7 The acquisition orders for mutexes are as follows:
9 - kvm->lock is taken outside vcpu->mutex
11 - kvm->lock is taken outside kvm->slots_lock and kvm->irq_lock
13 - kvm->slots_lock is taken outside kvm->irq_lock, though acquiring
14 them together is quite rare.
16 On x86, vcpu->mutex is taken outside kvm->arch.hyperv.hv_lock.
18 Everything else is a leaf: no other lock is taken inside the critical
26 Fast page fault is the fast path which fixes the guest page fault out of
27 the mmu-lock on x86. Currently, the page fault can be fast in one of the
30 1. Access Tracking: The SPTE is not present, but it is marked for access
31 tracking i.e. the SPTE_SPECIAL_MASK is set. That means we need to
32 restore the saved R/X bits. This is described in more detail later below.
34 2. Write-Protection: The SPTE is present and the fault is
35 caused by write-protect. That means we just need to change the W bit of the
38 What we use to avoid all the race is the SPTE_HOST_WRITEABLE bit and
39 SPTE_MMU_WRITEABLE bit on the spte:
40 - SPTE_HOST_WRITEABLE means the gfn is writable on host.
41 - SPTE_MMU_WRITEABLE means the gfn is writable on mmu. The bit is set when
42 the gfn is writable on guest mmu and it is not write-protected by shadow
43 page write-protection.
45 On fast page fault path, we will use cmpxchg to atomically set the spte W
46 bit if spte.SPTE_HOST_WRITEABLE = 1 and spte.SPTE_WRITE_PROTECT = 1, or
47 restore the saved R/X bits if VMX_EPT_TRACK_ACCESS mask is set, or both. This
48 is safe because whenever changing these bits can be detected by cmpxchg.
50 But we need carefully check these cases:
51 1): The mapping from gfn to pfn
52 The mapping from gfn to pfn may be changed since we can only ensure the pfn
53 is not changed during cmpxchg. This is a ABA problem, for example, below case
58 gfn1 is mapped to pfn1 on host
59 spte is the shadow page table entry corresponding with gpte and
63 on fast page fault path:
69 pfn1 is re-alloced for gfn2.
71 gpte is changed to point to
75 if (cmpxchg(spte, old_spte, old_spte+W)
76 mark_page_dirty(vcpu->kvm, gfn1)
79 We dirty-log for gfn1, that means gfn2 is lost in dirty-bitmap.
81 For direct sp, we can easily avoid it since the spte of direct sp is fixed
82 to gfn. For indirect sp, before we do cmpxchg, we call gfn_to_pfn_atomic()
83 to pin gfn to pfn, because after gfn_to_pfn_atomic():
84 - We have held the refcount of pfn that means the pfn can not be freed and
85 be reused for another gfn.
86 - The pfn is writable that means it can not be shared between different gfns
89 Then, we can ensure the dirty bitmaps is correctly set for a gfn.
91 Currently, to simplify the whole things, we disable fast page fault for
94 2): Dirty bit tracking
95 In the origin code, the spte can be fast updated (non-atomically) if the
96 spte is read-only and the Accessed bit has already been set since the
97 Accessed bit and Dirty bit can not be lost.
99 But it is not true after fast page fault since the spte can be marked
100 writable between reading spte and updating spte. Like below case:
107 In mmu_spte_clear_track_bits():
111 /* 'if' condition is satisfied. */
112 if (old_spte.Accessed == 1 &&
115 on fast page fault path:
117 memory write on the spte:
122 old_spte = xchg(spte, 0ull)
125 if (old_spte.Accessed == 1)
126 kvm_set_pfn_accessed(spte.pfn);
127 if (old_spte.Dirty == 1)
128 kvm_set_pfn_dirty(spte.pfn);
131 The Dirty bit is lost in this case.
133 In order to avoid this kind of issue, we always treat the spte as "volatile"
134 if it can be updated out of mmu-lock, see spte_has_volatile_bits(), it means,
135 the spte is always atomically updated in this case.
137 3): flush tlbs due to spte updated
138 If the spte is updated from writable to readonly, we should flush all TLBs,
139 otherwise rmap_write_protect will find a read-only spte, even though the
140 writable spte might be cached on a CPU's TLB.
142 As mentioned before, the spte can be updated to writable out of mmu-lock on
143 fast page fault path, in order to easily audit the path, we see if TLBs need
144 be flushed caused by this reason in mmu_spte_update() since this is a common
145 function to update spte (present -> present).
147 Since the spte is "volatile" if it can be updated out of mmu-lock, we always
148 atomically update the spte, the race caused by fast page fault can be avoided,
149 See the comments in spte_has_volatile_bits() and mmu_spte_update().
151 Lockless Access Tracking:
153 This is used for Intel CPUs that are using EPT but do not support the EPT A/D
154 bits. In this case, when the KVM MMU notifier is called to track accesses to a
155 page (via kvm_mmu_notifier_clear_flush_young), it marks the PTE as not-present
156 by clearing the RWX bits in the PTE and storing the original R & X bits in
157 some unused/ignored bits. In addition, the SPTE_SPECIAL_MASK is also set on the
158 PTE (using the ignored bit 62). When the VM tries to access the page later on,
159 a fault is generated and the fast page fault mechanism described above is used
160 to atomically restore the PTE to a Present state. The W bit is not saved when
161 the PTE is marked for access tracking and during restoration to the Present
162 state, the W bit is set depending on whether or not it was a write access. If
163 it wasn't, then the W bit will remain clear until a write access happens, at
164 which time it will be set using the Dirty tracking mechanism described above.
177 Protects: - hardware virtualization enable/disable
178 Comment: 'raw' because hardware enabling/disabling must be atomic /wrt
181 Name: kvm_arch::tsc_write_lock
184 Protects: - kvm_arch::{last_tsc_write,last_tsc_nsec,last_tsc_offset}
186 Comment: 'raw' because updating the tsc offsets must not be preempted.
191 Protects: -shadow page/shadow tlb entry
192 Comment: it is a spinlock since it is used in mmu notifier.
197 Protects: - kvm->memslots
199 Comment: The srcu read lock must be held while accessing memslots (e.g.
200 when using gfn_to_* functions) and while accessing in-kernel
201 MMIO/PIO address->device structure mapping (kvm->buses).
202 The srcu index can be stored in kvm_vcpu->srcu_idx per vcpu
203 if it is needed by multiple functions.
205 Name: blocked_vcpu_on_cpu_lock
208 Protects: blocked_vcpu_on_cpu
209 Comment: This is a per-CPU lock and it is used for VT-d posted-interrupts.
210 When VT-d posted-interrupts is supported and the VM has assigned
211 devices, we put the blocked vCPU on the list blocked_vcpu_on_cpu
212 protected by blocked_vcpu_on_cpu_lock, when VT-d hardware issues
213 wakeup notification event since external interrupts from the
214 assigned devices happens, we will find the vCPU on the list to