On Tue, May 03, 2022 at 09:11:02PM -0300, Jason Gunthorpe wrote:
> This is based on Robins draft here:
>
> https://lore.kernel.org/linux-iommu/18831161-473f-e04f-4a81-1c7062ad1...@arm.com/
>
> With some rework. I re-organized the call chains instead of introducing
> iommu_group_user_attached(), fixed a recursive locking for
> iommu_group_get_purgatory(), and made a proper commit message.
>
> Still only compile tested, so RFCish.
>
> Nicolin/Lu? What do you think, can you check it?
I am able to repro the issue on ARM64 and give this a quick try.
But the patch seems to need to include the following change too.
diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index 94d99768023c..9bb108d01baa 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -2040,7 +2040,8 @@ static int __iommu_attach_group(struct iommu_domain
*domain,
{
int ret;
- if (group->domain && group->domain != group->default_domain)
+ if (group->domain && group->domain != group->default_domain &&
+ group->domain != group->blocking_domain)
return -EBUSY;
ret = __iommu_group_for_each_dev(group, domain,
> @@ -2072,38 +2072,66 @@ static int iommu_group_do_detach_device(struct device
> *dev, void *data)
> return 0;
> }
>
> -static void __iommu_detach_group(struct iommu_domain *domain,
> - struct iommu_group *group)
> +static int __iommu_group_attach_domain(struct iommu_group *group,
> + struct iommu_domain *new_domain)
> {
> int ret;
>
> + if (group->domain == new_domain)
> + return 0;
> +
> /*
> - * If the group has been claimed already, do not re-attach the default
> - * domain.
> + * A NULL domain means to call the detach_dev() op. New drivers should
> + * use a IOMMU_DOMAIN_IDENTITY domain instead of a NULL default_domain
an IOMMU_DOMAIN_IDENTITY?
Just a nit here. I will take a closer look at the change tomorrow.
Thanks
Nic
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu
