1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 2
3 ================= 3 =================
4 KVM Lock Overview 4 KVM Lock Overview
5 ================= 5 =================
6 6
7 1. Acquisition Orders 7 1. Acquisition Orders
8 --------------------- 8 ---------------------
9 9
10 The acquisition orders for mutexes are as foll 10 The acquisition orders for mutexes are as follows:
11 11
12 - cpus_read_lock() is taken outside kvm_lock 12 - cpus_read_lock() is taken outside kvm_lock
13 13
14 - kvm_usage_lock is taken outside cpus_read_lo <<
15 <<
16 - kvm->lock is taken outside vcpu->mutex 14 - kvm->lock is taken outside vcpu->mutex
17 15
18 - kvm->lock is taken outside kvm->slots_lock a 16 - kvm->lock is taken outside kvm->slots_lock and kvm->irq_lock
19 17
20 - kvm->slots_lock is taken outside kvm->irq_lo 18 - kvm->slots_lock is taken outside kvm->irq_lock, though acquiring
21 them together is quite rare. 19 them together is quite rare.
22 20
23 - kvm->mn_active_invalidate_count ensures that 21 - kvm->mn_active_invalidate_count ensures that pairs of
24 invalidate_range_start() and invalidate_rang 22 invalidate_range_start() and invalidate_range_end() callbacks
25 use the same memslots array. kvm->slots_loc 23 use the same memslots array. kvm->slots_lock and kvm->slots_arch_lock
26 are taken on the waiting side when modifying 24 are taken on the waiting side when modifying memslots, so MMU notifiers
27 must not take either kvm->slots_lock or kvm- 25 must not take either kvm->slots_lock or kvm->slots_arch_lock.
28 26
29 cpus_read_lock() vs kvm_lock: <<
30 <<
31 - Taking cpus_read_lock() outside of kvm_lock <<
32 being the official ordering, as it is quite <<
33 cpus_read_lock() while holding kvm_lock. Us <<
34 e.g. avoid complex operations when possible. <<
35 <<
36 For SRCU: 27 For SRCU:
37 28
38 - ``synchronize_srcu(&kvm->srcu)`` is called i 29 - ``synchronize_srcu(&kvm->srcu)`` is called inside critical sections
39 for kvm->lock, vcpu->mutex and kvm->slots_lo 30 for kvm->lock, vcpu->mutex and kvm->slots_lock. These locks _cannot_
40 be taken inside a kvm->srcu read-side critic 31 be taken inside a kvm->srcu read-side critical section; that is, the
41 following is broken:: 32 following is broken::
42 33
43 srcu_read_lock(&kvm->srcu); 34 srcu_read_lock(&kvm->srcu);
44 mutex_lock(&kvm->slots_lock); 35 mutex_lock(&kvm->slots_lock);
45 36
46 - kvm->slots_arch_lock instead is released bef 37 - kvm->slots_arch_lock instead is released before the call to
47 ``synchronize_srcu()``. It _can_ therefore 38 ``synchronize_srcu()``. It _can_ therefore be taken inside a
48 kvm->srcu read-side critical section, for ex 39 kvm->srcu read-side critical section, for example while processing
49 a vmexit. 40 a vmexit.
50 41
51 On x86: 42 On x86:
52 43
53 - vcpu->mutex is taken outside kvm->arch.hyper 44 - vcpu->mutex is taken outside kvm->arch.hyperv.hv_lock and kvm->arch.xen.xen_lock
54 45
55 - kvm->arch.mmu_lock is an rwlock; critical se 46 - kvm->arch.mmu_lock is an rwlock; critical sections for
56 kvm->arch.tdp_mmu_pages_lock and kvm->arch.m 47 kvm->arch.tdp_mmu_pages_lock and kvm->arch.mmu_unsync_pages_lock must
57 also take kvm->arch.mmu_lock 48 also take kvm->arch.mmu_lock
58 49
59 Everything else is a leaf: no other lock is ta 50 Everything else is a leaf: no other lock is taken inside the critical
60 sections. 51 sections.
61 52
62 2. Exception 53 2. Exception
63 ------------ 54 ------------
64 55
65 Fast page fault: 56 Fast page fault:
66 57
67 Fast page fault is the fast path which fixes t 58 Fast page fault is the fast path which fixes the guest page fault out of
68 the mmu-lock on x86. Currently, the page fault 59 the mmu-lock on x86. Currently, the page fault can be fast in one of the
69 following two cases: 60 following two cases:
70 61
71 1. Access Tracking: The SPTE is not present, b 62 1. Access Tracking: The SPTE is not present, but it is marked for access
72 tracking. That means we need to restore the 63 tracking. That means we need to restore the saved R/X bits. This is
73 described in more detail later below. 64 described in more detail later below.
74 65
75 2. Write-Protection: The SPTE is present and t 66 2. Write-Protection: The SPTE is present and the fault is caused by
76 write-protect. That means we just need to c 67 write-protect. That means we just need to change the W bit of the spte.
77 68
78 What we use to avoid all the races is the Host 69 What we use to avoid all the races is the Host-writable bit and MMU-writable bit
79 on the spte: 70 on the spte:
80 71
81 - Host-writable means the gfn is writable in t 72 - Host-writable means the gfn is writable in the host kernel page tables and in
82 its KVM memslot. 73 its KVM memslot.
83 - MMU-writable means the gfn is writable in th 74 - MMU-writable means the gfn is writable in the guest's mmu and it is not
84 write-protected by shadow page write-protect 75 write-protected by shadow page write-protection.
85 76
86 On fast page fault path, we will use cmpxchg t 77 On fast page fault path, we will use cmpxchg to atomically set the spte W
87 bit if spte.HOST_WRITEABLE = 1 and spte.WRITE_ 78 bit if spte.HOST_WRITEABLE = 1 and spte.WRITE_PROTECT = 1, to restore the saved
88 R/X bits if for an access-traced spte, or both 79 R/X bits if for an access-traced spte, or both. This is safe because whenever
89 changing these bits can be detected by cmpxchg 80 changing these bits can be detected by cmpxchg.
90 81
91 But we need carefully check these cases: 82 But we need carefully check these cases:
92 83
93 1) The mapping from gfn to pfn 84 1) The mapping from gfn to pfn
94 85
95 The mapping from gfn to pfn may be changed sin 86 The mapping from gfn to pfn may be changed since we can only ensure the pfn
96 is not changed during cmpxchg. This is a ABA p 87 is not changed during cmpxchg. This is a ABA problem, for example, below case
97 will happen: 88 will happen:
98 89
99 +--------------------------------------------- 90 +------------------------------------------------------------------------+
100 | At the beginning:: 91 | At the beginning:: |
101 | 92 | |
102 | gpte = gfn1 93 | gpte = gfn1 |
103 | gfn1 is mapped to pfn1 on host 94 | gfn1 is mapped to pfn1 on host |
104 | spte is the shadow page table entry co 95 | spte is the shadow page table entry corresponding with gpte and |
105 | spte = pfn1 96 | spte = pfn1 |
106 +--------------------------------------------- 97 +------------------------------------------------------------------------+
107 | On fast page fault path: 98 | On fast page fault path: |
108 +------------------------------------+-------- 99 +------------------------------------+-----------------------------------+
109 | CPU 0: | CPU 1: 100 | CPU 0: | CPU 1: |
110 +------------------------------------+-------- 101 +------------------------------------+-----------------------------------+
111 | :: | 102 | :: | |
112 | | 103 | | |
113 | old_spte = *spte; | 104 | old_spte = *spte; | |
114 +------------------------------------+-------- 105 +------------------------------------+-----------------------------------+
115 | | pfn1 is 106 | | pfn1 is swapped out:: |
116 | | 107 | | |
117 | | spte 108 | | spte = 0; |
118 | | 109 | | |
119 | | pfn1 is 110 | | pfn1 is re-alloced for gfn2. |
120 | | 111 | | |
121 | | gpte is 112 | | gpte is changed to point to |
122 | | gfn2 by 113 | | gfn2 by the guest:: |
123 | | 114 | | |
124 | | spte 115 | | spte = pfn1; |
125 +------------------------------------+-------- 116 +------------------------------------+-----------------------------------+
126 | :: 117 | :: |
127 | 118 | |
128 | if (cmpxchg(spte, old_spte, old_spte+W) 119 | if (cmpxchg(spte, old_spte, old_spte+W) |
129 | mark_page_dirty(vcpu->kvm, gfn1) 120 | mark_page_dirty(vcpu->kvm, gfn1) |
130 | OOPS!!! 121 | OOPS!!! |
131 +--------------------------------------------- 122 +------------------------------------------------------------------------+
132 123
133 We dirty-log for gfn1, that means gfn2 is lost 124 We dirty-log for gfn1, that means gfn2 is lost in dirty-bitmap.
134 125
135 For direct sp, we can easily avoid it since th 126 For direct sp, we can easily avoid it since the spte of direct sp is fixed
136 to gfn. For indirect sp, we disabled fast pag 127 to gfn. For indirect sp, we disabled fast page fault for simplicity.
137 128
138 A solution for indirect sp could be to pin the 129 A solution for indirect sp could be to pin the gfn, for example via
139 gfn_to_pfn_memslot_atomic, before the cmpxchg. !! 130 kvm_vcpu_gfn_to_pfn_atomic, before the cmpxchg. After the pinning:
140 131
141 - We have held the refcount of pfn; that means 132 - We have held the refcount of pfn; that means the pfn can not be freed and
142 be reused for another gfn. 133 be reused for another gfn.
143 - The pfn is writable and therefore it cannot 134 - The pfn is writable and therefore it cannot be shared between different gfns
144 by KSM. 135 by KSM.
145 136
146 Then, we can ensure the dirty bitmaps is corre 137 Then, we can ensure the dirty bitmaps is correctly set for a gfn.
147 138
148 2) Dirty bit tracking 139 2) Dirty bit tracking
149 140
150 In the origin code, the spte can be fast updat 141 In the origin code, the spte can be fast updated (non-atomically) if the
151 spte is read-only and the Accessed bit has alr 142 spte is read-only and the Accessed bit has already been set since the
152 Accessed bit and Dirty bit can not be lost. 143 Accessed bit and Dirty bit can not be lost.
153 144
154 But it is not true after fast page fault since 145 But it is not true after fast page fault since the spte can be marked
155 writable between reading spte and updating spt 146 writable between reading spte and updating spte. Like below case:
156 147
157 +--------------------------------------------- 148 +------------------------------------------------------------------------+
158 | At the beginning:: 149 | At the beginning:: |
159 | 150 | |
160 | spte.W = 0 151 | spte.W = 0 |
161 | spte.Accessed = 1 152 | spte.Accessed = 1 |
162 +------------------------------------+-------- 153 +------------------------------------+-----------------------------------+
163 | CPU 0: | CPU 1: 154 | CPU 0: | CPU 1: |
164 +------------------------------------+-------- 155 +------------------------------------+-----------------------------------+
165 | In mmu_spte_clear_track_bits():: | 156 | In mmu_spte_clear_track_bits():: | |
166 | | 157 | | |
167 | old_spte = *spte; | 158 | old_spte = *spte; | |
168 | | 159 | | |
169 | | 160 | | |
170 | /* 'if' condition is satisfied. */| 161 | /* 'if' condition is satisfied. */| |
171 | if (old_spte.Accessed == 1 && | 162 | if (old_spte.Accessed == 1 && | |
172 | old_spte.W == 0) | 163 | old_spte.W == 0) | |
173 | spte = 0ull; | 164 | spte = 0ull; | |
174 +------------------------------------+-------- 165 +------------------------------------+-----------------------------------+
175 | | on fast 166 | | on fast page fault path:: |
176 | | 167 | | |
177 | | spte 168 | | spte.W = 1 |
178 | | 169 | | |
179 | | memory 170 | | memory write on the spte:: |
180 | | 171 | | |
181 | | spte 172 | | spte.Dirty = 1 |
182 +------------------------------------+-------- 173 +------------------------------------+-----------------------------------+
183 | :: | 174 | :: | |
184 | | 175 | | |
185 | else | 176 | else | |
186 | old_spte = xchg(spte, 0ull) | 177 | old_spte = xchg(spte, 0ull) | |
187 | if (old_spte.Accessed == 1) | 178 | if (old_spte.Accessed == 1) | |
188 | kvm_set_pfn_accessed(spte.pfn);| 179 | kvm_set_pfn_accessed(spte.pfn);| |
189 | if (old_spte.Dirty == 1) | 180 | if (old_spte.Dirty == 1) | |
190 | kvm_set_pfn_dirty(spte.pfn); | 181 | kvm_set_pfn_dirty(spte.pfn); | |
191 | OOPS!!! | 182 | OOPS!!! | |
192 +------------------------------------+-------- 183 +------------------------------------+-----------------------------------+
193 184
194 The Dirty bit is lost in this case. 185 The Dirty bit is lost in this case.
195 186
196 In order to avoid this kind of issue, we alway 187 In order to avoid this kind of issue, we always treat the spte as "volatile"
197 if it can be updated out of mmu-lock [see spte 188 if it can be updated out of mmu-lock [see spte_has_volatile_bits()]; it means
198 the spte is always atomically updated in this 189 the spte is always atomically updated in this case.
199 190
200 3) flush tlbs due to spte updated 191 3) flush tlbs due to spte updated
201 192
202 If the spte is updated from writable to read-o 193 If the spte is updated from writable to read-only, we should flush all TLBs,
203 otherwise rmap_write_protect will find a read- 194 otherwise rmap_write_protect will find a read-only spte, even though the
204 writable spte might be cached on a CPU's TLB. 195 writable spte might be cached on a CPU's TLB.
205 196
206 As mentioned before, the spte can be updated t 197 As mentioned before, the spte can be updated to writable out of mmu-lock on
207 fast page fault path. In order to easily audit 198 fast page fault path. In order to easily audit the path, we see if TLBs needing
208 to be flushed caused this reason in mmu_spte_u 199 to be flushed caused this reason in mmu_spte_update() since this is a common
209 function to update spte (present -> present). 200 function to update spte (present -> present).
210 201
211 Since the spte is "volatile" if it can be upda 202 Since the spte is "volatile" if it can be updated out of mmu-lock, we always
212 atomically update the spte and the race caused 203 atomically update the spte and the race caused by fast page fault can be avoided.
213 See the comments in spte_has_volatile_bits() a 204 See the comments in spte_has_volatile_bits() and mmu_spte_update().
214 205
215 Lockless Access Tracking: 206 Lockless Access Tracking:
216 207
217 This is used for Intel CPUs that are using EPT 208 This is used for Intel CPUs that are using EPT but do not support the EPT A/D
218 bits. In this case, PTEs are tagged as A/D dis 209 bits. In this case, PTEs are tagged as A/D disabled (using ignored bits), and
219 when the KVM MMU notifier is called to track a 210 when the KVM MMU notifier is called to track accesses to a page (via
220 kvm_mmu_notifier_clear_flush_young), it marks 211 kvm_mmu_notifier_clear_flush_young), it marks the PTE not-present in hardware
221 by clearing the RWX bits in the PTE and storin 212 by clearing the RWX bits in the PTE and storing the original R & X bits in more
222 unused/ignored bits. When the VM tries to acce 213 unused/ignored bits. When the VM tries to access the page later on, a fault is
223 generated and the fast page fault mechanism de 214 generated and the fast page fault mechanism described above is used to
224 atomically restore the PTE to a Present state. 215 atomically restore the PTE to a Present state. The W bit is not saved when the
225 PTE is marked for access tracking and during r 216 PTE is marked for access tracking and during restoration to the Present state,
226 the W bit is set depending on whether or not i 217 the W bit is set depending on whether or not it was a write access. If it
227 wasn't, then the W bit will remain clear until 218 wasn't, then the W bit will remain clear until a write access happens, at which
228 time it will be set using the Dirty tracking m 219 time it will be set using the Dirty tracking mechanism described above.
229 220
230 3. Reference 221 3. Reference
231 ------------ 222 ------------
232 223
233 ``kvm_lock`` 224 ``kvm_lock``
234 ^^^^^^^^^^^^ 225 ^^^^^^^^^^^^
235 226
236 :Type: mutex 227 :Type: mutex
237 :Arch: any 228 :Arch: any
238 :Protects: - vm_list 229 :Protects: - vm_list
239 !! 230 - kvm_usage_count
240 ``kvm_usage_lock`` <<
241 ^^^^^^^^^^^^^^^^^^ <<
242 <<
243 :Type: mutex <<
244 :Arch: any <<
245 :Protects: - kvm_usage_count <<
246 - hardware virtualization enab 231 - hardware virtualization enable/disable
247 :Comment: Exists to allow taking cpus_re !! 232 :Comment: KVM also disables CPU hotplug via cpus_read_lock() during
248 protected, which simplifies th !! 233 enable/disable.
249 234
250 ``kvm->mn_invalidate_lock`` 235 ``kvm->mn_invalidate_lock``
251 ^^^^^^^^^^^^^^^^^^^^^^^^^^^ 236 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
252 237
253 :Type: spinlock_t 238 :Type: spinlock_t
254 :Arch: any 239 :Arch: any
255 :Protects: mn_active_invalidate_count, mn 240 :Protects: mn_active_invalidate_count, mn_memslots_update_rcuwait
256 241
257 ``kvm_arch::tsc_write_lock`` 242 ``kvm_arch::tsc_write_lock``
258 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 243 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
259 244
260 :Type: raw_spinlock_t 245 :Type: raw_spinlock_t
261 :Arch: x86 246 :Arch: x86
262 :Protects: - kvm_arch::{last_tsc_write,la 247 :Protects: - kvm_arch::{last_tsc_write,last_tsc_nsec,last_tsc_offset}
263 - tsc offset in vmcb 248 - tsc offset in vmcb
264 :Comment: 'raw' because updating the tsc 249 :Comment: 'raw' because updating the tsc offsets must not be preempted.
265 250
266 ``kvm->mmu_lock`` 251 ``kvm->mmu_lock``
267 ^^^^^^^^^^^^^^^^^ 252 ^^^^^^^^^^^^^^^^^
268 :Type: spinlock_t or rwlock_t 253 :Type: spinlock_t or rwlock_t
269 :Arch: any 254 :Arch: any
270 :Protects: -shadow page/shadow tlb entry 255 :Protects: -shadow page/shadow tlb entry
271 :Comment: it is a spinlock since it is u 256 :Comment: it is a spinlock since it is used in mmu notifier.
272 257
273 ``kvm->srcu`` 258 ``kvm->srcu``
274 ^^^^^^^^^^^^^ 259 ^^^^^^^^^^^^^
275 :Type: srcu lock 260 :Type: srcu lock
276 :Arch: any 261 :Arch: any
277 :Protects: - kvm->memslots 262 :Protects: - kvm->memslots
278 - kvm->buses 263 - kvm->buses
279 :Comment: The srcu read lock must be hel 264 :Comment: The srcu read lock must be held while accessing memslots (e.g.
280 when using gfn_to_* functions) 265 when using gfn_to_* functions) and while accessing in-kernel
281 MMIO/PIO address->device struc 266 MMIO/PIO address->device structure mapping (kvm->buses).
282 The srcu index can be stored i 267 The srcu index can be stored in kvm_vcpu->srcu_idx per vcpu
283 if it is needed by multiple fu 268 if it is needed by multiple functions.
284 269
285 ``kvm->slots_arch_lock`` 270 ``kvm->slots_arch_lock``
286 ^^^^^^^^^^^^^^^^^^^^^^^^ 271 ^^^^^^^^^^^^^^^^^^^^^^^^
287 :Type: mutex 272 :Type: mutex
288 :Arch: any (only needed on x86 though 273 :Arch: any (only needed on x86 though)
289 :Protects: any arch-specific fields of me 274 :Protects: any arch-specific fields of memslots that have to be modified
290 in a ``kvm->srcu`` read-side c 275 in a ``kvm->srcu`` read-side critical section.
291 :Comment: must be held before reading th 276 :Comment: must be held before reading the pointer to the current memslots,
292 until after all changes to the 277 until after all changes to the memslots are complete
293 278
294 ``wakeup_vcpus_on_cpu_lock`` 279 ``wakeup_vcpus_on_cpu_lock``
295 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 280 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
296 :Type: spinlock_t 281 :Type: spinlock_t
297 :Arch: x86 282 :Arch: x86
298 :Protects: wakeup_vcpus_on_cpu 283 :Protects: wakeup_vcpus_on_cpu
299 :Comment: This is a per-CPU lock and it 284 :Comment: This is a per-CPU lock and it is used for VT-d posted-interrupts.
300 When VT-d posted-interrupts ar 285 When VT-d posted-interrupts are supported and the VM has assigned
301 devices, we put the blocked vC 286 devices, we put the blocked vCPU on the list blocked_vcpu_on_cpu
302 protected by blocked_vcpu_on_c 287 protected by blocked_vcpu_on_cpu_lock. When VT-d hardware issues
303 wakeup notification event sinc 288 wakeup notification event since external interrupts from the
304 assigned devices happens, we w 289 assigned devices happens, we will find the vCPU on the list to
305 wakeup. 290 wakeup.
306 291
307 ``vendor_module_lock`` 292 ``vendor_module_lock``
308 ^^^^^^^^^^^^^^^^^^^^^^ !! 293 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
309 :Type: mutex 294 :Type: mutex
310 :Arch: x86 295 :Arch: x86
311 :Protects: loading a vendor module (kvm_a 296 :Protects: loading a vendor module (kvm_amd or kvm_intel)
312 :Comment: Exists because using kvm_lock !! 297 :Comment: Exists because using kvm_lock leads to deadlock. cpu_hotplug_lock is
313 in notifiers, e.g. __kvmclock_cpufreq_noti !! 298 taken outside of kvm_lock, e.g. in KVM's CPU online/offline callbacks, and
314 cpu_hotplug_lock is held, e.g. from cpufre !! 299 many operations need to take cpu_hotplug_lock when loading a vendor module,
315 operations need to take cpu_hotplug_lock w !! 300 e.g. updating static calls.
316 updating static calls. <<
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.