Date:   Thu, 11 Nov 2021 18:50:24 +0000
From:   David Matlack <dmatlack@google.com>
To:     Ben Gardon <bgardon@google.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Peter Xu <peterx@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        Peter Shier <pshier@google.com>,
        Mingwei Zhang <mizhang@google.com>,
        Yulei Zhang <yulei.kernel@gmail.com>,
        Wanpeng Li <kernellwp@gmail.com>,
        Xiao Guangrong <xiaoguangrong.eric@gmail.com>,
        Kai Huang <kai.huang@intel.com>,
        Keqian Zhu <zhukeqian1@huawei.com>,
        David Hildenbrand <david@redhat.com>
Subject: Re: [RFC 04/19] KVM: x86/mmu: Yield while processing disconnected_sps
Message-ID: <YY1l8EMsQyTDAAkT@google.com>
References: <20211110223010.1392399-1-bgardon@google.com>
 <20211110223010.1392399-5-bgardon@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20211110223010.1392399-5-bgardon@google.com>
Precedence: bulk

On Wed, Nov 10, 2021 at 02:29:55PM -0800, Ben Gardon wrote:
> When preparing to free disconnected SPs, the list can accumulate many
> entries; enough that it is likely necessary to yeild while queuing RCU
> callbacks to free the SPs.
> 
> Signed-off-by: Ben Gardon <bgardon@google.com>
> ---
>  arch/x86/kvm/mmu/tdp_mmu.c | 18 +++++++++++++++---
>  1 file changed, 15 insertions(+), 3 deletions(-)
> 
> diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> index a448f0f2d993..c2a9f7acf8ef 100644
> --- a/arch/x86/kvm/mmu/tdp_mmu.c
> +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> @@ -513,7 +513,8 @@ static void handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
>   * being removed from the paging structure and this function being called.
>   */
>  static void handle_disconnected_sps(struct kvm *kvm,
> -				    struct list_head *disconnected_sps)
> +				    struct list_head *disconnected_sps,
> +				    bool can_yield, bool shared)
>  {
>  	struct kvm_mmu_page *sp;
>  	struct kvm_mmu_page *next;
> @@ -521,6 +522,16 @@ static void handle_disconnected_sps(struct kvm *kvm,
>  	list_for_each_entry_safe(sp, next, disconnected_sps, link) {
>  		list_del(&sp->link);
>  		call_rcu(&sp->rcu_head, tdp_mmu_free_sp_rcu_callback);
> +
> +		if (can_yield &&
> +		    (need_resched() || rwlock_needbreak(&kvm->mmu_lock))) {
> +			rcu_read_unlock();
> +			if (shared)
> +				cond_resched_rwlock_read(&kvm->mmu_lock);
> +			else
> +				cond_resched_rwlock_write(&kvm->mmu_lock);
> +			rcu_read_lock();
> +		}

What about something like this to cut down on the duplicate code?

diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
index c2a9f7acf8ef..2fd010f2421e 100644
--- a/arch/x86/kvm/mmu/tdp_mmu.c
+++ b/arch/x86/kvm/mmu/tdp_mmu.c
@@ -508,6 +508,26 @@ static void handle_changed_spte(struct kvm *kvm, int as_id, gfn_t gfn,
                                      new_spte, level);
 }

+static inline bool tdp_mmu_need_resched(struct kvm *kvm)
+{
+       return need_resched() || rwlock_needbreak(&kvm->mmu_lock);
+}
+
+static void tdp_mmu_cond_resched(struct kvm *kvm, bool shared, bool flush)
+{
+       rcu_read_unlock()
+
+       if (flush)
+               kvm_flush_remote_tlbs(kvm);
+
+       if (shared)
+               cond_resched_rwlock_read(&kvm->mmu_lock);
+       else
+               cond_resched_rwlock_write(&kvm->mmu_lock);
+
+       rcu_read_lock();
+}
+
 /*
  * The TLBs must be flushed between the pages linked from disconnected_sps
  * being removed from the paging structure and this function being called.
@@ -523,15 +543,8 @@ static void handle_disconnected_sps(struct kvm *kvm,
                list_del(&sp->link);
                call_rcu(&sp->rcu_head, tdp_mmu_free_sp_rcu_callback);

-               if (can_yield &&
-                   (need_resched() || rwlock_needbreak(&kvm->mmu_lock))) {
-                       rcu_read_unlock();
-                       if (shared)
-                               cond_resched_rwlock_read(&kvm->mmu_lock);
-                       else
-                               cond_resched_rwlock_write(&kvm->mmu_lock);
-                       rcu_read_lock();
-               }
+               if (can_yield && tdp_mmu_need_resched(kvm))
+                       tdp_mmu_cond_resched(kvm, shared, false);
        }
 }

@@ -724,18 +737,8 @@ static inline bool tdp_mmu_iter_cond_resched(struct kvm *kvm,
        if (iter->next_last_level_gfn == iter->yielded_gfn)
                return false;

-       if (need_resched() || rwlock_needbreak(&kvm->mmu_lock)) {
-               rcu_read_unlock();
-
-               if (flush)
-                       kvm_flush_remote_tlbs(kvm);
-
-               if (shared)
-                       cond_resched_rwlock_read(&kvm->mmu_lock);
-               else
-                       cond_resched_rwlock_write(&kvm->mmu_lock);
-
-               rcu_read_lock();
+       if (tdp_mmu_need_resched(kvm)) {
+               tdp_mmu_cond_resched(kvm, shared, flush);

                WARN_ON(iter->gfn > iter->next_last_level_gfn);

>  	}
>  }
>  
> @@ -599,7 +610,7 @@ static inline bool tdp_mmu_zap_spte_atomic(struct kvm *kvm,
>  	 */
>  	WRITE_ONCE(*rcu_dereference(iter->sptep), 0);
>  
> -	handle_disconnected_sps(kvm, &disconnected_sps);
> +	handle_disconnected_sps(kvm, &disconnected_sps, false, true);
>  
>  	return true;
>  }
> @@ -817,7 +828,8 @@ static bool zap_gfn_range(struct kvm *kvm, struct kvm_mmu_page *root,
>  
>  	if (!list_empty(&disconnected_sps)) {
>  		kvm_flush_remote_tlbs(kvm);
> -		handle_disconnected_sps(kvm, &disconnected_sps);
> +		handle_disconnected_sps(kvm, &disconnected_sps,
> +					can_yield, shared);
>  		flush = false;
>  	}
>  
> -- 
> 2.34.0.rc0.344.g81b53c2807-goog
>