RE: [PATCH 2/5] X86: Hyper-V: Enable IPI enlightenments

["Michael Kelley (EOSG)" <Michael.H.Kelley@microsoft.com>]

> -----Original Message-----
> From: kys@linuxonhyperv.com <kys@linuxonhyperv.com>
> Sent: Wednesday, April 25, 2018 11:13 AM
> To: x86@kernel.org; gregkh@linuxfoundation.org; linux-kernel@vger.kernel.org;
> devel@linuxdriverproject.org; olaf@aepfle.de; apw@canonical.com; jasowang@redhat.com;
> tglx@linutronix.de; hpa@zytor.com; Stephen Hemminger <sthemmin@microsoft.com>;
> Michael Kelley (EOSG) <Michael.H.Kelley@microsoft.com>; vkuznets@redhat.com
> Cc: KY Srinivasan <kys@microsoft.com>
> Subject: [PATCH 2/5] X86: Hyper-V: Enable IPI enlightenments
> 
> From: "K. Y. Srinivasan" <kys@microsoft.com>
> 
> Hyper-V supports hypercalls to implement IPI; use them.
> 
> Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
> ---
>  arch/x86/hyperv/hv_apic.c          | 125 +++++++++++++++++++++++++++++++++++++
>  arch/x86/hyperv/hv_init.c          |  17 +++++
>  arch/x86/include/asm/hyperv-tlfs.h |   9 +++
>  arch/x86/include/asm/mshyperv.h    |   1 +
>  4 files changed, 152 insertions(+)
> 
> diff --git a/arch/x86/hyperv/hv_apic.c b/arch/x86/hyperv/hv_apic.c
> index e0a5b36208fc..7f3322ecfb01 100644
> --- a/arch/x86/hyperv/hv_apic.c
> +++ b/arch/x86/hyperv/hv_apic.c
> @@ -30,6 +30,14 @@
>  #include <linux/slab.h>
>  #include <linux/cpuhotplug.h>
> 
> +struct ipi_arg_non_ex {
> +	u32 vector;
> +	u32 reserved;
> +	u64 cpu_mask;
> +};

I think we'd like to put structures like this, which are defined in the
Hyper-V Top Level Functional Spec, in hyperv-tlfs.h.  Also, the 5.0b
version of the TLFS, which is latest, shows this structure on page 100:

	u32 vector;
	u8  targetvtl;
	u8  reserved[3];
	u64 cpu_mask;

> +
> +static struct apic orig_apic;
> +
>  static u64 hv_apic_icr_read(void)
>  {
>  	u64 reg_val;
> @@ -85,8 +93,125 @@ static void hv_apic_eoi_write(u32 reg, u32 val)
>  	wrmsr(HV_X64_MSR_EOI, val, 0);
>  }
> 
> +/*
> + * IPI implementation on Hyper-V.
> + */
> +
> +static int __send_ipi_mask(const struct cpumask *mask, int vector)
> +{
> +	int cur_cpu, vcpu;
> +	struct ipi_arg_non_ex **arg;
> +	struct ipi_arg_non_ex *ipi_arg;
> +	int ret = 1;
> +	unsigned long flags;
> +
> +	if (cpumask_empty(mask))
> +		return 0;
> +
> +	if (!hv_hypercall_pg)
> +		return ret;
> +
> +	if ((vector < HV_IPI_LOW_VECTOR) || (vector > HV_IPI_HIGH_VECTOR))
> +		return ret;
> +
> +	local_irq_save(flags);
> +	arg = (struct ipi_arg_non_ex **)this_cpu_ptr(hyperv_pcpu_input_arg);
> +
> +	ipi_arg = *arg;
> +	if (unlikely(!ipi_arg))
> +		goto ipi_mask_done;
> +
> +
> +	ipi_arg->vector = vector;
> +	ipi_arg->reserved = 0;
> +	ipi_arg->cpu_mask = 0;
> +
> +	for_each_cpu(cur_cpu, mask) {
> +		vcpu = hv_cpu_number_to_vp_number(cur_cpu);
> +		if (vcpu >= 64)
> +			goto ipi_mask_done;
> +
> +		__set_bit(vcpu, (unsigned long *)&ipi_arg->cpu_mask);
> +	}
> +
> +	ret = hv_do_hypercall(HVCALL_SEND_IPI, ipi_arg, NULL);
> +
> +ipi_mask_done:
> +	local_irq_restore(flags);
> +	return ret;
> +}
> +
> +static int __send_ipi_one(int cpu, int vector)
> +{
> +	struct cpumask mask = CPU_MASK_NONE;
> +
> +	cpumask_set_cpu(cpu, &mask);
> +	return __send_ipi_mask(&mask, vector);
> +}
> +
> +static void hv_send_ipi(int cpu, int vector)
> +{
> +	if (__send_ipi_one(cpu, vector))
> +		orig_apic.send_IPI(cpu, vector);
> +}
> +
> +static void hv_send_ipi_mask(const struct cpumask *mask, int vector)
> +{
> +	if (__send_ipi_mask(mask, vector))
> +		orig_apic.send_IPI_mask(mask, vector);
> +}
> +
> +static void hv_send_ipi_mask_allbutself(const struct cpumask *mask, int vector)
> +{
> +	unsigned int this_cpu = smp_processor_id();
> +	struct cpumask new_mask;
> +	const struct cpumask *local_mask;
> +
> +	cpumask_copy(&new_mask, mask);
> +	cpumask_clear_cpu(this_cpu, &new_mask);
> +	local_mask = &new_mask;
> +	if (__send_ipi_mask(local_mask, vector))
> +		orig_apic.send_IPI_mask_allbutself(mask, vector);
> +}
> +
> +static void hv_send_ipi_allbutself(int vector)
> +{
> +	hv_send_ipi_mask_allbutself(cpu_online_mask, vector);
> +}
> +
> +static void hv_send_ipi_all(int vector)
> +{
> +	if (__send_ipi_mask(cpu_online_mask, vector))
> +		orig_apic.send_IPI_all(vector);
> +}
> +
> +static void hv_send_ipi_self(int vector)
> +{
> +	if (__send_ipi_one(smp_processor_id(), vector))
> +		orig_apic.send_IPI_self(vector);
> +}
> +
>  void __init hv_apic_init(void)
>  {
> +	if (ms_hyperv.hints & HV_X64_CLUSTER_IPI_RECOMMENDED) {
> +		if (hyperv_pcpu_input_arg == NULL)
> +			goto msr_based_access;
> +
> +		pr_info("Hyper-V: Using IPI hypercalls\n");
> +		/*
> +		 * Set the IPI entry points.
> +		 */
> +		orig_apic = *apic;
> +
> +		apic->send_IPI = hv_send_ipi;
> +		apic->send_IPI_mask = hv_send_ipi_mask;
> +		apic->send_IPI_mask_allbutself = hv_send_ipi_mask_allbutself;
> +		apic->send_IPI_allbutself = hv_send_ipi_allbutself;
> +		apic->send_IPI_all = hv_send_ipi_all;
> +		apic->send_IPI_self = hv_send_ipi_self;
> +	}
> +
> +msr_based_access:
>  	if (ms_hyperv.hints & HV_X64_APIC_ACCESS_RECOMMENDED) {
>  		pr_info("Hyper-V: Using MSR ased APIC access\n");
>  		apic_set_eoi_write(hv_apic_eoi_write);
> diff --git a/arch/x86/hyperv/hv_init.c b/arch/x86/hyperv/hv_init.c
> index 71e50fc2b7ef..a895662b6b4c 100644
> --- a/arch/x86/hyperv/hv_init.c
> +++ b/arch/x86/hyperv/hv_init.c
> @@ -91,12 +91,19 @@ EXPORT_SYMBOL_GPL(hv_vp_index);
>  struct hv_vp_assist_page **hv_vp_assist_page;
>  EXPORT_SYMBOL_GPL(hv_vp_assist_page);
> 
> +void  __percpu **hyperv_pcpu_input_arg;
> +EXPORT_SYMBOL_GPL(hyperv_pcpu_input_arg);
> +
>  u32 hv_max_vp_index;
> 
>  static int hv_cpu_init(unsigned int cpu)
>  {
>  	u64 msr_vp_index;
>  	struct hv_vp_assist_page **hvp = &hv_vp_assist_page[smp_processor_id()];
> +	void **input_arg;
> +
> +	input_arg = (void **)this_cpu_ptr(hyperv_pcpu_input_arg);
> +	*input_arg = page_address(alloc_page(GFP_ATOMIC));
> 
>  	hv_get_vp_index(msr_vp_index);
> 
> @@ -217,6 +224,10 @@ static int hv_cpu_die(unsigned int cpu)
>  {
>  	struct hv_reenlightenment_control re_ctrl;
>  	unsigned int new_cpu;
> +	void **input_arg;
> +
> +	input_arg = (void **)this_cpu_ptr(hyperv_pcpu_input_arg);
> +	free_page((unsigned long)*input_arg);
> 
>  	if (hv_vp_assist_page && hv_vp_assist_page[cpu])
>  		wrmsrl(HV_X64_MSR_VP_ASSIST_PAGE, 0);
> @@ -260,6 +271,12 @@ void __init hyperv_init(void)
>  	if ((ms_hyperv.features & required_msrs) != required_msrs)
>  		return;
> 
> +	/* Allocate the per-CPU state for the hypercall input arg */
> +	hyperv_pcpu_input_arg = alloc_percpu(void  *);
> +
> +	if (hyperv_pcpu_input_arg == NULL)
> +		return;
> +
>  	/* Allocate percpu VP index */
>  	hv_vp_index = kmalloc_array(num_possible_cpus(), sizeof(*hv_vp_index),
>  				    GFP_KERNEL);
> diff --git a/arch/x86/include/asm/hyperv-tlfs.h b/arch/x86/include/asm/hyperv-tlfs.h
> index 416cb0e0c496..646cf2ca2aaa 100644
> --- a/arch/x86/include/asm/hyperv-tlfs.h
> +++ b/arch/x86/include/asm/hyperv-tlfs.h
> @@ -164,6 +164,11 @@
>   */
>  #define HV_X64_DEPRECATING_AEOI_RECOMMENDED	(1 << 9)
> 
> +/*
> + * Recommend using cluster IPI hypercalls.
> + */
> +#define HV_X64_CLUSTER_IPI_RECOMMENDED         (1 << 10)
> +
>  /* Recommend using the newer ExProcessorMasks interface */
>  #define HV_X64_EX_PROCESSOR_MASKS_RECOMMENDED	(1 << 11)
> 
> @@ -329,10 +334,14 @@ struct hv_tsc_emulation_status {
>  #define HV_X64_MSR_HYPERCALL_PAGE_ADDRESS_MASK	\
>  		(~((1ull << HV_X64_MSR_HYPERCALL_PAGE_ADDRESS_SHIFT) - 1))
> 
> +#define HV_IPI_LOW_VECTOR	0x10
> +#define HV_IPI_HIGH_VECTOR	0xff
> +
>  /* Declare the various hypercall operations. */
>  #define HVCALL_FLUSH_VIRTUAL_ADDRESS_SPACE	0x0002
>  #define HVCALL_FLUSH_VIRTUAL_ADDRESS_LIST	0x0003
>  #define HVCALL_NOTIFY_LONG_SPIN_WAIT		0x0008
> +#define HVCALL_SEND_IPI				0x000b
>  #define HVCALL_FLUSH_VIRTUAL_ADDRESS_SPACE_EX  0x0013
>  #define HVCALL_FLUSH_VIRTUAL_ADDRESS_LIST_EX   0x0014
>  #define HVCALL_POST_MESSAGE			0x005c
> diff --git a/arch/x86/include/asm/mshyperv.h b/arch/x86/include/asm/mshyperv.h
> index bcced50037c1..f6045f3611de 100644
> --- a/arch/x86/include/asm/mshyperv.h
> +++ b/arch/x86/include/asm/mshyperv.h
> @@ -122,6 +122,7 @@ static inline void hv_disable_stimer0_percpu_irq(int irq) {}
>  #if IS_ENABLED(CONFIG_HYPERV)
>  extern struct clocksource *hyperv_cs;
>  extern void *hv_hypercall_pg;
> +extern void  __percpu  **hyperv_pcpu_input_arg;
> 
>  static inline u64 hv_do_hypercall(u64 control, void *input, void *output)
>  {
> --
> 2.15.1