[PATCH 2/3] KVM: PPC: BOOK3S: HV: Use unlock variant with memory barrier

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

We switch to unlock variant with memory barriers in the error path
and also in code path where we had implicit dependency on previous
functions calling lwsync/ptesync. In most of the cases we don't really
need an explicit barrier, but using the variant make sure we don't make
mistakes later with code movements. We also document why a
non-barrier variant is ok in performance critical path.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/kvm/book3s_64_mmu_hv.c | 10 +++++-----
 arch/powerpc/kvm/book3s_hv_rm_mmu.c | 15 ++++++++++-----
 2 files changed, 15 insertions(+), 10 deletions(-)

diff --git a/arch/powerpc/kvm/book3s_64_mmu_hv.c b/arch/powerpc/kvm/book3s_64_mmu_hv.c
index 5ea4b2b6a157..c97690ffb5f6 100644
--- a/arch/powerpc/kvm/book3s_64_mmu_hv.c
+++ b/arch/powerpc/kvm/book3s_64_mmu_hv.c
@@ -774,7 +774,7 @@ int kvmppc_book3s_hv_page_fault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 	return ret;
 
  out_unlock:
-	__unlock_hpte(hptep, be64_to_cpu(hptep[0]));
+	unlock_hpte(hptep, be64_to_cpu(hptep[0]));
 	preempt_enable();
 	goto out_put;
 }
@@ -903,8 +903,8 @@ static int kvm_unmap_rmapp(struct kvm *kvm, unsigned long *rmapp,
 				note_hpte_modification(kvm, &rev[i]);
 			}
 		}
+		unlock_hpte(hptep, be64_to_cpu(hptep[0]));
 		unlock_rmap(rmapp);
-		__unlock_hpte(hptep, be64_to_cpu(hptep[0]));
 	}
 	return 0;
 }
@@ -992,7 +992,7 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
 			}
 			ret = 1;
 		}
-		__unlock_hpte(hptep, be64_to_cpu(hptep[0]));
+		unlock_hpte(hptep, be64_to_cpu(hptep[0]));
 	} while ((i = j) != head);
 
 	unlock_rmap(rmapp);
@@ -1115,7 +1115,7 @@ static int kvm_test_clear_dirty_npages(struct kvm *kvm, unsigned long *rmapp)
 
 		/* Now check and modify the HPTE */
 		if (!(hptep[0] & cpu_to_be64(HPTE_V_VALID))) {
-			__unlock_hpte(hptep, be64_to_cpu(hptep[0]));
+			unlock_hpte(hptep, be64_to_cpu(hptep[0]));
 			continue;
 		}
 		/* need to make it temporarily absent so C is stable */
@@ -1137,7 +1137,7 @@ static int kvm_test_clear_dirty_npages(struct kvm *kvm, unsigned long *rmapp)
 		}
 		v &= ~HPTE_V_ABSENT;
 		v |= HPTE_V_VALID;
-		__unlock_hpte(hptep, v);
+		unlock_hpte(hptep, v);
 	} while ((i = j) != head);
 
 	unlock_rmap(rmapp);
diff --git a/arch/powerpc/kvm/book3s_hv_rm_mmu.c b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
index 769a5d4c0430..78e689b066f1 100644
--- a/arch/powerpc/kvm/book3s_hv_rm_mmu.c
+++ b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
@@ -292,6 +292,9 @@ long kvmppc_do_h_enter(struct kvm *kvm, unsigned long flags,
 				pte = be64_to_cpu(hpte[0]);
 				if (!(pte & (HPTE_V_VALID | HPTE_V_ABSENT)))
 					break;
+				/*
+				 * Data dependency will avoid re-ordering
+				 */
 				__unlock_hpte(hpte, pte);
 				hpte += 2;
 			}
@@ -310,7 +313,7 @@ long kvmppc_do_h_enter(struct kvm *kvm, unsigned long flags,
 				cpu_relax();
 			pte = be64_to_cpu(hpte[0]);
 			if (pte & (HPTE_V_VALID | HPTE_V_ABSENT)) {
-				__unlock_hpte(hpte, pte);
+				unlock_hpte(hpte, pte);
 				return H_PTEG_FULL;
 			}
 		}
@@ -481,7 +484,7 @@ long kvmppc_do_h_remove(struct kvm *kvm, unsigned long flags,
 	if ((pte & (HPTE_V_ABSENT | HPTE_V_VALID)) == 0 ||
 	    ((flags & H_AVPN) && (pte & ~0x7fUL) != avpn) ||
 	    ((flags & H_ANDCOND) && (pte & avpn) != 0)) {
-		__unlock_hpte(hpte, pte);
+		unlock_hpte(hpte, pte);
 		return H_NOT_FOUND;
 	}
 
@@ -617,7 +620,7 @@ long kvmppc_h_bulk_remove(struct kvm_vcpu *vcpu)
 				be64_to_cpu(hp[0]), be64_to_cpu(hp[1]));
 			rcbits = rev->guest_rpte & (HPTE_R_R|HPTE_R_C);
 			args[j] |= rcbits << (56 - 5);
-			__unlock_hpte(hp, 0);
+			unlock_hpte(hp, 0);
 		}
 	}
 
@@ -643,7 +646,7 @@ long kvmppc_h_protect(struct kvm_vcpu *vcpu, unsigned long flags,
 	pte = be64_to_cpu(hpte[0]);
 	if ((pte & (HPTE_V_ABSENT | HPTE_V_VALID)) == 0 ||
 	    ((flags & H_AVPN) && (pte & ~0x7fUL) != avpn)) {
-		__unlock_hpte(hpte, pte);
+		unlock_hpte(hpte, pte);
 		return H_NOT_FOUND;
 	}
 
@@ -834,7 +837,9 @@ long kvmppc_hv_find_lock_hpte(struct kvm *kvm, gva_t eaddr, unsigned long slb_v,
 			    hpte_base_page_size(v, r) == (1ul << pshift))
 				/* Return with the HPTE still locked */
 				return (hash << 3) + (i >> 1);
-
+			/*
+			 * Data dependency should avoid re-ordering
+			 */
 			__unlock_hpte(&hpte[i], v);
 		}
 
-- 
1.9.1