Skip to contentSkip to conversation

Fix the allocation of pv cpu mask

PATCH In review [email protected] 2 participants 6 messages Sep 14, 2020 – Sep 23, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover Fix the allocation of pv cpu mask

From: Haiwei Li <[email protected]>

Hi,

There is a old version patch of 'KVM: Check the allocation of pv cpu mask' in upstream. The v2 and what is discussed is in url:

https://lore.kernel.org/kvm/[email protected]/

In this patch, i fix the build error and make changes as suggested.

Haiwei Li (2):
KVM: Fix the build error
KVM: Check if __pv_cpu_mask was allocated instead of assigning ops too
late

 arch/x86/kernel/kvm.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)
Signature · 1 line
2.18.4
#1 lore raw

PATCH 1/2 KVM: Fix the build error

From: Haiwei Li <[email protected]>

When CONFIG_SMP is not set, an build error occurs with message "error:
use of undeclared identifier 'kvm_send_ipi_mask_allbutself'"

Fixes 0f990222108d ("KVM: Check the allocation of pv cpu mask", 2020-09-01)
---
 arch/x86/kernel/kvm.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kernel/kvm.c +2 -0
index 1b51b727b140..7e8be0421720 100644
@@ -797,7 +797,9 @@ static __init int kvm_alloc_cpumask(void)
 			}
 		}
 
+#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;
 	return 0;
 
Signature · 1 line
2.18.4
maintainer
#2 lore raw

Re: [PATCH 1/2] KVM: Fix the build error

Quoted 18 lines · On 14/09/20 11:11, [email protected] wrote:

From: Haiwei Li <[email protected]>

When CONFIG_SMP is not set, an build error occurs with message "error:
use of undeclared identifier 'kvm_send_ipi_mask_allbutself'"

Fixes 0f990222108d ("KVM: Check the allocation of pv cpu mask", 2020-09-01)
---
 arch/x86/kernel/kvm.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kernel/kvm.c +0 -0
index 1b51b727b140..7e8be0421720 100644

arch/x86/kernel/kvm.c static __init int kvm_alloc_cpumask(void) · line 805

Quoted 9 lines
arch/x86/kernel/kvm.c +2 -0
@@ -797,7 +797,9 @@ static __init int kvm_alloc_cpumask(void)
 			}
 		}
 
+#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;
 	return 0;
 

If CONFIG_SMP is not set you don't need kvm_alloc_cpumask or pv_ops.mmu.flush_tlb_others at all. Can you squash these two into the original patch and re-submit for 5.10?

Paolo

author
#3 lore raw

Re: [PATCH 1/2] KVM: Fix the build error

Quoted 21 lines · Paolo Bonzini
Quoted 20 lines · On 14/09/20 11:11, [email protected] wrote:

From: Haiwei Li <[email protected]>

When CONFIG_SMP is not set, an build error occurs with message "error:
use of undeclared identifier 'kvm_send_ipi_mask_allbutself'"

Fixes 0f990222108d ("KVM: Check the allocation of pv cpu mask", 2020-09-01)
---
  arch/x86/kernel/kvm.c | 2 ++
  1 file changed, 2 insertions(+)
arch/x86/kernel/kvm.c +0 -0
index 1b51b727b140..7e8be0421720 100644
@@ -797,7 +797,9 @@ static __init int kvm_alloc_cpumask(void)
  			}
  		}
 
Quoted 11 lines

arch/x86/kernel/kvm.c static __init int kvm_alloc_cpumask(void) · line 805

Quoted 7 lines
arch/x86/kernel/kvm.c +2 -0
+#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;
  	return 0;
 

If CONFIG_SMP is not set you don't need kvm_alloc_cpumask or pv_ops.mmu.flush_tlb_others at all. Can you squash these two into the original patch and re-submit for 5.10?

I will, thanks.

Haiwei Li

author
#4 lore raw

Re: [PATCH 1/2] KVM: Fix the build error

In reply to Paolo Bonzini
Quoted 21 lines · Paolo Bonzini
Quoted 20 lines · On 14/09/20 11:11, [email protected] wrote:

From: Haiwei Li <[email protected]>

When CONFIG_SMP is not set, an build error occurs with message "error:
use of undeclared identifier 'kvm_send_ipi_mask_allbutself'"

Fixes 0f990222108d ("KVM: Check the allocation of pv cpu mask", 2020-09-01)
---
  arch/x86/kernel/kvm.c | 2 ++
  1 file changed, 2 insertions(+)
arch/x86/kernel/kvm.c +0 -0
index 1b51b727b140..7e8be0421720 100644
@@ -797,7 +797,9 @@ static __init int kvm_alloc_cpumask(void)
  			}
  		}
 
Quoted 11 lines

arch/x86/kernel/kvm.c static __init int kvm_alloc_cpumask(void) · line 805

Quoted 7 lines
arch/x86/kernel/kvm.c +2 -0
+#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;
  	return 0;
 

If CONFIG_SMP is not set you don't need kvm_alloc_cpumask or pv_ops.mmu.flush_tlb_others at all. Can you squash these two into the original patch and re-submit for 5.10?

Hi, Paolo

I'm a little confused. Function kvm_flush_tlb_others doesn't seem to be related to CONFIG_SMP.

And my patch like:

---
 arch/x86/kernel/kvm.c | 27 ++++++++++++++++++++++++---
 1 file changed, 24 insertions(+), 3 deletions(-)
arch/x86/kernel/kvm.c +24 -3
index 9663ba31347c..1e5da6db519c 100644
@@ -553,7 +553,6 @@ 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");
 }
 
@@ -619,6 +618,11 @@ static void kvm_flush_tlb_others(const struct cpumask *cpumask,
 	struct kvm_steal_time *src;
 	struct cpumask *flushmask = this_cpu_cpumask_var_ptr(__pv_cpu_mask);
 
+	if (unlikely(!flushmask)) {
+		native_flush_tlb_others(cpumask, info);
+		return;
+	}
+
 	cpumask_copy(flushmask, cpumask);
 	/*
 	 * We have to call flush only on online vCPUs. And
@@ -765,6 +769,14 @@ static __init int activate_jump_labels(void)
 }
 arch_initcall(activate_jump_labels);
 
+static void kvm_free_cpumask(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;
@@ -783,11 +795,20 @@ 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
 	return 0;
+
+zalloc_cpumask_fail:
+	kvm_free_cpumask();
+	return -ENOMEM;
 }
 arch_initcall(kvm_alloc_cpumask);
 

--
2.18.4

Do you have any suggestion? Thanks.

Haiwei

#5 lore raw

PATCH 2/2 KVM: Check if __pv_cpu_mask was allocated instead of assigning ops too late

In reply to [email protected]

From: Haiwei Li <[email protected]>

Without this patch, kvm_flush_tlb_others is not called. We can always check if __pv_cpu_mask was allocated and revert back to the architectural path if not.

---
 arch/x86/kernel/kvm.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
arch/x86/kernel/kvm.c +6 -2
index 7e8be0421720..3c6c516d9828 100644
@@ -553,7 +553,6 @@ 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");
 }
 
@@ -619,6 +618,11 @@ static void kvm_flush_tlb_others(const struct cpumask *cpumask,
 	struct kvm_steal_time *src;
 	struct cpumask *flushmask = this_cpu_cpumask_var_ptr(__pv_cpu_mask);
 
+	if (unlikely(!flushmask)) {
+		native_flush_tlb_others(cpumask, info);
+		return;
+	}
+
 	cpumask_copy(flushmask, cpumask);
 	/*
 	 * We have to call flush only on online vCPUs. And
@@ -652,6 +656,7 @@ 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");
 	}
@@ -800,7 +805,6 @@ static __init int kvm_alloc_cpumask(void)
 #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;
 	return 0;
 
 zalloc_cpumask_fail:
Signature · 1 line
2.18.4