Re: [PATCH v5 09/10] x86/hyper-v: support extended CPU ranges for TLB flush hypercalls

[Vitaly Kuznetsov]

Jork Loeser <Jork.Loeser@xxxxxxxxxxxxx> writes:

>> -----Original Message-----
>> From: Vitaly Kuznetsov [mailto:vkuznets@xxxxxxxxxx]
>> Sent: Tuesday, May 30, 2017 04:34
>> To: x86@xxxxxxxxxx; devel@xxxxxxxxxxxxxxxxxxxxxx
>> Cc: linux-kernel@xxxxxxxxxxxxxxx; KY Srinivasan <kys@xxxxxxxxxxxxx>; Haiyang
>> Zhang <haiyangz@xxxxxxxxxxxxx>; Stephen Hemminger
>> <sthemmin@xxxxxxxxxxxxx>; Thomas Gleixner <tglx@xxxxxxxxxxxxx>; Ingo
>> Molnar <mingo@xxxxxxxxxx>; H. Peter Anvin <hpa@xxxxxxxxx>; Steven
>> Rostedt <rostedt@xxxxxxxxxxx>; Jork Loeser <Jork.Loeser@xxxxxxxxxxxxx>;
>> Simon Xiao <sixiao@xxxxxxxxxxxxx>; Andy Lutomirski <luto@xxxxxxxxxx>; Andy
>> Shevchenko <andy.shevchenko@xxxxxxxxx>
>> Subject: [PATCH v5 09/10] x86/hyper-v: support extended CPU ranges for TLB
>> flush hypercalls
>> 
>> Hyper-V hosts may support more than 64 vCPUs, we need to use
>> HVCALL_FLUSH_VIRTUAL_ADDRESS_SPACE_EX/LIST_EX hypercalls in this case.
>
>> +static inline int cpumask_to_vp_set(struct hv_flush_pcpu_ex *flush,
>> +				    const struct cpumask *cpus)
>> +{
>> +	int cpu, vcpu, vcpu_bank, vcpu_offset, cur_bank, nr_bank = 0;
>> +	bool has_cpus;
>> +
>> +	/*
>> +	 * We can't be sure that translated vCPU numbers will always be
>> +	 * in ascending order, so iterate over all possible banks and
>> +	 * check all vCPUs in it instead.
>> +	 */
>> +	for (cur_bank = 0; cur_bank < ms_hyperv.max_vp_index/64;
>> cur_bank++) {
>> +		has_cpus = false;
>> +		for_each_cpu(cpu, cpus) {
>> +			vcpu = hv_cpu_number_to_vp_number(cpu);
>> +			vcpu_bank = vcpu / 64;
>> +			vcpu_offset = vcpu % 64;
>> +
>> +			if (vcpu_bank != cur_bank)
>> +				continue;
>> +			__set_bit(vcpu_offset, (unsigned long *)
>> +				  &flush->hv_vp_set.bank_contents[nr_bank]);
>> +			if (!has_cpus) {
>> +				__set_bit(vcpu_bank, (unsigned long *)
>> +					  &flush->hv_vp_set.valid_bank_mask);
>> +				has_cpus = true;
>> +			}
>> +		}
>> +		if (has_cpus)
>> +			nr_bank++;
>> +	}
>> +
>> +	return nr_bank;
>> +}
>
> Note that the HV_VP_SET may contain empty banks. As such, consider
> enabling all (5) bits in the valid_bank_mask, and a single
> for_each(cpu, cpus) pass - setting the bits as per
> hv_cpu_number_to_vp_number(cpu).
>

Oh, I didn't know that! I'll switch to doing a single
pass. Unfortunately I have no Hyper-V setup with > 64 vCPUs so I can't
really test if it works or not (I mean empty banks, *_EX hypercalls
work).

-- 
  Vitaly