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