Message ID | 20250122073456.2950-1-lirongqing@baidu.com (mailing list archive) |
---|---|
State | New |
Headers | show |
Series | KVM: x86: use kvfree_rcu to simplify the code | expand |
On 1/22/2025 1:04 PM, lirongqing wrote: > From: Li RongQing <lirongqing@baidu.com> > > The callback function of call_rcu() just calls kvfree(), so we can > use kvfree_rcu() instead of call_rcu() + callback function. > Reword to imperative style? The callback function of call_rcu() just calls kvfree(). So, use kvfree_rcu() instead of call_rcu() + callback function. > Signed-off-by: Li RongQing <lirongqing@baidu.com> > --- > arch/x86/kvm/lapic.c | 9 +-------- > 1 file changed, 1 insertion(+), 8 deletions(-) > > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c > index 3c83951..d6e62a2 100644 > --- a/arch/x86/kvm/lapic.c > +++ b/arch/x86/kvm/lapic.c > @@ -221,13 +221,6 @@ static inline bool kvm_apic_map_get_logical_dest(struct kvm_apic_map *map, > } > } > > -static void kvm_apic_map_free(struct rcu_head *rcu) > -{ > - struct kvm_apic_map *map = container_of(rcu, struct kvm_apic_map, rcu); > - > - kvfree(map); > -} > - > static int kvm_recalculate_phys_map(struct kvm_apic_map *new, > struct kvm_vcpu *vcpu, > bool *xapic_id_mismatch) > @@ -489,7 +482,7 @@ static void kvm_recalculate_apic_map(struct kvm *kvm) > mutex_unlock(&kvm->arch.apic_map_lock); > > if (old) > - call_rcu(&old->rcu, kvm_apic_map_free); > + kvfree_rcu(old, rcu); > Reviewed-by: Neeraj Upadhyay <Neeraj.Upadhyay@amd.com> Just a note: rcu_barrier() does not wait for kvfree_rcu() callbacks to complete. It only waits for call_rcu() callbacks. For kvfree_rcu() callbacks, kvfree_rcu_barrier() is required to wait for previously enqueued kvfree_rcu() callbacks to complete. As rcu_barrier() is not used for apic_map free callbacks, this looks fine. - Neeraj > kvm_make_scan_ioapic_request(kvm); > }
On Wed, 22 Jan 2025 15:34:56 +0800, lirongqing wrote: > The callback function of call_rcu() just calls kvfree(), so we can > use kvfree_rcu() instead of call_rcu() + callback function. Applied to kvm-x86 misc, with a heavily rewritten changelog to call out the rcu_barrier() wrinkles (I'm pretty sure it's, very technically, a bug fix). Thanks! [1/1] KVM: x86: use kvfree_rcu to simplify the code https://github.com/kvm-x86/linux/commit/82c470121c7b -- https://github.com/kvm-x86/linux/tree/next
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c index 3c83951..d6e62a2 100644 --- a/arch/x86/kvm/lapic.c +++ b/arch/x86/kvm/lapic.c @@ -221,13 +221,6 @@ static inline bool kvm_apic_map_get_logical_dest(struct kvm_apic_map *map, } } -static void kvm_apic_map_free(struct rcu_head *rcu) -{ - struct kvm_apic_map *map = container_of(rcu, struct kvm_apic_map, rcu); - - kvfree(map); -} - static int kvm_recalculate_phys_map(struct kvm_apic_map *new, struct kvm_vcpu *vcpu, bool *xapic_id_mismatch) @@ -489,7 +482,7 @@ static void kvm_recalculate_apic_map(struct kvm *kvm) mutex_unlock(&kvm->arch.apic_map_lock); if (old) - call_rcu(&old->rcu, kvm_apic_map_free); + kvfree_rcu(old, rcu); kvm_make_scan_ioapic_request(kvm); }