Re: [PATCH v2] KVM: Check the allocation of pv cpu mask

[Haiwei Li]

On 20/9/3 18:39, Vitaly Kuznetsov wrote:
> Haiwei Li <lihaiwei.kernel@xxxxxxxxx> writes:
>
> > From: Haiwei Li <lihaiwei@xxxxxxxxxxx>
> >
> > check the allocation of per-cpu __pv_cpu_mask. Initialize ops only when
> > successful.
> >
> > Signed-off-by: Haiwei Li <lihaiwei@xxxxxxxxxxx>
> > ---
> >    arch/x86/kernel/kvm.c | 24 ++++++++++++++++++++----
> >    1 file changed, 20 insertions(+), 4 deletions(-)
> >
> > diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
> > index 08320b0b2b27..d3c062e551d7 100644
> > --- a/arch/x86/kernel/kvm.c
> > +++ b/arch/x86/kernel/kvm.c
> > @@ -555,7 +555,6 @@ static void kvm_send_ipi_mask_allbutself(const
> > struct cpumask *mask, int vector)
> >    static void kvm_setup_pv_ipi(void)
> >    {
> >    	apic->send_IPI_mask = kvm_send_ipi_mask;
> > -	apic->send_IPI_mask_allbutself = kvm_send_ipi_mask_allbutself;
> >    	pr_info("setup PV IPIs\n");
> >    }
> >
> > @@ -654,7 +653,6 @@ static void __init kvm_guest_init(void)
> >    	}
> >
> >    	if (pv_tlb_flush_supported()) {
> > -		pv_ops.mmu.flush_tlb_others = kvm_flush_tlb_others;
> >    		pv_ops.mmu.tlb_remove_table = tlb_remove_table;
> >    		pr_info("KVM setup pv remote TLB flush\n");
> >    	}
> > @@ -767,6 +765,14 @@ static __init int activate_jump_labels(void)
> >    }
> >    arch_initcall(activate_jump_labels);
> >
> > +static void kvm_free_pv_cpu_mask(void)
> > +{
> > +	unsigned int cpu;
> > +
> > +	for_each_possible_cpu(cpu)
> > +		free_cpumask_var(per_cpu(__pv_cpu_mask, cpu));
> > +}
> > +
> >    static __init int kvm_alloc_cpumask(void)
> >    {
> >    	int cpu;
> > @@ -785,11 +791,21 @@ static __init int kvm_alloc_cpumask(void)
> >
> >    	if (alloc)
> >    		for_each_possible_cpu(cpu) {
> > -			zalloc_cpumask_var_node(per_cpu_ptr(&__pv_cpu_mask, cpu),
> > -				GFP_KERNEL, cpu_to_node(cpu));
> > +			if (!zalloc_cpumask_var_node(
> > +				per_cpu_ptr(&__pv_cpu_mask, cpu),
> > +				GFP_KERNEL, cpu_to_node(cpu)))
> > +				goto zalloc_cpumask_fail;
> >    		}
> >
> > +#if defined(CONFIG_SMP)
> > +	apic->send_IPI_mask_allbutself = kvm_send_ipi_mask_allbutself;
> > +#endif
> > +	pv_ops.mmu.flush_tlb_others = kvm_flush_tlb_others;
>
> This is too late I'm afraid. If I'm not mistaken PV patching happens
> earlier, so .init.guest_late_init (kvm_guest_init()) is good and
> arch_initcall() is bad.

.init.guest_late_init (kvm_guest_init()) is called before 
arch_initcall() and kvm_flush_tlb_others && kvm_send_ipi_mask_allbutself 
rely on __pv_cpu_mask.  So, i can not put this assign in kvm_guest_init().

> Have you checked that with this patch kvm_flush_tlb_others() is still
> being called?

yes. I add a printk and i get the log.

> Actually, there is no need to assign kvm_flush_tlb_others() so late. We
> can always check if __pv_cpu_mask was allocated and revert back to the
> architectural path if not.
I am sorry i don't really understand. Can you explain in more detail? Thx.

> >    	return 0;
> > +
> > +zalloc_cpumask_fail:
> > +	kvm_free_pv_cpu_mask();
> > +	return -ENOMEM;
> >    }
> >    arch_initcall(kvm_alloc_cpumask);
> >
> > --
> > 2.18.4