[PATCH 0/2] page hinting add passthrough support

[PATCH 0/2] page hinting add passthrough support

[weiqi]

From: wei qi <weiqi4@huawei.com>


I just implemented dynamically updating the iommu table to support pass-through,
It seen to work fine.

Test:
start a 4G vm with 2M hugetlb and ixgbevf passthrough, 
    GuestOS: linux-5.2.6 + 	(mm / virtio: Provide support for free page reporting)
    HostOS: 5.5-rc4
    Host: Intel(R) Xeon(R) Gold 6161 CPU @ 2.20GHz 

after enable page hinting, free pages at GuestOS can be free at host. 


before,
 # cat /sys/devices/system/node/node*/hugepages/hugepages-2048kB/free_hugepages 
 5620
 5620
after start VM,
 # numastat -c qemu

 Per-node process memory usage (in MBs)
 PID              Node 0 Node 1 Total
 ---------------  ------ ------ -----
 24463 (qemu_hotr      6      6    12
 24479 (qemu_tls_      0      8     8
 70718 (qemu-syst     58    539   597
 ---------------  ------ ------ -----
 Total                64    553   616
 # cat /sys/devices/system/node/node*/hugepages/hugepages-2048kB/free_hugepages 
 5595
 5366

the modify at qemu,
 +int kvm_discard_range(struct kvm_discard_msg discard_msg)  
 +{
 +    return kvm_vm_ioctl(kvm_state, KVM_DISCARD_RANGE, &discard_msg);
 +}

 static void virtio_balloon_handle_report(VirtIODevice *vdev, VirtQueue *vq)
 {
            ..................
 +           discard_msg.in_addr = elem->in_addr[i];
 +           discard_msg.iov_len = elem->in_sg[i].iov_len;

            ram_block_discard_range(rb, ram_offset, size);
 +           kvm_discard_range(discard_msg);

then, further test network bandwidth, performance seem ok.
 
Is there any hidden problem in this implementation?
And, is there plan to support pass-throughyour?

wei qi (2):
  vfio: add mmap/munmap API for page hinting
  KVM: add support for page hinting

 arch/x86/kvm/mmu/mmu.c          |  79 ++++++++++++++++++++
 arch/x86/kvm/x86.c              |  96 ++++++++++++++++++++++++
 drivers/vfio/vfio.c             | 109 ++++++++++++++++++++++++++++
 drivers/vfio/vfio_iommu_type1.c | 157 +++++++++++++++++++++++++++++++++++++++-
 include/linux/kvm_host.h        |  41 +++++++++++
 include/linux/vfio.h            |  17 ++++-
 include/uapi/linux/kvm.h        |   7 ++
 virt/kvm/vfio.c                 |  11 ---
 8 files changed, 503 insertions(+), 14 deletions(-)

-- 
1.8.3.1

[PATCH 1/2] vfio: add mmap/munmap API for page hinting

[weiqi]

From: wei qi <weiqi4@huawei.com>

add mmap/munmap API for page hinting.

Signed-off-by: wei qi <weiqi4@huawei.com>
---
 drivers/vfio/vfio.c             | 109 ++++++++++++++++++++++++++++
 drivers/vfio/vfio_iommu_type1.c | 157 +++++++++++++++++++++++++++++++++++++++-
 include/linux/vfio.h            |  17 ++++-
 3 files changed, 280 insertions(+), 3 deletions(-)

diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
index c848262..c7e9103 100644
--- a/drivers/vfio/vfio.c
+++ b/drivers/vfio/vfio.c
@@ -1866,6 +1866,115 @@ int vfio_set_irqs_validate_and_prepare(struct vfio_irq_set *hdr, int num_irqs,
 }
 EXPORT_SYMBOL(vfio_set_irqs_validate_and_prepare);
 
+int vfio_mmap_pages(struct device *dev, unsigned long user_pfn,
+			unsigned long page_size, int prot,
+			unsigned long pfn)
+{
+	struct vfio_container *container;
+	struct vfio_group *group;
+	struct vfio_iommu_driver *driver;
+	int ret;
+
+	if (!dev || !user_pfn || !page_size)
+		return -EINVAL;
+
+	group = vfio_group_get_from_dev(dev);
+	if (!group)
+		return -ENODEV;
+
+	ret = vfio_group_add_container_user(group);
+	if (ret)
+		goto err_pin_pages;
+
+	container = group->container;
+	driver = container->iommu_driver;
+	if (likely(driver && driver->ops->mmap_pages))
+		ret = driver->ops->mmap_pages(container->iommu_data, user_pfn,
+					page_size, prot, pfn);
+	else
+		ret = -ENOTTY;
+
+	vfio_group_try_dissolve_container(group);
+
+err_pin_pages:
+	vfio_group_put(group);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(vfio_mmap_pages);
+
+int vfio_munmap_pages(struct device *dev, unsigned long user_pfn,
+			unsigned long page_size)
+{
+	struct vfio_container *container;
+	struct vfio_group *group;
+	struct vfio_iommu_driver *driver;
+	int ret;
+
+	if (!dev || !user_pfn || !page_size)
+		return -EINVAL;
+
+	group = vfio_group_get_from_dev(dev);
+	if (!group)
+		return -ENODEV;
+
+	ret = vfio_group_add_container_user(group);
+	if (ret)
+		goto err_pin_pages;
+
+	container = group->container;
+	driver = container->iommu_driver;
+	if (likely(driver && driver->ops->munmap_pages))
+		ret = driver->ops->munmap_pages(container->iommu_data, user_pfn,
+						page_size);
+	else
+		ret = -ENOTTY;
+
+	vfio_group_try_dissolve_container(group);
+
+err_pin_pages:
+	vfio_group_put(group);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(vfio_munmap_pages);
+
+int vfio_dma_find(struct device *dev, unsigned long user_pfn, int npage,
+		unsigned long *phys_pfn)
+{
+	struct vfio_container *container;
+	struct vfio_group *group;
+	struct vfio_iommu_driver *driver;
+	int ret;
+
+	if (!dev || !user_pfn || !npage || !phys_pfn)
+		return -EINVAL;
+
+	if (npage > VFIO_PIN_PAGES_MAX_ENTRIES)
+		return -E2BIG;
+
+	group = vfio_group_get_from_dev(dev);
+	if (!group)
+		return -ENODEV;
+
+	ret = vfio_group_add_container_user(group);
+	if (ret)
+		goto err_pin_pages;
+
+	container = group->container;
+	driver = container->iommu_driver;
+	if (driver && driver->ops->dma_find)
+		ret = driver->ops->dma_find(container->iommu_data, user_pfn,
+					npage, phys_pfn);
+	else
+		ret = -ENOTTY;
+
+	vfio_group_try_dissolve_container(group);
+
+err_pin_pages:
+	vfio_group_put(group);
+	return ret;
+}
+EXPORT_SYMBOL(vfio_dma_find);
+
 /*
  * Pin a set of guest PFNs and return their associated host PFNs for local
  * domain only.
diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index 2ada8e6..df115dc 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -414,7 +414,7 @@ static long vfio_pin_pages_remote(struct vfio_dma *dma, unsigned long vaddr,
 		goto out;
 
 	/* Lock all the consecutive pages from pfn_base */
-	for (vaddr += PAGE_SIZE, iova += PAGE_SIZE; pinned < npage;
+	for (vaddr += PAGE_SIZE, iova += PAGE_SIZE; (pinned < npage && pinned < 512);
 	     pinned++, vaddr += PAGE_SIZE, iova += PAGE_SIZE) {
 		ret = vaddr_get_pfn(current->mm, vaddr, dma->prot, &pfn);
 		if (ret)
@@ -768,7 +768,7 @@ static long vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
 		phys_addr_t phys, next;
 
 		phys = iommu_iova_to_phys(domain->domain, iova);
-		if (WARN_ON(!phys)) {
+		if (!phys) {
 			iova += PAGE_SIZE;
 			continue;
 		}
@@ -1154,6 +1154,156 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu,
 	return ret;
 }
 
+static int vfio_iommu_type1_munmap_pages(void *iommu_data,
+					unsigned long user_pfn,
+					unsigned long page_size)
+{
+	struct vfio_iommu *iommu = iommu_data;
+	struct vfio_domain *domain;
+	struct vfio_dma *dma;
+	dma_addr_t iova = user_pfn  << PAGE_SHIFT;
+	int ret = 0;
+	phys_addr_t phys;
+	size_t unmapped;
+	long unlocked = 0;
+
+	if (!iommu || !user_pfn || !page_size)
+		return -EINVAL;
+
+	/* Supported for v2 version only */
+	if (!iommu->v2)
+		return -EACCES;
+
+	mutex_lock(&iommu->lock);
+	dma = vfio_find_dma(iommu, iova, page_size);
+	if (!dma) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	domain = list_first_entry(&iommu->domain_list,
+			struct vfio_domain, next);
+	phys = iommu_iova_to_phys(domain->domain, iova);
+	if (!phys) {
+		goto out_unlock;
+	} else {
+		unmapped = iommu_unmap(domain->domain, iova, page_size);
+		unlocked = vfio_unpin_pages_remote(dma, iova,
+					phys >> PAGE_SHIFT,
+					unmapped >> PAGE_SHIFT, true);
+	}
+
+out_unlock:
+	mutex_unlock(&iommu->lock);
+	return ret;
+}
+
+static int vfio_iommu_type1_mmap_pages(void *iommu_data,
+				unsigned long user_pfn,
+				unsigned long page_size, int prot,
+				unsigned long pfn)
+{
+	struct vfio_iommu *iommu = iommu_data;
+	struct vfio_domain *domain;
+	struct vfio_dma *dma;
+	dma_addr_t iova = user_pfn  << PAGE_SHIFT;
+	int ret = 0;
+	size_t unmapped;
+	phys_addr_t phys;
+	long unlocked = 0;
+
+	if (!iommu || !user_pfn || !page_size || !pfn)
+		return -EINVAL;
+
+	/* Supported for v2 version only */
+	if (!iommu->v2)
+		return -EACCES;
+
+	mutex_lock(&iommu->lock);
+
+	if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu)) {
+		ret = -EACCES;
+		goto out_unlock;
+	}
+
+	dma = vfio_find_dma(iommu, iova, page_size);
+	if (!dma) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	domain = list_first_entry(&iommu->domain_list,
+		struct vfio_domain, next);
+
+	phys = iommu_iova_to_phys(domain->domain, iova);
+	if (phys) {
+		unmapped = iommu_unmap(domain->domain, iova, page_size);
+		unlocked = vfio_unpin_pages_remote(dma, iova,
+					phys >> PAGE_SHIFT,
+					unmapped >> PAGE_SHIFT, false);
+	}
+
+	ret = vfio_iommu_map(iommu, iova, pfn, page_size >> PAGE_SHIFT, prot);
+	if (ret) {
+		pr_warn("%s: gfn: %lx, pfn: %lx, npages：%lu\n", __func__,
+			user_pfn, pfn, page_size >> PAGE_SHIFT);
+	}
+
+out_unlock:
+	mutex_unlock(&iommu->lock);
+	return ret;
+}
+
+u64 vfio_iommu_iova_to_phys(struct vfio_iommu *iommu, dma_addr_t iova)
+{
+	struct vfio_domain *d;
+	u64 phys;
+
+	list_for_each_entry(d, &iommu->domain_list, next) {
+		phys = iommu_iova_to_phys(d->domain, iova);
+		if (phys)
+			return phys;
+	}
+	return 0;
+}
+
+static int vfio_iommu_type1_dma_find(void *iommu_data,
+					unsigned long user_pfn,
+					int npage, unsigned long *phys_pfn)
+{
+	struct vfio_iommu *iommu = iommu_data;
+	int i = 0;
+	struct vfio_dma *dma;
+	u64 phys;
+	dma_addr_t iova;
+
+	if (!iommu || !user_pfn)
+		return -EINVAL;
+
+	/* Supported for v2 version only */
+	if (!iommu->v2)
+		return -EACCES;
+
+	 mutex_lock(&iommu->lock);
+
+	iova = user_pfn << PAGE_SHIFT;
+	dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
+	if (!dma)
+		goto unpin_exit;
+
+	if (((user_pfn + npage) << PAGE_SHIFT) <= (dma->iova + dma->size))
+		i = npage;
+	else
+		goto unpin_exit;
+
+	phys = vfio_iommu_iova_to_phys(iommu, iova);
+	*phys_pfn = phys >> PAGE_SHIFT;
+
+unpin_exit:
+	mutex_unlock(&iommu->lock);
+	return i;
+}
+
 static int vfio_bus_type(struct device *dev, void *data)
 {
 	struct bus_type **bus = data;
@@ -2336,6 +2486,9 @@ static int vfio_iommu_type1_unregister_notifier(void *iommu_data,
 	.detach_group		= vfio_iommu_type1_detach_group,
 	.pin_pages		= vfio_iommu_type1_pin_pages,
 	.unpin_pages		= vfio_iommu_type1_unpin_pages,
+	.mmap_pages             = vfio_iommu_type1_mmap_pages,
+	.munmap_pages           = vfio_iommu_type1_munmap_pages,
+	.dma_find		= vfio_iommu_type1_dma_find,
 	.register_notifier	= vfio_iommu_type1_register_notifier,
 	.unregister_notifier	= vfio_iommu_type1_unregister_notifier,
 };
diff --git a/include/linux/vfio.h b/include/linux/vfio.h
index e42a711..d7df495 100644
--- a/include/linux/vfio.h
+++ b/include/linux/vfio.h
@@ -77,6 +77,15 @@ struct vfio_iommu_driver_ops {
 				     unsigned long *phys_pfn);
 	int		(*unpin_pages)(void *iommu_data,
 				       unsigned long *user_pfn, int npage);
+	int		(*mmap_pages)(void *iommu_data,
+					unsigned long user_pfn,
+					unsigned long page_size,
+					int prot, unsigned long pfn);
+	int		(*munmap_pages)(void *iommu_data,
+					unsigned long user_pfn,
+					unsigned long page_size);
+	int		(*dma_find)(void *iommu_data, unsigned long user_pfn,
+					int npage, unsigned long *phys_pfn);
 	int		(*register_notifier)(void *iommu_data,
 					     unsigned long *events,
 					     struct notifier_block *nb);
@@ -106,7 +115,13 @@ extern int vfio_pin_pages(struct device *dev, unsigned long *user_pfn,
 			  int npage, int prot, unsigned long *phys_pfn);
 extern int vfio_unpin_pages(struct device *dev, unsigned long *user_pfn,
 			    int npage);
-
+extern int vfio_dma_find(struct device *dev, unsigned long user_pfn, int npage,
+			unsigned long *phys_pfn);
+extern int vfio_mmap_pages(struct device *dev, unsigned long user_pfn,
+			unsigned long page_size, int prot,
+			unsigned long pfn);
+extern int vfio_munmap_pages(struct device *dev, unsigned long user_pfn,
+			unsigned long page_size);
 /* each type has independent events */
 enum vfio_notify_type {
 	VFIO_IOMMU_NOTIFY = 0,
-- 
1.8.3.1

Re: [PATCH 1/2] vfio: add mmap/munmap API for page hinting

[Alex Williamson]

On Tue, 7 Jan 2020 22:46:38 +0800
weiqi <weiqi4@huawei.com> wrote:

> From: wei qi <weiqi4@huawei.com>
> 
> add mmap/munmap API for page hinting.

AIUI, this is arbitrarily chunking IOMMU mappings into 512 pages (what
happens with 1G pages?) and creating a back channel for KVM to map and
unmap ranges that the user has mapped (why's it called "mmap"?).  Can't
we do this via the existing user API rather than directed via another
module?  For example, userspace can choose to map chunks of IOVA space
in whatever granularity they choose.  Clearly they can then unmap and
re-map chunks from those previous mappings.  Why can't KVM tell
userspace how and when to do this?  I'm really not in favor of back
channel paths like this, especially to unmap what a user has told us to
map.  Thanks,

Alex

> Signed-off-by: wei qi <weiqi4@huawei.com>
> ---
>  drivers/vfio/vfio.c             | 109 ++++++++++++++++++++++++++++
>  drivers/vfio/vfio_iommu_type1.c | 157 +++++++++++++++++++++++++++++++++++++++-
>  include/linux/vfio.h            |  17 ++++-
>  3 files changed, 280 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index c848262..c7e9103 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -1866,6 +1866,115 @@ int vfio_set_irqs_validate_and_prepare(struct vfio_irq_set *hdr, int num_irqs,
>  }
>  EXPORT_SYMBOL(vfio_set_irqs_validate_and_prepare);
>  
> +int vfio_mmap_pages(struct device *dev, unsigned long user_pfn,
> +			unsigned long page_size, int prot,
> +			unsigned long pfn)
> +{
> +	struct vfio_container *container;
> +	struct vfio_group *group;
> +	struct vfio_iommu_driver *driver;
> +	int ret;
> +
> +	if (!dev || !user_pfn || !page_size)
> +		return -EINVAL;
> +
> +	group = vfio_group_get_from_dev(dev);
> +	if (!group)
> +		return -ENODEV;
> +
> +	ret = vfio_group_add_container_user(group);
> +	if (ret)
> +		goto err_pin_pages;
> +
> +	container = group->container;
> +	driver = container->iommu_driver;
> +	if (likely(driver && driver->ops->mmap_pages))
> +		ret = driver->ops->mmap_pages(container->iommu_data, user_pfn,
> +					page_size, prot, pfn);
> +	else
> +		ret = -ENOTTY;
> +
> +	vfio_group_try_dissolve_container(group);
> +
> +err_pin_pages:
> +	vfio_group_put(group);
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(vfio_mmap_pages);
> +
> +int vfio_munmap_pages(struct device *dev, unsigned long user_pfn,
> +			unsigned long page_size)
> +{
> +	struct vfio_container *container;
> +	struct vfio_group *group;
> +	struct vfio_iommu_driver *driver;
> +	int ret;
> +
> +	if (!dev || !user_pfn || !page_size)
> +		return -EINVAL;
> +
> +	group = vfio_group_get_from_dev(dev);
> +	if (!group)
> +		return -ENODEV;
> +
> +	ret = vfio_group_add_container_user(group);
> +	if (ret)
> +		goto err_pin_pages;
> +
> +	container = group->container;
> +	driver = container->iommu_driver;
> +	if (likely(driver && driver->ops->munmap_pages))
> +		ret = driver->ops->munmap_pages(container->iommu_data, user_pfn,
> +						page_size);
> +	else
> +		ret = -ENOTTY;
> +
> +	vfio_group_try_dissolve_container(group);
> +
> +err_pin_pages:
> +	vfio_group_put(group);
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(vfio_munmap_pages);
> +
> +int vfio_dma_find(struct device *dev, unsigned long user_pfn, int npage,
> +		unsigned long *phys_pfn)
> +{
> +	struct vfio_container *container;
> +	struct vfio_group *group;
> +	struct vfio_iommu_driver *driver;
> +	int ret;
> +
> +	if (!dev || !user_pfn || !npage || !phys_pfn)
> +		return -EINVAL;
> +
> +	if (npage > VFIO_PIN_PAGES_MAX_ENTRIES)
> +		return -E2BIG;
> +
> +	group = vfio_group_get_from_dev(dev);
> +	if (!group)
> +		return -ENODEV;
> +
> +	ret = vfio_group_add_container_user(group);
> +	if (ret)
> +		goto err_pin_pages;
> +
> +	container = group->container;
> +	driver = container->iommu_driver;
> +	if (driver && driver->ops->dma_find)
> +		ret = driver->ops->dma_find(container->iommu_data, user_pfn,
> +					npage, phys_pfn);
> +	else
> +		ret = -ENOTTY;
> +
> +	vfio_group_try_dissolve_container(group);
> +
> +err_pin_pages:
> +	vfio_group_put(group);
> +	return ret;
> +}
> +EXPORT_SYMBOL(vfio_dma_find);
> +
>  /*
>   * Pin a set of guest PFNs and return their associated host PFNs for local
>   * domain only.
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index 2ada8e6..df115dc 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -414,7 +414,7 @@ static long vfio_pin_pages_remote(struct vfio_dma *dma, unsigned long vaddr,
>  		goto out;
>  
>  	/* Lock all the consecutive pages from pfn_base */
> -	for (vaddr += PAGE_SIZE, iova += PAGE_SIZE; pinned < npage;
> +	for (vaddr += PAGE_SIZE, iova += PAGE_SIZE; (pinned < npage && pinned < 512);
>  	     pinned++, vaddr += PAGE_SIZE, iova += PAGE_SIZE) {
>  		ret = vaddr_get_pfn(current->mm, vaddr, dma->prot, &pfn);
>  		if (ret)
> @@ -768,7 +768,7 @@ static long vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
>  		phys_addr_t phys, next;
>  
>  		phys = iommu_iova_to_phys(domain->domain, iova);
> -		if (WARN_ON(!phys)) {
> +		if (!phys) {
>  			iova += PAGE_SIZE;
>  			continue;
>  		}
> @@ -1154,6 +1154,156 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu,
>  	return ret;
>  }
>  
> +static int vfio_iommu_type1_munmap_pages(void *iommu_data,
> +					unsigned long user_pfn,
> +					unsigned long page_size)
> +{
> +	struct vfio_iommu *iommu = iommu_data;
> +	struct vfio_domain *domain;
> +	struct vfio_dma *dma;
> +	dma_addr_t iova = user_pfn  << PAGE_SHIFT;
> +	int ret = 0;
> +	phys_addr_t phys;
> +	size_t unmapped;
> +	long unlocked = 0;
> +
> +	if (!iommu || !user_pfn || !page_size)
> +		return -EINVAL;
> +
> +	/* Supported for v2 version only */
> +	if (!iommu->v2)
> +		return -EACCES;
> +
> +	mutex_lock(&iommu->lock);
> +	dma = vfio_find_dma(iommu, iova, page_size);
> +	if (!dma) {
> +		ret = -EINVAL;
> +		goto out_unlock;
> +	}
> +
> +	domain = list_first_entry(&iommu->domain_list,
> +			struct vfio_domain, next);
> +	phys = iommu_iova_to_phys(domain->domain, iova);
> +	if (!phys) {
> +		goto out_unlock;
> +	} else {
> +		unmapped = iommu_unmap(domain->domain, iova, page_size);
> +		unlocked = vfio_unpin_pages_remote(dma, iova,
> +					phys >> PAGE_SHIFT,
> +					unmapped >> PAGE_SHIFT, true);
> +	}
> +
> +out_unlock:
> +	mutex_unlock(&iommu->lock);
> +	return ret;
> +}
> +
> +static int vfio_iommu_type1_mmap_pages(void *iommu_data,
> +				unsigned long user_pfn,
> +				unsigned long page_size, int prot,
> +				unsigned long pfn)
> +{
> +	struct vfio_iommu *iommu = iommu_data;
> +	struct vfio_domain *domain;
> +	struct vfio_dma *dma;
> +	dma_addr_t iova = user_pfn  << PAGE_SHIFT;
> +	int ret = 0;
> +	size_t unmapped;
> +	phys_addr_t phys;
> +	long unlocked = 0;
> +
> +	if (!iommu || !user_pfn || !page_size || !pfn)
> +		return -EINVAL;
> +
> +	/* Supported for v2 version only */
> +	if (!iommu->v2)
> +		return -EACCES;
> +
> +	mutex_lock(&iommu->lock);
> +
> +	if (!IS_IOMMU_CAP_DOMAIN_IN_CONTAINER(iommu)) {
> +		ret = -EACCES;
> +		goto out_unlock;
> +	}
> +
> +	dma = vfio_find_dma(iommu, iova, page_size);
> +	if (!dma) {
> +		ret = -EINVAL;
> +		goto out_unlock;
> +	}
> +
> +	domain = list_first_entry(&iommu->domain_list,
> +		struct vfio_domain, next);
> +
> +	phys = iommu_iova_to_phys(domain->domain, iova);
> +	if (phys) {
> +		unmapped = iommu_unmap(domain->domain, iova, page_size);
> +		unlocked = vfio_unpin_pages_remote(dma, iova,
> +					phys >> PAGE_SHIFT,
> +					unmapped >> PAGE_SHIFT, false);
> +	}
> +
> +	ret = vfio_iommu_map(iommu, iova, pfn, page_size >> PAGE_SHIFT, prot);
> +	if (ret) {
> +		pr_warn("%s: gfn: %lx, pfn: %lx, npages：%lu\n", __func__,
> +			user_pfn, pfn, page_size >> PAGE_SHIFT);
> +	}
> +
> +out_unlock:
> +	mutex_unlock(&iommu->lock);
> +	return ret;
> +}
> +
> +u64 vfio_iommu_iova_to_phys(struct vfio_iommu *iommu, dma_addr_t iova)
> +{
> +	struct vfio_domain *d;
> +	u64 phys;
> +
> +	list_for_each_entry(d, &iommu->domain_list, next) {
> +		phys = iommu_iova_to_phys(d->domain, iova);
> +		if (phys)
> +			return phys;
> +	}
> +	return 0;
> +}
> +
> +static int vfio_iommu_type1_dma_find(void *iommu_data,
> +					unsigned long user_pfn,
> +					int npage, unsigned long *phys_pfn)
> +{
> +	struct vfio_iommu *iommu = iommu_data;
> +	int i = 0;
> +	struct vfio_dma *dma;
> +	u64 phys;
> +	dma_addr_t iova;
> +
> +	if (!iommu || !user_pfn)
> +		return -EINVAL;
> +
> +	/* Supported for v2 version only */
> +	if (!iommu->v2)
> +		return -EACCES;
> +
> +	 mutex_lock(&iommu->lock);
> +
> +	iova = user_pfn << PAGE_SHIFT;
> +	dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
> +	if (!dma)
> +		goto unpin_exit;
> +
> +	if (((user_pfn + npage) << PAGE_SHIFT) <= (dma->iova + dma->size))
> +		i = npage;
> +	else
> +		goto unpin_exit;
> +
> +	phys = vfio_iommu_iova_to_phys(iommu, iova);
> +	*phys_pfn = phys >> PAGE_SHIFT;
> +
> +unpin_exit:
> +	mutex_unlock(&iommu->lock);
> +	return i;
> +}
> +
>  static int vfio_bus_type(struct device *dev, void *data)
>  {
>  	struct bus_type **bus = data;
> @@ -2336,6 +2486,9 @@ static int vfio_iommu_type1_unregister_notifier(void *iommu_data,
>  	.detach_group		= vfio_iommu_type1_detach_group,
>  	.pin_pages		= vfio_iommu_type1_pin_pages,
>  	.unpin_pages		= vfio_iommu_type1_unpin_pages,
> +	.mmap_pages             = vfio_iommu_type1_mmap_pages,
> +	.munmap_pages           = vfio_iommu_type1_munmap_pages,
> +	.dma_find		= vfio_iommu_type1_dma_find,
>  	.register_notifier	= vfio_iommu_type1_register_notifier,
>  	.unregister_notifier	= vfio_iommu_type1_unregister_notifier,
>  };
> diff --git a/include/linux/vfio.h b/include/linux/vfio.h
> index e42a711..d7df495 100644
> --- a/include/linux/vfio.h
> +++ b/include/linux/vfio.h
> @@ -77,6 +77,15 @@ struct vfio_iommu_driver_ops {
>  				     unsigned long *phys_pfn);
>  	int		(*unpin_pages)(void *iommu_data,
>  				       unsigned long *user_pfn, int npage);
> +	int		(*mmap_pages)(void *iommu_data,
> +					unsigned long user_pfn,
> +					unsigned long page_size,
> +					int prot, unsigned long pfn);
> +	int		(*munmap_pages)(void *iommu_data,
> +					unsigned long user_pfn,
> +					unsigned long page_size);
> +	int		(*dma_find)(void *iommu_data, unsigned long user_pfn,
> +					int npage, unsigned long *phys_pfn);
>  	int		(*register_notifier)(void *iommu_data,
>  					     unsigned long *events,
>  					     struct notifier_block *nb);
> @@ -106,7 +115,13 @@ extern int vfio_pin_pages(struct device *dev, unsigned long *user_pfn,
>  			  int npage, int prot, unsigned long *phys_pfn);
>  extern int vfio_unpin_pages(struct device *dev, unsigned long *user_pfn,
>  			    int npage);
> -
> +extern int vfio_dma_find(struct device *dev, unsigned long user_pfn, int npage,
> +			unsigned long *phys_pfn);
> +extern int vfio_mmap_pages(struct device *dev, unsigned long user_pfn,
> +			unsigned long page_size, int prot,
> +			unsigned long pfn);
> +extern int vfio_munmap_pages(struct device *dev, unsigned long user_pfn,
> +			unsigned long page_size);
>  /* each type has independent events */
>  enum vfio_notify_type {
>  	VFIO_IOMMU_NOTIFY = 0,

Re: [PATCH 1/2] vfio: add mmap/munmap API for page hinting

[kbuild test robot]

Hi weiqi,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on kvm/linux-next]
[also build test WARNING on vfio/next vhost/linux-next v5.5-rc5 next-20200109]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/weiqi/page-hinting-add-passthrough-support/20200108-152941
base:   https://git.kernel.org/pub/scm/virt/kvm/kvm.git linux-next
reproduce:
        # apt-get install sparse
        # sparse version: v0.6.1-129-g341daf20-dirty
        make ARCH=x86_64 allmodconfig
        make C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__'

If you fix the issue, kindly add following tag
Reported-by: kbuild test robot <lkp@intel.com>


sparse warnings: (new ones prefixed by >>)

>> drivers/vfio/vfio_iommu_type1.c:1275:5: sparse: sparse: symbol 'vfio_iommu_iova_to_phys' was not declared. Should it be static?

Please review and possibly fold the followup patch.

---
0-DAY kernel test infrastructure                 Open Source Technology Center
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org Intel Corporation

[RFC PATCH] vfio: vfio_iommu_iova_to_phys() can be static

[kbuild test robot]

Fixes: 3b764b3397df ("vfio: add mmap/munmap API for page hinting")
Signed-off-by: kbuild test robot <lkp@intel.com>
---
 vfio_iommu_type1.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
index 3939f9573f74f..f25c107e2c709 100644
--- a/drivers/vfio/vfio_iommu_type1.c
+++ b/drivers/vfio/vfio_iommu_type1.c
@@ -1272,7 +1272,7 @@ static int vfio_iommu_type1_mmap_pages(void *iommu_data,
 	return ret;
 }
 
-u64 vfio_iommu_iova_to_phys(struct vfio_iommu *iommu, dma_addr_t iova)
+static u64 vfio_iommu_iova_to_phys(struct vfio_iommu *iommu, dma_addr_t iova)
 {
 	struct vfio_domain *d;
 	u64 phys;

[PATCH 2/2] KVM: add support for page hinting

[weiqi]

From: wei qi <weiqi4@huawei.com>

add support for page hinting.

Signed-off-by: wei qi <weiqi4@huawei.com>
---
 arch/x86/kvm/mmu/mmu.c   | 79 +++++++++++++++++++++++++++++++++++++++
 arch/x86/kvm/x86.c       | 96 ++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/kvm_host.h | 41 +++++++++++++++++++++
 include/uapi/linux/kvm.h |  7 ++++
 virt/kvm/vfio.c          | 11 ------
 5 files changed, 223 insertions(+), 11 deletions(-)

diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
index 6f92b40..0cf2584 100644
--- a/arch/x86/kvm/mmu/mmu.c
+++ b/arch/x86/kvm/mmu/mmu.c
@@ -4259,6 +4259,71 @@ int kvm_handle_page_fault(struct kvm_vcpu *vcpu, u64 error_code,
 	return kvm_mtrr_check_gfn_range_consistency(vcpu, gfn, page_num);
 }
 
+#include <linux/vfio.h>
+static void kvm_vfio_mmap_range(struct kvm_vcpu *vcpu, struct kvm_device *tmp,
+				gfn_t gfn, kvm_pfn_t pfn)
+{
+	struct kvm_vfio *kv = tmp->private;
+	struct kvm_vfio_group *kvg;
+
+	list_for_each_entry(kvg, &kv->group_list, node) {
+		struct vfio_group *group = kvg->vfio_group;
+		struct vfio_device *it, *device = NULL;
+
+		list_for_each_entry(it, &group->device_list, group_next) {
+			int flags;
+			unsigned long page_size;
+			struct kvm_memory_slot *memslot;
+			int size;
+			unsigned long old_pfn = 0;
+			gfn_t gfn_base;
+			kvm_pfn_t pfn_base;
+
+			device = it;
+			memslot = kvm_vcpu_gfn_to_memslot(vcpu, gfn);
+			page_size = kvm_host_page_size(vcpu->kvm, gfn);
+
+			/* only discard free pages, just check 2M hugetlb */
+			if (page_size >> PAGE_SHIFT != 512)
+				return;
+
+			gfn_base = ((gfn << PAGE_SHIFT) & (~(page_size - 1)))
+					>> PAGE_SHIFT;
+			pfn_base = ((pfn << PAGE_SHIFT) & (~(page_size - 1)))
+					>> PAGE_SHIFT;
+
+			while ((gfn << PAGE_SHIFT) & (page_size - 1))
+				page_size >>= 1;
+
+			while (__gfn_to_hva_memslot(memslot, gfn) &
+						(page_size - 1))
+				page_size >>= 1;
+
+			size = vfio_dma_find(device->dev, gfn_base,
+					page_size >> PAGE_SHIFT, &old_pfn);
+			if (!size) {
+				pr_err("%s:not find dma: gfn: %llx, size: %lu.\n",
+					__func__, gfn_base,
+					page_size >> PAGE_SHIFT);
+				return;
+			}
+			if (!old_pfn)
+				pr_err("%s: not find pfn: gfn: %llx, size: %lu.\n",
+					__func__, gfn_base,
+					page_size >> PAGE_SHIFT);
+
+			if (pfn_base == old_pfn)
+				return;
+
+			flags = IOMMU_READ;
+			if (!(memslot->flags & KVM_MEM_READONLY))
+				flags |= IOMMU_WRITE;
+			vfio_mmap_pages(device->dev, gfn,
+					page_size, flags, pfn);
+		}
+	}
+}
+
 static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
 			  bool prefault)
 {
@@ -4317,6 +4382,20 @@ static int tdp_page_fault(struct kvm_vcpu *vcpu, gva_t gpa, u32 error_code,
 			 prefault, lpage_disallowed);
 out_unlock:
 	spin_unlock(&vcpu->kvm->mmu_lock);
+
+	if (!is_noslot_pfn(pfn) && gfn) {
+		struct kvm_device *tmp;
+
+		list_for_each_entry(tmp, &vcpu->kvm->devices, vm_node) {
+			if (tmp->ops && tmp->ops->name &&
+				(!strcmp(tmp->ops->name, "kvm-vfio"))) {
+				spin_lock(&vcpu->kvm->discard_lock);
+				kvm_vfio_mmap_range(vcpu, tmp, gfn, pfn);
+				spin_unlock(&vcpu->kvm->discard_lock);
+			}
+		}
+	}
+
 	kvm_release_pfn_clean(pfn);
 	return r;
 }
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index cf91713..264c65e 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -4837,6 +4837,92 @@ int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
 	return r;
 }
 
+#include <linux/vfio.h>
+static void kvm_vfio_ummap_range(struct kvm *kvm, struct kvm_device *tmp,
+				gfn_t gfn, int npages, unsigned long hva)
+{
+	struct kvm_vfio *kv = tmp->private;
+	struct kvm_vfio_group *kvg;
+
+	list_for_each_entry(kvg, &kv->group_list, node) {
+		struct vfio_group *group = kvg->vfio_group;
+		struct vfio_device *it, *device = NULL;
+
+		list_for_each_entry(it, &group->device_list, group_next) {
+			unsigned long page_size, page_size_base;
+			unsigned long addr;
+			int size;
+			unsigned long old_pfn = 0;
+			int ret = 0;
+			size_t unmapped = npages;
+			gfn_t iova_gfn = gfn;
+			unsigned long iova_hva = hva;
+
+			device = it;
+			while (unmapped) {
+				addr = gfn_to_hva(kvm, iova_gfn);
+				page_size_base = page_size =
+						kvm_host_page_size(kvm,
+						iova_gfn);
+
+				if (addr != iova_hva)
+					return;
+
+				while ((iova_gfn << PAGE_SHIFT) &
+					(page_size - 1))
+					page_size >>= 1;
+
+				while (addr & (page_size - 1))
+					page_size >>= 1;
+
+				if (page_size_base != page_size)
+					return;
+
+				size = vfio_dma_find(device->dev, iova_gfn,
+						page_size >> PAGE_SHIFT,
+						&old_pfn);
+				if (!size)
+					return;
+
+				if (!old_pfn)
+					return;
+
+				ret = vfio_munmap_pages(device->dev,
+						iova_gfn, page_size);
+				unmapped -= page_size >> PAGE_SHIFT;
+				iova_hva += page_size;
+				iova_gfn += page_size >> PAGE_SHIFT;
+			}
+		}
+	}
+}
+
+
+static int kvm_vm_ioctl_discard_range(struct kvm *kvm,
+				struct kvm_discard_msg *msg)
+{
+	gfn_t gfn, end_gfn;
+	int idx;
+	struct kvm_device *tmp;
+	unsigned long hva = msg->iov_base;
+	int npages = msg->iov_len >> PAGE_SHIFT;
+
+	gfn = gpa_to_gfn(msg->in_addr);
+	end_gfn = gpa_to_gfn(msg->in_addr + msg->iov_len);
+
+	idx = srcu_read_lock(&kvm->srcu);
+
+	list_for_each_entry(tmp, &kvm->devices, vm_node) {
+		if (tmp->ops->name && (!strcmp(tmp->ops->name, "kvm-vfio"))) {
+			spin_lock(&kvm->discard_lock);
+			kvm_vfio_ummap_range(kvm, tmp, gfn, npages, hva);
+			spin_unlock(&kvm->discard_lock);
+		}
+	}
+	srcu_read_unlock(&kvm->srcu, idx);
+	return 0;
+}
+
 long kvm_arch_vm_ioctl(struct file *filp,
 		       unsigned int ioctl, unsigned long arg)
 {
@@ -5134,6 +5220,16 @@ long kvm_arch_vm_ioctl(struct file *filp,
 	case KVM_SET_PMU_EVENT_FILTER:
 		r = kvm_vm_ioctl_set_pmu_event_filter(kvm, argp);
 		break;
+	case KVM_DISCARD_RANGE: {
+		struct kvm_discard_msg discard_msg;
+
+		r = -EFAULT;
+		if (copy_from_user(&discard_msg, argp, sizeof(discard_msg)))
+			goto out;
+
+		r = kvm_vm_ioctl_discard_range(kvm, &discard_msg);
+		break;
+	}
 	default:
 		r = -ENOTTY;
 	}
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 538c25e..6667e6b 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -442,6 +442,7 @@ struct kvm_memslots {
 
 struct kvm {
 	spinlock_t mmu_lock;
+	spinlock_t discard_lock;
 	struct mutex slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots __rcu *memslots[KVM_ADDRESS_SPACE_NUM];
@@ -502,6 +503,46 @@ struct kvm {
 	struct srcu_struct irq_srcu;
 	pid_t userspace_pid;
 };
+struct vfio_device {
+	struct kref                     kref;
+	struct device                   *dev;
+	const struct vfio_device_ops    *ops;
+	struct vfio_group               *group;
+	struct list_head                group_next;
+	void                            *device_data;
+};
+
+struct vfio_group {
+	struct kref                     kref;
+	int                             minor;
+	atomic_t                        container_users;
+	struct iommu_group              *iommu_group;
+	struct vfio_container           *container;
+	struct list_head                device_list;
+	struct mutex                    device_lock;
+	struct device                   *dev;
+	struct notifier_block           nb;
+	struct list_head                vfio_next;
+	struct list_head                container_next;
+	struct list_head                unbound_list;
+	struct mutex                    unbound_lock;
+	atomic_t                        opened;
+	wait_queue_head_t               container_q;
+	bool                            noiommu;
+	struct kvm                      *kvm;
+	struct blocking_notifier_head   notifier;
+};
+
+struct kvm_vfio_group {
+	struct list_head node;
+	struct vfio_group *vfio_group;
+};
+
+struct kvm_vfio {
+	struct list_head group_list;
+	struct mutex lock;
+	bool noncoherent;
+};
 
 #define kvm_err(fmt, ...) \
 	pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index f0a16b4..53331fe 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -1264,6 +1264,13 @@ struct kvm_vfio_spapr_tce {
 					struct kvm_userspace_memory_region)
 #define KVM_SET_TSS_ADDR          _IO(KVMIO,   0x47)
 #define KVM_SET_IDENTITY_MAP_ADDR _IOW(KVMIO,  0x48, __u64)
+struct kvm_discard_msg {
+	__u64 iov_len;
+	__u64 iov_base;
+	__u64 in_addr;
+};
+#define KVM_DISCARD_RANGE   _IOW(KVMIO, 0x49, struct kvm_discard_msg)
+
 
 /* enable ucontrol for s390 */
 struct kvm_s390_ucas_mapping {
diff --git a/virt/kvm/vfio.c b/virt/kvm/vfio.c
index 8fcbc50..f6dc61e 100644
--- a/virt/kvm/vfio.c
+++ b/virt/kvm/vfio.c
@@ -21,17 +21,6 @@
 #include <asm/kvm_ppc.h>
 #endif
 
-struct kvm_vfio_group {
-	struct list_head node;
-	struct vfio_group *vfio_group;
-};
-
-struct kvm_vfio {
-	struct list_head group_list;
-	struct mutex lock;
-	bool noncoherent;
-};
-
 static struct vfio_group *kvm_vfio_group_get_external_user(struct file *filep)
 {
 	struct vfio_group *vfio_group;
-- 
1.8.3.1

Re: [PATCH 2/2] KVM: add support for page hinting

[kbuild test robot]

[-- Attachment #1: Type: text/plain, Size: 3281 bytes --]

Hi weiqi,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on kvm/linux-next]
[also build test ERROR on vfio/next vhost/linux-next]
[cannot apply to v5.6-rc2 next-20200217]
[if your patch is applied to the wrong git tree, please drop us a note to help
improve the system. BTW, we also suggest to use '--base' option to specify the
base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/weiqi/page-hinting-add-passthrough-support/20200108-152941
base:   https://git.kernel.org/pub/scm/virt/kvm/kvm.git linux-next
config: x86_64-lkp (attached as .config)
compiler: gcc-7 (Debian 7.5.0-3) 7.5.0
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

If you fix the issue, kindly add following tag
Reported-by: kbuild test robot <lkp@intel.com>

All errors (new ones prefixed by >>):

   ld: arch/x86/kvm/x86.o: in function `kvm_vfio_ummap_range':
>> arch/x86/kvm/x86.c:4880: undefined reference to `vfio_dma_find'
>> ld: arch/x86/kvm/x86.c:4889: undefined reference to `vfio_munmap_pages'
   ld: arch/x86/kvm/mmu/mmu.o: in function `kvm_vfio_mmap_range':
>> arch/x86/kvm/mmu/mmu.c:4302: undefined reference to `vfio_dma_find'
>> ld: arch/x86/kvm/mmu/mmu.c:4321: undefined reference to `vfio_mmap_pages'

vim +4880 arch/x86/kvm/x86.c

  4838	
  4839	#include <linux/vfio.h>
  4840	static void kvm_vfio_ummap_range(struct kvm *kvm, struct kvm_device *tmp,
  4841					gfn_t gfn, int npages, unsigned long hva)
  4842	{
  4843		struct kvm_vfio *kv = tmp->private;
  4844		struct kvm_vfio_group *kvg;
  4845	
  4846		list_for_each_entry(kvg, &kv->group_list, node) {
  4847			struct vfio_group *group = kvg->vfio_group;
  4848			struct vfio_device *it, *device = NULL;
  4849	
  4850			list_for_each_entry(it, &group->device_list, group_next) {
  4851				unsigned long page_size, page_size_base;
  4852				unsigned long addr;
  4853				int size;
  4854				unsigned long old_pfn = 0;
  4855				int ret = 0;
  4856				size_t unmapped = npages;
  4857				gfn_t iova_gfn = gfn;
  4858				unsigned long iova_hva = hva;
  4859	
  4860				device = it;
  4861				while (unmapped) {
  4862					addr = gfn_to_hva(kvm, iova_gfn);
  4863					page_size_base = page_size =
  4864							kvm_host_page_size(kvm,
  4865							iova_gfn);
  4866	
  4867					if (addr != iova_hva)
  4868						return;
  4869	
  4870					while ((iova_gfn << PAGE_SHIFT) &
  4871						(page_size - 1))
  4872						page_size >>= 1;
  4873	
  4874					while (addr & (page_size - 1))
  4875						page_size >>= 1;
  4876	
  4877					if (page_size_base != page_size)
  4878						return;
  4879	
> 4880					size = vfio_dma_find(device->dev, iova_gfn,
  4881							page_size >> PAGE_SHIFT,
  4882							&old_pfn);
  4883					if (!size)
  4884						return;
  4885	
  4886					if (!old_pfn)
  4887						return;
  4888	
> 4889					ret = vfio_munmap_pages(device->dev,
  4890							iova_gfn, page_size);
  4891					unmapped -= page_size >> PAGE_SHIFT;
  4892					iova_hva += page_size;
  4893					iova_gfn += page_size >> PAGE_SHIFT;
  4894				}
  4895			}
  4896		}
  4897	}
  4898	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org

[-- Attachment #2: .config.gz --]
[-- Type: application/gzip, Size: 28600 bytes --]

Re: [PATCH 0/2] page hinting add passthrough support

[Alexander Duyck]

On Tue, 2020-01-07 at 22:46 +0800, weiqi wrote:
> From: wei qi <weiqi4@huawei.com>
> 
> 
> I just implemented dynamically updating the iommu table to support pass-through,
> It seen to work fine.
> 
> Test:
> start a 4G vm with 2M hugetlb and ixgbevf passthrough, 
>     GuestOS: linux-5.2.6 + 	(mm / virtio: Provide support for free page reporting)
>     HostOS: 5.5-rc4
>     Host: Intel(R) Xeon(R) Gold 6161 CPU @ 2.20GHz 
> 
> after enable page hinting, free pages at GuestOS can be free at host. 
> 
> 
> before,
>  # cat /sys/devices/system/node/node*/hugepages/hugepages-2048kB/free_hugepages 
>  5620
>  5620
> after start VM,
>  # numastat -c qemu
> 
>  Per-node process memory usage (in MBs)
>  PID              Node 0 Node 1 Total
>  ---------------  ------ ------ -----
>  24463 (qemu_hotr      6      6    12
>  24479 (qemu_tls_      0      8     8
>  70718 (qemu-syst     58    539   597
>  ---------------  ------ ------ -----
>  Total                64    553   616
>  # cat /sys/devices/system/node/node*/hugepages/hugepages-2048kB/free_hugepages 
>  5595
>  5366
> 
> the modify at qemu,
>  +int kvm_discard_range(struct kvm_discard_msg discard_msg)  
>  +{
>  +    return kvm_vm_ioctl(kvm_state, KVM_DISCARD_RANGE, &discard_msg);
>  +}
> 
>  static void virtio_balloon_handle_report(VirtIODevice *vdev, VirtQueue *vq)
>  {
>             ..................
>  +           discard_msg.in_addr = elem->in_addr[i];
>  +           discard_msg.iov_len = elem->in_sg[i].iov_len;
> 
>             ram_block_discard_range(rb, ram_offset, size);
>  +           kvm_discard_range(discard_msg);
> 
> then, further test network bandwidth, performance seem ok.
>  
> Is there any hidden problem in this implementation?

How is it you are avoiding triggering the call to qemu_balloon_inhibit in
QEMU?

> And, is there plan to support pass-throughyour?

It wasn't something I was immediately planning to do. Before we got there
we would need to really address the fact that the host has no idea what
pages the device could be accessing since normally the entire guest is
pinned. I guess these patches are a step toward addressing that?