Re: [PATCH] vfio/type1: Fix vfio_find_dma_valid return

[Anthony Yznaga <anthony.yznaga@oracle.com>]

On 8/20/21 9:24 AM, Alex Williamson wrote:
> On Thu, 19 Aug 2021 16:53:57 -0700
> Anthony Yznaga <anthony.yznaga@oracle.com> wrote:
>
>> Fix vfio_find_dma_valid to return WAITED on success if it was necessary
>> to wait which mean iommu lock was dropped and reacquired.  This allows
>> vfio_iommu_type1_pin_pages to recheck vaddr_invalid_count and possibly
>> avoid the checking the validity of every vaddr in its list.
>>
>> Signed-off-by: Anthony Yznaga <anthony.yznaga@oracle.com>
>> ---
>>  drivers/vfio/vfio_iommu_type1.c | 5 +++--
>>  1 file changed, 3 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
>> index a3e925a41b0d..7ca8c4e95da4 100644
>> --- a/drivers/vfio/vfio_iommu_type1.c
>> +++ b/drivers/vfio/vfio_iommu_type1.c
>> @@ -612,6 +612,7 @@ static int vfio_find_dma_valid(struct vfio_iommu *iommu, dma_addr_t start,
>>  			       size_t size, struct vfio_dma **dma_p)
>>  {
>>  	int ret;
>> +	int waited = 0;
>>  
>>  	do {
>>  		*dma_p = vfio_find_dma(iommu, start, size);
>> @@ -620,10 +621,10 @@ static int vfio_find_dma_valid(struct vfio_iommu *iommu, dma_addr_t start,
>>  		else if (!(*dma_p)->vaddr_invalid)
>>  			ret = 0;
>>  		else
>> -			ret = vfio_wait(iommu);
>> +			ret = waited = vfio_wait(iommu);
>>  	} while (ret > 0);
>>  
>> -	return ret;
>> +	return ret ? ret : waited;
>>  }
>>  
>>  /*
> How about...
>
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index 0b4f7c174c7a..0e9217687f5c 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -612,17 +612,17 @@ static int vfio_wait(struct vfio_iommu *iommu)
>  static int vfio_find_dma_valid(struct vfio_iommu *iommu, dma_addr_t start,
>  			       size_t size, struct vfio_dma **dma_p)
>  {
> -	int ret;
> +	int ret = 0;
>  
>  	do {
>  		*dma_p = vfio_find_dma(iommu, start, size);
>  		if (!*dma_p)
> -			ret = -EINVAL;
> +			return -EINVAL;
>  		else if (!(*dma_p)->vaddr_invalid)
> -			ret = 0;
> +			return ret;
>  		else
>  			ret = vfio_wait(iommu);
> -	} while (ret > 0);
> +	} while (ret == WAITED);
>  
>  	return ret;
>  }
>

Even better.  Should I send a new patch?

Thanks,
Anthony