2 * Copyright (c) 2011 NetApp, Inc.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #include <sys/cdefs.h>
30 __FBSDID("$FreeBSD$");
32 #include <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/cpuset.h>
37 #include <machine/clock.h>
38 #include <machine/cpufunc.h>
39 #include <machine/md_var.h>
40 #include <machine/segments.h>
41 #include <machine/specialreg.h>
43 #include <machine/vmm.h>
48 #define CPUID_VM_HIGH 0x40000000
50 static const char bhyve_id[12] = "bhyve bhyve ";
52 static uint64_t bhyve_xcpuids;
55 x86_emulate_cpuid(struct vm *vm, int vcpu_id,
56 uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx)
58 const struct xsave_limits *limits;
60 int error, enable_invpcid;
61 unsigned int func, regs[4];
62 enum x2apic_state x2apic_state;
65 * Requests for invalid CPUID levels should map to the highest
66 * available level instead.
68 if (cpu_exthigh != 0 && *eax >= 0x80000000) {
69 if (*eax > cpu_exthigh)
71 } else if (*eax >= 0x40000000) {
72 if (*eax > CPUID_VM_HIGH)
74 } else if (*eax > cpu_high) {
81 * In general the approach used for CPU topology is to
82 * advertise a flat topology where all CPUs are packages with
83 * no multi-core or SMT.
87 * Pass these through to the guest
98 cpuid_count(*eax, *ecx, regs);
101 case CPUID_8000_0001:
103 * Hide rdtscp/ia32_tsc_aux until we know how
106 cpuid_count(*eax, *ecx, regs);
107 regs[3] &= ~AMDID_RDTSCP;
110 case CPUID_8000_0007:
111 cpuid_count(*eax, *ecx, regs);
113 * If the host TSCs are not synchronized across
114 * physical cpus then we cannot advertise an
115 * invariant tsc to a vcpu.
117 * XXX This still falls short because the vcpu
118 * can observe the TSC moving backwards as it
119 * migrates across physical cpus. But at least
120 * it should discourage the guest from using the
121 * TSC to keep track of time.
124 regs[3] &= ~AMDPM_TSC_INVARIANT;
127 case CPUID_0000_0001:
130 error = vm_get_x2apic_state(vm, vcpu_id, &x2apic_state);
132 panic("x86_emulate_cpuid: error %d "
133 "fetching x2apic state", error);
137 * Override the APIC ID only in ebx
139 regs[1] &= ~(CPUID_LOCAL_APIC_ID);
140 regs[1] |= (vcpu_id << CPUID_0000_0001_APICID_SHIFT);
143 * Don't expose VMX, SpeedStep or TME capability.
144 * Advertise x2APIC capability and Hypervisor guest.
146 regs[2] &= ~(CPUID2_VMX | CPUID2_EST | CPUID2_TM2);
148 regs[2] |= CPUID2_HV;
150 if (x2apic_state != X2APIC_DISABLED)
151 regs[2] |= CPUID2_X2APIC;
153 regs[2] &= ~CPUID2_X2APIC;
156 * Only advertise CPUID2_XSAVE in the guest if
157 * the host is using XSAVE.
159 if (!(regs[2] & CPUID2_OSXSAVE))
160 regs[2] &= ~CPUID2_XSAVE;
163 * If CPUID2_XSAVE is being advertised and the
164 * guest has set CR4_XSAVE, set
167 regs[2] &= ~CPUID2_OSXSAVE;
168 if (regs[2] & CPUID2_XSAVE) {
169 error = vm_get_register(vm, vcpu_id,
170 VM_REG_GUEST_CR4, &cr4);
172 panic("x86_emulate_cpuid: error %d "
173 "fetching %%cr4", error);
175 regs[2] |= CPUID2_OSXSAVE;
179 * Hide monitor/mwait until we know how to deal with
180 * these instructions.
182 regs[2] &= ~CPUID2_MON;
185 * Hide the performance and debug features.
187 regs[2] &= ~CPUID2_PDCM;
190 * No TSC deadline support in the APIC yet
192 regs[2] &= ~CPUID2_TSCDLT;
195 * Hide thermal monitoring
197 regs[3] &= ~(CPUID_ACPI | CPUID_TM);
200 * Machine check handling is done in the host.
201 * Hide MTRR capability.
203 regs[3] &= ~(CPUID_MCA | CPUID_MCE | CPUID_MTRR);
206 * Hide the debug store capability.
208 regs[3] &= ~CPUID_DS;
211 * Disable multi-core.
213 regs[1] &= ~CPUID_HTT_CORES;
214 regs[3] &= ~CPUID_HTT;
217 case CPUID_0000_0004:
221 * Do not expose topology.
223 regs[0] &= 0xffff8000;
224 regs[0] |= 0x04008000;
227 case CPUID_0000_0007:
235 error = vm_get_capability(vm, vcpu_id,
236 VM_CAP_ENABLE_INVPCID, &enable_invpcid);
237 if (error == 0 && enable_invpcid)
238 regs[1] |= CPUID_STDEXT_INVPCID;
242 case CPUID_0000_0006:
243 case CPUID_0000_000A:
245 * Handle the access, but report 0 for
254 case CPUID_0000_000B:
256 * Processor topology enumeration
260 regs[2] = *ecx & 0xff;
264 case CPUID_0000_000D:
265 limits = vmm_get_xsave_limits();
266 if (!limits->xsave_enabled) {
274 cpuid_count(*eax, *ecx, regs);
278 * Only permit the guest to use bits
279 * that are active in the host in
280 * %xcr0. Also, claim that the
281 * maximum save area size is
282 * equivalent to the host's current
283 * save area size. Since this runs
284 * "inside" of vmrun(), it runs with
285 * the guest's xcr0, so the current
286 * save area size is correct as-is.
288 regs[0] &= limits->xcr0_allowed;
289 regs[2] = limits->xsave_max_size;
290 regs[3] &= (limits->xcr0_allowed >> 32);
293 /* Only permit XSAVEOPT. */
294 regs[0] &= CPUID_EXTSTATE_XSAVEOPT;
301 * If the leaf is for a permitted feature,
302 * pass through as-is, otherwise return
305 if (!(limits->xcr0_allowed & (1ul << *ecx))) {
316 regs[0] = CPUID_VM_HIGH;
317 bcopy(bhyve_id, ®s[1], 4);
318 bcopy(bhyve_id + 4, ®s[2], 4);
319 bcopy(bhyve_id + 8, ®s[3], 4);
324 * The leaf value has already been clamped so
325 * simply pass this through, keeping count of
326 * how many unhandled leaf values have been seen.
328 atomic_add_long(&bhyve_xcpuids, 1);
329 cpuid_count(*eax, *ecx, regs);