmx->nested.current_vmcs12_page); | |
5333 | nested_release_page(vmx->nested.current_vmcs12_page); | |
5334 | vmx->nested.current_vmptr = -1ull; | |
5335 | vmx->nested.current_vmcs12 = NULL; | |
5336 | } | |
5337 | ||
5338 | page = nested_get_page(vcpu, vmptr); | |
5339 | if (page == NULL) { | |
5340 | /* | |
5341 | * For accurate processor emulation, VMCLEAR beyond available | |
5342 | * physical memory should do nothing at all. However, it is | |
5343 | * possible that a nested vmx bug, not a guest hypervisor bug, | |
5344 | * resulted in this case, so let's shut down before doing any | |
5345 | * more damage: | |
5346 | */ | |
5347 | kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu); | |
5348 | return 1; | |
5349 | } | |
5350 | vmcs12 = kmap(page); | |
5351 | vmcs12->launch_state = 0; | |
5352 | kunmap(page); | |
5353 | nested_release_page(page); | |
5354 | ||
5355 | nested_free_vmcs02(vmx, vmptr); | |
5356 | ||
5357 | skip_emulated_instruction(vcpu); | |
5358 | nested_vmx_succeed(vcpu); | |
5359 | return 1; | |
5360 | } | |
5361 | ||
cd232ad0 NHE |
5362 | static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch); |
5363 | ||
5364 | /* Emulate the VMLAUNCH instruction */ | |
5365 | static int handle_vmlaunch(struct kvm_vcpu *vcpu) | |
5366 | { | |
5367 | return nested_vmx_run(vcpu, true); | |
5368 | } | |
5369 | ||
5370 | /* Emulate the VMRESUME instruction */ | |
5371 | static int handle_vmresume(struct kvm_vcpu *vcpu) | |
5372 | { | |
5373 | ||
5374 | return nested_vmx_run(vcpu, false); | |
5375 | } | |
5376 | ||
49f705c5 NHE |
5377 | enum vmcs_field_type { |
5378 | VMCS_FIELD_TYPE_U16 = 0, | |
5379 | VMCS_FIELD_TYPE_U64 = 1, | |
5380 | VMCS_FIELD_TYPE_U32 = 2, | |
5381 | VMCS_FIELD_TYPE_NATURAL_WIDTH = 3 | |
5382 | }; | |
5383 | ||
5384 | static inline int vmcs_field_type(unsigned long field) | |
5385 | { | |
5386 | if (0x1 & field) /* the *_HIGH fields are all 32 bit */ | |
5387 | return VMCS_FIELD_TYPE_U32; | |
5388 | return (field >> 13) & 0x3 ; | |
5389 | } | |
5390 | ||
5391 | static inline int vmcs_field_readonly(unsigned long field) | |
5392 | { | |
5393 | return (((field >> 10) & 0x3) == 1); | |
5394 | } | |
5395 | ||
5396 | /* | |
5397 | * Read a vmcs12 field. Since these can have varying lengths and we return | |
5398 | * one type, we chose the biggest type (u64) and zero-extend the return value | |
5399 | * to that size. Note that the caller, handle_vmread, might need to use only | |
5400 | * some of the bits we return here (e.g., on 32-bit guests, only 32 bits of | |
5401 | * 64-bit fields are to be returned). | |
5402 | */ | |
5403 | static inline bool vmcs12_read_any(struct kvm_vcpu *vcpu, | |
5404 | unsigned long field, u64 *ret) | |
5405 | { | |
5406 | short offset = vmcs_field_to_offset(field); | |
5407 | char *p; | |
5408 | ||
5409 | if (offset < 0) | |
5410 | return 0; | |
5411 | ||
5412 | p = ((char *)(get_vmcs12(vcpu))) + offset; | |
5413 | ||
5414 | switch (vmcs_field_type(field)) { | |
5415 | case VMCS_FIELD_TYPE_NATURAL_WIDTH: | |
5416 | *ret = *((natural_width *)p); | |
5417 | return 1; | |
5418 | case VMCS_FIELD_TYPE_U16: | |
5419 | *ret = *((u16 *)p); | |
5420 | return 1; | |
5421 | case VMCS_FIELD_TYPE_U32: | |
5422 | *ret = *((u32 *)p); | |
5423 | return 1; | |
5424 | case VMCS_FIELD_TYPE_U64: | |
5425 | *ret = *((u64 *)p); | |
5426 | return 1; | |
5427 | default: | |
5428 | return 0; /* can never happen. */ | |
5429 | } | |
5430 | } | |
5431 | ||
5432 | /* | |
5433 | * VMX instructions which assume a current vmcs12 (i.e., that VMPTRLD was | |
5434 | * used before) all generate the same failure when it is missing. | |
5435 | */ | |
5436 | static int nested_vmx_check_vmcs12(struct kvm_vcpu *vcpu) | |
5437 | { | |
5438 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
5439 | if (vmx->nested.current_vmptr == -1ull) { | |
5440 | nested_vmx_failInvalid(vcpu); | |
5441 | skip_emulated_instruction(vcpu); | |
5442 | return 0; | |
5443 | } | |
5444 | return 1; | |
5445 | } | |
5446 | ||
5447 | static int handle_vmread(struct kvm_vcpu *vcpu) | |
5448 | { | |
5449 | unsigned long field; | |
5450 | u64 field_value; | |
5451 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
5452 | u32 vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO); | |
5453 | gva_t gva = 0; | |
5454 | ||
5455 | if (!nested_vmx_check_permission(vcpu) || | |
5456 | !nested_vmx_check_vmcs12(vcpu)) | |
5457 | return 1; | |
5458 | ||
5459 | /* Decode instruction info and find the field to read */ | |
5460 | field = kvm_register_read(vcpu, (((vmx_instruction_info) >> 28) & 0xf)); | |
5461 | /* Read the field, zero-extended to a u64 field_value */ | |
5462 | if (!vmcs12_read_any(vcpu, field, &field_value)) { | |
5463 | nested_vmx_failValid(vcpu, VMXERR_UNSUPPORTED_VMCS_COMPONENT); | |
5464 | skip_emulated_instruction(vcpu); | |
5465 | return 1; | |
5466 | } | |
5467 | /* | |
5468 | * Now copy part of this value to register or memory, as requested. | |
5469 | * Note that the number of bits actually copied is 32 or 64 depending | |
5470 | * on the guest's mode (32 or 64 bit), not on the given field's length. | |
5471 | */ | |
5472 | if (vmx_instruction_info & (1u << 10)) { | |
5473 | kvm_register_write(vcpu, (((vmx_instruction_info) >> 3) & 0xf), | |
5474 | field_value); | |
5475 | } else { | |
5476 | if (get_vmx_mem_address(vcpu, exit_qualification, | |
5477 | vmx_instruction_info, &gva)) | |
5478 | return 1; | |
5479 | /* _system ok, as nested_vmx_check_permission verified cpl=0 */ | |
5480 | kvm_write_guest_virt_system(&vcpu->arch.emulate_ctxt, gva, | |
5481 | &field_value, (is_long_mode(vcpu) ? 8 : 4), NULL); | |
5482 | } | |
5483 | ||
5484 | nested_vmx_succeed(vcpu); | |
5485 | skip_emulated_instruction(vcpu); | |
5486 | return 1; | |
5487 | } | |
5488 | ||
5489 | ||
5490 | static int handle_vmwrite(struct kvm_vcpu *vcpu) | |
5491 | { | |
5492 | unsigned long field; | |
5493 | gva_t gva; | |
5494 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
5495 | u32 vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO); | |
5496 | char *p; | |
5497 | short offset; | |
5498 | /* The value to write might be 32 or 64 bits, depending on L1's long | |
5499 | * mode, and eventually we need to write that into a field of several | |
5500 | * possible lengths. The code below first zero-extends the value to 64 | |
5501 | * bit (field_value), and then copies only the approriate number of | |
5502 | * bits into the vmcs12 field. | |
5503 | */ | |
5504 | u64 field_value = 0; | |
5505 | struct x86_exception e; | |
5506 | ||
5507 | if (!nested_vmx_check_permission(vcpu) || | |
5508 | !nested_vmx_check_vmcs12(vcpu)) | |
5509 | return 1; | |
5510 | ||
5511 | if (vmx_instruction_info & (1u << 10)) | |
5512 | field_value = kvm_register_read(vcpu, | |
5513 | (((vmx_instruction_info) >> 3) & 0xf)); | |
5514 | else { | |
5515 | if (get_vmx_mem_address(vcpu, exit_qualification, | |
5516 | vmx_instruction_info, &gva)) | |
5517 | return 1; | |
5518 | if (kvm_read_guest_virt(&vcpu->arch.emulate_ctxt, gva, | |
5519 | &field_value, (is_long_mode(vcpu) ? 8 : 4), &e)) { | |
5520 | kvm_inject_page_fault(vcpu, &e); | |
5521 | return 1; | |
5522 | } | |
5523 | } | |
5524 | ||
5525 | ||
5526 | field = kvm_register_read(vcpu, (((vmx_instruction_info) >> 28) & 0xf)); | |
5527 | if (vmcs_field_readonly(field)) { | |
5528 | nested_vmx_failValid(vcpu, | |
5529 | VMXERR_VMWRITE_READ_ONLY_VMCS_COMPONENT); | |
5530 | skip_emulated_instruction(vcpu); | |
5531 | return 1; | |
5532 | } | |
5533 | ||
5534 | offset = vmcs_field_to_offset(field); | |
5535 | if (offset < 0) { | |
5536 | nested_vmx_failValid(vcpu, VMXERR_UNSUPPORTED_VMCS_COMPONENT); | |
5537 | skip_emulated_instruction(vcpu); | |
5538 | return 1; | |
5539 | } | |
5540 | p = ((char *) get_vmcs12(vcpu)) + offset; | |
5541 | ||
5542 | switch (vmcs_field_type(field)) { | |
5543 | case VMCS_FIELD_TYPE_U16: | |
5544 | *(u16 *)p = field_value; | |
5545 | break; | |
5546 | case VMCS_FIELD_TYPE_U32: | |
5547 | *(u32 *)p = field_value; | |
5548 | break; | |
5549 | case VMCS_FIELD_TYPE_U64: | |
5550 | *(u64 *)p = field_value; | |
5551 | break; | |
5552 | case VMCS_FIELD_TYPE_NATURAL_WIDTH: | |
5553 | *(natural_width *)p = field_value; | |
5554 | break; | |
5555 | default: | |
5556 | nested_vmx_failValid(vcpu, VMXERR_UNSUPPORTED_VMCS_COMPONENT); | |
5557 | skip_emulated_instruction(vcpu); | |
5558 | return 1; | |
5559 | } | |
5560 | ||
5561 | nested_vmx_succeed(vcpu); | |
5562 | skip_emulated_instruction(vcpu); | |
5563 | return 1; | |
5564 | } | |
5565 | ||
63846663 NHE |
5566 | /* Emulate the VMPTRLD instruction */ |
5567 | static int handle_vmptrld(struct kvm_vcpu *vcpu) | |
5568 | { | |
5569 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
5570 | gva_t gva; | |
5571 | gpa_t vmptr; | |
5572 | struct x86_exception e; | |
5573 | ||
5574 | if (!nested_vmx_check_permission(vcpu)) | |
5575 | return 1; | |
5576 | ||
5577 | if (get_vmx_mem_address(vcpu, vmcs_readl(EXIT_QUALIFICATION), | |
5578 | vmcs_read32(VMX_INSTRUCTION_INFO), &gva)) | |
5579 | return 1; | |
5580 | ||
5581 | if (kvm_read_guest_virt(&vcpu->arch.emulate_ctxt, gva, &vmptr, | |
5582 | sizeof(vmptr), &e)) { | |
5583 | kvm_inject_page_fault(vcpu, &e); | |
5584 | return 1; | |
5585 | } | |
5586 | ||
5587 | if (!IS_ALIGNED(vmptr, PAGE_SIZE)) { | |
5588 | nested_vmx_failValid(vcpu, VMXERR_VMPTRLD_INVALID_ADDRESS); | |
5589 | skip_emulated_instruction(vcpu); | |
5590 | return 1; | |
5591 | } | |
5592 | ||
5593 | if (vmx->nested.current_vmptr != vmptr) { | |
5594 | struct vmcs12 *new_vmcs12; | |
5595 | struct page *page; | |
5596 | page = nested_get_page(vcpu, vmptr); | |
5597 | if (page == NULL) { | |
5598 | nested_vmx_failInvalid(vcpu); | |
5599 | skip_emulated_instruction(vcpu); | |
5600 | return 1; | |
5601 | } | |
5602 | new_vmcs12 = kmap(page); | |
5603 | if (new_vmcs12->revision_id != VMCS12_REVISION) { | |
5604 | kunmap(page); | |
5605 | nested_release_page_clean(page); | |
5606 | nested_vmx_failValid(vcpu, | |
5607 | VMXERR_VMPTRLD_INCORRECT_VMCS_REVISION_ID); | |
5608 | skip_emulated_instruction(vcpu); | |
5609 | return 1; | |
5610 | } | |
5611 | if (vmx->nested.current_vmptr != -1ull) { | |
5612 | kunmap(vmx->nested.current_vmcs12_page); | |
5613 | nested_release_page(vmx->nested.current_vmcs12_page); | |
5614 | } | |
5615 | ||
5616 | vmx->nested.current_vmptr = vmptr; | |
5617 | vmx->nested.current_vmcs12 = new_vmcs12; | |
5618 | vmx->nested.current_vmcs12_page = page; | |
5619 | } | |
5620 | ||
5621 | nested_vmx_succeed(vcpu); | |
5622 | skip_emulated_instruction(vcpu); | |
5623 | return 1; | |
5624 | } | |
5625 | ||
6a4d7550 NHE |
5626 | /* Emulate the VMPTRST instruction */ |
5627 | static int handle_vmptrst(struct kvm_vcpu *vcpu) | |
5628 | { | |
5629 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
5630 | u32 vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO); | |
5631 | gva_t vmcs_gva; | |
5632 | struct x86_exception e; | |
5633 | ||
5634 | if (!nested_vmx_check_permission(vcpu)) | |
5635 | return 1; | |
5636 | ||
5637 | if (get_vmx_mem_address(vcpu, exit_qualification, | |
5638 | vmx_instruction_info, &vmcs_gva)) | |
5639 | return 1; | |
5640 | /* ok to use *_system, as nested_vmx_check_permission verified cpl=0 */ | |
5641 | if (kvm_write_guest_virt_system(&vcpu->arch.emulate_ctxt, vmcs_gva, | |
5642 | (void *)&to_vmx(vcpu)->nested.current_vmptr, | |
5643 | sizeof(u64), &e)) { | |
5644 | kvm_inject_page_fault(vcpu, &e); | |
5645 | return 1; | |
5646 | } | |
5647 | nested_vmx_succeed(vcpu); | |
5648 | skip_emulated_instruction(vcpu); | |
5649 | return 1; | |
5650 | } | |
5651 | ||
6aa8b732 AK |
5652 | /* |
5653 | * The exit handlers return 1 if the exit was handled fully and guest execution | |
5654 | * may resume. Otherwise they set the kvm_run parameter to indicate what needs | |
5655 | * to be done to userspace and return 0. | |
5656 | */ | |
851ba692 | 5657 | static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = { |
6aa8b732 AK |
5658 | [EXIT_REASON_EXCEPTION_NMI] = handle_exception, |
5659 | [EXIT_REASON_EXTERNAL_INTERRUPT] = handle_external_interrupt, | |
988ad74f | 5660 | [EXIT_REASON_TRIPLE_FAULT] = handle_triple_fault, |
f08864b4 | 5661 | [EXIT_REASON_NMI_WINDOW] = handle_nmi_window, |
6aa8b732 | 5662 | [EXIT_REASON_IO_INSTRUCTION] = handle_io, |
6aa8b732 AK |
5663 | [EXIT_REASON_CR_ACCESS] = handle_cr, |
5664 | [EXIT_REASON_DR_ACCESS] = handle_dr, | |
5665 | [EXIT_REASON_CPUID] = handle_cpuid, | |
5666 | [EXIT_REASON_MSR_READ] = handle_rdmsr, | |
5667 | [EXIT_REASON_MSR_WRITE] = handle_wrmsr, | |
5668 | [EXIT_REASON_PENDING_INTERRUPT] = handle_interrupt_window, | |
5669 | [EXIT_REASON_HLT] = handle_halt, | |
ec25d5e6 | 5670 | [EXIT_REASON_INVD] = handle_invd, |
a7052897 | 5671 | [EXIT_REASON_INVLPG] = handle_invlpg, |
fee84b07 | 5672 | [EXIT_REASON_RDPMC] = handle_rdpmc, |
c21415e8 | 5673 | [EXIT_REASON_VMCALL] = handle_vmcall, |
27d6c865 | 5674 | [EXIT_REASON_VMCLEAR] = handle_vmclear, |
cd232ad0 | 5675 | [EXIT_REASON_VMLAUNCH] = handle_vmlaunch, |
63846663 | 5676 | [EXIT_REASON_VMPTRLD] = handle_vmptrld, |
6a4d7550 | 5677 | [EXIT_REASON_VMPTRST] = handle_vmptrst, |
49f705c5 | 5678 | [EXIT_REASON_VMREAD] = handle_vmread, |
cd232ad0 | 5679 | [EXIT_REASON_VMRESUME] = handle_vmresume, |
49f705c5 | 5680 | [EXIT_REASON_VMWRITE] = handle_vmwrite, |
ec378aee NHE |
5681 | [EXIT_REASON_VMOFF] = handle_vmoff, |
5682 | [EXIT_REASON_VMON] = handle_vmon, | |
f78e0e2e SY |
5683 | [EXIT_REASON_TPR_BELOW_THRESHOLD] = handle_tpr_below_threshold, |
5684 | [EXIT_REASON_APIC_ACCESS] = handle_apic_access, | |
e5edaa01 | 5685 | [EXIT_REASON_WBINVD] = handle_wbinvd, |
2acf923e | 5686 | [EXIT_REASON_XSETBV] = handle_xsetbv, |
37817f29 | 5687 | [EXIT_REASON_TASK_SWITCH] = handle_task_switch, |
a0861c02 | 5688 | [EXIT_REASON_MCE_DURING_VMENTRY] = handle_machine_check, |
68f89400 MT |
5689 | [EXIT_REASON_EPT_VIOLATION] = handle_ept_violation, |
5690 | [EXIT_REASON_EPT_MISCONFIG] = handle_ept_misconfig, | |
4b8d54f9 | 5691 | [EXIT_REASON_PAUSE_INSTRUCTION] = handle_pause, |
59708670 SY |
5692 | [EXIT_REASON_MWAIT_INSTRUCTION] = handle_invalid_op, |
5693 | [EXIT_REASON_MONITOR_INSTRUCTION] = handle_invalid_op, | |
6aa8b732 AK |
5694 | }; |
5695 | ||
5696 | static const int kvm_vmx_max_exit_handlers = | |
50a3485c | 5697 | ARRAY_SIZE(kvm_vmx_exit_handlers); |
6aa8b732 | 5698 | |
644d711a NHE |
5699 | /* |
5700 | * Return 1 if we should exit from L2 to L1 to handle an MSR access access, | |
5701 | * rather than handle it ourselves in L0. I.e., check whether L1 expressed | |
5702 | * disinterest in the current event (read or write a specific MSR) by using an | |
5703 | * MSR bitmap. This may be the case even when L0 doesn't use MSR bitmaps. | |
5704 | */ | |
5705 | static bool nested_vmx_exit_handled_msr(struct kvm_vcpu *vcpu, | |
5706 | struct vmcs12 *vmcs12, u32 exit_reason) | |
5707 | { | |
5708 | u32 msr_index = vcpu->arch.regs[VCPU_REGS_RCX]; | |
5709 | gpa_t bitmap; | |
5710 | ||
5711 | if (!nested_cpu_has(get_vmcs12(vcpu), CPU_BASED_USE_MSR_BITMAPS)) | |
5712 | return 1; | |
5713 | ||
5714 | /* | |
5715 | * The MSR_BITMAP page is divided into four 1024-byte bitmaps, | |
5716 | * for the four combinations of read/write and low/high MSR numbers. | |
5717 | * First we need to figure out which of the four to use: | |
5718 | */ | |
5719 | bitmap = vmcs12->msr_bitmap; | |
5720 | if (exit_reason == EXIT_REASON_MSR_WRITE) | |
5721 | bitmap += 2048; | |
5722 | if (msr_index >= 0xc0000000) { | |
5723 | msr_index -= 0xc0000000; | |
5724 | bitmap += 1024; | |
5725 | } | |
5726 | ||
5727 | /* Then read the msr_index'th bit from this bitmap: */ | |
5728 | if (msr_index < 1024*8) { | |
5729 | unsigned char b; | |
5730 | kvm_read_guest(vcpu->kvm, bitmap + msr_index/8, &b, 1); | |
5731 | return 1 & (b >> (msr_index & 7)); | |
5732 | } else | |
5733 | return 1; /* let L1 handle the wrong parameter */ | |
5734 | } | |
5735 | ||
5736 | /* | |
5737 | * Return 1 if we should exit from L2 to L1 to handle a CR access exit, | |
5738 | * rather than handle it ourselves in L0. I.e., check if L1 wanted to | |
5739 | * intercept (via guest_host_mask etc.) the current event. | |
5740 | */ | |
5741 | static bool nested_vmx_exit_handled_cr(struct kvm_vcpu *vcpu, | |
5742 | struct vmcs12 *vmcs12) | |
5743 | { | |
5744 | unsigned long exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
5745 | int cr = exit_qualification & 15; | |
5746 | int reg = (exit_qualification >> 8) & 15; | |
5747 | unsigned long val = kvm_register_read(vcpu, reg); | |
5748 | ||
5749 | switch ((exit_qualification >> 4) & 3) { | |
5750 | case 0: /* mov to cr */ | |
5751 | switch (cr) { | |
5752 | case 0: | |
5753 | if (vmcs12->cr0_guest_host_mask & | |
5754 | (val ^ vmcs12->cr0_read_shadow)) | |
5755 | return 1; | |
5756 | break; | |
5757 | case 3: | |
5758 | if ((vmcs12->cr3_target_count >= 1 && | |
5759 | vmcs12->cr3_target_value0 == val) || | |
5760 | (vmcs12->cr3_target_count >= 2 && | |
5761 | vmcs12->cr3_target_value1 == val) || | |
5762 | (vmcs12->cr3_target_count >= 3 && | |
5763 | vmcs12->cr3_target_value2 == val) || | |
5764 | (vmcs12->cr3_target_count >= 4 && | |
5765 | vmcs12->cr3_target_value3 == val)) | |
5766 | return 0; | |
5767 | if (nested_cpu_has(vmcs12, CPU_BASED_CR3_LOAD_EXITING)) | |
5768 | return 1; | |
5769 | break; | |
5770 | case 4: | |
5771 | if (vmcs12->cr4_guest_host_mask & | |
5772 | (vmcs12->cr4_read_shadow ^ val)) | |
5773 | return 1; | |
5774 | break; | |
5775 | case 8: | |
5776 | if (nested_cpu_has(vmcs12, CPU_BASED_CR8_LOAD_EXITING)) | |
5777 | return 1; | |
5778 | break; | |
5779 | } | |
5780 | break; | |
5781 | case 2: /* clts */ | |
5782 | if ((vmcs12->cr0_guest_host_mask & X86_CR0_TS) && | |
5783 | (vmcs12->cr0_read_shadow & X86_CR0_TS)) | |
5784 | return 1; | |
5785 | break; | |
5786 | case 1: /* mov from cr */ | |
5787 | switch (cr) { | |
5788 | case 3: | |
5789 | if (vmcs12->cpu_based_vm_exec_control & | |
5790 | CPU_BASED_CR3_STORE_EXITING) | |
5791 | return 1; | |
5792 | break; | |
5793 | case 8: | |
5794 | if (vmcs12->cpu_based_vm_exec_control & | |
5795 | CPU_BASED_CR8_STORE_EXITING) | |
5796 | return 1; | |
5797 | break; | |
5798 | } | |
5799 | break; | |
5800 | case 3: /* lmsw */ | |
5801 | /* | |
5802 | * lmsw can change bits 1..3 of cr0, and only set bit 0 of | |
5803 | * cr0. Other attempted changes are ignored, with no exit. | |
5804 | */ | |
5805 | if (vmcs12->cr0_guest_host_mask & 0xe & | |
5806 | (val ^ vmcs12->cr0_read_shadow)) | |
5807 | return 1; | |
5808 | if ((vmcs12->cr0_guest_host_mask & 0x1) && | |
5809 | !(vmcs12->cr0_read_shadow & 0x1) && | |
5810 | (val & 0x1)) | |
5811 | return 1; | |
5812 | break; | |
5813 | } | |
5814 | return 0; | |
5815 | } | |
5816 | ||
5817 | /* | |
5818 | * Return 1 if we should exit from L2 to L1 to handle an exit, or 0 if we | |
5819 | * should handle it ourselves in L0 (and then continue L2). Only call this | |
5820 | * when in is_guest_mode (L2). | |
5821 | */ | |
5822 | static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu) | |
5823 | { | |
5824 | u32 exit_reason = vmcs_read32(VM_EXIT_REASON); | |
5825 | u32 intr_info = vmcs_read32(VM_EXIT_INTR_INFO); | |
5826 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
5827 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); | |
5828 | ||
5829 | if (vmx->nested.nested_run_pending) | |
5830 | return 0; | |
5831 | ||
5832 | if (unlikely(vmx->fail)) { | |
bd80158a JK |
5833 | pr_info_ratelimited("%s failed vm entry %x\n", __func__, |
5834 | vmcs_read32(VM_INSTRUCTION_ERROR)); | |
644d711a NHE |
5835 | return 1; |
5836 | } | |
5837 | ||
5838 | switch (exit_reason) { | |
5839 | case EXIT_REASON_EXCEPTION_NMI: | |
5840 | if (!is_exception(intr_info)) | |
5841 | return 0; | |
5842 | else if (is_page_fault(intr_info)) | |
5843 | return enable_ept; | |
5844 | return vmcs12->exception_bitmap & | |
5845 | (1u << (intr_info & INTR_INFO_VECTOR_MASK)); | |
5846 | case EXIT_REASON_EXTERNAL_INTERRUPT: | |
5847 | return 0; | |
5848 | case EXIT_REASON_TRIPLE_FAULT: | |
5849 | return 1; | |
5850 | case EXIT_REASON_PENDING_INTERRUPT: | |
5851 | case EXIT_REASON_NMI_WINDOW: | |
5852 | /* | |
5853 | * prepare_vmcs02() set the CPU_BASED_VIRTUAL_INTR_PENDING bit | |
5854 | * (aka Interrupt Window Exiting) only when L1 turned it on, | |
5855 | * so if we got a PENDING_INTERRUPT exit, this must be for L1. | |
5856 | * Same for NMI Window Exiting. | |
5857 | */ | |
5858 | return 1; | |
5859 | case EXIT_REASON_TASK_SWITCH: | |
5860 | return 1; | |
5861 | case EXIT_REASON_CPUID: | |
5862 | return 1; | |
5863 | case EXIT_REASON_HLT: | |
5864 | return nested_cpu_has(vmcs12, CPU_BASED_HLT_EXITING); | |
5865 | case EXIT_REASON_INVD: | |
5866 | return 1; | |
5867 | case EXIT_REASON_INVLPG: | |
5868 | return nested_cpu_has(vmcs12, CPU_BASED_INVLPG_EXITING); | |
5869 | case EXIT_REASON_RDPMC: | |
5870 | return nested_cpu_has(vmcs12, CPU_BASED_RDPMC_EXITING); | |
5871 | case EXIT_REASON_RDTSC: | |
5872 | return nested_cpu_has(vmcs12, CPU_BASED_RDTSC_EXITING); | |
5873 | case EXIT_REASON_VMCALL: case EXIT_REASON_VMCLEAR: | |
5874 | case EXIT_REASON_VMLAUNCH: case EXIT_REASON_VMPTRLD: | |
5875 | case EXIT_REASON_VMPTRST: case EXIT_REASON_VMREAD: | |
5876 | case EXIT_REASON_VMRESUME: case EXIT_REASON_VMWRITE: | |
5877 | case EXIT_REASON_VMOFF: case EXIT_REASON_VMON: | |
5878 | /* | |
5879 | * VMX instructions trap unconditionally. This allows L1 to | |
5880 | * emulate them for its L2 guest, i.e., allows 3-level nesting! | |
5881 | */ | |
5882 | return 1; | |
5883 | case EXIT_REASON_CR_ACCESS: | |
5884 | return nested_vmx_exit_handled_cr(vcpu, vmcs12); | |
5885 | case EXIT_REASON_DR_ACCESS: | |
5886 | return nested_cpu_has(vmcs12, CPU_BASED_MOV_DR_EXITING); | |
5887 | case EXIT_REASON_IO_INSTRUCTION: | |
5888 | /* TODO: support IO bitmaps */ | |
5889 | return 1; | |
5890 | case EXIT_REASON_MSR_READ: | |
5891 | case EXIT_REASON_MSR_WRITE: | |
5892 | return nested_vmx_exit_handled_msr(vcpu, vmcs12, exit_reason); | |
5893 | case EXIT_REASON_INVALID_STATE: | |
5894 | return 1; | |
5895 | case EXIT_REASON_MWAIT_INSTRUCTION: | |
5896 | return nested_cpu_has(vmcs12, CPU_BASED_MWAIT_EXITING); | |
5897 | case EXIT_REASON_MONITOR_INSTRUCTION: | |
5898 | return nested_cpu_has(vmcs12, CPU_BASED_MONITOR_EXITING); | |
5899 | case EXIT_REASON_PAUSE_INSTRUCTION: | |
5900 | return nested_cpu_has(vmcs12, CPU_BASED_PAUSE_EXITING) || | |
5901 | nested_cpu_has2(vmcs12, | |
5902 | SECONDARY_EXEC_PAUSE_LOOP_EXITING); | |
5903 | case EXIT_REASON_MCE_DURING_VMENTRY: | |
5904 | return 0; | |
5905 | case EXIT_REASON_TPR_BELOW_THRESHOLD: | |
5906 | return 1; | |
5907 | case EXIT_REASON_APIC_ACCESS: | |
5908 | return nested_cpu_has2(vmcs12, | |
5909 | SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES); | |
5910 | case EXIT_REASON_EPT_VIOLATION: | |
5911 | case EXIT_REASON_EPT_MISCONFIG: | |
5912 | return 0; | |
5913 | case EXIT_REASON_WBINVD: | |
5914 | return nested_cpu_has2(vmcs12, SECONDARY_EXEC_WBINVD_EXITING); | |
5915 | case EXIT_REASON_XSETBV: | |
5916 | return 1; | |
5917 | default: | |
5918 | return 1; | |
5919 | } | |
5920 | } | |
5921 | ||
586f9607 AK |
5922 | static void vmx_get_exit_info(struct kvm_vcpu *vcpu, u64 *info1, u64 *info2) |
5923 | { | |
5924 | *info1 = vmcs_readl(EXIT_QUALIFICATION); | |
5925 | *info2 = vmcs_read32(VM_EXIT_INTR_INFO); | |
5926 | } | |
5927 | ||
6aa8b732 AK |
5928 | /* |
5929 | * The guest has exited. See if we can fix it or if we need userspace | |
5930 | * assistance. | |
5931 | */ | |
851ba692 | 5932 | static int vmx_handle_exit(struct kvm_vcpu *vcpu) |
6aa8b732 | 5933 | { |
29bd8a78 | 5934 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
a0861c02 | 5935 | u32 exit_reason = vmx->exit_reason; |
1155f76a | 5936 | u32 vectoring_info = vmx->idt_vectoring_info; |
29bd8a78 | 5937 | |
80ced186 MG |
5938 | /* If guest state is invalid, start emulating */ |
5939 | if (vmx->emulation_required && emulate_invalid_guest_state) | |
5940 | return handle_invalid_guest_state(vcpu); | |
1d5a4d9b | 5941 | |
b6f1250e NHE |
5942 | /* |
5943 | * the KVM_REQ_EVENT optimization bit is only on for one entry, and if | |
5944 | * we did not inject a still-pending event to L1 now because of | |
5945 | * nested_run_pending, we need to re-enable this bit. | |
5946 | */ | |
5947 | if (vmx->nested.nested_run_pending) | |
5948 | kvm_make_request(KVM_REQ_EVENT, vcpu); | |
5949 | ||
509c75ea NHE |
5950 | if (!is_guest_mode(vcpu) && (exit_reason == EXIT_REASON_VMLAUNCH || |
5951 | exit_reason == EXIT_REASON_VMRESUME)) | |
644d711a NHE |
5952 | vmx->nested.nested_run_pending = 1; |
5953 | else | |
5954 | vmx->nested.nested_run_pending = 0; | |
5955 | ||
5956 | if (is_guest_mode(vcpu) && nested_vmx_exit_handled(vcpu)) { | |
5957 | nested_vmx_vmexit(vcpu); | |
5958 | return 1; | |
5959 | } | |
5960 | ||
5120702e MG |
5961 | if (exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY) { |
5962 | vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; | |
5963 | vcpu->run->fail_entry.hardware_entry_failure_reason | |
5964 | = exit_reason; | |
5965 | return 0; | |
5966 | } | |
5967 | ||
29bd8a78 | 5968 | if (unlikely(vmx->fail)) { |
851ba692 AK |
5969 | vcpu->run->exit_reason = KVM_EXIT_FAIL_ENTRY; |
5970 | vcpu->run->fail_entry.hardware_entry_failure_reason | |
29bd8a78 AK |
5971 | = vmcs_read32(VM_INSTRUCTION_ERROR); |
5972 | return 0; | |
5973 | } | |
6aa8b732 | 5974 | |
d77c26fc | 5975 | if ((vectoring_info & VECTORING_INFO_VALID_MASK) && |
1439442c | 5976 | (exit_reason != EXIT_REASON_EXCEPTION_NMI && |
60637aac JK |
5977 | exit_reason != EXIT_REASON_EPT_VIOLATION && |
5978 | exit_reason != EXIT_REASON_TASK_SWITCH)) | |
5979 | printk(KERN_WARNING "%s: unexpected, valid vectoring info " | |
5980 | "(0x%x) and exit reason is 0x%x\n", | |
5981 | __func__, vectoring_info, exit_reason); | |
3b86cd99 | 5982 | |
644d711a NHE |
5983 | if (unlikely(!cpu_has_virtual_nmis() && vmx->soft_vnmi_blocked && |
5984 | !(is_guest_mode(vcpu) && nested_cpu_has_virtual_nmis( | |
5985 | get_vmcs12(vcpu), vcpu)))) { | |
c4282df9 | 5986 | if (vmx_interrupt_allowed(vcpu)) { |
3b86cd99 | 5987 | vmx->soft_vnmi_blocked = 0; |
3b86cd99 | 5988 | } else if (vmx->vnmi_blocked_time > 1000000000LL && |
4531220b | 5989 | vcpu->arch.nmi_pending) { |
3b86cd99 JK |
5990 | /* |
5991 | * This CPU don't support us in finding the end of an | |
5992 | * NMI-blocked window if the guest runs with IRQs | |
5993 | * disabled. So we pull the trigger after 1 s of | |
5994 | * futile waiting, but inform the user about this. | |
5995 | */ | |
5996 | printk(KERN_WARNING "%s: Breaking out of NMI-blocked " | |
5997 | "state on VCPU %d after 1 s timeout\n", | |
5998 | __func__, vcpu->vcpu_id); | |
5999 | vmx->soft_vnmi_blocked = 0; | |
3b86cd99 | 6000 | } |
3b86cd99 JK |
6001 | } |
6002 | ||
6aa8b732 AK |
6003 | if (exit_reason < kvm_vmx_max_exit_handlers |
6004 | && kvm_vmx_exit_handlers[exit_reason]) | |
851ba692 | 6005 | return kvm_vmx_exit_handlers[exit_reason](vcpu); |
6aa8b732 | 6006 | else { |
851ba692 AK |
6007 | vcpu->run->exit_reason = KVM_EXIT_UNKNOWN; |
6008 | vcpu->run->hw.hardware_exit_reason = exit_reason; | |
6aa8b732 AK |
6009 | } |
6010 | return 0; | |
6011 | } | |
6012 | ||
95ba8273 | 6013 | static void update_cr8_intercept(struct kvm_vcpu *vcpu, int tpr, int irr) |
6e5d865c | 6014 | { |
95ba8273 | 6015 | if (irr == -1 || tpr < irr) { |
6e5d865c YS |
6016 | vmcs_write32(TPR_THRESHOLD, 0); |
6017 | return; | |
6018 | } | |
6019 | ||
95ba8273 | 6020 | vmcs_write32(TPR_THRESHOLD, irr); |
6e5d865c YS |
6021 | } |
6022 | ||
51aa01d1 | 6023 | static void vmx_complete_atomic_exit(struct vcpu_vmx *vmx) |
cf393f75 | 6024 | { |
00eba012 AK |
6025 | u32 exit_intr_info; |
6026 | ||
6027 | if (!(vmx->exit_reason == EXIT_REASON_MCE_DURING_VMENTRY | |
6028 | || vmx->exit_reason == EXIT_REASON_EXCEPTION_NMI)) | |
6029 | return; | |
6030 | ||
c5ca8e57 | 6031 | vmx->exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); |
00eba012 | 6032 | exit_intr_info = vmx->exit_intr_info; |
a0861c02 AK |
6033 | |
6034 | /* Handle machine checks before interrupts are enabled */ | |
00eba012 | 6035 | if (is_machine_check(exit_intr_info)) |
a0861c02 AK |
6036 | kvm_machine_check(); |
6037 | ||
20f65983 | 6038 | /* We need to handle NMIs before interrupts are enabled */ |
00eba012 | 6039 | if ((exit_intr_info & INTR_INFO_INTR_TYPE_MASK) == INTR_TYPE_NMI_INTR && |
ff9d07a0 ZY |
6040 | (exit_intr_info & INTR_INFO_VALID_MASK)) { |
6041 | kvm_before_handle_nmi(&vmx->vcpu); | |
20f65983 | 6042 | asm("int $2"); |
ff9d07a0 ZY |
6043 | kvm_after_handle_nmi(&vmx->vcpu); |
6044 | } | |
51aa01d1 | 6045 | } |
20f65983 | 6046 | |
51aa01d1 AK |
6047 | static void vmx_recover_nmi_blocking(struct vcpu_vmx *vmx) |
6048 | { | |
c5ca8e57 | 6049 | u32 exit_intr_info; |
51aa01d1 AK |
6050 | bool unblock_nmi; |
6051 | u8 vector; | |
6052 | bool idtv_info_valid; | |
6053 | ||
6054 | idtv_info_valid = vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK; | |
20f65983 | 6055 | |
cf393f75 | 6056 | if (cpu_has_virtual_nmis()) { |
9d58b931 AK |
6057 | if (vmx->nmi_known_unmasked) |
6058 | return; | |
c5ca8e57 AK |
6059 | /* |
6060 | * Can't use vmx->exit_intr_info since we're not sure what | |
6061 | * the exit reason is. | |
6062 | */ | |
6063 | exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); | |
cf393f75 AK |
6064 | unblock_nmi = (exit_intr_info & INTR_INFO_UNBLOCK_NMI) != 0; |
6065 | vector = exit_intr_info & INTR_INFO_VECTOR_MASK; | |
6066 | /* | |
7b4a25cb | 6067 | * SDM 3: 27.7.1.2 (September 2008) |
cf393f75 AK |
6068 | * Re-set bit "block by NMI" before VM entry if vmexit caused by |
6069 | * a guest IRET fault. | |
7b4a25cb GN |
6070 | * SDM 3: 23.2.2 (September 2008) |
6071 | * Bit 12 is undefined in any of the following cases: | |
6072 | * If the VM exit sets the valid bit in the IDT-vectoring | |
6073 | * information field. | |
6074 | * If the VM exit is due to a double fault. | |
cf393f75 | 6075 | */ |
7b4a25cb GN |
6076 | if ((exit_intr_info & INTR_INFO_VALID_MASK) && unblock_nmi && |
6077 | vector != DF_VECTOR && !idtv_info_valid) | |
cf393f75 AK |
6078 | vmcs_set_bits(GUEST_INTERRUPTIBILITY_INFO, |
6079 | GUEST_INTR_STATE_NMI); | |
9d58b931 AK |
6080 | else |
6081 | vmx->nmi_known_unmasked = | |
6082 | !(vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) | |
6083 | & GUEST_INTR_STATE_NMI); | |
3b86cd99 JK |
6084 | } else if (unlikely(vmx->soft_vnmi_blocked)) |
6085 | vmx->vnmi_blocked_time += | |
6086 | ktime_to_ns(ktime_sub(ktime_get(), vmx->entry_time)); | |
51aa01d1 AK |
6087 | } |
6088 | ||
83422e17 AK |
6089 | static void __vmx_complete_interrupts(struct vcpu_vmx *vmx, |
6090 | u32 idt_vectoring_info, | |
6091 | int instr_len_field, | |
6092 | int error_code_field) | |
51aa01d1 | 6093 | { |
51aa01d1 AK |
6094 | u8 vector; |
6095 | int type; | |
6096 | bool idtv_info_valid; | |
6097 | ||
6098 | idtv_info_valid = idt_vectoring_info & VECTORING_INFO_VALID_MASK; | |
668f612f | 6099 | |
37b96e98 GN |
6100 | vmx->vcpu.arch.nmi_injected = false; |
6101 | kvm_clear_exception_queue(&vmx->vcpu); | |
6102 | kvm_clear_interrupt_queue(&vmx->vcpu); | |
6103 | ||
6104 | if (!idtv_info_valid) | |
6105 | return; | |
6106 | ||
3842d135 AK |
6107 | kvm_make_request(KVM_REQ_EVENT, &vmx->vcpu); |
6108 | ||
668f612f AK |
6109 | vector = idt_vectoring_info & VECTORING_INFO_VECTOR_MASK; |
6110 | type = idt_vectoring_info & VECTORING_INFO_TYPE_MASK; | |
37b96e98 | 6111 | |
64a7ec06 | 6112 | switch (type) { |
37b96e98 GN |
6113 | case INTR_TYPE_NMI_INTR: |
6114 | vmx->vcpu.arch.nmi_injected = true; | |
668f612f | 6115 | /* |
7b4a25cb | 6116 | * SDM 3: 27.7.1.2 (September 2008) |
37b96e98 GN |
6117 | * Clear bit "block by NMI" before VM entry if a NMI |
6118 | * delivery faulted. | |
668f612f | 6119 | */ |
654f06fc | 6120 | vmx_set_nmi_mask(&vmx->vcpu, false); |
37b96e98 | 6121 | break; |
37b96e98 | 6122 | case INTR_TYPE_SOFT_EXCEPTION: |
66fd3f7f | 6123 | vmx->vcpu.arch.event_exit_inst_len = |
83422e17 | 6124 | vmcs_read32(instr_len_field); |
66fd3f7f GN |
6125 | /* fall through */ |
6126 | case INTR_TYPE_HARD_EXCEPTION: | |
35920a35 | 6127 | if (idt_vectoring_info & VECTORING_INFO_DELIVER_CODE_MASK) { |
83422e17 | 6128 | u32 err = vmcs_read32(error_code_field); |
37b96e98 | 6129 | kvm_queue_exception_e(&vmx->vcpu, vector, err); |
35920a35 AK |
6130 | } else |
6131 | kvm_queue_exception(&vmx->vcpu, vector); | |
37b96e98 | 6132 | break; |
66fd3f7f GN |
6133 | case INTR_TYPE_SOFT_INTR: |
6134 | vmx->vcpu.arch.event_exit_inst_len = | |
83422e17 | 6135 | vmcs_read32(instr_len_field); |
66fd3f7f | 6136 | /* fall through */ |
37b96e98 | 6137 | case INTR_TYPE_EXT_INTR: |
66fd3f7f GN |
6138 | kvm_queue_interrupt(&vmx->vcpu, vector, |
6139 | type == INTR_TYPE_SOFT_INTR); | |
37b96e98 GN |
6140 | break; |
6141 | default: | |
6142 | break; | |
f7d9238f | 6143 | } |
cf393f75 AK |
6144 | } |
6145 | ||
83422e17 AK |
6146 | static void vmx_complete_interrupts(struct vcpu_vmx *vmx) |
6147 | { | |
66c78ae4 NHE |
6148 | if (is_guest_mode(&vmx->vcpu)) |
6149 | return; | |
83422e17 AK |
6150 | __vmx_complete_interrupts(vmx, vmx->idt_vectoring_info, |
6151 | VM_EXIT_INSTRUCTION_LEN, | |
6152 | IDT_VECTORING_ERROR_CODE); | |
6153 | } | |
6154 | ||
b463a6f7 AK |
6155 | static void vmx_cancel_injection(struct kvm_vcpu *vcpu) |
6156 | { | |
66c78ae4 NHE |
6157 | if (is_guest_mode(vcpu)) |
6158 | return; | |
b463a6f7 AK |
6159 | __vmx_complete_interrupts(to_vmx(vcpu), |
6160 | vmcs_read32(VM_ENTRY_INTR_INFO_FIELD), | |
6161 | VM_ENTRY_INSTRUCTION_LEN, | |
6162 | VM_ENTRY_EXCEPTION_ERROR_CODE); | |
6163 | ||
6164 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0); | |
6165 | } | |
6166 | ||
d7cd9796 GN |
6167 | static void atomic_switch_perf_msrs(struct vcpu_vmx *vmx) |
6168 | { | |
6169 | int i, nr_msrs; | |
6170 | struct perf_guest_switch_msr *msrs; | |
6171 | ||
6172 | msrs = perf_guest_get_msrs(&nr_msrs); | |
6173 | ||
6174 | if (!msrs) | |
6175 | return; | |
6176 | ||
6177 | for (i = 0; i < nr_msrs; i++) | |
6178 | if (msrs[i].host == msrs[i].guest) | |
6179 | clear_atomic_switch_msr(vmx, msrs[i].msr); | |
6180 | else | |
6181 | add_atomic_switch_msr(vmx, msrs[i].msr, msrs[i].guest, | |
6182 | msrs[i].host); | |
6183 | } | |
6184 | ||
c801949d AK |
6185 | #ifdef CONFIG_X86_64 |
6186 | #define R "r" | |
6187 | #define Q "q" | |
6188 | #else | |
6189 | #define R "e" | |
6190 | #define Q "l" | |
6191 | #endif | |
6192 | ||
a3b5ba49 | 6193 | static void __noclone vmx_vcpu_run(struct kvm_vcpu *vcpu) |
6aa8b732 | 6194 | { |
a2fa3e9f | 6195 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
2a7921b7 | 6196 | unsigned long debugctlmsr; |
104f226b | 6197 | |
66c78ae4 NHE |
6198 | if (is_guest_mode(vcpu) && !vmx->nested.nested_run_pending) { |
6199 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); | |
6200 | if (vmcs12->idt_vectoring_info_field & | |
6201 | VECTORING_INFO_VALID_MASK) { | |
6202 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, | |
6203 | vmcs12->idt_vectoring_info_field); | |
6204 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, | |
6205 | vmcs12->vm_exit_instruction_len); | |
6206 | if (vmcs12->idt_vectoring_info_field & | |
6207 | VECTORING_INFO_DELIVER_CODE_MASK) | |
6208 | vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, | |
6209 | vmcs12->idt_vectoring_error_code); | |
6210 | } | |
6211 | } | |
6212 | ||
104f226b AK |
6213 | /* Record the guest's net vcpu time for enforced NMI injections. */ |
6214 | if (unlikely(!cpu_has_virtual_nmis() && vmx->soft_vnmi_blocked)) | |
6215 | vmx->entry_time = ktime_get(); | |
6216 | ||
6217 | /* Don't enter VMX if guest state is invalid, let the exit handler | |
6218 | start emulation until we arrive back to a valid state */ | |
6219 | if (vmx->emulation_required && emulate_invalid_guest_state) | |
6220 | return; | |
6221 | ||
6222 | if (test_bit(VCPU_REGS_RSP, (unsigned long *)&vcpu->arch.regs_dirty)) | |
6223 | vmcs_writel(GUEST_RSP, vcpu->arch.regs[VCPU_REGS_RSP]); | |
6224 | if (test_bit(VCPU_REGS_RIP, (unsigned long *)&vcpu->arch.regs_dirty)) | |
6225 | vmcs_writel(GUEST_RIP, vcpu->arch.regs[VCPU_REGS_RIP]); | |
6226 | ||
6227 | /* When single-stepping over STI and MOV SS, we must clear the | |
6228 | * corresponding interruptibility bits in the guest state. Otherwise | |
6229 | * vmentry fails as it then expects bit 14 (BS) in pending debug | |
6230 | * exceptions being set, but that's not correct for the guest debugging | |
6231 | * case. */ | |
6232 | if (vcpu->guest_debug & KVM_GUESTDBG_SINGLESTEP) | |
6233 | vmx_set_interrupt_shadow(vcpu, 0); | |
6234 | ||
d7cd9796 | 6235 | atomic_switch_perf_msrs(vmx); |
2a7921b7 | 6236 | debugctlmsr = get_debugctlmsr(); |
d7cd9796 | 6237 | |
d462b819 | 6238 | vmx->__launched = vmx->loaded_vmcs->launched; |
104f226b | 6239 | asm( |
6aa8b732 | 6240 | /* Store host registers */ |
c801949d | 6241 | "push %%"R"dx; push %%"R"bp;" |
40712fae | 6242 | "push %%"R"cx \n\t" /* placeholder for guest rcx */ |
c801949d | 6243 | "push %%"R"cx \n\t" |
313dbd49 AK |
6244 | "cmp %%"R"sp, %c[host_rsp](%0) \n\t" |
6245 | "je 1f \n\t" | |
6246 | "mov %%"R"sp, %c[host_rsp](%0) \n\t" | |
4ecac3fd | 6247 | __ex(ASM_VMX_VMWRITE_RSP_RDX) "\n\t" |
313dbd49 | 6248 | "1: \n\t" |
d3edefc0 AK |
6249 | /* Reload cr2 if changed */ |
6250 | "mov %c[cr2](%0), %%"R"ax \n\t" | |
6251 | "mov %%cr2, %%"R"dx \n\t" | |
6252 | "cmp %%"R"ax, %%"R"dx \n\t" | |
6253 | "je 2f \n\t" | |
6254 | "mov %%"R"ax, %%cr2 \n\t" | |
6255 | "2: \n\t" | |
6aa8b732 | 6256 | /* Check if vmlaunch of vmresume is needed */ |
e08aa78a | 6257 | "cmpl $0, %c[launched](%0) \n\t" |
6aa8b732 | 6258 | /* Load guest registers. Don't clobber flags. */ |
c801949d AK |
6259 | "mov %c[rax](%0), %%"R"ax \n\t" |
6260 | "mov %c[rbx](%0), %%"R"bx \n\t" | |
6261 | "mov %c[rdx](%0), %%"R"dx \n\t" | |
6262 | "mov %c[rsi](%0), %%"R"si \n\t" | |
6263 | "mov %c[rdi](%0), %%"R"di \n\t" | |
6264 | "mov %c[rbp](%0), %%"R"bp \n\t" | |
05b3e0c2 | 6265 | #ifdef CONFIG_X86_64 |
e08aa78a AK |
6266 | "mov %c[r8](%0), %%r8 \n\t" |
6267 | "mov %c[r9](%0), %%r9 \n\t" | |
6268 | "mov %c[r10](%0), %%r10 \n\t" | |
6269 | "mov %c[r11](%0), %%r11 \n\t" | |
6270 | "mov %c[r12](%0), %%r12 \n\t" | |
6271 | "mov %c[r13](%0), %%r13 \n\t" | |
6272 | "mov %c[r14](%0), %%r14 \n\t" | |
6273 | "mov %c[r15](%0), %%r15 \n\t" | |
6aa8b732 | 6274 | #endif |
c801949d AK |
6275 | "mov %c[rcx](%0), %%"R"cx \n\t" /* kills %0 (ecx) */ |
6276 | ||
6aa8b732 | 6277 | /* Enter guest mode */ |
cd2276a7 | 6278 | "jne .Llaunched \n\t" |
4ecac3fd | 6279 | __ex(ASM_VMX_VMLAUNCH) "\n\t" |
cd2276a7 | 6280 | "jmp .Lkvm_vmx_return \n\t" |
4ecac3fd | 6281 | ".Llaunched: " __ex(ASM_VMX_VMRESUME) "\n\t" |
cd2276a7 | 6282 | ".Lkvm_vmx_return: " |
6aa8b732 | 6283 | /* Save guest registers, load host registers, keep flags */ |
40712fae AK |
6284 | "mov %0, %c[wordsize](%%"R"sp) \n\t" |
6285 | "pop %0 \n\t" | |
c801949d AK |
6286 | "mov %%"R"ax, %c[rax](%0) \n\t" |
6287 | "mov %%"R"bx, %c[rbx](%0) \n\t" | |
1c696d0e | 6288 | "pop"Q" %c[rcx](%0) \n\t" |
c801949d AK |
6289 | "mov %%"R"dx, %c[rdx](%0) \n\t" |
6290 | "mov %%"R"si, %c[rsi](%0) \n\t" | |
6291 | "mov %%"R"di, %c[rdi](%0) \n\t" | |
6292 | "mov %%"R"bp, %c[rbp](%0) \n\t" | |
05b3e0c2 | 6293 | #ifdef CONFIG_X86_64 |
e08aa78a AK |
6294 | "mov %%r8, %c[r8](%0) \n\t" |
6295 | "mov %%r9, %c[r9](%0) \n\t" | |
6296 | "mov %%r10, %c[r10](%0) \n\t" | |
6297 | "mov %%r11, %c[r11](%0) \n\t" | |
6298 | "mov %%r12, %c[r12](%0) \n\t" | |
6299 | "mov %%r13, %c[r13](%0) \n\t" | |
6300 | "mov %%r14, %c[r14](%0) \n\t" | |
6301 | "mov %%r15, %c[r15](%0) \n\t" | |
6aa8b732 | 6302 | #endif |
c801949d AK |
6303 | "mov %%cr2, %%"R"ax \n\t" |
6304 | "mov %%"R"ax, %c[cr2](%0) \n\t" | |
6305 | ||
1c696d0e | 6306 | "pop %%"R"bp; pop %%"R"dx \n\t" |
e08aa78a AK |
6307 | "setbe %c[fail](%0) \n\t" |
6308 | : : "c"(vmx), "d"((unsigned long)HOST_RSP), | |
d462b819 | 6309 | [launched]"i"(offsetof(struct vcpu_vmx, __launched)), |
e08aa78a | 6310 | [fail]"i"(offsetof(struct vcpu_vmx, fail)), |
313dbd49 | 6311 | [host_rsp]"i"(offsetof(struct vcpu_vmx, host_rsp)), |
ad312c7c ZX |
6312 | [rax]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RAX])), |
6313 | [rbx]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RBX])), | |
6314 | [rcx]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RCX])), | |
6315 | [rdx]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RDX])), | |
6316 | [rsi]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RSI])), | |
6317 | [rdi]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RDI])), | |
6318 | [rbp]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_RBP])), | |
05b3e0c2 | 6319 | #ifdef CONFIG_X86_64 |
ad312c7c ZX |
6320 | [r8]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R8])), |
6321 | [r9]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R9])), | |
6322 | [r10]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R10])), | |
6323 | [r11]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R11])), | |
6324 | [r12]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R12])), | |
6325 | [r13]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R13])), | |
6326 | [r14]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R14])), | |
6327 | [r15]"i"(offsetof(struct vcpu_vmx, vcpu.arch.regs[VCPU_REGS_R15])), | |
6aa8b732 | 6328 | #endif |
40712fae AK |
6329 | [cr2]"i"(offsetof(struct vcpu_vmx, vcpu.arch.cr2)), |
6330 | [wordsize]"i"(sizeof(ulong)) | |
c2036300 | 6331 | : "cc", "memory" |
07d6f555 | 6332 | , R"ax", R"bx", R"di", R"si" |
c2036300 | 6333 | #ifdef CONFIG_X86_64 |
c2036300 LV |
6334 | , "r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15" |
6335 | #endif | |
6336 | ); | |
6aa8b732 | 6337 | |
2a7921b7 GN |
6338 | /* MSR_IA32_DEBUGCTLMSR is zeroed on vmexit. Restore it if needed */ |
6339 | if (debugctlmsr) | |
6340 | update_debugctlmsr(debugctlmsr); | |
6341 | ||
aa67f609 AK |
6342 | #ifndef CONFIG_X86_64 |
6343 | /* | |
6344 | * The sysexit path does not restore ds/es, so we must set them to | |
6345 | * a reasonable value ourselves. | |
6346 | * | |
6347 | * We can't defer this to vmx_load_host_state() since that function | |
6348 | * may be executed in interrupt context, which saves and restore segments | |
6349 | * around it, nullifying its effect. | |
6350 | */ | |
6351 | loadsegment(ds, __USER_DS); | |
6352 | loadsegment(es, __USER_DS); | |
6353 | #endif | |
6354 | ||
6de4f3ad | 6355 | vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP) |
6de12732 | 6356 | | (1 << VCPU_EXREG_RFLAGS) |
69c73028 | 6357 | | (1 << VCPU_EXREG_CPL) |
aff48baa | 6358 | | (1 << VCPU_EXREG_PDPTR) |
2fb92db1 | 6359 | | (1 << VCPU_EXREG_SEGMENTS) |
aff48baa | 6360 | | (1 << VCPU_EXREG_CR3)); |
5fdbf976 MT |
6361 | vcpu->arch.regs_dirty = 0; |
6362 | ||
1155f76a AK |
6363 | vmx->idt_vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD); |
6364 | ||
66c78ae4 NHE |
6365 | if (is_guest_mode(vcpu)) { |
6366 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); | |
6367 | vmcs12->idt_vectoring_info_field = vmx->idt_vectoring_info; | |
6368 | if (vmx->idt_vectoring_info & VECTORING_INFO_VALID_MASK) { | |
6369 | vmcs12->idt_vectoring_error_code = | |
6370 | vmcs_read32(IDT_VECTORING_ERROR_CODE); | |
6371 | vmcs12->vm_exit_instruction_len = | |
6372 | vmcs_read32(VM_EXIT_INSTRUCTION_LEN); | |
6373 | } | |
6374 | } | |
6375 | ||
d462b819 | 6376 | vmx->loaded_vmcs->launched = 1; |
1b6269db | 6377 | |
51aa01d1 | 6378 | vmx->exit_reason = vmcs_read32(VM_EXIT_REASON); |
1e2b1dd7 | 6379 | trace_kvm_exit(vmx->exit_reason, vcpu, KVM_ISA_VMX); |
51aa01d1 AK |
6380 | |
6381 | vmx_complete_atomic_exit(vmx); | |
6382 | vmx_recover_nmi_blocking(vmx); | |
cf393f75 | 6383 | vmx_complete_interrupts(vmx); |
6aa8b732 AK |
6384 | } |
6385 | ||
c801949d AK |
6386 | #undef R |
6387 | #undef Q | |
6388 | ||
6aa8b732 AK |
6389 | static void vmx_free_vcpu(struct kvm_vcpu *vcpu) |
6390 | { | |
fb3f0f51 RR |
6391 | struct vcpu_vmx *vmx = to_vmx(vcpu); |
6392 | ||
cdbecfc3 | 6393 | free_vpid(vmx); |
ec378aee | 6394 | free_nested(vmx); |
d462b819 | 6395 | free_loaded_vmcs(vmx->loaded_vmcs); |
fb3f0f51 RR |
6396 | kfree(vmx->guest_msrs); |
6397 | kvm_vcpu_uninit(vcpu); | |
a4770347 | 6398 | kmem_cache_free(kvm_vcpu_cache, vmx); |
6aa8b732 AK |
6399 | } |
6400 | ||
fb3f0f51 | 6401 | static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id) |
6aa8b732 | 6402 | { |
fb3f0f51 | 6403 | int err; |
c16f862d | 6404 | struct vcpu_vmx *vmx = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL); |
15ad7146 | 6405 | int cpu; |
6aa8b732 | 6406 | |
a2fa3e9f | 6407 | if (!vmx) |
fb3f0f51 RR |
6408 | return ERR_PTR(-ENOMEM); |
6409 | ||
2384d2b3 SY |
6410 | allocate_vpid(vmx); |
6411 | ||
fb3f0f51 RR |
6412 | err = kvm_vcpu_init(&vmx->vcpu, kvm, id); |
6413 | if (err) | |
6414 | goto free_vcpu; | |
965b58a5 | 6415 | |
a2fa3e9f | 6416 | vmx->guest_msrs = kmalloc(PAGE_SIZE, GFP_KERNEL); |
be6d05cf | 6417 | err = -ENOMEM; |
fb3f0f51 | 6418 | if (!vmx->guest_msrs) { |
fb3f0f51 RR |
6419 | goto uninit_vcpu; |
6420 | } | |
965b58a5 | 6421 | |
d462b819 NHE |
6422 | vmx->loaded_vmcs = &vmx->vmcs01; |
6423 | vmx->loaded_vmcs->vmcs = alloc_vmcs(); | |
6424 | if (!vmx->loaded_vmcs->vmcs) | |
fb3f0f51 | 6425 | goto free_msrs; |
d462b819 NHE |
6426 | if (!vmm_exclusive) |
6427 | kvm_cpu_vmxon(__pa(per_cpu(vmxarea, raw_smp_processor_id()))); | |
6428 | loaded_vmcs_init(vmx->loaded_vmcs); | |
6429 | if (!vmm_exclusive) | |
6430 | kvm_cpu_vmxoff(); | |
a2fa3e9f | 6431 | |
15ad7146 AK |
6432 | cpu = get_cpu(); |
6433 | vmx_vcpu_load(&vmx->vcpu, cpu); | |
e48672fa | 6434 | vmx->vcpu.cpu = cpu; |
8b9cf98c | 6435 | err = vmx_vcpu_setup(vmx); |
fb3f0f51 | 6436 | vmx_vcpu_put(&vmx->vcpu); |
15ad7146 | 6437 | put_cpu(); |
fb3f0f51 RR |
6438 | if (err) |
6439 | goto free_vmcs; | |
5e4a0b3c | 6440 | if (vm_need_virtualize_apic_accesses(kvm)) |
be6d05cf JK |
6441 | err = alloc_apic_access_page(kvm); |
6442 | if (err) | |
5e4a0b3c | 6443 | goto free_vmcs; |
fb3f0f51 | 6444 | |
b927a3ce SY |
6445 | if (enable_ept) { |
6446 | if (!kvm->arch.ept_identity_map_addr) | |
6447 | kvm->arch.ept_identity_map_addr = | |
6448 | VMX_EPT_IDENTITY_PAGETABLE_ADDR; | |
93ea5388 | 6449 | err = -ENOMEM; |
b7ebfb05 SY |
6450 | if (alloc_identity_pagetable(kvm) != 0) |
6451 | goto free_vmcs; | |
93ea5388 GN |
6452 | if (!init_rmode_identity_map(kvm)) |
6453 | goto free_vmcs; | |
b927a3ce | 6454 | } |
b7ebfb05 | 6455 | |
a9d30f33 NHE |
6456 | vmx->nested.current_vmptr = -1ull; |
6457 | vmx->nested.current_vmcs12 = NULL; | |
6458 | ||
fb3f0f51 RR |
6459 | return &vmx->vcpu; |
6460 | ||
6461 | free_vmcs: | |
5f3fbc34 | 6462 | free_loaded_vmcs(vmx->loaded_vmcs); |
fb3f0f51 | 6463 | free_msrs: |
fb3f0f51 RR |
6464 | kfree(vmx->guest_msrs); |
6465 | uninit_vcpu: | |
6466 | kvm_vcpu_uninit(&vmx->vcpu); | |
6467 | free_vcpu: | |
cdbecfc3 | 6468 | free_vpid(vmx); |
a4770347 | 6469 | kmem_cache_free(kvm_vcpu_cache, vmx); |
fb3f0f51 | 6470 | return ERR_PTR(err); |
6aa8b732 AK |
6471 | } |
6472 | ||
002c7f7c YS |
6473 | static void __init vmx_check_processor_compat(void *rtn) |
6474 | { | |
6475 | struct vmcs_config vmcs_conf; | |
6476 | ||
6477 | *(int *)rtn = 0; | |
6478 | if (setup_vmcs_config(&vmcs_conf) < 0) | |
6479 | *(int *)rtn = -EIO; | |
6480 | if (memcmp(&vmcs_config, &vmcs_conf, sizeof(struct vmcs_config)) != 0) { | |
6481 | printk(KERN_ERR "kvm: CPU %d feature inconsistency!\n", | |
6482 | smp_processor_id()); | |
6483 | *(int *)rtn = -EIO; | |
6484 | } | |
6485 | } | |
6486 | ||
67253af5 SY |
6487 | static int get_ept_level(void) |
6488 | { | |
6489 | return VMX_EPT_DEFAULT_GAW + 1; | |
6490 | } | |
6491 | ||
4b12f0de | 6492 | static u64 vmx_get_mt_mask(struct kvm_vcpu *vcpu, gfn_t gfn, bool is_mmio) |
64d4d521 | 6493 | { |
4b12f0de SY |
6494 | u64 ret; |
6495 | ||
522c68c4 SY |
6496 | /* For VT-d and EPT combination |
6497 | * 1. MMIO: always map as UC | |
6498 | * 2. EPT with VT-d: | |
6499 | * a. VT-d without snooping control feature: can't guarantee the | |
6500 | * result, try to trust guest. | |
6501 | * b. VT-d with snooping control feature: snooping control feature of | |
6502 | * VT-d engine can guarantee the cache correctness. Just set it | |
6503 | * to WB to keep consistent with host. So the same as item 3. | |
a19a6d11 | 6504 | * 3. EPT without VT-d: always map as WB and set IPAT=1 to keep |
522c68c4 SY |
6505 | * consistent with host MTRR |
6506 | */ | |
4b12f0de SY |
6507 | if (is_mmio) |
6508 | ret = MTRR_TYPE_UNCACHABLE << VMX_EPT_MT_EPTE_SHIFT; | |
522c68c4 SY |
6509 | else if (vcpu->kvm->arch.iommu_domain && |
6510 | !(vcpu->kvm->arch.iommu_flags & KVM_IOMMU_CACHE_COHERENCY)) | |
6511 | ret = kvm_get_guest_memory_type(vcpu, gfn) << | |
6512 | VMX_EPT_MT_EPTE_SHIFT; | |
4b12f0de | 6513 | else |
522c68c4 | 6514 | ret = (MTRR_TYPE_WRBACK << VMX_EPT_MT_EPTE_SHIFT) |
a19a6d11 | 6515 | | VMX_EPT_IPAT_BIT; |
4b12f0de SY |
6516 | |
6517 | return ret; | |
64d4d521 SY |
6518 | } |
6519 | ||
17cc3935 | 6520 | static int vmx_get_lpage_level(void) |
344f414f | 6521 | { |
878403b7 SY |
6522 | if (enable_ept && !cpu_has_vmx_ept_1g_page()) |
6523 | return PT_DIRECTORY_LEVEL; | |
6524 | else | |
6525 | /* For shadow and EPT supported 1GB page */ | |
6526 | return PT_PDPE_LEVEL; | |
344f414f JR |
6527 | } |
6528 | ||
0e851880 SY |
6529 | static void vmx_cpuid_update(struct kvm_vcpu *vcpu) |
6530 | { | |
4e47c7a6 SY |
6531 | struct kvm_cpuid_entry2 *best; |
6532 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
6533 | u32 exec_control; | |
6534 | ||
6535 | vmx->rdtscp_enabled = false; | |
6536 | if (vmx_rdtscp_supported()) { | |
6537 | exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL); | |
6538 | if (exec_control & SECONDARY_EXEC_RDTSCP) { | |
6539 | best = kvm_find_cpuid_entry(vcpu, 0x80000001, 0); | |
6540 | if (best && (best->edx & bit(X86_FEATURE_RDTSCP))) | |
6541 | vmx->rdtscp_enabled = true; | |
6542 | else { | |
6543 | exec_control &= ~SECONDARY_EXEC_RDTSCP; | |
6544 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, | |
6545 | exec_control); | |
6546 | } | |
6547 | } | |
6548 | } | |
ad756a16 MJ |
6549 | |
6550 | exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL); | |
6551 | /* Exposing INVPCID only when PCID is exposed */ | |
6552 | best = kvm_find_cpuid_entry(vcpu, 0x7, 0); | |
6553 | if (vmx_invpcid_supported() && | |
6554 | best && (best->ecx & bit(X86_FEATURE_INVPCID)) && | |
6555 | guest_cpuid_has_pcid(vcpu)) { | |
6556 | exec_control |= SECONDARY_EXEC_ENABLE_INVPCID; | |
6557 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, | |
6558 | exec_control); | |
6559 | } else { | |
6560 | exec_control &= ~SECONDARY_EXEC_ENABLE_INVPCID; | |
6561 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, | |
6562 | exec_control); | |
6563 | if (best) | |
6564 | best->ecx &= ~bit(X86_FEATURE_INVPCID); | |
6565 | } | |
0e851880 SY |
6566 | } |
6567 | ||
d4330ef2 JR |
6568 | static void vmx_set_supported_cpuid(u32 func, struct kvm_cpuid_entry2 *entry) |
6569 | { | |
7b8050f5 NHE |
6570 | if (func == 1 && nested) |
6571 | entry->ecx |= bit(X86_FEATURE_VMX); | |
d4330ef2 JR |
6572 | } |
6573 | ||
fe3ef05c NHE |
6574 | /* |
6575 | * prepare_vmcs02 is called when the L1 guest hypervisor runs its nested | |
6576 | * L2 guest. L1 has a vmcs for L2 (vmcs12), and this function "merges" it | |
6577 | * with L0's requirements for its guest (a.k.a. vmsc01), so we can run the L2 | |
6578 | * guest in a way that will both be appropriate to L1's requests, and our | |
6579 | * needs. In addition to modifying the active vmcs (which is vmcs02), this | |
6580 | * function also has additional necessary side-effects, like setting various | |
6581 | * vcpu->arch fields. | |
6582 | */ | |
6583 | static void prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12) | |
6584 | { | |
6585 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
6586 | u32 exec_control; | |
6587 | ||
6588 | vmcs_write16(GUEST_ES_SELECTOR, vmcs12->guest_es_selector); | |
6589 | vmcs_write16(GUEST_CS_SELECTOR, vmcs12->guest_cs_selector); | |
6590 | vmcs_write16(GUEST_SS_SELECTOR, vmcs12->guest_ss_selector); | |
6591 | vmcs_write16(GUEST_DS_SELECTOR, vmcs12->guest_ds_selector); | |
6592 | vmcs_write16(GUEST_FS_SELECTOR, vmcs12->guest_fs_selector); | |
6593 | vmcs_write16(GUEST_GS_SELECTOR, vmcs12->guest_gs_selector); | |
6594 | vmcs_write16(GUEST_LDTR_SELECTOR, vmcs12->guest_ldtr_selector); | |
6595 | vmcs_write16(GUEST_TR_SELECTOR, vmcs12->guest_tr_selector); | |
6596 | vmcs_write32(GUEST_ES_LIMIT, vmcs12->guest_es_limit); | |
6597 | vmcs_write32(GUEST_CS_LIMIT, vmcs12->guest_cs_limit); | |
6598 | vmcs_write32(GUEST_SS_LIMIT, vmcs12->guest_ss_limit); | |
6599 | vmcs_write32(GUEST_DS_LIMIT, vmcs12->guest_ds_limit); | |
6600 | vmcs_write32(GUEST_FS_LIMIT, vmcs12->guest_fs_limit); | |
6601 | vmcs_write32(GUEST_GS_LIMIT, vmcs12->guest_gs_limit); | |
6602 | vmcs_write32(GUEST_LDTR_LIMIT, vmcs12->guest_ldtr_limit); | |
6603 | vmcs_write32(GUEST_TR_LIMIT, vmcs12->guest_tr_limit); | |
6604 | vmcs_write32(GUEST_GDTR_LIMIT, vmcs12->guest_gdtr_limit); | |
6605 | vmcs_write32(GUEST_IDTR_LIMIT, vmcs12->guest_idtr_limit); | |
6606 | vmcs_write32(GUEST_ES_AR_BYTES, vmcs12->guest_es_ar_bytes); | |
6607 | vmcs_write32(GUEST_CS_AR_BYTES, vmcs12->guest_cs_ar_bytes); | |
6608 | vmcs_write32(GUEST_SS_AR_BYTES, vmcs12->guest_ss_ar_bytes); | |
6609 | vmcs_write32(GUEST_DS_AR_BYTES, vmcs12->guest_ds_ar_bytes); | |
6610 | vmcs_write32(GUEST_FS_AR_BYTES, vmcs12->guest_fs_ar_bytes); | |
6611 | vmcs_write32(GUEST_GS_AR_BYTES, vmcs12->guest_gs_ar_bytes); | |
6612 | vmcs_write32(GUEST_LDTR_AR_BYTES, vmcs12->guest_ldtr_ar_bytes); | |
6613 | vmcs_write32(GUEST_TR_AR_BYTES, vmcs12->guest_tr_ar_bytes); | |
6614 | vmcs_writel(GUEST_ES_BASE, vmcs12->guest_es_base); | |
6615 | vmcs_writel(GUEST_CS_BASE, vmcs12->guest_cs_base); | |
6616 | vmcs_writel(GUEST_SS_BASE, vmcs12->guest_ss_base); | |
6617 | vmcs_writel(GUEST_DS_BASE, vmcs12->guest_ds_base); | |
6618 | vmcs_writel(GUEST_FS_BASE, vmcs12->guest_fs_base); | |
6619 | vmcs_writel(GUEST_GS_BASE, vmcs12->guest_gs_base); | |
6620 | vmcs_writel(GUEST_LDTR_BASE, vmcs12->guest_ldtr_base); | |
6621 | vmcs_writel(GUEST_TR_BASE, vmcs12->guest_tr_base); | |
6622 | vmcs_writel(GUEST_GDTR_BASE, vmcs12->guest_gdtr_base); | |
6623 | vmcs_writel(GUEST_IDTR_BASE, vmcs12->guest_idtr_base); | |
6624 | ||
6625 | vmcs_write64(GUEST_IA32_DEBUGCTL, vmcs12->guest_ia32_debugctl); | |
6626 | vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, | |
6627 | vmcs12->vm_entry_intr_info_field); | |
6628 | vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, | |
6629 | vmcs12->vm_entry_exception_error_code); | |
6630 | vmcs_write32(VM_ENTRY_INSTRUCTION_LEN, | |
6631 | vmcs12->vm_entry_instruction_len); | |
6632 | vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, | |
6633 | vmcs12->guest_interruptibility_info); | |
6634 | vmcs_write32(GUEST_ACTIVITY_STATE, vmcs12->guest_activity_state); | |
6635 | vmcs_write32(GUEST_SYSENTER_CS, vmcs12->guest_sysenter_cs); | |
6636 | vmcs_writel(GUEST_DR7, vmcs12->guest_dr7); | |
6637 | vmcs_writel(GUEST_RFLAGS, vmcs12->guest_rflags); | |
6638 | vmcs_writel(GUEST_PENDING_DBG_EXCEPTIONS, | |
6639 | vmcs12->guest_pending_dbg_exceptions); | |
6640 | vmcs_writel(GUEST_SYSENTER_ESP, vmcs12->guest_sysenter_esp); | |
6641 | vmcs_writel(GUEST_SYSENTER_EIP, vmcs12->guest_sysenter_eip); | |
6642 | ||
6643 | vmcs_write64(VMCS_LINK_POINTER, -1ull); | |
6644 | ||
6645 | vmcs_write32(PIN_BASED_VM_EXEC_CONTROL, | |
6646 | (vmcs_config.pin_based_exec_ctrl | | |
6647 | vmcs12->pin_based_vm_exec_control)); | |
6648 | ||
6649 | /* | |
6650 | * Whether page-faults are trapped is determined by a combination of | |
6651 | * 3 settings: PFEC_MASK, PFEC_MATCH and EXCEPTION_BITMAP.PF. | |
6652 | * If enable_ept, L0 doesn't care about page faults and we should | |
6653 | * set all of these to L1's desires. However, if !enable_ept, L0 does | |
6654 | * care about (at least some) page faults, and because it is not easy | |
6655 | * (if at all possible?) to merge L0 and L1's desires, we simply ask | |
6656 | * to exit on each and every L2 page fault. This is done by setting | |
6657 | * MASK=MATCH=0 and (see below) EB.PF=1. | |
6658 | * Note that below we don't need special code to set EB.PF beyond the | |
6659 | * "or"ing of the EB of vmcs01 and vmcs12, because when enable_ept, | |
6660 | * vmcs01's EB.PF is 0 so the "or" will take vmcs12's value, and when | |
6661 | * !enable_ept, EB.PF is 1, so the "or" will always be 1. | |
6662 | * | |
6663 | * A problem with this approach (when !enable_ept) is that L1 may be | |
6664 | * injected with more page faults than it asked for. This could have | |
6665 | * caused problems, but in practice existing hypervisors don't care. | |
6666 | * To fix this, we will need to emulate the PFEC checking (on the L1 | |
6667 | * page tables), using walk_addr(), when injecting PFs to L1. | |
6668 | */ | |
6669 | vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, | |
6670 | enable_ept ? vmcs12->page_fault_error_code_mask : 0); | |
6671 | vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, | |
6672 | enable_ept ? vmcs12->page_fault_error_code_match : 0); | |
6673 | ||
6674 | if (cpu_has_secondary_exec_ctrls()) { | |
6675 | u32 exec_control = vmx_secondary_exec_control(vmx); | |
6676 | if (!vmx->rdtscp_enabled) | |
6677 | exec_control &= ~SECONDARY_EXEC_RDTSCP; | |
6678 | /* Take the following fields only from vmcs12 */ | |
6679 | exec_control &= ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
6680 | if (nested_cpu_has(vmcs12, | |
6681 | CPU_BASED_ACTIVATE_SECONDARY_CONTROLS)) | |
6682 | exec_control |= vmcs12->secondary_vm_exec_control; | |
6683 | ||
6684 | if (exec_control & SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES) { | |
6685 | /* | |
6686 | * Translate L1 physical address to host physical | |
6687 | * address for vmcs02. Keep the page pinned, so this | |
6688 | * physical address remains valid. We keep a reference | |
6689 | * to it so we can release it later. | |
6690 | */ | |
6691 | if (vmx->nested.apic_access_page) /* shouldn't happen */ | |
6692 | nested_release_page(vmx->nested.apic_access_page); | |
6693 | vmx->nested.apic_access_page = | |
6694 | nested_get_page(vcpu, vmcs12->apic_access_addr); | |
6695 | /* | |
6696 | * If translation failed, no matter: This feature asks | |
6697 | * to exit when accessing the given address, and if it | |
6698 | * can never be accessed, this feature won't do | |
6699 | * anything anyway. | |
6700 | */ | |
6701 | if (!vmx->nested.apic_access_page) | |
6702 | exec_control &= | |
6703 | ~SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES; | |
6704 | else | |
6705 | vmcs_write64(APIC_ACCESS_ADDR, | |
6706 | page_to_phys(vmx->nested.apic_access_page)); | |
6707 | } | |
6708 | ||
6709 | vmcs_write32(SECONDARY_VM_EXEC_CONTROL, exec_control); | |
6710 | } | |
6711 | ||
6712 | ||
6713 | /* | |
6714 | * Set host-state according to L0's settings (vmcs12 is irrelevant here) | |
6715 | * Some constant fields are set here by vmx_set_constant_host_state(). | |
6716 | * Other fields are different per CPU, and will be set later when | |
6717 | * vmx_vcpu_load() is called, and when vmx_save_host_state() is called. | |
6718 | */ | |
6719 | vmx_set_constant_host_state(); | |
6720 | ||
6721 | /* | |
6722 | * HOST_RSP is normally set correctly in vmx_vcpu_run() just before | |
6723 | * entry, but only if the current (host) sp changed from the value | |
6724 | * we wrote last (vmx->host_rsp). This cache is no longer relevant | |
6725 | * if we switch vmcs, and rather than hold a separate cache per vmcs, | |
6726 | * here we just force the write to happen on entry. | |
6727 | */ | |
6728 | vmx->host_rsp = 0; | |
6729 | ||
6730 | exec_control = vmx_exec_control(vmx); /* L0's desires */ | |
6731 | exec_control &= ~CPU_BASED_VIRTUAL_INTR_PENDING; | |
6732 | exec_control &= ~CPU_BASED_VIRTUAL_NMI_PENDING; | |
6733 | exec_control &= ~CPU_BASED_TPR_SHADOW; | |
6734 | exec_control |= vmcs12->cpu_based_vm_exec_control; | |
6735 | /* | |
6736 | * Merging of IO and MSR bitmaps not currently supported. | |
6737 | * Rather, exit every time. | |
6738 | */ | |
6739 | exec_control &= ~CPU_BASED_USE_MSR_BITMAPS; | |
6740 | exec_control &= ~CPU_BASED_USE_IO_BITMAPS; | |
6741 | exec_control |= CPU_BASED_UNCOND_IO_EXITING; | |
6742 | ||
6743 | vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, exec_control); | |
6744 | ||
6745 | /* EXCEPTION_BITMAP and CR0_GUEST_HOST_MASK should basically be the | |
6746 | * bitwise-or of what L1 wants to trap for L2, and what we want to | |
6747 | * trap. Note that CR0.TS also needs updating - we do this later. | |
6748 | */ | |
6749 | update_exception_bitmap(vcpu); | |
6750 | vcpu->arch.cr0_guest_owned_bits &= ~vmcs12->cr0_guest_host_mask; | |
6751 | vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits); | |
6752 | ||
6753 | /* Note: IA32_MODE, LOAD_IA32_EFER are modified by vmx_set_efer below */ | |
6754 | vmcs_write32(VM_EXIT_CONTROLS, | |
6755 | vmcs12->vm_exit_controls | vmcs_config.vmexit_ctrl); | |
6756 | vmcs_write32(VM_ENTRY_CONTROLS, vmcs12->vm_entry_controls | | |
6757 | (vmcs_config.vmentry_ctrl & ~VM_ENTRY_IA32E_MODE)); | |
6758 | ||
6759 | if (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_PAT) | |
6760 | vmcs_write64(GUEST_IA32_PAT, vmcs12->guest_ia32_pat); | |
6761 | else if (vmcs_config.vmentry_ctrl & VM_ENTRY_LOAD_IA32_PAT) | |
6762 | vmcs_write64(GUEST_IA32_PAT, vmx->vcpu.arch.pat); | |
6763 | ||
6764 | ||
6765 | set_cr4_guest_host_mask(vmx); | |
6766 | ||
27fc51b2 NHE |
6767 | if (vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_TSC_OFFSETING) |
6768 | vmcs_write64(TSC_OFFSET, | |
6769 | vmx->nested.vmcs01_tsc_offset + vmcs12->tsc_offset); | |
6770 | else | |
6771 | vmcs_write64(TSC_OFFSET, vmx->nested.vmcs01_tsc_offset); | |
fe3ef05c NHE |
6772 | |
6773 | if (enable_vpid) { | |
6774 | /* | |
6775 | * Trivially support vpid by letting L2s share their parent | |
6776 | * L1's vpid. TODO: move to a more elaborate solution, giving | |
6777 | * each L2 its own vpid and exposing the vpid feature to L1. | |
6778 | */ | |
6779 | vmcs_write16(VIRTUAL_PROCESSOR_ID, vmx->vpid); | |
6780 | vmx_flush_tlb(vcpu); | |
6781 | } | |
6782 | ||
6783 | if (vmcs12->vm_entry_controls & VM_ENTRY_LOAD_IA32_EFER) | |
6784 | vcpu->arch.efer = vmcs12->guest_ia32_efer; | |
6785 | if (vmcs12->vm_entry_controls & VM_ENTRY_IA32E_MODE) | |
6786 | vcpu->arch.efer |= (EFER_LMA | EFER_LME); | |
6787 | else | |
6788 | vcpu->arch.efer &= ~(EFER_LMA | EFER_LME); | |
6789 | /* Note: modifies VM_ENTRY/EXIT_CONTROLS and GUEST/HOST_IA32_EFER */ | |
6790 | vmx_set_efer(vcpu, vcpu->arch.efer); | |
6791 | ||
6792 | /* | |
6793 | * This sets GUEST_CR0 to vmcs12->guest_cr0, with possibly a modified | |
6794 | * TS bit (for lazy fpu) and bits which we consider mandatory enabled. | |
6795 | * The CR0_READ_SHADOW is what L2 should have expected to read given | |
6796 | * the specifications by L1; It's not enough to take | |
6797 | * vmcs12->cr0_read_shadow because on our cr0_guest_host_mask we we | |
6798 | * have more bits than L1 expected. | |
6799 | */ | |
6800 | vmx_set_cr0(vcpu, vmcs12->guest_cr0); | |
6801 | vmcs_writel(CR0_READ_SHADOW, nested_read_cr0(vmcs12)); | |
6802 | ||
6803 | vmx_set_cr4(vcpu, vmcs12->guest_cr4); | |
6804 | vmcs_writel(CR4_READ_SHADOW, nested_read_cr4(vmcs12)); | |
6805 | ||
6806 | /* shadow page tables on either EPT or shadow page tables */ | |
6807 | kvm_set_cr3(vcpu, vmcs12->guest_cr3); | |
6808 | kvm_mmu_reset_context(vcpu); | |
6809 | ||
6810 | kvm_register_write(vcpu, VCPU_REGS_RSP, vmcs12->guest_rsp); | |
6811 | kvm_register_write(vcpu, VCPU_REGS_RIP, vmcs12->guest_rip); | |
6812 | } | |
6813 | ||
cd232ad0 NHE |
6814 | /* |
6815 | * nested_vmx_run() handles a nested entry, i.e., a VMLAUNCH or VMRESUME on L1 | |
6816 | * for running an L2 nested guest. | |
6817 | */ | |
6818 | static int nested_vmx_run(struct kvm_vcpu *vcpu, bool launch) | |
6819 | { | |
6820 | struct vmcs12 *vmcs12; | |
6821 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
6822 | int cpu; | |
6823 | struct loaded_vmcs *vmcs02; | |
6824 | ||
6825 | if (!nested_vmx_check_permission(vcpu) || | |
6826 | !nested_vmx_check_vmcs12(vcpu)) | |
6827 | return 1; | |
6828 | ||
6829 | skip_emulated_instruction(vcpu); | |
6830 | vmcs12 = get_vmcs12(vcpu); | |
6831 | ||
7c177938 NHE |
6832 | /* |
6833 | * The nested entry process starts with enforcing various prerequisites | |
6834 | * on vmcs12 as required by the Intel SDM, and act appropriately when | |
6835 | * they fail: As the SDM explains, some conditions should cause the | |
6836 | * instruction to fail, while others will cause the instruction to seem | |
6837 | * to succeed, but return an EXIT_REASON_INVALID_STATE. | |
6838 | * To speed up the normal (success) code path, we should avoid checking | |
6839 | * for misconfigurations which will anyway be caught by the processor | |
6840 | * when using the merged vmcs02. | |
6841 | */ | |
6842 | if (vmcs12->launch_state == launch) { | |
6843 | nested_vmx_failValid(vcpu, | |
6844 | launch ? VMXERR_VMLAUNCH_NONCLEAR_VMCS | |
6845 | : VMXERR_VMRESUME_NONLAUNCHED_VMCS); | |
6846 | return 1; | |
6847 | } | |
6848 | ||
6849 | if ((vmcs12->cpu_based_vm_exec_control & CPU_BASED_USE_MSR_BITMAPS) && | |
6850 | !IS_ALIGNED(vmcs12->msr_bitmap, PAGE_SIZE)) { | |
6851 | /*TODO: Also verify bits beyond physical address width are 0*/ | |
6852 | nested_vmx_failValid(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD); | |
6853 | return 1; | |
6854 | } | |
6855 | ||
6856 | if (nested_cpu_has2(vmcs12, SECONDARY_EXEC_VIRTUALIZE_APIC_ACCESSES) && | |
6857 | !IS_ALIGNED(vmcs12->apic_access_addr, PAGE_SIZE)) { | |
6858 | /*TODO: Also verify bits beyond physical address width are 0*/ | |
6859 | nested_vmx_failValid(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD); | |
6860 | return 1; | |
6861 | } | |
6862 | ||
6863 | if (vmcs12->vm_entry_msr_load_count > 0 || | |
6864 | vmcs12->vm_exit_msr_load_count > 0 || | |
6865 | vmcs12->vm_exit_msr_store_count > 0) { | |
bd80158a JK |
6866 | pr_warn_ratelimited("%s: VMCS MSR_{LOAD,STORE} unsupported\n", |
6867 | __func__); | |
7c177938 NHE |
6868 | nested_vmx_failValid(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD); |
6869 | return 1; | |
6870 | } | |
6871 | ||
6872 | if (!vmx_control_verify(vmcs12->cpu_based_vm_exec_control, | |
6873 | nested_vmx_procbased_ctls_low, nested_vmx_procbased_ctls_high) || | |
6874 | !vmx_control_verify(vmcs12->secondary_vm_exec_control, | |
6875 | nested_vmx_secondary_ctls_low, nested_vmx_secondary_ctls_high) || | |
6876 | !vmx_control_verify(vmcs12->pin_based_vm_exec_control, | |
6877 | nested_vmx_pinbased_ctls_low, nested_vmx_pinbased_ctls_high) || | |
6878 | !vmx_control_verify(vmcs12->vm_exit_controls, | |
6879 | nested_vmx_exit_ctls_low, nested_vmx_exit_ctls_high) || | |
6880 | !vmx_control_verify(vmcs12->vm_entry_controls, | |
6881 | nested_vmx_entry_ctls_low, nested_vmx_entry_ctls_high)) | |
6882 | { | |
6883 | nested_vmx_failValid(vcpu, VMXERR_ENTRY_INVALID_CONTROL_FIELD); | |
6884 | return 1; | |
6885 | } | |
6886 | ||
6887 | if (((vmcs12->host_cr0 & VMXON_CR0_ALWAYSON) != VMXON_CR0_ALWAYSON) || | |
6888 | ((vmcs12->host_cr4 & VMXON_CR4_ALWAYSON) != VMXON_CR4_ALWAYSON)) { | |
6889 | nested_vmx_failValid(vcpu, | |
6890 | VMXERR_ENTRY_INVALID_HOST_STATE_FIELD); | |
6891 | return 1; | |
6892 | } | |
6893 | ||
6894 | if (((vmcs12->guest_cr0 & VMXON_CR0_ALWAYSON) != VMXON_CR0_ALWAYSON) || | |
6895 | ((vmcs12->guest_cr4 & VMXON_CR4_ALWAYSON) != VMXON_CR4_ALWAYSON)) { | |
6896 | nested_vmx_entry_failure(vcpu, vmcs12, | |
6897 | EXIT_REASON_INVALID_STATE, ENTRY_FAIL_DEFAULT); | |
6898 | return 1; | |
6899 | } | |
6900 | if (vmcs12->vmcs_link_pointer != -1ull) { | |
6901 | nested_vmx_entry_failure(vcpu, vmcs12, | |
6902 | EXIT_REASON_INVALID_STATE, ENTRY_FAIL_VMCS_LINK_PTR); | |
6903 | return 1; | |
6904 | } | |
6905 | ||
6906 | /* | |
6907 | * We're finally done with prerequisite checking, and can start with | |
6908 | * the nested entry. | |
6909 | */ | |
6910 | ||
cd232ad0 NHE |
6911 | vmcs02 = nested_get_current_vmcs02(vmx); |
6912 | if (!vmcs02) | |
6913 | return -ENOMEM; | |
6914 | ||
6915 | enter_guest_mode(vcpu); | |
6916 | ||
6917 | vmx->nested.vmcs01_tsc_offset = vmcs_read64(TSC_OFFSET); | |
6918 | ||
6919 | cpu = get_cpu(); | |
6920 | vmx->loaded_vmcs = vmcs02; | |
6921 | vmx_vcpu_put(vcpu); | |
6922 | vmx_vcpu_load(vcpu, cpu); | |
6923 | vcpu->cpu = cpu; | |
6924 | put_cpu(); | |
6925 | ||
6926 | vmcs12->launch_state = 1; | |
6927 | ||
6928 | prepare_vmcs02(vcpu, vmcs12); | |
6929 | ||
6930 | /* | |
6931 | * Note no nested_vmx_succeed or nested_vmx_fail here. At this point | |
6932 | * we are no longer running L1, and VMLAUNCH/VMRESUME has not yet | |
6933 | * returned as far as L1 is concerned. It will only return (and set | |
6934 | * the success flag) when L2 exits (see nested_vmx_vmexit()). | |
6935 | */ | |
6936 | return 1; | |
6937 | } | |
6938 | ||
4704d0be NHE |
6939 | /* |
6940 | * On a nested exit from L2 to L1, vmcs12.guest_cr0 might not be up-to-date | |
6941 | * because L2 may have changed some cr0 bits directly (CRO_GUEST_HOST_MASK). | |
6942 | * This function returns the new value we should put in vmcs12.guest_cr0. | |
6943 | * It's not enough to just return the vmcs02 GUEST_CR0. Rather, | |
6944 | * 1. Bits that neither L0 nor L1 trapped, were set directly by L2 and are now | |
6945 | * available in vmcs02 GUEST_CR0. (Note: It's enough to check that L0 | |
6946 | * didn't trap the bit, because if L1 did, so would L0). | |
6947 | * 2. Bits that L1 asked to trap (and therefore L0 also did) could not have | |
6948 | * been modified by L2, and L1 knows it. So just leave the old value of | |
6949 | * the bit from vmcs12.guest_cr0. Note that the bit from vmcs02 GUEST_CR0 | |
6950 | * isn't relevant, because if L0 traps this bit it can set it to anything. | |
6951 | * 3. Bits that L1 didn't trap, but L0 did. L1 believes the guest could have | |
6952 | * changed these bits, and therefore they need to be updated, but L0 | |
6953 | * didn't necessarily allow them to be changed in GUEST_CR0 - and rather | |
6954 | * put them in vmcs02 CR0_READ_SHADOW. So take these bits from there. | |
6955 | */ | |
6956 | static inline unsigned long | |
6957 | vmcs12_guest_cr0(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12) | |
6958 | { | |
6959 | return | |
6960 | /*1*/ (vmcs_readl(GUEST_CR0) & vcpu->arch.cr0_guest_owned_bits) | | |
6961 | /*2*/ (vmcs12->guest_cr0 & vmcs12->cr0_guest_host_mask) | | |
6962 | /*3*/ (vmcs_readl(CR0_READ_SHADOW) & ~(vmcs12->cr0_guest_host_mask | | |
6963 | vcpu->arch.cr0_guest_owned_bits)); | |
6964 | } | |
6965 | ||
6966 | static inline unsigned long | |
6967 | vmcs12_guest_cr4(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12) | |
6968 | { | |
6969 | return | |
6970 | /*1*/ (vmcs_readl(GUEST_CR4) & vcpu->arch.cr4_guest_owned_bits) | | |
6971 | /*2*/ (vmcs12->guest_cr4 & vmcs12->cr4_guest_host_mask) | | |
6972 | /*3*/ (vmcs_readl(CR4_READ_SHADOW) & ~(vmcs12->cr4_guest_host_mask | | |
6973 | vcpu->arch.cr4_guest_owned_bits)); | |
6974 | } | |
6975 | ||
6976 | /* | |
6977 | * prepare_vmcs12 is part of what we need to do when the nested L2 guest exits | |
6978 | * and we want to prepare to run its L1 parent. L1 keeps a vmcs for L2 (vmcs12), | |
6979 | * and this function updates it to reflect the changes to the guest state while | |
6980 | * L2 was running (and perhaps made some exits which were handled directly by L0 | |
6981 | * without going back to L1), and to reflect the exit reason. | |
6982 | * Note that we do not have to copy here all VMCS fields, just those that | |
6983 | * could have changed by the L2 guest or the exit - i.e., the guest-state and | |
6984 | * exit-information fields only. Other fields are modified by L1 with VMWRITE, | |
6985 | * which already writes to vmcs12 directly. | |
6986 | */ | |
6987 | void prepare_vmcs12(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12) | |
6988 | { | |
6989 | /* update guest state fields: */ | |
6990 | vmcs12->guest_cr0 = vmcs12_guest_cr0(vcpu, vmcs12); | |
6991 | vmcs12->guest_cr4 = vmcs12_guest_cr4(vcpu, vmcs12); | |
6992 | ||
6993 | kvm_get_dr(vcpu, 7, (unsigned long *)&vmcs12->guest_dr7); | |
6994 | vmcs12->guest_rsp = kvm_register_read(vcpu, VCPU_REGS_RSP); | |
6995 | vmcs12->guest_rip = kvm_register_read(vcpu, VCPU_REGS_RIP); | |
6996 | vmcs12->guest_rflags = vmcs_readl(GUEST_RFLAGS); | |
6997 | ||
6998 | vmcs12->guest_es_selector = vmcs_read16(GUEST_ES_SELECTOR); | |
6999 | vmcs12->guest_cs_selector = vmcs_read16(GUEST_CS_SELECTOR); | |
7000 | vmcs12->guest_ss_selector = vmcs_read16(GUEST_SS_SELECTOR); | |
7001 | vmcs12->guest_ds_selector = vmcs_read16(GUEST_DS_SELECTOR); | |
7002 | vmcs12->guest_fs_selector = vmcs_read16(GUEST_FS_SELECTOR); | |
7003 | vmcs12->guest_gs_selector = vmcs_read16(GUEST_GS_SELECTOR); | |
7004 | vmcs12->guest_ldtr_selector = vmcs_read16(GUEST_LDTR_SELECTOR); | |
7005 | vmcs12->guest_tr_selector = vmcs_read16(GUEST_TR_SELECTOR); | |
7006 | vmcs12->guest_es_limit = vmcs_read32(GUEST_ES_LIMIT); | |
7007 | vmcs12->guest_cs_limit = vmcs_read32(GUEST_CS_LIMIT); | |
7008 | vmcs12->guest_ss_limit = vmcs_read32(GUEST_SS_LIMIT); | |
7009 | vmcs12->guest_ds_limit = vmcs_read32(GUEST_DS_LIMIT); | |
7010 | vmcs12->guest_fs_limit = vmcs_read32(GUEST_FS_LIMIT); | |
7011 | vmcs12->guest_gs_limit = vmcs_read32(GUEST_GS_LIMIT); | |
7012 | vmcs12->guest_ldtr_limit = vmcs_read32(GUEST_LDTR_LIMIT); | |
7013 | vmcs12->guest_tr_limit = vmcs_read32(GUEST_TR_LIMIT); | |
7014 | vmcs12->guest_gdtr_limit = vmcs_read32(GUEST_GDTR_LIMIT); | |
7015 | vmcs12->guest_idtr_limit = vmcs_read32(GUEST_IDTR_LIMIT); | |
7016 | vmcs12->guest_es_ar_bytes = vmcs_read32(GUEST_ES_AR_BYTES); | |
7017 | vmcs12->guest_cs_ar_bytes = vmcs_read32(GUEST_CS_AR_BYTES); | |
7018 | vmcs12->guest_ss_ar_bytes = vmcs_read32(GUEST_SS_AR_BYTES); | |
7019 | vmcs12->guest_ds_ar_bytes = vmcs_read32(GUEST_DS_AR_BYTES); | |
7020 | vmcs12->guest_fs_ar_bytes = vmcs_read32(GUEST_FS_AR_BYTES); | |
7021 | vmcs12->guest_gs_ar_bytes = vmcs_read32(GUEST_GS_AR_BYTES); | |
7022 | vmcs12->guest_ldtr_ar_bytes = vmcs_read32(GUEST_LDTR_AR_BYTES); | |
7023 | vmcs12->guest_tr_ar_bytes = vmcs_read32(GUEST_TR_AR_BYTES); | |
7024 | vmcs12->guest_es_base = vmcs_readl(GUEST_ES_BASE); | |
7025 | vmcs12->guest_cs_base = vmcs_readl(GUEST_CS_BASE); | |
7026 | vmcs12->guest_ss_base = vmcs_readl(GUEST_SS_BASE); | |
7027 | vmcs12->guest_ds_base = vmcs_readl(GUEST_DS_BASE); | |
7028 | vmcs12->guest_fs_base = vmcs_readl(GUEST_FS_BASE); | |
7029 | vmcs12->guest_gs_base = vmcs_readl(GUEST_GS_BASE); | |
7030 | vmcs12->guest_ldtr_base = vmcs_readl(GUEST_LDTR_BASE); | |
7031 | vmcs12->guest_tr_base = vmcs_readl(GUEST_TR_BASE); | |
7032 | vmcs12->guest_gdtr_base = vmcs_readl(GUEST_GDTR_BASE); | |
7033 | vmcs12->guest_idtr_base = vmcs_readl(GUEST_IDTR_BASE); | |
7034 | ||
7035 | vmcs12->guest_activity_state = vmcs_read32(GUEST_ACTIVITY_STATE); | |
7036 | vmcs12->guest_interruptibility_info = | |
7037 | vmcs_read32(GUEST_INTERRUPTIBILITY_INFO); | |
7038 | vmcs12->guest_pending_dbg_exceptions = | |
7039 | vmcs_readl(GUEST_PENDING_DBG_EXCEPTIONS); | |
7040 | ||
7041 | /* TODO: These cannot have changed unless we have MSR bitmaps and | |
7042 | * the relevant bit asks not to trap the change */ | |
7043 | vmcs12->guest_ia32_debugctl = vmcs_read64(GUEST_IA32_DEBUGCTL); | |
7044 | if (vmcs12->vm_entry_controls & VM_EXIT_SAVE_IA32_PAT) | |
7045 | vmcs12->guest_ia32_pat = vmcs_read64(GUEST_IA32_PAT); | |
7046 | vmcs12->guest_sysenter_cs = vmcs_read32(GUEST_SYSENTER_CS); | |
7047 | vmcs12->guest_sysenter_esp = vmcs_readl(GUEST_SYSENTER_ESP); | |
7048 | vmcs12->guest_sysenter_eip = vmcs_readl(GUEST_SYSENTER_EIP); | |
7049 | ||
7050 | /* update exit information fields: */ | |
7051 | ||
7052 | vmcs12->vm_exit_reason = vmcs_read32(VM_EXIT_REASON); | |
7053 | vmcs12->exit_qualification = vmcs_readl(EXIT_QUALIFICATION); | |
7054 | ||
7055 | vmcs12->vm_exit_intr_info = vmcs_read32(VM_EXIT_INTR_INFO); | |
7056 | vmcs12->vm_exit_intr_error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE); | |
7057 | vmcs12->idt_vectoring_info_field = | |
7058 | vmcs_read32(IDT_VECTORING_INFO_FIELD); | |
7059 | vmcs12->idt_vectoring_error_code = | |
7060 | vmcs_read32(IDT_VECTORING_ERROR_CODE); | |
7061 | vmcs12->vm_exit_instruction_len = vmcs_read32(VM_EXIT_INSTRUCTION_LEN); | |
7062 | vmcs12->vmx_instruction_info = vmcs_read32(VMX_INSTRUCTION_INFO); | |
7063 | ||
7064 | /* clear vm-entry fields which are to be cleared on exit */ | |
7065 | if (!(vmcs12->vm_exit_reason & VMX_EXIT_REASONS_FAILED_VMENTRY)) | |
7066 | vmcs12->vm_entry_intr_info_field &= ~INTR_INFO_VALID_MASK; | |
7067 | } | |
7068 | ||
7069 | /* | |
7070 | * A part of what we need to when the nested L2 guest exits and we want to | |
7071 | * run its L1 parent, is to reset L1's guest state to the host state specified | |
7072 | * in vmcs12. | |
7073 | * This function is to be called not only on normal nested exit, but also on | |
7074 | * a nested entry failure, as explained in Intel's spec, 3B.23.7 ("VM-Entry | |
7075 | * Failures During or After Loading Guest State"). | |
7076 | * This function should be called when the active VMCS is L1's (vmcs01). | |
7077 | */ | |
7078 | void load_vmcs12_host_state(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12) | |
7079 | { | |
7080 | if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_EFER) | |
7081 | vcpu->arch.efer = vmcs12->host_ia32_efer; | |
7082 | if (vmcs12->vm_exit_controls & VM_EXIT_HOST_ADDR_SPACE_SIZE) | |
7083 | vcpu->arch.efer |= (EFER_LMA | EFER_LME); | |
7084 | else | |
7085 | vcpu->arch.efer &= ~(EFER_LMA | EFER_LME); | |
7086 | vmx_set_efer(vcpu, vcpu->arch.efer); | |
7087 | ||
7088 | kvm_register_write(vcpu, VCPU_REGS_RSP, vmcs12->host_rsp); | |
7089 | kvm_register_write(vcpu, VCPU_REGS_RIP, vmcs12->host_rip); | |
7090 | /* | |
7091 | * Note that calling vmx_set_cr0 is important, even if cr0 hasn't | |
7092 | * actually changed, because it depends on the current state of | |
7093 | * fpu_active (which may have changed). | |
7094 | * Note that vmx_set_cr0 refers to efer set above. | |
7095 | */ | |
7096 | kvm_set_cr0(vcpu, vmcs12->host_cr0); | |
7097 | /* | |
7098 | * If we did fpu_activate()/fpu_deactivate() during L2's run, we need | |
7099 | * to apply the same changes to L1's vmcs. We just set cr0 correctly, | |
7100 | * but we also need to update cr0_guest_host_mask and exception_bitmap. | |
7101 | */ | |
7102 | update_exception_bitmap(vcpu); | |
7103 | vcpu->arch.cr0_guest_owned_bits = (vcpu->fpu_active ? X86_CR0_TS : 0); | |
7104 | vmcs_writel(CR0_GUEST_HOST_MASK, ~vcpu->arch.cr0_guest_owned_bits); | |
7105 | ||
7106 | /* | |
7107 | * Note that CR4_GUEST_HOST_MASK is already set in the original vmcs01 | |
7108 | * (KVM doesn't change it)- no reason to call set_cr4_guest_host_mask(); | |
7109 | */ | |
7110 | vcpu->arch.cr4_guest_owned_bits = ~vmcs_readl(CR4_GUEST_HOST_MASK); | |
7111 | kvm_set_cr4(vcpu, vmcs12->host_cr4); | |
7112 | ||
7113 | /* shadow page tables on either EPT or shadow page tables */ | |
7114 | kvm_set_cr3(vcpu, vmcs12->host_cr3); | |
7115 | kvm_mmu_reset_context(vcpu); | |
7116 | ||
7117 | if (enable_vpid) { | |
7118 | /* | |
7119 | * Trivially support vpid by letting L2s share their parent | |
7120 | * L1's vpid. TODO: move to a more elaborate solution, giving | |
7121 | * each L2 its own vpid and exposing the vpid feature to L1. | |
7122 | */ | |
7123 | vmx_flush_tlb(vcpu); | |
7124 | } | |
7125 | ||
7126 | ||
7127 | vmcs_write32(GUEST_SYSENTER_CS, vmcs12->host_ia32_sysenter_cs); | |
7128 | vmcs_writel(GUEST_SYSENTER_ESP, vmcs12->host_ia32_sysenter_esp); | |
7129 | vmcs_writel(GUEST_SYSENTER_EIP, vmcs12->host_ia32_sysenter_eip); | |
7130 | vmcs_writel(GUEST_IDTR_BASE, vmcs12->host_idtr_base); | |
7131 | vmcs_writel(GUEST_GDTR_BASE, vmcs12->host_gdtr_base); | |
7132 | vmcs_writel(GUEST_TR_BASE, vmcs12->host_tr_base); | |
7133 | vmcs_writel(GUEST_GS_BASE, vmcs12->host_gs_base); | |
7134 | vmcs_writel(GUEST_FS_BASE, vmcs12->host_fs_base); | |
7135 | vmcs_write16(GUEST_ES_SELECTOR, vmcs12->host_es_selector); | |
7136 | vmcs_write16(GUEST_CS_SELECTOR, vmcs12->host_cs_selector); | |
7137 | vmcs_write16(GUEST_SS_SELECTOR, vmcs12->host_ss_selector); | |
7138 | vmcs_write16(GUEST_DS_SELECTOR, vmcs12->host_ds_selector); | |
7139 | vmcs_write16(GUEST_FS_SELECTOR, vmcs12->host_fs_selector); | |
7140 | vmcs_write16(GUEST_GS_SELECTOR, vmcs12->host_gs_selector); | |
7141 | vmcs_write16(GUEST_TR_SELECTOR, vmcs12->host_tr_selector); | |
7142 | ||
7143 | if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PAT) | |
7144 | vmcs_write64(GUEST_IA32_PAT, vmcs12->host_ia32_pat); | |
7145 | if (vmcs12->vm_exit_controls & VM_EXIT_LOAD_IA32_PERF_GLOBAL_CTRL) | |
7146 | vmcs_write64(GUEST_IA32_PERF_GLOBAL_CTRL, | |
7147 | vmcs12->host_ia32_perf_global_ctrl); | |
7148 | } | |
7149 | ||
7150 | /* | |
7151 | * Emulate an exit from nested guest (L2) to L1, i.e., prepare to run L1 | |
7152 | * and modify vmcs12 to make it see what it would expect to see there if | |
7153 | * L2 was its real guest. Must only be called when in L2 (is_guest_mode()) | |
7154 | */ | |
7155 | static void nested_vmx_vmexit(struct kvm_vcpu *vcpu) | |
7156 | { | |
7157 | struct vcpu_vmx *vmx = to_vmx(vcpu); | |
7158 | int cpu; | |
7159 | struct vmcs12 *vmcs12 = get_vmcs12(vcpu); | |
7160 | ||
7161 | leave_guest_mode(vcpu); | |
7162 | prepare_vmcs12(vcpu, vmcs12); | |
7163 | ||
7164 | cpu = get_cpu(); | |
7165 | vmx->loaded_vmcs = &vmx->vmcs01; | |
7166 | vmx_vcpu_put(vcpu); | |
7167 | vmx_vcpu_load(vcpu, cpu); | |
7168 | vcpu->cpu = cpu; | |
7169 | put_cpu(); | |
7170 | ||
7171 | /* if no vmcs02 cache requested, remove the one we used */ | |
7172 | if (VMCS02_POOL_SIZE == 0) | |
7173 | nested_free_vmcs02(vmx, vmx->nested.current_vmptr); | |
7174 | ||
7175 | load_vmcs12_host_state(vcpu, vmcs12); | |
7176 | ||
27fc51b2 | 7177 | /* Update TSC_OFFSET if TSC was changed while L2 ran */ |
4704d0be NHE |
7178 | vmcs_write64(TSC_OFFSET, vmx->nested.vmcs01_tsc_offset); |
7179 | ||
7180 | /* This is needed for same reason as it was needed in prepare_vmcs02 */ | |
7181 | vmx->host_rsp = 0; | |
7182 | ||
7183 | /* Unpin physical memory we referred to in vmcs02 */ | |
7184 | if (vmx->nested.apic_access_page) { | |
7185 | nested_release_page(vmx->nested.apic_access_page); | |
7186 | vmx->nested.apic_access_page = 0; | |
7187 | } | |
7188 | ||
7189 | /* | |
7190 | * Exiting from L2 to L1, we're now back to L1 which thinks it just | |
7191 | * finished a VMLAUNCH or VMRESUME instruction, so we need to set the | |
7192 | * success or failure flag accordingly. | |
7193 | */ | |
7194 | if (unlikely(vmx->fail)) { | |
7195 | vmx->fail = 0; | |
7196 | nested_vmx_failValid(vcpu, vmcs_read32(VM_INSTRUCTION_ERROR)); | |
7197 | } else | |
7198 | nested_vmx_succeed(vcpu); | |
7199 | } | |
7200 | ||
7c177938 NHE |
7201 | /* |
7202 | * L1's failure to enter L2 is a subset of a normal exit, as explained in | |
7203 | * 23.7 "VM-entry failures during or after loading guest state" (this also | |
7204 | * lists the acceptable exit-reason and exit-qualification parameters). | |
7205 | * It should only be called before L2 actually succeeded to run, and when | |
7206 | * vmcs01 is current (it doesn't leave_guest_mode() or switch vmcss). | |
7207 | */ | |
7208 | static void nested_vmx_entry_failure(struct kvm_vcpu *vcpu, | |
7209 | struct vmcs12 *vmcs12, | |
7210 | u32 reason, unsigned long qualification) | |
7211 | { | |
7212 | load_vmcs12_host_state(vcpu, vmcs12); | |
7213 | vmcs12->vm_exit_reason = reason | VMX_EXIT_REASONS_FAILED_VMENTRY; | |
7214 | vmcs12->exit_qualification = qualification; | |
7215 | nested_vmx_succeed(vcpu); | |
7216 | } | |
7217 | ||
8a76d7f2 JR |
7218 | static int vmx_check_intercept(struct kvm_vcpu *vcpu, |
7219 | struct x86_instruction_info *info, | |
7220 | enum x86_intercept_stage stage) | |
7221 | { | |
7222 | return X86EMUL_CONTINUE; | |
7223 | } | |
7224 | ||
cbdd1bea | 7225 | static struct kvm_x86_ops vmx_x86_ops = { |
6aa8b732 AK |
7226 | .cpu_has_kvm_support = cpu_has_kvm_support, |
7227 | .disabled_by_bios = vmx_disabled_by_bios, | |
7228 | .hardware_setup = hardware_setup, | |
7229 | .hardware_unsetup = hardware_unsetup, | |
002c7f7c | 7230 | .check_processor_compatibility = vmx_check_processor_compat, |
6aa8b732 AK |
7231 | .hardware_enable = hardware_enable, |
7232 | .hardware_disable = hardware_disable, | |
04547156 | 7233 | .cpu_has_accelerated_tpr = report_flexpriority, |
6aa8b732 AK |
7234 | |
7235 | .vcpu_create = vmx_create_vcpu, | |
7236 | .vcpu_free = vmx_free_vcpu, | |
04d2cc77 | 7237 | .vcpu_reset = vmx_vcpu_reset, |
6aa8b732 | 7238 | |
04d2cc77 | 7239 | .prepare_guest_switch = vmx_save_host_state, |
6aa8b732 AK |
7240 | .vcpu_load = vmx_vcpu_load, |
7241 | .vcpu_put = vmx_vcpu_put, | |
7242 | ||
7243 | .set_guest_debug = set_guest_debug, | |
7244 | .get_msr = vmx_get_msr, | |
7245 | .set_msr = vmx_set_msr, | |
7246 | .get_segment_base = vmx_get_segment_base, | |
7247 | .get_segment = vmx_get_segment, | |
7248 | .set_segment = vmx_set_segment, | |
2e4d2653 | 7249 | .get_cpl = vmx_get_cpl, |
6aa8b732 | 7250 | .get_cs_db_l_bits = vmx_get_cs_db_l_bits, |
e8467fda | 7251 | .decache_cr0_guest_bits = vmx_decache_cr0_guest_bits, |
aff48baa | 7252 | .decache_cr3 = vmx_decache_cr3, |
25c4c276 | 7253 | .decache_cr4_guest_bits = vmx_decache_cr4_guest_bits, |
6aa8b732 | 7254 | .set_cr0 = vmx_set_cr0, |
6aa8b732 AK |
7255 | .set_cr3 = vmx_set_cr3, |
7256 | .set_cr4 = vmx_set_cr4, | |
6aa8b732 | 7257 | .set_efer = vmx_set_efer, |
6aa8b732 AK |
7258 | .get_idt = vmx_get_idt, |
7259 | .set_idt = vmx_set_idt, | |
7260 | .get_gdt = vmx_get_gdt, | |
7261 | .set_gdt = vmx_set_gdt, | |
020df079 | 7262 | .set_dr7 = vmx_set_dr7, |
5fdbf976 | 7263 | .cache_reg = vmx_cache_reg, |
6aa8b732 AK |
7264 | .get_rflags = vmx_get_rflags, |
7265 | .set_rflags = vmx_set_rflags, | |
ebcbab4c | 7266 | .fpu_activate = vmx_fpu_activate, |
02daab21 | 7267 | .fpu_deactivate = vmx_fpu_deactivate, |
6aa8b732 AK |
7268 | |
7269 | .tlb_flush = vmx_flush_tlb, | |
6aa8b732 | 7270 | |
6aa8b732 | 7271 | .run = vmx_vcpu_run, |
6062d012 | 7272 | .handle_exit = vmx_handle_exit, |
6aa8b732 | 7273 | .skip_emulated_instruction = skip_emulated_instruction, |
2809f5d2 GC |
7274 | .set_interrupt_shadow = vmx_set_interrupt_shadow, |
7275 | .get_interrupt_shadow = vmx_get_interrupt_shadow, | |
102d8325 | 7276 | .patch_hypercall = vmx_patch_hypercall, |
2a8067f1 | 7277 | .set_irq = vmx_inject_irq, |
95ba8273 | 7278 | .set_nmi = vmx_inject_nmi, |
298101da | 7279 | .queue_exception = vmx_queue_exception, |
b463a6f7 | 7280 | .cancel_injection = vmx_cancel_injection, |
78646121 | 7281 | .interrupt_allowed = vmx_interrupt_allowed, |
95ba8273 | 7282 | .nmi_allowed = vmx_nmi_allowed, |
3cfc3092 JK |
7283 | .get_nmi_mask = vmx_get_nmi_mask, |
7284 | .set_nmi_mask = vmx_set_nmi_mask, | |
95ba8273 GN |
7285 | .enable_nmi_window = enable_nmi_window, |
7286 | .enable_irq_window = enable_irq_window, | |
7287 | .update_cr8_intercept = update_cr8_intercept, | |
95ba8273 | 7288 | |
cbc94022 | 7289 | .set_tss_addr = vmx_set_tss_addr, |
67253af5 | 7290 | .get_tdp_level = get_ept_level, |
4b12f0de | 7291 | .get_mt_mask = vmx_get_mt_mask, |
229456fc | 7292 | |
586f9607 | 7293 | .get_exit_info = vmx_get_exit_info, |
586f9607 | 7294 | |
17cc3935 | 7295 | .get_lpage_level = vmx_get_lpage_level, |
0e851880 SY |
7296 | |
7297 | .cpuid_update = vmx_cpuid_update, | |
4e47c7a6 SY |
7298 | |
7299 | .rdtscp_supported = vmx_rdtscp_supported, | |
ad756a16 | 7300 | .invpcid_supported = vmx_invpcid_supported, |
d4330ef2 JR |
7301 | |
7302 | .set_supported_cpuid = vmx_set_supported_cpuid, | |
f5f48ee1 SY |
7303 | |
7304 | .has_wbinvd_exit = cpu_has_vmx_wbinvd_exit, | |
99e3e30a | 7305 | |
4051b188 | 7306 | .set_tsc_khz = vmx_set_tsc_khz, |
99e3e30a | 7307 | .write_tsc_offset = vmx_write_tsc_offset, |
e48672fa | 7308 | .adjust_tsc_offset = vmx_adjust_tsc_offset, |
857e4099 | 7309 | .compute_tsc_offset = vmx_compute_tsc_offset, |
d5c1785d | 7310 | .read_l1_tsc = vmx_read_l1_tsc, |
1c97f0a0 JR |
7311 | |
7312 | .set_tdp_cr3 = vmx_set_cr3, | |
8a76d7f2 JR |
7313 | |
7314 | .check_intercept = vmx_check_intercept, | |
6aa8b732 AK |
7315 | }; |
7316 | ||
7317 | static int __init vmx_init(void) | |
7318 | { | |
26bb0981 AK |
7319 | int r, i; |
7320 | ||
7321 | rdmsrl_safe(MSR_EFER, &host_efer); | |
7322 | ||
7323 | for (i = 0; i < NR_VMX_MSR; ++i) | |
7324 | kvm_define_shared_msr(i, vmx_msr_index[i]); | |
fdef3ad1 | 7325 | |
3e7c73e9 | 7326 | vmx_io_bitmap_a = (unsigned long *)__get_free_page(GFP_KERNEL); |
fdef3ad1 HQ |
7327 | if (!vmx_io_bitmap_a) |
7328 | return -ENOMEM; | |
7329 | ||
2106a548 GC |
7330 | r = -ENOMEM; |
7331 | ||
3e7c73e9 | 7332 | vmx_io_bitmap_b = (unsigned long *)__get_free_page(GFP_KERNEL); |
2106a548 | 7333 | if (!vmx_io_bitmap_b) |
fdef3ad1 | 7334 | goto out; |
fdef3ad1 | 7335 | |
5897297b | 7336 | vmx_msr_bitmap_legacy = (unsigned long *)__get_free_page(GFP_KERNEL); |
2106a548 | 7337 | if (!vmx_msr_bitmap_legacy) |
25c5f225 | 7338 | goto out1; |
2106a548 | 7339 | |
25c5f225 | 7340 | |
5897297b | 7341 | vmx_msr_bitmap_longmode = (unsigned long *)__get_free_page(GFP_KERNEL); |
2106a548 | 7342 | if (!vmx_msr_bitmap_longmode) |
5897297b | 7343 | goto out2; |
2106a548 | 7344 | |
5897297b | 7345 | |
fdef3ad1 HQ |
7346 | /* |
7347 | * Allow direct access to the PC debug port (it is often used for I/O | |
7348 | * delays, but the vmexits simply slow things down). | |
7349 | */ | |
3e7c73e9 AK |
7350 | memset(vmx_io_bitmap_a, 0xff, PAGE_SIZE); |
7351 | clear_bit(0x80, vmx_io_bitmap_a); | |
fdef3ad1 | 7352 | |
3e7c73e9 | 7353 | memset(vmx_io_bitmap_b, 0xff, PAGE_SIZE); |
fdef3ad1 | 7354 | |
5897297b AK |
7355 | memset(vmx_msr_bitmap_legacy, 0xff, PAGE_SIZE); |
7356 | memset(vmx_msr_bitmap_longmode, 0xff, PAGE_SIZE); | |
25c5f225 | 7357 | |
2384d2b3 SY |
7358 | set_bit(0, vmx_vpid_bitmap); /* 0 is reserved for host */ |
7359 | ||
0ee75bea AK |
7360 | r = kvm_init(&vmx_x86_ops, sizeof(struct vcpu_vmx), |
7361 | __alignof__(struct vcpu_vmx), THIS_MODULE); | |
fdef3ad1 | 7362 | if (r) |
5897297b | 7363 | goto out3; |
25c5f225 | 7364 | |
5897297b AK |
7365 | vmx_disable_intercept_for_msr(MSR_FS_BASE, false); |
7366 | vmx_disable_intercept_for_msr(MSR_GS_BASE, false); | |
7367 | vmx_disable_intercept_for_msr(MSR_KERNEL_GS_BASE, true); | |
7368 | vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_CS, false); | |
7369 | vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_ESP, false); | |
7370 | vmx_disable_intercept_for_msr(MSR_IA32_SYSENTER_EIP, false); | |
fdef3ad1 | 7371 | |
089d034e | 7372 | if (enable_ept) { |
3f6d8c8a XH |
7373 | kvm_mmu_set_mask_ptes(0ull, |
7374 | (enable_ept_ad_bits) ? VMX_EPT_ACCESS_BIT : 0ull, | |
7375 | (enable_ept_ad_bits) ? VMX_EPT_DIRTY_BIT : 0ull, | |
7376 | 0ull, VMX_EPT_EXECUTABLE_MASK); | |
ce88decf | 7377 | ept_set_mmio_spte_mask(); |
5fdbcb9d SY |
7378 | kvm_enable_tdp(); |
7379 | } else | |
7380 | kvm_disable_tdp(); | |
1439442c | 7381 | |
fdef3ad1 HQ |
7382 | return 0; |
7383 | ||
5897297b AK |
7384 | out3: |
7385 | free_page((unsigned long)vmx_msr_bitmap_longmode); | |
25c5f225 | 7386 | out2: |
5897297b | 7387 | free_page((unsigned long)vmx_msr_bitmap_legacy); |
fdef3ad1 | 7388 | out1: |
3e7c73e9 | 7389 | free_page((unsigned long)vmx_io_bitmap_b); |
fdef3ad1 | 7390 | out: |
3e7c73e9 | 7391 | free_page((unsigned long)vmx_io_bitmap_a); |
fdef3ad1 | 7392 | return r; |
6aa8b732 AK |
7393 | } |
7394 | ||
7395 | static void __exit vmx_exit(void) | |
7396 | { | |
5897297b AK |
7397 | free_page((unsigned long)vmx_msr_bitmap_legacy); |
7398 | free_page((unsigned long)vmx_msr_bitmap_longmode); | |
3e7c73e9 AK |
7399 | free_page((unsigned long)vmx_io_bitmap_b); |
7400 | free_page((unsigned long)vmx_io_bitmap_a); | |
fdef3ad1 | 7401 | |
cb498ea2 | 7402 | kvm_exit(); |
6aa8b732 AK |
7403 | } |
7404 | ||
7405 | module_init(vmx_init) | |
7406 | module_exit(vmx_exit) |