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;
From:   Vitaly Kuznetsov <vkuznets@redhat.com>
To:     Radim =?utf-8?B?S3LEjW3DocWZ?= <rkrcmar@redhat.com>
Cc:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org, x86@kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        "K. Y. Srinivasan" <kys@microsoft.com>,
        "Michael Kelley \(EOSG\)" <Michael.H.Kelley@microsoft.com>,
        Roman Kagan <rkagan@virtuozzo.com>,
        "Denis V . Lunev" <den@openvz.org>
Subject: Re: [PATCH v2 2/3] x86/kvm/hyper-v: remove stale entries from vec_bitmap/auto_eoi_bitmap on vector change
References: <20180301141514.3482-1-vkuznets@redhat.com>
        <20180301141514.3482-3-vkuznets@redhat.com>
        <20180308210804.GN12290@flask>
Date:   Fri, 09 Mar 2018 16:21:19 +0100
In-Reply-To: <20180308210804.GN12290@flask> ("Radim \=\?utf-8\?B\?S3LEjW3DocWZ\?\=
 \=\?utf-8\?B\?Iidz\?\= message of "Thu,
        8 Mar 2018 22:08:04 +0100")
Message-ID: <87muzhe0ls.fsf@vitty.brq.redhat.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/25.3 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Radim Krčmář <rkrcmar@redhat.com> writes:

> 2018-03-01 15:15+0100, Vitaly Kuznetsov:
>> When a new vector is written to SINx we update vec_bitmap/auto_eoi_bitmap
>> but we forget to remove old vector from these masks (in case it is not
>> present in some other SINTx).
>> 
>> Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
>> Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
>> ---
>>  arch/x86/include/uapi/asm/hyperv.h |  2 ++
>>  arch/x86/kvm/hyperv.c              | 32 ++++++++++++++++++++++----------
>>  2 files changed, 24 insertions(+), 10 deletions(-)
>> 
>> diff --git a/arch/x86/include/uapi/asm/hyperv.h b/arch/x86/include/uapi/asm/hyperv.h
>> index 197c2e6c7376..62c778a303a1 100644
>> --- a/arch/x86/include/uapi/asm/hyperv.h
>> +++ b/arch/x86/include/uapi/asm/hyperv.h
>> @@ -318,6 +318,8 @@ typedef struct _HV_REFERENCE_TSC_PAGE {
>>  #define HV_SYNIC_SINT_COUNT		(16)
>>  /* Define the expected SynIC version. */
>>  #define HV_SYNIC_VERSION_1		(0x1)
>> +/* Valid SynIC vectors are 16-255. */
>> +#define HV_SYNIC_FIRST_VALID_VECTOR	(16)
>>  
>>  #define HV_SYNIC_CONTROL_ENABLE		(1ULL << 0)
>>  #define HV_SYNIC_SIMP_ENABLE		(1ULL << 0)
>> diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
>> index 05f414525538..6d14f808145d 100644
>> --- a/arch/x86/kvm/hyperv.c
>> +++ b/arch/x86/kvm/hyperv.c
>> @@ -74,13 +74,30 @@ static bool synic_has_vector_auto_eoi(struct kvm_vcpu_hv_synic *synic,
>>  	return false;
>>  }
>>  
>> +static void synic_update_vector(struct kvm_vcpu_hv_synic *synic,
>> +				int vector)
>> +{
>> +	if (vector < HV_SYNIC_FIRST_VALID_VECTOR)
>> +		return;
>> +
>> +	if (synic_has_vector_connected(synic, vector))
>> +		__set_bit(vector, synic->vec_bitmap);
>> +	else
>> +		__clear_bit(vector, synic->vec_bitmap);
>> +
>> +	if (synic_has_vector_auto_eoi(synic, vector))
>> +		__set_bit(vector, synic->auto_eoi_bitmap);
>> +	else
>> +		__clear_bit(vector, synic->auto_eoi_bitmap);
>> +}
>> +
>>  static int synic_set_sint(struct kvm_vcpu_hv_synic *synic, int sint,
>>  			  u64 data, bool host)
>>  {
>> -	int vector;
>> +	int vector, old_vector;
>>  
>>  	vector = data & HV_SYNIC_SINT_VECTOR_MASK;
>> -	if (vector < 16 && !host)
>> +	if (vector < HV_SYNIC_FIRST_VALID_VECTOR && !host)
>>  		return 1;
>>  	/*
>>  	 * Guest may configure multiple SINTs to use the same vector, so
>> @@ -88,18 +105,13 @@ static int synic_set_sint(struct kvm_vcpu_hv_synic *synic, int sint,
>>  	 * bitmap of vectors with auto-eoi behavior.  The bitmaps are
>>  	 * updated here, and atomically queried on fast paths.
>>  	 */
>> +	old_vector = synic_read_sint(synic, sint) & HV_SYNIC_SINT_VECTOR_MASK;
>>  
>>  	atomic64_set(&synic->sint[sint], data);
>>  
>> -	if (synic_has_vector_connected(synic, vector))
>> -		__set_bit(vector, synic->vec_bitmap);
>> -	else
>> -		__clear_bit(vector, synic->vec_bitmap);
>> +	synic_update_vector(synic, old_vector);
>> 
>> -	if (synic_has_vector_auto_eoi(synic, vector))
>> -		__set_bit(vector, synic->auto_eoi_bitmap);
>> -	else
>> -		__clear_bit(vector, synic->auto_eoi_bitmap);
>> +	synic_update_vector(synic, vector);
>
> This looks like it solves the problem when we get two SINTs with the
> same vector back-to-back , but shouldn't these bits really be cleared on
> EOI (either auto or manual)?

Hmm,

I was trying to address the following issue: guest programs SynIC's
SINTx with some vector but later re-programs it with a different
one. Without the patch synic->vec_bitmap and synic->auto_eoi_bitmap keep
stale data. If there's no concurrent interrupt than we're safe, but what
happens if there is one...

kvm_hv_synic_send_eoi() already goes through all SINTx but we already
updated vector so it won't find any. We could've added something like
'old_vector' but what if the request with this vector came _after_ we
re-programed SynIC (and, so, it wasn't meant to be serviced by SynIC?)?

-- 
  Vitaly