Hi Steve,
On 2019/12/17 22:40, Steven Price wrote:
> On Tue, Dec 17, 2019 at 01:55:49PM +, yezengr...@huawei.com wrote:
>> From: Zengruan Ye
>>
>> Support the vcpu_is_preempted() functionality under KVM/arm64. This will
>> enhance lock performance on overcommitted hosts (more runnable vcpus
>> than physical cpus in the system) as doing busy waits for preempted
>> vcpus will hurt system performance far worse than early yielding.
>>
>> unix benchmark result:
>> host: kernel 5.5.0-rc1, HiSilicon Kunpeng920, 8 cpus
>> guest: kernel 5.5.0-rc1, 16 vcpus
>>
>>test-case|after-patch| before-patch
>> +---+--
>> Dhrystone 2 using register variables | 334600751.0 lps | 335319028.3 lps
>> Double-Precision Whetstone | 32856.1 MWIPS | 32849.6
>> MWIPS
>> Execl Throughput | 3662.1 lps | 2718.0 lps
>> File Copy 1024 bufsize 2000 maxblocks |432906.4 KBps |158011.8
>> KBps
>> File Copy 256 bufsize 500 maxblocks|116023.0 KBps | 37664.0
>> KBps
>> File Copy 4096 bufsize 8000 maxblocks | 1432769.8 KBps |441108.8
>> KBps
>> Pipe Throughput| 6405029.6 lps | 6021457.6 lps
>> Pipe-based Context Switching |185872.7 lps |184255.3 lps
>> Process Creation | 4025.7 lps | 3706.6 lps
>> Shell Scripts (1 concurrent) | 6745.6 lpm | 6436.1 lpm
>> Shell Scripts (8 concurrent) | 998.7 lpm | 931.1 lpm
>> System Call Overhead | 3913363.1 lps | 3883287.8 lps
>> +---+--
>> System Benchmarks Index Score | 1835.1 | 1327.6
>>
>> Signed-off-by: Zengruan Ye
>> ---
>> arch/arm64/include/asm/paravirt.h | 3 +
>> arch/arm64/kernel/paravirt.c | 91 +++
>> arch/arm64/kernel/setup.c | 2 +
>> include/linux/cpuhotplug.h| 1 +
>> 4 files changed, 97 insertions(+)
>>
>> diff --git a/arch/arm64/include/asm/paravirt.h
>> b/arch/arm64/include/asm/paravirt.h
>> index 7b1c81b544bb..a2cd0183bbef 100644
>> --- a/arch/arm64/include/asm/paravirt.h
>> +++ b/arch/arm64/include/asm/paravirt.h
>> @@ -29,6 +29,8 @@ static inline u64 paravirt_steal_clock(int cpu)
>>
>> int __init pv_time_init(void);
>>
>> +int __init kvm_guest_init(void);
>> +
>
> This is a *very* generic name - I suggest something like pv_lock_init()
> so it's clear what the function actually does.
>
>> __visible bool __native_vcpu_is_preempted(int cpu);
>>
>> static inline bool pv_vcpu_is_preempted(int cpu)
>> @@ -39,6 +41,7 @@ static inline bool pv_vcpu_is_preempted(int cpu)
>> #else
>>
>> #define pv_time_init() do {} while (0)
>> +#define kvm_guest_init() do {} while (0)
>>
>> #endif // CONFIG_PARAVIRT
>>
>> diff --git a/arch/arm64/kernel/paravirt.c b/arch/arm64/kernel/paravirt.c
>> index d8f1ba8c22ce..a86dead40473 100644
>> --- a/arch/arm64/kernel/paravirt.c
>> +++ b/arch/arm64/kernel/paravirt.c
>> @@ -22,6 +22,7 @@
>> #include
>> #include
>> #include
>> +#include
>>
>> struct static_key paravirt_steal_enabled;
>> struct static_key paravirt_steal_rq_enabled;
>> @@ -158,3 +159,93 @@ int __init pv_time_init(void)
>>
>> return 0;
>> }
>> +
>> +DEFINE_PER_CPU(struct pvlock_vcpu_state, pvlock_vcpu_region) __aligned(64);
>> +EXPORT_PER_CPU_SYMBOL(pvlock_vcpu_region);
>> +
>> +static int pvlock_vcpu_state_dying_cpu(unsigned int cpu)
>> +{
>> +struct pvlock_vcpu_state *reg;
>> +
>> +reg = this_cpu_ptr(_vcpu_region);
>> +if (!reg)
>> +return -EFAULT;
>> +
>> +memset(reg, 0, sizeof(*reg));
>
> I might be missing something obvious here - but I don't see the point of
> this. The hypervisor might immediately overwrite the structure again.
> Indeed you should conside a mechanism for the guest to "unregister" the
> region - otherwise you will face issues with the likes of kexec.
>
> For pv_time the memory is allocated by the hypervisor not the guest to
> avoid lifetime issues about kexec.
Thanks for pointing it out to me! I'll update the memory allocation
mechanism of the PV lock structure to avoid lifetime issues about
kexec.
>
>> +
>> +return 0;
>> +}
>> +
>> +static int init_pvlock_vcpu_state(unsigned int cpu)
>> +{
>> +struct pvlock_vcpu_state *reg;
>> +struct arm_smccc_res res;
>> +
>> +reg = this_cpu_ptr(_vcpu_region);
>> +if (!reg)
>> +return -EFAULT;
>> +
>> +/* Pass the memory address to host via hypercall */
>> +arm_smccc_1_1_invoke(ARM_SMCCC_HV_PV_LOCK_PREEMPTED,
>> + virt_to_phys(reg), );
>> +
>> +return 0;
>> +}
>> +
>> +static bool kvm_vcpu_is_preempted(int cpu)
>> +{
>> +struct pvlock_vcpu_state *reg = _cpu(pvlock_vcpu_region, cpu);