Re: [PATCH 03/18] TSC reset compensation

[Zachary Amsden <zamsden@redhat.com>]

On 07/18/2010 04:34 AM, Avi Kivity wrote:
> On 07/13/2010 05:25 AM, Zachary Amsden wrote:
>> Attempt to synchronize TSCs which are reset to the same value.  In the
>> case of a reliable hardware TSC, we can just re-use the same offset, but
>> on non-reliable hardware, we can get closer by adjusting the offset to
>> match the elapsed time.
>>
>> diff --git a/arch/x86/include/asm/kvm_host.h 
>> b/arch/x86/include/asm/kvm_host.h
>> index 3b4efe2..4b42893 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -396,6 +396,9 @@ struct kvm_arch {
>>       unsigned long irq_sources_bitmap;
>>       s64 kvmclock_offset;
>>       spinlock_t tsc_write_lock;
>> +    u64 last_tsc_nsec;
>> +    u64 last_tsc_offset;
>> +    u64 last_tsc_write;
>
> So that we know what the lock protects, let's have
>
>    struct kvm_global_tsc {
>          spinlock_t lock;
>          ...
>     } tsc;
>
>> @@ -896,10 +896,39 @@ static DEFINE_PER_CPU(unsigned long, cpu_tsc_khz);
>>   void guest_write_tsc(struct kvm_vcpu *vcpu, u64 data)
>>   {
>>       struct kvm *kvm = vcpu->kvm;
>> -    u64 offset;
>> +    u64 offset, ns, elapsed;
>> +    struct timespec ts;
>>
>>       spin_lock(&kvm->arch.tsc_write_lock);
>>       offset = data - native_read_tsc();
>> +    ktime_get_ts(&ts);
>> +    monotonic_to_bootbased(&ts);
>> +    ns = timespec_to_ns(&ts);
>> +    elapsed = ns - kvm->arch.last_tsc_nsec;
>> +
>> +    /*
>> +     * Special case: identical write to TSC within 5 seconds of
>> +     * another CPU is interpreted as an attempt to synchronize
>> +     * (the 5 seconds is to accomodate host load / swapping).
>> +     *
>> +     * In that case, for a reliable TSC, we can match TSC offsets,
>> +     * or make a best guest using kernel_ns value.
>> +     */
>> +    if (data == kvm->arch.last_tsc_write&&  elapsed<  5ULL * 
>> NSEC_PER_SEC) {
>> +        if (!check_tsc_unstable()) {
>> +            offset = kvm->arch.last_tsc_offset;
>> +            pr_debug("kvm: matched tsc offset for %llu\n", data);
>> +        } else {
>> +            u64 tsc_delta = elapsed * __get_cpu_var(cpu_tsc_khz);
>> +            tsc_delta = tsc_delta / USEC_PER_SEC;
>> +            offset += tsc_delta;
>> +            pr_debug("kvm: adjusted tsc offset by %llu\n", tsc_delta);
>> +        }
>> +        ns = kvm->arch.last_tsc_nsec;
>> +    }
>> +    kvm->arch.last_tsc_nsec = ns;
>> +    kvm->arch.last_tsc_write = data;
>> +    kvm->arch.last_tsc_offset = offset;
>
> We'd have a false alarm here during a reset within 5 seconds of boot.  
> Does it matter?  Easy to work around by forgetting the state during 
> reset.
>

Not forgetting, but ignoring; reset within 5 seconds will not reset TSC, 
which normally is fine.  The problem is that one CPU could reset within 
5 seconds and one slightly after.  Forgetting during reset is a good 
solution.