blob: 8ddcfe84c09a3df07809015e47b5e5290fedc14f [file] [log] [blame]
Glauber Costad2d7a612010-06-01 08:22:48 -04001KVM-specific MSRs.
2Glauber Costa <glommer@redhat.com>, Red Hat Inc, 2010
3=====================================================
4
5KVM makes use of some custom MSRs to service some requests.
6At present, this facility is only used by kvmclock.
7
8Custom MSRs have a range reserved for them, that goes from
90x4b564d00 to 0x4b564dff. There are MSRs outside this area,
10but they are deprecated and their use is discouraged.
11
12Custom MSR list
13--------
14
15The current supported Custom MSR list is:
16
17MSR_KVM_WALL_CLOCK_NEW: 0x4b564d00
18
19 data: 4-byte alignment physical address of a memory area which must be
20 in guest RAM. This memory is expected to hold a copy of the following
21 structure:
22
23 struct pvclock_wall_clock {
24 u32 version;
25 u32 sec;
26 u32 nsec;
27 } __attribute__((__packed__));
28
29 whose data will be filled in by the hypervisor. The hypervisor is only
30 guaranteed to update this data at the moment of MSR write.
31 Users that want to reliably query this information more than once have
32 to write more than once to this MSR. Fields have the following meanings:
33
34 version: guest has to check version before and after grabbing
35 time information and check that they are both equal and even.
36 An odd version indicates an in-progress update.
37
38 sec: number of seconds for wallclock.
39
40 nsec: number of nanoseconds for wallclock.
41
42 Note that although MSRs are per-CPU entities, the effect of this
43 particular MSR is global.
44
45 Availability of this MSR must be checked via bit 3 in 0x4000001 cpuid
46 leaf prior to usage.
47
48MSR_KVM_SYSTEM_TIME_NEW: 0x4b564d01
49
50 data: 4-byte aligned physical address of a memory area which must be in
51 guest RAM, plus an enable bit in bit 0. This memory is expected to hold
52 a copy of the following structure:
53
54 struct pvclock_vcpu_time_info {
55 u32 version;
56 u32 pad0;
57 u64 tsc_timestamp;
58 u64 system_time;
59 u32 tsc_to_system_mul;
60 s8 tsc_shift;
61 u8 flags;
62 u8 pad[2];
63 } __attribute__((__packed__)); /* 32 bytes */
64
65 whose data will be filled in by the hypervisor periodically. Only one
66 write, or registration, is needed for each VCPU. The interval between
67 updates of this structure is arbitrary and implementation-dependent.
68 The hypervisor may update this structure at any time it sees fit until
69 anything with bit0 == 0 is written to it.
70
71 Fields have the following meanings:
72
73 version: guest has to check version before and after grabbing
74 time information and check that they are both equal and even.
75 An odd version indicates an in-progress update.
76
77 tsc_timestamp: the tsc value at the current VCPU at the time
78 of the update of this structure. Guests can subtract this value
79 from current tsc to derive a notion of elapsed time since the
80 structure update.
81
82 system_time: a host notion of monotonic time, including sleep
83 time at the time this structure was last updated. Unit is
84 nanoseconds.
85
86 tsc_to_system_mul: a function of the tsc frequency. One has
87 to multiply any tsc-related quantity by this value to get
88 a value in nanoseconds, besides dividing by 2^tsc_shift
89
90 tsc_shift: cycle to nanosecond divider, as a power of two, to
91 allow for shift rights. One has to shift right any tsc-related
92 quantity by this value to get a value in nanoseconds, besides
93 multiplying by tsc_to_system_mul.
94
95 With this information, guests can derive per-CPU time by
96 doing:
97
98 time = (current_tsc - tsc_timestamp)
99 time = (time * tsc_to_system_mul) >> tsc_shift
100 time = time + system_time
101
102 flags: bits in this field indicate extended capabilities
103 coordinated between the guest and the hypervisor. Availability
104 of specific flags has to be checked in 0x40000001 cpuid leaf.
105 Current flags are:
106
107 flag bit | cpuid bit | meaning
108 -------------------------------------------------------------
109 | | time measures taken across
110 0 | 24 | multiple cpus are guaranteed to
111 | | be monotonic
112 -------------------------------------------------------------
113
114 Availability of this MSR must be checked via bit 3 in 0x4000001 cpuid
115 leaf prior to usage.
116
117
118MSR_KVM_WALL_CLOCK: 0x11
119
120 data and functioning: same as MSR_KVM_WALL_CLOCK_NEW. Use that instead.
121
122 This MSR falls outside the reserved KVM range and may be removed in the
123 future. Its usage is deprecated.
124
125 Availability of this MSR must be checked via bit 0 in 0x4000001 cpuid
126 leaf prior to usage.
127
128MSR_KVM_SYSTEM_TIME: 0x12
129
130 data and functioning: same as MSR_KVM_SYSTEM_TIME_NEW. Use that instead.
131
132 This MSR falls outside the reserved KVM range and may be removed in the
133 future. Its usage is deprecated.
134
135 Availability of this MSR must be checked via bit 0 in 0x4000001 cpuid
136 leaf prior to usage.
137
138 The suggested algorithm for detecting kvmclock presence is then:
139
140 if (!kvm_para_available()) /* refer to cpuid.txt */
141 return NON_PRESENT;
142
143 flags = cpuid_eax(0x40000001);
144 if (flags & 3) {
145 msr_kvm_system_time = MSR_KVM_SYSTEM_TIME_NEW;
146 msr_kvm_wall_clock = MSR_KVM_WALL_CLOCK_NEW;
147 return PRESENT;
148 } else if (flags & 0) {
149 msr_kvm_system_time = MSR_KVM_SYSTEM_TIME;
150 msr_kvm_wall_clock = MSR_KVM_WALL_CLOCK;
151 return PRESENT;
152 } else
153 return NON_PRESENT;