Re: [RFC] vfio/type1: handle case where IOMMU does not support PAGE_SIZE size

From: Eric Auger
Date: Wed Oct 28 2015 - 14:00:37 EST


On 10/28/2015 06:55 PM, Will Deacon wrote:
> On Wed, Oct 28, 2015 at 06:48:41PM +0100, Eric Auger wrote:
>> On 10/28/2015 06:37 PM, Alex Williamson wrote:
>>> Ok, so with hopefully correcting my understand of what this does, isn't
>>> this effectively the same:
>>>
>>> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
>>> index 57d8c37..7db4f5a 100644
>>> --- a/drivers/vfio/vfio_iommu_type1.c
>>> +++ b/drivers/vfio/vfio_iommu_type1.c
>>> @@ -403,13 +403,19 @@ static void vfio_remove_dma(struct vfio_iommu *iommu, stru
>>> static unsigned long vfio_pgsize_bitmap(struct vfio_iommu *iommu)
>>> {
>>> struct vfio_domain *domain;
>>> - unsigned long bitmap = PAGE_MASK;
>>> + unsigned long bitmap = ULONG_MAX;
>>>
>>> mutex_lock(&iommu->lock);
>>> list_for_each_entry(domain, &iommu->domain_list, next)
>>> bitmap &= domain->domain->ops->pgsize_bitmap;
>>> mutex_unlock(&iommu->lock);
>>>
>>> + /* Some comment about how the IOMMU API splits requests */
>>> + if (bitmap & ~PAGE_MASK) {
>>> + bitmap &= PAGE_MASK;
>>> + bitmap |= PAGE_SIZE;
>>> + }
>>> +
>>> return bitmap;
>>> }
>> Yes, to me it is indeed the same
>>>
>>> This would also expose to the user that we're accepting PAGE_SIZE, which
>>> we weren't before, so it was not quite right to just let them do it
>>> anyway. I don't think we even need to get rid of the WARN_ONs, do we?
>>> Thanks,
>>
>> The end-user might be afraid of those latter. Personally I would get rid
>> of them but that's definitively up to you.
>
> I think Alex's point is that the WARN_ON's won't trigger with this patch,
> because he clears those lower bits in the bitmap.
ah yes sure!

Thanks

Eric
>
> Will
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/