Re: [PATCH v3 12/15] KVM: MMU: fast invalid all shadow pages

[Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>]

On 04/18/2013 08:05 AM, Marcelo Tosatti wrote:
> On Tue, Apr 16, 2013 at 02:32:50PM +0800, Xiao Guangrong wrote:
>> The current kvm_mmu_zap_all is really slow - it is holding mmu-lock to
>> walk and zap all shadow pages one by one, also it need to zap all guest
>> page's rmap and all shadow page's parent spte list. Particularly, things
>> become worse if guest uses more memory or vcpus. It is not good for
>> scalability.
>>
>> In this patch, we introduce a faster way to invalid all shadow pages.
>> KVM maintains a global mmu invalid generation-number which is stored in
>> kvm->arch.mmu_valid_gen and every shadow page stores the current global
>> generation-number into sp->mmu_valid_gen when it is created.
>>
>> When KVM need zap all shadow pages sptes, it just simply increase the
>> global generation-number then reload root shadow pages on all vcpus.
>> Vcpu will create a new shadow page table according to current kvm's
>> generation-number. It ensures the old pages are not used any more.
>>
>> The invalid-gen pages (sp->mmu_valid_gen != kvm->arch.mmu_valid_gen)
>> are keeped in mmu-cache until page allocator reclaims page.
>>
>> If the invalidation is due to memslot changed, its rmap amd lpage-info
>> will be freed soon, in order to avoiding use invalid memory, we unmap
>> all sptes on its rmap and always reset the large-info all memslots so
>> that rmap and lpage info can be safely freed.
>>
>> Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
>> ---
>>  arch/x86/include/asm/kvm_host.h |    2 +
>>  arch/x86/kvm/mmu.c              |   85 +++++++++++++++++++++++++++++++++++++-
>>  arch/x86/kvm/mmu.h              |    4 ++
>>  arch/x86/kvm/x86.c              |    6 +++
>>  4 files changed, 94 insertions(+), 3 deletions(-)
>>
>> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
>> index 1ad9a34..6f8ee18 100644
>> --- a/arch/x86/include/asm/kvm_host.h
>> +++ b/arch/x86/include/asm/kvm_host.h
>> @@ -223,6 +223,7 @@ struct kvm_mmu_page {
>>  	int root_count;          /* Currently serving as active root */
>>  	unsigned int unsync_children;
>>  	unsigned long parent_ptes;	/* Reverse mapping for parent_pte */
>> +	unsigned long mmu_valid_gen;
>>  	DECLARE_BITMAP(unsync_child_bitmap, 512);
>>  
>>  #ifdef CONFIG_X86_32
>> @@ -531,6 +532,7 @@ struct kvm_arch {
>>  	unsigned int n_requested_mmu_pages;
>>  	unsigned int n_max_mmu_pages;
>>  	unsigned int indirect_shadow_pages;
>> +	unsigned long mmu_valid_gen;
>>  	struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
>>  	/*
>>  	 * Hash table of struct kvm_mmu_page.
>> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
>> index 9ac584f..12129b7 100644
>> --- a/arch/x86/kvm/mmu.c
>> +++ b/arch/x86/kvm/mmu.c
>> @@ -1732,6 +1732,11 @@ static struct rmap_operations invalid_rmap_ops = {
>>  	.rmap_unmap = kvm_unmap_invalid_rmapp
>>  };
>>  
>> +static void init_invalid_memslot_rmap_ops(struct kvm_memory_slot *slot)
>> +{
>> +	slot->arch.ops = &invalid_rmap_ops;
>> +}
>> +
>>  typedef void (*handle_rmap_fun)(unsigned long *rmapp, void *data);
>>  static void walk_memslot_rmap_nolock(struct kvm_memory_slot *slot,
>>  				     handle_rmap_fun fun, void *data)
>> @@ -1812,6 +1817,65 @@ void free_meslot_rmap_desc_nolock(struct kvm_memory_slot *slot)
>>  	walk_memslot_rmap_nolock(slot, free_rmap_desc_nolock, NULL);
>>  }
>>  
>> +/*
>> + * Fast invalid all shadow pages belong to @slot.
>> + *
>> + * @slot != NULL means the invalidation is caused the memslot specified
>> + * by @slot is being deleted, in this case, we should ensure that rmap
>> + * and lpage-info of the @slot can not be used after calling the function.
>> + * Specially, if @slot is INVALID_ALL_SLOTS, all slots will be deleted
>> + * soon, it always happens when kvm is being destroyed.
>> + *
>> + * @slot == NULL means the invalidation due to other reasons, we need
>> + * not care rmap and lpage-info since they are still valid after calling
>> + * the function.
>> + */
>> +void kvm_mmu_invalid_memslot_pages(struct kvm *kvm,
>> +				   struct kvm_memory_slot *slot)
>> +{
>> +	struct kvm_memory_slot *each_slot;
>> +
>> +	spin_lock(&kvm->mmu_lock);
>> +	kvm->arch.mmu_valid_gen++;
>> +
>> +	if (slot == INVALID_ALL_SLOTS)
>> +		kvm_for_each_memslot(each_slot, kvm_memslots(kvm))
>> +			init_invalid_memslot_rmap_ops(each_slot);
>> +	else if (slot)
>> +		init_invalid_memslot_rmap_ops(slot);
>> +
>> +	/*
>> +	 * All shadow paes are invalid, reset the large page info,
>> +	 * then we can safely desotry the memslot, it is also good
>> +	 * for large page used.
>> +	 */
>> +	kvm_clear_all_lpage_info(kvm);
>> +
>> +	/*
>> +	 * Notify all vcpus to reload its shadow page table
>> +	 * and flush TLB. Then all vcpus will switch to new
>> +	 * shadow page table with the new mmu_valid_gen.
>> +	 *
>> +	 * Note: we should do this under the protection of
>> +	 * mmu-lock, otherwise, vcpu would purge shadow page
>> +	 * but miss tlb flush.
>> +	 */
>> +	kvm_reload_remote_mmus(kvm);
>> +	spin_unlock(&kvm->mmu_lock);
>> +
>> +	if (slot == INVALID_ALL_SLOTS)
>> +		kvm_for_each_memslot(each_slot, kvm_memslots(kvm))
>> +			unmap_memslot_rmap_nolock(kvm, each_slot);
>> +	else if (slot)
>> +		unmap_memslot_rmap_nolock(kvm, slot);
> 
> What is the justification for this? 

We want the rmap of being deleted memslot is removed-only that is
needed for unmapping rmap out of mmu-lock.

======
1) do not corrupt the rmap
2) keep pte-list-descs available
3) keep shadow page available

Resolve 1):
we make the invalid rmap be remove-only that means we only delete and
clear spte from the rmap, no new sptes can be added to it.
This is reasonable since kvm can not do address translation on invalid rmap
(gfn_to_pfn is failed on invalid memslot) and all sptes on invalid rmap can
not be reused (they belong to invalid shadow page).
======

clear_flush_young / test_young / change_pte of mmu-notify can rewrite
rmap with the present-spte (P bit is set), we should umap rmap in
these handlers.

> 
>> +
>> +	/*
>> +	 * To ensure that all vcpus and mmu-notify are not clearing
>> +	 * spte and rmap entry.
>> +	 */
>> +	synchronize_srcu_expedited(&kvm->srcu);
>> +}
>> +
>>  #ifdef MMU_DEBUG
>>  static int is_empty_shadow_page(u64 *spt)
>>  {
>> @@ -2219,6 +2283,11 @@ static void clear_sp_write_flooding_count(u64 *spte)
>>  	__clear_sp_write_flooding_count(sp);
>>  }
>>  
>> +static bool is_valid_sp(struct kvm *kvm, struct kvm_mmu_page *sp)
>> +{
>> +	return likely(sp->mmu_valid_gen == kvm->arch.mmu_valid_gen);
>> +}
>> +
>>  static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>>  					     gfn_t gfn,
>>  					     gva_t gaddr,
>> @@ -2245,6 +2314,9 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>>  		role.quadrant = quadrant;
>>  	}
>>  	for_each_gfn_sp(vcpu->kvm, sp, gfn) {
>> +		if (!is_valid_sp(vcpu->kvm, sp))
>> +			continue;
>> +
>>  		if (!need_sync && sp->unsync)
>>  			need_sync = true;
>>  
>> @@ -2281,6 +2353,7 @@ static struct kvm_mmu_page *kvm_mmu_get_page(struct kvm_vcpu *vcpu,
>>  
>>  		account_shadowed(vcpu->kvm, gfn);
>>  	}
>> +	sp->mmu_valid_gen = vcpu->kvm->arch.mmu_valid_gen;
>>  	init_shadow_page_table(sp);
>>  	trace_kvm_mmu_get_page(sp, true);
>>  	return sp;
>> @@ -2451,8 +2524,12 @@ static int kvm_mmu_prepare_zap_page(struct kvm *kvm, struct kvm_mmu_page *sp,
>>  	ret = mmu_zap_unsync_children(kvm, sp, invalid_list);
>>  	kvm_mmu_page_unlink_children(kvm, sp);
> 
> The rmaps[] arrays linked to !is_valid_sp() shadow pages should not be
> accessed (as they have been freed already).
> 
> I suppose the is_valid_sp() conditional below should be moved earlier,
> before kvm_mmu_unlink_parents or any other rmap access.
> 
> This is fine: the !is_valid_sp() shadow pages are only reachable
> by SLAB and the hypervisor itself.

Unfortunately we can not do this. :(

The sptes in shadow pape can linked to many slots, if the spte is linked
to the rmap of being deleted memslot, it is ok, otherwise, the rmap of
still used memslot is miss updated.

For example, slot 0 is being deleted, sp->spte[0] is linked on slot[0].rmap,
sp->spte[1] is linked on slot[1].rmap. If we do not access rmap of this 'sp',
the already-freed spte[1] is still linked on slot[1].rmap.

We can let kvm update the rmap for sp->spte[1] and do not unlink sp->spte[0].
This is also not allowed since mmu-notify can access the invalid rmap before
the memslot is destroyed, then mmu-notify will get already-freed spte on
the rmap or page Access/Dirty is miss tracked (if let mmu-notify do not access
the invalid rmap).

> 
>>  	kvm_mmu_unlink_parents(kvm, sp);
>> -	if (!sp->role.invalid && !sp->role.direct)
>> +
>> +	if (!sp->role.invalid && !sp->role.direct &&
>> +	      /* Invalid-gen pages are not accounted. */
>> +	      is_valid_sp(kvm, sp))
>>  		unaccount_shadowed(kvm, sp->gfn);
>> +
>>  	if (sp->unsync)
>>  		kvm_unlink_unsync_page(kvm, sp);
>>  	if (!sp->root_count) {
>> @@ -3256,7 +3333,8 @@ static void mmu_free_roots(struct kvm_vcpu *vcpu)
>>  
>>  		sp = page_header(root);
>>  		--sp->root_count;
>> -		if (!sp->root_count && sp->role.invalid) {
>> +		if (!is_valid_sp(vcpu->kvm, sp) ||
>> +		      (!sp->root_count && sp->role.invalid)) {
>>  			kvm_mmu_prepare_zap_page(vcpu->kvm, sp, &invalid_list);
>>  			kvm_mmu_commit_zap_page(vcpu->kvm, &invalid_list);
> 
> Why can't kvm_mmu_prepare_zap_page+kvm_mmu_commit_zap_page handle
> invalid shadow pages?

Oh, this is unnecessary. Anyway vcpu can reset its root when it is handling
KVM_REQ_MMU_RELOAD

> 
> Perhaps another name is useful to avoid confusion with role.invalid.
> Maybe 'defunct'.

Yes, this is better.