Temporarily drop kvm->srcu before invoking kvm_vcpu_block() on a vCPU
that hasn't yet been initialized. Best case scenario, blocking while
holding kvm->srcu will degrade guest performance. Worst case scenario,
the vCPU will never get a wake event and the VM's tasks will hang
indefinitely on synchronize_srcu(), e.g. when trying update memslots.
E.g. running the "apic" KVM unit test often results in the test hanging
==> 6409/stack <==
[<0>] __synchronize_srcu.part.0+0x7a/0xa0
[<0>] kvm_swap_active_memslots+0x141/0x180
[<0>] kvm_set_memslot+0x2f9/0x470
[<0>] kvm_set_memory_region+0x29/0x40
[<0>] kvm_vm_ioctl+0x2c3/0xd70
[<0>] __x64_sys_ioctl+0x83/0xb0
[<0>] do_syscall_64+0x3b/0xc0
[<0>] entry_SYSCALL_64_after_hwframe+0x44/0xae
==> 6410/stack <==
[<0>] kvm_vcpu_block+0x36/0x80
[<0>] kvm_arch_vcpu_ioctl_run+0x17b1/0x1f50
[<0>] kvm_vcpu_ioctl+0x279/0x690
[<0>] __x64_sys_ioctl+0x83/0xb0
[<0>] do_syscall_64+0x3b/0xc0
[<0>] entry_SYSCALL_64_after_hwframe+0x44/0xae
While it's tempting to never acquire kvm->srcu for an uninitialized vCPU,
practically speaking there's no penalty to acquiring kvm->srcu "early"
as the KVM_MP_STATE_UNINITIALIZED path is a one-time thing per vCPU. On
the other hand, seemingly innocuous helpers like kvm_apic_accept_events()
and sync_regs() can theoretically reach code that might access
SRCU-protected data structures, e.g. sync_regs() can trigger forced
existing of nested mode via kvm_vcpu_ioctl_x86_set_vcpu_events().
Fixes: 5d8d2bfc5e65 ("KVM: x86: pull kvm->srcu read-side to kvm_arch_vcpu_ioctl_run")
Cc: Like Xu <like.xu.linux@xxxxxxxxx>
Signed-off-by: Sean Christopherson <seanjc@xxxxxxxxxx>
---
arch/x86/kvm/x86.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index e55de9b48d1a..8fd60887f38d 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -10411,7 +10411,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu)
* use before KVM has ever run the vCPU.
*/
WARN_ON_ONCE(kvm_lapic_hv_timer_in_use(vcpu));
+
+ srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
kvm_vcpu_block(vcpu);
+ vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
+
if (kvm_apic_accept_events(vcpu) < 0) {
r = 0;
goto out;
base-commit: 991f988b43c5ee82ef681907bfe979bee93a55c2
--
2.35.1.574.g5d30c73bfb-goog
