iommufd: Add a flag to skip clearing of IOPTE dirty

VFIO has an operation where it unmaps an IOVA while returning a bitmap with
the dirty data. In reality the operation doesn't quite query the IO
pagetables that the PTE was dirty or not. Instead it marks as dirty on
anything that was mapped, and doing so in one syscall.

In IOMMUFD the equivalent is done in two operations by querying with
GET_DIRTY_IOVA followed by UNMAP_IOVA. However, this would incur two TLB
flushes given that after clearing dirty bits IOMMU implementations require
invalidating their IOTLB, plus another invalidation needed for the UNMAP.
To allow dirty bits to be queried faster, add a flag
(IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR) that requests to not clear the dirty
bits from the PTE (but just reading them), under the expectation that the
next operation is the unmap. An alternative is to unmap and just
perpectually mark as dirty as that's the same behaviour as today. So here
equivalent functionally can be provided with unmap alone, and if real dirty
info is required it will amortize the cost while querying.

There's still a race against DMA where in theory the unmap of the IOVA
(when the guest invalidates the IOTLB via emulated iommu) would race
against the VF performing DMA on the same IOVA. As discussed in [0], we are
accepting to resolve this race as throwing away the DMA and it doesn't
matter if it hit physical DRAM or not, the VM can't tell if we threw it
away because the DMA was blocked or because we failed to copy the DRAM.

[0] https://lore.kernel.org/linux-iommu/20220502185239.GR8364@nvidia.com/

Link: https://lore.kernel.org/r/20231024135109.73787-10-joao.m.martins@oracle.com
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

2 files changed, 9 insertions, 3 deletions

diff --git a/drivers/iommu/iommufd/hw_pagetable.c b/drivers/iommu/iommufd/hw_pagetable.c
index 7316f69110ef..72a5269984b0 100644
--- a/drivers/iommu/iommufd/hw_pagetable.c
+++ b/drivers/iommu/iommufd/hw_pagetable.c
@@ -228,7 +228,8 @@ int iommufd_hwpt_get_dirty_bitmap(struct iommufd_ucmd *ucmd)
 	struct iommufd_ioas *ioas;
 	int rc = -EOPNOTSUPP;
 
-	if ((cmd->flags || cmd->__reserved))
+	if ((cmd->flags & ~(IOMMU_HWPT_GET_DIRTY_BITMAP_NO_CLEAR)) ||
+	    cmd->__reserved)
 		return -EOPNOTSUPP;
 
 	hwpt = iommufd_get_hwpt(ucmd, cmd->hwpt_id);
diff --git a/drivers/iommu/iommufd/io_pagetable.c b/drivers/iommu/iommufd/io_pagetable.c
index 255264e796fb..9f060abe53b6 100644
--- a/drivers/iommu/iommufd/io_pagetable.c
+++ b/drivers/iommu/iommufd/io_pagetable.c
@@ -414,6 +414,7 @@ int iopt_map_user_pages(struct iommufd_ctx *ictx, struct io_pagetable *iopt,
 }
 
 struct iova_bitmap_fn_arg {
+	unsigned long flags;
 	struct io_pagetable *iopt;
 	struct iommu_domain *domain;
 	struct iommu_dirty_bitmap *dirty;
@@ -430,13 +431,14 @@ static int __iommu_read_and_clear_dirty(struct iova_bitmap *bitmap,
 	struct iommu_dirty_bitmap *dirty = arg->dirty;
 	const struct iommu_dirty_ops *ops = domain->dirty_ops;
 	unsigned long last_iova = iova + length - 1;
+	unsigned long flags = arg->flags;
 	int ret;
 
 	iopt_for_each_contig_area(&iter, area, arg->iopt, iova, last_iova) {
 		unsigned long last = min(last_iova, iopt_area_last_iova(area));
 
 		ret = ops->read_and_clear_dirty(domain, iter.cur_iova,
-						last - iter.cur_iova + 1, 0,
+						last - iter.cur_iova + 1, flags,
 						dirty);
 		if (ret)
 			return ret;
@@ -470,12 +472,15 @@ iommu_read_and_clear_dirty(struct iommu_domain *domain,
 
 	iommu_dirty_bitmap_init(&dirty, iter, &gather);
 
+	arg.flags = flags;
 	arg.iopt = iopt;
 	arg.domain = domain;
 	arg.dirty = &dirty;
 	iova_bitmap_for_each(iter, &arg, __iommu_read_and_clear_dirty);
 
-	iommu_iotlb_sync(domain, &gather);
+	if (!(flags & IOMMU_DIRTY_NO_CLEAR))
+		iommu_iotlb_sync(domain, &gather);
+
 	iova_bitmap_free(iter);
 
 	return ret;