Virtualization - Dec 2020 - [PATCH v3 05/19] vdpa_sim: remove the limit of IOTLB entries

On 2020/12/4 ??1:04, Stefano Garzarella wrote:
> The simulated devices can support multiple queues, so this limit
> should be defined according to the number of queues supported by
> the device.
>
> Since we are in a simulator, let's simply remove that limit.
>
> Suggested-by: Jason Wang <jasowang at redhat.com>
> Acked-by: Jason Wang <jasowang at redhat.com>
> Signed-off-by: Stefano Garzarella <sgarzare at redhat.com>

Rethink about this, since simulator can be used by VM, so the allocation 
is actually guest trigger-able when vIOMMU is enabled.

This means we need a limit somehow, (e.g I remember swiotlb is about 
64MB by default). Or having a module parameter for this.

Btw, have you met any issue when using 2048, I guess it can happen when 
we run several processes in parallel?

> ---
> v3:
> - used VHOST_IOTLB_UNLIMITED macro [Jason]
> v2:
> - added VDPASIM_IOTLB_LIMIT macro [Jason]
> ---
>   drivers/vdpa/vdpa_sim/vdpa_sim.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim.c
b/drivers/vdpa/vdpa_sim/vdpa_sim.c
> index 295a770caac0..688aceaa6543 100644
> --- a/drivers/vdpa/vdpa_sim/vdpa_sim.c
> +++ b/drivers/vdpa/vdpa_sim/vdpa_sim.c
> @@ -368,7 +368,7 @@ static struct vdpasim *vdpasim_create(void)
>   	if (!vdpasim->vqs)
>   		goto err_iommu;
>   
> -	vdpasim->iommu = vhost_iotlb_alloc(2048, 0);
> +	vdpasim->iommu = vhost_iotlb_alloc(VHOST_IOTLB_UNLIMITED, 0);
>   	if (!vdpasim->iommu)
>   		goto err_iommu;
>

On Mon, Dec 07, 2020 at 12:00:07PM +0800, Jason Wang
wrote:
>
>On 2020/12/4 ??1:04, Stefano Garzarella wrote:
>>The simulated devices can support multiple queues, so this limit
>>should be defined according to the number of queues supported by
>>the device.
>>
>>Since we are in a simulator, let's simply remove that limit.
>>
>>Suggested-by: Jason Wang <jasowang at redhat.com>
>>Acked-by: Jason Wang <jasowang at redhat.com>
>>Signed-off-by: Stefano Garzarella <sgarzare at redhat.com>
>
>
>Rethink about this, since simulator can be used by VM, so the 
>allocation is actually guest trigger-able when vIOMMU is enabled.
>
>This means we need a limit somehow, (e.g I remember swiotlb is about 
>64MB by default). Or having a module parameter for this.
>
>Btw, have you met any issue when using 2048, I guess it can happen 
>when we run several processes in parallel?
>
No, I didn't try with the limit.
This came from the reviews to Max's patches.

Anyway I can add a module parameter to control that limit, do you think 
is better to set a limit per queue (the parameter per number of queues), 
or just a value for the entire device?

Thanks,
Stefano