linux-kernel - Re: Re: [PATCH] virtio_ring: use u32 for virtio_max_dma

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <8254f1bb-dfd0-5ff0-651f-7f664bca65ba@bytedance.com>
Date:   Wed, 10 May 2023 15:52:13 +0800
From:   zhenwei pi <pizhenwei@...edance.com>
To:     Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
Cc:     virtualization@...ts.linux-foundation.org,
        linux-kernel@...r.kernel.org, Joerg Roedel <jroedel@...e.de>,
        mst@...hat.com, jasowang@...hat.com
Subject: Re: Re: [PATCH] virtio_ring: use u32 for virtio_max_dma_size



On 5/10/23 11:26, Xuan Zhuo wrote:
> On Wed, 10 May 2023 10:54:37 +0800, zhenwei pi <pizhenwei@...edance.com> wrote:
>> Both split ring and packed ring use 32bits to describe the length of
>> a descriptor: see struct vring_desc and struct vring_packed_desc.
>> This means the max segment size supported by virtio is U32_MAX.
>>
>> An example of virtio_max_dma_size in virtio_blk.c:
>>    u32 v, max_size;
>>
>>    max_size = virtio_max_dma_size(vdev);  -> implicit convert
>>    err = virtio_cread_feature(vdev, VIRTIO_BLK_F_SIZE_MAX,
>>                               struct virtio_blk_config, size_max, &v);
>>    max_size = min(max_size, v);
>>
>> There is a risk during implicit convert here, once virtio_max_dma_size
>> returns 4G, max_size becomes 0.
>>
>> Fixes: e6d6dd6c875e ("virtio: Introduce virtio_max_dma_size()")
>> Cc: Joerg Roedel <jroedel@...e.de>
>> Signed-off-by: zhenwei pi <pizhenwei@...edance.com>
>> ---
>>   drivers/virtio/virtio_ring.c | 12 ++++++++----
>>   include/linux/virtio.h       |  2 +-
>>   2 files changed, 9 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
>> index c5310eaf8b46..55cfecf030a1 100644
>> --- a/drivers/virtio/virtio_ring.c
>> +++ b/drivers/virtio/virtio_ring.c
>> @@ -289,12 +289,16 @@ static bool vring_use_dma_api(const struct virtio_device *vdev)
>>   	return false;
>>   }
>>
>> -size_t virtio_max_dma_size(const struct virtio_device *vdev)
>> +u32 virtio_max_dma_size(const struct virtio_device *vdev)
> 
> 
> LGTM
> 
> But, should we change the parameter to vq, then use the dma_dev?
> 
> @Jason
> 
> Thanks.
> 

The max DMA size is a attribute of a virtio device rather than any VQ, 
so I guess virtio_max_dma_size(const struct virtio_device *vdev) is clear.

On the other hand, if changing the parameter to vq, we need select a VQ, 
then the question is:
1, which VQ to select? VQ0 or a random one? this leads confusing.
2, The virtio spec defines: Each device can have zero or more virtqueues


> 
>>   {
>> -	size_t max_segment_size = SIZE_MAX;
>> +	u32 max_segment_size = U32_MAX;
>>
>> -	if (vring_use_dma_api(vdev))
>> -		max_segment_size = dma_max_mapping_size(vdev->dev.parent);
>> +	if (vring_use_dma_api(vdev)) {
>> +		size_t max_dma_size = dma_max_mapping_size(vdev->dev.parent);
>> +
>> +		if (max_dma_size < max_segment_size)
>> +			max_segment_size = max_dma_size;
>> +	}
>>
>>   	return max_segment_size;
>>   }
>> diff --git a/include/linux/virtio.h b/include/linux/virtio.h
>> index b93238db94e3..1a605f408329 100644
>> --- a/include/linux/virtio.h
>> +++ b/include/linux/virtio.h
>> @@ -147,7 +147,7 @@ int virtio_device_restore(struct virtio_device *dev);
>>   #endif
>>   void virtio_reset_device(struct virtio_device *dev);
>>
>> -size_t virtio_max_dma_size(const struct virtio_device *vdev);
>> +u32 virtio_max_dma_size(const struct virtio_device *vdev);
>>
>>   #define virtio_device_for_each_vq(vdev, vq) \
>>   	list_for_each_entry(vq, &vdev->vqs, list)
>> --
>> 2.20.1
>>

-- 
zhenwei pi