2019-06-09 13:43:46

by Jacob Pan

[permalink] [raw]
Subject: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

When supporting guest SVA with emulated IOMMU, the guest PASID
table is shadowed in VMM. Updates to guest vIOMMU PASID table
will result in PASID cache flush which will be passed down to
the host as bind guest PASID calls.

For the SL page tables, it will be harvested from device's
default domain (request w/o PASID), or aux domain in case of
mediated device.

.-------------. .---------------------------.
| vIOMMU | | Guest process CR3, FL only|
| | '---------------------------'
.----------------/
| PASID Entry |--- PASID cache flush -
'-------------' |
| | V
| | CR3 in GPA
'-------------'
Guest
------| Shadow |--------------------------|--------
v v v
Host
.-------------. .----------------------.
| pIOMMU | | Bind FL for GVA-GPA |
| | '----------------------'
.----------------/ |
| PASID Entry | V (Nested xlate)
'----------------\.------------------------------.
| | |SL for GPA-HPA, default domain|
| | '------------------------------'
'-------------'
Where:
- FL = First level/stage one page tables
- SL = Second level/stage two page tables

Signed-off-by: Jacob Pan <[email protected]>
Signed-off-by: Liu, Yi L <[email protected]>
---
drivers/iommu/intel-iommu.c | 4 +
drivers/iommu/intel-svm.c | 187 ++++++++++++++++++++++++++++++++++++++++++++
include/linux/intel-iommu.h | 13 ++-
include/linux/intel-svm.h | 17 ++++
4 files changed, 219 insertions(+), 2 deletions(-)

diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
index 7cfa0eb..3b4d712 100644
--- a/drivers/iommu/intel-iommu.c
+++ b/drivers/iommu/intel-iommu.c
@@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
.dev_enable_feat = intel_iommu_dev_enable_feat,
.dev_disable_feat = intel_iommu_dev_disable_feat,
.pgsize_bitmap = INTEL_IOMMU_PGSIZES,
+#ifdef CONFIG_INTEL_IOMMU_SVM
+ .sva_bind_gpasid = intel_svm_bind_gpasid,
+ .sva_unbind_gpasid = intel_svm_unbind_gpasid,
+#endif
};

static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
index 66d98e1..f06a82f 100644
--- a/drivers/iommu/intel-svm.c
+++ b/drivers/iommu/intel-svm.c
@@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
list_for_each_entry(sdev, &svm->devs, list) \
if (dev == sdev->dev) \

+int intel_svm_bind_gpasid(struct iommu_domain *domain,
+ struct device *dev,
+ struct gpasid_bind_data *data)
+{
+ struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
+ struct intel_svm_dev *sdev;
+ struct intel_svm *svm = NULL;
+ struct dmar_domain *ddomain;
+ int ret = 0;
+
+ if (WARN_ON(!iommu) || !data)
+ return -EINVAL;
+
+ if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
+ data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
+ return -EINVAL;
+
+ if (dev_is_pci(dev)) {
+ /* VT-d supports devices with full 20 bit PASIDs only */
+ if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
+ return -EINVAL;
+ }
+
+ /*
+ * We only check host PASID range, we have no knowledge to check
+ * guest PASID range nor do we use the guest PASID.
+ */
+ if (data->hpasid <= 0 || data->hpasid >= PASID_MAX)
+ return -EINVAL;
+
+ ddomain = to_dmar_domain(domain);
+ /* REVISIT:
+ * Sanity check adddress width and paging mode support
+ * width matching in two dimensions:
+ * 1. paging mode CPU <= IOMMU
+ * 2. address width Guest <= Host.
+ */
+ mutex_lock(&pasid_mutex);
+ svm = ioasid_find(NULL, data->hpasid, NULL);
+ if (IS_ERR(svm)) {
+ ret = PTR_ERR(svm);
+ goto out;
+ }
+ if (svm) {
+ /*
+ * If we found svm for the PASID, there must be at
+ * least one device bond, otherwise svm should be freed.
+ */
+ BUG_ON(list_empty(&svm->devs));
+
+ for_each_svm_dev() {
+ /* In case of multiple sub-devices of the same pdev assigned, we should
+ * allow multiple bind calls with the same PASID and pdev.
+ */
+ sdev->users++;
+ goto out;
+ }
+ } else {
+ /* We come here when PASID has never been bond to a device. */
+ svm = kzalloc(sizeof(*svm), GFP_KERNEL);
+ if (!svm) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ /* REVISIT: upper layer/VFIO can track host process that bind the PASID.
+ * ioasid_set = mm might be sufficient for vfio to check pasid VMM
+ * ownership.
+ */
+ svm->mm = get_task_mm(current);
+ svm->pasid = data->hpasid;
+ if (data->flags & IOMMU_SVA_GPASID_VAL) {
+ svm->gpasid = data->gpasid;
+ svm->flags &= SVM_FLAG_GUEST_PASID;
+ }
+ refcount_set(&svm->refs, 0);
+ ioasid_set_data(data->hpasid, svm);
+ INIT_LIST_HEAD_RCU(&svm->devs);
+ INIT_LIST_HEAD(&svm->list);
+
+ mmput(svm->mm);
+ }
+ sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
+ if (!sdev) {
+ ret = -ENOMEM;
+ goto out;
+ }
+ sdev->dev = dev;
+ sdev->users = 1;
+
+ /* Set up device context entry for PASID if not enabled already */
+ ret = intel_iommu_enable_pasid(iommu, sdev->dev);
+ if (ret) {
+ dev_err(dev, "Failed to enable PASID capability\n");
+ kfree(sdev);
+ goto out;
+ }
+
+ /*
+ * For guest bind, we need to set up PASID table entry as follows:
+ * - FLPM matches guest paging mode
+ * - turn on nested mode
+ * - SL guest address width matching
+ */
+ ret = intel_pasid_setup_nested(iommu,
+ dev,
+ (pgd_t *)data->gpgd,
+ data->hpasid,
+ data->flags,
+ ddomain,
+ data->addr_width);
+ if (ret) {
+ dev_err(dev, "Failed to set up PASID %llu in nested mode, Err %d\n",
+ data->hpasid, ret);
+ kfree(sdev);
+ goto out;
+ }
+ svm->flags |= SVM_FLAG_GUEST_MODE;
+
+ init_rcu_head(&sdev->rcu);
+ refcount_inc(&svm->refs);
+ list_add_rcu(&sdev->list, &svm->devs);
+ out:
+ mutex_unlock(&pasid_mutex);
+ return ret;
+}
+
+int intel_svm_unbind_gpasid(struct device *dev, int pasid)
+{
+ struct intel_svm_dev *sdev;
+ struct intel_iommu *iommu;
+ struct intel_svm *svm;
+ int ret = -EINVAL;
+
+ mutex_lock(&pasid_mutex);
+ iommu = intel_svm_device_to_iommu(dev);
+ if (!iommu)
+ goto out;
+
+ svm = ioasid_find(NULL, pasid, NULL);
+ if (IS_ERR(svm)) {
+ ret = PTR_ERR(svm);
+ goto out;
+ }
+
+ if (!svm)
+ goto out;
+
+ for_each_svm_dev() {
+ ret = 0;
+ sdev->users--;
+ if (!sdev->users) {
+ list_del_rcu(&sdev->list);
+ intel_pasid_tear_down_entry(iommu, dev, svm->pasid);
+ /* TODO: Drain in flight PRQ for the PASID since it
+ * may get reused soon, we don't want to
+ * confuse with its previous live.
+ * intel_svm_drain_prq(dev, pasid);
+ */
+ kfree_rcu(sdev, rcu);
+
+ if (list_empty(&svm->devs)) {
+ list_del(&svm->list);
+ kfree(svm);
+ /*
+ * We do not free PASID here until explicit call
+ * from VFIO to free. The PASID life cycle
+ * management is largely tied to VFIO management
+ * of assigned device life cycles. In case of
+ * guest exit without a explicit free PASID call,
+ * the responsibility lies in VFIO layer to free
+ * the PASIDs allocated for the guest.
+ * For security reasons, VFIO has to track the
+ * PASID ownership per guest anyway to ensure
+ * that PASID allocated by one guest cannot be
+ * used by another.
+ */
+ ioasid_set_data(pasid, NULL);
+ }
+ }
+ break;
+ }
+ out:
+ mutex_unlock(&pasid_mutex);
+
+ return ret;
+}
+
int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ops *ops)
{
struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
diff --git a/include/linux/intel-iommu.h b/include/linux/intel-iommu.h
index b75f17d..94d3a9a 100644
--- a/include/linux/intel-iommu.h
+++ b/include/linux/intel-iommu.h
@@ -677,7 +677,9 @@ int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct device *dev);
int intel_svm_init(struct intel_iommu *iommu);
extern int intel_svm_enable_prq(struct intel_iommu *iommu);
extern int intel_svm_finish_prq(struct intel_iommu *iommu);
-
+extern int intel_svm_bind_gpasid(struct iommu_domain *domain,
+ struct device *dev, struct gpasid_bind_data *data);
+extern int intel_svm_unbind_gpasid(struct device *dev, int pasid);
struct svm_dev_ops;

struct intel_svm_dev {
@@ -693,12 +695,19 @@ struct intel_svm_dev {

struct intel_svm {
struct mmu_notifier notifier;
- struct mm_struct *mm;
+ union {
+ struct mm_struct *mm;
+ u64 gcr3;
+ };
struct intel_iommu *iommu;
int flags;
int pasid;
+ int gpasid; /* Guest PASID in case of vSVA bind with non-identity host
+ * to guest PASID mapping.
+ */
struct list_head devs;
struct list_head list;
+ refcount_t refs; /* Number of devices sharing this PASID */
};

extern struct intel_iommu *intel_svm_device_to_iommu(struct device *dev);
diff --git a/include/linux/intel-svm.h b/include/linux/intel-svm.h
index e3f7631..577d5df 100644
--- a/include/linux/intel-svm.h
+++ b/include/linux/intel-svm.h
@@ -52,6 +52,23 @@ struct svm_dev_ops {
* do such IOTLB flushes automatically.
*/
#define SVM_FLAG_SUPERVISOR_MODE (1<<1)
+/*
+ * The SVM_FLAG_GUEST_MODE flag is used when a guest process bind to a device.
+ * In this case the mm_struct is in the guest kernel or userspace, its life
+ * cycle is managed by VMM and VFIO layer. For IOMMU driver, this API provides
+ * means to bind/unbind guest CR3 with PASIDs allocated for a device.
+ */
+#define SVM_FLAG_GUEST_MODE (1<<2)
+/*
+ * The SVM_FLAG_GUEST_PASID flag is used when a guest has its own PASID space,
+ * which requires guest and host PASID translation at both directions. We keep
+ * track of guest PASID in order to provide lookup service to device drivers.
+ * One such example is a physical function (PF) driver that supports mediated
+ * device (mdev) assignment. Guest programming of mdev configuration space can
+ * only be done with guest PASID, therefore PF driver needs to find the matching
+ * host PASID to program the real hardware.
+ */
+#define SVM_FLAG_GUEST_PASID (1<<3)

#ifdef CONFIG_INTEL_IOMMU_SVM

--
2.7.4


2019-06-18 16:45:40

by Jonathan Cameron

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

On Sun, 9 Jun 2019 06:44:20 -0700
Jacob Pan <[email protected]> wrote:

> When supporting guest SVA with emulated IOMMU, the guest PASID
> table is shadowed in VMM. Updates to guest vIOMMU PASID table
> will result in PASID cache flush which will be passed down to
> the host as bind guest PASID calls.
>
> For the SL page tables, it will be harvested from device's
> default domain (request w/o PASID), or aux domain in case of
> mediated device.
>
> .-------------. .---------------------------.
> | vIOMMU | | Guest process CR3, FL only|
> | | '---------------------------'
> .----------------/
> | PASID Entry |--- PASID cache flush -
> '-------------' |
> | | V
> | | CR3 in GPA
> '-------------'
> Guest
> ------| Shadow |--------------------------|--------
> v v v
> Host
> .-------------. .----------------------.
> | pIOMMU | | Bind FL for GVA-GPA |
> | | '----------------------'
> .----------------/ |
> | PASID Entry | V (Nested xlate)
> '----------------\.------------------------------.
> | | |SL for GPA-HPA, default domain|
> | | '------------------------------'
> '-------------'
> Where:
> - FL = First level/stage one page tables
> - SL = Second level/stage two page tables
>
> Signed-off-by: Jacob Pan <[email protected]>
> Signed-off-by: Liu, Yi L <[email protected]>


A few trivial bits inline. As far as I can tell looks good but I'm not that
familiar with the hardware.

Jonathan

> ---
> drivers/iommu/intel-iommu.c | 4 +
> drivers/iommu/intel-svm.c | 187 ++++++++++++++++++++++++++++++++++++++++++++
> include/linux/intel-iommu.h | 13 ++-
> include/linux/intel-svm.h | 17 ++++
> 4 files changed, 219 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
> index 7cfa0eb..3b4d712 100644
> --- a/drivers/iommu/intel-iommu.c
> +++ b/drivers/iommu/intel-iommu.c
> @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
> .dev_enable_feat = intel_iommu_dev_enable_feat,
> .dev_disable_feat = intel_iommu_dev_disable_feat,
> .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
> +#ifdef CONFIG_INTEL_IOMMU_SVM
> + .sva_bind_gpasid = intel_svm_bind_gpasid,
> + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
> +#endif
> };
>
> static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
> diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> index 66d98e1..f06a82f 100644
> --- a/drivers/iommu/intel-svm.c
> +++ b/drivers/iommu/intel-svm.c
> @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
> list_for_each_entry(sdev, &svm->devs, list) \
> if (dev == sdev->dev) \
>
> +int intel_svm_bind_gpasid(struct iommu_domain *domain,
> + struct device *dev,
> + struct gpasid_bind_data *data)
> +{
> + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> + struct intel_svm_dev *sdev;
> + struct intel_svm *svm = NULL;
I think this is set in all the paths that use it..

> + struct dmar_domain *ddomain;
> + int ret = 0;
> +
> + if (WARN_ON(!iommu) || !data)
> + return -EINVAL;
> +
> + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
> + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
> + return -EINVAL;
> +
> + if (dev_is_pci(dev)) {
> + /* VT-d supports devices with full 20 bit PASIDs only */
> + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
> + return -EINVAL;
> + }
> +
> + /*
> + * We only check host PASID range, we have no knowledge to check
> + * guest PASID range nor do we use the guest PASID.
> + */
> + if (data->hpasid <= 0 || data->hpasid >= PASID_MAX)
> + return -EINVAL;
> +
> + ddomain = to_dmar_domain(domain);
> + /* REVISIT:
> + * Sanity check adddress width and paging mode support
> + * width matching in two dimensions:
> + * 1. paging mode CPU <= IOMMU
> + * 2. address width Guest <= Host.
> + */
> + mutex_lock(&pasid_mutex);
> + svm = ioasid_find(NULL, data->hpasid, NULL);
> + if (IS_ERR(svm)) {
> + ret = PTR_ERR(svm);
> + goto out;
> + }
> + if (svm) {
> + /*
> + * If we found svm for the PASID, there must be at
> + * least one device bond, otherwise svm should be freed.
> + */
> + BUG_ON(list_empty(&svm->devs));
> +
> + for_each_svm_dev() {
> + /* In case of multiple sub-devices of the same pdev assigned, we should
> + * allow multiple bind calls with the same PASID and pdev.
> + */
> + sdev->users++;
> + goto out;
> + }
> + } else {
> + /* We come here when PASID has never been bond to a device. */
> + svm = kzalloc(sizeof(*svm), GFP_KERNEL);
> + if (!svm) {
> + ret = -ENOMEM;
> + goto out;
> + }
> + /* REVISIT: upper layer/VFIO can track host process that bind the PASID.
> + * ioasid_set = mm might be sufficient for vfio to check pasid VMM
> + * ownership.
> + */
> + svm->mm = get_task_mm(current);
> + svm->pasid = data->hpasid;
> + if (data->flags & IOMMU_SVA_GPASID_VAL) {
> + svm->gpasid = data->gpasid;
> + svm->flags &= SVM_FLAG_GUEST_PASID;
> + }
> + refcount_set(&svm->refs, 0);
> + ioasid_set_data(data->hpasid, svm);
> + INIT_LIST_HEAD_RCU(&svm->devs);
> + INIT_LIST_HEAD(&svm->list);
> +
> + mmput(svm->mm);
> + }
> + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> + if (!sdev) {
> + ret = -ENOMEM;
> + goto out;
> + }
> + sdev->dev = dev;
> + sdev->users = 1;
> +
> + /* Set up device context entry for PASID if not enabled already */
> + ret = intel_iommu_enable_pasid(iommu, sdev->dev);
> + if (ret) {
> + dev_err(dev, "Failed to enable PASID capability\n");
> + kfree(sdev);
> + goto out;
> + }
> +
> + /*
> + * For guest bind, we need to set up PASID table entry as follows:
> + * - FLPM matches guest paging mode
> + * - turn on nested mode
> + * - SL guest address width matching
> + */
> + ret = intel_pasid_setup_nested(iommu,
> + dev,
> + (pgd_t *)data->gpgd,
> + data->hpasid,
> + data->flags,
> + ddomain,
> + data->addr_width);
> + if (ret) {
> + dev_err(dev, "Failed to set up PASID %llu in nested mode, Err %d\n",
> + data->hpasid, ret);
> + kfree(sdev);
> + goto out;
> + }
> + svm->flags |= SVM_FLAG_GUEST_MODE;
> +
> + init_rcu_head(&sdev->rcu);
> + refcount_inc(&svm->refs);
> + list_add_rcu(&sdev->list, &svm->devs);
> + out:
> + mutex_unlock(&pasid_mutex);
> + return ret;
> +}
> +
> +int intel_svm_unbind_gpasid(struct device *dev, int pasid)
> +{
> + struct intel_svm_dev *sdev;
> + struct intel_iommu *iommu;
> + struct intel_svm *svm;
> + int ret = -EINVAL;
> +
> + mutex_lock(&pasid_mutex);
> + iommu = intel_svm_device_to_iommu(dev);
> + if (!iommu)
> + goto out;
> +
> + svm = ioasid_find(NULL, pasid, NULL);
> + if (IS_ERR(svm)) {
> + ret = PTR_ERR(svm);
> + goto out;
> + }
> +
> + if (!svm)
> + goto out;
> +
> + for_each_svm_dev() {
> + ret = 0;
> + sdev->users--;
> + if (!sdev->users) {
> + list_del_rcu(&sdev->list);
> + intel_pasid_tear_down_entry(iommu, dev, svm->pasid);
> + /* TODO: Drain in flight PRQ for the PASID since it
> + * may get reused soon, we don't want to
> + * confuse with its previous live.
life?

> + * intel_svm_drain_prq(dev, pasid);
> + */
> + kfree_rcu(sdev, rcu);
> +
> + if (list_empty(&svm->devs)) {
> + list_del(&svm->list);
> + kfree(svm);
> + /*
> + * We do not free PASID here until explicit call
> + * from VFIO to free. The PASID life cycle
> + * management is largely tied to VFIO management
> + * of assigned device life cycles. In case of
> + * guest exit without a explicit free PASID call,
> + * the responsibility lies in VFIO layer to free
> + * the PASIDs allocated for the guest.
> + * For security reasons, VFIO has to track the
> + * PASID ownership per guest anyway to ensure
> + * that PASID allocated by one guest cannot be
> + * used by another.
> + */
> + ioasid_set_data(pasid, NULL);
> + }
> + }
> + break;
> + }
> + out:
> + mutex_unlock(&pasid_mutex);
> +
> + return ret;
> +}
> +
> int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ops *ops)
> {
> struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> diff --git a/include/linux/intel-iommu.h b/include/linux/intel-iommu.h
> index b75f17d..94d3a9a 100644
> --- a/include/linux/intel-iommu.h
> +++ b/include/linux/intel-iommu.h
> @@ -677,7 +677,9 @@ int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct device *dev);
> int intel_svm_init(struct intel_iommu *iommu);
> extern int intel_svm_enable_prq(struct intel_iommu *iommu);
> extern int intel_svm_finish_prq(struct intel_iommu *iommu);
> -
> +extern int intel_svm_bind_gpasid(struct iommu_domain *domain,
> + struct device *dev, struct gpasid_bind_data *data);
> +extern int intel_svm_unbind_gpasid(struct device *dev, int pasid);
> struct svm_dev_ops;
>
> struct intel_svm_dev {
> @@ -693,12 +695,19 @@ struct intel_svm_dev {
>
> struct intel_svm {
> struct mmu_notifier notifier;
> - struct mm_struct *mm;
> + union {
> + struct mm_struct *mm;
> + u64 gcr3;
> + };
> struct intel_iommu *iommu;
> int flags;
> int pasid;
> + int gpasid; /* Guest PASID in case of vSVA bind with non-identity host
> + * to guest PASID mapping.
> + */
> struct list_head devs;
> struct list_head list;
> + refcount_t refs; /* Number of devices sharing this PASID */
> };
>
> extern struct intel_iommu *intel_svm_device_to_iommu(struct device *dev);
> diff --git a/include/linux/intel-svm.h b/include/linux/intel-svm.h
> index e3f7631..577d5df 100644
> --- a/include/linux/intel-svm.h
> +++ b/include/linux/intel-svm.h
> @@ -52,6 +52,23 @@ struct svm_dev_ops {
> * do such IOTLB flushes automatically.
> */
> #define SVM_FLAG_SUPERVISOR_MODE (1<<1)
> +/*
> + * The SVM_FLAG_GUEST_MODE flag is used when a guest process bind to a device.
> + * In this case the mm_struct is in the guest kernel or userspace, its life
> + * cycle is managed by VMM and VFIO layer. For IOMMU driver, this API provides
> + * means to bind/unbind guest CR3 with PASIDs allocated for a device.
> + */
> +#define SVM_FLAG_GUEST_MODE (1<<2)
> +/*
> + * The SVM_FLAG_GUEST_PASID flag is used when a guest has its own PASID space,
> + * which requires guest and host PASID translation at both directions. We keep
> + * track of guest PASID in order to provide lookup service to device drivers.
> + * One such example is a physical function (PF) driver that supports mediated
> + * device (mdev) assignment. Guest programming of mdev configuration space can
> + * only be done with guest PASID, therefore PF driver needs to find the matching
> + * host PASID to program the real hardware.
> + */
> +#define SVM_FLAG_GUEST_PASID (1<<3)
>
> #ifdef CONFIG_INTEL_IOMMU_SVM
>


2019-06-24 22:39:18

by Jacob Pan

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

On Tue, 18 Jun 2019 17:44:49 +0100
Jonathan Cameron <[email protected]> wrote:

> On Sun, 9 Jun 2019 06:44:20 -0700
> Jacob Pan <[email protected]> wrote:
>
> > When supporting guest SVA with emulated IOMMU, the guest PASID
> > table is shadowed in VMM. Updates to guest vIOMMU PASID table
> > will result in PASID cache flush which will be passed down to
> > the host as bind guest PASID calls.
> >
> > For the SL page tables, it will be harvested from device's
> > default domain (request w/o PASID), or aux domain in case of
> > mediated device.
> >
> > .-------------. .---------------------------.
> > | vIOMMU | | Guest process CR3, FL only|
> > | | '---------------------------'
> > .----------------/
> > | PASID Entry |--- PASID cache flush -
> > '-------------' |
> > | | V
> > | | CR3 in GPA
> > '-------------'
> > Guest
> > ------| Shadow |--------------------------|--------
> > v v v
> > Host
> > .-------------. .----------------------.
> > | pIOMMU | | Bind FL for GVA-GPA |
> > | | '----------------------'
> > .----------------/ |
> > | PASID Entry | V (Nested xlate)
> > '----------------\.------------------------------.
> > | | |SL for GPA-HPA, default domain|
> > | | '------------------------------'
> > '-------------'
> > Where:
> > - FL = First level/stage one page tables
> > - SL = Second level/stage two page tables
> >
> > Signed-off-by: Jacob Pan <[email protected]>
> > Signed-off-by: Liu, Yi L <[email protected]>
>
>
> A few trivial bits inline. As far as I can tell looks good but I'm
> not that familiar with the hardware.
>
Thank you so much for the review.

> Jonathan
>
> > ---
> > drivers/iommu/intel-iommu.c | 4 +
> > drivers/iommu/intel-svm.c | 187
> > ++++++++++++++++++++++++++++++++++++++++++++
> > include/linux/intel-iommu.h | 13 ++- include/linux/intel-svm.h
> > | 17 ++++ 4 files changed, 219 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/iommu/intel-iommu.c
> > b/drivers/iommu/intel-iommu.c index 7cfa0eb..3b4d712 100644
> > --- a/drivers/iommu/intel-iommu.c
> > +++ b/drivers/iommu/intel-iommu.c
> > @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
> > .dev_enable_feat = intel_iommu_dev_enable_feat,
> > .dev_disable_feat = intel_iommu_dev_disable_feat,
> > .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
> > +#ifdef CONFIG_INTEL_IOMMU_SVM
> > + .sva_bind_gpasid = intel_svm_bind_gpasid,
> > + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
> > +#endif
> > };
> >
> > static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
> > diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> > index 66d98e1..f06a82f 100644
> > --- a/drivers/iommu/intel-svm.c
> > +++ b/drivers/iommu/intel-svm.c
> > @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
> > list_for_each_entry(sdev, &svm->devs, list) \
> > if (dev == sdev->dev) \
> >
> > +int intel_svm_bind_gpasid(struct iommu_domain *domain,
> > + struct device *dev,
> > + struct gpasid_bind_data *data)
> > +{
> > + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> > + struct intel_svm_dev *sdev;
> > + struct intel_svm *svm = NULL;
> I think this is set in all the paths that use it..
>
indeed. will remove.
> > + struct dmar_domain *ddomain;
> > + int ret = 0;
> > +
> > + if (WARN_ON(!iommu) || !data)
> > + return -EINVAL;
> > +
> > + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
> > + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
> > + return -EINVAL;
> > +
> > + if (dev_is_pci(dev)) {
> > + /* VT-d supports devices with full 20 bit PASIDs
> > only */
> > + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
> > + return -EINVAL;
> > + }
> > +
> > + /*
> > + * We only check host PASID range, we have no knowledge to
> > check
> > + * guest PASID range nor do we use the guest PASID.
> > + */
> > + if (data->hpasid <= 0 || data->hpasid >= PASID_MAX)
> > + return -EINVAL;
> > +
> > + ddomain = to_dmar_domain(domain);
> > + /* REVISIT:
> > + * Sanity check adddress width and paging mode support
> > + * width matching in two dimensions:
> > + * 1. paging mode CPU <= IOMMU
> > + * 2. address width Guest <= Host.
> > + */
> > + mutex_lock(&pasid_mutex);
> > + svm = ioasid_find(NULL, data->hpasid, NULL);
> > + if (IS_ERR(svm)) {
> > + ret = PTR_ERR(svm);
> > + goto out;
> > + }
> > + if (svm) {
> > + /*
> > + * If we found svm for the PASID, there must be at
> > + * least one device bond, otherwise svm should be
> > freed.
> > + */
> > + BUG_ON(list_empty(&svm->devs));
> > +
> > + for_each_svm_dev() {
> > + /* In case of multiple sub-devices of the
> > same pdev assigned, we should
> > + * allow multiple bind calls with the same
> > PASID and pdev.
> > + */
> > + sdev->users++;
> > + goto out;
> > + }
> > + } else {
> > + /* We come here when PASID has never been bond to
> > a device. */
> > + svm = kzalloc(sizeof(*svm), GFP_KERNEL);
> > + if (!svm) {
> > + ret = -ENOMEM;
> > + goto out;
> > + }
> > + /* REVISIT: upper layer/VFIO can track host
> > process that bind the PASID.
> > + * ioasid_set = mm might be sufficient for vfio to
> > check pasid VMM
> > + * ownership.
> > + */
> > + svm->mm = get_task_mm(current);
> > + svm->pasid = data->hpasid;
> > + if (data->flags & IOMMU_SVA_GPASID_VAL) {
> > + svm->gpasid = data->gpasid;
> > + svm->flags &= SVM_FLAG_GUEST_PASID;
> > + }
> > + refcount_set(&svm->refs, 0);
> > + ioasid_set_data(data->hpasid, svm);
> > + INIT_LIST_HEAD_RCU(&svm->devs);
> > + INIT_LIST_HEAD(&svm->list);
> > +
> > + mmput(svm->mm);
> > + }
> > + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> > + if (!sdev) {
> > + ret = -ENOMEM;
> > + goto out;
> > + }
> > + sdev->dev = dev;
> > + sdev->users = 1;
> > +
> > + /* Set up device context entry for PASID if not enabled
> > already */
> > + ret = intel_iommu_enable_pasid(iommu, sdev->dev);
> > + if (ret) {
> > + dev_err(dev, "Failed to enable PASID
> > capability\n");
> > + kfree(sdev);
> > + goto out;
> > + }
> > +
> > + /*
> > + * For guest bind, we need to set up PASID table entry as
> > follows:
> > + * - FLPM matches guest paging mode
> > + * - turn on nested mode
> > + * - SL guest address width matching
> > + */
> > + ret = intel_pasid_setup_nested(iommu,
> > + dev,
> > + (pgd_t *)data->gpgd,
> > + data->hpasid,
> > + data->flags,
> > + ddomain,
> > + data->addr_width);
> > + if (ret) {
> > + dev_err(dev, "Failed to set up PASID %llu in
> > nested mode, Err %d\n",
> > + data->hpasid, ret);
> > + kfree(sdev);
> > + goto out;
> > + }
> > + svm->flags |= SVM_FLAG_GUEST_MODE;
> > +
> > + init_rcu_head(&sdev->rcu);
> > + refcount_inc(&svm->refs);
> > + list_add_rcu(&sdev->list, &svm->devs);
> > + out:
> > + mutex_unlock(&pasid_mutex);
> > + return ret;
> > +}
> > +
> > +int intel_svm_unbind_gpasid(struct device *dev, int pasid)
> > +{
> > + struct intel_svm_dev *sdev;
> > + struct intel_iommu *iommu;
> > + struct intel_svm *svm;
> > + int ret = -EINVAL;
> > +
> > + mutex_lock(&pasid_mutex);
> > + iommu = intel_svm_device_to_iommu(dev);
> > + if (!iommu)
> > + goto out;
> > +
> > + svm = ioasid_find(NULL, pasid, NULL);
> > + if (IS_ERR(svm)) {
> > + ret = PTR_ERR(svm);
> > + goto out;
> > + }
> > +
> > + if (!svm)
> > + goto out;
> > +
> > + for_each_svm_dev() {
> > + ret = 0;
> > + sdev->users--;
> > + if (!sdev->users) {
> > + list_del_rcu(&sdev->list);
> > + intel_pasid_tear_down_entry(iommu, dev,
> > svm->pasid);
> > + /* TODO: Drain in flight PRQ for the PASID
> > since it
> > + * may get reused soon, we don't want to
> > + * confuse with its previous live.
> life?
>
right. thanks.
> > + * intel_svm_drain_prq(dev, pasid);
> > + */
> > + kfree_rcu(sdev, rcu);
> > +
> > + if (list_empty(&svm->devs)) {
> > + list_del(&svm->list);
> > + kfree(svm);
> > + /*
> > + * We do not free PASID here until
> > explicit call
> > + * from VFIO to free. The PASID
> > life cycle
> > + * management is largely tied to
> > VFIO management
> > + * of assigned device life cycles.
> > In case of
> > + * guest exit without a explicit
> > free PASID call,
> > + * the responsibility lies in VFIO
> > layer to free
> > + * the PASIDs allocated for the
> > guest.
> > + * For security reasons, VFIO has
> > to track the
> > + * PASID ownership per guest
> > anyway to ensure
> > + * that PASID allocated by one
> > guest cannot be
> > + * used by another.
> > + */
> > + ioasid_set_data(pasid, NULL);
> > + }
> > + }
> > + break;
> > + }
> > + out:
> > + mutex_unlock(&pasid_mutex);
> > +
> > + return ret;
> > +}
> > +
> > int intel_svm_bind_mm(struct device *dev, int *pasid, int flags,
> > struct svm_dev_ops *ops) {
> > struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> > diff --git a/include/linux/intel-iommu.h
> > b/include/linux/intel-iommu.h index b75f17d..94d3a9a 100644
> > --- a/include/linux/intel-iommu.h
> > +++ b/include/linux/intel-iommu.h
> > @@ -677,7 +677,9 @@ int intel_iommu_enable_pasid(struct intel_iommu
> > *iommu, struct device *dev); int intel_svm_init(struct intel_iommu
> > *iommu); extern int intel_svm_enable_prq(struct intel_iommu *iommu);
> > extern int intel_svm_finish_prq(struct intel_iommu *iommu);
> > -
> > +extern int intel_svm_bind_gpasid(struct iommu_domain *domain,
> > + struct device *dev, struct gpasid_bind_data *data);
> > +extern int intel_svm_unbind_gpasid(struct device *dev, int pasid);
> > struct svm_dev_ops;
> >
> > struct intel_svm_dev {
> > @@ -693,12 +695,19 @@ struct intel_svm_dev {
> >
> > struct intel_svm {
> > struct mmu_notifier notifier;
> > - struct mm_struct *mm;
> > + union {
> > + struct mm_struct *mm;
> > + u64 gcr3;
> > + };
> > struct intel_iommu *iommu;
> > int flags;
> > int pasid;
> > + int gpasid; /* Guest PASID in case of vSVA bind with
> > non-identity host
> > + * to guest PASID mapping.
> > + */
> > struct list_head devs;
> > struct list_head list;
> > + refcount_t refs; /* Number of devices sharing this PASID */
> > };
> >
> > extern struct intel_iommu *intel_svm_device_to_iommu(struct device
> > *dev); diff --git a/include/linux/intel-svm.h
> > b/include/linux/intel-svm.h index e3f7631..577d5df 100644
> > --- a/include/linux/intel-svm.h
> > +++ b/include/linux/intel-svm.h
> > @@ -52,6 +52,23 @@ struct svm_dev_ops {
> > * do such IOTLB flushes automatically.
> > */
> > #define SVM_FLAG_SUPERVISOR_MODE (1<<1)
> > +/*
> > + * The SVM_FLAG_GUEST_MODE flag is used when a guest process bind
> > to a device.
> > + * In this case the mm_struct is in the guest kernel or userspace,
> > its life
> > + * cycle is managed by VMM and VFIO layer. For IOMMU driver, this
> > API provides
> > + * means to bind/unbind guest CR3 with PASIDs allocated for a
> > device.
> > + */
> > +#define SVM_FLAG_GUEST_MODE (1<<2)
> > +/*
> > + * The SVM_FLAG_GUEST_PASID flag is used when a guest has its own
> > PASID space,
> > + * which requires guest and host PASID translation at both
> > directions. We keep
> > + * track of guest PASID in order to provide lookup service to
> > device drivers.
> > + * One such example is a physical function (PF) driver that
> > supports mediated
> > + * device (mdev) assignment. Guest programming of mdev
> > configuration space can
> > + * only be done with guest PASID, therefore PF driver needs to
> > find the matching
> > + * host PASID to program the real hardware.
> > + */
> > +#define SVM_FLAG_GUEST_PASID (1<<3)
> >
> > #ifdef CONFIG_INTEL_IOMMU_SVM
> >
>
>

[Jacob Pan]

2019-06-27 02:58:32

by Lu Baolu

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

Hi Jacob and Yi,

On 6/9/19 9:44 PM, Jacob Pan wrote:
> When supporting guest SVA with emulated IOMMU, the guest PASID
> table is shadowed in VMM. Updates to guest vIOMMU PASID table
> will result in PASID cache flush which will be passed down to
> the host as bind guest PASID calls.
>
> For the SL page tables, it will be harvested from device's
> default domain (request w/o PASID), or aux domain in case of
> mediated device.
>
> .-------------. .---------------------------.
> | vIOMMU | | Guest process CR3, FL only|
> | | '---------------------------'
> .----------------/
> | PASID Entry |--- PASID cache flush -
> '-------------' |
> | | V
> | | CR3 in GPA
> '-------------'
> Guest
> ------| Shadow |--------------------------|--------
> v v v
> Host
> .-------------. .----------------------.
> | pIOMMU | | Bind FL for GVA-GPA |
> | | '----------------------'
> .----------------/ |
> | PASID Entry | V (Nested xlate)
> '----------------\.------------------------------.
> | | |SL for GPA-HPA, default domain|
> | | '------------------------------'
> '-------------'
> Where:
> - FL = First level/stage one page tables
> - SL = Second level/stage two page tables
>
> Signed-off-by: Jacob Pan <[email protected]>
> Signed-off-by: Liu, Yi L <[email protected]>
> ---
> drivers/iommu/intel-iommu.c | 4 +
> drivers/iommu/intel-svm.c | 187 ++++++++++++++++++++++++++++++++++++++++++++
> include/linux/intel-iommu.h | 13 ++-
> include/linux/intel-svm.h | 17 ++++
> 4 files changed, 219 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
> index 7cfa0eb..3b4d712 100644
> --- a/drivers/iommu/intel-iommu.c
> +++ b/drivers/iommu/intel-iommu.c
> @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
> .dev_enable_feat = intel_iommu_dev_enable_feat,
> .dev_disable_feat = intel_iommu_dev_disable_feat,
> .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
> +#ifdef CONFIG_INTEL_IOMMU_SVM
> + .sva_bind_gpasid = intel_svm_bind_gpasid,
> + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
> +#endif
> };
>
> static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
> diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> index 66d98e1..f06a82f 100644
> --- a/drivers/iommu/intel-svm.c
> +++ b/drivers/iommu/intel-svm.c
> @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
> list_for_each_entry(sdev, &svm->devs, list) \
> if (dev == sdev->dev) \
>
> +int intel_svm_bind_gpasid(struct iommu_domain *domain,
> + struct device *dev,
> + struct gpasid_bind_data *data)
> +{
> + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> + struct intel_svm_dev *sdev;
> + struct intel_svm *svm = NULL;
> + struct dmar_domain *ddomain;
> + int ret = 0;
> +
> + if (WARN_ON(!iommu) || !data)
> + return -EINVAL;
> +
> + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
> + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
> + return -EINVAL;
> +
> + if (dev_is_pci(dev)) {
> + /* VT-d supports devices with full 20 bit PASIDs only */
> + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
> + return -EINVAL;
> + }
> +
> + /*
> + * We only check host PASID range, we have no knowledge to check
> + * guest PASID range nor do we use the guest PASID.
> + */
> + if (data->hpasid <= 0 || data->hpasid >= PASID_MAX)
> + return -EINVAL;
> +
> + ddomain = to_dmar_domain(domain);
> + /* REVISIT:
> + * Sanity check adddress width and paging mode support
> + * width matching in two dimensions:
> + * 1. paging mode CPU <= IOMMU
> + * 2. address width Guest <= Host.
> + */
> + mutex_lock(&pasid_mutex);
> + svm = ioasid_find(NULL, data->hpasid, NULL);
> + if (IS_ERR(svm)) {
> + ret = PTR_ERR(svm);
> + goto out;
> + }
> + if (svm) {
> + /*
> + * If we found svm for the PASID, there must be at
> + * least one device bond, otherwise svm should be freed.
> + */
> + BUG_ON(list_empty(&svm->devs));
> +
> + for_each_svm_dev() {
> + /* In case of multiple sub-devices of the same pdev assigned, we should
> + * allow multiple bind calls with the same PASID and pdev.
> + */
> + sdev->users++;
> + goto out;
> + }
> + } else {
> + /* We come here when PASID has never been bond to a device. */
> + svm = kzalloc(sizeof(*svm), GFP_KERNEL);
> + if (!svm) {
> + ret = -ENOMEM;
> + goto out;
> + }
> + /* REVISIT: upper layer/VFIO can track host process that bind the PASID.
> + * ioasid_set = mm might be sufficient for vfio to check pasid VMM
> + * ownership.
> + */
> + svm->mm = get_task_mm(current);
> + svm->pasid = data->hpasid;
> + if (data->flags & IOMMU_SVA_GPASID_VAL) {
> + svm->gpasid = data->gpasid;
> + svm->flags &= SVM_FLAG_GUEST_PASID;

I am guessing that you want to set this flag bit, so it should be

svm->flags |= SVM_FLAG_GUEST_PASID;

> + }
> + refcount_set(&svm->refs, 0);
> + ioasid_set_data(data->hpasid, svm);
> + INIT_LIST_HEAD_RCU(&svm->devs);
> + INIT_LIST_HEAD(&svm->list);
> +
> + mmput(svm->mm);
> + }
> + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> + if (!sdev) {
> + ret = -ENOMEM;
> + goto out;

I think you need to clean up svm if its device list is empty here, as
you said above:

+ if (svm) {
+ /*
+ * If we found svm for the PASID, there must be at
+ * least one device bond, otherwise svm should be freed.
+ */
+ BUG_ON(list_empty(&svm->devs));
+

> + }
> + sdev->dev = dev;
> + sdev->users = 1;
> +
> + /* Set up device context entry for PASID if not enabled already */
> + ret = intel_iommu_enable_pasid(iommu, sdev->dev);
> + if (ret) {
> + dev_err(dev, "Failed to enable PASID capability\n");
> + kfree(sdev);
> + goto out;
> + }
> +
> + /*
> + * For guest bind, we need to set up PASID table entry as follows:
> + * - FLPM matches guest paging mode
> + * - turn on nested mode
> + * - SL guest address width matching
> + */
> + ret = intel_pasid_setup_nested(iommu,
> + dev,
> + (pgd_t *)data->gpgd,
> + data->hpasid,
> + data->flags,
> + ddomain,
> + data->addr_width);
> + if (ret) {
> + dev_err(dev, "Failed to set up PASID %llu in nested mode, Err %d\n",
> + data->hpasid, ret);
> + kfree(sdev);
> + goto out;
> + }
> + svm->flags |= SVM_FLAG_GUEST_MODE;
> +
> + init_rcu_head(&sdev->rcu);
> + refcount_inc(&svm->refs);
> + list_add_rcu(&sdev->list, &svm->devs);
> + out:
> + mutex_unlock(&pasid_mutex);
> + return ret;
> +}
> +
> +int intel_svm_unbind_gpasid(struct device *dev, int pasid)
> +{
> + struct intel_svm_dev *sdev;
> + struct intel_iommu *iommu;
> + struct intel_svm *svm;
> + int ret = -EINVAL;
> +
> + mutex_lock(&pasid_mutex);
> + iommu = intel_svm_device_to_iommu(dev);
> + if (!iommu)
> + goto out;
> +
> + svm = ioasid_find(NULL, pasid, NULL);
> + if (IS_ERR(svm)) {
> + ret = PTR_ERR(svm);
> + goto out;
> + }
> +
> + if (!svm)
> + goto out;
> +
> + for_each_svm_dev() {
> + ret = 0;
> + sdev->users--;
> + if (!sdev->users) {
> + list_del_rcu(&sdev->list);
> + intel_pasid_tear_down_entry(iommu, dev, svm->pasid);
> + /* TODO: Drain in flight PRQ for the PASID since it
> + * may get reused soon, we don't want to
> + * confuse with its previous live.
> + * intel_svm_drain_prq(dev, pasid);
> + */
> + kfree_rcu(sdev, rcu);
> +
> + if (list_empty(&svm->devs)) {
> + list_del(&svm->list);
> + kfree(svm);
> + /*
> + * We do not free PASID here until explicit call
> + * from VFIO to free. The PASID life cycle
> + * management is largely tied to VFIO management
> + * of assigned device life cycles. In case of
> + * guest exit without a explicit free PASID call,
> + * the responsibility lies in VFIO layer to free
> + * the PASIDs allocated for the guest.
> + * For security reasons, VFIO has to track the
> + * PASID ownership per guest anyway to ensure
> + * that PASID allocated by one guest cannot be
> + * used by another.
> + */
> + ioasid_set_data(pasid, NULL);
> + }
> + }
> + break;
> + }
> + out:
> + mutex_unlock(&pasid_mutex);
> +
> + return ret;
> +}
> +
> int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ops *ops)
> {
> struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> diff --git a/include/linux/intel-iommu.h b/include/linux/intel-iommu.h
> index b75f17d..94d3a9a 100644
> --- a/include/linux/intel-iommu.h
> +++ b/include/linux/intel-iommu.h
> @@ -677,7 +677,9 @@ int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct device *dev);
> int intel_svm_init(struct intel_iommu *iommu);
> extern int intel_svm_enable_prq(struct intel_iommu *iommu);
> extern int intel_svm_finish_prq(struct intel_iommu *iommu);
> -
> +extern int intel_svm_bind_gpasid(struct iommu_domain *domain,
> + struct device *dev, struct gpasid_bind_data *data);
> +extern int intel_svm_unbind_gpasid(struct device *dev, int pasid);
> struct svm_dev_ops;
>
> struct intel_svm_dev {
> @@ -693,12 +695,19 @@ struct intel_svm_dev {
>
> struct intel_svm {
> struct mmu_notifier notifier;
> - struct mm_struct *mm;
> + union {
> + struct mm_struct *mm;
> + u64 gcr3;

I didn't see gcr3 being used anywhere? Anything I missed?

> + };
> struct intel_iommu *iommu;
> int flags;
> int pasid;
> + int gpasid; /* Guest PASID in case of vSVA bind with non-identity host
> + * to guest PASID mapping.
> + */
> struct list_head devs;
> struct list_head list;
> + refcount_t refs; /* Number of devices sharing this PASID */
> };
>
> extern struct intel_iommu *intel_svm_device_to_iommu(struct device *dev);
> diff --git a/include/linux/intel-svm.h b/include/linux/intel-svm.h
> index e3f7631..577d5df 100644
> --- a/include/linux/intel-svm.h
> +++ b/include/linux/intel-svm.h
> @@ -52,6 +52,23 @@ struct svm_dev_ops {
> * do such IOTLB flushes automatically.
> */
> #define SVM_FLAG_SUPERVISOR_MODE (1<<1)
> +/*
> + * The SVM_FLAG_GUEST_MODE flag is used when a guest process bind to a device.
> + * In this case the mm_struct is in the guest kernel or userspace, its life
> + * cycle is managed by VMM and VFIO layer. For IOMMU driver, this API provides
> + * means to bind/unbind guest CR3 with PASIDs allocated for a device.
> + */
> +#define SVM_FLAG_GUEST_MODE (1<<2)
> +/*
> + * The SVM_FLAG_GUEST_PASID flag is used when a guest has its own PASID space,
> + * which requires guest and host PASID translation at both directions. We keep
> + * track of guest PASID in order to provide lookup service to device drivers.
> + * One such example is a physical function (PF) driver that supports mediated
> + * device (mdev) assignment. Guest programming of mdev configuration space can
> + * only be done with guest PASID, therefore PF driver needs to find the matching
> + * host PASID to program the real hardware.
> + */
> +#define SVM_FLAG_GUEST_PASID (1<<3)
>
> #ifdef CONFIG_INTEL_IOMMU_SVM
>
>

Best regards,
Baolu

2019-06-27 20:19:59

by Jacob Pan

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

On Thu, 27 Jun 2019 10:50:11 +0800
Lu Baolu <[email protected]> wrote:

> Hi Jacob and Yi,
>
> On 6/9/19 9:44 PM, Jacob Pan wrote:
> > When supporting guest SVA with emulated IOMMU, the guest PASID
> > table is shadowed in VMM. Updates to guest vIOMMU PASID table
> > will result in PASID cache flush which will be passed down to
> > the host as bind guest PASID calls.
> >
> > For the SL page tables, it will be harvested from device's
> > default domain (request w/o PASID), or aux domain in case of
> > mediated device.
> >
> > .-------------. .---------------------------.
> > | vIOMMU | | Guest process CR3, FL only|
> > | | '---------------------------'
> > .----------------/
> > | PASID Entry |--- PASID cache flush -
> > '-------------' |
> > | | V
> > | | CR3 in GPA
> > '-------------'
> > Guest
> > ------| Shadow |--------------------------|--------
> > v v v
> > Host
> > .-------------. .----------------------.
> > | pIOMMU | | Bind FL for GVA-GPA |
> > | | '----------------------'
> > .----------------/ |
> > | PASID Entry | V (Nested xlate)
> > '----------------\.------------------------------.
> > | | |SL for GPA-HPA, default domain|
> > | | '------------------------------'
> > '-------------'
> > Where:
> > - FL = First level/stage one page tables
> > - SL = Second level/stage two page tables
> >
> > Signed-off-by: Jacob Pan <[email protected]>
> > Signed-off-by: Liu, Yi L <[email protected]>
> > ---
> > drivers/iommu/intel-iommu.c | 4 +
> > drivers/iommu/intel-svm.c | 187
> > ++++++++++++++++++++++++++++++++++++++++++++
> > include/linux/intel-iommu.h | 13 ++- include/linux/intel-svm.h
> > | 17 ++++ 4 files changed, 219 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/iommu/intel-iommu.c
> > b/drivers/iommu/intel-iommu.c index 7cfa0eb..3b4d712 100644
> > --- a/drivers/iommu/intel-iommu.c
> > +++ b/drivers/iommu/intel-iommu.c
> > @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
> > .dev_enable_feat = intel_iommu_dev_enable_feat,
> > .dev_disable_feat = intel_iommu_dev_disable_feat,
> > .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
> > +#ifdef CONFIG_INTEL_IOMMU_SVM
> > + .sva_bind_gpasid = intel_svm_bind_gpasid,
> > + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
> > +#endif
> > };
> >
> > static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
> > diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> > index 66d98e1..f06a82f 100644
> > --- a/drivers/iommu/intel-svm.c
> > +++ b/drivers/iommu/intel-svm.c
> > @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
> > list_for_each_entry(sdev, &svm->devs, list) \
> > if (dev == sdev->dev) \
> >
> > +int intel_svm_bind_gpasid(struct iommu_domain *domain,
> > + struct device *dev,
> > + struct gpasid_bind_data *data)
> > +{
> > + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> > + struct intel_svm_dev *sdev;
> > + struct intel_svm *svm = NULL;
> > + struct dmar_domain *ddomain;
> > + int ret = 0;
> > +
> > + if (WARN_ON(!iommu) || !data)
> > + return -EINVAL;
> > +
> > + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
> > + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
> > + return -EINVAL;
> > +
> > + if (dev_is_pci(dev)) {
> > + /* VT-d supports devices with full 20 bit PASIDs
> > only */
> > + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
> > + return -EINVAL;
> > + }
> > +
> > + /*
> > + * We only check host PASID range, we have no knowledge to
> > check
> > + * guest PASID range nor do we use the guest PASID.
> > + */
> > + if (data->hpasid <= 0 || data->hpasid >= PASID_MAX)
> > + return -EINVAL;
> > +
> > + ddomain = to_dmar_domain(domain);
> > + /* REVISIT:
> > + * Sanity check adddress width and paging mode support
> > + * width matching in two dimensions:
> > + * 1. paging mode CPU <= IOMMU
> > + * 2. address width Guest <= Host.
> > + */
> > + mutex_lock(&pasid_mutex);
> > + svm = ioasid_find(NULL, data->hpasid, NULL);
> > + if (IS_ERR(svm)) {
> > + ret = PTR_ERR(svm);
> > + goto out;
> > + }
> > + if (svm) {
> > + /*
> > + * If we found svm for the PASID, there must be at
> > + * least one device bond, otherwise svm should be
> > freed.
> > + */
> > + BUG_ON(list_empty(&svm->devs));
> > +
> > + for_each_svm_dev() {
> > + /* In case of multiple sub-devices of the
> > same pdev assigned, we should
> > + * allow multiple bind calls with the same
> > PASID and pdev.
> > + */
> > + sdev->users++;
> > + goto out;
> > + }
> > + } else {
> > + /* We come here when PASID has never been bond to
> > a device. */
> > + svm = kzalloc(sizeof(*svm), GFP_KERNEL);
> > + if (!svm) {
> > + ret = -ENOMEM;
> > + goto out;
> > + }
> > + /* REVISIT: upper layer/VFIO can track host
> > process that bind the PASID.
> > + * ioasid_set = mm might be sufficient for vfio to
> > check pasid VMM
> > + * ownership.
> > + */
> > + svm->mm = get_task_mm(current);
> > + svm->pasid = data->hpasid;
> > + if (data->flags & IOMMU_SVA_GPASID_VAL) {
> > + svm->gpasid = data->gpasid;
> > + svm->flags &= SVM_FLAG_GUEST_PASID;
>
> I am guessing that you want to set this flag bit, so it should be
>
> svm->flags |= SVM_FLAG_GUEST_PASID;
>
you are right, this will be used when we have non-identity G-H pasid
mapping. thanks.
> > + }
> > + refcount_set(&svm->refs, 0);
> > + ioasid_set_data(data->hpasid, svm);
> > + INIT_LIST_HEAD_RCU(&svm->devs);
> > + INIT_LIST_HEAD(&svm->list);
> > +
> > + mmput(svm->mm);
> > + }
> > + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> > + if (!sdev) {
> > + ret = -ENOMEM;
> > + goto out;
>
> I think you need to clean up svm if its device list is empty here, as
> you said above:
>
No, we come here only if the device list is not empty and the new
device to bind is different than any existing device in the list. If we
cannot allocate memory for the new device, should not free the existing
SVM, right?

> + if (svm) {
> + /*
> + * If we found svm for the PASID, there must be at
> + * least one device bond, otherwise svm should be
> freed.
> + */
> + BUG_ON(list_empty(&svm->devs));
> +
>
> > + }
> > + sdev->dev = dev;
> > + sdev->users = 1;
> > +
> > + /* Set up device context entry for PASID if not enabled
> > already */
> > + ret = intel_iommu_enable_pasid(iommu, sdev->dev);
> > + if (ret) {
> > + dev_err(dev, "Failed to enable PASID
> > capability\n");
> > + kfree(sdev);
> > + goto out;
> > + }
> > +
> > + /*
> > + * For guest bind, we need to set up PASID table entry as
> > follows:
> > + * - FLPM matches guest paging mode
> > + * - turn on nested mode
> > + * - SL guest address width matching
> > + */
> > + ret = intel_pasid_setup_nested(iommu,
> > + dev,
> > + (pgd_t *)data->gpgd,
> > + data->hpasid,
> > + data->flags,
> > + ddomain,
> > + data->addr_width);
> > + if (ret) {
> > + dev_err(dev, "Failed to set up PASID %llu in
> > nested mode, Err %d\n",
> > + data->hpasid, ret);
> > + kfree(sdev);
> > + goto out;
> > + }
> > + svm->flags |= SVM_FLAG_GUEST_MODE;
> > +
> > + init_rcu_head(&sdev->rcu);
> > + refcount_inc(&svm->refs);
> > + list_add_rcu(&sdev->list, &svm->devs);
> > + out:
> > + mutex_unlock(&pasid_mutex);
> > + return ret;
> > +}
> > +
> > +int intel_svm_unbind_gpasid(struct device *dev, int pasid)
> > +{
> > + struct intel_svm_dev *sdev;
> > + struct intel_iommu *iommu;
> > + struct intel_svm *svm;
> > + int ret = -EINVAL;
> > +
> > + mutex_lock(&pasid_mutex);
> > + iommu = intel_svm_device_to_iommu(dev);
> > + if (!iommu)
> > + goto out;
> > +
> > + svm = ioasid_find(NULL, pasid, NULL);
> > + if (IS_ERR(svm)) {
> > + ret = PTR_ERR(svm);
> > + goto out;
> > + }
> > +
> > + if (!svm)
> > + goto out;
> > +
> > + for_each_svm_dev() {
> > + ret = 0;
> > + sdev->users--;
> > + if (!sdev->users) {
> > + list_del_rcu(&sdev->list);
> > + intel_pasid_tear_down_entry(iommu, dev,
> > svm->pasid);
> > + /* TODO: Drain in flight PRQ for the PASID
> > since it
> > + * may get reused soon, we don't want to
> > + * confuse with its previous live.
> > + * intel_svm_drain_prq(dev, pasid);
> > + */
> > + kfree_rcu(sdev, rcu);
> > +
> > + if (list_empty(&svm->devs)) {
> > + list_del(&svm->list);
> > + kfree(svm);
> > + /*
> > + * We do not free PASID here until
> > explicit call
> > + * from VFIO to free. The PASID
> > life cycle
> > + * management is largely tied to
> > VFIO management
> > + * of assigned device life cycles.
> > In case of
> > + * guest exit without a explicit
> > free PASID call,
> > + * the responsibility lies in VFIO
> > layer to free
> > + * the PASIDs allocated for the
> > guest.
> > + * For security reasons, VFIO has
> > to track the
> > + * PASID ownership per guest
> > anyway to ensure
> > + * that PASID allocated by one
> > guest cannot be
> > + * used by another.
> > + */
> > + ioasid_set_data(pasid, NULL);
> > + }
> > + }
> > + break;
> > + }
> > + out:
> > + mutex_unlock(&pasid_mutex);
> > +
> > + return ret;
> > +}
> > +
> > int intel_svm_bind_mm(struct device *dev, int *pasid, int flags,
> > struct svm_dev_ops *ops) {
> > struct intel_iommu *iommu =
> > intel_svm_device_to_iommu(dev); diff --git
> > a/include/linux/intel-iommu.h b/include/linux/intel-iommu.h index
> > b75f17d..94d3a9a 100644 --- a/include/linux/intel-iommu.h
> > +++ b/include/linux/intel-iommu.h
> > @@ -677,7 +677,9 @@ int intel_iommu_enable_pasid(struct intel_iommu
> > *iommu, struct device *dev); int intel_svm_init(struct intel_iommu
> > *iommu); extern int intel_svm_enable_prq(struct intel_iommu *iommu);
> > extern int intel_svm_finish_prq(struct intel_iommu *iommu);
> > -
> > +extern int intel_svm_bind_gpasid(struct iommu_domain *domain,
> > + struct device *dev, struct gpasid_bind_data *data);
> > +extern int intel_svm_unbind_gpasid(struct device *dev, int pasid);
> > struct svm_dev_ops;
> >
> > struct intel_svm_dev {
> > @@ -693,12 +695,19 @@ struct intel_svm_dev {
> >
> > struct intel_svm {
> > struct mmu_notifier notifier;
> > - struct mm_struct *mm;
> > + union {
> > + struct mm_struct *mm;
> > + u64 gcr3;
>
> I didn't see gcr3 being used anywhere? Anything I missed?
>
You are right, not used for now. We directly used guest PGD from passed
in parameters to program first level pointer. I guess we don't need to
store it. I will remove it. Thanks.

> > + };
> > struct intel_iommu *iommu;
> > int flags;
> > int pasid;
> > + int gpasid; /* Guest PASID in case of vSVA bind with
> > non-identity host
> > + * to guest PASID mapping.
> > + */
> > struct list_head devs;
> > struct list_head list;
> > + refcount_t refs; /* Number of devices sharing this PASID */
> > };
> >
> > extern struct intel_iommu *intel_svm_device_to_iommu(struct
> > device *dev); diff --git a/include/linux/intel-svm.h
> > b/include/linux/intel-svm.h index e3f7631..577d5df 100644
> > --- a/include/linux/intel-svm.h
> > +++ b/include/linux/intel-svm.h
> > @@ -52,6 +52,23 @@ struct svm_dev_ops {
> > * do such IOTLB flushes automatically.
> > */
> > #define SVM_FLAG_SUPERVISOR_MODE (1<<1)
> > +/*
> > + * The SVM_FLAG_GUEST_MODE flag is used when a guest process bind
> > to a device.
> > + * In this case the mm_struct is in the guest kernel or userspace,
> > its life
> > + * cycle is managed by VMM and VFIO layer. For IOMMU driver, this
> > API provides
> > + * means to bind/unbind guest CR3 with PASIDs allocated for a
> > device.
> > + */
> > +#define SVM_FLAG_GUEST_MODE (1<<2)
> > +/*
> > + * The SVM_FLAG_GUEST_PASID flag is used when a guest has its own
> > PASID space,
> > + * which requires guest and host PASID translation at both
> > directions. We keep
> > + * track of guest PASID in order to provide lookup service to
> > device drivers.
> > + * One such example is a physical function (PF) driver that
> > supports mediated
> > + * device (mdev) assignment. Guest programming of mdev
> > configuration space can
> > + * only be done with guest PASID, therefore PF driver needs to
> > find the matching
> > + * host PASID to program the real hardware.
> > + */
> > +#define SVM_FLAG_GUEST_PASID (1<<3)
> >
> > #ifdef CONFIG_INTEL_IOMMU_SVM
> >
> >
>
> Best regards,
> Baolu

Thank you!
[Jacob Pan]

2019-07-05 02:36:19

by Lu Baolu

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

Hi Jacob,

On 6/28/19 4:22 AM, Jacob Pan wrote:
>>> + }
>>> + refcount_set(&svm->refs, 0);
>>> + ioasid_set_data(data->hpasid, svm);
>>> + INIT_LIST_HEAD_RCU(&svm->devs);
>>> + INIT_LIST_HEAD(&svm->list);
>>> +
>>> + mmput(svm->mm);
>>> + }
>>> + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
>>> + if (!sdev) {
>>> + ret = -ENOMEM;
>>> + goto out;
>> I think you need to clean up svm if its device list is empty here, as
>> you said above:
>>
> No, we come here only if the device list is not empty and the new
> device to bind is different than any existing device in the list. If we
> cannot allocate memory for the new device, should not free the existing
> SVM, right?
>

I'm sorry, but the code doesn't show this. We come here even an svm data
structure was newly created with an empty device list. I post the code
below to ensure that we are reading a same piece of code.

mutex_lock(&pasid_mutex);
svm = ioasid_find(NULL, data->hpasid, NULL);
if (IS_ERR(svm)) {
ret = PTR_ERR(svm);
goto out;
}
if (svm) {
/*
* If we found svm for the PASID, there must be at
* least one device bond, otherwise svm should be freed.
*/
BUG_ON(list_empty(&svm->devs));

for_each_svm_dev() {
/* In case of multiple sub-devices of the same
pdev assigned, we should
* allow multiple bind calls with the same
PASID and pdev.
*/
sdev->users++;
goto out;
}
} else {
/* We come here when PASID has never been bond to a
device. */
svm = kzalloc(sizeof(*svm), GFP_KERNEL);
if (!svm) {
ret = -ENOMEM;
goto out;
}
/* REVISIT: upper layer/VFIO can track host process
that bind the PASID.
* ioasid_set = mm might be sufficient for vfio to
check pasid VMM
* ownership.
*/
svm->mm = get_task_mm(current);
svm->pasid = data->hpasid;
if (data->flags & IOMMU_SVA_GPASID_VAL) {
svm->gpasid = data->gpasid;
svm->flags &= SVM_FLAG_GUEST_PASID;
}
refcount_set(&svm->refs, 0);
ioasid_set_data(data->hpasid, svm);
INIT_LIST_HEAD_RCU(&svm->devs);
INIT_LIST_HEAD(&svm->list);

mmput(svm->mm);
}
sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
if (!sdev) {
ret = -ENOMEM;
goto out;
}
sdev->dev = dev;
sdev->users = 1;

Best regards,
Baolu

2019-07-16 16:46:45

by Eric Auger

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

Hi Jacob,

On 6/9/19 3:44 PM, Jacob Pan wrote:
> When supporting guest SVA with emulated IOMMU, the guest PASID
> table is shadowed in VMM. Updates to guest vIOMMU PASID table
> will result in PASID cache flush which will be passed down to
> the host as bind guest PASID calls.
>
> For the SL page tables, it will be harvested from device's
> default domain (request w/o PASID), or aux domain in case of
> mediated device.
>
> .-------------. .---------------------------.
> | vIOMMU | | Guest process CR3, FL only|
> | | '---------------------------'
> .----------------/
> | PASID Entry |--- PASID cache flush -
> '-------------' |
> | | V
> | | CR3 in GPA
> '-------------'
> Guest
> ------| Shadow |--------------------------|--------
> v v v
> Host
> .-------------. .----------------------.
> | pIOMMU | | Bind FL for GVA-GPA |
> | | '----------------------'
> .----------------/ |
> | PASID Entry | V (Nested xlate)
> '----------------\.------------------------------.
> | | |SL for GPA-HPA, default domain|
> | | '------------------------------'
> '-------------'
> Where:
> - FL = First level/stage one page tables
> - SL = Second level/stage two page tables
>
> Signed-off-by: Jacob Pan <[email protected]>
> Signed-off-by: Liu, Yi L <[email protected]>
> ---
> drivers/iommu/intel-iommu.c | 4 +
> drivers/iommu/intel-svm.c | 187 ++++++++++++++++++++++++++++++++++++++++++++
> include/linux/intel-iommu.h | 13 ++-
> include/linux/intel-svm.h | 17 ++++
> 4 files changed, 219 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
> index 7cfa0eb..3b4d712 100644
> --- a/drivers/iommu/intel-iommu.c
> +++ b/drivers/iommu/intel-iommu.c
> @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
> .dev_enable_feat = intel_iommu_dev_enable_feat,
> .dev_disable_feat = intel_iommu_dev_disable_feat,
> .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
> +#ifdef CONFIG_INTEL_IOMMU_SVM
> + .sva_bind_gpasid = intel_svm_bind_gpasid,
> + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
> +#endif
> };
>
> static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
> diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> index 66d98e1..f06a82f 100644
> --- a/drivers/iommu/intel-svm.c
> +++ b/drivers/iommu/intel-svm.c
> @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
> list_for_each_entry(sdev, &svm->devs, list) \
> if (dev == sdev->dev) \
>
> +int intel_svm_bind_gpasid(struct iommu_domain *domain,
> + struct device *dev,
> + struct gpasid_bind_data *data)
> +{
> + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> + struct intel_svm_dev *sdev;
> + struct intel_svm *svm = NULL;
not requested
> + struct dmar_domain *ddomain;
> + int ret = 0;
> +
> + if (WARN_ON(!iommu) || !data)
> + return -EINVAL;
> +
> + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
> + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
> + return -EINVAL;
> +
> + if (dev_is_pci(dev)) {
> + /* VT-d supports devices with full 20 bit PASIDs only */
> + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
> + return -EINVAL;
> + }
> +
> + /*
> + * We only check host PASID range, we have no knowledge to check
> + * guest PASID range nor do we use the guest PASID.
guest pasid is set below in svm->gpasid.
So I am confused, do you handle gpasid or not? If you don't use gpasid
at the moment, then you may return -EINVAL if data->flags &
IOMMU_SVA_GPASID_VAL.

I confess I don't really understand gpasid as I thought you use
enlightened PASID allocation to use a system wide PASID allocator on
host so in which case guest and host PASID do differ?
> + */
> + if (data->hpasid <= 0 || data->hpasid >= PASID_MAX)
> + return -EINVAL;
> +
> + ddomain = to_dmar_domain(domain);
> + /* REVISIT:
> + * Sanity check adddress width and paging mode support
> + * width matching in two dimensions:
> + * 1. paging mode CPU <= IOMMU
> + * 2. address width Guest <= Host.
> + */
> + mutex_lock(&pasid_mutex);
> + svm = ioasid_find(NULL, data->hpasid, NULL);
> + if (IS_ERR(svm)) {
> + ret = PTR_ERR(svm);
> + goto out;
> + }
> + if (svm) {
> + /*
> + * If we found svm for the PASID, there must be at
> + * least one device bond, otherwise svm should be freed.
> + */
> + BUG_ON(list_empty(&svm->devs));
> +
> + for_each_svm_dev() {
> + /* In case of multiple sub-devices of the same pdev assigned, we should
> + * allow multiple bind calls with the same PASID and pdev.
> + */
> + sdev->users++;
> + goto out;
> + }
> + } else {
> + /* We come here when PASID has never been bond to a device. */
> + svm = kzalloc(sizeof(*svm), GFP_KERNEL);
> + if (!svm) {
> + ret = -ENOMEM;
> + goto out;
> + }
> + /* REVISIT: upper layer/VFIO can track host process that bind the PASID.
> + * ioasid_set = mm might be sufficient for vfio to check pasid VMM
> + * ownership.
> + */
> + svm->mm = get_task_mm(current);
> + svm->pasid = data->hpasid;
> + if (data->flags & IOMMU_SVA_GPASID_VAL) {
> + svm->gpasid = data->gpasid;
> + svm->flags &= SVM_FLAG_GUEST_PASID;
I guest you mean |= ?
> + }
> + refcount_set(&svm->refs, 0);
> + ioasid_set_data(data->hpasid, svm);
> + INIT_LIST_HEAD_RCU(&svm->devs);
> + INIT_LIST_HEAD(&svm->list);
> +
> + mmput(svm->mm);
> + }
> + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> + if (!sdev) {
> + ret = -ENOMEM;
> + goto out;
from now on, if any error occurs below and if you've just created the
svm object, no sdev will be attached to the svm object => opposed to
comment above and likely to hit BUG_ON() on next call.
> + }
> + sdev->dev = dev;
> + sdev->users = 1;
> +
> + /* Set up device context entry for PASID if not enabled already */
> + ret = intel_iommu_enable_pasid(iommu, sdev->dev);
> + if (ret) {
> + dev_err(dev, "Failed to enable PASID capability\n");
> + kfree(sdev);
> + goto out;
> + }
> +
> + /*
> + * For guest bind, we need to set up PASID table entry as follows:
> + * - FLPM matches guest paging mode
> + * - turn on nested mode
> + * - SL guest address width matching
> + */
> + ret = intel_pasid_setup_nested(iommu,
> + dev,
> + (pgd_t *)data->gpgd,
> + data->hpasid,
> + data->flags,
> + ddomain,
> + data->addr_width);
So at the moment you don't seem to use the data->vtd field and pass it
to intel_pasid_setup_nested(). Don't you need to perform some sanity
checks on some fields against the extended capability reg?


wouldn't it be simpler to pass a struct gpasid_bind_data pointer to
intel_pasid_setup_nested.
> + if (ret) {
> + dev_err(dev, "Failed to set up PASID %llu in nested mode, Err %d\n",
> + data->hpasid, ret);
> + kfree(sdev);
> + goto out;
> + }
> + svm->flags |= SVM_FLAG_GUEST_MODE;
> +
> + init_rcu_head(&sdev->rcu);
> + refcount_inc(&svm->refs);
> + list_add_rcu(&sdev->list, &svm->devs);
> + out:
> + mutex_unlock(&pasid_mutex);
> + return ret;
> +}
> +
> +int intel_svm_unbind_gpasid(struct device *dev, int pasid)
> +{
> + struct intel_svm_dev *sdev;
> + struct intel_iommu *iommu;
> + struct intel_svm *svm;
> + int ret = -EINVAL;
> +
> + mutex_lock(&pasid_mutex);
> + iommu = intel_svm_device_to_iommu(dev);
> + if (!iommu)
> + goto out;
> +
> + svm = ioasid_find(NULL, pasid, NULL);
> + if (IS_ERR(svm)) {
> + ret = PTR_ERR(svm);
> + goto out;
> + }
> +
> + if (!svm)
> + goto out;
IS_ERR_OR_NULL as advised by Lu in another patch
> +
> + for_each_svm_dev() {
> + ret = 0;
> + sdev->users--;
> + if (!sdev->users) {
> + list_del_rcu(&sdev->list);
> + intel_pasid_tear_down_entry(iommu, dev, svm->pasid);
> + /* TODO: Drain in flight PRQ for the PASID since it
> + * may get reused soon, we don't want to
> + * confuse with its previous live.
life
> + * intel_svm_drain_prq(dev, pasid);
> + */
> + kfree_rcu(sdev, rcu);
> +
> + if (list_empty(&svm->devs)) {
> + list_del(&svm->list);
> + kfree(svm);
> + /*
> + * We do not free PASID here until explicit call
> + * from VFIO to free. The PASID life cycle
> + * management is largely tied to VFIO management
> + * of assigned device life cycles. In case of
> + * guest exit without a explicit free PASID call,
> + * the responsibility lies in VFIO layer to free
> + * the PASIDs allocated for the guest.
> + * For security reasons, VFIO has to track the
> + * PASID ownership per guest anyway to ensure
> + * that PASID allocated by one guest cannot be
> + * used by another.
> + */
> + ioasid_set_data(pasid, NULL);
> + }
> + }
> + break;
> + }
> + out:
> + mutex_unlock(&pasid_mutex);
> +
> + return ret;
> +}
> +
> int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ops *ops)
> {
> struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> diff --git a/include/linux/intel-iommu.h b/include/linux/intel-iommu.h
> index b75f17d..94d3a9a 100644
> --- a/include/linux/intel-iommu.h
> +++ b/include/linux/intel-iommu.h
> @@ -677,7 +677,9 @@ int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct device *dev);
> int intel_svm_init(struct intel_iommu *iommu);
> extern int intel_svm_enable_prq(struct intel_iommu *iommu);
> extern int intel_svm_finish_prq(struct intel_iommu *iommu);
> -
> +extern int intel_svm_bind_gpasid(struct iommu_domain *domain,
> + struct device *dev, struct gpasid_bind_data *data);
> +extern int intel_svm_unbind_gpasid(struct device *dev, int pasid);
> struct svm_dev_ops;
>
> struct intel_svm_dev {
> @@ -693,12 +695,19 @@ struct intel_svm_dev {
>
> struct intel_svm {
> struct mmu_notifier notifier;
> - struct mm_struct *mm;
> + union {
> + struct mm_struct *mm;
> + u64 gcr3;
> + };
not related to that patch?
> struct intel_iommu *iommu;
> int flags;
> int pasid;
> + int gpasid; /* Guest PASID in case of vSVA bind with non-identity host
> + * to guest PASID mapping.
> + */
> struct list_head devs;
> struct list_head list;
> + refcount_t refs; /* Number of devices sharing this PASID */
> };
>
> extern struct intel_iommu *intel_svm_device_to_iommu(struct device *dev);
> diff --git a/include/linux/intel-svm.h b/include/linux/intel-svm.h
> index e3f7631..577d5df 100644
> --- a/include/linux/intel-svm.h
> +++ b/include/linux/intel-svm.h
> @@ -52,6 +52,23 @@ struct svm_dev_ops {
> * do such IOTLB flushes automatically.
> */
> #define SVM_FLAG_SUPERVISOR_MODE (1<<1)
> +/*
> + * The SVM_FLAG_GUEST_MODE flag is used when a guest process bind to a device.
> + * In this case the mm_struct is in the guest kernel or userspace, its life
> + * cycle is managed by VMM and VFIO layer. For IOMMU driver, this API provides
> + * means to bind/unbind guest CR3 with PASIDs allocated for a device.
> + */
> +#define SVM_FLAG_GUEST_MODE (1<<2)
> +/*
> + * The SVM_FLAG_GUEST_PASID flag is used when a guest has its own PASID space,
> + * which requires guest and host PASID translation at both directions. We keep
> + * track of guest PASID in order to provide lookup service to device drivers.
> + * One such example is a physical function (PF) driver that supports mediated
> + * device (mdev) assignment. Guest programming of mdev configuration space can
> + * only be done with guest PASID, therefore PF driver needs to find the matching
> + * host PASID to program the real hardware.
> + */
> +#define SVM_FLAG_GUEST_PASID (1<<3)
>
> #ifdef CONFIG_INTEL_IOMMU_SVM
>
>

Thanks

Eric

2019-07-16 17:06:08

by Ashok Raj

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

Hi Eric

Jacob is on sabbatical, so i'll give it my best shot :-)

Yi/Kevin can jump in...

On Tue, Jul 16, 2019 at 06:45:51PM +0200, Auger Eric wrote:
> Hi Jacob,
>
> On 6/9/19 3:44 PM, Jacob Pan wrote:
> > When supporting guest SVA with emulated IOMMU, the guest PASID
> > table is shadowed in VMM. Updates to guest vIOMMU PASID table
> > will result in PASID cache flush which will be passed down to
> > the host as bind guest PASID calls.
> >
> > For the SL page tables, it will be harvested from device's
> > default domain (request w/o PASID), or aux domain in case of
> > mediated device.
> >
> > .-------------. .---------------------------.
> > | vIOMMU | | Guest process CR3, FL only|
> > | | '---------------------------'
> > .----------------/
> > | PASID Entry |--- PASID cache flush -
> > '-------------' |
> > | | V
> > | | CR3 in GPA
> > '-------------'
> > Guest
> > ------| Shadow |--------------------------|--------
> > v v v
> > Host
> > .-------------. .----------------------.
> > | pIOMMU | | Bind FL for GVA-GPA |
> > | | '----------------------'
> > .----------------/ |
> > | PASID Entry | V (Nested xlate)
> > '----------------\.------------------------------.
> > | | |SL for GPA-HPA, default domain|
> > | | '------------------------------'
> > '-------------'
> > Where:
> > - FL = First level/stage one page tables
> > - SL = Second level/stage two page tables
> >
> > Signed-off-by: Jacob Pan <[email protected]>
> > Signed-off-by: Liu, Yi L <[email protected]>
> > ---
> > drivers/iommu/intel-iommu.c | 4 +
> > drivers/iommu/intel-svm.c | 187 ++++++++++++++++++++++++++++++++++++++++++++
> > include/linux/intel-iommu.h | 13 ++-
> > include/linux/intel-svm.h | 17 ++++
> > 4 files changed, 219 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
> > index 7cfa0eb..3b4d712 100644
> > --- a/drivers/iommu/intel-iommu.c
> > +++ b/drivers/iommu/intel-iommu.c
> > @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
> > .dev_enable_feat = intel_iommu_dev_enable_feat,
> > .dev_disable_feat = intel_iommu_dev_disable_feat,
> > .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
> > +#ifdef CONFIG_INTEL_IOMMU_SVM
> > + .sva_bind_gpasid = intel_svm_bind_gpasid,
> > + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
> > +#endif
> > };
> >
> > static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
> > diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
> > index 66d98e1..f06a82f 100644
> > --- a/drivers/iommu/intel-svm.c
> > +++ b/drivers/iommu/intel-svm.c
> > @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
> > list_for_each_entry(sdev, &svm->devs, list) \
> > if (dev == sdev->dev) \
> >
> > +int intel_svm_bind_gpasid(struct iommu_domain *domain,
> > + struct device *dev,
> > + struct gpasid_bind_data *data)
> > +{
> > + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
> > + struct intel_svm_dev *sdev;
> > + struct intel_svm *svm = NULL;
> not requested
> > + struct dmar_domain *ddomain;
> > + int ret = 0;
> > +
> > + if (WARN_ON(!iommu) || !data)
> > + return -EINVAL;
> > +
> > + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
> > + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
> > + return -EINVAL;
> > +
> > + if (dev_is_pci(dev)) {
> > + /* VT-d supports devices with full 20 bit PASIDs only */
> > + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
> > + return -EINVAL;
> > + }
> > +
> > + /*
> > + * We only check host PASID range, we have no knowledge to check
> > + * guest PASID range nor do we use the guest PASID.
> guest pasid is set below in svm->gpasid.
> So I am confused, do you handle gpasid or not? If you don't use gpasid
> at the moment, then you may return -EINVAL if data->flags &
> IOMMU_SVA_GPASID_VAL.
>
> I confess I don't really understand gpasid as I thought you use
> enlightened PASID allocation to use a system wide PASID allocator on
> host so in which case guest and host PASID do differ?

Correct, from within the guest, we only use the enlightned allocator.

Guest PASID can be managed via Qemu, and it will also associate
guest pasid's with appropriate host pasids. Sort of how guest bdf
and host bdf are managed for page-request etc.

2019-07-18 07:48:35

by Eric Auger

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

Hi Ashok,

On 7/16/19 7:04 PM, Raj, Ashok wrote:
> Hi Eric
>
> Jacob is on sabbatical, so i'll give it my best shot :-)
>
> Yi/Kevin can jump in...
>
> On Tue, Jul 16, 2019 at 06:45:51PM +0200, Auger Eric wrote:
>> Hi Jacob,
>>
>> On 6/9/19 3:44 PM, Jacob Pan wrote:
>>> When supporting guest SVA with emulated IOMMU, the guest PASID
>>> table is shadowed in VMM. Updates to guest vIOMMU PASID table
>>> will result in PASID cache flush which will be passed down to
>>> the host as bind guest PASID calls.
>>>
>>> For the SL page tables, it will be harvested from device's
>>> default domain (request w/o PASID), or aux domain in case of
>>> mediated device.
>>>
>>> .-------------. .---------------------------.
>>> | vIOMMU | | Guest process CR3, FL only|
>>> | | '---------------------------'
>>> .----------------/
>>> | PASID Entry |--- PASID cache flush -
>>> '-------------' |
>>> | | V
>>> | | CR3 in GPA
>>> '-------------'
>>> Guest
>>> ------| Shadow |--------------------------|--------
>>> v v v
>>> Host
>>> .-------------. .----------------------.
>>> | pIOMMU | | Bind FL for GVA-GPA |
>>> | | '----------------------'
>>> .----------------/ |
>>> | PASID Entry | V (Nested xlate)
>>> '----------------\.------------------------------.
>>> | | |SL for GPA-HPA, default domain|
>>> | | '------------------------------'
>>> '-------------'
>>> Where:
>>> - FL = First level/stage one page tables
>>> - SL = Second level/stage two page tables
>>>
>>> Signed-off-by: Jacob Pan <[email protected]>
>>> Signed-off-by: Liu, Yi L <[email protected]>
>>> ---
>>> drivers/iommu/intel-iommu.c | 4 +
>>> drivers/iommu/intel-svm.c | 187 ++++++++++++++++++++++++++++++++++++++++++++
>>> include/linux/intel-iommu.h | 13 ++-
>>> include/linux/intel-svm.h | 17 ++++
>>> 4 files changed, 219 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/iommu/intel-iommu.c b/drivers/iommu/intel-iommu.c
>>> index 7cfa0eb..3b4d712 100644
>>> --- a/drivers/iommu/intel-iommu.c
>>> +++ b/drivers/iommu/intel-iommu.c
>>> @@ -5782,6 +5782,10 @@ const struct iommu_ops intel_iommu_ops = {
>>> .dev_enable_feat = intel_iommu_dev_enable_feat,
>>> .dev_disable_feat = intel_iommu_dev_disable_feat,
>>> .pgsize_bitmap = INTEL_IOMMU_PGSIZES,
>>> +#ifdef CONFIG_INTEL_IOMMU_SVM
>>> + .sva_bind_gpasid = intel_svm_bind_gpasid,
>>> + .sva_unbind_gpasid = intel_svm_unbind_gpasid,
>>> +#endif
>>> };
>>>
>>> static void quirk_iommu_g4x_gfx(struct pci_dev *dev)
>>> diff --git a/drivers/iommu/intel-svm.c b/drivers/iommu/intel-svm.c
>>> index 66d98e1..f06a82f 100644
>>> --- a/drivers/iommu/intel-svm.c
>>> +++ b/drivers/iommu/intel-svm.c
>>> @@ -229,6 +229,193 @@ static LIST_HEAD(global_svm_list);
>>> list_for_each_entry(sdev, &svm->devs, list) \
>>> if (dev == sdev->dev) \
>>>
>>> +int intel_svm_bind_gpasid(struct iommu_domain *domain,
>>> + struct device *dev,
>>> + struct gpasid_bind_data *data)
>>> +{
>>> + struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
>>> + struct intel_svm_dev *sdev;
>>> + struct intel_svm *svm = NULL;
>> not requested
>>> + struct dmar_domain *ddomain;
>>> + int ret = 0;
>>> +
>>> + if (WARN_ON(!iommu) || !data)
>>> + return -EINVAL;
>>> +
>>> + if (data->version != IOMMU_GPASID_BIND_VERSION_1 ||
>>> + data->format != IOMMU_PASID_FORMAT_INTEL_VTD)
>>> + return -EINVAL;
>>> +
>>> + if (dev_is_pci(dev)) {
>>> + /* VT-d supports devices with full 20 bit PASIDs only */
>>> + if (pci_max_pasids(to_pci_dev(dev)) != PASID_MAX)
>>> + return -EINVAL;
>>> + }
>>> +
>>> + /*
>>> + * We only check host PASID range, we have no knowledge to check
>>> + * guest PASID range nor do we use the guest PASID.
>> guest pasid is set below in svm->gpasid.
>> So I am confused, do you handle gpasid or not? If you don't use gpasid
>> at the moment, then you may return -EINVAL if data->flags &
>> IOMMU_SVA_GPASID_VAL.
>>
>> I confess I don't really understand gpasid as I thought you use
>> enlightened PASID allocation to use a system wide PASID allocator on
>> host so in which case guest and host PASID do differ?
>
> Correct, from within the guest, we only use the enlightned allocator.
>
> Guest PASID can be managed via Qemu, and it will also associate
> guest pasid's with appropriate host pasids. Sort of how guest bdf
> and host bdf are managed for page-request etc.

OK I will look at the QEMU side then.

Thanks

Eric
>

2019-08-14 17:17:45

by Jacob Pan

[permalink] [raw]
Subject: Re: [PATCH v4 20/22] iommu/vt-d: Add bind guest PASID support

On Fri, 5 Jul 2019 10:21:27 +0800
Lu Baolu <[email protected]> wrote:

> Hi Jacob,
>
> On 6/28/19 4:22 AM, Jacob Pan wrote:
> >>> + }
> >>> + refcount_set(&svm->refs, 0);
> >>> + ioasid_set_data(data->hpasid, svm);
> >>> + INIT_LIST_HEAD_RCU(&svm->devs);
> >>> + INIT_LIST_HEAD(&svm->list);
> >>> +
> >>> + mmput(svm->mm);
> >>> + }
> >>> + sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> >>> + if (!sdev) {
> >>> + ret = -ENOMEM;
> >>> + goto out;
> >> I think you need to clean up svm if its device list is empty here,
> >> as you said above:
> >>
> > No, we come here only if the device list is not empty and the new
> > device to bind is different than any existing device in the list.
> > If we cannot allocate memory for the new device, should not free
> > the existing SVM, right?
> >
>
> I'm sorry, but the code doesn't show this. We come here even an svm
> data structure was newly created with an empty device list. I post
> the code below to ensure that we are reading a same piece of code.
>
Sorry for the delay. You are right, I need to clean up svm if device
list is empty.

Thanks!
> mutex_lock(&pasid_mutex);
> svm = ioasid_find(NULL, data->hpasid, NULL);
> if (IS_ERR(svm)) {
> ret = PTR_ERR(svm);
> goto out;
> }
> if (svm) {
> /*
> * If we found svm for the PASID, there must be at
> * least one device bond, otherwise svm should be
> freed. */
> BUG_ON(list_empty(&svm->devs));
>
> for_each_svm_dev() {
> /* In case of multiple sub-devices of the
> same pdev assigned, we should
> * allow multiple bind calls with the same
> PASID and pdev.
> */
> sdev->users++;
> goto out;
> }
> } else {
> /* We come here when PASID has never been bond to a
> device. */
> svm = kzalloc(sizeof(*svm), GFP_KERNEL);
> if (!svm) {
> ret = -ENOMEM;
> goto out;
> }
> /* REVISIT: upper layer/VFIO can track host process
> that bind the PASID.
> * ioasid_set = mm might be sufficient for vfio to
> check pasid VMM
> * ownership.
> */
> svm->mm = get_task_mm(current);
> svm->pasid = data->hpasid;
> if (data->flags & IOMMU_SVA_GPASID_VAL) {
> svm->gpasid = data->gpasid;
> svm->flags &= SVM_FLAG_GUEST_PASID;
> }
> refcount_set(&svm->refs, 0);
> ioasid_set_data(data->hpasid, svm);
> INIT_LIST_HEAD_RCU(&svm->devs);
> INIT_LIST_HEAD(&svm->list);
>
> mmput(svm->mm);
> }
> sdev = kzalloc(sizeof(*sdev), GFP_KERNEL);
> if (!sdev) {
> ret = -ENOMEM;
> goto out;
> }
> sdev->dev = dev;
> sdev->users = 1;
>
> Best regards,
> Baolu

[Jacob Pan]