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: <0c830e36-fcf2-fab6-aed9-7b6a6736140f@redhat.com>
Date:   Thu, 31 Mar 2022 11:34:35 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.7.0
Subject: Re: [PATCH] KVM: MMU: propagate alloc_workqueue failure
Content-Language: en-US
To:     David Matlack <dmatlack@google.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org
References: <20220330165510.213111-1-pbonzini@redhat.com>
 <YkTs5BU24zrw30hK@google.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
In-Reply-To: <YkTs5BU24zrw30hK@google.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 3/31/22 01:51, David Matlack wrote:
>> -void kvm_mmu_init_vm(struct kvm *kvm)
>> +int kvm_mmu_init_vm(struct kvm *kvm)
>>   {
>>   	struct kvm_page_track_notifier_node *node = &kvm->arch.mmu_sp_tracker;
>> +	int r;
>>   
>> +	INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
>> +	INIT_LIST_HEAD(&kvm->arch.zapped_obsolete_pages);
>> +	INIT_LIST_HEAD(&kvm->arch.lpage_disallowed_mmu_pages);
> 
> I agree with moving these but that should probably be done in a separate
> commit.

Ok.

>> -	kvm->arch.tdp_mmu_zap_wq =
>> -		alloc_workqueue("kvm", WQ_UNBOUND|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE, 0);
>> -
>> -	return true;
>> +	kvm->arch.tdp_mmu_zap_wq = wq;
> 
> Suggest moving this to just after checking the return value of
> alloc_workqueue().

This is intentional, in case we have other future allocations, to avoid 
having to NULL out the field in the unwind path.  It's a matter of taste 
I guess.

>> +	return 1;
> 
> Perhaps return 0 until we have a reason to differentiate the 2 cases.

Yeah, though I wanted to preserve the previous behavior.

>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index fe2171b11441..89b6efb7f504 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -11629,12 +11629,13 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
>>   
>>   	ret = kvm_page_track_init(kvm);
>>   	if (ret)
>> -		return ret;
>> +		goto out;
> 
> nit: This goto is unnecessary.

True, but I prefer to be consistent in using "goto" so that any future 
additions are careful about preserving the chain.

Paolo