iommu/sva: Tighten SVA bind API with explicit flags

drivers/iommu/iommu-sva-lib.c +6 -5

index bd41405..bd99f6b 100644
@@ -12,27 +12,33 @@ static DECLARE_IOASID_SET(iommu_sva_pasid);
 
 /**
  * iommu_sva_alloc_pasid - Allocate a PASID for the mm
- * @mm: the mm
  * @min: minimum PASID value (inclusive)
  * @max: maximum PASID value (inclusive)
  *
- * Try to allocate a PASID for this mm, or take a reference to the existing one
- * provided it fits within the [@min, @max] range. On success the PASID is
- * available in mm->pasid, and must be released with iommu_sva_free_pasid().
+ * Try to allocate a PASID for the current mm, or take a reference to the
+ * existing one provided it fits within the [@min, @max] range. On success
+ * the PASID is available in the current mm->pasid, and must be released with
+ * iommu_sva_free_pasid().
  * @min must be greater than 0, because 0 indicates an unused mm->pasid.
  *
  * Returns 0 on success and < 0 on error.
  */
-int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max)
+int iommu_sva_alloc_pasid(ioasid_t min, ioasid_t max)
 {
 	int ret = 0;
 	ioasid_t pasid;
+	struct mm_struct *mm;
 

drivers/iommu/iommu-sva-lib.c +3 -1

 	if (mm->pasid) {
 		if (mm->pasid >= min && mm->pasid <= max)
 			ioasid_get(mm->pasid);

@@ -45,22 +51,32 @@ int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max)
 		else
 			mm->pasid = pasid;
 	}
+	mmput(mm);
+out_unlock:
 	mutex_unlock(&iommu_sva_lock);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(iommu_sva_alloc_pasid);
 
 /**
- * iommu_sva_free_pasid - Release the mm's PASID
+ * iommu_sva_free_pasid - Release the current mm's PASID
  * @mm: the mm
  *
  * Drop one reference to a PASID allocated with iommu_sva_alloc_pasid()
  */

drivers/iommu/iommu-sva-lib.c +2 -0

 	if (ioasid_put(mm->pasid))
 		mm->pasid = 0;
+	mmput(mm);
+out_unlock:
 	mutex_unlock(&iommu_sva_lock);
 }
 EXPORT_SYMBOL_GPL(iommu_sva_free_pasid);

drivers/iommu/iommu-sva-lib.h +2 -2

index b40990a..278b8b4 100644
@@ -8,8 +8,8 @@
 #include <linux/ioasid.h>
 #include <linux/mm_types.h>
 
-int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max);
-void iommu_sva_free_pasid(struct mm_struct *mm);
+int iommu_sva_alloc_pasid(ioasid_t min, ioasid_t max);
+void iommu_sva_free_pasid(void);
 struct mm_struct *iommu_sva_find(ioasid_t pasid);
 
 #endif /* _IOMMU_SVA_LIB_H */

drivers/iommu/iommu.c +1 -0

index bf0a20f..25840e6 100644
@@ -23,6 +23,7 @@
 #include <linux/property.h>
 #include <linux/fsl/mc.h>
 #include <linux/module.h>
+#include <linux/sched/mm.h>
 #include <trace/events/iommu.h>
 
 static struct kset *iommu_group_kset;

drivers/iommu/iommu.c +9 -3

@@ -2975,9 +2975,10 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
  * On error, returns an ERR_PTR value.
  */
 struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags)
+iommu_sva_bind_device(struct device *dev, unsigned int flags)
 {
 	struct iommu_group *group;
+	struct mm_struct *mm = NULL;
 	struct iommu_sva *handle = ERR_PTR(-EINVAL);
 	const struct iommu_ops *ops = dev->bus->iommu_ops;
 

@@ -2989,8 +2990,11 @@ iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int fla
 		return ERR_PTR(-ENODEV);
 
 	/* Supervisor SVA does not need the current mm */
-	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && mm)
-		return ERR_PTR(-EINVAL);
+	if (!(flags & IOMMU_SVA_BIND_SUPERVISOR)) {
+		mm = get_task_mm(current);
+		if (!mm)
+			return ERR_PTR(-EINVAL);
+	}
 	/* Ensure device count and domain don't change while we're binding */
 	mutex_lock(&group->mutex);
 

@@ -3004,6 +3008,8 @@ iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int fla
 		goto out_unlock;
 
 	handle = ops->sva_bind(dev, mm, flags);
+	if (mm)
+		mmput(mm);
 out_unlock:
 	mutex_unlock(&group->mutex);
 	iommu_group_put(group);

drivers/misc/uacce/uacce.c +1 -1

index 27e0e04..da4401a 100644
@@ -99,7 +99,7 @@ static int uacce_bind_queue(struct uacce_device *uacce, struct uacce_queue *q)
 	if (!(uacce->flags & UACCE_DEV_SVA))
 		return 0;
 
-	handle = iommu_sva_bind_device(uacce->parent, current->mm, 0);
+	handle = iommu_sva_bind_device(uacce->parent, 0);
 	if (IS_ERR(handle))
 		return PTR_ERR(handle);
 

include/linux/iommu.h +0 -0

index a3fbaa2..cf752f3 100644

include/linux/iommu.h +1 -2

  * @sva_get_pasid: Get PASID associated to a SVA handle
  * @page_response: handle page request response
  * @cache_invalidate: invalidate translation caches

@@ -652,7 +652,6 @@ void iommu_aux_detach_device(struct iommu_domain *domain, struct device *dev);
 int iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev);
 
 struct iommu_sva *iommu_sva_bind_device(struct device *dev,
-					struct mm_struct *mm,
 					unsigned int flags);
 void iommu_sva_unbind_device(struct iommu_sva *handle);
 u32 iommu_sva_get_pasid(struct iommu_sva *handle);

@@ -1028,7 +1027,7 @@ iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev)
 }
 
 static inline struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags)
+iommu_sva_bind_device(struct device *dev, unsigned int flags)
 {
 	return NULL;
 }

Signature · 1 line

2.7.4

Quoted 40 lines · Jacob Pan

(quoted diff) +0 -0

b/drivers/iommu/iommu-sva-lib.c index bd41405..bd99f6b 100644

drivers/iommu/iommu-sva-lib.c +0 -2

@@ -12,27 +12,33 @@ static DECLARE_IOASID_SET(iommu_sva_pasid);
 
 /**
  * iommu_sva_alloc_pasid - Allocate a PASID for the mm
- * @mm: the mm
  * @min: minimum PASID value (inclusive)
  * @max: maximum PASID value (inclusive)
  *
- * Try to allocate a PASID for this mm, or take a reference to the

existing one
- * provided it fits within the [@min, @max] range. On success the PASID is
- * available in mm->pasid, and must be released with iommu_sva_free_pasid().
+ * Try to allocate a PASID for the current mm, or take a reference to the
+ * existing one provided it fits within the [@min, @max] range. On success
+ * the PASID is available in the current mm->pasid, and must be released with
+ * iommu_sva_free_pasid().
* @min must be greater than 0, because 0 indicates an unused
mm->pasid. *
* Returns 0 on success and < 0 on error.
*/
-int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max) +int iommu_sva_alloc_pasid(ioasid_t min, ioasid_t max)
{

drivers/iommu/iommu-sva-lib.c +1 -0

 	int ret = 0;
 	ioasid_t pasid;
+	struct mm_struct *mm;
 
 	if (min == INVALID_IOASID || max == INVALID_IOASID ||
 	    min == 0 || max < min)
 		return -EINVAL;

drivers/iommu/iommu-sva-lib.c static DECLARE_IOASID_SET(iommu_sva_pasid); · line 41

Quoted 6 lines

drivers/iommu/iommu-sva-lib.c +5 -0

 	mutex_lock(&iommu_sva_lock);
+	mm = get_task_mm(current);
+	if (!mm) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}  

I still think it would be more elegant to keep the choice of context in iommu_sva_bind_device() and pass it down to leaf functions such as iommu_sva_alloc_pasid(). The patch is trying to solve two separate

problems:

* We don't have a use-case for binding the mm of a remote process (and it's supposedly difficult for device drivers to do it securely). So OK, we remove the mm argument from iommu_sva_bind_device() and use the current mm. But the IOMMU driver isn't going to do get_task_mm(current) every time it needs the mm being bound, it will take it from iommu_sva_bind_device(). Likewise iommu_sva_alloc_pasid() shouldn't need to bother with get_task_mm().

* cgroup accounting for IOASIDs needs to be on the current task. Removing the mm parameter from iommu_sva_alloc_pasid() doesn't help with that. Sure it indicates that iommu_sva_alloc_pasid() needs a specific task context but that's only for cgroup purpose, and I'd rather pass the cgroup down from iommu_sva_bind_device() anyway (but am fine with keeping it within ioasid_alloc() for now). Plus it's an internal helper, easy for us to check that the callers are doing the right thing.

Quoted 21 lines

drivers/iommu/iommu-sva-lib.c +0 -0

 	if (mm->pasid) {
 		if (mm->pasid >= min && mm->pasid <= max)
 			ioasid_get(mm->pasid);

@@ -45,22 +51,32 @@ int iommu_sva_alloc_pasid(struct mm_struct *mm,

ioasid_t min, ioasid_t max) else

drivers/iommu/iommu-sva-lib.c +3 -1

 			mm->pasid = pasid;
 	}
+	mmput(mm);
+out_unlock:
 	mutex_unlock(&iommu_sva_lock);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(iommu_sva_alloc_pasid);
 
 /**
- * iommu_sva_free_pasid - Release the mm's PASID
+ * iommu_sva_free_pasid - Release the current mm's PASID
  * @mm: the mm
  *
  * Drop one reference to a PASID allocated with iommu_sva_alloc_pasid()
  */

drivers/iommu/iommu-sva-lib.c int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max) · line 75

Quoted 10 lines

drivers/iommu/iommu-sva-lib.c +7 -1

-void iommu_sva_free_pasid(struct mm_struct *mm)
+void iommu_sva_free_pasid(void)
 {
+	struct mm_struct *mm;
+
 	mutex_lock(&iommu_sva_lock);
+	mm = get_task_mm(current);
+	if (!mm)
+		goto out_unlock;
+  

More importantly, could we at least dissociate free_pasid() from the current process? Otherwise drivers can't clean up from a workqueue (as amdkfd does) or from an rcu callback. Given that iommu_sva_unbind_device() takes the SVA handle owned by whomever did bind(), there shouldn't be any security issue. For the cgroup problem, ioasid.c could internally keep track of the cgroup used during allocation rather than assuming the context of ioasid_put() is the same as ioasid_get()

Quoted 37 lines

drivers/iommu/iommu-sva-lib.c +2 -0

 	if (ioasid_put(mm->pasid))
 		mm->pasid = 0;
+	mmput(mm);
+out_unlock:
 	mutex_unlock(&iommu_sva_lock);
 }
 EXPORT_SYMBOL_GPL(iommu_sva_free_pasid);

(quoted diff) +0 -0

b/drivers/iommu/iommu-sva-lib.h index b40990a..278b8b4 100644

drivers/iommu/iommu-sva-lib.h +2 -1

@@ -8,8 +8,8 @@
 #include <linux/ioasid.h>
 #include <linux/mm_types.h>
 
-int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t
 max); -void iommu_sva_free_pasid(struct mm_struct *mm);
+int iommu_sva_alloc_pasid(ioasid_t min, ioasid_t max);
+void iommu_sva_free_pasid(void);
 struct mm_struct *iommu_sva_find(ioasid_t pasid);
 
 #endif /* _IOMMU_SVA_LIB_H */

drivers/iommu/iommu.c +1 -0

index bf0a20f..25840e6 100644
@@ -23,6 +23,7 @@
 #include <linux/property.h>
 #include <linux/fsl/mc.h>
 #include <linux/module.h>
+#include <linux/sched/mm.h>
 #include <trace/events/iommu.h>
 
 static struct kset *iommu_group_kset;

@@ -2959,9 +2960,8 @@ int iommu_aux_get_pasid(struct iommu_domain
 *domain, struct device *dev) EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
 

Quoted 10 lines

drivers/iommu/iommu.c +1 -1

 /**
- * iommu_sva_bind_device() - Bind a process address space to a device
+ * iommu_sva_bind_device() - Bind the current process address space to

a device
* @dev: the device
- * @mm: the mm to bind, caller must hold a reference to it
* @flags: options for the bind operation
*
* Create a bond between device and address space, allowing the device
to access

There is another reference to @mm to remove in the function description

Quoted 48 lines

drivers/iommu/iommu.c +0 -1

@@ -2975,9 +2975,10 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
  * On error, returns an ERR_PTR value.
  */
 struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm,

unsigned int flags) +iommu_sva_bind_device(struct device *dev, unsigned int flags) {

drivers/iommu/iommu.c +8 -2

 	struct iommu_group *group;
+	struct mm_struct *mm = NULL;
 	struct iommu_sva *handle = ERR_PTR(-EINVAL);
 	const struct iommu_ops *ops = dev->bus->iommu_ops;

@@ -2989,8 +2990,11 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, unsigned int fla return ERR_PTR(-ENODEV);
 
 	/* Supervisor SVA does not need the current mm */
-	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && mm)
-		return ERR_PTR(-EINVAL);
+	if (!(flags & IOMMU_SVA_BIND_SUPERVISOR)) {
+		mm = get_task_mm(current);
+		if (!mm)
+			return ERR_PTR(-EINVAL);
+	}
 	/* Ensure device count and domain don't change while we're
 binding */ mutex_lock(&group->mutex);
 

@@ -3004,6 +3008,8 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, unsigned int fla goto out_unlock;
 
 	handle = ops->sva_bind(dev, mm, flags);
+	if (mm)
+		mmput(mm);
 out_unlock:
 	mutex_unlock(&group->mutex);
 	iommu_group_put(group);

drivers/misc/uacce/uacce.c +1 -1

index 27e0e04..da4401a 100644
@@ -99,7 +99,7 @@ static int uacce_bind_queue(struct uacce_device
 *uacce, struct uacce_queue *q) if (!(uacce->flags & UACCE_DEV_SVA))
 		return 0;
 
-	handle = iommu_sva_bind_device(uacce->parent, current->mm, 0);
+	handle = iommu_sva_bind_device(uacce->parent, 0);
 	if (IS_ERR(handle))
 		return PTR_ERR(handle);
 

include/linux/iommu.h struct iommu_iotlb_gather { · line 235

Quoted 12 lines

include/linux/iommu.h +2 -2

index a3fbaa2..cf752f3 100644
@@ -231,8 +231,8 @@ struct iommu_iotlb_gather {
  * @dev_feat_enabled: check enabled feature
  * @aux_attach/detach_dev: aux-domain specific attach/detach entries.
  * @aux_get_pasid: get the pasid given an aux-domain
- * @sva_bind: Bind process address space to device
- * @sva_unbind: Unbind process address space from device
+ * @sva_bind: Bind the current process address space to device
+ * @sva_unbind: Unbind the current process address space from device  

These don't need changing since we're still passing the mm down to the drivers

Thanks,
Jean

Quoted 14 lines

include/linux/iommu.h +0 -1

  * @sva_get_pasid: Get PASID associated to a SVA handle
  * @page_response: handle page request response
  * @cache_invalidate: invalidate translation caches

@@ -652,7 +652,6 @@ void iommu_aux_detach_device(struct iommu_domain
 *domain, struct device *dev); int iommu_aux_get_pasid(struct
 iommu_domain *domain, struct device *dev); 
 struct iommu_sva *iommu_sva_bind_device(struct device *dev,
-					struct mm_struct *mm,
 					unsigned int flags);
 void iommu_sva_unbind_device(struct iommu_sva *handle);
 u32 iommu_sva_get_pasid(struct iommu_sva *handle);

@@ -1028,7 +1027,7 @@ iommu_aux_get_pasid(struct iommu_domain *domain,
 struct device *dev) }
 

Quoted 9 lines

include/linux/iommu.h +0 -1

 static inline struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm,

unsigned int flags) +iommu_sva_bind_device(struct device *dev, unsigned int flags) {
return NULL;
}

Signature · 1 line

2.7.4

Quoted 10 lines

problems:

* We don't have a use-case for binding the mm of a remote process (and it's supposedly difficult for device drivers to do it securely). So
OK, we remove the mm argument from iommu_sva_bind_device() and use the
current mm. But the IOMMU driver isn't going to do
get_task_mm(current) every time it needs the mm being bound, it will take it from iommu_sva_bind_device(). Likewise iommu_sva_alloc_pasid() shouldn't need to bother with get_task_mm().

Quoted 8 lines

* cgroup accounting for IOASIDs needs to be on the current task. Removing the mm parameter from iommu_sva_alloc_pasid() doesn't help with that. Sure it indicates that iommu_sva_alloc_pasid() needs a specific task context but that's only for cgroup purpose, and I'd rather pass the cgroup down from iommu_sva_bind_device() anyway (but am fine with keeping it within ioasid_alloc() for now). Plus it's an internal helper, easy for us to check that the callers are doing the right thing.

With the above split, we really just have one allocation function:
ioasid_alloc(), so it can manage current cgroup accounting within. Would this work?

Hi Jean,

Quoted 18 lines · Jacob Pan

Quoted 18 lines

problems:

* We don't have a use-case for binding the mm of a remote process (and it's supposedly difficult for device drivers to do it securely). So
OK, we remove the mm argument from iommu_sva_bind_device() and use the
current mm. But the IOMMU driver isn't going to do
get_task_mm(current) every time it needs the mm being bound, it will take it from iommu_sva_bind_device(). Likewise iommu_sva_alloc_pasid() shouldn't need to bother with get_task_mm().

* cgroup accounting for IOASIDs needs to be on the current task. Removing the mm parameter from iommu_sva_alloc_pasid() doesn't help with that. Sure it indicates that iommu_sva_alloc_pasid() needs a specific task context but that's only for cgroup purpose, and I'd rather pass the cgroup down from iommu_sva_bind_device() anyway (but am fine with keeping it within ioasid_alloc() for now). Plus it's an internal helper, easy for us to check that the callers are doing the right thing.

Quoted 3 lines

With the above split, we really just have one allocation function:
ioasid_alloc(), so it can manage current cgroup accounting within. Would this work?

After a few attempts, I don't think the split can work better. I will restore the mm parameter and add a warning if mm != current->mm.

Thanks,

Jacob

I still worry about supervisor pasid allocation.

If we use iommu_sva_alloc_pasid() to allocate a supervisor pasid, which mm should the pasid be set? I've ever thought about passing &init_mm to iommu_sva_alloc_pasid(). But if you add "mm != current->mm", this seems not to work. Or do you prefer a separated interface for supervisor pasid allocation/free?

Quoted 2 lines · Lu Baolu

I still worry about supervisor pasid allocation.

Quoted 5 lines

If we use iommu_sva_alloc_pasid() to allocate a supervisor pasid, which mm should the pasid be set? I've ever thought about passing &init_mm to iommu_sva_alloc_pasid(). But if you add "mm != current->mm", this seems not to work. Or do you prefer a separated interface for supervisor pasid allocation/free?

Without a mm_struct it is not SVA, so don't use SVA APIs for whatever a 'supervisor pasid' is

Hi Jason,

Quoted 10 lines · Jason Gunthorpe

Quoted 7 lines · Lu Baolu

I still worry about supervisor pasid allocation.

If we use iommu_sva_alloc_pasid() to allocate a supervisor pasid, which mm should the pasid be set? I've ever thought about passing &init_mm to iommu_sva_alloc_pasid(). But if you add "mm != current->mm", this seems not to work. Or do you prefer a separated interface for supervisor pasid allocation/free?

Quoted 2 lines

Without a mm_struct it is not SVA, so don't use SVA APIs for whatever a 'supervisor pasid' is

The supervisor PASID has its mm_struct. The only difference is that the device will set priv=1 in its DMA transactions with the PASID.

/**
* iommu_sva_alloc_pasid - Allocate a PASID for the mm
- * @mm: the mm
* @min: minimum PASID value (inclusive)
* @max: maximum PASID value (inclusive)
*
- * Try to allocate a PASID for this mm, or take a reference to the existing one
- * provided it fits within the [@min, @max] range. On success the PASID is
- * available in mm->pasid, and must be released with iommu_sva_free_pasid().
+ * Try to allocate a PASID for the current mm, or take a reference to the
+ * existing one provided it fits within the [@min, @max] range. On success
+ * the PASID is available in the current mm->pasid, and must be released with
+ * iommu_sva_free_pasid().
* @min must be greater than 0, because 0 indicates an unused mm->pasid. *
* Returns 0 on success and < 0 on error.
*/

drivers/iommu/iommu-sva-lib.c +1 -1

-int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max)
+int iommu_sva_alloc_pasid(ioasid_t min, ioasid_t max)
  {

drivers/iommu/iommu-sva-lib.c +4 -0

+	if (!mm) {
+		ret = -EINVAL;
+		goto out_unlock;
+	}
  	if (mm->pasid) {
  		if (mm->pasid >= min && mm->pasid <= max)
  			ioasid_get(mm->pasid);

Quoted 11 lines

-int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max) +int iommu_sva_alloc_pasid(ioasid_t min, ioasid_t max)
{
int ret = 0;
ioasid_t pasid;

drivers/iommu/iommu-sva-lib.c +1 -0

+	struct mm_struct *mm;
 
  	if (min == INVALID_IOASID || max == INVALID_IOASID ||
  	    min == 0 || max < min)
  		return -EINVAL;

Review

Quoted 2 lines

  	mutex_lock(&iommu_sva_lock);

(quoted diff) +1 -0

+	mm = get_task_mm(current);  

How could we allocate a supervisor PASID through iommu_sva_alloc_pasid() if we always use current->mm here?

The void* drvdata parameter isn't really used in iommu_sva_bind_device() API,

the current IDXD code "borrows" the drvdata for a VT-d private flag for supervisor SVA usage.

Supervisor/Privileged mode request is a generic feature. It should be promoted from the VT-d vendor driver to the generic code.

This patch replaces void* drvdata with a unsigned int flags parameter and adjusts callers accordingly.

Link https://lore.kernel.org/linux-iommu/YFhiMLR35WWMW%2FHu@myrica/ open here

Suggested-by Jean-Philippe Brucker <[email protected]>

Signed-off-by Jacob Pan <[email protected]>

---
 drivers/dma/idxd/cdev.c                         |  2 +-
 drivers/dma/idxd/init.c                         |  6 +++---
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c |  2 +-
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h     |  4 ++--
 drivers/iommu/intel/Kconfig                     |  1 +
 drivers/iommu/intel/svm.c                       | 18 ++++++------------
 drivers/iommu/iommu.c                           |  9 ++++++---
 drivers/misc/uacce/uacce.c                      |  2 +-
 include/linux/intel-iommu.h                     |  2 +-
 include/linux/intel-svm.h                       | 17 ++---------------
 include/linux/iommu.h                           | 19 ++++++++++++++++---
 11 files changed, 40 insertions(+), 42 deletions(-)

drivers/dma/idxd/cdev.c +1 -1

index 0db9b82..21ec82b 100644
@@ -103,7 +103,7 @@ static int idxd_cdev_open(struct inode *inode, struct file *filp)
 	filp->private_data = ctx;
 
 	if (device_pasid_enabled(idxd)) {
-		sva = iommu_sva_bind_device(dev, current->mm, NULL);
+		sva = iommu_sva_bind_device(dev, current->mm, 0);
 		if (IS_ERR(sva)) {
 			rc = PTR_ERR(sva);
 			dev_err(dev, "pasid allocation failed: %d\n", rc);

drivers/dma/idxd/init.c +3 -3

index 085a0c3..cdc85f1 100644
@@ -300,13 +300,13 @@ static struct idxd_device *idxd_alloc(struct pci_dev *pdev)
 
 static int idxd_enable_system_pasid(struct idxd_device *idxd)
 {
-	int flags;
+	unsigned int flags;
 	unsigned int pasid;
 	struct iommu_sva *sva;
 
-	flags = SVM_FLAG_SUPERVISOR_MODE;
+	flags = IOMMU_SVA_BIND_SUPERVISOR;
 
-	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, &flags);
+	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, flags);
 	if (IS_ERR(sva)) {
 		dev_warn(&idxd->pdev->dev,
 			 "iommu sva bind failed: %ld\n", PTR_ERR(sva));

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +0 -0

index bb251ca..23e287e 100644

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +0 -0

 {
 	struct iommu_sva *handle;
 	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h +2 -2

index f985817..b971d4d 100644
@@ -711,7 +711,7 @@ bool arm_smmu_master_sva_enabled(struct arm_smmu_master *master);
 int arm_smmu_master_enable_sva(struct arm_smmu_master *master);
 int arm_smmu_master_disable_sva(struct arm_smmu_master *master);
 struct iommu_sva *arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm,
-				    void *drvdata);
+				    unsigned int flags);
 void arm_smmu_sva_unbind(struct iommu_sva *handle);
 u32 arm_smmu_sva_get_pasid(struct iommu_sva *handle);
 void arm_smmu_sva_notifier_synchronize(void);

@@ -742,7 +742,7 @@ static inline int arm_smmu_master_disable_sva(struct arm_smmu_master *master)
 }
 
 static inline struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, unsigned int flags)
 {
 	return ERR_PTR(-ENODEV);
 }

drivers/iommu/intel/Kconfig +0 -0

index 28a3d15..5415052 100644

drivers/iommu/intel/Kconfig +0 -0

 	help
 	  Shared Virtual Memory (SVM) provides a facility for devices
 	  to access DMA resources through process address space by

drivers/iommu/intel/svm.c +6 -12

index 574a7e6..4b5f8b0 100644
@@ -486,12 +486,9 @@ intel_svm_bind_mm(struct device *dev, unsigned int flags,
 	} else
 		pasid_max = 1 << 20;
 
-	/* Bind supervisor PASID shuld have mm = NULL */
-	if (flags & SVM_FLAG_SUPERVISOR_MODE) {
-		if (!ecap_srs(iommu->ecap) || mm) {
-			pr_err("Supervisor PASID with user provided mm.\n");
-			return -EINVAL;
-		}
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && !ecap_srs(iommu->ecap)) {
+		pr_err("Supervisor PASID not supported.\n");
+		return -EINVAL;
 	}
 
 	if (!(flags & SVM_FLAG_PRIVATE_PASID)) {

@@ -593,7 +590,7 @@ intel_svm_bind_mm(struct device *dev, unsigned int flags,
 		ret = intel_pasid_setup_first_level(iommu, dev,
 				mm ? mm->pgd : init_mm.pgd,
 				svm->pasid, FLPT_DEFAULT_DID,
-				(mm ? 0 : PASID_FLAG_SUPERVISOR_MODE) |
+				(mm ? 0 : IOMMU_SVA_BIND_SUPERVISOR) |
 				(cpu_feature_enabled(X86_FEATURE_LA57) ?
 				 PASID_FLAG_FL5LP : 0));
 		spin_unlock_irqrestore(&iommu->lock, iflags);

@@ -620,7 +617,7 @@ intel_svm_bind_mm(struct device *dev, unsigned int flags,
 		ret = intel_pasid_setup_first_level(iommu, dev,
 						mm ? mm->pgd : init_mm.pgd,
 						svm->pasid, FLPT_DEFAULT_DID,
-						(mm ? 0 : PASID_FLAG_SUPERVISOR_MODE) |
+						(mm ? 0 : IOMMU_SVA_BIND_SUPERVISOR) |
 						(cpu_feature_enabled(X86_FEATURE_LA57) ?
 						PASID_FLAG_FL5LP : 0));
 		spin_unlock_irqrestore(&iommu->lock, iflags);

@@ -1059,11 +1056,10 @@ static irqreturn_t prq_event_thread(int irq, void *d)
 
 #define to_intel_svm_dev(handle) container_of(handle, struct intel_svm_dev, sva)
 struct iommu_sva *
-intel_svm_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+intel_svm_bind(struct device *dev, struct mm_struct *mm, unsigned int flags)
 {
 	struct iommu_sva *sva = ERR_PTR(-EINVAL);
 	struct intel_svm_dev *sdev = NULL;
-	unsigned int flags = 0;
 	int ret;
 
 	/*

@@ -1071,8 +1067,6 @@ intel_svm_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
 	 * It will require shared SVM data structures, i.e. combine io_mm
 	 * and intel_svm etc.
 	 */
-	if (drvdata)
-		flags = *(unsigned int *)drvdata;
 	mutex_lock(&pasid_mutex);
 	ret = intel_svm_bind_mm(dev, flags, NULL, mm, &sdev);
 	if (ret)

drivers/iommu/iommu.c +0 -0

index d0b0a15..bf0a20f 100644

drivers/iommu/iommu.c +5 -3

  *
  * Create a bond between device and address space, allowing the device to access
  * the mm using the returned PASID. If a bond already exists between @device and

@@ -2974,7 +2975,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
  * On error, returns an ERR_PTR value.
  */
 struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
+iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags)
 {
 	struct iommu_group *group;
 	struct iommu_sva *handle = ERR_PTR(-EINVAL);

@@ -2987,6 +2988,9 @@ iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
 	if (!group)
 		return ERR_PTR(-ENODEV);
 
+	/* Supervisor SVA does not need the current mm */
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && mm)
+		return ERR_PTR(-EINVAL);
 	/* Ensure device count and domain don't change while we're binding */
 	mutex_lock(&group->mutex);
 

@@ -2999,8 +3003,7 @@ iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
 	if (iommu_group_device_count(group) != 1)
 		goto out_unlock;
 
-	handle = ops->sva_bind(dev, mm, drvdata);
-
+	handle = ops->sva_bind(dev, mm, flags);
 out_unlock:
 	mutex_unlock(&group->mutex);
 	iommu_group_put(group);

drivers/misc/uacce/uacce.c +1 -1

index d07af4e..27e0e04 100644
@@ -99,7 +99,7 @@ static int uacce_bind_queue(struct uacce_device *uacce, struct uacce_queue *q)
 	if (!(uacce->flags & UACCE_DEV_SVA))
 		return 0;
 
-	handle = iommu_sva_bind_device(uacce->parent, current->mm, NULL);
+	handle = iommu_sva_bind_device(uacce->parent, current->mm, 0);
 	if (IS_ERR(handle))
 		return PTR_ERR(handle);
 

include/linux/intel-iommu.h +1 -1

index 1bc46b8..cdff752 100644
@@ -757,7 +757,7 @@ int intel_svm_bind_gpasid(struct iommu_domain *domain, struct device *dev,
 			  struct iommu_gpasid_bind_data *data);
 int intel_svm_unbind_gpasid(struct device *dev, u32 pasid);
 struct iommu_sva *intel_svm_bind(struct device *dev, struct mm_struct *mm,
-				 void *drvdata);
+				 unsigned int flags);
 void intel_svm_unbind(struct iommu_sva *handle);
 u32 intel_svm_get_pasid(struct iommu_sva *handle);
 int intel_svm_page_response(struct device *dev, struct iommu_fault_event *evt,

include/linux/intel-svm.h +2 -15

index 39d368a..ef6b753 100644
@@ -30,30 +30,17 @@ struct svm_dev_ops {
  * if there is no other way to do so. It should be used sparingly, if at all.
  */
 #define SVM_FLAG_PRIVATE_PASID		(1<<0)
-
-/*
- * The SVM_FLAG_SUPERVISOR_MODE flag requests a PASID which can be used only
- * for access to kernel addresses. No IOTLB flushes are automatically done
- * for kernel mappings; it is valid only for access to the kernel's static
- * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
- * A future API addition may permit the use of such ranges, by means of an
- * explicit IOTLB flush call (akin to the DMA API's unmap method).
- *
- * It is unlikely that we will ever hook into flush_tlb_kernel_range() to
- * do such IOTLB flushes automatically.
- */
-#define SVM_FLAG_SUPERVISOR_MODE	(1<<1)
 /*
  * The SVM_FLAG_GUEST_MODE flag is used when a PASID bind is for guest
  * processes. Compared to the host bind, the primary differences are:
  * 1. mm life cycle management
  * 2. fault reporting
  */
-#define SVM_FLAG_GUEST_MODE		(1<<2)
+#define SVM_FLAG_GUEST_MODE		(1<<1)
 /*
  * 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.
  */
-#define SVM_FLAG_GUEST_PASID		(1<<3)
+#define SVM_FLAG_GUEST_PASID		(1<<2)
 
 #endif /* __INTEL_SVM_H__ */

include/linux/iommu.h +3 -0

index 5e7fe51..a3fbaa2 100644
@@ -166,6 +166,19 @@ enum iommu_dev_features {
 
 #ifdef CONFIG_IOMMU_API
 
+/*
+ * The IOMMU_SVA_BIND_SUPERVISOR flag requests a PASID which can be used only
+ * for access to kernel addresses. No IOTLB flushes are automatically done

include/linux/iommu.h +3 -3

 /**
  * struct iommu_iotlb_gather - Range information for a pending IOTLB flush
  *

@@ -287,7 +300,7 @@ struct iommu_ops {
 	int (*aux_get_pasid)(struct iommu_domain *domain, struct device *dev);
 
 	struct iommu_sva *(*sva_bind)(struct device *dev, struct mm_struct *mm,
-				      void *drvdata);
+				      unsigned int flags);
 	void (*sva_unbind)(struct iommu_sva *handle);
 	u32 (*sva_get_pasid)(struct iommu_sva *handle);
 

@@ -640,7 +653,7 @@ int iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev);
 
 struct iommu_sva *iommu_sva_bind_device(struct device *dev,
 					struct mm_struct *mm,
-					void *drvdata);
+					unsigned int flags);
 void iommu_sva_unbind_device(struct iommu_sva *handle);
 u32 iommu_sva_get_pasid(struct iommu_sva *handle);
 

@@ -1015,7 +1028,7 @@ iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev)
 }
 
 static inline struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
+iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags)
 {
 	return NULL;
 }

Signature · 1 line

2.7.4

Quoted 2 lines · Jacob Pan

The void* drvdata parameter isn't really used in iommu_sva_bind_device() API,

Right, it used to be a cookie passed to the device driver in the exit_mm() callback, but that went away with edcc40d2ab5f ("iommu: Remove iommu_sva_ops::mm_exit()")

Quoted 6 lines

the current IDXD code "borrows" the drvdata for a VT-d private flag for supervisor SVA usage.

Supervisor/Privileged mode request is a generic feature. It should be promoted from the VT-d vendor driver to the generic code.

Quoted 2 lines

This patch replaces void* drvdata with a unsigned int flags parameter and adjusts callers accordingly.

Thanks for cleaning this up. Making flags unsigned long seems more common (I suggested int without thinking). But it doesn't matter much, we won't get to 32 flags.

Quoted 59 lines

Link https://lore.kernel.org/linux-iommu/YFhiMLR35WWMW%2FHu@myrica/ open here

Suggested-by Jean-Philippe Brucker <[email protected]>

Signed-off-by Jacob Pan <[email protected]>

---
 drivers/dma/idxd/cdev.c                         |  2 +-
 drivers/dma/idxd/init.c                         |  6 +++---
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c |  2 +-
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h     |  4 ++--
 drivers/iommu/intel/Kconfig                     |  1 +
 drivers/iommu/intel/svm.c                       | 18 ++++++------------
 drivers/iommu/iommu.c                           |  9 ++++++---
 drivers/misc/uacce/uacce.c                      |  2 +-
 include/linux/intel-iommu.h                     |  2 +-
 include/linux/intel-svm.h                       | 17 ++---------------
 include/linux/iommu.h                           | 19

++++++++++++++++--- 11 files changed, 40 insertions(+), 42 deletions(-)

drivers/dma/idxd/cdev.c +1 -1

index 0db9b82..21ec82b 100644
@@ -103,7 +103,7 @@ static int idxd_cdev_open(struct inode *inode,
 struct file *filp) filp->private_data = ctx;
 
 	if (device_pasid_enabled(idxd)) {
-		sva = iommu_sva_bind_device(dev, current->mm, NULL);
+		sva = iommu_sva_bind_device(dev, current->mm, 0);
 		if (IS_ERR(sva)) {
 			rc = PTR_ERR(sva);
 			dev_err(dev, "pasid allocation failed: %d\n",

rc); diff --git a/drivers/dma/idxd/init.c b/drivers/dma/idxd/init.c index 085a0c3..cdc85f1 100644

drivers/dma/idxd/init.c +3 -3

@@ -300,13 +300,13 @@ static struct idxd_device *idxd_alloc(struct
 pci_dev *pdev) 
 static int idxd_enable_system_pasid(struct idxd_device *idxd)
 {
-	int flags;
+	unsigned int flags;
 	unsigned int pasid;
 	struct iommu_sva *sva;
 
-	flags = SVM_FLAG_SUPERVISOR_MODE;
+	flags = IOMMU_SVA_BIND_SUPERVISOR;
 
-	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, &flags);
+	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, flags);
 	if (IS_ERR(sva)) {
 		dev_warn(&idxd->pdev->dev,
 			 "iommu sva bind failed: %ld\n", PTR_ERR(sva));

(quoted diff) +0 -0

b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c index bb251ca..23e287e 100644 --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c @@ -354,7 +354,7 @@ __arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm) }

Quoted 4 lines

struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void *drvdata) +arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, unsigned int flags)

Could you add a check on flags:

(quoted diff) +0 -0

b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c index bb251cab61f3..145ceb2fc5da 100644 --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c @@ -354,12 +354,15 @@ __arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm) }

struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void *drvdata) +arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, unsigned int flags) {

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +0 -0

 	if (smmu_domain->stage != ARM_SMMU_DOMAIN_S1)
 		return ERR_PTR(-EINVAL);

Quoted 32 lines

{

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +0 -0

 	struct iommu_sva *handle;
 	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);

(quoted diff) +0 -0

b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h index f985817..b971d4d 100644 --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h

(quoted diff) +0 -0

@@ -711,7 +711,7 @@ bool arm_smmu_master_sva_enabled(struct

arm_smmu_master *master); int arm_smmu_master_enable_sva(struct arm_smmu_master *master); int arm_smmu_master_disable_sva(struct arm_smmu_master *master); struct iommu_sva *arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm,

(quoted diff) +1 -2

-				    void *drvdata);
+				    unsigned int flags);
 void arm_smmu_sva_unbind(struct iommu_sva *handle);
 u32 arm_smmu_sva_get_pasid(struct iommu_sva *handle);
 void arm_smmu_sva_notifier_synchronize(void);

@@ -742,7 +742,7 @@ static inline int
 arm_smmu_master_disable_sva(struct arm_smmu_master *master) }
 
 static inline struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void
 *drvdata) +arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm,
 unsigned int flags) {
 	return ERR_PTR(-ENODEV);
 }

drivers/iommu/intel/Kconfig +0 -1

index 28a3d15..5415052 100644
--- a/drivers/iommu/intel/Kconfig

drivers/iommu/intel/Kconfig config INTEL_IOMMU_SVM · line 44

Quoted 6 lines

+++ b/drivers/iommu/intel/Kconfig

drivers/iommu/intel/Kconfig +1 -0

@@ -41,6 +41,7 @@ config INTEL_IOMMU_SVM
 	select PCI_PRI
 	select MMU_NOTIFIER
 	select IOASID
+	select IOMMU_SVA_LIB  

Not needed here?

Quoted 72 lines

drivers/iommu/intel/Kconfig +0 -0

 	help
 	  Shared Virtual Memory (SVM) provides a facility for devices
 	  to access DMA resources through process address space by

drivers/iommu/intel/svm.c +0 -0

index 574a7e6..4b5f8b0 100644
@@ -486,12 +486,9 @@ intel_svm_bind_mm(struct device *dev, unsigned int

flags, } else

drivers/iommu/intel/svm.c +0 -4

 		pasid_max = 1 << 20;
 
-	/* Bind supervisor PASID shuld have mm = NULL */
-	if (flags & SVM_FLAG_SUPERVISOR_MODE) {
-		if (!ecap_srs(iommu->ecap) || mm) {
-			pr_err("Supervisor PASID with user provided

mm.\n");

drivers/iommu/intel/svm.c +1 -2

-			return -EINVAL;
-		}
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) &&

!ecap_srs(iommu->ecap)) {

drivers/iommu/intel/svm.c +3 -2

+		pr_err("Supervisor PASID not supported.\n");
+		return -EINVAL;
 	}
 
 	if (!(flags & SVM_FLAG_PRIVATE_PASID)) {

@@ -593,7 +590,7 @@ intel_svm_bind_mm(struct device *dev, unsigned int
 flags, ret = intel_pasid_setup_first_level(iommu, dev,
 				mm ? mm->pgd : init_mm.pgd,
 				svm->pasid, FLPT_DEFAULT_DID,
-				(mm ? 0 : PASID_FLAG_SUPERVISOR_MODE) |
+				(mm ? 0 : IOMMU_SVA_BIND_SUPERVISOR) |
 				(cpu_feature_enabled(X86_FEATURE_LA57)
 ? PASID_FLAG_FL5LP : 0));
 		spin_unlock_irqrestore(&iommu->lock, iflags);

@@ -620,7 +617,7 @@ intel_svm_bind_mm(struct device *dev, unsigned int
 flags, ret = intel_pasid_setup_first_level(iommu, dev,
 						mm ? mm->pgd :
 init_mm.pgd, svm->pasid, FLPT_DEFAULT_DID,
-						(mm ? 0 :

PASID_FLAG_SUPERVISOR_MODE) |

drivers/iommu/intel/svm.c +1 -0

+						(mm ? 0 :

IOMMU_SVA_BIND_SUPERVISOR) | (cpu_feature_enabled(X86_FEATURE_LA57) ?

drivers/iommu/intel/svm.c +0 -0

 						PASID_FLAG_FL5LP : 0));
 		spin_unlock_irqrestore(&iommu->lock, iflags);

@@ -1059,11 +1056,10 @@ static irqreturn_t prq_event_thread(int irq,
 void *d) 
 #define to_intel_svm_dev(handle) container_of(handle, struct

intel_svm_dev, sva) struct iommu_sva *

drivers/iommu/intel/svm.c +1 -1

-intel_svm_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+intel_svm_bind(struct device *dev, struct mm_struct *mm, unsigned int

flags) {

drivers/iommu/intel/svm.c +0 -1

 	struct iommu_sva *sva = ERR_PTR(-EINVAL);
 	struct intel_svm_dev *sdev = NULL;
-	unsigned int flags = 0;
 	int ret;
 
 	/*

@@ -1071,8 +1067,6 @@ intel_svm_bind(struct device *dev, struct
 mm_struct *mm, void *drvdata)
 	 * It will require shared SVM data structures, i.e. combine

io_mm

drivers/iommu/intel/svm.c +0 -2

 	 * and intel_svm etc.
 	 */
-	if (drvdata)
-		flags = *(unsigned int *)drvdata;
 	mutex_lock(&pasid_mutex);
 	ret = intel_svm_bind_mm(dev, flags, NULL, mm, &sdev);
 	if (ret)

drivers/iommu/iommu.c +0 -1

index d0b0a15..bf0a20f 100644
--- a/drivers/iommu/iommu.c

drivers/iommu/iommu.c EXPORT_SYMBOL_GPL(iommu_aux_get_pasid); · line 2965

Quoted 6 lines

+++ b/drivers/iommu/iommu.c

drivers/iommu/iommu.c +1 -0

@@ -2962,6 +2962,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
  * iommu_sva_bind_device() - Bind a process address space to a device
  * @dev: the device
  * @mm: the mm to bind, caller must hold a reference to it
+ * @flags: options for the bind operation  

Could also specify valid flags "IOMMU_SVA_BIND_*"

Quoted 128 lines

drivers/iommu/iommu.c +0 -0

  *
  * Create a bond between device and address space, allowing the device

to access
* the mm using the returned PASID. If a bond already exists between
@device and @@ -2974,7 +2975,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
* On error, returns an ERR_PTR value.
*/
struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata) +iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags) {

drivers/iommu/iommu.c +4 -2

 	struct iommu_group *group;
 	struct iommu_sva *handle = ERR_PTR(-EINVAL);

@@ -2987,6 +2988,9 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, void *drvdata) if (!group)
 		return ERR_PTR(-ENODEV);
 
+	/* Supervisor SVA does not need the current mm */
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && mm)
+		return ERR_PTR(-EINVAL);
 	/* Ensure device count and domain don't change while we're
 binding */ mutex_lock(&group->mutex);
 

@@ -2999,8 +3003,7 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, void *drvdata) if (iommu_group_device_count(group) != 1)
 		goto out_unlock;
 
-	handle = ops->sva_bind(dev, mm, drvdata);
-
+	handle = ops->sva_bind(dev, mm, flags);
 out_unlock:
 	mutex_unlock(&group->mutex);
 	iommu_group_put(group);

drivers/misc/uacce/uacce.c +1 -1

index d07af4e..27e0e04 100644
@@ -99,7 +99,7 @@ static int uacce_bind_queue(struct uacce_device
 *uacce, struct uacce_queue *q) if (!(uacce->flags & UACCE_DEV_SVA))
 		return 0;
 
-	handle = iommu_sva_bind_device(uacce->parent, current->mm,
 NULL);
+	handle = iommu_sva_bind_device(uacce->parent, current->mm, 0);
 	if (IS_ERR(handle))
 		return PTR_ERR(handle);

include/linux/intel-iommu.h +1 -1

index 1bc46b8..cdff752 100644
@@ -757,7 +757,7 @@ int intel_svm_bind_gpasid(struct iommu_domain
 *domain, struct device *dev, struct iommu_gpasid_bind_data *data);
 int intel_svm_unbind_gpasid(struct device *dev, u32 pasid);
 struct iommu_sva *intel_svm_bind(struct device *dev, struct mm_struct
 *mm,
-				 void *drvdata);
+				 unsigned int flags);
 void intel_svm_unbind(struct iommu_sva *handle);
 u32 intel_svm_get_pasid(struct iommu_sva *handle);

int intel_svm_page_response(struct device *dev, struct
iommu_fault_event *evt, diff --git a/include/linux/intel-svm.h b/include/linux/intel-svm.h index 39d368a..ef6b753 100644

include/linux/intel-svm.h +0 -0

@@ -30,30 +30,17 @@ struct svm_dev_ops {
  * if there is no other way to do so. It should be used sparingly, if

at all. */
#define SVM_FLAG_PRIVATE_PASID (1<<0)

include/linux/intel-svm.h +0 -3

-
-/*
- * The SVM_FLAG_SUPERVISOR_MODE flag requests a PASID which can be

used only
- * for access to kernel addresses. No IOTLB flushes are automatically done
- * for kernel mappings; it is valid only for access to the kernel's static
- * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
- * A future API addition may permit the use of such ranges, by means of an
- * explicit IOTLB flush call (akin to the DMA API's unmap method).
- *
- * It is unlikely that we will ever hook into flush_tlb_kernel_range() to
- * do such IOTLB flushes automatically.
- */

include/linux/intel-svm.h +1 -2

-#define SVM_FLAG_SUPERVISOR_MODE	(1<<1)
 /*
  * The SVM_FLAG_GUEST_MODE flag is used when a PASID bind is for guest
  * processes. Compared to the host bind, the primary differences are:
  * 1. mm life cycle management
  * 2. fault reporting
  */
-#define SVM_FLAG_GUEST_MODE		(1<<2)
+#define SVM_FLAG_GUEST_MODE		(1<<1)
 /*
  * 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. */

include/linux/intel-svm.h +1 -1

-#define SVM_FLAG_GUEST_PASID		(1<<3)
+#define SVM_FLAG_GUEST_PASID		(1<<2)
 
 #endif /* __INTEL_SVM_H__ */

include/linux/iommu.h +2 -0

index 5e7fe51..a3fbaa2 100644
@@ -166,6 +166,19 @@ enum iommu_dev_features {
 
 #ifdef CONFIG_IOMMU_API
 
+/*
+ * The IOMMU_SVA_BIND_SUPERVISOR flag requests a PASID which can be

used only
+ * for access to kernel addresses. No IOTLB flushes are automatically done
+ * for kernel mappings; it is valid only for access to the kernel's static
+ * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
+ * A future API addition may permit the use of such ranges, by means of an
+ * explicit IOTLB flush call (akin to the DMA API's unmap method).
+ *
+ * It is unlikely that we will ever hook into flush_tlb_kernel_range()

include/linux/iommu.h

Quoted 5 lines

to
+ * do such IOTLB flushes automatically.
+ */

include/linux/iommu.h +2 -0

+#define IOMMU_SVA_BIND_SUPERVISOR       BIT(0)
+  

It needs to be defined before the #ifdef CONFIG_IOMMU_API, otherwise drivers using it won't build with !CONFIG_IOMMU_API

Thanks,
Jean

Quoted 25 lines

/**
* struct iommu_iotlb_gather - Range information for a pending IOTLB
flush *

include/linux/iommu.h +2 -2

@@ -287,7 +300,7 @@ struct iommu_ops {
 	int (*aux_get_pasid)(struct iommu_domain *domain, struct
 device *dev); 
 	struct iommu_sva *(*sva_bind)(struct device *dev, struct
 mm_struct *mm,
-				      void *drvdata);
+				      unsigned int flags);
 	void (*sva_unbind)(struct iommu_sva *handle);
 	u32 (*sva_get_pasid)(struct iommu_sva *handle);

@@ -640,7 +653,7 @@ int iommu_aux_get_pasid(struct iommu_domain
 *domain, struct device *dev); 
 struct iommu_sva *iommu_sva_bind_device(struct device *dev,
 					struct mm_struct *mm,
-					void *drvdata);
+					unsigned int flags);
 void iommu_sva_unbind_device(struct iommu_sva *handle);
 u32 iommu_sva_get_pasid(struct iommu_sva *handle);
 

@@ -1015,7 +1028,7 @@ iommu_aux_get_pasid(struct iommu_domain *domain,
 struct device *dev) }
 

Quoted 9 lines

include/linux/iommu.h +0 -1

 static inline struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void
 *drvdata) +iommu_sva_bind_device(struct device *dev, struct mm_struct
 *mm, unsigned int flags) {
 	return NULL;
 }

Signature · 1 line

2.7.4

The void* drvdata parameter isn't really used in iommu_sva_bind_device() API, the current IDXD code "borrows" the drvdata for a VT-d private flag for supervisor SVA usage.

Supervisor/Privileged mode request is a generic feature. It should be promoted from the VT-d vendor driver to the generic code.

This patch replaces void* drvdata with a unsigned int flags parameter and adjusts callers accordingly.

Link https://lore.kernel.org/linux-iommu/YFhiMLR35WWMW%2FHu@myrica/ open here

Suggested-by Jean-Philippe Brucker <[email protected]>

Signed-off-by Jacob Pan <[email protected]>

---
  drivers/dma/idxd/cdev.c                         |  2 +-
  drivers/dma/idxd/init.c                         |  6 +++---
  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c |  2 +-
  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h     |  4 ++--
  drivers/iommu/intel/Kconfig                     |  1 +
  drivers/iommu/intel/svm.c                       | 18 ++++++------------
  drivers/iommu/iommu.c                           |  9 ++++++---
  drivers/misc/uacce/uacce.c                      |  2 +-
  include/linux/intel-iommu.h                     |  2 +-
  include/linux/intel-svm.h                       | 17 ++---------------
  include/linux/iommu.h                           | 19 ++++++++++++++++---
  11 files changed, 40 insertions(+), 42 deletions(-)

drivers/dma/idxd/cdev.c +1 -1

index 0db9b82..21ec82b 100644
@@ -103,7 +103,7 @@ static int idxd_cdev_open(struct inode *inode, struct file *filp)
  	filp->private_data = ctx;
 
  	if (device_pasid_enabled(idxd)) {
-		sva = iommu_sva_bind_device(dev, current->mm, NULL);
+		sva = iommu_sva_bind_device(dev, current->mm, 0);
  		if (IS_ERR(sva)) {
  			rc = PTR_ERR(sva);
  			dev_err(dev, "pasid allocation failed: %d\n", rc);

drivers/dma/idxd/init.c +0 -0

index 085a0c3..cdc85f1 100644

drivers/dma/idxd/init.c +1 -1

-	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, &flags);
+	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, flags);
  	if (IS_ERR(sva)) {
  		dev_warn(&idxd->pdev->dev,
  			 "iommu sva bind failed: %ld\n", PTR_ERR(sva));

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +1 -1

index bb251ca..23e287e 100644
@@ -354,7 +354,7 @@ __arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm)
  }
 
  struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, unsigned int flags)
  {
  	struct iommu_sva *handle;
  	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);

drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h +2 -2

index f985817..b971d4d 100644
@@ -711,7 +711,7 @@ bool arm_smmu_master_sva_enabled(struct arm_smmu_master *master);
  int arm_smmu_master_enable_sva(struct arm_smmu_master *master);
  int arm_smmu_master_disable_sva(struct arm_smmu_master *master);
  struct iommu_sva *arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm,
-				    void *drvdata);
+				    unsigned int flags);
  void arm_smmu_sva_unbind(struct iommu_sva *handle);
  u32 arm_smmu_sva_get_pasid(struct iommu_sva *handle);
  void arm_smmu_sva_notifier_synchronize(void);

@@ -742,7 +742,7 @@ static inline int arm_smmu_master_disable_sva(struct arm_smmu_master *master)
  }
 
  static inline struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, unsigned int flags)
  {
  	return ERR_PTR(-ENODEV);
  }

drivers/iommu/intel/Kconfig +0 -0

index 28a3d15..5415052 100644

drivers/iommu/intel/Kconfig +0 -0

  	help
  	  Shared Virtual Memory (SVM) provides a facility for devices
  	  to access DMA resources through process address space by

drivers/iommu/intel/svm.c +3 -6

index 574a7e6..4b5f8b0 100644
@@ -486,12 +486,9 @@ intel_svm_bind_mm(struct device *dev, unsigned int flags,
  	} else
  		pasid_max = 1 << 20;
 
-	/* Bind supervisor PASID shuld have mm = NULL */
-	if (flags & SVM_FLAG_SUPERVISOR_MODE) {
-		if (!ecap_srs(iommu->ecap) || mm) {
-			pr_err("Supervisor PASID with user provided mm.\n");
-			return -EINVAL;
-		}
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && !ecap_srs(iommu->ecap)) {
+		pr_err("Supervisor PASID not supported.\n");
+		return -EINVAL;
  	}
 
  	if (!(flags & SVM_FLAG_PRIVATE_PASID)) {

drivers/iommu/intel/svm.c +1 -4

  						(cpu_feature_enabled(X86_FEATURE_LA57) ?
  						PASID_FLAG_FL5LP : 0));
  		spin_unlock_irqrestore(&iommu->lock, iflags);

@@ -1059,11 +1056,10 @@ static irqreturn_t prq_event_thread(int irq, void *d)
 
  #define to_intel_svm_dev(handle) container_of(handle, struct intel_svm_dev, sva)
  struct iommu_sva *
-intel_svm_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+intel_svm_bind(struct device *dev, struct mm_struct *mm, unsigned int flags)
  {
  	struct iommu_sva *sva = ERR_PTR(-EINVAL);
  	struct intel_svm_dev *sdev = NULL;
-	unsigned int flags = 0;
  	int ret;
 
  	/*

@@ -1071,8 +1067,6 @@ intel_svm_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
  	 * It will require shared SVM data structures, i.e. combine io_mm
  	 * and intel_svm etc.
  	 */
-	if (drvdata)
-		flags = *(unsigned int *)drvdata;
  	mutex_lock(&pasid_mutex);
  	ret = intel_svm_bind_mm(dev, flags, NULL, mm, &sdev);
  	if (ret)

drivers/iommu/iommu.c +6 -3

index d0b0a15..bf0a20f 100644
@@ -2962,6 +2962,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
   * iommu_sva_bind_device() - Bind a process address space to a device
   * @dev: the device
   * @mm: the mm to bind, caller must hold a reference to it
+ * @flags: options for the bind operation
   *
   * Create a bond between device and address space, allowing the device to access
   * the mm using the returned PASID. If a bond already exists between @device and

@@ -2974,7 +2975,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
   * On error, returns an ERR_PTR value.
   */
  struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
+iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags)
  {
  	struct iommu_group *group;
  	struct iommu_sva *handle = ERR_PTR(-EINVAL);

@@ -2987,6 +2988,9 @@ iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
  	if (!group)
  		return ERR_PTR(-ENODEV);
 
+	/* Supervisor SVA does not need the current mm */
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && mm)
+		return ERR_PTR(-EINVAL);
  	/* Ensure device count and domain don't change while we're binding */
  	mutex_lock(&group->mutex);
 

@@ -2999,8 +3003,7 @@ iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata)
  	if (iommu_group_device_count(group) != 1)
  		goto out_unlock;
 
-	handle = ops->sva_bind(dev, mm, drvdata);
-
+	handle = ops->sva_bind(dev, mm, flags);
  out_unlock:
  	mutex_unlock(&group->mutex);
  	iommu_group_put(group);

drivers/misc/uacce/uacce.c +1 -1

index d07af4e..27e0e04 100644
@@ -99,7 +99,7 @@ static int uacce_bind_queue(struct uacce_device *uacce, struct uacce_queue *q)
  	if (!(uacce->flags & UACCE_DEV_SVA))
  		return 0;
 
-	handle = iommu_sva_bind_device(uacce->parent, current->mm, NULL);
+	handle = iommu_sva_bind_device(uacce->parent, current->mm, 0);
  	if (IS_ERR(handle))
  		return PTR_ERR(handle);
 

include/linux/intel-iommu.h +1 -1

index 1bc46b8..cdff752 100644
@@ -757,7 +757,7 @@ int intel_svm_bind_gpasid(struct iommu_domain *domain, struct device *dev,
  			  struct iommu_gpasid_bind_data *data);
  int intel_svm_unbind_gpasid(struct device *dev, u32 pasid);
  struct iommu_sva *intel_svm_bind(struct device *dev, struct mm_struct *mm,
-				 void *drvdata);
+				 unsigned int flags);
  void intel_svm_unbind(struct iommu_sva *handle);
  u32 intel_svm_get_pasid(struct iommu_sva *handle);
  int intel_svm_page_response(struct device *dev, struct iommu_fault_event *evt,

include/linux/intel-svm.h +2 -15

index 39d368a..ef6b753 100644
@@ -30,30 +30,17 @@ struct svm_dev_ops {
   * if there is no other way to do so. It should be used sparingly, if at all.
   */
  #define SVM_FLAG_PRIVATE_PASID		(1<<0)
-
-/*
- * The SVM_FLAG_SUPERVISOR_MODE flag requests a PASID which can be used only
- * for access to kernel addresses. No IOTLB flushes are automatically done
- * for kernel mappings; it is valid only for access to the kernel's static
- * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
- * A future API addition may permit the use of such ranges, by means of an
- * explicit IOTLB flush call (akin to the DMA API's unmap method).
- *
- * It is unlikely that we will ever hook into flush_tlb_kernel_range() to
- * do such IOTLB flushes automatically.
- */
-#define SVM_FLAG_SUPERVISOR_MODE	(1<<1)
  /*
   * The SVM_FLAG_GUEST_MODE flag is used when a PASID bind is for guest
   * processes. Compared to the host bind, the primary differences are:
   * 1. mm life cycle management
   * 2. fault reporting
   */
-#define SVM_FLAG_GUEST_MODE		(1<<2)
+#define SVM_FLAG_GUEST_MODE		(1<<1)
  /*
   * 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.
   */
-#define SVM_FLAG_GUEST_PASID		(1<<3)
+#define SVM_FLAG_GUEST_PASID		(1<<2)
 
  #endif /* __INTEL_SVM_H__ */

include/linux/iommu.h +15 -2

index 5e7fe51..a3fbaa2 100644
@@ -166,6 +166,19 @@ enum iommu_dev_features {
 
  #ifdef CONFIG_IOMMU_API
 
+/*
+ * The IOMMU_SVA_BIND_SUPERVISOR flag requests a PASID which can be used only
+ * for access to kernel addresses. No IOTLB flushes are automatically done
+ * for kernel mappings; it is valid only for access to the kernel's static
+ * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
+ * A future API addition may permit the use of such ranges, by means of an
+ * explicit IOTLB flush call (akin to the DMA API's unmap method).
+ *
+ * It is unlikely that we will ever hook into flush_tlb_kernel_range() to
+ * do such IOTLB flushes automatically.
+ */
+#define IOMMU_SVA_BIND_SUPERVISOR       BIT(0)
+
  /**
   * struct iommu_iotlb_gather - Range information for a pending IOTLB flush
   *

@@ -287,7 +300,7 @@ struct iommu_ops {
  	int (*aux_get_pasid)(struct iommu_domain *domain, struct device *dev);
 
  	struct iommu_sva *(*sva_bind)(struct device *dev, struct mm_struct *mm,
-				      void *drvdata);
+				      unsigned int flags);
  	void (*sva_unbind)(struct iommu_sva *handle);
  	u32 (*sva_get_pasid)(struct iommu_sva *handle);
 

@@ -640,7 +653,7 @@ int iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev);
 
  struct iommu_sva *iommu_sva_bind_device(struct device *dev,
  					struct mm_struct *mm,
-					void *drvdata);
+					unsigned int flags);
  void iommu_sva_unbind_device(struct iommu_sva *handle);
  u32 iommu_sva_get_pasid(struct iommu_sva *handle);
 

Hi Jacob,

Quoted 53 lines · Jacob Pan

The void* drvdata parameter isn't really used in iommu_sva_bind_device() API, the current IDXD code "borrows" the drvdata for a VT-d private flag for supervisor SVA usage.

Supervisor/Privileged mode request is a generic feature. It should be promoted from the VT-d vendor driver to the generic code.

This patch replaces void* drvdata with a unsigned int flags parameter and adjusts callers accordingly.

Link https://lore.kernel.org/linux-iommu/YFhiMLR35WWMW%2FHu@myrica/ open here

Suggested-by Jean-Philippe Brucker <[email protected]>

Signed-off-by Jacob Pan <[email protected]>

---
  drivers/dma/idxd/cdev.c                         |  2 +-
  drivers/dma/idxd/init.c                         |  6 +++---
  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c |  2 +-
  drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h     |  4 ++--
  drivers/iommu/intel/Kconfig                     |  1 +
  drivers/iommu/intel/svm.c                       | 18
++++++------------ drivers/iommu/iommu.c                           |  9
++++++--- drivers/misc/uacce/uacce.c                      |  2 +-
  include/linux/intel-iommu.h                     |  2 +-
  include/linux/intel-svm.h                       | 17 ++---------------
  include/linux/iommu.h                           | 19

++++++++++++++++--- 11 files changed, 40 insertions(+), 42 deletions(-)

drivers/dma/idxd/cdev.c +1 -1

index 0db9b82..21ec82b 100644
@@ -103,7 +103,7 @@ static int idxd_cdev_open(struct inode *inode,
 struct file *filp) filp->private_data = ctx;
 
  	if (device_pasid_enabled(idxd)) {
-		sva = iommu_sva_bind_device(dev, current->mm, NULL);
+		sva = iommu_sva_bind_device(dev, current->mm, 0);
  		if (IS_ERR(sva)) {
  			rc = PTR_ERR(sva);
  			dev_err(dev, "pasid allocation failed: %d\n",

rc); diff --git a/drivers/dma/idxd/init.c b/drivers/dma/idxd/init.c index 085a0c3..cdc85f1 100644

drivers/dma/idxd/init.c +1 -1

@@ -300,13 +300,13 @@ static struct idxd_device *idxd_alloc(struct
 pci_dev *pdev) 
  static int idxd_enable_system_pasid(struct idxd_device *idxd)
  {
-	int flags;
+	unsigned int flags;
  	unsigned int pasid;
  	struct iommu_sva *sva;
 

drivers/dma/idxd/init.c static struct idxd_device *idxd_alloc(struct pci_dev *pdev) · line 307

Quoted 2 lines

drivers/dma/idxd/init.c +1 -1

-	flags = SVM_FLAG_SUPERVISOR_MODE;
+	flags = IOMMU_SVA_BIND_SUPERVISOR;  

With SVM_FLAG_SUPERVISOR_MODE removed, I guess

#include <linux/intel-svm.h>

in this file could be removed as well.

Quoted 46 lines

drivers/dma/idxd/init.c +1 -1

-	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, &flags);
+	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, flags);
  	if (IS_ERR(sva)) {
  		dev_warn(&idxd->pdev->dev,
  			 "iommu sva bind failed: %ld\n",

PTR_ERR(sva)); diff --git a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c index bb251ca..23e287e 100644 --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c @@ -354,7 +354,7 @@ __arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm) }
struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void *drvdata) +arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, unsigned int flags) {
struct iommu_sva *handle;
struct iommu_domain *domain = iommu_get_domain_for_dev(dev);

(quoted diff) +0 -0

b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h index f985817..b971d4d 100644 --- a/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h +++ b/drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h

drivers/dma/idxd/init.c +0 -0

@@ -711,7 +711,7 @@ bool arm_smmu_master_sva_enabled(struct

arm_smmu_master *master); int arm_smmu_master_enable_sva(struct arm_smmu_master *master); int arm_smmu_master_disable_sva(struct arm_smmu_master *master); struct iommu_sva *arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm,

drivers/dma/idxd/init.c +1 -2

-				    void *drvdata);
+				    unsigned int flags);
  void arm_smmu_sva_unbind(struct iommu_sva *handle);
  u32 arm_smmu_sva_get_pasid(struct iommu_sva *handle);
  void arm_smmu_sva_notifier_synchronize(void);

@@ -742,7 +742,7 @@ static inline int
 arm_smmu_master_disable_sva(struct arm_smmu_master *master) }
 
  static inline struct iommu_sva *
-arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm, void
 *drvdata) +arm_smmu_sva_bind(struct device *dev, struct mm_struct *mm,
 unsigned int flags) {
  	return ERR_PTR(-ENODEV);
  }

drivers/iommu/intel/Kconfig +0 -1

index 28a3d15..5415052 100644
--- a/drivers/iommu/intel/Kconfig

drivers/iommu/intel/Kconfig config INTEL_IOMMU_SVM · line 44

Quoted 6 lines

+++ b/drivers/iommu/intel/Kconfig

drivers/iommu/intel/Kconfig +1 -0

@@ -41,6 +41,7 @@ config INTEL_IOMMU_SVM
  	select PCI_PRI
  	select MMU_NOTIFIER
  	select IOASID
+	select IOMMU_SVA_LIB  

Why?

Quoted 25 lines

drivers/iommu/intel/Kconfig +0 -0

  	help
  	  Shared Virtual Memory (SVM) provides a facility for devices
  	  to access DMA resources through process address space by

drivers/iommu/intel/svm.c +0 -0

index 574a7e6..4b5f8b0 100644
@@ -486,12 +486,9 @@ intel_svm_bind_mm(struct device *dev, unsigned int

flags, } else
pasid_max = 1 << 20;

drivers/iommu/intel/svm.c +0 -4

-	/* Bind supervisor PASID shuld have mm = NULL */
-	if (flags & SVM_FLAG_SUPERVISOR_MODE) {
-		if (!ecap_srs(iommu->ecap) || mm) {
-			pr_err("Supervisor PASID with user provided

mm.\n");

drivers/iommu/intel/svm.c +1 -2

-			return -EINVAL;
-		}
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) &&

!ecap_srs(iommu->ecap)) {

drivers/iommu/intel/svm.c +2 -0

+		pr_err("Supervisor PASID not supported.\n");
+		return -EINVAL;
  	}

drivers/iommu/intel/svm.c intel_svm_bind_mm(struct device *dev, unsigned int · line 594

Quoted 8 lines

  	if (!(flags & SVM_FLAG_PRIVATE_PASID)) {

drivers/iommu/intel/svm.c +1 -1

@@ -593,7 +590,7 @@ intel_svm_bind_mm(struct device *dev, unsigned int
 flags, ret = intel_pasid_setup_first_level(iommu, dev,
  				mm ? mm->pgd : init_mm.pgd,
  				svm->pasid, FLPT_DEFAULT_DID,
-				(mm ? 0 : PASID_FLAG_SUPERVISOR_MODE) |
+				(mm ? 0 : IOMMU_SVA_BIND_SUPERVISOR) |
 

NO. PASID_FLAG_SUPERVISOR_MODE is an internal flag to tell the helper whether the SRE bit should be set in the PASID entry.

Quoted 5 lines

drivers/iommu/intel/svm.c +0 -0

  				(cpu_feature_enabled(X86_FEATURE_LA57)
 ? PASID_FLAG_FL5LP : 0));

  		spin_unlock_irqrestore(&iommu->lock, iflags);

drivers/iommu/intel/svm.c +0 -0

@@ -620,7 +617,7 @@ intel_svm_bind_mm(struct device *dev, unsigned int
 flags, ret = intel_pasid_setup_first_level(iommu, dev,

Quoted 6 lines

drivers/iommu/intel/svm.c +0 -1

  						mm ? mm->pgd :
 init_mm.pgd, svm->pasid, FLPT_DEFAULT_DID,
-						(mm ? 0 :

PASID_FLAG_SUPERVISOR_MODE) |

drivers/iommu/intel/svm.c +1 -0

+						(mm ? 0 :

IOMMU_SVA_BIND_SUPERVISOR) |

The same as above.

Quoted 194 lines

(cpu_feature_enabled(X86_FEATURE_LA57)
? PASID_FLAG_FL5LP : 0));
spin_unlock_irqrestore(&iommu->lock, iflags);

drivers/iommu/intel/svm.c +0 -0

@@ -1059,11 +1056,10 @@ static irqreturn_t prq_event_thread(int irq,
 void *d) 
  #define to_intel_svm_dev(handle) container_of(handle, struct

intel_svm_dev, sva) struct iommu_sva *

drivers/iommu/intel/svm.c +1 -1

-intel_svm_bind(struct device *dev, struct mm_struct *mm, void *drvdata)
+intel_svm_bind(struct device *dev, struct mm_struct *mm, unsigned int

flags) {
struct iommu_sva *sva = ERR_PTR(-EINVAL);
struct intel_svm_dev *sdev = NULL;

drivers/iommu/intel/svm.c +0 -1

-	unsigned int flags = 0;
  	int ret;
 
  	/*

@@ -1071,8 +1067,6 @@ intel_svm_bind(struct device *dev, struct
 mm_struct *mm, void *drvdata)
  	 * It will require shared SVM data structures, i.e. combine

io_mm
* and intel_svm etc.
*/

drivers/iommu/intel/svm.c +0 -2

-	if (drvdata)
-		flags = *(unsigned int *)drvdata;
  	mutex_lock(&pasid_mutex);
  	ret = intel_svm_bind_mm(dev, flags, NULL, mm, &sdev);
  	if (ret)

drivers/iommu/iommu.c +1 -0

index d0b0a15..bf0a20f 100644
@@ -2962,6 +2962,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
   * iommu_sva_bind_device() - Bind a process address space to a device
   * @dev: the device
   * @mm: the mm to bind, caller must hold a reference to it
+ * @flags: options for the bind operation
   *
   * Create a bond between device and address space, allowing the

device to access
* the mm using the returned PASID. If a bond already exists between
@device and @@ -2974,7 +2975,7 @@ EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
* On error, returns an ERR_PTR value.
*/
struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void *drvdata) +iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int flags) {
struct iommu_group *group;
struct iommu_sva *handle = ERR_PTR(-EINVAL);

drivers/iommu/iommu.c +4 -2

@@ -2987,6 +2988,9 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, void *drvdata) if (!group)
  		return ERR_PTR(-ENODEV);
 
+	/* Supervisor SVA does not need the current mm */
+	if ((flags & IOMMU_SVA_BIND_SUPERVISOR) && mm)
+		return ERR_PTR(-EINVAL);
  	/* Ensure device count and domain don't change while we're
 binding */ mutex_lock(&group->mutex);
 

@@ -2999,8 +3003,7 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, void *drvdata) if (iommu_group_device_count(group) != 1)
  		goto out_unlock;
 
-	handle = ops->sva_bind(dev, mm, drvdata);
-
+	handle = ops->sva_bind(dev, mm, flags);
  out_unlock:
  	mutex_unlock(&group->mutex);
  	iommu_group_put(group);

drivers/misc/uacce/uacce.c +1 -1

index d07af4e..27e0e04 100644
@@ -99,7 +99,7 @@ static int uacce_bind_queue(struct uacce_device
 *uacce, struct uacce_queue *q) if (!(uacce->flags & UACCE_DEV_SVA))
  		return 0;
 
-	handle = iommu_sva_bind_device(uacce->parent, current->mm,
 NULL);
+	handle = iommu_sva_bind_device(uacce->parent, current->mm, 0);
  	if (IS_ERR(handle))
  		return PTR_ERR(handle);

include/linux/intel-iommu.h +1 -1

index 1bc46b8..cdff752 100644
@@ -757,7 +757,7 @@ int intel_svm_bind_gpasid(struct iommu_domain
 *domain, struct device *dev, struct iommu_gpasid_bind_data *data);
  int intel_svm_unbind_gpasid(struct device *dev, u32 pasid);
  struct iommu_sva *intel_svm_bind(struct device *dev, struct mm_struct
 *mm,
-				 void *drvdata);
+				 unsigned int flags);
  void intel_svm_unbind(struct iommu_sva *handle);
  u32 intel_svm_get_pasid(struct iommu_sva *handle);

int intel_svm_page_response(struct device *dev, struct
iommu_fault_event *evt, diff --git a/include/linux/intel-svm.h b/include/linux/intel-svm.h index 39d368a..ef6b753 100644

include/linux/intel-svm.h +0 -0

@@ -30,30 +30,17 @@ struct svm_dev_ops {
   * if there is no other way to do so. It should be used sparingly, if

at all. */
#define SVM_FLAG_PRIVATE_PASID (1<<0)

include/linux/intel-svm.h +0 -3

-
-/*
- * The SVM_FLAG_SUPERVISOR_MODE flag requests a PASID which can be

used only
- * for access to kernel addresses. No IOTLB flushes are automatically done
- * for kernel mappings; it is valid only for access to the kernel's static
- * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
- * A future API addition may permit the use of such ranges, by means of an
- * explicit IOTLB flush call (akin to the DMA API's unmap method).
- *
- * It is unlikely that we will ever hook into flush_tlb_kernel_range() to
- * do such IOTLB flushes automatically.
- */

include/linux/intel-svm.h +1 -2

-#define SVM_FLAG_SUPERVISOR_MODE	(1<<1)
  /*
   * The SVM_FLAG_GUEST_MODE flag is used when a PASID bind is for guest
   * processes. Compared to the host bind, the primary differences are:
   * 1. mm life cycle management
   * 2. fault reporting
   */
-#define SVM_FLAG_GUEST_MODE		(1<<2)
+#define SVM_FLAG_GUEST_MODE		(1<<1)
  /*
   * 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. */

include/linux/intel-svm.h +1 -1

-#define SVM_FLAG_GUEST_PASID		(1<<3)
+#define SVM_FLAG_GUEST_PASID		(1<<2)
 
  #endif /* __INTEL_SVM_H__ */

include/linux/iommu.h +2 -0

index 5e7fe51..a3fbaa2 100644
@@ -166,6 +166,19 @@ enum iommu_dev_features {
 
  #ifdef CONFIG_IOMMU_API
 
+/*
+ * The IOMMU_SVA_BIND_SUPERVISOR flag requests a PASID which can be

used only
+ * for access to kernel addresses. No IOTLB flushes are automatically done
+ * for kernel mappings; it is valid only for access to the kernel's static
+ * 1:1 mapping of physical memory — not to vmalloc or even module mappings.
+ * A future API addition may permit the use of such ranges, by means of an
+ * explicit IOTLB flush call (akin to the DMA API's unmap method).
+ *
+ * It is unlikely that we will ever hook into flush_tlb_kernel_range() to
+ * do such IOTLB flushes automatically.
+ */

include/linux/iommu.h +2 -0

+#define IOMMU_SVA_BIND_SUPERVISOR       BIT(0)
+
  /**
   * struct iommu_iotlb_gather - Range information for a pending IOTLB

flush *

include/linux/iommu.h +2 -2

@@ -287,7 +300,7 @@ struct iommu_ops {
  	int (*aux_get_pasid)(struct iommu_domain *domain, struct
 device *dev); 
  	struct iommu_sva *(*sva_bind)(struct device *dev, struct
 mm_struct *mm,
-				      void *drvdata);
+				      unsigned int flags);
  	void (*sva_unbind)(struct iommu_sva *handle);
  	u32 (*sva_get_pasid)(struct iommu_sva *handle);

@@ -640,7 +653,7 @@ int iommu_aux_get_pasid(struct iommu_domain
 *domain, struct device *dev); 
  struct iommu_sva *iommu_sva_bind_device(struct device *dev,
  					struct mm_struct *mm,
-					void *drvdata);
+					unsigned int flags);
  void iommu_sva_unbind_device(struct iommu_sva *handle);
  u32 iommu_sva_get_pasid(struct iommu_sva *handle);
 

@@ -1015,7 +1028,7 @@ iommu_aux_get_pasid(struct iommu_domain *domain,
 struct device *dev) }
 

include/linux/iommu.h iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev) · line 1034

Quoted 7 lines

include/linux/iommu.h +0 -1

  static inline struct iommu_sva *
-iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, void
 *drvdata) +iommu_sva_bind_device(struct device *dev, struct mm_struct
 *mm, unsigned int flags) {
  	return NULL;
  }

Best regards,
baolu