Skip to contentSkip to conversation

Simplify and restrict IOMMU SVA APIs

PATCH Superseded Jacob Pan 2 participants 7 messages Apr 15, 2021

Subsystems (8)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Jacob Pan

author
#0 lore raw

v2 cover Simplify and restrict IOMMU SVA APIs

A couple of small changes to simplify and restrict SVA APIs. The motivation is to make PASID allocation palatable for cgroup consumptions. Misc cgroup is merged for v5.13, it can be extended for IOASID as another scalar resource.

I have not tested on ARM platforms due to availability. Would appreciate if someone could help with the testing on ARM.

Thanks,

Jacob

ChangeLog:
V2
- retained mm argument in iommu_sva_alloc_pasid()
- keep generic supervisor flag separated from vt-d's SRE
- move flag declaration out of CONFIG_IOMMU_API

Jacob Pan (2):
iommu/sva: Tighten SVA bind API with explicit flags iommu/sva: Remove mm parameter from SVA bind API

 drivers/dma/idxd/cdev.c                       |  2 +-
 drivers/dma/idxd/init.c                       |  7 +++---
 .../iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c   |  5 +++-
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h   |  4 ++--
 drivers/iommu/intel/svm.c                     | 14 ++++-------
 drivers/iommu/iommu-sva-lib.c                 | 11 +++++----
 drivers/iommu/iommu.c                         | 23 +++++++++++++------
 drivers/misc/uacce/uacce.c                    |  2 +-
 include/linux/intel-iommu.h                   |  2 +-
 include/linux/intel-svm.h                     | 17 ++------------
 include/linux/iommu.h                         | 20 ++++++++++++----
 11 files changed, 57 insertions(+), 50 deletions(-)

base-commit: e49d033bddf5b565044e2abe4241353959bc9120

Signature · 1 line
2.25.1
author
#1 lore raw

PATCH v2 1/2 iommu/sva: Tighten SVA bind API with explicit flags

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.

---
 drivers/dma/idxd/cdev.c                       |  2 +-
 drivers/dma/idxd/init.c                       |  7 +++----
 .../iommu/arm/arm-smmu-v3/arm-smmu-v3-sva.c   |  5 ++++-
 drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h   |  4 ++--
 drivers/iommu/intel/svm.c                     | 14 ++++----------
 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 ++++++++++++++++---
 10 files changed, 40 insertions(+), 41 deletions(-)
drivers/dma/idxd/cdev.c +1 -1
index 0db9b82ed8cf..21ec82bc47b6 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 -4
index 085a0c3b62c6..82a0985ad6dc 100644
@@ -14,7 +14,6 @@
 #include <linux/io-64-nonatomic-lo-hi.h>
 #include <linux/device.h>
 #include <linux/idr.h>
-#include <linux/intel-svm.h>
 #include <linux/iommu.h>
 #include <uapi/linux/idxd.h>
 #include <linux/dmaengine.h>
@@ -300,13 +299,13 @@ 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 +4 -1
index bb251cab61f3..145ceb2fc5da 100644
@@ -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)
 {
 	struct iommu_sva *handle;
 	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);
 	struct arm_smmu_domain *smmu_domain = to_smmu_domain(domain);
 
+	if (flags)
+		return ERR_PTR(-EINVAL);
+
 	if (smmu_domain->stage != ARM_SMMU_DOMAIN_S1)
 		return ERR_PTR(-EINVAL);
 
drivers/iommu/arm/arm-smmu-v3/arm-smmu-v3.h +2 -2
index f985817c967a..b971d4dcf090 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/svm.c +4 -10
index 574a7e657a9a..d4840821f7b5 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)) {
@@ -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 d0b0a15dba84..eefa541d8674 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 defined as IOMMU_SVA_BIND_*
  *
  * 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 d07af4edfcac..27e0e04dfcab 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 1bc46b88711a..cdff752aacdd 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 39d368a810b8..ef6b75326a2b 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 +16 -3
index 5e7fe519430a..ba81cec0b086 100644
@@ -164,6 +164,19 @@ enum iommu_dev_features {
 
 #define IOMMU_PASID_INVALID	(-1U)
 
+/*
+ * 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)
+
 #ifdef CONFIG_IOMMU_API
 
 /**
@@ -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
2.25.1
author
#4 lore raw

PATCH v2 2/2 iommu/sva: Remove mm parameter from SVA bind API

In reply to Jacob Pan

The mm parameter in iommu_sva_bind_device() is intended for privileged process perform bind() on behalf of other processes. This use case has yet to be materialized, let alone potential security implications of adding kernel hooks without explicit user consent.
In addition, with the agreement that IOASID allocation shall be subject cgroup limit. It will be inline with misc cgroup proposal if IOASID allocation as part of the SVA bind is limited to the current task.

---
 drivers/dma/idxd/cdev.c       |  2 +-
 drivers/dma/idxd/init.c       |  2 +-
 drivers/iommu/iommu-sva-lib.c | 11 +++++++----
 drivers/iommu/iommu.c         | 20 +++++++++++++-------
 drivers/misc/uacce/uacce.c    |  2 +-
 include/linux/iommu.h         |  3 +--
 6 files changed, 24 insertions(+), 16 deletions(-)
drivers/dma/idxd/cdev.c +1 -1
index 21ec82bc47b6..8c3347c8930c 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, 0);
+		sva = iommu_sva_bind_device(dev, 0);
 		if (IS_ERR(sva)) {
 			rc = PTR_ERR(sva);
 			dev_err(dev, "pasid allocation failed: %d\n", rc);
drivers/dma/idxd/init.c +1 -1
index 82a0985ad6dc..a92fa625f3b5 100644
@@ -305,7 +305,7 @@ static int idxd_enable_system_pasid(struct idxd_device *idxd)
 
 	flags = IOMMU_SVA_BIND_SUPERVISOR;
 
-	sva = iommu_sva_bind_device(&idxd->pdev->dev, NULL, flags);
+	sva = iommu_sva_bind_device(&idxd->pdev->dev, flags);
 	if (IS_ERR(sva)) {
 		dev_warn(&idxd->pdev->dev,
 			 "iommu sva bind failed: %ld\n", PTR_ERR(sva));
drivers/iommu/iommu-sva-lib.c +7 -4
index bd41405d34e9..6e3d1a010d47 100644
@@ -12,13 +12,13 @@ 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.
@@ -28,6 +28,9 @@ int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max)
 	int ret = 0;
 	ioasid_t pasid;
 
+	if (mm != current->mm)
+		return -EINVAL;
+
 	if (min == INVALID_IOASID || max == INVALID_IOASID ||
 	    min == 0 || max < min)
 		return -EINVAL;
drivers/iommu/iommu.c +13 -7
index eefa541d8674..5bbc35c395a6 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;
@@ -2959,15 +2960,14 @@ int iommu_aux_get_pasid(struct iommu_domain *domain, struct device *dev)
 EXPORT_SYMBOL_GPL(iommu_aux_get_pasid);
 
 /**
- * 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 defined as IOMMU_SVA_BIND_*
  *
  * 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
- * @mm, it is returned and an additional reference is taken. Caller must call
- * iommu_sva_unbind_device() to release each reference.
+ * the current mm, it is returned and an additional reference is taken. Caller
+ * must call iommu_sva_unbind_device() to release each reference.
  *
  * iommu_dev_enable_feature(dev, IOMMU_DEV_FEAT_SVA) must be called first, to
  * initialize the required SVA features.
@@ -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 27e0e04dfcab..da4401a3d8f5 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 +1 -2
index ba81cec0b086..4e846950a6b9 100644
@@ -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 · 6 lines
2.25.1

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
#2 lore raw

Re: [PATCH v2 1/2] iommu/sva: Tighten SVA bind API with explicit flags

In reply to Jacob Pan

drivers/dma/idxd/init.c static struct idxd_device *idxd_alloc(struct pci_dev *pdev) · line 308

Quoted 12 lines · Jacob Pan

static int idxd_enable_system_pasid(struct idxd_device *idxd) {

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

Please also remove the now pointless flags variable.

Quoted 1 line

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

Pleae avoid the pointless overly long line.

include/linux/intel-svm.h struct svm_dev_ops { · line 44

Quoted 2 lines
include/linux/intel-svm.h +1 -1
-#define SVM_FLAG_GUEST_PASID		(1<<3)
+#define SVM_FLAG_GUEST_PASID		(1<<2)

This flag is entirely unused, please just remove it in a prep patch rather than renumbering it.

Review

Quoted 3 lines

static inline struct iommu_sva *

(quoted diff) +1 -1
-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)

Same overy long line here.

#5 lore raw

Re: [PATCH v2 2/2] iommu/sva: Remove mm parameter from SVA bind API

In reply to Jacob Pan

drivers/iommu/iommu-sva-lib.c int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max) · line 33

Quoted 9 lines

*
* Returns 0 on success and < 0 on error.

drivers/iommu/iommu-sva-lib.c +3 -0
@@ -28,6 +28,9 @@ int iommu_sva_alloc_pasid(struct mm_struct *mm, ioasid_t min, ioasid_t max)
 	int ret = 0;
 	ioasid_t pasid;
 
+	if (mm != current->mm)
+		return -EINVAL;
+

Why not remove the parameter entirely?

drivers/iommu/iommu.c iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int fla · line 2997

Quoted 11 lines
drivers/iommu/iommu.c +5 -2
@@ -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);
+	}

I don't see why we need the reference. I think we should just stop passing the mm to ->sva_bind and let the low-level driver deal with any reference to current->mm where needed.

author
#3 lore raw

Re: [PATCH v2 1/2] iommu/sva: Tighten SVA bind API with explicit flags

In reply to Christoph Hellwig

Hi Christoph,

Thanks for the review.

Quoted 8 lines · Christoph Hellwig
Quoted 7 lines · Jacob Pan

static int idxd_enable_system_pasid(struct idxd_device *idxd) {

drivers/dma/idxd/init.c +1 -1
-	int flags;
+	unsigned int flags;
 	unsigned int pasid;
 	struct iommu_sva *sva;
Quoted 8 lines

drivers/dma/idxd/init.c static struct idxd_device *idxd_alloc(struct pci_dev *pdev) · line 308

Quoted 5 lines
drivers/dma/idxd/init.c +2 -2
-	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);  

Please also remove the now pointless flags variable.

Good catch.

Quoted 11 lines
Quoted 2 lines

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

Pleae avoid the pointless overly long line.

include/linux/intel-svm.h struct svm_dev_ops { · line 44

Quoted 2 lines
include/linux/intel-svm.h +1 -1
-#define SVM_FLAG_GUEST_PASID		(1<<3)
+#define SVM_FLAG_GUEST_PASID		(1<<2)  

This flag is entirely unused, please just remove it in a prep patch rather than renumbering it.

You are right. The flag was set and intended to be used by the guest IO page request patches by Baolu.

As you might be aware, we are restructuring the guest SVA uAPI according to Jason's proposal, can we wait until we have a clear solution? We may refactor lots of code.

Quoted 6 lines
Quoted 4 lines

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)

Same overy long line here.

This is temporary as the mm parameter will be removed in the next patch.

Thanks,

Jacob
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

author
#6 lore raw

Re: [PATCH v2 2/2] iommu/sva: Remove mm parameter from SVA bind API

In reply to Christoph Hellwig

Hi Christoph,

Quoted 11 lines · Christoph Hellwig

drivers/iommu/iommu-sva-lib.c int iommu_sva_alloc_pasid(struct mm_struct *mm, · line 33

Quoted 9 lines

*
* Returns 0 on success and < 0 on error.

drivers/iommu/iommu-sva-lib.c +3 -0
@@ -28,6 +28,9 @@ int iommu_sva_alloc_pasid(struct mm_struct *mm,
 ioasid_t min, ioasid_t max) int ret = 0;
 	ioasid_t pasid;
 
+	if (mm != current->mm)
+		return -EINVAL;
+  

Why not remove the parameter entirely?

It was removed in my v1 but thought it would be cleaner if we treat iommu_sva_alloc_pasid() as a leaf function of iommu_sva_bind_device(). Then we don't have to do get_task_mm() every time. But to your point below, it is better to get low-level driver handle it.

Quoted 4 lines
Quoted 4 lines
(quoted diff) +0 -0
@@ -2989,8 +2990,11 @@ iommu_sva_bind_device(struct device *dev, struct
 mm_struct *mm, unsigned int fla return ERR_PTR(-ENODEV);
 
Quoted 12 lines

drivers/iommu/iommu.c iommu_sva_bind_device(struct device *dev, struct mm_struct *mm, unsigned int fla · line 2997

Quoted 8 lines
drivers/iommu/iommu.c +5 -2
 	/* 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);
+	}  

I don't see why we need the reference. I think we should just stop passing the mm to ->sva_bind and let the low-level driver deal with any reference to current->mm where needed.

The mm users reference is just for precaution, in case low level driver use kthread etc.
I agree it is cleaner to just remove mm here, let the low-level driver deal with it.
Let me give it a spin.

Thanks,

Jacob