From 61985a040f17c03b09a2772508ee02729571365b Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 6 Apr 2011 17:46:49 +0200
Subject: x86/amd-iommu: Remove command buffer resetting logic

The logic to reset the command buffer caused more problems
than it actually helped. The logic jumped in when the IOMMU
hardware doesn't execute commands anymore but the reasons
for this are usually not fixed by just resetting the command
buffer. So the code can be removed to reduce complexity.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h | 3 ---
 1 file changed, 3 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index e3509fc303bf..878ae008eb04 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -409,9 +409,6 @@ struct amd_iommu {
 	/* if one, we need to send a completion wait command */
 	bool need_sync;
 
-	/* becomes true if a command buffer reset is running */
-	bool reset_in_progress;
-
 	/* default dma_ops domain for that IOMMU */
 	struct dma_ops_domain *default_dom;
 
-- 
cgit v1.2.3


From 7d0c5cc5be73f7ce26fdcca7b8ec2203f661eb93 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 7 Apr 2011 08:16:10 +0200
Subject: x86/amd-iommu: Flush all internal TLBs when IOMMUs are enabled

The old code only flushed a DTE or a domain TLB before it is
actually used by the IOMMU driver. While this is efficient
and works when done right it is more likely to introduce new
bugs when changing code (which happened in the past).
This patch adds code to flush all DTEs and all domain TLBs
in each IOMMU right after it is enabled (at boot and after
resume). This reduces the complexity of the driver and makes
it less likely to introduce stale-TLB bugs in the future.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_proto.h |  2 -
 arch/x86/kernel/amd_iommu.c            | 75 +++++++++++++++-------------------
 arch/x86/kernel/amd_iommu_init.c       | 11 ++++-
 3 files changed, 43 insertions(+), 45 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_proto.h b/arch/x86/include/asm/amd_iommu_proto.h
index 916bc8111a01..1223c0fe03f5 100644
--- a/arch/x86/include/asm/amd_iommu_proto.h
+++ b/arch/x86/include/asm/amd_iommu_proto.h
@@ -24,8 +24,6 @@ struct amd_iommu;
 extern int amd_iommu_init_dma_ops(void);
 extern int amd_iommu_init_passthrough(void);
 extern irqreturn_t amd_iommu_int_handler(int irq, void *data);
-extern void amd_iommu_flush_all_domains(void);
-extern void amd_iommu_flush_all_devices(void);
 extern void amd_iommu_apply_erratum_63(u16 devid);
 extern void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu);
 extern int amd_iommu_init_devices(void);
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 3557f223f40b..bcf58ea55cfa 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -539,6 +539,40 @@ static int iommu_flush_dte(struct amd_iommu *iommu, u16 devid)
 	return iommu_queue_command(iommu, &cmd);
 }
 
+static void iommu_flush_dte_all(struct amd_iommu *iommu)
+{
+	u32 devid;
+
+	for (devid = 0; devid <= 0xffff; ++devid)
+		iommu_flush_dte(iommu, devid);
+
+	iommu_completion_wait(iommu);
+}
+
+/*
+ * This function uses heavy locking and may disable irqs for some time. But
+ * this is no issue because it is only called during resume.
+ */
+static void iommu_flush_tlb_all(struct amd_iommu *iommu)
+{
+	u32 dom_id;
+
+	for (dom_id = 0; dom_id <= 0xffff; ++dom_id) {
+		struct iommu_cmd cmd;
+		build_inv_iommu_pages(&cmd, 0, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
+				      dom_id, 1);
+		iommu_queue_command(iommu, &cmd);
+	}
+
+	iommu_completion_wait(iommu);
+}
+
+void iommu_flush_all_caches(struct amd_iommu *iommu)
+{
+	iommu_flush_dte_all(iommu);
+	iommu_flush_tlb_all(iommu);
+}
+
 /*
  * Command send function for invalidating a device table entry
  */
@@ -631,47 +665,6 @@ static void domain_flush_devices(struct protection_domain *domain)
 	spin_unlock_irqrestore(&domain->lock, flags);
 }
 
-static void iommu_flush_all_domain_devices(void)
-{
-	struct protection_domain *domain;
-	unsigned long flags;
-
-	spin_lock_irqsave(&amd_iommu_pd_lock, flags);
-
-	list_for_each_entry(domain, &amd_iommu_pd_list, list) {
-		domain_flush_devices(domain);
-		domain_flush_complete(domain);
-	}
-
-	spin_unlock_irqrestore(&amd_iommu_pd_lock, flags);
-}
-
-void amd_iommu_flush_all_devices(void)
-{
-	iommu_flush_all_domain_devices();
-}
-
-/*
- * This function uses heavy locking and may disable irqs for some time. But
- * this is no issue because it is only called during resume.
- */
-void amd_iommu_flush_all_domains(void)
-{
-	struct protection_domain *domain;
-	unsigned long flags;
-
-	spin_lock_irqsave(&amd_iommu_pd_lock, flags);
-
-	list_for_each_entry(domain, &amd_iommu_pd_list, list) {
-		spin_lock(&domain->lock);
-		domain_flush_tlb_pde(domain);
-		domain_flush_complete(domain);
-		spin_unlock(&domain->lock);
-	}
-
-	spin_unlock_irqrestore(&amd_iommu_pd_lock, flags);
-}
-
 /****************************************************************************
  *
  * The functions below are used the create the page table mappings for
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 246d727b65b7..8848dda808e2 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -180,6 +180,12 @@ static u32 dev_table_size;	/* size of the device table */
 static u32 alias_table_size;	/* size of the alias table */
 static u32 rlookup_table_size;	/* size if the rlookup table */
 
+/*
+ * This function flushes all internal caches of
+ * the IOMMU used by this driver.
+ */
+extern void iommu_flush_all_caches(struct amd_iommu *iommu);
+
 static inline void update_last_devid(u16 devid)
 {
 	if (devid > amd_iommu_last_bdf)
@@ -1244,6 +1250,7 @@ static void enable_iommus(void)
 		iommu_set_exclusion_range(iommu);
 		iommu_init_msi(iommu);
 		iommu_enable(iommu);
+		iommu_flush_all_caches(iommu);
 	}
 }
 
@@ -1274,8 +1281,8 @@ static void amd_iommu_resume(void)
 	 * we have to flush after the IOMMUs are enabled because a
 	 * disabled IOMMU will never execute the commands we send
 	 */
-	amd_iommu_flush_all_devices();
-	amd_iommu_flush_all_domains();
+	for_each_iommu(iommu)
+		iommu_flush_all_caches(iommu);
 }
 
 static int amd_iommu_suspend(void)
-- 
cgit v1.2.3


From cb41ed85efa01e633388314c03a4f3004c6b783b Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 5 Apr 2011 11:00:53 +0200
Subject: x86/amd-iommu: Flush device IOTLB if ATS is enabled

This patch implements a function to flush the IOTLB on
devices supporting ATS and makes sure that this TLB is also
flushed if necessary.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  3 +-
 arch/x86/kernel/amd_iommu.c            | 74 +++++++++++++++++++++++++++++++++-
 2 files changed, 75 insertions(+), 2 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 878ae008eb04..f5d184e7d5be 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -113,7 +113,8 @@
 /* command specific defines */
 #define CMD_COMPL_WAIT          0x01
 #define CMD_INV_DEV_ENTRY       0x02
-#define CMD_INV_IOMMU_PAGES     0x03
+#define CMD_INV_IOMMU_PAGES	0x03
+#define CMD_INV_IOTLB_PAGES	0x04
 
 #define CMD_COMPL_WAIT_STORE_MASK	0x01
 #define CMD_COMPL_WAIT_INT_MASK		0x02
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index bcf58ea55cfa..f3ce4338dade 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -18,6 +18,7 @@
  */
 
 #include <linux/pci.h>
+#include <linux/pci-ats.h>
 #include <linux/bitmap.h>
 #include <linux/slab.h>
 #include <linux/debugfs.h>
@@ -463,6 +464,37 @@ static void build_inv_iommu_pages(struct iommu_cmd *cmd, u64 address,
 		cmd->data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK;
 }
 
+static void build_inv_iotlb_pages(struct iommu_cmd *cmd, u16 devid, int qdep,
+				  u64 address, size_t size)
+{
+	u64 pages;
+	int s;
+
+	pages = iommu_num_pages(address, size, PAGE_SIZE);
+	s     = 0;
+
+	if (pages > 1) {
+		/*
+		 * If we have to flush more than one page, flush all
+		 * TLB entries for this domain
+		 */
+		address = CMD_INV_IOMMU_ALL_PAGES_ADDRESS;
+		s = 1;
+	}
+
+	address &= PAGE_MASK;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->data[0]  = devid;
+	cmd->data[0] |= (qdep & 0xff) << 24;
+	cmd->data[1]  = devid;
+	cmd->data[2]  = lower_32_bits(address);
+	cmd->data[3]  = upper_32_bits(address);
+	CMD_SET_TYPE(cmd, CMD_INV_IOTLB_PAGES);
+	if (s)
+		cmd->data[2] |= CMD_INV_IOMMU_PAGES_SIZE_MASK;
+}
+
 /*
  * Writes the command to the IOMMUs command buffer and informs the
  * hardware about the new command.
@@ -573,18 +605,48 @@ void iommu_flush_all_caches(struct amd_iommu *iommu)
 	iommu_flush_tlb_all(iommu);
 }
 
+/*
+ * Command send function for flushing on-device TLB
+ */
+static int device_flush_iotlb(struct device *dev, u64 address, size_t size)
+{
+	struct pci_dev *pdev = to_pci_dev(dev);
+	struct amd_iommu *iommu;
+	struct iommu_cmd cmd;
+	u16 devid;
+	int qdep;
+
+	qdep  = pci_ats_queue_depth(pdev);
+	devid = get_device_id(dev);
+	iommu = amd_iommu_rlookup_table[devid];
+
+	build_inv_iotlb_pages(&cmd, devid, qdep, address, size);
+
+	return iommu_queue_command(iommu, &cmd);
+}
+
 /*
  * Command send function for invalidating a device table entry
  */
 static int device_flush_dte(struct device *dev)
 {
 	struct amd_iommu *iommu;
+	struct pci_dev *pdev;
 	u16 devid;
+	int ret;
 
+	pdev  = to_pci_dev(dev);
 	devid = get_device_id(dev);
 	iommu = amd_iommu_rlookup_table[devid];
 
-	return iommu_flush_dte(iommu, devid);
+	ret = iommu_flush_dte(iommu, devid);
+	if (ret)
+		return ret;
+
+	if (pci_ats_enabled(pdev))
+		ret = device_flush_iotlb(dev, 0, ~0UL);
+
+	return ret;
 }
 
 /*
@@ -595,6 +657,7 @@ static int device_flush_dte(struct device *dev)
 static void __domain_flush_pages(struct protection_domain *domain,
 				 u64 address, size_t size, int pde)
 {
+	struct iommu_dev_data *dev_data;
 	struct iommu_cmd cmd;
 	int ret = 0, i;
 
@@ -611,6 +674,15 @@ static void __domain_flush_pages(struct protection_domain *domain,
 		ret |= iommu_queue_command(amd_iommus[i], &cmd);
 	}
 
+	list_for_each_entry(dev_data, &domain->dev_list, list) {
+		struct pci_dev *pdev = to_pci_dev(dev_data->dev);
+
+		if (!pci_ats_enabled(pdev))
+			continue;
+
+		ret |= device_flush_iotlb(dev_data->dev, address, size);
+	}
+
 	WARN_ON(ret);
 }
 
-- 
cgit v1.2.3


From 60f723b4117507c05c8b0b5c8b98ecc12a76878e Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 5 Apr 2011 12:50:24 +0200
Subject: x86/amd-iommu: Add flag to indicate IOTLB support

This patch adds a flag to the AMD IOMMU driver to indicate
that all IOMMUs present in the system support device IOTLBs.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h | 2 ++
 arch/x86/kernel/amd_iommu_init.c       | 4 ++++
 2 files changed, 6 insertions(+)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index f5d184e7d5be..cb811c965548 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -250,6 +250,8 @@ extern bool amd_iommu_dump;
 
 /* global flag if IOMMUs cache non-present entries */
 extern bool amd_iommu_np_cache;
+/* Only true if all IOMMUs support device IOTLBs */
+extern bool amd_iommu_iotlb_sup;
 
 /*
  * Make iterating over all IOMMUs easier
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 8848dda808e2..b6c634f3dc07 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -137,6 +137,7 @@ int amd_iommus_present;
 
 /* IOMMUs have a non-present cache? */
 bool amd_iommu_np_cache __read_mostly;
+bool amd_iommu_iotlb_sup __read_mostly = true;
 
 /*
  * The ACPI table parsing functions set this variable on an error
@@ -673,6 +674,9 @@ static void __init init_iommu_from_pci(struct amd_iommu *iommu)
 					MMIO_GET_LD(range));
 	iommu->evt_msi_num = MMIO_MSI_NUM(misc);
 
+	if (!(iommu->cap & (1 << IOMMU_CAP_IOTLB)))
+		amd_iommu_iotlb_sup = false;
+
 	if (!is_rd890_iommu(iommu->dev))
 		return;
 
-- 
cgit v1.2.3


From fd7b5535e10ce820f030842da3f289f80ec0d4f3 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 5 Apr 2011 15:31:08 +0200
Subject: x86/amd-iommu: Add ATS enable/disable code

This patch adds the necessary code to the AMD IOMMU driver
for enabling and disabling the ATS capability on a device
and to setup the IOMMU data structures correctly.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  2 ++
 arch/x86/kernel/amd_iommu.c            | 34 ++++++++++++++++++++++++++++------
 2 files changed, 30 insertions(+), 6 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index cb811c965548..7434377b2ab9 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -216,6 +216,8 @@
 #define IOMMU_PTE_IR (1ULL << 61)
 #define IOMMU_PTE_IW (1ULL << 62)
 
+#define DTE_FLAG_IOTLB	0x01
+
 #define IOMMU_PAGE_MASK (((1ULL << 52) - 1) & ~0xfffULL)
 #define IOMMU_PTE_PRESENT(pte) ((pte) & IOMMU_PTE_P)
 #define IOMMU_PTE_PAGE(pte) (phys_to_virt((pte) & IOMMU_PAGE_MASK))
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index f3ce4338dade..e4791f66aa38 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1452,17 +1452,22 @@ static bool dma_ops_domain(struct protection_domain *domain)
 	return domain->flags & PD_DMA_OPS_MASK;
 }
 
-static void set_dte_entry(u16 devid, struct protection_domain *domain)
+static void set_dte_entry(u16 devid, struct protection_domain *domain, bool ats)
 {
 	u64 pte_root = virt_to_phys(domain->pt_root);
+	u32 flags = 0;
 
 	pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK)
 		    << DEV_ENTRY_MODE_SHIFT;
 	pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV;
 
-	amd_iommu_dev_table[devid].data[2] = domain->id;
-	amd_iommu_dev_table[devid].data[1] = upper_32_bits(pte_root);
-	amd_iommu_dev_table[devid].data[0] = lower_32_bits(pte_root);
+	if (ats)
+		flags |= DTE_FLAG_IOTLB;
+
+	amd_iommu_dev_table[devid].data[3] |= flags;
+	amd_iommu_dev_table[devid].data[2]  = domain->id;
+	amd_iommu_dev_table[devid].data[1]  = upper_32_bits(pte_root);
+	amd_iommu_dev_table[devid].data[0]  = lower_32_bits(pte_root);
 }
 
 static void clear_dte_entry(u16 devid)
@@ -1479,16 +1484,22 @@ static void do_attach(struct device *dev, struct protection_domain *domain)
 {
 	struct iommu_dev_data *dev_data;
 	struct amd_iommu *iommu;
+	struct pci_dev *pdev;
+	bool ats = false;
 	u16 devid;
 
 	devid    = get_device_id(dev);
 	iommu    = amd_iommu_rlookup_table[devid];
 	dev_data = get_dev_data(dev);
+	pdev     = to_pci_dev(dev);
+
+	if (amd_iommu_iotlb_sup)
+		ats = pci_ats_enabled(pdev);
 
 	/* Update data structures */
 	dev_data->domain = domain;
 	list_add(&dev_data->list, &domain->dev_list);
-	set_dte_entry(devid, domain);
+	set_dte_entry(devid, domain, ats);
 
 	/* Do reference counting */
 	domain->dev_iommu[iommu->index] += 1;
@@ -1502,11 +1513,13 @@ static void do_detach(struct device *dev)
 {
 	struct iommu_dev_data *dev_data;
 	struct amd_iommu *iommu;
+	struct pci_dev *pdev;
 	u16 devid;
 
 	devid    = get_device_id(dev);
 	iommu    = amd_iommu_rlookup_table[devid];
 	dev_data = get_dev_data(dev);
+	pdev     = to_pci_dev(dev);
 
 	/* decrease reference counters */
 	dev_data->domain->dev_iommu[iommu->index] -= 1;
@@ -1581,9 +1594,13 @@ out_unlock:
 static int attach_device(struct device *dev,
 			 struct protection_domain *domain)
 {
+	struct pci_dev *pdev = to_pci_dev(dev);
 	unsigned long flags;
 	int ret;
 
+	if (amd_iommu_iotlb_sup)
+		pci_enable_ats(pdev, PAGE_SHIFT);
+
 	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
 	ret = __attach_device(dev, domain);
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
@@ -1640,12 +1657,16 @@ static void __detach_device(struct device *dev)
  */
 static void detach_device(struct device *dev)
 {
+	struct pci_dev *pdev = to_pci_dev(dev);
 	unsigned long flags;
 
 	/* lock device table */
 	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
 	__detach_device(dev);
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+
+	if (amd_iommu_iotlb_sup && pci_ats_enabled(pdev))
+		pci_disable_ats(pdev);
 }
 
 /*
@@ -1795,8 +1816,9 @@ static void update_device_table(struct protection_domain *domain)
 	struct iommu_dev_data *dev_data;
 
 	list_for_each_entry(dev_data, &domain->dev_list, list) {
+		struct pci_dev *pdev = to_pci_dev(dev_data->dev);
 		u16 devid = get_device_id(dev_data->dev);
-		set_dte_entry(devid, domain);
+		set_dte_entry(devid, domain, pci_ats_enabled(pdev));
 	}
 }
 
-- 
cgit v1.2.3


From d99ddec3eee0be8a43b2c1ff624b9dfaaa26b959 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Mon, 11 Apr 2011 11:03:18 +0200
Subject: x86/amd-iommu: Add extended feature detection

This patch adds detection of the extended features of an
AMD IOMMU. The available features are printed to dmesg on
boot.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_proto.h | 10 +++++++++-
 arch/x86/include/asm/amd_iommu_types.h | 17 +++++++++++++++++
 arch/x86/kernel/amd_iommu_init.c       | 24 ++++++++++++++++++++++--
 3 files changed, 48 insertions(+), 3 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_proto.h b/arch/x86/include/asm/amd_iommu_proto.h
index 1223c0fe03f5..a4ae6c3875eb 100644
--- a/arch/x86/include/asm/amd_iommu_proto.h
+++ b/arch/x86/include/asm/amd_iommu_proto.h
@@ -19,7 +19,7 @@
 #ifndef _ASM_X86_AMD_IOMMU_PROTO_H
 #define _ASM_X86_AMD_IOMMU_PROTO_H
 
-struct amd_iommu;
+#include <asm/amd_iommu_types.h>
 
 extern int amd_iommu_init_dma_ops(void);
 extern int amd_iommu_init_passthrough(void);
@@ -42,4 +42,12 @@ static inline bool is_rd890_iommu(struct pci_dev *pdev)
 	       (pdev->device == PCI_DEVICE_ID_RD890_IOMMU);
 }
 
+static inline bool iommu_feature(struct amd_iommu *iommu, u64 f)
+{
+	if (!(iommu->cap & (1 << IOMMU_CAP_EFR)))
+		return false;
+
+	return !!(iommu->features & f);
+}
+
 #endif /* _ASM_X86_AMD_IOMMU_PROTO_H  */
diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 878ae008eb04..5c24e4652347 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -68,12 +68,25 @@
 #define MMIO_CONTROL_OFFSET     0x0018
 #define MMIO_EXCL_BASE_OFFSET   0x0020
 #define MMIO_EXCL_LIMIT_OFFSET  0x0028
+#define MMIO_EXT_FEATURES	0x0030
 #define MMIO_CMD_HEAD_OFFSET	0x2000
 #define MMIO_CMD_TAIL_OFFSET	0x2008
 #define MMIO_EVT_HEAD_OFFSET	0x2010
 #define MMIO_EVT_TAIL_OFFSET	0x2018
 #define MMIO_STATUS_OFFSET	0x2020
 
+
+/* Extended Feature Bits */
+#define FEATURE_PREFETCH	(1ULL<<0)
+#define FEATURE_PPR		(1ULL<<1)
+#define FEATURE_X2APIC		(1ULL<<2)
+#define FEATURE_NX		(1ULL<<3)
+#define FEATURE_GT		(1ULL<<4)
+#define FEATURE_IA		(1ULL<<6)
+#define FEATURE_GA		(1ULL<<7)
+#define FEATURE_HE		(1ULL<<8)
+#define FEATURE_PC		(1ULL<<9)
+
 /* MMIO status bits */
 #define MMIO_STATUS_COM_WAIT_INT_MASK	0x04
 
@@ -227,6 +240,7 @@
 /* IOMMU capabilities */
 #define IOMMU_CAP_IOTLB   24
 #define IOMMU_CAP_NPCACHE 26
+#define IOMMU_CAP_EFR     27
 
 #define MAX_DOMAIN_ID 65536
 
@@ -371,6 +385,9 @@ struct amd_iommu {
 	/* flags read from acpi table */
 	u8 acpi_flags;
 
+	/* Extended features */
+	u64 features;
+
 	/*
 	 * Capability pointer. There could be more than one IOMMU per PCI
 	 * device function if there are more than one AMD IOMMU capability
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 8848dda808e2..047905dc3e14 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -299,9 +299,23 @@ static void iommu_feature_disable(struct amd_iommu *iommu, u8 bit)
 /* Function to enable the hardware */
 static void iommu_enable(struct amd_iommu *iommu)
 {
-	printk(KERN_INFO "AMD-Vi: Enabling IOMMU at %s cap 0x%hx\n",
+	static const char * const feat_str[] = {
+		"PreF", "PPR", "X2APIC", "NX", "GT", "[5]",
+		"IA", "GA", "HE", "PC", NULL
+	};
+	int i;
+
+	printk(KERN_INFO "AMD-Vi: Enabling IOMMU at %s cap 0x%hx",
 	       dev_name(&iommu->dev->dev), iommu->cap_ptr);
 
+	if (iommu->cap & (1 << IOMMU_CAP_EFR)) {
+		printk(KERN_CONT " extended features: ");
+		for (i = 0; feat_str[i]; ++i)
+			if (iommu_feature(iommu, (1ULL << i)))
+				printk(KERN_CONT " %s", feat_str[i]);
+	}
+	printk(KERN_CONT "\n");
+
 	iommu_feature_enable(iommu, CONTROL_IOMMU_EN);
 }
 
@@ -657,7 +671,7 @@ static void __init set_device_exclusion_range(u16 devid, struct ivmd_header *m)
 static void __init init_iommu_from_pci(struct amd_iommu *iommu)
 {
 	int cap_ptr = iommu->cap_ptr;
-	u32 range, misc;
+	u32 range, misc, low, high;
 	int i, j;
 
 	pci_read_config_dword(iommu->dev, cap_ptr + MMIO_CAP_HDR_OFFSET,
@@ -673,6 +687,12 @@ static void __init init_iommu_from_pci(struct amd_iommu *iommu)
 					MMIO_GET_LD(range));
 	iommu->evt_msi_num = MMIO_MSI_NUM(misc);
 
+	/* read extended feature bits */
+	low  = readl(iommu->mmio_base + MMIO_EXT_FEATURES);
+	high = readl(iommu->mmio_base + MMIO_EXT_FEATURES + 4);
+
+	iommu->features = ((u64)high << 32) | low;
+
 	if (!is_rd890_iommu(iommu->dev))
 		return;
 
-- 
cgit v1.2.3


From 58fc7f1419560efa9c426b829c195050e0147d7f Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Mon, 11 Apr 2011 11:13:24 +0200
Subject: x86/amd-iommu: Add support for invalidate_all command

This patch adds support for the invalidate_all command
present in new versions of the AMD IOMMU.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  1 +
 arch/x86/kernel/amd_iommu.c            | 24 ++++++++++++++++++++++--
 2 files changed, 23 insertions(+), 2 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 5c24e4652347..df62d26ed2ab 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -127,6 +127,7 @@
 #define CMD_COMPL_WAIT          0x01
 #define CMD_INV_DEV_ENTRY       0x02
 #define CMD_INV_IOMMU_PAGES     0x03
+#define CMD_INV_ALL		0x08
 
 #define CMD_COMPL_WAIT_STORE_MASK	0x01
 #define CMD_COMPL_WAIT_INT_MASK		0x02
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index bcf58ea55cfa..d6192bcf9f09 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -463,6 +463,12 @@ static void build_inv_iommu_pages(struct iommu_cmd *cmd, u64 address,
 		cmd->data[2] |= CMD_INV_IOMMU_PAGES_PDE_MASK;
 }
 
+static void build_inv_all(struct iommu_cmd *cmd)
+{
+	memset(cmd, 0, sizeof(*cmd));
+	CMD_SET_TYPE(cmd, CMD_INV_ALL);
+}
+
 /*
  * Writes the command to the IOMMUs command buffer and informs the
  * hardware about the new command.
@@ -567,10 +573,24 @@ static void iommu_flush_tlb_all(struct amd_iommu *iommu)
 	iommu_completion_wait(iommu);
 }
 
+static void iommu_flush_all(struct amd_iommu *iommu)
+{
+	struct iommu_cmd cmd;
+
+	build_inv_all(&cmd);
+
+	iommu_queue_command(iommu, &cmd);
+	iommu_completion_wait(iommu);
+}
+
 void iommu_flush_all_caches(struct amd_iommu *iommu)
 {
-	iommu_flush_dte_all(iommu);
-	iommu_flush_tlb_all(iommu);
+	if (iommu_feature(iommu, FEATURE_IA)) {
+		iommu_flush_all(iommu);
+	} else {
+		iommu_flush_dte_all(iommu);
+		iommu_flush_tlb_all(iommu);
+	}
 }
 
 /*
-- 
cgit v1.2.3


From 72fe00f01f9a3240a1073be27aeaf4fc476cc662 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 10 May 2011 10:50:42 +0200
Subject: x86/amd-iommu: Use threaded interupt handler

Move the interupt handling for the iommu into the interupt
thread to reduce latencies and prepare interupt handling for
pri handling.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_proto.h | 1 +
 arch/x86/kernel/amd_iommu.c            | 7 ++++++-
 arch/x86/kernel/amd_iommu_init.c       | 9 +++++----
 3 files changed, 12 insertions(+), 5 deletions(-)

(limited to 'arch/x86/include/asm')

diff --git a/arch/x86/include/asm/amd_iommu_proto.h b/arch/x86/include/asm/amd_iommu_proto.h
index a4ae6c3875eb..55d95eb789b3 100644
--- a/arch/x86/include/asm/amd_iommu_proto.h
+++ b/arch/x86/include/asm/amd_iommu_proto.h
@@ -23,6 +23,7 @@
 
 extern int amd_iommu_init_dma_ops(void);
 extern int amd_iommu_init_passthrough(void);
+extern irqreturn_t amd_iommu_int_thread(int irq, void *data);
 extern irqreturn_t amd_iommu_int_handler(int irq, void *data);
 extern void amd_iommu_apply_erratum_63(u16 devid);
 extern void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu);
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index dc5dddafe5c2..873e7e1ead7b 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -366,7 +366,7 @@ static void iommu_poll_events(struct amd_iommu *iommu)
 	spin_unlock_irqrestore(&iommu->lock, flags);
 }
 
-irqreturn_t amd_iommu_int_handler(int irq, void *data)
+irqreturn_t amd_iommu_int_thread(int irq, void *data)
 {
 	struct amd_iommu *iommu;
 
@@ -376,6 +376,11 @@ irqreturn_t amd_iommu_int_handler(int irq, void *data)
 	return IRQ_HANDLED;
 }
 
+irqreturn_t amd_iommu_int_handler(int irq, void *data)
+{
+	return IRQ_WAKE_THREAD;
+}
+
 /****************************************************************************
  *
  * IOMMU command queuing functions
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 28b078133688..9179c21120a8 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -1034,10 +1034,11 @@ static int iommu_setup_msi(struct amd_iommu *iommu)
 	if (pci_enable_msi(iommu->dev))
 		return 1;
 
-	r = request_irq(iommu->dev->irq, amd_iommu_int_handler,
-			IRQF_SAMPLE_RANDOM,
-			"AMD-Vi",
-			NULL);
+	r = request_threaded_irq(iommu->dev->irq,
+				 amd_iommu_int_handler,
+				 amd_iommu_int_thread,
+				 0, "AMD-Vi",
+				 iommu->dev);
 
 	if (r) {
 		pci_disable_msi(iommu->dev);
-- 
cgit v1.2.3