| Message ID | 3-v1-8060f06462cc+c0a39-dart_paging_jgg@nvidia.com (mailing list archive) |
|---|---|
| State | Handled Elsewhere |
| Headers | show |
| Series | iommu: Convert dart & iommufd to the new domain_alloc_paging() | expand |
On 9/23/23 1:07 AM, Jason Gunthorpe wrote: > Trivially migrate to the ops->blocked_domain for the existing global > static. > > Signed-off-by: Jason Gunthorpe<jgg@nvidia.com> > --- > drivers/iommu/intel/iommu.c | 3 +-- > 1 file changed, 1 insertion(+), 2 deletions(-) Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com> P.S. We can further do the same thing to the identity domain. I will clean it up after all patches are landed. Best regards, baolu
On Mon, Sep 25, 2023 at 10:29:52AM +0800, Baolu Lu wrote: > On 9/23/23 1:07 AM, Jason Gunthorpe wrote: > > Trivially migrate to the ops->blocked_domain for the existing global > > static. > > > > Signed-off-by: Jason Gunthorpe<jgg@nvidia.com> > > --- > > drivers/iommu/intel/iommu.c | 3 +-- > > 1 file changed, 1 insertion(+), 2 deletions(-) > > Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com> > > P.S. We can further do the same thing to the identity domain. I will > clean it up after all patches are landed. I looked at that, and it is not trivial.. Both the Intel and virtio-iommu drivers create an "identity" domain out of a paging domain and pass that off as a true "identity" domain. So neither can set the global static since the determination is at runtime.. What I was thinking about doing is consolidating that code so that the core logic is the thing turning a paging domain into an identity domain. Jason
On 9/25/23 7:41 PM, Jason Gunthorpe wrote: > On Mon, Sep 25, 2023 at 10:29:52AM +0800, Baolu Lu wrote: >> On 9/23/23 1:07 AM, Jason Gunthorpe wrote: >>> Trivially migrate to the ops->blocked_domain for the existing global >>> static. >>> >>> Signed-off-by: Jason Gunthorpe<jgg@nvidia.com> >>> --- >>> drivers/iommu/intel/iommu.c | 3 +-- >>> 1 file changed, 1 insertion(+), 2 deletions(-) >> >> Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com> >> >> P.S. We can further do the same thing to the identity domain. I will >> clean it up after all patches are landed. > > I looked at that, and it is not trivial.. > > Both the Intel and virtio-iommu drivers create an "identity" domain > out of a paging domain and pass that off as a true "identity" > domain. So neither can set the global static since the determination > is at runtime.. Emm, yes. The early hardware requires a real 1:1 mapped page table. The recent implementations are no longer needed. > > What I was thinking about doing is consolidating that code so that the > core logic is the thing turning a paging domain into an identity > domain. Yes. It's not trivial. Needs a separated series with some refactoring efforts. Best regards, baolu
diff --git a/drivers/iommu/intel/iommu.c b/drivers/iommu/intel/iommu.c index 0d0cb2534972a2..b79188d0c67e95 100644 --- a/drivers/iommu/intel/iommu.c +++ b/drivers/iommu/intel/iommu.c @@ -4041,8 +4041,6 @@ static struct iommu_domain *intel_iommu_domain_alloc(unsigned type) struct iommu_domain *domain; switch (type) { - case IOMMU_DOMAIN_BLOCKED: - return &blocking_domain; case IOMMU_DOMAIN_DMA: case IOMMU_DOMAIN_UNMANAGED: dmar_domain = alloc_domain(type); @@ -4803,6 +4801,7 @@ static void *intel_iommu_hw_info(struct device *dev, u32 *length, u32 *type) } const struct iommu_ops intel_iommu_ops = { + .blocked_domain = &blocking_domain, .capable = intel_iommu_capable, .hw_info = intel_iommu_hw_info, .domain_alloc = intel_iommu_domain_alloc,
Trivially migrate to the ops->blocked_domain for the existing global static. Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> --- drivers/iommu/intel/iommu.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)