| Message ID | 20210107044401.19828-4-zhukeqian1@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | vfio/iommu_type1: Some optimizations about dirty tracking | expand |
On Thu, 7 Jan 2021 12:43:58 +0800 Keqian Zhu <zhukeqian1@huawei.com> wrote: > For now there are 3 ways to promote the pinned_page_dirty_scope > status of vfio_iommu: > > 1. Through vfio pin interface. > 2. Detach a group without pinned_dirty_scope. > 3. Attach a group with pinned_dirty_scope. > > For point 3, the only chance to promote the pinned_page_dirty_scope > status is when vfio_iommu is newly created. As we can safely set > empty vfio_iommu to be at pinned status, then the point 3 can be > removed to reduce operations. > > Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> > --- > drivers/vfio/vfio_iommu_type1.c | 4 +--- > 1 file changed, 1 insertion(+), 3 deletions(-) > > diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c > index 110ada24ee91..b596c482487b 100644 > --- a/drivers/vfio/vfio_iommu_type1.c > +++ b/drivers/vfio/vfio_iommu_type1.c > @@ -2045,11 +2045,8 @@ static int vfio_iommu_type1_attach_group(void *iommu_data, > * Non-iommu backed group cannot dirty memory directly, > * it can only use interfaces that provide dirty > * tracking. > - * The iommu scope can only be promoted with the > - * addition of a dirty tracking group. > */ > group->pinned_page_dirty_scope = true; > - promote_pinned_page_dirty_scope(iommu); > mutex_unlock(&iommu->lock); > > return 0; > @@ -2436,6 +2433,7 @@ static void *vfio_iommu_type1_open(unsigned long arg) > INIT_LIST_HEAD(&iommu->iova_list); > iommu->dma_list = RB_ROOT; > iommu->dma_avail = dma_entry_limit; > + iommu->pinned_page_dirty_scope = true; > mutex_init(&iommu->lock); > BLOCKING_INIT_NOTIFIER_HEAD(&iommu->notifier); > This would be resolved automatically if we used the counter approach I mentioned on the previous patch, adding a pinned-page scope group simply wouldn't increment the iommu counter, which would initially be zero indicating no "all-dma" groups. Thanks, Alex
On 2021/1/16 7:30, Alex Williamson wrote: > On Thu, 7 Jan 2021 12:43:58 +0800 > Keqian Zhu <zhukeqian1@huawei.com> wrote: > >> For now there are 3 ways to promote the pinned_page_dirty_scope >> status of vfio_iommu: >> >> 1. Through vfio pin interface. >> 2. Detach a group without pinned_dirty_scope. >> 3. Attach a group with pinned_dirty_scope. >> >> For point 3, the only chance to promote the pinned_page_dirty_scope >> status is when vfio_iommu is newly created. As we can safely set >> empty vfio_iommu to be at pinned status, then the point 3 can be >> removed to reduce operations. >> >> Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> >> --- >> drivers/vfio/vfio_iommu_type1.c | 4 +--- >> 1 file changed, 1 insertion(+), 3 deletions(-) >> >> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c >> index 110ada24ee91..b596c482487b 100644 >> --- a/drivers/vfio/vfio_iommu_type1.c >> +++ b/drivers/vfio/vfio_iommu_type1.c >> @@ -2045,11 +2045,8 @@ static int vfio_iommu_type1_attach_group(void *iommu_data, >> * Non-iommu backed group cannot dirty memory directly, >> * it can only use interfaces that provide dirty >> * tracking. >> - * The iommu scope can only be promoted with the >> - * addition of a dirty tracking group. >> */ >> group->pinned_page_dirty_scope = true; >> - promote_pinned_page_dirty_scope(iommu); >> mutex_unlock(&iommu->lock); >> >> return 0; >> @@ -2436,6 +2433,7 @@ static void *vfio_iommu_type1_open(unsigned long arg) >> INIT_LIST_HEAD(&iommu->iova_list); >> iommu->dma_list = RB_ROOT; >> iommu->dma_avail = dma_entry_limit; >> + iommu->pinned_page_dirty_scope = true; >> mutex_init(&iommu->lock); >> BLOCKING_INIT_NOTIFIER_HEAD(&iommu->notifier); >> > > This would be resolved automatically if we used the counter approach I > mentioned on the previous patch, adding a pinned-page scope group simply > wouldn't increment the iommu counter, which would initially be zero > indicating no "all-dma" groups. Thanks, > Sure, I will do that, thanks. Keqian. > Alex > > . >
diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c index 110ada24ee91..b596c482487b 100644 --- a/drivers/vfio/vfio_iommu_type1.c +++ b/drivers/vfio/vfio_iommu_type1.c @@ -2045,11 +2045,8 @@ static int vfio_iommu_type1_attach_group(void *iommu_data, * Non-iommu backed group cannot dirty memory directly, * it can only use interfaces that provide dirty * tracking. - * The iommu scope can only be promoted with the - * addition of a dirty tracking group. */ group->pinned_page_dirty_scope = true; - promote_pinned_page_dirty_scope(iommu); mutex_unlock(&iommu->lock); return 0; @@ -2436,6 +2433,7 @@ static void *vfio_iommu_type1_open(unsigned long arg) INIT_LIST_HEAD(&iommu->iova_list); iommu->dma_list = RB_ROOT; iommu->dma_avail = dma_entry_limit; + iommu->pinned_page_dirty_scope = true; mutex_init(&iommu->lock); BLOCKING_INIT_NOTIFIER_HEAD(&iommu->notifier);
For now there are 3 ways to promote the pinned_page_dirty_scope status of vfio_iommu: 1. Through vfio pin interface. 2. Detach a group without pinned_dirty_scope. 3. Attach a group with pinned_dirty_scope. For point 3, the only chance to promote the pinned_page_dirty_scope status is when vfio_iommu is newly created. As we can safely set empty vfio_iommu to be at pinned status, then the point 3 can be removed to reduce operations. Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com> --- drivers/vfio/vfio_iommu_type1.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)