7 On Intel processors, KVM uses Intel's VMX (Virtual-Machine eXtensions)
8 to easily and efficiently run guest operating systems. Normally, these guests
9 *cannot* themselves be hypervisors running their own guests, because in VMX,
10 guests cannot use VMX instructions.
12 The "Nested VMX" feature adds this missing capability - of running guest
13 hypervisors (which use VMX) with their own nested guests. It does so by
14 allowing a guest to use VMX instructions, and correctly and efficiently
15 emulating them using the single level of VMX available in the hardware.
17 We describe in much greater detail the theory behind the nested VMX feature,
18 its implementation and its performance characteristics, in the OSDI 2010 paper
19 "The Turtles Project: Design and Implementation of Nested Virtualization",
22 http://www.usenix.org/events/osdi10/tech/full_papers/Ben-Yehuda.pdf
28 Single-level virtualization has two levels - the host (KVM) and the guests.
29 In nested virtualization, we have three levels: The host (KVM), which we call
30 L0, the guest hypervisor, which we call L1, and its nested guest, which we
37 The nested VMX feature is disabled by default. It can be enabled by giving
38 the "nested=1" option to the kvm-intel module.
40 No modifications are required to user space (qemu). However, qemu's default
41 emulated CPU type (qemu64) does not list the "VMX" CPU feature, so it must be
42 explicitly enabled, by giving qemu one of the following options:
44 -cpu host (emulated CPU has all features of the real CPU)
46 -cpu qemu64,+vmx (add just the vmx feature to a named CPU type)
52 Nested VMX aims to present a standard and (eventually) fully-functional VMX
53 implementation for the a guest hypervisor to use. As such, the official
54 specification of the ABI that it provides is Intel's VMX specification,
55 namely volume 3B of their "Intel 64 and IA-32 Architectures Software
56 Developer's Manual". Not all of VMX's features are currently fully supported,
57 but the goal is to eventually support them all, starting with the VMX features
58 which are used in practice by popular hypervisors (KVM and others).
60 As a VMX implementation, nested VMX presents a VMCS structure to L1.
61 As mandated by the spec, other than the two fields revision_id and abort,
62 this structure is *opaque* to its user, who is not supposed to know or care
63 about its internal structure. Rather, the structure is accessed through the
64 VMREAD and VMWRITE instructions.
65 Still, for debugging purposes, KVM developers might be interested to know the
66 internals of this structure; This is struct vmcs12 from arch/x86/kvm/vmx.c.
68 The name "vmcs12" refers to the VMCS that L1 builds for L2. In the code we
69 also have "vmcs01", the VMCS that L0 built for L1, and "vmcs02" is the VMCS
70 which L0 builds to actually run L2 - how this is done is explained in the
73 For convenience, we repeat the content of struct vmcs12 here. If the internals
74 of this structure changes, this can break live migration across KVM versions.
75 VMCS12_REVISION (from vmx.c) should be changed if struct vmcs12 or its inner
76 struct shadow_vmcs is ever changed.
78 typedef u64 natural_width;
79 struct __packed vmcs12 {
80 /* According to the Intel spec, a VMCS region must start with
81 * these two user-visible fields */
85 u32 launch_state; /* set to 0 by VMCLEAR, to 1 by VMLAUNCH */
86 u32 padding[7]; /* room for future expansion */
91 u64 vm_exit_msr_store_addr;
92 u64 vm_exit_msr_load_addr;
93 u64 vm_entry_msr_load_addr;
95 u64 virtual_apic_page_addr;
98 u64 guest_physical_address;
99 u64 vmcs_link_pointer;
100 u64 guest_ia32_debugctl;
109 u64 padding64[8]; /* room for future expansion */
110 natural_width cr0_guest_host_mask;
111 natural_width cr4_guest_host_mask;
112 natural_width cr0_read_shadow;
113 natural_width cr4_read_shadow;
114 natural_width cr3_target_value0;
115 natural_width cr3_target_value1;
116 natural_width cr3_target_value2;
117 natural_width cr3_target_value3;
118 natural_width exit_qualification;
119 natural_width guest_linear_address;
120 natural_width guest_cr0;
121 natural_width guest_cr3;
122 natural_width guest_cr4;
123 natural_width guest_es_base;
124 natural_width guest_cs_base;
125 natural_width guest_ss_base;
126 natural_width guest_ds_base;
127 natural_width guest_fs_base;
128 natural_width guest_gs_base;
129 natural_width guest_ldtr_base;
130 natural_width guest_tr_base;
131 natural_width guest_gdtr_base;
132 natural_width guest_idtr_base;
133 natural_width guest_dr7;
134 natural_width guest_rsp;
135 natural_width guest_rip;
136 natural_width guest_rflags;
137 natural_width guest_pending_dbg_exceptions;
138 natural_width guest_sysenter_esp;
139 natural_width guest_sysenter_eip;
140 natural_width host_cr0;
141 natural_width host_cr3;
142 natural_width host_cr4;
143 natural_width host_fs_base;
144 natural_width host_gs_base;
145 natural_width host_tr_base;
146 natural_width host_gdtr_base;
147 natural_width host_idtr_base;
148 natural_width host_ia32_sysenter_esp;
149 natural_width host_ia32_sysenter_eip;
150 natural_width host_rsp;
151 natural_width host_rip;
152 natural_width paddingl[8]; /* room for future expansion */
153 u32 pin_based_vm_exec_control;
154 u32 cpu_based_vm_exec_control;
155 u32 exception_bitmap;
156 u32 page_fault_error_code_mask;
157 u32 page_fault_error_code_match;
158 u32 cr3_target_count;
159 u32 vm_exit_controls;
160 u32 vm_exit_msr_store_count;
161 u32 vm_exit_msr_load_count;
162 u32 vm_entry_controls;
163 u32 vm_entry_msr_load_count;
164 u32 vm_entry_intr_info_field;
165 u32 vm_entry_exception_error_code;
166 u32 vm_entry_instruction_len;
168 u32 secondary_vm_exec_control;
169 u32 vm_instruction_error;
171 u32 vm_exit_intr_info;
172 u32 vm_exit_intr_error_code;
173 u32 idt_vectoring_info_field;
174 u32 idt_vectoring_error_code;
175 u32 vm_exit_instruction_len;
176 u32 vmx_instruction_info;
183 u32 guest_ldtr_limit;
185 u32 guest_gdtr_limit;
186 u32 guest_idtr_limit;
187 u32 guest_es_ar_bytes;
188 u32 guest_cs_ar_bytes;
189 u32 guest_ss_ar_bytes;
190 u32 guest_ds_ar_bytes;
191 u32 guest_fs_ar_bytes;
192 u32 guest_gs_ar_bytes;
193 u32 guest_ldtr_ar_bytes;
194 u32 guest_tr_ar_bytes;
195 u32 guest_interruptibility_info;
196 u32 guest_activity_state;
197 u32 guest_sysenter_cs;
198 u32 host_ia32_sysenter_cs;
199 u32 padding32[8]; /* room for future expansion */
200 u16 virtual_processor_id;
201 u16 guest_es_selector;
202 u16 guest_cs_selector;
203 u16 guest_ss_selector;
204 u16 guest_ds_selector;
205 u16 guest_fs_selector;
206 u16 guest_gs_selector;
207 u16 guest_ldtr_selector;
208 u16 guest_tr_selector;
209 u16 host_es_selector;
210 u16 host_cs_selector;
211 u16 host_ss_selector;
212 u16 host_ds_selector;
213 u16 host_fs_selector;
214 u16 host_gs_selector;
215 u16 host_tr_selector;
222 These patches were written by:
223 Abel Gordon, abelg <at> il.ibm.com
224 Nadav Har'El, nyh <at> il.ibm.com
225 Orit Wasserman, oritw <at> il.ibm.com
226 Ben-Ami Yassor, benami <at> il.ibm.com
227 Muli Ben-Yehuda, muli <at> il.ibm.com
229 With contributions by:
230 Anthony Liguori, aliguori <at> us.ibm.com
231 Mike Day, mdday <at> us.ibm.com
232 Michael Factor, factor <at> il.ibm.com
233 Zvi Dubitzky, dubi <at> il.ibm.com
235 And valuable reviews by:
236 Avi Kivity, avi <at> redhat.com
237 Gleb Natapov, gleb <at> redhat.com
238 Marcelo Tosatti, mtosatti <at> redhat.com
239 Kevin Tian, kevin.tian <at> intel.com