[PATCH] virtio_vring: Fix maximum DMA mapping size computation

[PATCH] virtio_vring: Fix maximum DMA mapping size computation

[Suzuki K Poulose]

The virtio vring is a bit messy about the DMA allocations. We
use the parent device for any DMA alloc/free. However, we use the
child device for calculating the maximum size mapped by the DMA API
for the device. This leads to issues where the max size is set to
SIZE_MAX, and if the virtio device (e.g, virtio-blk) is bound to
swiotlb (e.g., via restricted-dma pool), we end up trying to allocate
larger chunks of memory, which are never satisfied by the swiotlb and
eventually fails.

Fix this by using the parent device (just like all the other
DMA API calls) for the max mapping size.

Fixes: e6d6dd6c875e ("virtio: Introduce virtio_max_dma_size()")
Cc: Joerg Roedel <jroedel@suse.de>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Co-developed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
---
 drivers/virtio/virtio_ring.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 6d2614e34470..028b05d44546 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -268,7 +268,7 @@ size_t virtio_max_dma_size(struct virtio_device *vdev)
 	size_t max_segment_size = SIZE_MAX;
 
 	if (vring_use_dma_api(vdev))
-		max_segment_size = dma_max_mapping_size(&vdev->dev);
+		max_segment_size = dma_max_mapping_size(vdev->dev.parent);
 
 	return max_segment_size;
 }
-- 
2.25.4

Re: [PATCH] virtio_vring: Fix maximum DMA mapping size computation

[Marc Zyngier]

On 2021-12-03 12:16, Suzuki K Poulose wrote:
> The virtio vring is a bit messy about the DMA allocations. We
> use the parent device for any DMA alloc/free. However, we use the
> child device for calculating the maximum size mapped by the DMA API
> for the device. This leads to issues where the max size is set to
> SIZE_MAX, and if the virtio device (e.g, virtio-blk) is bound to
> swiotlb (e.g., via restricted-dma pool), we end up trying to allocate
> larger chunks of memory, which are never satisfied by the swiotlb and
> eventually fails.
> 
> Fix this by using the parent device (just like all the other
> DMA API calls) for the max mapping size.
> 
> Fixes: e6d6dd6c875e ("virtio: Introduce virtio_max_dma_size()")
> Cc: Joerg Roedel <jroedel@suse.de>
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> Cc: Christoph Hellwig <hch@lst.de>
> Cc: Michael S. Tsirkin <mst@redhat.com>
> Cc: Robin Murphy <robin.murphy@arm.com>
> Signed-off-by: Steven Price <steven.price@arm.com>
> Co-developed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
> ---
>  drivers/virtio/virtio_ring.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/virtio/virtio_ring.c 
> b/drivers/virtio/virtio_ring.c
> index 6d2614e34470..028b05d44546 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -268,7 +268,7 @@ size_t virtio_max_dma_size(struct virtio_device 
> *vdev)
>  	size_t max_segment_size = SIZE_MAX;
> 
>  	if (vring_use_dma_api(vdev))
> -		max_segment_size = dma_max_mapping_size(&vdev->dev);
> +		max_segment_size = dma_max_mapping_size(vdev->dev.parent);
> 
>  	return max_segment_size;
>  }

Ah, it looks like we all hit the same thing[1] at the same time!

         M.

[1] https://lore.kernel.org/r/20211201112018.25276-1-will@kernel.org
-- 
Jazz is not dead. It just smells funny...

Re: [PATCH] virtio_vring: Fix maximum DMA mapping size computation

[Suzuki K Poulose]

On 03/12/2021 12:26, Marc Zyngier wrote:
> On 2021-12-03 12:16, Suzuki K Poulose wrote:
>> The virtio vring is a bit messy about the DMA allocations. We
>> use the parent device for any DMA alloc/free. However, we use the
>> child device for calculating the maximum size mapped by the DMA API
>> for the device. This leads to issues where the max size is set to
>> SIZE_MAX, and if the virtio device (e.g, virtio-blk) is bound to
>> swiotlb (e.g., via restricted-dma pool), we end up trying to allocate
>> larger chunks of memory, which are never satisfied by the swiotlb and
>> eventually fails.
>>
>> Fix this by using the parent device (just like all the other
>> DMA API calls) for the max mapping size.
>>
>> Fixes: e6d6dd6c875e ("virtio: Introduce virtio_max_dma_size()")
>> Cc: Joerg Roedel <jroedel@suse.de>
>> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
>> Cc: Christoph Hellwig <hch@lst.de>
>> Cc: Michael S. Tsirkin <mst@redhat.com>
>> Cc: Robin Murphy <robin.murphy@arm.com>
>> Signed-off-by: Steven Price <steven.price@arm.com>
>> Co-developed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
>> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
>> ---
>>  drivers/virtio/virtio_ring.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
>> index 6d2614e34470..028b05d44546 100644
>> --- a/drivers/virtio/virtio_ring.c
>> +++ b/drivers/virtio/virtio_ring.c
>> @@ -268,7 +268,7 @@ size_t virtio_max_dma_size(struct virtio_device 
>> *vdev)
>>      size_t max_segment_size = SIZE_MAX;
>>
>>      if (vring_use_dma_api(vdev))
>> -        max_segment_size = dma_max_mapping_size(&vdev->dev);
>> +        max_segment_size = dma_max_mapping_size(vdev->dev.parent);
>>
>>      return max_segment_size;
>>  }
> 
> Ah, it looks like we all hit the same thing[1] at the same time!

Aha ! Thanks for the pointer Marc.

Suzuki

> 
>          M.
> 
> [1] https://lore.kernel.org/r/20211201112018.25276-1-will@kernel.org