2021-01-07 04:48:30

by Keqian Zhu

[permalink] [raw]
Subject: [PATCH 0/6] vfio/iommu_type1: Some optimizations about dirty tracking

Hi,

In patch series[1], I put forward some fixes and optimizations for
vfio_iommu_type1. This extracts and improves the optimization part
of it.

[1] https://lore.kernel.org/linux-iommu/[email protected]/T/#t

Thanks,
Keqian


Keqian Zhu (6):
vfio/iommu_type1: Make an explicit "promote" semantic
vfio/iommu_type1: Ignore external domain when promote pinned_scope
vfio/iommu_type1: Initially set the pinned_page_dirty_scope
vfio/iommu_type1: Drop parameter "pgsize" of vfio_dma_bitmap_alloc_all
vfio/iommu_type1: Drop parameter "pgsize" of vfio_iova_dirty_bitmap
vfio/iommu_type1: Drop parameter "pgsize" of update_user_bitmap

drivers/vfio/vfio_iommu_type1.c | 67 +++++++++++++--------------------
1 file changed, 26 insertions(+), 41 deletions(-)

--
2.19.1


2021-01-07 04:48:42

by Keqian Zhu

[permalink] [raw]
Subject: [PATCH 4/6] vfio/iommu_type1: Drop parameter "pgsize" of vfio_dma_bitmap_alloc_all

We always use the smallest supported page size of vfio_iommu as
pgsize. Remove parameter "pgsize" of vfio_dma_bitmap_alloc_all.

Signed-off-by: Keqian Zhu <[email protected]>
---
drivers/vfio/vfio_iommu_type1.c | 8 +++-----
1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index b596c482487b..080c05b129ee 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -236,9 +236,10 @@ static void vfio_dma_populate_bitmap(struct vfio_dma *dma, size_t pgsize)
}
}

-static int vfio_dma_bitmap_alloc_all(struct vfio_iommu *iommu, size_t pgsize)
+static int vfio_dma_bitmap_alloc_all(struct vfio_iommu *iommu)
{
struct rb_node *n;
+ size_t pgsize = (size_t)1 << __ffs(iommu->pgsize_bitmap);

for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
@@ -2761,12 +2762,9 @@ static int vfio_iommu_type1_dirty_pages(struct vfio_iommu *iommu,
return -EINVAL;

if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_START) {
- size_t pgsize;
-
mutex_lock(&iommu->lock);
- pgsize = 1 << __ffs(iommu->pgsize_bitmap);
if (!iommu->dirty_page_tracking) {
- ret = vfio_dma_bitmap_alloc_all(iommu, pgsize);
+ ret = vfio_dma_bitmap_alloc_all(iommu);
if (!ret)
iommu->dirty_page_tracking = true;
}
--
2.19.1

2021-01-15 23:40:26

by Alex Williamson

[permalink] [raw]
Subject: Re: [PATCH 4/6] vfio/iommu_type1: Drop parameter "pgsize" of vfio_dma_bitmap_alloc_all

On Thu, 7 Jan 2021 12:43:59 +0800
Keqian Zhu <[email protected]> wrote:

> We always use the smallest supported page size of vfio_iommu as
> pgsize. Remove parameter "pgsize" of vfio_dma_bitmap_alloc_all.
>
> Signed-off-by: Keqian Zhu <[email protected]>
> ---
> drivers/vfio/vfio_iommu_type1.c | 8 +++-----
> 1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index b596c482487b..080c05b129ee 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -236,9 +236,10 @@ static void vfio_dma_populate_bitmap(struct vfio_dma *dma, size_t pgsize)
> }
> }
>
> -static int vfio_dma_bitmap_alloc_all(struct vfio_iommu *iommu, size_t pgsize)
> +static int vfio_dma_bitmap_alloc_all(struct vfio_iommu *iommu)
> {
> struct rb_node *n;
> + size_t pgsize = (size_t)1 << __ffs(iommu->pgsize_bitmap);
>
> for (n = rb_first(&iommu->dma_list); n; n = rb_next(n)) {
> struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
> @@ -2761,12 +2762,9 @@ static int vfio_iommu_type1_dirty_pages(struct vfio_iommu *iommu,
> return -EINVAL;
>
> if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_START) {
> - size_t pgsize;
> -
> mutex_lock(&iommu->lock);
> - pgsize = 1 << __ffs(iommu->pgsize_bitmap);
> if (!iommu->dirty_page_tracking) {
> - ret = vfio_dma_bitmap_alloc_all(iommu, pgsize);
> + ret = vfio_dma_bitmap_alloc_all(iommu);
> if (!ret)
> iommu->dirty_page_tracking = true;
> }

This just moves the same calculation from one place to another, what's
the point? Thanks,

Alex