Skip to contentSkip to conversation

iommu/vt-d: Improve PASID id and table management

PATCH Superseded Lu Baolu 2 participants 12 messages Jun 19, 2018 – Jul 7, 2018 IOMMU

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Lu Baolu

author
#0 lore raw

v3 cover iommu/vt-d: Improve PASID id and table management

Hi,

This patch set improves the PASID id and PASID table management for Intel IOMMU driver.

PATCH 1~3 replace per IOMMU idr name space with a global one. Current per IOMMU idr doesn't work in some cases where one application (associated with a PASID) might talk to two physical devices simultaneously while the two devices could reside behind two different IOMMU units.

PATCH 4~9 implement per device PASID table. Current per IOMMU PASID table implementation is insecure in the cases where multiple devices under one single IOMMU unit support PASID feature. With per domain PASID table, we can achieve finer protection and isolation granularity. This has been discussed at https://lkml.org/lkml/2018/5/16/154.

Best regards,
Lu Baolu

Change log:
v1->v2:
- Patches have been reviewed by "Liu Yi L <[email protected]>".
- An error case handling was added in PATCH 6/9.
- Some commit messages are refined to be more accurate.
v2->v3:
- Patches rebased on top of v4.18-rc1.
- Replace per-domain with per-device pasid table.

Lu Baolu (9):
iommu/vt-d: Global PASID name space iommu/vt-d: Avoid using idr_for_each_entry() iommu/vt-d: Apply global PASID in SVA iommu/vt-d: Move device_domain_info to header iommu/vt-d: Add for_each_device_domain() helper iommu/vt-d: Per PCI device pasid table interfaces iommu/vt-d: Allocate and free pasid table iommu/vt-d: Apply per pci device pasid table in SVA iommu/vt-d: Remove the obsolete per iommu pasid tables

 drivers/iommu/Makefile      |   2 +-
 drivers/iommu/intel-iommu.c | 142 ++++++++++++--------------
 drivers/iommu/intel-pasid.c | 238 ++++++++++++++++++++++++++++++++++++++++++++
 drivers/iommu/intel-pasid.h |  39 ++++++++
 drivers/iommu/intel-svm.c   |  79 ++++++---------
 include/linux/intel-iommu.h |  79 ++++++++++++++-
 6 files changed, 447 insertions(+), 132 deletions(-)
 create mode 100644 drivers/iommu/intel-pasid.c
 create mode 100644 drivers/iommu/intel-pasid.h
Signature · 1 line
2.7.4
author
#1 lore raw

PATCH v3 1/9 iommu/vt-d: Global PASID name space

This adds the system wide PASID name space for the PASID allocation. Currently we are using per IOMMU PASID name spaces which are not suitable for some use cases. For an example, one application (associated with a PASID) might talk to two physical devices simultaneously while the two devices could reside behind two different IOMMU units.

---
 drivers/iommu/Makefile      |  2 +-
 drivers/iommu/intel-iommu.c | 13 ++++++++++
 drivers/iommu/intel-pasid.c | 60 +++++++++++++++++++++++++++++++++++++++++++++
 drivers/iommu/intel-pasid.h | 21 ++++++++++++++++
 4 files changed, 95 insertions(+), 1 deletion(-)
 create mode 100644 drivers/iommu/intel-pasid.c
 create mode 100644 drivers/iommu/intel-pasid.h
drivers/iommu/Makefile +1 -1
index 1fb6958..0a190b4 100644
@@ -14,7 +14,7 @@ obj-$(CONFIG_AMD_IOMMU_V2) += amd_iommu_v2.o
 obj-$(CONFIG_ARM_SMMU) += arm-smmu.o
 obj-$(CONFIG_ARM_SMMU_V3) += arm-smmu-v3.o
 obj-$(CONFIG_DMAR_TABLE) += dmar.o
-obj-$(CONFIG_INTEL_IOMMU) += intel-iommu.o
+obj-$(CONFIG_INTEL_IOMMU) += intel-iommu.o intel-pasid.o
 obj-$(CONFIG_INTEL_IOMMU_SVM) += intel-svm.o
 obj-$(CONFIG_IPMMU_VMSA) += ipmmu-vmsa.o
 obj-$(CONFIG_IRQ_REMAP) += intel_irq_remapping.o irq_remapping.o
drivers/iommu/intel-iommu.c +13 -0
index 14e4b37..3653ebc 100644
@@ -53,6 +53,7 @@
 #include <asm/iommu.h>
 
 #include "irq_remapping.h"
+#include "intel-pasid.h"
 
 #define ROOT_SIZE		VTD_PAGE_SIZE
 #define CONTEXT_SIZE		VTD_PAGE_SIZE
@@ -3277,6 +3278,18 @@ static int __init init_dmars(void)
 	}
 
 	for_each_active_iommu(iommu, drhd) {
+		/*
+		 * Find the max pasid size of all IOMMU's in the system.
+		 * We need to ensure the system pasid table is no bigger
+		 * than the smallest supported.
+		 */
+		if (pasid_enabled(iommu)) {
+			u32 temp = 2 << ecap_pss(iommu->ecap);
+
+			intel_pasid_max_id = min_t(u32, temp,
+						   intel_pasid_max_id);
+		}
+
 		g_iommus[iommu->seq_id] = iommu;
 
 		intel_iommu_init_qi(iommu);
drivers/iommu/intel-pasid.c new file +60 -0

60 lines not on this page · raw message

drivers/iommu/intel-pasid.h new file +21 -0
new file mode 100644
index 0000000..d5feb3d
@@ -0,0 +1,21 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * intel-pasid.h - PASID idr, table and entry header
+ *
+ * Copyright (C) 2018 Intel Corporation
+ *
+ * Author: Lu Baolu <[email protected]>
+ */
+
+#ifndef __INTEL_PASID_H
+#define __INTEL_PASID_H
+
+#define PASID_MIN			0x1
+#define PASID_MAX			0x20000
+
+extern u32 intel_pasid_max_id;
+int intel_pasid_alloc_id(void *ptr, int start, int end, gfp_t gfp);
+void intel_pasid_free_id(int pasid);
+void *intel_pasid_lookup_id(int pasid);
+
+#endif /* __INTEL_PASID_H */
Signature · 1 line
2.7.4
author
#2 lore raw

PATCH v3 2/9 iommu/vt-d: Avoid using idr_for_each_entry()

In reply to Lu Baolu

idr_for_each_entry() is used to iteratte over idr elements of a given type. It isn't suitable for the globle pasid idr since the pasid idr consumer could specify different types of pointers to bind with a pasid.

Cc Ashok Raj <[email protected]>
Cc Jacob Pan <jacob.jun.pan-VuQAYsv1563Yd54FQh9/[email protected]>
Cc Kevin Tian <[email protected]>
Cc Liu Yi L <[email protected]>
---
 drivers/iommu/intel-svm.c   | 14 ++++++++++----
 include/linux/intel-iommu.h |  1 +
 2 files changed, 11 insertions(+), 4 deletions(-)
drivers/iommu/intel-svm.c +10 -4
index 2cc0aac..36cc1d1 100644
@@ -298,6 +298,7 @@ static const struct mmu_notifier_ops intel_mmuops = {
 };
 
 static DEFINE_MUTEX(pasid_mutex);
+static LIST_HEAD(global_svm_list);
 
 int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ops *ops)
 {
@@ -329,13 +330,13 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 
 	mutex_lock(&pasid_mutex);
 	if (pasid && !(flags & SVM_FLAG_PRIVATE_PASID)) {
-		int i;
+		struct intel_svm *t;
 
-		idr_for_each_entry(&iommu->pasid_idr, svm, i) {
-			if (svm->mm != mm ||
-			    (svm->flags & SVM_FLAG_PRIVATE_PASID))
+		list_for_each_entry(t, &global_svm_list, list) {
+			if (t->mm != mm || (t->flags & SVM_FLAG_PRIVATE_PASID))
 				continue;
 
+			svm = t;
 			if (svm->pasid >= pasid_max) {
 				dev_warn(dev,
 					 "Limited PASID width. Cannot use existing PASID %d\n",
@@ -404,6 +405,7 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 		svm->mm = mm;
 		svm->flags = flags;
 		INIT_LIST_HEAD_RCU(&svm->devs);
+		INIT_LIST_HEAD(&svm->list);
 		ret = -ENOMEM;
 		if (mm) {
 			ret = mmu_notifier_register(&svm->notifier, mm);
@@ -430,6 +432,8 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 		 */
 		if (cap_caching_mode(iommu->cap))
 			intel_flush_pasid_dev(svm, sdev, svm->pasid);
+
+		list_add_tail(&svm->list, &global_svm_list);
 	}
 	list_add_rcu(&sdev->list, &svm->devs);
 
@@ -485,6 +489,8 @@ int intel_svm_unbind_mm(struct device *dev, int pasid)
 					if (svm->mm)
 						mmu_notifier_unregister(&svm->notifier, svm->mm);
 
+					list_del(&svm->list);
+
 					/* We mandate that no page faults may be outstanding
 					 * for the PASID when intel_svm_unbind_mm() is called.
 					 * If that is not obeyed, subtle errors will happen.
include/linux/intel-iommu.h +1 -0
index 1df9401..06f04e2 100644
@@ -485,6 +485,7 @@ struct intel_svm {
 	int flags;
 	int pasid;
 	struct list_head devs;
+	struct list_head list;
 };
 
 extern int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct intel_svm_dev *sdev);
Signature · 1 line
2.7.4
author
#3 lore raw

PATCH v3 3/9 iommu/vt-d: Apply global PASID in SVA

In reply to Lu Baolu

This patch applies the global pasid name space in the shared virtual address (SVA) implementation.

---
 drivers/iommu/intel-svm.c   | 22 +++++++++++-----------
 include/linux/intel-iommu.h |  1 -
 2 files changed, 11 insertions(+), 12 deletions(-)
drivers/iommu/intel-svm.c +11 -11
index 36cc1d1..56e65d0 100644
@@ -26,6 +26,8 @@
 #include <linux/interrupt.h>
 #include <asm/page.h>
 
+#include "intel-pasid.h"
+
 #define PASID_ENTRY_P		BIT_ULL(0)
 #define PASID_ENTRY_FLPM_5LP	BIT_ULL(9)
 #define PASID_ENTRY_SRE		BIT_ULL(11)
@@ -85,8 +87,6 @@ int intel_svm_alloc_pasid_tables(struct intel_iommu *iommu)
 				iommu->name);
 	}
 
-	idr_init(&iommu->pasid_idr);
-
 	return 0;
 }
 
@@ -102,7 +102,7 @@ int intel_svm_free_pasid_tables(struct intel_iommu *iommu)
 		free_pages((unsigned long)iommu->pasid_state_table, order);
 		iommu->pasid_state_table = NULL;
 	}
-	idr_destroy(&iommu->pasid_idr);
+
 	return 0;
 }
 
@@ -392,9 +392,9 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 			pasid_max = iommu->pasid_max;
 
 		/* Do not use PASID 0 in caching mode (virtualised IOMMU) */
-		ret = idr_alloc(&iommu->pasid_idr, svm,
-				!!cap_caching_mode(iommu->cap),
-				pasid_max - 1, GFP_KERNEL);
+		ret = intel_pasid_alloc_id(svm,
+					   !!cap_caching_mode(iommu->cap),
+					   pasid_max - 1, GFP_KERNEL);
 		if (ret < 0) {
 			kfree(svm);
 			kfree(sdev);
@@ -410,7 +410,7 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 		if (mm) {
 			ret = mmu_notifier_register(&svm->notifier, mm);
 			if (ret) {
-				idr_remove(&svm->iommu->pasid_idr, svm->pasid);
+				intel_pasid_free_id(svm->pasid);
 				kfree(svm);
 				kfree(sdev);
 				goto out;
@@ -460,7 +460,7 @@ int intel_svm_unbind_mm(struct device *dev, int pasid)
 	if (!iommu || !iommu->pasid_table)
 		goto out;
 
-	svm = idr_find(&iommu->pasid_idr, pasid);
+	svm = intel_pasid_lookup_id(pasid);
 	if (!svm)
 		goto out;
 
@@ -485,7 +485,7 @@ int intel_svm_unbind_mm(struct device *dev, int pasid)
 					svm->iommu->pasid_table[svm->pasid].val = 0;
 					wmb();
 
-					idr_remove(&svm->iommu->pasid_idr, svm->pasid);
+					intel_pasid_free_id(svm->pasid);
 					if (svm->mm)
 						mmu_notifier_unregister(&svm->notifier, svm->mm);
 
@@ -520,7 +520,7 @@ int intel_svm_is_pasid_valid(struct device *dev, int pasid)
 	if (!iommu || !iommu->pasid_table)
 		goto out;
 
-	svm = idr_find(&iommu->pasid_idr, pasid);
+	svm = intel_pasid_lookup_id(pasid);
 	if (!svm)
 		goto out;
 
@@ -618,7 +618,7 @@ static irqreturn_t prq_event_thread(int irq, void *d)
 
 		if (!svm || svm->pasid != req->pasid) {
 			rcu_read_lock();
-			svm = idr_find(&iommu->pasid_idr, req->pasid);
+			svm = intel_pasid_lookup_id(req->pasid);
 			/* It *can't* go away, because the driver is not permitted
 			 * to unbind the mm while any page faults are outstanding.
 			 * So we only need RCU to protect the internal idr code. */
include/linux/intel-iommu.h +0 -1
index 06f04e2..0cfac9d 100644
@@ -417,7 +417,6 @@ struct intel_iommu {
 	struct pasid_state_entry *pasid_state_table;
 	struct page_req_dsc *prq;
 	unsigned char prq_name[16];    /* Name for PRQ interrupt */
-	struct idr pasid_idr;
 	u32 pasid_max;
 #endif
 	struct q_inval  *qi;            /* Queued invalidation info */
Signature · 1 line
2.7.4
author
#4 lore raw

PATCH v3 4/9 iommu/vt-d: Move device_domain_info to header

In reply to Lu Baolu

This allows the per device iommu data and some helpers to be used in other files.

---
 drivers/iommu/intel-iommu.c | 62 +++--------------------------------------
 include/linux/intel-iommu.h | 68 +++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 72 insertions(+), 58 deletions(-)
drivers/iommu/intel-iommu.c +4 -58
index 3653ebc..d9a8222 100644
@@ -381,60 +381,6 @@ static int hw_pass_through = 1;
 	for (idx = 0; idx < g_num_of_iommus; idx++)		\
 		if (domain->iommu_refcnt[idx])
 
-struct dmar_domain {
-	int	nid;			/* node id */
-
-	unsigned	iommu_refcnt[DMAR_UNITS_SUPPORTED];
-					/* Refcount of devices per iommu */
-
-
-	u16		iommu_did[DMAR_UNITS_SUPPORTED];
-					/* Domain ids per IOMMU. Use u16 since
-					 * domain ids are 16 bit wide according
-					 * to VT-d spec, section 9.3 */
-
-	bool has_iotlb_device;
-	struct list_head devices;	/* all devices' list */
-	struct iova_domain iovad;	/* iova's that belong to this domain */
-
-	struct dma_pte	*pgd;		/* virtual address */
-	int		gaw;		/* max guest address width */
-
-	/* adjusted guest address width, 0 is level 2 30-bit */
-	int		agaw;
-
-	int		flags;		/* flags to find out type of domain */
-
-	int		iommu_coherency;/* indicate coherency of iommu access */
-	int		iommu_snooping; /* indicate snooping control feature*/
-	int		iommu_count;	/* reference count of iommu */
-	int		iommu_superpage;/* Level of superpages supported:
-					   0 == 4KiB (no superpages), 1 == 2MiB,
-					   2 == 1GiB, 3 == 512GiB, 4 == 1TiB */
-	u64		max_addr;	/* maximum mapped address */
-
-	struct iommu_domain domain;	/* generic domain data structure for
-					   iommu core */
-};
-
-/* PCI domain-device relationship */
-struct device_domain_info {
-	struct list_head link;	/* link to domain siblings */
-	struct list_head global; /* link to global list */
-	u8 bus;			/* PCI bus number */
-	u8 devfn;		/* PCI devfn number */
-	u8 pasid_supported:3;
-	u8 pasid_enabled:1;
-	u8 pri_supported:1;
-	u8 pri_enabled:1;
-	u8 ats_supported:1;
-	u8 ats_enabled:1;
-	u8 ats_qdep;
-	struct device *dev; /* it's NULL for PCIe-to-PCI bridge */
-	struct intel_iommu *iommu; /* IOMMU used by this device */
-	struct dmar_domain *domain; /* pointer to domain */
-};
-
 struct dmar_rmrr_unit {
 	struct list_head list;		/* list of rmrr units	*/
 	struct acpi_dmar_header *hdr;	/* ACPI header		*/
@@ -603,7 +549,7 @@ static void set_iommu_domain(struct intel_iommu *iommu, u16 did,
 		domains[did & 0xff] = domain;
 }
 
-static inline void *alloc_pgtable_page(int node)
+void *alloc_pgtable_page(int node)
 {
 	struct page *page;
 	void *vaddr = NULL;
@@ -614,7 +560,7 @@ static inline void *alloc_pgtable_page(int node)
 	return vaddr;
 }
 
-static inline void free_pgtable_page(void *vaddr)
+void free_pgtable_page(void *vaddr)
 {
 	free_page((unsigned long)vaddr);
 }
@@ -697,7 +643,7 @@ int iommu_calculate_agaw(struct intel_iommu *iommu)
 }
 
 /* This functionin only returns single iommu in a domain */
-static struct intel_iommu *domain_get_iommu(struct dmar_domain *domain)
+struct intel_iommu *domain_get_iommu(struct dmar_domain *domain)
 {
 	int iommu_id;
 
@@ -3512,7 +3458,7 @@ static unsigned long intel_alloc_iova(struct device *dev,
 	return iova_pfn;
 }
 
-static struct dmar_domain *get_valid_domain_for_dev(struct device *dev)
+struct dmar_domain *get_valid_domain_for_dev(struct device *dev)
 {
 	struct dmar_domain *domain, *tmp;
 	struct dmar_rmrr_unit *rmrr;
include/linux/intel-iommu.h +68 -0

92 lines not on this page · raw message

Signature · 1 line
2.7.4
author
#5 lore raw

PATCH v3 5/9 iommu/vt-d: Add for_each_device_domain() helper

In reply to Lu Baolu

This adds a helper named for_each_device_domain() to iterate over the elements in device_domain_list and invoke a callback against each element. This allows to search the device_domain list in other source files.

---
 drivers/iommu/intel-iommu.c | 21 +++++++++++++++++++++
 include/linux/intel-iommu.h |  2 ++
 2 files changed, 23 insertions(+)
drivers/iommu/intel-iommu.c +21 -0
index d9a8222..aea182d 100644
@@ -448,6 +448,27 @@ EXPORT_SYMBOL_GPL(intel_iommu_gfx_mapped);
 static DEFINE_SPINLOCK(device_domain_lock);
 static LIST_HEAD(device_domain_list);
 
+/*
+ * Iterate over elements in device_domain_list and call the specified
+ * callback @fn against each element. This helper should only be used
+ * in the context where the device_domain_lock has already been holden.
+ */
+int for_each_device_domain(int (*fn)(struct device_domain_info *info,
+				     void *data), void *data)
+{
+	int ret = 0;
+	struct device_domain_info *info;
+
+	assert_spin_locked(&device_domain_lock);
+	list_for_each_entry(info, &device_domain_list, global) {
+		ret = fn(info, data);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
 const struct iommu_ops intel_iommu_ops;
 
 static bool translation_pre_enabled(struct intel_iommu *iommu)
include/linux/intel-iommu.h +2 -0
index 3598dc0f..d52013e 100644
@@ -525,6 +525,8 @@ struct dmar_domain *get_valid_domain_for_dev(struct device *dev);
 void *alloc_pgtable_page(int node);
 void free_pgtable_page(void *vaddr);
 struct intel_iommu *domain_get_iommu(struct dmar_domain *domain);
+int for_each_device_domain(int (*fn)(struct device_domain_info *info,
+				     void *data), void *data);
 
 #ifdef CONFIG_INTEL_IOMMU_SVM
 extern int intel_svm_alloc_pasid_tables(struct intel_iommu *iommu);
Signature · 1 line
2.7.4
author
#6 lore raw

PATCH v3 6/9 iommu/vt-d: Per PCI device pasid table interfaces

In reply to Lu Baolu

This patch adds the interfaces for per PCI device pasid table management. Currently we allocate one pasid table for all PCI devices under the scope of an IOMMU. It's insecure in some cases where multiple devices under one single IOMMU unit support PASID features. With per PCI device pasid table, we can achieve finer protection and isolation granularity.

---
 drivers/iommu/intel-iommu.c |   1 +
 drivers/iommu/intel-pasid.c | 178 ++++++++++++++++++++++++++++++++++++++++++++
 drivers/iommu/intel-pasid.h |  18 +++++
 drivers/iommu/intel-svm.c   |   4 -
 include/linux/intel-iommu.h |   2 +
 5 files changed, 199 insertions(+), 4 deletions(-)
drivers/iommu/intel-iommu.c +1 -0
index aea182d..44ce77a 100644
@@ -2436,6 +2436,7 @@ static struct dmar_domain *dmar_insert_one_dev_info(struct intel_iommu *iommu,
 	info->dev = dev;
 	info->domain = domain;
 	info->iommu = iommu;
+	info->pasid_table = NULL;
 
 	if (dev && dev_is_pci(dev)) {
 		struct pci_dev *pdev = to_pci_dev(info->dev);
drivers/iommu/intel-pasid.c +178 -0

187 lines not on this page · raw message

drivers/iommu/intel-pasid.h +18 -0
index d5feb3d..1fb5e12 100644
@@ -13,9 +13,27 @@
 #define PASID_MIN			0x1
 #define PASID_MAX			0x20000
 
+struct pasid_entry {
+	u64 val;
+};
+
+/* The representative of a PASID table */
+struct pasid_table {
+	void			*table;		/* pasid table pointer */
+	int			order;		/* page order of pasid table */
+	int			max_pasid;	/* max pasid */
+	struct list_head	dev;		/* device list */
+};
+
 extern u32 intel_pasid_max_id;
 int intel_pasid_alloc_id(void *ptr, int start, int end, gfp_t gfp);
 void intel_pasid_free_id(int pasid);
 void *intel_pasid_lookup_id(int pasid);
+int intel_pasid_alloc_table(struct device *dev);
+void intel_pasid_free_table(struct device *dev);
+struct pasid_table *intel_pasid_get_table(struct device *dev);
+int intel_pasid_get_dev_max_id(struct device *dev);
+struct pasid_entry *intel_pasid_get_entry(struct device *dev, int pasid);
+void intel_pasid_clear_entry(struct device *dev, int pasid);
 
 #endif /* __INTEL_PASID_H */
drivers/iommu/intel-svm.c +0 -4
index 56e65d0..9b5dc72 100644
@@ -34,10 +34,6 @@
 
 static irqreturn_t prq_event_thread(int irq, void *d);
 
-struct pasid_entry {
-	u64 val;
-};
-
 struct pasid_state_entry {
 	u64 val;
 };
include/linux/intel-iommu.h +2 -0
index d52013e..8ec22b3 100644
@@ -481,6 +481,7 @@ struct intel_iommu {
 struct device_domain_info {
 	struct list_head link;	/* link to domain siblings */
 	struct list_head global; /* link to global list */
+	struct list_head table;	/* link to pasid table */
 	u8 bus;			/* PCI bus number */
 	u8 devfn;		/* PCI devfn number */
 	u8 pasid_supported:3;
@@ -493,6 +494,7 @@ struct device_domain_info {
 	struct device *dev; /* it's NULL for PCIe-to-PCI bridge */
 	struct intel_iommu *iommu; /* IOMMU used by this device */
 	struct dmar_domain *domain; /* pointer to domain */
+	struct pasid_table *pasid_table; /* pasid table */
 };
 
 static inline void __iommu_flush_cache(
Signature · 1 line
2.7.4
author
#7 lore raw

PATCH v3 7/9 iommu/vt-d: Allocate and free pasid table

In reply to Lu Baolu

This patch allocates a PASID table for a PCI device at the time when the dmar dev_info is attached to dev->archdata.iommu, and free it in the opposite case.

---
 drivers/iommu/intel-iommu.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
drivers/iommu/intel-iommu.c +10 -0
index 44ce77a..43bf151 100644
@@ -2494,6 +2494,15 @@ static struct dmar_domain *dmar_insert_one_dev_info(struct intel_iommu *iommu,
 	list_add(&info->global, &device_domain_list);
 	if (dev)
 		dev->archdata.iommu = info;
+
+	if (dev && dev_is_pci(dev) && info->pasid_supported) {
+		ret = intel_pasid_alloc_table(dev);
+		if (ret) {
+			__dmar_remove_one_dev_info(info);
+			spin_unlock_irqrestore(&device_domain_lock, flags);
+			return NULL;
+		}
+	}
 	spin_unlock_irqrestore(&device_domain_lock, flags);
 
 	if (dev && domain_context_mapping(domain, dev)) {
@@ -4828,6 +4837,7 @@ static void __dmar_remove_one_dev_info(struct device_domain_info *info)
 	if (info->dev) {
 		iommu_disable_dev_iotlb(info);
 		domain_context_clear(iommu, info->dev);
+		intel_pasid_free_table(info->dev);
 	}
 
 	unlink_domain_info(info);
Signature · 1 line
2.7.4
author
#8 lore raw

PATCH v3 8/9 iommu/vt-d: Apply per pci device pasid table in SVA

In reply to Lu Baolu

This patch applies the per pci device pasid table in the Shared Virtual Address (SVA) implementation.

---
 drivers/iommu/intel-iommu.c | 29 +++++++++--------------------
 drivers/iommu/intel-svm.c   | 22 ++++++++++------------
 2 files changed, 19 insertions(+), 32 deletions(-)
drivers/iommu/intel-iommu.c +9 -20
index 43bf151..b0651a2 100644
@@ -5163,22 +5163,16 @@ static void intel_iommu_put_resv_regions(struct device *dev,
 
 #ifdef CONFIG_INTEL_IOMMU_SVM
 #define MAX_NR_PASID_BITS (20)
-static inline unsigned long intel_iommu_get_pts(struct intel_iommu *iommu)
+static inline unsigned long intel_iommu_get_pts(struct device *dev)
 {
-	/*
-	 * Convert ecap_pss to extend context entry pts encoding, also
-	 * respect the soft pasid_max value set by the iommu.
-	 * - number of PASID bits = ecap_pss + 1
-	 * - number of PASID table entries = 2^(pts + 5)
-	 * Therefore, pts = ecap_pss - 4
-	 * e.g. KBL ecap_pss = 0x13, PASID has 20 bits, pts = 15
-	 */
-	if (ecap_pss(iommu->ecap) < 5)
+	int pts, max_pasid;
+
+	max_pasid = intel_pasid_get_dev_max_id(dev);
+	pts = find_first_bit((unsigned long *)&max_pasid, MAX_NR_PASID_BITS);
+	if (pts < 5)
 		return 0;
 
-	/* pasid_max is encoded as actual number of entries not the bits */
-	return find_first_bit((unsigned long *)&iommu->pasid_max,
-			MAX_NR_PASID_BITS) - 5;
+	return pts - 5;
 }
 
 int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct intel_svm_dev *sdev)
@@ -5214,8 +5208,8 @@ int intel_iommu_enable_pasid(struct intel_iommu *iommu, struct intel_svm_dev *sd
 	if (!(ctx_lo & CONTEXT_PASIDE)) {
 		if (iommu->pasid_state_table)
 			context[1].hi = (u64)virt_to_phys(iommu->pasid_state_table);
-		context[1].lo = (u64)virt_to_phys(iommu->pasid_table) |
-			intel_iommu_get_pts(iommu);
+		context[1].lo = (u64)virt_to_phys(info->pasid_table->table) |
+			intel_iommu_get_pts(sdev->dev);
 
 		wmb();
 		/* CONTEXT_TT_MULTI_LEVEL and CONTEXT_TT_DEV_IOTLB are both
@@ -5282,11 +5276,6 @@ struct intel_iommu *intel_svm_device_to_iommu(struct device *dev)
 		return NULL;
 	}
 
-	if (!iommu->pasid_table) {
-		dev_err(dev, "PASID not enabled on IOMMU; cannot enable SVM\n");
-		return NULL;
-	}
-
 	return iommu;
 }
 #endif /* CONFIG_INTEL_IOMMU_SVM */
drivers/iommu/intel-svm.c +10 -12
index 9b5dc72..a253cde 100644
@@ -274,11 +274,9 @@ static void intel_mm_release(struct mmu_notifier *mn, struct mm_struct *mm)
 	 * page) so that we end up taking a fault that the hardware really
 	 * *has* to handle gracefully without affecting other processes.
 	 */
-	svm->iommu->pasid_table[svm->pasid].val = 0;
-	wmb();
-
 	rcu_read_lock();
 	list_for_each_entry_rcu(sdev, &svm->devs, list) {
+		intel_pasid_clear_entry(sdev->dev, svm->pasid);
 		intel_flush_pasid_dev(svm, sdev, svm->pasid);
 		intel_flush_svm_range_dev(svm, sdev, 0, -1, 0, !svm->mm);
 	}
@@ -299,6 +297,7 @@ static LIST_HEAD(global_svm_list);
 int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_ops *ops)
 {
 	struct intel_iommu *iommu = intel_svm_device_to_iommu(dev);
+	struct pasid_entry *entry;
 	struct intel_svm_dev *sdev;
 	struct intel_svm *svm = NULL;
 	struct mm_struct *mm = NULL;
@@ -306,7 +305,7 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 	int pasid_max;
 	int ret;
 
-	if (!iommu || !iommu->pasid_table)
+	if (!iommu)
 		return -EINVAL;
 
 	if (dev_is_pci(dev)) {
@@ -384,8 +383,8 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 		}
 		svm->iommu = iommu;
 
-		if (pasid_max > iommu->pasid_max)
-			pasid_max = iommu->pasid_max;
+		if (pasid_max > intel_pasid_max_id)
+			pasid_max = intel_pasid_max_id;
 
 		/* Do not use PASID 0 in caching mode (virtualised IOMMU) */
 		ret = intel_pasid_alloc_id(svm,
@@ -418,7 +417,8 @@ int intel_svm_bind_mm(struct device *dev, int *pasid, int flags, struct svm_dev_
 		if (cpu_feature_enabled(X86_FEATURE_LA57))
 			pasid_entry_val |= PASID_ENTRY_FLPM_5LP;
 
-		iommu->pasid_table[svm->pasid].val = pasid_entry_val;
+		entry = intel_pasid_get_entry(dev, svm->pasid);
+		entry->val = pasid_entry_val;
 
 		wmb();
 
@@ -453,7 +453,7 @@ int intel_svm_unbind_mm(struct device *dev, int pasid)
 
 	mutex_lock(&pasid_mutex);
 	iommu = intel_svm_device_to_iommu(dev);
-	if (!iommu || !iommu->pasid_table)
+	if (!iommu)
 		goto out;
 
 	svm = intel_pasid_lookup_id(pasid);
@@ -476,11 +476,9 @@ int intel_svm_unbind_mm(struct device *dev, int pasid)
 				intel_flush_pasid_dev(svm, sdev, svm->pasid);
 				intel_flush_svm_range_dev(svm, sdev, 0, -1, 0, !svm->mm);
 				kfree_rcu(sdev, rcu);
+				intel_pasid_clear_entry(dev, svm->pasid);
 
 				if (list_empty(&svm->devs)) {
-					svm->iommu->pasid_table[svm->pasid].val = 0;
-					wmb();
-
 					intel_pasid_free_id(svm->pasid);
 					if (svm->mm)
 						mmu_notifier_unregister(&svm->notifier, svm->mm);
@@ -513,7 +511,7 @@ int intel_svm_is_pasid_valid(struct device *dev, int pasid)
 
 	mutex_lock(&pasid_mutex);
 	iommu = intel_svm_device_to_iommu(dev);
-	if (!iommu || !iommu->pasid_table)
+	if (!iommu)
 		goto out;
 
 	svm = intel_pasid_lookup_id(pasid);
Signature · 1 line
2.7.4
author
#9 lore raw

PATCH v3 9/9 iommu/vt-d: Remove the obsolete per iommu pasid tables

In reply to Lu Baolu

The obsolete per iommu pasid tables are no longer used. Hence, clean up them.

---
 drivers/iommu/intel-iommu.c |  6 +++---
 drivers/iommu/intel-svm.c   | 17 ++---------------
 include/linux/intel-iommu.h |  5 ++---
 3 files changed, 7 insertions(+), 21 deletions(-)
drivers/iommu/intel-iommu.c +3 -3
index b0651a2..0703ca3 100644
@@ -1741,7 +1741,7 @@ static void free_dmar_iommu(struct intel_iommu *iommu)
 	if (pasid_enabled(iommu)) {
 		if (ecap_prs(iommu->ecap))
 			intel_svm_finish_prq(iommu);
-		intel_svm_free_pasid_tables(iommu);
+		intel_svm_exit(iommu);
 	}
 #endif
 }
@@ -3322,7 +3322,7 @@ static int __init init_dmars(void)
 			hw_pass_through = 0;
 #ifdef CONFIG_INTEL_IOMMU_SVM
 		if (pasid_enabled(iommu))
-			intel_svm_alloc_pasid_tables(iommu);
+			intel_svm_init(iommu);
 #endif
 	}
 
@@ -4285,7 +4285,7 @@ static int intel_iommu_add(struct dmar_drhd_unit *dmaru)
 
 #ifdef CONFIG_INTEL_IOMMU_SVM
 	if (pasid_enabled(iommu))
-		intel_svm_alloc_pasid_tables(iommu);
+		intel_svm_init(iommu);
 #endif
 
 	if (dmaru->ignored) {
drivers/iommu/intel-svm.c +2 -15
index a253cde..eb30836 100644
@@ -38,7 +38,7 @@ struct pasid_state_entry {
 	u64 val;
 };
 
-int intel_svm_alloc_pasid_tables(struct intel_iommu *iommu)
+int intel_svm_init(struct intel_iommu *iommu)
 {
 	struct page *pages;
 	int order;
@@ -63,15 +63,6 @@ int intel_svm_alloc_pasid_tables(struct intel_iommu *iommu)
 		iommu->pasid_max = 0x20000;
 
 	order = get_order(sizeof(struct pasid_entry) * iommu->pasid_max);
-	pages = alloc_pages(GFP_KERNEL | __GFP_ZERO, order);
-	if (!pages) {
-		pr_warn("IOMMU: %s: Failed to allocate PASID table\n",
-			iommu->name);
-		return -ENOMEM;
-	}
-	iommu->pasid_table = page_address(pages);
-	pr_info("%s: Allocated order %d PASID table.\n", iommu->name, order);
-
 	if (ecap_dis(iommu->ecap)) {
 		/* Just making it explicit... */
 		BUILD_BUG_ON(sizeof(struct pasid_entry) != sizeof(struct pasid_state_entry));
@@ -86,14 +77,10 @@ int intel_svm_alloc_pasid_tables(struct intel_iommu *iommu)
 	return 0;
 }
 
-int intel_svm_free_pasid_tables(struct intel_iommu *iommu)
+int intel_svm_exit(struct intel_iommu *iommu)
 {
 	int order = get_order(sizeof(struct pasid_entry) * iommu->pasid_max);
 
-	if (iommu->pasid_table) {
-		free_pages((unsigned long)iommu->pasid_table, order);
-		iommu->pasid_table = NULL;
-	}
 	if (iommu->pasid_state_table) {
 		free_pages((unsigned long)iommu->pasid_state_table, order);
 		iommu->pasid_state_table = NULL;
include/linux/intel-iommu.h +2 -3
index 8ec22b3..8051bc5 100644
@@ -458,7 +458,6 @@ struct intel_iommu {
 	 * devices away to userspace processes (e.g. for DPDK) and don't
 	 * want to trust that userspace will use *only* the PASID it was
 	 * told to. But while it's all driver-arbitrated, we're fine. */
-	struct pasid_entry *pasid_table;
 	struct pasid_state_entry *pasid_state_table;
 	struct page_req_dsc *prq;
 	unsigned char prq_name[16];    /* Name for PRQ interrupt */
@@ -531,8 +530,8 @@ int for_each_device_domain(int (*fn)(struct device_domain_info *info,
 				     void *data), void *data);
 
 #ifdef CONFIG_INTEL_IOMMU_SVM
-extern int intel_svm_alloc_pasid_tables(struct intel_iommu *iommu);
-extern int intel_svm_free_pasid_tables(struct intel_iommu *iommu);
+int intel_svm_init(struct intel_iommu *iommu);
+int intel_svm_exit(struct intel_iommu *iommu);
 extern int intel_svm_enable_prq(struct intel_iommu *iommu);
 extern int intel_svm_finish_prq(struct intel_iommu *iommu);
 
Signature · 1 line
2.7.4
maintainer
#10 lore raw
In reply to Lu Baolu

Hey Lu,

the patches do not apply cleanly to my x86/vt-d branch. Can you rebase against that branch after the weekend when I pushed it out and re-send, please?

Thanks a lot,

Joerg

author
#11 lore raw

Hi Joerg,

Quoted 5 lines · Joerg Roedel

Hey Lu,

the patches do not apply cleanly to my x86/vt-d branch. Can you rebase against that branch after the weekend when I pushed it out and re-send, please?

Sure.

Best regards,
Lu Baolu