Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <2f80a561-a338-cdb3-e641-d8803b859ce0@redhat.com>
Date:   Wed, 28 Sep 2022 19:29:22 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.2.1
Content-Language: en-US
To:     Emanuele Giuseppe Esposito <eesposit@redhat.com>,
        kvm@vger.kernel.org
Cc:     Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>,
        David Hildenbrand <david@redhat.com>,
        Maxim Levitsky <mlevitsk@redhat.com>, x86@kernel.org,
        "H. Peter Anvin" <hpa@zytor.com>, linux-kernel@vger.kernel.org
References: <20220909104506.738478-1-eesposit@redhat.com>
 <20220909104506.738478-10-eesposit@redhat.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [RFC PATCH 9/9] kvm_main.c: handle atomic memslot update
In-Reply-To: <20220909104506.738478-10-eesposit@redhat.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 9/9/22 12:45, Emanuele Giuseppe Esposito wrote:
> @@ -1782,7 +1782,8 @@ static void kvm_update_flags_memslot(struct kvm *kvm,
>   
>   /*
>    * Takes kvm->slots_arch_lock, and releases it only if
> - * invalid_slot allocation or kvm_prepare_memory_region failed.
> + * invalid_slot allocation, kvm_prepare_memory_region failed
> + * or batch->is_move_delete is true.
>    */

This _is_ getting out of hand, though. :)  It is not clear to me why you 
need to do multiple swaps.  If you did a single swap, you could do all 
the allocations of invalid slots in a separate loop, called with 
slots_arch_lock taken and not released until the final swap.  In other 
words, something like

	mutex_lock(&kvm->slots_arch_lock);
	r = create_invalid_slots();
	if (r < 0)
		return r;

	replace_old_slots();

	// also handles abort on failure
	prepare_memory_regions();
	if (r < 0)
		return r;
	swap();
	finish_memslots();

where each function is little more than a loop over the corresponding 
function called by KVM_SET_MEMORY_REGION.

Paolo