Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Subject: Re: [PATCH 2/2] KVM: Fix rcu splat if vm creation fails
To:     Wanpeng Li <kernellwp@gmail.com>
Cc:     LKML <linux-kernel@vger.kernel.org>, kvm <kvm@vger.kernel.org>,
        =?UTF-8?B?UmFkaW0gS3LEjW3DocWZ?= <rkrcmar@redhat.com>,
        Sean Christopherson <sean.j.christopherson@intel.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>
References: <1572848879-21011-1-git-send-email-wanpengli@tencent.com>
 <1572848879-21011-2-git-send-email-wanpengli@tencent.com>
 <c32d632b-8fb0-f7c6-4937-07c30769b924@redhat.com>
 <CANRm+CzkbrbE2C2yFKL1=mQCBCZMfVH8Tue3eXXqTL5Z1VUB5w@mail.gmail.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
Openpgp: preference=signencrypt
Message-ID: <ee896a34-0914-8d3c-bcdd-5aede1743190@redhat.com>
Date:   Mon, 4 Nov 2019 13:21:52 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.8.0
MIME-Version: 1.0
In-Reply-To: <CANRm+CzkbrbE2C2yFKL1=mQCBCZMfVH8Tue3eXXqTL5Z1VUB5w@mail.gmail.com>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 04/11/19 13:16, Wanpeng Li wrote:
>> I don't understand this one, hasn't
>>
>>         WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
>>
>> decreased the conut already?  With your patch the refcount would then
>> underflow.
> 
> r = kvm_arch_init_vm(kvm, type);
> if (r)
>     goto out_err_no_arch_destroy_vm;
> 
> out_err_no_disable:
>     kvm_arch_destroy_vm(kvm);
>     WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
> out_err_no_arch_destroy_vm:
> 
> So, if kvm_arch_init_vm() fails, we will not execute
> WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));

Uuh of course.  But I'd rather do the opposite: move the refcount_set
earlier so that refcount_dec_and_test can be moved after
no_arch_destroy_vm.  Moving the refcount_set is not strictly necessary,
but avoids the introduction of yet another label.

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index e22ff63e5b1a..e7a07132cd7f 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -650,6 +650,7 @@ static struct kvm *kvm_create_vm(unsigned long type)
 	if (init_srcu_struct(&kvm->irq_srcu))
 		goto out_err_no_irq_srcu;

+	refcount_set(&kvm->users_count, 1);
 	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++) {
 		struct kvm_memslots *slots = kvm_alloc_memslots();

@@ -667,7 +668,6 @@ static struct kvm *kvm_create_vm(unsigned long type)
 			goto out_err_no_arch_destroy_vm;
 	}

-	refcount_set(&kvm->users_count, 1);
 	r = kvm_arch_init_vm(kvm, type);
 	if (r)
 		goto out_err_no_arch_destroy_vm;
@@ -696,8 +696,8 @@ static struct kvm *kvm_create_vm(unsigned long type)
 	hardware_disable_all();
 out_err_no_disable:
 	kvm_arch_destroy_vm(kvm);
-	WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
 out_err_no_arch_destroy_vm:
+	WARN_ON_ONCE(!refcount_dec_and_test(&kvm->users_count));
 	for (i = 0; i < KVM_NR_BUSES; i++)
 		kfree(kvm_get_bus(kvm, i));
 	for (i = 0; i < KVM_ADDRESS_SPACE_NUM; i++)