Currently we cannot do the isolation of mmu pages, i.e. deleting the current hash_link node by hlist_del(), in this function, because we may call it while traversing the linked list; we cannot solve the problem by hlist_for_each_entry_safe as zapping can happen recursively.
Since the isolation must be done before releasing mmu_lock, we are now forced to call kvm_mmu_isolate_page() for each mmu page found in the invalid_list in kvm_mmu_commit_zap_page(). This patch adds a new parameter to kvm_mmu_prepare_zap_page() as a preparation for solving this issue: all callers just pass NULL now. Note: the abstraction, the introduction of sp_next_pos, makes it possible to support the other list later. Signed-off-by: Takuya Yoshikawa <[email protected]> --- arch/x86/kvm/mmu.c | 41 +++++++++++++++++++++++++++-------------- 1 files changed, 27 insertions(+), 14 deletions(-) diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c index 46b1435..2a48533 100644 --- a/arch/x86/kvm/mmu.c +++ b/arch/x86/kvm/mmu.c @@ -1653,8 +1653,18 @@ static void kvm_unlink_unsync_page(struct kvm *kvm, struct kvm_mmu_page *sp) --kvm->stat.mmu_unsync; } +/* + * Used to hold a pointer to the next mmu page's node when traversing through + * one of the linked lists. This must be updated correctly when deleting any + * entries from the list. + */ +struct sp_next_pos { + struct hlist_node *hn; /* next hash_link node */ +}; + static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp, - struct list_head *invalid_list); + struct list_head *invalid_list, + struct sp_next_pos *npos); static void kvm_mmu_commit_zap_page(struct kvm *kvm, struct list_head *invalid_list); @@ -1672,7 +1682,7 @@ static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, struct list_head *invalid_list, bool clear_unsync) { if (sp->role.cr4_pae != !!is_pae(vcpu)) { - kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list); + kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list, NULL); return 1; } @@ -1680,7 +1690,7 @@ static int __kvm_sync_page(struct kvm_vcpu *vcpu, struct kvm_mmu_page *sp, kvm_unlink_unsync_page(vcpu->kvm, sp); if (vcpu->arch.mmu.sync_page(vcpu, sp)) { - kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list); + kvm_mmu_prepare_zap_page(vcpu->kvm, sp, invalid_list, NULL); return 1; } @@ -1730,7 +1740,8 @@ static void kvm_sync_pages(struct kvm_vcpu *vcpu, gfn_t gfn) kvm_unlink_unsync_page(vcpu->kvm, s); if ((s->role.cr4_pae != !!is_pae(vcpu)) || (vcpu->arch.mmu.sync_page(vcpu, s))) { - kvm_mmu_prepare_zap_page(vcpu->kvm, s, &invalid_list); + kvm_mmu_prepare_zap_page(vcpu->kvm, s, + &invalid_list, NULL); continue; } flush = true; @@ -2062,7 +2073,7 @@ static int mmu_zap_unsync_children(struct kvm *kvm, struct kvm_mmu_page *sp; for_each_sp(pages, sp, parents, i) { - kvm_mmu_prepare_zap_page(kvm, sp, invalid_list); + kvm_mmu_prepare_zap_page(kvm, sp, invalid_list, NULL); mmu_pages_clear_parents(&parents); zapped++; } @@ -2073,7 +2084,8 @@ static int mmu_zap_unsync_children(struct kvm *kvm, } static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp, - struct list_head *invalid_list) + struct list_head *invalid_list, + struct sp_next_pos *npos) { int ret; @@ -2149,7 +2161,7 @@ void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int goal_nr_mmu_pages) page = container_of(kvm->arch.active_mmu_pages.prev, struct kvm_mmu_page, link); - kvm_mmu_prepare_zap_page(kvm, page, &invalid_list); + kvm_mmu_prepare_zap_page(kvm, page, &invalid_list, NULL); } kvm_mmu_commit_zap_page(kvm, &invalid_list); goal_nr_mmu_pages = kvm->arch.n_used_mmu_pages; @@ -2174,7 +2186,7 @@ int kvm_mmu_unprotect_page(struct kvm *kvm, gfn_t gfn) pgprintk("%s: gfn %llx role %x\n", __func__, gfn, sp->role.word); r = 1; - kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list); + kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list, NULL); } kvm_mmu_commit_zap_page(kvm, &invalid_list); spin_unlock(&kvm->mmu_lock); @@ -2894,7 +2906,8 @@ static void mmu_free_roots(struct kvm_vcpu *vcpu) sp = page_header(root); --sp->root_count; if (!sp->root_count && sp->role.invalid) { - kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list); + kvm_mmu_prepare_zap_page(vcpu->kvm, sp, + &invalid_list, NULL); kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list); } vcpu->arch.mmu.root_hpa = INVALID_PAGE; @@ -2910,7 +2923,7 @@ static void mmu_free_roots(struct kvm_vcpu *vcpu) --sp->root_count; if (!sp->root_count && sp->role.invalid) kvm_mmu_prepare_zap_page(vcpu->kvm, sp, - &invalid_list); + &invalid_list, NULL); } vcpu->arch.mmu.pae_root[i] = INVALID_PAGE; } @@ -3987,7 +4000,7 @@ void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa, if (detect_write_misaligned(sp, gpa, bytes) || detect_write_flooding(sp)) { zap_page |= !!kvm_mmu_prepare_zap_page(vcpu->kvm, sp, - &invalid_list); + &invalid_list, NULL); ++vcpu->kvm->stat.mmu_flooded; continue; } @@ -4041,7 +4054,7 @@ void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu) sp = container_of(vcpu->kvm->arch.active_mmu_pages.prev, struct kvm_mmu_page, link); - kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list); + kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list, NULL); ++vcpu->kvm->stat.mmu_recycled; } kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list); @@ -4203,7 +4216,7 @@ void kvm_mmu_zap_all(struct kvm *kvm) spin_lock(&kvm->mmu_lock); restart: list_for_each_entry_safe(sp, node, &kvm->arch.active_mmu_pages, link) - if (kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list)) + if (kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list, NULL)) goto restart; kvm_mmu_commit_zap_page(kvm, &invalid_list); @@ -4220,7 +4233,7 @@ static void kvm_mmu_remove_some_alloc_mmu_pages(struct kvm *kvm, page = container_of(kvm->arch.active_mmu_pages.prev, struct kvm_mmu_page, link); - kvm_mmu_prepare_zap_page(kvm, page, invalid_list); + kvm_mmu_prepare_zap_page(kvm, page, invalid_list, NULL); } static int mmu_shrink(struct shrinker *shrink, struct shrink_control *sc) -- 1.7.5.4 -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
