Re: [PATCH 09/13] KVM: x86/mmu: Allow zap gfn range to operate under the mmu read lock

[Paolo Bonzini <pbonzini@redhat.com>]

On 31/03/21 23:08, Ben Gardon wrote:
> To reduce lock contention and interference with page fault handlers,
> allow the TDP MMU function to zap a GFN range to operate under the MMU
> read lock.
> 
> Signed-off-by: Ben Gardon <bgardon@google.com>
> ---
>   arch/x86/kvm/mmu/mmu.c     |  15 ++++--
>   arch/x86/kvm/mmu/tdp_mmu.c | 102 ++++++++++++++++++++++++++-----------
>   arch/x86/kvm/mmu/tdp_mmu.h |   6 ++-
>   3 files changed, 87 insertions(+), 36 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index 667d64daa82c..dcbfc784cf2f 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -3155,7 +3155,7 @@ static void mmu_free_root_page(struct kvm *kvm, hpa_t *root_hpa,
>   	sp = to_shadow_page(*root_hpa & PT64_BASE_ADDR_MASK);
>   
>   	if (is_tdp_mmu_page(sp))
> -		kvm_tdp_mmu_put_root(kvm, sp);
> +		kvm_tdp_mmu_put_root(kvm, sp, false);
>   	else if (!--sp->root_count && sp->role.invalid)
>   		kvm_mmu_prepare_zap_page(kvm, sp, invalid_list);
>   
> @@ -5514,13 +5514,17 @@ void kvm_zap_gfn_range(struct kvm *kvm, gfn_t gfn_start, gfn_t gfn_end)
>   		}
>   	}
>   
> +	write_unlock(&kvm->mmu_lock);
> +
>   	if (is_tdp_mmu_enabled(kvm)) {
> -		flush = kvm_tdp_mmu_zap_gfn_range(kvm, gfn_start, gfn_end);
> +		read_lock(&kvm->mmu_lock);
> +		flush = kvm_tdp_mmu_zap_gfn_range(kvm, gfn_start, gfn_end,
> +						  true);
>   		if (flush)
>   			kvm_flush_remote_tlbs(kvm);
> -	}
>   
> -	write_unlock(&kvm->mmu_lock);
> +		read_unlock(&kvm->mmu_lock);
> +	}
>   }

This will conflict with Sean's MMU notifier series patches:

KVM: x86/mmu: Pass address space ID to __kvm_tdp_mmu_zap_gfn_range()

What I can do for now is change the mmu.c part of that patch to

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index e6e02360ef67..9882bbd9b742 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -5510,15 +5510,15 @@ void kvm_zap_gfn_range(struct kvm *kvm, gfn_t gfn_start, gfn_t gfn_end)
  		}
  	}
  
-	if (flush)
-		kvm_flush_remote_tlbs_with_address(kvm, gfn_start, gfn_end);
-
  	if (is_tdp_mmu_enabled(kvm)) {
-		flush = kvm_tdp_mmu_zap_gfn_range(kvm, gfn_start, gfn_end);
-		if (flush)
-			kvm_flush_remote_tlbs(kvm);
+		for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)
+			flush = kvm_tdp_mmu_zap_gfn_range(kvm, i, gfn_start,
+							  gfn_end, flush);
  	}
  
+	if (flush)
+		kvm_flush_remote_tlbs_with_address(kvm, gfn_start, gfn_end);
+
  	write_unlock(&kvm->mmu_lock);
  }
  
  
but you will have to add a separate "if (flush)" when moving the write_unlock
earlier, since there's no downgrade function for rwlocks.  In practice it's
not a huge deal since unless running nested there will be only one active MMU.

Paolo

>   static bool slot_rmap_write_protect(struct kvm *kvm,
> @@ -5959,7 +5963,8 @@ static void kvm_recover_nx_lpages(struct kvm *kvm)
>   		WARN_ON_ONCE(!sp->lpage_disallowed);
>   		if (is_tdp_mmu_page(sp)) {
>   			kvm_tdp_mmu_zap_gfn_range(kvm, sp->gfn,
> -				sp->gfn + KVM_PAGES_PER_HPAGE(sp->role.level));
> +				sp->gfn + KVM_PAGES_PER_HPAGE(sp->role.level),
> +				false);
>   		} else {
>   			kvm_mmu_prepare_zap_page(kvm, sp, &invalid_list);
>   			WARN_ON_ONCE(sp->lpage_disallowed);
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index d255125059c4..0e99e4675dd4 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -27,6 +27,15 @@ void kvm_mmu_init_tdp_mmu(struct kvm *kvm)
>   	INIT_LIST_HEAD(&kvm->arch.tdp_mmu_pages);
>   }
>   
> +static __always_inline void kvm_lockdep_assert_mmu_lock_held(struct kvm *kvm,
> +							     bool shared)
> +{
> +	if (shared)
> +		lockdep_assert_held_read(&kvm->mmu_lock);
> +	else
> +		lockdep_assert_held_write(&kvm->mmu_lock);
> +}
> +
>   void kvm_mmu_uninit_tdp_mmu(struct kvm *kvm)
>   {
>   	if (!kvm->arch.tdp_mmu_enabled)
> @@ -42,7 +51,7 @@ void kvm_mmu_uninit_tdp_mmu(struct kvm *kvm)
>   }
>   
>   static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
> -			  gfn_t start, gfn_t end, bool can_yield);
> +			  gfn_t start, gfn_t end, bool can_yield, bool shared);
>   
>   static void tdp_mmu_free_sp(struct kvm_mmu_page *sp)
>   {
> @@ -66,11 +75,12 @@ static void tdp_mmu_free_sp_rcu_callback(struct rcu_head *head)
>   	tdp_mmu_free_sp(sp);
>   }
>   
> -void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root)
> +void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root,
> +			  bool shared)
>   {
>   	gfn_t max_gfn = 1ULL << (shadow_phys_bits - PAGE_SHIFT);
>   
> -	lockdep_assert_held_write(&kvm->mmu_lock);
> +	kvm_lockdep_assert_mmu_lock_held(kvm, shared);
>   
>   	if (!refcount_dec_and_test(&root->tdp_mmu_root_count))
>   		return;
> @@ -81,7 +91,7 @@ void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root)
>   	list_del_rcu(&root->link);
>   	spin_unlock(&kvm->arch.tdp_mmu_pages_lock);
>   
> -	zap_gfn_range(kvm, root, 0, max_gfn, false);
> +	zap_gfn_range(kvm, root, 0, max_gfn, false, shared);
>   
>   	call_rcu(&root->rcu_head, tdp_mmu_free_sp_rcu_callback);
>   }
> @@ -94,11 +104,11 @@ void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root)
>    * function will return NULL.
>    */
>   static struct kvm_mmu_page *tdp_mmu_next_root(struct kvm *kvm,
> -					      struct kvm_mmu_page *prev_root)
> +					      struct kvm_mmu_page *prev_root,
> +					      bool shared)
>   {
>   	struct kvm_mmu_page *next_root;
>   
> -	lockdep_assert_held_write(&kvm->mmu_lock);
>   
>   	rcu_read_lock();
>   
> @@ -117,7 +127,7 @@ static struct kvm_mmu_page *tdp_mmu_next_root(struct kvm *kvm,
>   	rcu_read_unlock();
>   
>   	if (prev_root)
> -		kvm_tdp_mmu_put_root(kvm, prev_root);
> +		kvm_tdp_mmu_put_root(kvm, prev_root, shared);
>   
>   	return next_root;
>   }
> @@ -127,11 +137,15 @@ static struct kvm_mmu_page *tdp_mmu_next_root(struct kvm *kvm,
>    * This makes it safe to release the MMU lock and yield within the loop, but
>    * if exiting the loop early, the caller must drop the reference to the most
>    * recent root. (Unless keeping a live reference is desirable.)
> + *
> + * If shared is set, this function is operating under the MMU lock in read
> + * mode. In the unlikely event that this thread must free a root, the lock
> + * will be temporarily dropped and reacquired in write mode.
>    */
> -#define for_each_tdp_mmu_root_yield_safe(_kvm, _root)	\
> -	for (_root = tdp_mmu_next_root(_kvm, NULL);	\
> -	     _root;					\
> -	     _root = tdp_mmu_next_root(_kvm, _root))
> +#define for_each_tdp_mmu_root_yield_safe(_kvm, _root, _shared)	\
> +	for (_root = tdp_mmu_next_root(_kvm, NULL, _shared);	\
> +	     _root;						\
> +	     _root = tdp_mmu_next_root(_kvm, _root, _shared))
>   
>   /* Only safe under the MMU lock in write mode, without yielding. */
>   #define for_each_tdp_mmu_root(_kvm, _root)				\
> @@ -632,7 +646,8 @@ static inline void tdp_mmu_set_spte_no_dirty_log(struct kvm *kvm,
>    * Return false if a yield was not needed.
>    */
>   static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
> -					     struct tdp_iter *iter, bool flush)
> +					     struct tdp_iter *iter, bool flush,
> +					     bool shared)
>   {
>   	/* Ensure forward progress has been made before yielding. */
>   	if (iter->next_last_level_gfn == iter->yielded_gfn)
> @@ -644,7 +659,11 @@ static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
>   		if (flush)
>   			kvm_flush_remote_tlbs(kvm);
>   
> -		cond_resched_rwlock_write(&kvm->mmu_lock);
> +		if (shared)
> +			cond_resched_rwlock_read(&kvm->mmu_lock);
> +		else
> +			cond_resched_rwlock_write(&kvm->mmu_lock);
> +
>   		rcu_read_lock();
>   
>   		WARN_ON(iter->gfn > iter->next_last_level_gfn);
> @@ -662,23 +681,33 @@ static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
>    * non-root pages mapping GFNs strictly within that range. Returns true if
>    * SPTEs have been cleared and a TLB flush is needed before releasing the
>    * MMU lock.
> + *
>    * If can_yield is true, will release the MMU lock and reschedule if the
>    * scheduler needs the CPU or there is contention on the MMU lock. If this
>    * function cannot yield, it will not release the MMU lock or reschedule and
>    * the caller must ensure it does not supply too large a GFN range, or the
>    * operation can cause a soft lockup.
> + *
> + * If shared is true, this thread holds the MMU lock in read mode and must
> + * account for the possibility that other threads are modifying the paging
> + * structures concurrently. If shared is false, this thread should hold the
> + * MMU lock in write mode.
>    */
>   static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
> -			  gfn_t start, gfn_t end, bool can_yield)
> +			  gfn_t start, gfn_t end, bool can_yield, bool shared)
>   {
>   	struct tdp_iter iter;
>   	bool flush_needed = false;
>   
> +	kvm_lockdep_assert_mmu_lock_held(kvm, shared);
> +
>   	rcu_read_lock();
>   
>   	tdp_root_for_each_pte(iter, root, start, end) {
> +retry:
>   		if (can_yield &&
> -		    tdp_mmu_iter_cond_resched(kvm, &iter, flush_needed)) {
> +		    tdp_mmu_iter_cond_resched(kvm, &iter, flush_needed,
> +					      shared)) {
>   			flush_needed = false;
>   			continue;
>   		}
> @@ -696,8 +725,17 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>   		    !is_last_spte(iter.old_spte, iter.level))
>   			continue;
>   
> -		tdp_mmu_set_spte(kvm, &iter, 0);
> -		flush_needed = true;
> +		if (!shared) {
> +			tdp_mmu_set_spte(kvm, &iter, 0);
> +			flush_needed = true;
> +		} else if (!tdp_mmu_zap_spte_atomic(kvm, &iter)) {
> +			/*
> +			 * The iter must explicitly re-read the SPTE because
> +			 * the atomic cmpxchg failed.
> +			 */
> +			iter.old_spte = READ_ONCE(*rcu_dereference(iter.sptep));
> +			goto retry;
> +		}
>   	}
>   
>   	rcu_read_unlock();
> @@ -709,14 +747,20 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>    * non-root pages mapping GFNs strictly within that range. Returns true if
>    * SPTEs have been cleared and a TLB flush is needed before releasing the
>    * MMU lock.
> + *
> + * If shared is true, this thread holds the MMU lock in read mode and must
> + * account for the possibility that other threads are modifying the paging
> + * structures concurrently. If shared is false, this thread should hold the
> + * MMU in write mode.
>    */
> -bool kvm_tdp_mmu_zap_gfn_range(struct kvm *kvm, gfn_t start, gfn_t end)
> +bool kvm_tdp_mmu_zap_gfn_range(struct kvm *kvm, gfn_t start, gfn_t end,
> +			       bool shared)
>   {
>   	struct kvm_mmu_page *root;
>   	bool flush = false;
>   
> -	for_each_tdp_mmu_root_yield_safe(kvm, root)
> -		flush |= zap_gfn_range(kvm, root, start, end, true);
> +	for_each_tdp_mmu_root_yield_safe(kvm, root, shared)
> +		flush |= zap_gfn_range(kvm, root, start, end, true, shared);
>   
>   	return flush;
>   }
> @@ -726,7 +770,7 @@ void kvm_tdp_mmu_zap_all(struct kvm *kvm)
>   	gfn_t max_gfn = 1ULL << (shadow_phys_bits - PAGE_SHIFT);
>   	bool flush;
>   
> -	flush = kvm_tdp_mmu_zap_gfn_range(kvm, 0, max_gfn);
> +	flush = kvm_tdp_mmu_zap_gfn_range(kvm, 0, max_gfn, false);
>   	if (flush)
>   		kvm_flush_remote_tlbs(kvm);
>   }
> @@ -893,7 +937,7 @@ static __always_inline int kvm_tdp_mmu_handle_hva_range(struct kvm *kvm,
>   	int ret = 0;
>   	int as_id;
>   
> -	for_each_tdp_mmu_root_yield_safe(kvm, root) {
> +	for_each_tdp_mmu_root_yield_safe(kvm, root, false) {
>   		as_id = kvm_mmu_page_as_id(root);
>   		slots = __kvm_memslots(kvm, as_id);
>   		kvm_for_each_memslot(memslot, slots) {
> @@ -933,7 +977,7 @@ static int zap_gfn_range_hva_wrapper(struct kvm *kvm,
>   				     struct kvm_mmu_page *root, gfn_t start,
>   				     gfn_t end, unsigned long unused)
>   {
> -	return zap_gfn_range(kvm, root, start, end, false);
> +	return zap_gfn_range(kvm, root, start, end, false, false);
>   }
>   
>   int kvm_tdp_mmu_zap_hva_range(struct kvm *kvm, unsigned long start,
> @@ -1098,7 +1142,7 @@ static bool wrprot_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>   
>   	for_each_tdp_pte_min_level(iter, root->spt, root->role.level,
>   				   min_level, start, end) {
> -		if (tdp_mmu_iter_cond_resched(kvm, &iter, false))
> +		if (tdp_mmu_iter_cond_resched(kvm, &iter, false, false))
>   			continue;
>   
>   		if (!is_shadow_present_pte(iter.old_spte) ||
> @@ -1128,7 +1172,7 @@ bool kvm_tdp_mmu_wrprot_slot(struct kvm *kvm, struct kvm_memory_slot *slot,
>   	int root_as_id;
>   	bool spte_set = false;
>   
> -	for_each_tdp_mmu_root_yield_safe(kvm, root) {
> +	for_each_tdp_mmu_root_yield_safe(kvm, root, false) {
>   		root_as_id = kvm_mmu_page_as_id(root);
>   		if (root_as_id != slot->as_id)
>   			continue;
> @@ -1157,7 +1201,7 @@ static bool clear_dirty_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>   	rcu_read_lock();
>   
>   	tdp_root_for_each_leaf_pte(iter, root, start, end) {
> -		if (tdp_mmu_iter_cond_resched(kvm, &iter, false))
> +		if (tdp_mmu_iter_cond_resched(kvm, &iter, false, false))
>   			continue;
>   
>   		if (spte_ad_need_write_protect(iter.old_spte)) {
> @@ -1193,7 +1237,7 @@ bool kvm_tdp_mmu_clear_dirty_slot(struct kvm *kvm, struct kvm_memory_slot *slot)
>   	int root_as_id;
>   	bool spte_set = false;
>   
> -	for_each_tdp_mmu_root_yield_safe(kvm, root) {
> +	for_each_tdp_mmu_root_yield_safe(kvm, root, false) {
>   		root_as_id = kvm_mmu_page_as_id(root);
>   		if (root_as_id != slot->as_id)
>   			continue;
> @@ -1291,7 +1335,7 @@ static void zap_collapsible_spte_range(struct kvm *kvm,
>   	rcu_read_lock();
>   
>   	tdp_root_for_each_pte(iter, root, start, end) {
> -		if (tdp_mmu_iter_cond_resched(kvm, &iter, spte_set)) {
> +		if (tdp_mmu_iter_cond_resched(kvm, &iter, spte_set, false)) {
>   			spte_set = false;
>   			continue;
>   		}
> @@ -1326,7 +1370,7 @@ void kvm_tdp_mmu_zap_collapsible_sptes(struct kvm *kvm,
>   	struct kvm_mmu_page *root;
>   	int root_as_id;
>   
> -	for_each_tdp_mmu_root_yield_safe(kvm, root) {
> +	for_each_tdp_mmu_root_yield_safe(kvm, root, false) {
>   		root_as_id = kvm_mmu_page_as_id(root);
>   		if (root_as_id != slot->as_id)
>   			continue;
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.h b/arch/x86/kvm/mmu/tdp_mmu.h
> index 9961df505067..855e58856815 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.h
> +++ b/arch/x86/kvm/mmu/tdp_mmu.h
> @@ -13,9 +13,11 @@ __must_check static inline bool kvm_tdp_mmu_get_root(struct kvm *kvm,
>   	return refcount_inc_not_zero(&root->tdp_mmu_root_count);
>   }
>   
> -void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root);
> +void kvm_tdp_mmu_put_root(struct kvm *kvm, struct kvm_mmu_page *root,
> +			  bool shared);
>   
> -bool kvm_tdp_mmu_zap_gfn_range(struct kvm *kvm, gfn_t start, gfn_t end);
> +bool kvm_tdp_mmu_zap_gfn_range(struct kvm *kvm, gfn_t start, gfn_t end,
> +			       bool shared);
>   void kvm_tdp_mmu_zap_all(struct kvm *kvm);
>   
>   int kvm_tdp_mmu_map(struct kvm_vcpu *vcpu, gpa_t gpa, u32 error_code,
>