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: <95c28949-8732-8812-c255-79467dafb5c8@linux.ibm.com>
Date:   Wed, 23 Mar 2022 09:52:16 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.6.0
Subject: Re: [PATCH] KVM: s390: Fix lockdep issue in vm memop
Content-Language: en-US
To:     Janosch Frank <frankja@linux.ibm.com>,
        Christian Borntraeger <borntraeger@linux.ibm.com>,
        Claudio Imbrenda <imbrenda@linux.ibm.com>,
        Heiko Carstens <hca@linux.ibm.com>,
        Vasily Gorbik <gor@linux.ibm.com>,
        Alexander Gordeev <agordeev@linux.ibm.com>
Cc:     David Hildenbrand <david@redhat.com>,
        Sven Schnelle <svens@linux.ibm.com>, kvm@vger.kernel.org,
        linux-s390@vger.kernel.org, linux-kernel@vger.kernel.org
References: <20220322153204.2637400-1-scgl@linux.ibm.com>
 <44618f05-9aee-5aa5-b036-dd838285b26f@linux.ibm.com>
From:   Janis Schoetterl-Glausch <scgl@linux.ibm.com>
In-Reply-To: <44618f05-9aee-5aa5-b036-dd838285b26f@linux.ibm.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 3/23/22 08:58, Janosch Frank wrote:
> On 3/22/22 16:32, Janis Schoetterl-Glausch wrote:
>> Issuing a memop on a protected vm does not make sense,
> 
> Issuing a vm memop on a protected vm...
> 
> The cpu memop still makes sense, no?

The vcpu memop does hold the vcpu->lock, so no lockdep issue.
If you issue a vcpu memop while enabling protected virtualization,
the memop might find that the vcpu is not protected, while other vcpus
might already be, but I don't think there's a way to create secure memory
concurrent with the memop.
> 
>> neither is the memory readable/writable, nor does it make sense to check
>> storage keys. This is why the ioctl will return -EINVAL when it detects
>> the vm to be protected. However, in order to ensure that the vm cannot
>> become protected during the memop, the kvm->lock would need to be taken
>> for the duration of the ioctl. This is also required because
>> kvm_s390_pv_is_protected asserts that the lock must be held.
>> Instead, don't try to prevent this. If user space enables secure
>> execution concurrently with a memop it must accecpt the possibility of
>> the memop failing.
>> Still check if the vm is currently protected, but without locking and
>> consider it a heuristic.
>>
>> Fixes: ef11c9463ae0 ("KVM: s390: Add vm IOCTL for key checked guest absolute memory access")
>> Signed-off-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
> 
> Makes sense to me.
> 
> Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
> 
>> ---
>>   arch/s390/kvm/kvm-s390.c | 11 ++++++++++-
>>   1 file changed, 10 insertions(+), 1 deletion(-)
>>
>> diff --git a/arch/s390/kvm/kvm-s390.c b/arch/s390/kvm/kvm-s390.c
>> index ca96f84db2cc..53adbe86a68f 100644
>> --- a/arch/s390/kvm/kvm-s390.c
>> +++ b/arch/s390/kvm/kvm-s390.c
>> @@ -2385,7 +2385,16 @@ static int kvm_s390_vm_mem_op(struct kvm *kvm, struct kvm_s390_mem_op *mop)
>>           return -EINVAL;
>>       if (mop->size > MEM_OP_MAX_SIZE)
>>           return -E2BIG;
>> -    if (kvm_s390_pv_is_protected(kvm))
>> +    /*
>> +     * This is technically a heuristic only, if the kvm->lock is not
>> +     * taken, it is not guaranteed that the vm is/remains non-protected.
>> +     * This is ok from a kernel perspective, wrongdoing is detected
>> +     * on the access, -EFAULT is returned and the vm may crash the
>> +     * next time it accesses the memory in question.
>> +     * There is no sane usecase to do switching and a memop on two
>> +     * different CPUs at the same time.
>> +     */
>> +    if (kvm_s390_pv_get_handle(kvm))
>>           return -EINVAL;
>>       if (mop->flags & KVM_S390_MEMOP_F_SKEY_PROTECTION) {
>>           if (access_key_invalid(mop->key))
>>
>> base-commit: c9b8fecddb5bb4b67e351bbaeaa648a6f7456912
>