From 3885123da8335dc6b67387e5e626acbffc56f664 Mon Sep 17 00:00:00 2001
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date: Fri, 10 Jul 2009 10:04:50 +0900
Subject: swiotlb: remove unused swiotlb_alloc_boot()

Nobody uses swiotlb_alloc_boot().

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Becky Bruce <beckyb@kernel.crashing.org>
---
 arch/x86/kernel/pci-swiotlb.c | 5 -----
 1 file changed, 5 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index 6af96ee44200..0ac7cd524788 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -13,11 +13,6 @@
 
 int swiotlb __read_mostly;
 
-void * __init swiotlb_alloc_boot(size_t size, unsigned long nslabs)
-{
-	return alloc_bootmem_low_pages(size);
-}
-
 void *swiotlb_alloc(unsigned order, unsigned long nslabs)
 {
 	return (void *)__get_free_pages(GFP_DMA | __GFP_NOWARN, order);
-- 
cgit v1.2.3-59-g8ed1b


From bb52196be37ce154ddc50b1f39496146d181cbe7 Mon Sep 17 00:00:00 2001
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date: Fri, 10 Jul 2009 10:04:51 +0900
Subject: swiotlb: remove unused swiotlb_alloc()

Nobody uses swiotlb_alloc().

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Becky Bruce <beckyb@kernel.crashing.org>
---
 arch/x86/kernel/pci-swiotlb.c | 5 -----
 include/linux/swiotlb.h       | 2 --
 lib/swiotlb.c                 | 8 ++------
 3 files changed, 2 insertions(+), 13 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index 0ac7cd524788..ea675cfe76fe 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -13,11 +13,6 @@
 
 int swiotlb __read_mostly;
 
-void *swiotlb_alloc(unsigned order, unsigned long nslabs)
-{
-	return (void *)__get_free_pages(GFP_DMA | __GFP_NOWARN, order);
-}
-
 dma_addr_t swiotlb_phys_to_bus(struct device *hwdev, phys_addr_t paddr)
 {
 	return paddr;
diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
index 94db70444c17..6bc50944040f 100644
--- a/include/linux/swiotlb.h
+++ b/include/linux/swiotlb.h
@@ -23,8 +23,6 @@ struct scatterlist;
 extern void
 swiotlb_init(void);
 
-extern void *swiotlb_alloc(unsigned order, unsigned long nslabs);
-
 extern dma_addr_t swiotlb_phys_to_bus(struct device *hwdev,
 				      phys_addr_t address);
 extern phys_addr_t swiotlb_bus_to_phys(struct device *hwdev,
diff --git a/lib/swiotlb.c b/lib/swiotlb.c
index 9edfdd442edc..3c4c21cdf43d 100644
--- a/lib/swiotlb.c
+++ b/lib/swiotlb.c
@@ -114,11 +114,6 @@ setup_io_tlb_npages(char *str)
 __setup("swiotlb=", setup_io_tlb_npages);
 /* make io_tlb_overflow tunable too? */
 
-void * __weak swiotlb_alloc(unsigned order, unsigned long nslabs)
-{
-	return (void *)__get_free_pages(GFP_DMA | __GFP_NOWARN, order);
-}
-
 dma_addr_t __weak swiotlb_phys_to_bus(struct device *hwdev, phys_addr_t paddr)
 {
 	return paddr;
@@ -240,7 +235,8 @@ swiotlb_late_init_with_default_size(size_t default_size)
 	bytes = io_tlb_nslabs << IO_TLB_SHIFT;
 
 	while ((SLABS_PER_PAGE << order) > IO_TLB_MIN_SLABS) {
-		io_tlb_start = swiotlb_alloc(order, io_tlb_nslabs);
+		io_tlb_start = (void *)__get_free_pages(GFP_DMA | __GFP_NOWARN,
+							order);
 		if (io_tlb_start)
 			break;
 		order--;
-- 
cgit v1.2.3-59-g8ed1b


From cf56e3f2e8a8d5b7bc719980869b0e7985c256f3 Mon Sep 17 00:00:00 2001
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date: Fri, 10 Jul 2009 10:04:52 +0900
Subject: swiotlb: remove swiotlb_arch_range_needs_mapping

Nobody uses swiotlb_arch_range_needs_mapping().

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Becky Bruce <beckyb@kernel.crashing.org>
---
 arch/x86/kernel/pci-swiotlb.c |  5 -----
 include/linux/swiotlb.h       |  2 --
 lib/swiotlb.c                 | 15 ++-------------
 3 files changed, 2 insertions(+), 20 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index ea675cfe76fe..165bd7f93bb1 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -23,11 +23,6 @@ phys_addr_t swiotlb_bus_to_phys(struct device *hwdev, dma_addr_t baddr)
 	return baddr;
 }
 
-int __weak swiotlb_arch_range_needs_mapping(phys_addr_t paddr, size_t size)
-{
-	return 0;
-}
-
 static void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 					dma_addr_t *dma_handle, gfp_t flags)
 {
diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
index 6bc50944040f..a977da24f17c 100644
--- a/include/linux/swiotlb.h
+++ b/include/linux/swiotlb.h
@@ -28,8 +28,6 @@ extern dma_addr_t swiotlb_phys_to_bus(struct device *hwdev,
 extern phys_addr_t swiotlb_bus_to_phys(struct device *hwdev,
 				       dma_addr_t address);
 
-extern int swiotlb_arch_range_needs_mapping(phys_addr_t paddr, size_t size);
-
 extern void
 *swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 			dma_addr_t *dma_handle, gfp_t flags);
diff --git a/lib/swiotlb.c b/lib/swiotlb.c
index 3c4c21cdf43d..dc1cd1f5369e 100644
--- a/lib/swiotlb.c
+++ b/lib/swiotlb.c
@@ -141,11 +141,6 @@ int __weak swiotlb_arch_address_needs_mapping(struct device *hwdev,
 	return !is_buffer_dma_capable(dma_get_mask(hwdev), addr, size);
 }
 
-int __weak swiotlb_arch_range_needs_mapping(phys_addr_t paddr, size_t size)
-{
-	return 0;
-}
-
 static void swiotlb_print_info(unsigned long bytes)
 {
 	phys_addr_t pstart, pend;
@@ -312,11 +307,6 @@ address_needs_mapping(struct device *hwdev, dma_addr_t addr, size_t size)
 	return swiotlb_arch_address_needs_mapping(hwdev, addr, size);
 }
 
-static inline int range_needs_mapping(phys_addr_t paddr, size_t size)
-{
-	return swiotlb_force || swiotlb_arch_range_needs_mapping(paddr, size);
-}
-
 static int is_swiotlb_buffer(char *addr)
 {
 	return addr >= io_tlb_start && addr < io_tlb_end;
@@ -646,8 +636,7 @@ dma_addr_t swiotlb_map_page(struct device *dev, struct page *page,
 	 * we can safely return the device addr and not worry about bounce
 	 * buffering it.
 	 */
-	if (!address_needs_mapping(dev, dev_addr, size) &&
-	    !range_needs_mapping(phys, size))
+	if (!address_needs_mapping(dev, dev_addr, size) && !swiotlb_force)
 		return dev_addr;
 
 	/*
@@ -810,7 +799,7 @@ swiotlb_map_sg_attrs(struct device *hwdev, struct scatterlist *sgl, int nelems,
 		phys_addr_t paddr = sg_phys(sg);
 		dma_addr_t dev_addr = swiotlb_phys_to_bus(hwdev, paddr);
 
-		if (range_needs_mapping(paddr, sg->length) ||
+		if (swiotlb_force ||
 		    address_needs_mapping(hwdev, dev_addr, sg->length)) {
 			void *map = map_single(hwdev, sg_phys(sg),
 					       sg->length, dir);
-- 
cgit v1.2.3-59-g8ed1b


From a4c2baa6e148adfb27beaf16b6fb6d465b5b3acb Mon Sep 17 00:00:00 2001
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date: Fri, 10 Jul 2009 10:04:55 +0900
Subject: x86: replace is_buffer_dma_capable() with dma_capable

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
 arch/x86/kernel/pci-dma.c     | 2 +-
 arch/x86/kernel/pci-gart_64.c | 5 ++---
 arch/x86/kernel/pci-nommu.c   | 2 +-
 3 files changed, 4 insertions(+), 5 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c
index 1a041bcf506b..3c945c0b3501 100644
--- a/arch/x86/kernel/pci-dma.c
+++ b/arch/x86/kernel/pci-dma.c
@@ -147,7 +147,7 @@ again:
 		return NULL;
 
 	addr = page_to_phys(page);
-	if (!is_buffer_dma_capable(dma_mask, addr, size)) {
+	if (addr + size > dma_mask) {
 		__free_pages(page, get_order(size));
 
 		if (dma_mask < DMA_BIT_MASK(32) && !(flag & GFP_DMA)) {
diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index d2e56b8f48e7..98a827ee9ed7 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -190,14 +190,13 @@ static void iommu_full(struct device *dev, size_t size, int dir)
 static inline int
 need_iommu(struct device *dev, unsigned long addr, size_t size)
 {
-	return force_iommu ||
-		!is_buffer_dma_capable(*dev->dma_mask, addr, size);
+	return force_iommu || !dma_capable(dev, addr, size);
 }
 
 static inline int
 nonforced_iommu(struct device *dev, unsigned long addr, size_t size)
 {
-	return !is_buffer_dma_capable(*dev->dma_mask, addr, size);
+	return !dma_capable(dev, addr, size);
 }
 
 /* Map a single continuous physical area into the IOMMU.
diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c
index 71d412a09f30..c0a4222bf62b 100644
--- a/arch/x86/kernel/pci-nommu.c
+++ b/arch/x86/kernel/pci-nommu.c
@@ -14,7 +14,7 @@
 static int
 check_addr(char *name, struct device *hwdev, dma_addr_t bus, size_t size)
 {
-	if (hwdev && !is_buffer_dma_capable(*hwdev->dma_mask, bus, size)) {
+	if (hwdev && !dma_capable(hwdev, bus, size)) {
 		if (*hwdev->dma_mask >= DMA_BIT_MASK(32))
 			printk(KERN_ERR
 			    "nommu_%s: overflow %Lx+%zu of device mask %Lx\n",
-- 
cgit v1.2.3-59-g8ed1b


From b683d42693c4e92b838117f5c6f7b90bfa1525c9 Mon Sep 17 00:00:00 2001
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date: Fri, 10 Jul 2009 10:05:04 +0900
Subject: x86: remove unused swiotlb_phys_to_bus() and swiotlb_bus_to_phys()

phys_to_dma() and dma_to_phys() are used instead of
swiotlb_phys_to_bus() and swiotlb_bus_to_phys().

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
 arch/x86/kernel/pci-swiotlb.c | 10 ----------
 1 file changed, 10 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-swiotlb.c b/arch/x86/kernel/pci-swiotlb.c
index 165bd7f93bb1..e8a35016115f 100644
--- a/arch/x86/kernel/pci-swiotlb.c
+++ b/arch/x86/kernel/pci-swiotlb.c
@@ -13,16 +13,6 @@
 
 int swiotlb __read_mostly;
 
-dma_addr_t swiotlb_phys_to_bus(struct device *hwdev, phys_addr_t paddr)
-{
-	return paddr;
-}
-
-phys_addr_t swiotlb_bus_to_phys(struct device *hwdev, dma_addr_t baddr)
-{
-	return baddr;
-}
-
 static void *x86_swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 					dma_addr_t *dma_handle, gfp_t flags)
 {
-- 
cgit v1.2.3-59-g8ed1b


From a8ad568dd8ca122aa8048ea067d3599820d1c1b4 Mon Sep 17 00:00:00 2001
From: Arnd Bergmann <arnd@arndb.de>
Date: Mon, 10 Aug 2009 11:53:10 +0900
Subject: dma-ops: Remove flush_write_buffers() in dma-mapping-common.h

This moves flush_write_buffers() in
asm-generic/dma-mapping-common.h to
arch/x86/kernel/pci-nommu.c.

The purpose of this patch is that, we can avoid defining NULL
flush_write_buffers() on IA64 and SPARC.

dma-mapping-common.h is used by X86 and IA64 (and SPARC soon)
but only X86 with CONFIG_X86_OOSTORE or CONFIG_X86_PPRO_FENCE
actually uses flush_write_buffers(). CONFIG_X86_OOSTORE or
CONFIG_X86_PPRO_FENCE is usable with only kernel/pci-nommu.c
(that is, not usable with other X86 IOMMU implementations such
as SWIOTLB, VT-d, etc) so we can safely move
flush_write_buffers() in asm-generic/dma-mapping-common.h to
arch/x86/kernel/pci-nommu.c.

The further discussion is:

  http://lkml.org/lkml/2009/6/28/104

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: davem@davemloft.net
Cc: tony.luck@intel.com
Cc: fenghua.yu@intel.com
LKML-Reference: <1249872797-1314-2-git-send-email-fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 arch/x86/kernel/pci-nommu.c              | 27 ++++++++++++++++++++++-----
 include/asm-generic/dma-mapping-common.h |  6 ------
 2 files changed, 22 insertions(+), 11 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c
index c0a4222bf62b..a3933d4330cd 100644
--- a/arch/x86/kernel/pci-nommu.c
+++ b/arch/x86/kernel/pci-nommu.c
@@ -79,12 +79,29 @@ static void nommu_free_coherent(struct device *dev, size_t size, void *vaddr,
 	free_pages((unsigned long)vaddr, get_order(size));
 }
 
+static void nommu_sync_single_for_device(struct device *dev,
+			dma_addr_t addr, size_t size,
+			enum dma_data_direction dir)
+{
+	flush_write_buffers();
+}
+
+
+static void nommu_sync_sg_for_device(struct device *dev,
+			struct scatterlist *sg, int nelems,
+			enum dma_data_direction dir)
+{
+	flush_write_buffers();
+}
+
 struct dma_map_ops nommu_dma_ops = {
-	.alloc_coherent	= dma_generic_alloc_coherent,
-	.free_coherent	= nommu_free_coherent,
-	.map_sg		= nommu_map_sg,
-	.map_page	= nommu_map_page,
-	.is_phys	= 1,
+	.alloc_coherent		= dma_generic_alloc_coherent,
+	.free_coherent		= nommu_free_coherent,
+	.map_sg			= nommu_map_sg,
+	.map_page		= nommu_map_page,
+	.sync_single_for_device = nommu_sync_single_for_device,
+	.sync_sg_for_device	= nommu_sync_sg_for_device,
+	.is_phys		= 1,
 };
 
 void __init no_iommu_init(void)
diff --git a/include/asm-generic/dma-mapping-common.h b/include/asm-generic/dma-mapping-common.h
index 5406a601185c..e694263445f7 100644
--- a/include/asm-generic/dma-mapping-common.h
+++ b/include/asm-generic/dma-mapping-common.h
@@ -103,7 +103,6 @@ static inline void dma_sync_single_for_cpu(struct device *dev, dma_addr_t addr,
 	if (ops->sync_single_for_cpu)
 		ops->sync_single_for_cpu(dev, addr, size, dir);
 	debug_dma_sync_single_for_cpu(dev, addr, size, dir);
-	flush_write_buffers();
 }
 
 static inline void dma_sync_single_for_device(struct device *dev,
@@ -116,7 +115,6 @@ static inline void dma_sync_single_for_device(struct device *dev,
 	if (ops->sync_single_for_device)
 		ops->sync_single_for_device(dev, addr, size, dir);
 	debug_dma_sync_single_for_device(dev, addr, size, dir);
-	flush_write_buffers();
 }
 
 static inline void dma_sync_single_range_for_cpu(struct device *dev,
@@ -132,7 +130,6 @@ static inline void dma_sync_single_range_for_cpu(struct device *dev,
 		ops->sync_single_range_for_cpu(dev, addr, offset, size, dir);
 		debug_dma_sync_single_range_for_cpu(dev, addr, offset, size, dir);
 
-		flush_write_buffers();
 	} else
 		dma_sync_single_for_cpu(dev, addr, size, dir);
 }
@@ -150,7 +147,6 @@ static inline void dma_sync_single_range_for_device(struct device *dev,
 		ops->sync_single_range_for_device(dev, addr, offset, size, dir);
 		debug_dma_sync_single_range_for_device(dev, addr, offset, size, dir);
 
-		flush_write_buffers();
 	} else
 		dma_sync_single_for_device(dev, addr, size, dir);
 }
@@ -165,7 +161,6 @@ dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
 	if (ops->sync_sg_for_cpu)
 		ops->sync_sg_for_cpu(dev, sg, nelems, dir);
 	debug_dma_sync_sg_for_cpu(dev, sg, nelems, dir);
-	flush_write_buffers();
 }
 
 static inline void
@@ -179,7 +174,6 @@ dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
 		ops->sync_sg_for_device(dev, sg, nelems, dir);
 	debug_dma_sync_sg_for_device(dev, sg, nelems, dir);
 
-	flush_write_buffers();
 }
 
 #define dma_map_single(d, a, s, r) dma_map_single_attrs(d, a, s, r, NULL)
-- 
cgit v1.2.3-59-g8ed1b


From e3e59876e82a5e1a07f365d5474e7b6943524725 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 14:02:10 +0200
Subject: x86/amd-iommu: Dump fault entry on DTE error

This patch adds code to dump the content of the device table
entry which caused an ILLEGAL_DEV_TABLE_ENTRY error from the
IOMMU hardware.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 6c99f5037801..364c6de26374 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -138,6 +138,15 @@ static int iommu_has_npcache(struct amd_iommu *iommu)
  *
  ****************************************************************************/
 
+static void dump_dte_entry(u16 devid)
+{
+	int i;
+
+	for (i = 0; i < 8; ++i)
+		pr_err("AMD-Vi: DTE[%d]: %08x\n", i,
+			amd_iommu_dev_table[devid].data[i]);
+}
+
 static void iommu_print_event(void *__evt)
 {
 	u32 *event = __evt;
@@ -155,6 +164,7 @@ static void iommu_print_event(void *__evt)
 		       "address=0x%016llx flags=0x%04x]\n",
 		       PCI_BUS(devid), PCI_SLOT(devid), PCI_FUNC(devid),
 		       address, flags);
+		dump_dte_entry(devid);
 		break;
 	case EVENT_TYPE_IO_FAULT:
 		printk("IO_PAGE_FAULT device=%02x:%02x.%x "
-- 
cgit v1.2.3-59-g8ed1b


From 945b4ac44e5700acd3d974c176c8ace34b4d2e8e Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 14:25:02 +0200
Subject: x86/amd-iommu: Dump illegal command on ILLEGAL_COMMAND_ERROR

This patch adds code to dump the command which caused an
ILLEGAL_COMMAND_ERROR raised by the IOMMU hardware.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 364c6de26374..e62b35f5df1b 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -147,6 +147,15 @@ static void dump_dte_entry(u16 devid)
 			amd_iommu_dev_table[devid].data[i]);
 }
 
+static void dump_command(unsigned long phys_addr)
+{
+	struct iommu_cmd *cmd = phys_to_virt(phys_addr);
+	int i;
+
+	for (i = 0; i < 4; ++i)
+		pr_err("AMD-Vi: CMD[%d]: %08x\n", i, cmd->data[i]);
+}
+
 static void iommu_print_event(void *__evt)
 {
 	u32 *event = __evt;
@@ -186,6 +195,7 @@ static void iommu_print_event(void *__evt)
 		break;
 	case EVENT_TYPE_ILL_CMD:
 		printk("ILLEGAL_COMMAND_ERROR address=0x%016llx]\n", address);
+		dump_command(address);
 		break;
 	case EVENT_TYPE_CMD_HARD_ERR:
 		printk("COMMAND_HARDWARE_ERROR address=0x%016llx "
-- 
cgit v1.2.3-59-g8ed1b


From e394d72aa8b319211b8f947d151d9d50b0fde842 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 15:28:33 +0200
Subject: x86/amd-iommu: Introduce function for iommu-local domain flush

This patch introduces a function to flush all domain tlbs
for on one given IOMMU. This is required later to reset the
command buffer on one IOMMU.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 49 ++++++++++++++++++++++++++++++---------------
 1 file changed, 33 insertions(+), 16 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index e62b35f5df1b..64cc582feb9b 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -465,38 +465,55 @@ static void iommu_flush_tlb_pde(struct amd_iommu *iommu, u16 domid)
 }
 
 /*
- * This function is used to flush the IO/TLB for a given protection domain
- * on every IOMMU in the system
+ * This function flushes one domain on one IOMMU
  */
-static void iommu_flush_domain(u16 domid)
+static void flush_domain_on_iommu(struct amd_iommu *iommu, u16 domid)
 {
-	unsigned long flags;
-	struct amd_iommu *iommu;
 	struct iommu_cmd cmd;
-
-	INC_STATS_COUNTER(domain_flush_all);
+	unsigned long flags;
 
 	__iommu_build_inv_iommu_pages(&cmd, CMD_INV_IOMMU_ALL_PAGES_ADDRESS,
 				      domid, 1, 1);
 
-	for_each_iommu(iommu) {
-		spin_lock_irqsave(&iommu->lock, flags);
-		__iommu_queue_command(iommu, &cmd);
-		__iommu_completion_wait(iommu);
-		__iommu_wait_for_completion(iommu);
-		spin_unlock_irqrestore(&iommu->lock, flags);
-	}
+	spin_lock_irqsave(&iommu->lock, flags);
+	__iommu_queue_command(iommu, &cmd);
+	__iommu_completion_wait(iommu);
+	__iommu_wait_for_completion(iommu);
+	spin_unlock_irqrestore(&iommu->lock, flags);
 }
 
-void amd_iommu_flush_all_domains(void)
+static void flush_all_domains_on_iommu(struct amd_iommu *iommu)
 {
 	int i;
 
 	for (i = 1; i < MAX_DOMAIN_ID; ++i) {
 		if (!test_bit(i, amd_iommu_pd_alloc_bitmap))
 			continue;
-		iommu_flush_domain(i);
+		flush_domain_on_iommu(iommu, i);
 	}
+
+}
+
+/*
+ * This function is used to flush the IO/TLB for a given protection domain
+ * on every IOMMU in the system
+ */
+static void iommu_flush_domain(u16 domid)
+{
+	struct amd_iommu *iommu;
+
+	INC_STATS_COUNTER(domain_flush_all);
+
+	for_each_iommu(iommu)
+		flush_domain_on_iommu(iommu, domid);
+}
+
+void amd_iommu_flush_all_domains(void)
+{
+	struct amd_iommu *iommu;
+
+	for_each_iommu(iommu)
+		flush_all_domains_on_iommu(iommu);
 }
 
 void amd_iommu_flush_all_devices(void)
-- 
cgit v1.2.3-59-g8ed1b


From f2430bd104bec2706315e9e983a9d9f828ff9565 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 26 Aug 2009 12:10:19 +0200
Subject: x86/amd-iommu: Remove some merge helper code

This patch removes some left-overs which where put into the code to
simplify merging code which also depends on changes in other trees.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 6 ------
 1 file changed, 6 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 6c99f5037801..70fdef54e066 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -41,9 +41,7 @@ static DEFINE_RWLOCK(amd_iommu_devtable_lock);
 static LIST_HEAD(iommu_pd_list);
 static DEFINE_SPINLOCK(iommu_pd_list_lock);
 
-#ifdef CONFIG_IOMMU_API
 static struct iommu_ops amd_iommu_ops;
-#endif
 
 /*
  * general struct to manage commands send to an IOMMU
@@ -62,10 +60,6 @@ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 				      unsigned long start_page,
 				      unsigned int pages);
 
-#ifndef BUS_NOTIFY_UNBOUND_DRIVER
-#define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
-#endif
-
 #ifdef CONFIG_AMD_IOMMU_STATS
 
 /*
-- 
cgit v1.2.3-59-g8ed1b


From 4c6f40d4e0f0bba77a5f27eec4e1c6d1c457d324 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 16:43:58 +0200
Subject: x86/amd-iommu: replace "AMD IOMMU" by "AMD-Vi"

This patch replaces the "AMD IOMMU" printk strings with the
official name for the hardware: "AMD-Vi".

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  2 +-
 arch/x86/kernel/amd_iommu.c            |  2 +-
 arch/x86/kernel/amd_iommu_init.c       | 12 ++++++------
 3 files changed, 8 insertions(+), 8 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 0c878caaa0a2..106e1305ef86 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -198,7 +198,7 @@ extern bool amd_iommu_dump;
 #define DUMP_printk(format, arg...)					\
 	do {								\
 		if (amd_iommu_dump)						\
-			printk(KERN_INFO "AMD IOMMU: " format, ## arg);	\
+			printk(KERN_INFO "AMD-Vi: " format, ## arg);	\
 	} while(0);
 
 /*
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 70fdef54e066..3e62d7836522 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -141,7 +141,7 @@ static void iommu_print_event(void *__evt)
 	int flags = (event[1] >> EVENT_FLAGS_SHIFT) & EVENT_FLAGS_MASK;
 	u64 address = (u64)(((u64)event[3]) << 32) | event[2];
 
-	printk(KERN_ERR "AMD IOMMU: Event logged [");
+	printk(KERN_ERR "AMD-Vi: Event logged [");
 
 	switch (type) {
 	case EVENT_TYPE_ILL_DEV:
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index c1b17e97252e..169958ad6243 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -252,7 +252,7 @@ static void __init 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 IOMMU: Enabling IOMMU at %s cap 0x%hx\n",
+	printk(KERN_INFO "AMD-Vi: Enabling IOMMU at %s cap 0x%hx\n",
 	       dev_name(&iommu->dev->dev), iommu->cap_ptr);
 
 	iommu_feature_enable(iommu, CONTROL_IOMMU_EN);
@@ -902,7 +902,7 @@ static int __init iommu_setup_msi(struct amd_iommu *iommu)
 
 	r = request_irq(iommu->dev->irq, amd_iommu_int_handler,
 			IRQF_SAMPLE_RANDOM,
-			"AMD IOMMU",
+			"AMD-Vi",
 			NULL);
 
 	if (r) {
@@ -1150,7 +1150,7 @@ int __init amd_iommu_init(void)
 
 
 	if (no_iommu) {
-		printk(KERN_INFO "AMD IOMMU disabled by kernel command line\n");
+		printk(KERN_INFO "AMD-Vi disabled by kernel command line\n");
 		return 0;
 	}
 
@@ -1248,16 +1248,16 @@ int __init amd_iommu_init(void)
 
 	enable_iommus();
 
-	printk(KERN_INFO "AMD IOMMU: device isolation ");
+	printk(KERN_INFO "AMD-Vi: device isolation ");
 	if (amd_iommu_isolate)
 		printk("enabled\n");
 	else
 		printk("disabled\n");
 
 	if (amd_iommu_unmap_flush)
-		printk(KERN_INFO "AMD IOMMU: IO/TLB flush on unmap enabled\n");
+		printk(KERN_INFO "AMD-Vi: IO/TLB flush on unmap enabled\n");
 	else
-		printk(KERN_INFO "AMD IOMMU: Lazy IO/TLB flushing enabled\n");
+		printk(KERN_INFO "AMD-Vi: Lazy IO/TLB flushing enabled\n");
 
 out:
 	return ret;
-- 
cgit v1.2.3-59-g8ed1b


From ae908c22aa2b9f7d4b41bd02d14e473f79c22dd3 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 16:52:16 +0200
Subject: x86/amd-iommu: Remove redundant 'IOMMU' string

The 'IOMMU: ' prefix is not necessary because the
DUMP_printk macro already prints its own prefix.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu_init.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index 169958ad6243..264b3ef9dd6e 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -858,7 +858,7 @@ static int __init init_iommu_all(struct acpi_table_header *table)
 		switch (*p) {
 		case ACPI_IVHD_TYPE:
 
-			DUMP_printk("IOMMU: device: %02x:%02x.%01x cap: %04x "
+			DUMP_printk("device: %02x:%02x.%01x cap: %04x "
 				    "seg: %d flags: %01x info %04x\n",
 				    PCI_BUS(h->devid), PCI_SLOT(h->devid),
 				    PCI_FUNC(h->devid), h->cap_ptr,
-- 
cgit v1.2.3-59-g8ed1b


From e0faf54ee82bf9c07f0307b4391caad4020bd659 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 15:45:51 +0200
Subject: x86/amd-iommu: fix broken check in amd_iommu_flush_all_devices

The amd_iommu_pd_table is indexed by protection domain
number and not by device id. So this check is broken and
must be removed.

Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 2 --
 1 file changed, 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 3e62d7836522..009d722af009 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -479,8 +479,6 @@ void amd_iommu_flush_all_devices(void)
 	int i;
 
 	for (i = 0; i <= amd_iommu_last_bdf; ++i) {
-		if (amd_iommu_pd_table[i] == NULL)
-			continue;
 
 		iommu = amd_iommu_rlookup_table[i];
 		if (!iommu)
-- 
cgit v1.2.3-59-g8ed1b


From d586d7852ccd0cecb502bf4809f827e60c486af0 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 15:39:23 +0200
Subject: x86/amd-iommu: Add function to flush all DTEs on one IOMMU

This function flushes all DTE entries on one IOMMU for all
devices behind this IOMMU. This is required for command
buffer resetting later.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 13 +++++++++++++
 1 file changed, 13 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 64cc582feb9b..2dc093370d2d 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -516,6 +516,19 @@ void amd_iommu_flush_all_domains(void)
 		flush_all_domains_on_iommu(iommu);
 }
 
+static void flush_all_devices_for_iommu(struct amd_iommu *iommu)
+{
+	int i;
+
+	for (i = 0; i <= amd_iommu_last_bdf; ++i) {
+		if (iommu != amd_iommu_rlookup_table[i])
+			continue;
+
+		iommu_queue_inv_dev_entry(iommu, i);
+		iommu_completion_wait(iommu);
+	}
+}
+
 void amd_iommu_flush_all_devices(void)
 {
 	struct amd_iommu *iommu;
-- 
cgit v1.2.3-59-g8ed1b


From 93f1cc67cf3196174412adca87321b25c1c986b0 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 14:50:20 +0200
Subject: x86/amd-iommu: Add reset function for command buffers

This patch factors parts of the command buffer
initialization code into a seperate function which can be
used to reset the command buffer later.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  3 +++
 arch/x86/kernel/amd_iommu_init.c       | 20 +++++++++++++++-----
 2 files changed, 18 insertions(+), 5 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 0c878caaa0a2..c54bc979dc17 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -457,4 +457,7 @@ static inline void amd_iommu_stats_init(void) { }
 
 #endif /* CONFIG_AMD_IOMMU_STATS */
 
+/* some function prototypes */
+extern void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu);
+
 #endif /* _ASM_X86_AMD_IOMMU_TYPES_H */
diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index c1b17e97252e..1752afef948a 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -434,6 +434,20 @@ static u8 * __init alloc_command_buffer(struct amd_iommu *iommu)
 	return cmd_buf;
 }
 
+/*
+ * This function resets the command buffer if the IOMMU stopped fetching
+ * commands from it.
+ */
+void amd_iommu_reset_cmd_buffer(struct amd_iommu *iommu)
+{
+	iommu_feature_disable(iommu, CONTROL_CMDBUF_EN);
+
+	writel(0x00, iommu->mmio_base + MMIO_CMD_HEAD_OFFSET);
+	writel(0x00, iommu->mmio_base + MMIO_CMD_TAIL_OFFSET);
+
+	iommu_feature_enable(iommu, CONTROL_CMDBUF_EN);
+}
+
 /*
  * This function writes the command buffer address to the hardware and
  * enables it.
@@ -450,11 +464,7 @@ static void iommu_enable_command_buffer(struct amd_iommu *iommu)
 	memcpy_toio(iommu->mmio_base + MMIO_CMD_BUF_OFFSET,
 		    &entry, sizeof(entry));
 
-	/* set head and tail to zero manually */
-	writel(0x00, iommu->mmio_base + MMIO_CMD_HEAD_OFFSET);
-	writel(0x00, iommu->mmio_base + MMIO_CMD_TAIL_OFFSET);
-
-	iommu_feature_enable(iommu, CONTROL_CMDBUF_EN);
+	amd_iommu_reset_cmd_buffer(iommu);
 }
 
 static void __init free_command_buffer(struct amd_iommu *iommu)
-- 
cgit v1.2.3-59-g8ed1b


From a345b23b79f1900e7d87c3165182504419180de4 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 15:01:43 +0200
Subject: x86/amd-iommu: Reset command buffer on ILLEGAL_COMMAND_ERROR

On an ILLEGAL_COMMAND_ERROR the IOMMU stops executing
further commands. This patch changes the code to handle this
case better by resetting the command buffer in the IOMMU.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 2dc093370d2d..2333d615f5ee 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -61,6 +61,7 @@ static u64* alloc_pte(struct protection_domain *dom,
 static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 				      unsigned long start_page,
 				      unsigned int pages);
+static void reset_iommu_command_buffer(struct amd_iommu *iommu);
 
 #ifndef BUS_NOTIFY_UNBOUND_DRIVER
 #define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
@@ -156,7 +157,7 @@ static void dump_command(unsigned long phys_addr)
 		pr_err("AMD-Vi: CMD[%d]: %08x\n", i, cmd->data[i]);
 }
 
-static void iommu_print_event(void *__evt)
+static void iommu_print_event(struct amd_iommu *iommu, void *__evt)
 {
 	u32 *event = __evt;
 	int type  = (event[1] >> EVENT_TYPE_SHIFT)  & EVENT_TYPE_MASK;
@@ -195,6 +196,7 @@ static void iommu_print_event(void *__evt)
 		break;
 	case EVENT_TYPE_ILL_CMD:
 		printk("ILLEGAL_COMMAND_ERROR address=0x%016llx]\n", address);
+		reset_iommu_command_buffer(iommu);
 		dump_command(address);
 		break;
 	case EVENT_TYPE_CMD_HARD_ERR:
@@ -229,7 +231,7 @@ static void iommu_poll_events(struct amd_iommu *iommu)
 	tail = readl(iommu->mmio_base + MMIO_EVT_TAIL_OFFSET);
 
 	while (head != tail) {
-		iommu_print_event(iommu->evt_buf + head);
+		iommu_print_event(iommu, iommu->evt_buf + head);
 		head = (head + EVENT_ENTRY_SIZE) % iommu->evt_buf_size;
 	}
 
@@ -547,6 +549,15 @@ void amd_iommu_flush_all_devices(void)
 	}
 }
 
+static void reset_iommu_command_buffer(struct amd_iommu *iommu)
+{
+	pr_err("AMD-Vi: Resetting IOMMU command buffer\n");
+
+	amd_iommu_reset_cmd_buffer(iommu);
+	flush_all_devices_for_iommu(iommu);
+	flush_all_domains_on_iommu(iommu);
+}
+
 /****************************************************************************
  *
  * The functions below are used the create the page table mappings for
-- 
cgit v1.2.3-59-g8ed1b


From b26e81b871bd18184968f0bb3f12945906eadfce Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 15:08:09 +0200
Subject: x86/amd-iommu: Panic if IOMMU command buffer reset fails

To prevent the driver from doing recursive command buffer
resets, just panic when that recursion happens.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h | 3 +++
 arch/x86/kernel/amd_iommu.c            | 7 +++++++
 2 files changed, 10 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index c54bc979dc17..e676746a4a30 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -337,6 +337,9 @@ 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;
 };
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 2333d615f5ee..b62a2f64dfc5 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -553,9 +553,16 @@ static void reset_iommu_command_buffer(struct amd_iommu *iommu)
 {
 	pr_err("AMD-Vi: Resetting IOMMU command buffer\n");
 
+	if (iommu->reset_in_progress)
+		panic("AMD-Vi: ILLEGAL_COMMAND_ERROR while resetting command buffer\n");
+
+	iommu->reset_in_progress = true;
+
 	amd_iommu_reset_cmd_buffer(iommu);
 	flush_all_devices_for_iommu(iommu);
 	flush_all_domains_on_iommu(iommu);
+
+	iommu->reset_in_progress = false;
 }
 
 /****************************************************************************
-- 
cgit v1.2.3-59-g8ed1b


From 6a1eddd2f951656a6abbd42e2cddc2267c4a639d Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 15:15:10 +0200
Subject: x86/amd-iommu: Reset command buffer if wait loop fails

Instead of a panic on an comletion wait loop failure, try to
recover from that event from resetting the command buffer.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index b62a2f64dfc5..cfca80bfe75f 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -318,8 +318,11 @@ static void __iommu_wait_for_completion(struct amd_iommu *iommu)
 	status &= ~MMIO_STATUS_COM_WAIT_INT_MASK;
 	writel(status, iommu->mmio_base + MMIO_STATUS_OFFSET);
 
-	if (unlikely(i == EXIT_LOOP_COUNT))
-		panic("AMD IOMMU: Completion wait loop failed\n");
+	if (unlikely(i == EXIT_LOOP_COUNT)) {
+		spin_unlock(&iommu->lock);
+		reset_iommu_command_buffer(iommu);
+		spin_lock(&iommu->lock);
+	}
 }
 
 /*
-- 
cgit v1.2.3-59-g8ed1b


From 9355a08186e52b7c120adea91c984923b54efa10 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 14:24:08 +0200
Subject: x86/amd-iommu: Make fetch_pte aware of dynamic mapping levels

This patch changes the fetch_pte function in the AMD IOMMU
driver to support dynamic mapping levels.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  9 +++++++++
 arch/x86/kernel/amd_iommu.c            | 24 +++++++++++++-----------
 2 files changed, 22 insertions(+), 11 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 0c878caaa0a2..7fce4ef77bdd 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -146,12 +146,21 @@
 #define PAGE_MODE_1_LEVEL 0x01
 #define PAGE_MODE_2_LEVEL 0x02
 #define PAGE_MODE_3_LEVEL 0x03
+#define PAGE_MODE_4_LEVEL 0x04
+#define PAGE_MODE_5_LEVEL 0x05
+#define PAGE_MODE_6_LEVEL 0x06
 
 #define IOMMU_PDE_NL_0   0x000ULL
 #define IOMMU_PDE_NL_1   0x200ULL
 #define IOMMU_PDE_NL_2   0x400ULL
 #define IOMMU_PDE_NL_3   0x600ULL
 
+#define PM_LEVEL_SHIFT(x)	(12 + ((x) * 9))
+#define PM_LEVEL_SIZE(x)	(((x) < 6) ? \
+				  ((1ULL << PM_LEVEL_SHIFT((x))) - 1): \
+				   (0xffffffffffffffffULL))
+#define PM_LEVEL_INDEX(x, a)	(((a) >> PM_LEVEL_SHIFT((x))) & 0x1ffULL)
+
 #define IOMMU_PTE_L2_INDEX(address) (((address) >> 30) & 0x1ffULL)
 #define IOMMU_PTE_L1_INDEX(address) (((address) >> 21) & 0x1ffULL)
 #define IOMMU_PTE_L0_INDEX(address) (((address) >> 12) & 0x1ffULL)
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 6c99f5037801..29bcd358b6ce 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -61,6 +61,8 @@ static u64* alloc_pte(struct protection_domain *dom,
 static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 				      unsigned long start_page,
 				      unsigned int pages);
+static u64 *fetch_pte(struct protection_domain *domain,
+		      unsigned long address);
 
 #ifndef BUS_NOTIFY_UNBOUND_DRIVER
 #define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
@@ -670,24 +672,24 @@ static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom,
  * This function checks if there is a PTE for a given dma address. If
  * there is one, it returns the pointer to it.
  */
-static u64* fetch_pte(struct protection_domain *domain,
+static u64 *fetch_pte(struct protection_domain *domain,
 		      unsigned long address)
 {
+	int level;
 	u64 *pte;
 
-	pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(address)];
+	level =  domain->mode - 1;
+	pte   = &domain->pt_root[PM_LEVEL_INDEX(level, address)];
 
-	if (!IOMMU_PTE_PRESENT(*pte))
-		return NULL;
+	while (level > 0) {
+		if (!IOMMU_PTE_PRESENT(*pte))
+			return NULL;
 
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L1_INDEX(address)];
+		level -= 1;
 
-	if (!IOMMU_PTE_PRESENT(*pte))
-		return NULL;
-
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L0_INDEX(address)];
+		pte = IOMMU_PTE_PAGE(*pte);
+		pte = &pte[PM_LEVEL_INDEX(level, address)];
+	}
 
 	return pte;
 }
-- 
cgit v1.2.3-59-g8ed1b


From 38a76eeeafb251bf67d143a34b37a8105cff302e Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 17:02:47 +0200
Subject: x86/amd-iommu: Use fetch_pte in iommu_unmap_page

Instead of reimplementing existing logic use fetch_pte to
walk the page table in iommu_unmap_page.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 19 +++----------------
 1 file changed, 3 insertions(+), 16 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 29bcd358b6ce..5e5279899994 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -546,23 +546,10 @@ static int iommu_map_page(struct protection_domain *dom,
 static void iommu_unmap_page(struct protection_domain *dom,
 			     unsigned long bus_addr)
 {
-	u64 *pte;
-
-	pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(bus_addr)];
-
-	if (!IOMMU_PTE_PRESENT(*pte))
-		return;
-
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
-
-	if (!IOMMU_PTE_PRESENT(*pte))
-		return;
-
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L1_INDEX(bus_addr)];
+	u64 *pte = fetch_pte(dom, bus_addr);
 
-	*pte = 0;
+	if (pte)
+		*pte = 0;
 }
 
 /*
-- 
cgit v1.2.3-59-g8ed1b


From a6d41a4027b758a9473197a78fab45afb31003aa Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 17:08:55 +0200
Subject: x86/amd-iommu: Use fetch_pte in amd_iommu_iova_to_phys

Don't reimplement the page table walker in this function.
Use the generic one.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 16 ++--------------
 1 file changed, 2 insertions(+), 14 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 5e5279899994..4a54366d422d 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -2140,21 +2140,9 @@ static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
 	phys_addr_t paddr;
 	u64 *pte;
 
-	pte = &domain->pt_root[IOMMU_PTE_L2_INDEX(iova)];
+	pte = fetch_pte(domain, iova);
 
-	if (!IOMMU_PTE_PRESENT(*pte))
-		return 0;
-
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L1_INDEX(iova)];
-
-	if (!IOMMU_PTE_PRESENT(*pte))
-		return 0;
-
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L0_INDEX(iova)];
-
-	if (!IOMMU_PTE_PRESENT(*pte))
+	if (!pte || !IOMMU_PTE_PRESENT(*pte))
 		return 0;
 
 	paddr  = *pte & IOMMU_PAGE_MASK;
-- 
cgit v1.2.3-59-g8ed1b


From 6a0dbcbe4e612fbc9d73cd4dde8ebef19295058a Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 15:41:59 +0200
Subject: x86/amd-iommu: Add a gneric version of amd_iommu_flush_all_devices

This patch adds a generic variant of
amd_iommu_flush_all_devices function which flushes only the
DTEs for a given protection domain.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 4a54366d422d..5265dd17eb5c 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -481,13 +481,14 @@ void amd_iommu_flush_all_domains(void)
 	}
 }
 
-void amd_iommu_flush_all_devices(void)
+static void flush_devices_by_domain(struct protection_domain *domain)
 {
 	struct amd_iommu *iommu;
 	int i;
 
 	for (i = 0; i <= amd_iommu_last_bdf; ++i) {
-		if (amd_iommu_pd_table[i] == NULL)
+		if ((domain == NULL && amd_iommu_pd_table[i] == NULL) ||
+		    (amd_iommu_pd_table[i] != domain))
 			continue;
 
 		iommu = amd_iommu_rlookup_table[i];
@@ -499,6 +500,11 @@ void amd_iommu_flush_all_devices(void)
 	}
 }
 
+void amd_iommu_flush_all_devices(void)
+{
+	flush_devices_by_domain(NULL);
+}
+
 /****************************************************************************
  *
  * The functions below are used the create the page table mappings for
-- 
cgit v1.2.3-59-g8ed1b


From 407d733e30a97daf5ea6f9eb5f9ebbd42a0a9ef2 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 16:07:00 +0200
Subject: x86/amd-iommu: Introduce set_dte_entry function

This function factors out some logic of attach_device to a
seperate function. This new function will be used to update
device table entries when necessary.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 27 +++++++++++++++++----------
 1 file changed, 17 insertions(+), 10 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 5265dd17eb5c..0fab1f1d135e 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1058,18 +1058,10 @@ static struct protection_domain *domain_for_device(u16 devid)
 	return dom;
 }
 
-/*
- * If a device is not yet associated with a domain, this function does
- * assigns it visible for the hardware
- */
-static void attach_device(struct amd_iommu *iommu,
-			  struct protection_domain *domain,
-			  u16 devid)
+static void set_dte_entry(u16 devid, struct protection_domain *domain)
 {
-	unsigned long flags;
 	u64 pte_root = virt_to_phys(domain->pt_root);
-
-	domain->dev_cnt += 1;
+	unsigned long flags;
 
 	pte_root |= (domain->mode & DEV_ENTRY_MODE_MASK)
 		    << DEV_ENTRY_MODE_SHIFT;
@@ -1082,6 +1074,21 @@ static void attach_device(struct amd_iommu *iommu,
 
 	amd_iommu_pd_table[devid] = domain;
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+}
+
+/*
+ * If a device is not yet associated with a domain, this function does
+ * assigns it visible for the hardware
+ */
+static void attach_device(struct amd_iommu *iommu,
+			  struct protection_domain *domain,
+			  u16 devid)
+{
+	/* set the DTE entry */
+	set_dte_entry(devid, domain);
+
+	/* increase reference counter */
+	domain->dev_cnt += 1;
 
        /*
         * We might boot into a crash-kernel here. The crashed kernel
-- 
cgit v1.2.3-59-g8ed1b


From 04bfdd8406099fca2e6b8844748c4d6c5eba8c8d Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 16:00:23 +0200
Subject: x86/amd-iommu: Flush domains if address space size was increased

Thist patch introduces the update_domain function which
propagates the larger address space of a protection domain
to the device table and flushes all relevant DTEs and the
domain TLB.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  1 +
 arch/x86/kernel/amd_iommu.c            | 32 ++++++++++++++++++++++++++++++++
 2 files changed, 33 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 7fce4ef77bdd..97f3d09d3be5 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -235,6 +235,7 @@ struct protection_domain {
 	int mode;		/* paging mode (0-6 levels) */
 	u64 *pt_root;		/* page table root pointer */
 	unsigned long flags;	/* flags to find out type of domain */
+	bool updated;		/* complete domain flush required */
 	unsigned dev_cnt;	/* devices assigned to this domain */
 	void *priv;		/* private data */
 };
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 0fab1f1d135e..5eab6a84b9cc 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -63,6 +63,7 @@ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 				      unsigned int pages);
 static u64 *fetch_pte(struct protection_domain *domain,
 		      unsigned long address);
+static void update_domain(struct protection_domain *domain);
 
 #ifndef BUS_NOTIFY_UNBOUND_DRIVER
 #define BUS_NOTIFY_UNBOUND_DRIVER 0x0005
@@ -546,6 +547,8 @@ static int iommu_map_page(struct protection_domain *dom,
 
 	*pte = __pte;
 
+	update_domain(dom);
+
 	return 0;
 }
 
@@ -762,9 +765,13 @@ static int alloc_new_range(struct amd_iommu *iommu,
 		dma_ops_reserve_addresses(dma_dom, i << PAGE_SHIFT, 1);
 	}
 
+	update_domain(&dma_dom->domain);
+
 	return 0;
 
 out_free:
+	update_domain(&dma_dom->domain);
+
 	free_page((unsigned long)dma_dom->aperture[index]->bitmap);
 
 	kfree(dma_dom->aperture[index]);
@@ -1294,6 +1301,29 @@ static int get_device_resources(struct device *dev,
 	return 1;
 }
 
+static void update_device_table(struct protection_domain *domain)
+{
+	int i;
+
+	for (i = 0; i <= amd_iommu_last_bdf; ++i) {
+		if (amd_iommu_pd_table[i] != domain)
+			continue;
+		set_dte_entry(i, domain);
+	}
+}
+
+static void update_domain(struct protection_domain *domain)
+{
+	if (!domain->updated)
+		return;
+
+	update_device_table(domain);
+	flush_devices_by_domain(domain);
+	iommu_flush_domain(domain->id);
+
+	domain->updated = false;
+}
+
 /*
  * If the pte_page is not yet allocated this function is called
  */
@@ -1351,6 +1381,8 @@ static u64* dma_ops_get_pte(struct dma_ops_domain *dom,
 	} else
 		pte += IOMMU_PTE_L0_INDEX(address);
 
+	update_domain(&dom->domain);
+
 	return pte;
 }
 
-- 
cgit v1.2.3-59-g8ed1b


From 50020fb6324465e478d6c8cdbf3c695f0a60358d Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 15:38:40 +0200
Subject: x86/amd-iommu: Introduce increase_address_space function

This function will be used to increase the address space
size of a protection domain.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  4 ++++
 arch/x86/kernel/amd_iommu.c            | 27 +++++++++++++++++++++++++++
 2 files changed, 31 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 97f3d09d3be5..1b4b3d6c9f04 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -160,6 +160,10 @@
 				  ((1ULL << PM_LEVEL_SHIFT((x))) - 1): \
 				   (0xffffffffffffffffULL))
 #define PM_LEVEL_INDEX(x, a)	(((a) >> PM_LEVEL_SHIFT((x))) & 0x1ffULL)
+#define PM_LEVEL_ENC(x)		(((x) << 9) & 0xe00ULL)
+#define PM_LEVEL_PDE(x, a)	((a) | PM_LEVEL_ENC((x)) | \
+				 IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW)
+
 
 #define IOMMU_PTE_L2_INDEX(address) (((address) >> 30) & 0x1ffULL)
 #define IOMMU_PTE_L1_INDEX(address) (((address) >> 21) & 0x1ffULL)
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 5eab6a84b9cc..fc97b51f0287 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1324,6 +1324,33 @@ static void update_domain(struct protection_domain *domain)
 	domain->updated = false;
 }
 
+/*
+ * This function is used to add another level to an IO page table. Adding
+ * another level increases the size of the address space by 9 bits to a size up
+ * to 64 bits.
+ */
+static bool increase_address_space(struct protection_domain *domain,
+				   gfp_t gfp)
+{
+	u64 *pte;
+
+	if (domain->mode == PAGE_MODE_6_LEVEL)
+		/* address space already 64 bit large */
+		return false;
+
+	pte = (void *)get_zeroed_page(gfp);
+	if (!pte)
+		return false;
+
+	*pte             = PM_LEVEL_PDE(domain->mode,
+					virt_to_phys(domain->pt_root));
+	domain->pt_root  = pte;
+	domain->mode    += 1;
+	domain->updated  = true;
+
+	return true;
+}
+
 /*
  * If the pte_page is not yet allocated this function is called
  */
-- 
cgit v1.2.3-59-g8ed1b


From 8bc3e127421bf3b735edbde05135892c12c5f615 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 16:48:40 +0200
Subject: x86/amd-iommu: Change alloc_pte to support 64 bit address space

This patch changes the alloc_pte function to be able to map
pages into the whole 64 bit address space supported by AMD
IOMMU hardware from the old limit of 2**39 bytes.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 44 ++++++++++++++++++++------------------------
 1 file changed, 20 insertions(+), 24 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index fc97b51f0287..3be2b61fc31f 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -55,7 +55,7 @@ struct iommu_cmd {
 static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 			     struct unity_map_entry *e);
 static struct dma_ops_domain *find_protection_domain(u16 devid);
-static u64* alloc_pte(struct protection_domain *dom,
+static u64 *alloc_pte(struct protection_domain *domain,
 		      unsigned long address, u64
 		      **pte_page, gfp_t gfp);
 static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
@@ -1351,39 +1351,35 @@ static bool increase_address_space(struct protection_domain *domain,
 	return true;
 }
 
-/*
- * If the pte_page is not yet allocated this function is called
- */
-static u64* alloc_pte(struct protection_domain *dom,
+static u64 *alloc_pte(struct protection_domain *domain,
 		      unsigned long address, u64 **pte_page, gfp_t gfp)
 {
 	u64 *pte, *page;
+	int level;
 
-	pte = &dom->pt_root[IOMMU_PTE_L2_INDEX(address)];
+	while (address > PM_LEVEL_SIZE(domain->mode))
+		increase_address_space(domain, gfp);
 
-	if (!IOMMU_PTE_PRESENT(*pte)) {
-		page = (u64 *)get_zeroed_page(gfp);
-		if (!page)
-			return NULL;
-		*pte = IOMMU_L2_PDE(virt_to_phys(page));
-	}
+	level =  domain->mode - 1;
+	pte   = &domain->pt_root[PM_LEVEL_INDEX(level, address)];
 
-	pte = IOMMU_PTE_PAGE(*pte);
-	pte = &pte[IOMMU_PTE_L1_INDEX(address)];
+	while (level > 0) {
+		if (!IOMMU_PTE_PRESENT(*pte)) {
+			page = (u64 *)get_zeroed_page(gfp);
+			if (!page)
+				return NULL;
+			*pte = PM_LEVEL_PDE(level, virt_to_phys(page));
+		}
 
-	if (!IOMMU_PTE_PRESENT(*pte)) {
-		page = (u64 *)get_zeroed_page(gfp);
-		if (!page)
-			return NULL;
-		*pte = IOMMU_L1_PDE(virt_to_phys(page));
-	}
+		level -= 1;
 
-	pte = IOMMU_PTE_PAGE(*pte);
+		pte = IOMMU_PTE_PAGE(*pte);
 
-	if (pte_page)
-		*pte_page = pte;
+		if (pte_page && level == 0)
+			*pte_page = pte;
 
-	pte = &pte[IOMMU_PTE_L0_INDEX(address)];
+		pte = &pte[PM_LEVEL_INDEX(level, address)];
+	}
 
 	return pte;
 }
-- 
cgit v1.2.3-59-g8ed1b


From 8c8c143cdc95ebe50fd962917556e25e8912997b Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 17:30:00 +0200
Subject: x86/amd-iommu: Remove last usages of IOMMU_PTE_L0_INDEX

This change allows to remove these old macros later.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 3be2b61fc31f..ebc1c8443926 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1402,7 +1402,7 @@ static u64* dma_ops_get_pte(struct dma_ops_domain *dom,
 		pte = alloc_pte(&dom->domain, address, &pte_page, GFP_ATOMIC);
 		aperture->pte_pages[APERTURE_PAGE_INDEX(address)] = pte_page;
 	} else
-		pte += IOMMU_PTE_L0_INDEX(address);
+		pte += PM_LEVEL_INDEX(0, address);
 
 	update_domain(&dom->domain);
 
@@ -1466,7 +1466,7 @@ static void dma_ops_domain_unmap(struct amd_iommu *iommu,
 	if (!pte)
 		return;
 
-	pte += IOMMU_PTE_L0_INDEX(address);
+	pte += PM_LEVEL_INDEX(0, address);
 
 	WARN_ON(!*pte);
 
-- 
cgit v1.2.3-59-g8ed1b


From bad1cac28a707c69301a4d0612da9ccbecd51953 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 16:52:23 +0200
Subject: x86/amd-iommu: Remove bus_addr check in iommu_map_page

The driver now supports full 64 bit device address spaces.
So this check is not longer required.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index ebc1c8443926..6ffb3e637657 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -530,8 +530,7 @@ static int iommu_map_page(struct protection_domain *dom,
 	bus_addr  = PAGE_ALIGN(bus_addr);
 	phys_addr = PAGE_ALIGN(phys_addr);
 
-	/* only support 512GB address spaces for now */
-	if (bus_addr > IOMMU_MAP_SIZE_L3 || !(prot & IOMMU_PROT_MASK))
+	if (!(prot & IOMMU_PROT_MASK))
 		return -EINVAL;
 
 	pte = alloc_pte(dom, bus_addr, NULL, GFP_KERNEL);
-- 
cgit v1.2.3-59-g8ed1b


From 8f7a017ce05ed4522809448e169daa44fe6edeb1 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 2 Sep 2009 16:55:24 +0200
Subject: x86/amd-iommu: Use 2-level page tables for dma_ops domains

The driver now supports a dynamic number of levels for IO
page tables. This allows to reduce the number of levels for
dma_ops domains by one because a dma_ops domain has usually
an address space size between 128MB and 4G.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 6ffb3e637657..addf6588c366 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1010,7 +1010,7 @@ static struct dma_ops_domain *dma_ops_domain_alloc(struct amd_iommu *iommu)
 	dma_dom->domain.id = domain_id_alloc();
 	if (dma_dom->domain.id == 0)
 		goto free_dma_dom;
-	dma_dom->domain.mode = PAGE_MODE_3_LEVEL;
+	dma_dom->domain.mode = PAGE_MODE_2_LEVEL;
 	dma_dom->domain.pt_root = (void *)get_zeroed_page(GFP_KERNEL);
 	dma_dom->domain.flags = PD_DMA_OPS_MASK;
 	dma_dom->domain.priv = dma_dom;
-- 
cgit v1.2.3-59-g8ed1b


From a6b256b41357c33ccb2d105a4457e22bdc83e021 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 12:21:31 +0200
Subject: x86/amd-iommu: Support higher level PTEs in iommu_page_unmap

This patch changes fetch_pte and iommu_page_unmap to support
different page sizes too.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  1 +
 arch/x86/kernel/amd_iommu.c            | 21 +++++++++++++--------
 2 files changed, 14 insertions(+), 8 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index d66430de5f7c..351ca39ece05 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -158,6 +158,7 @@
 #define PM_LEVEL_ENC(x)		(((x) << 9) & 0xe00ULL)
 #define PM_LEVEL_PDE(x, a)	((a) | PM_LEVEL_ENC((x)) | \
 				 IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW)
+#define PM_PTE_LEVEL(pte)	(((pte) >> 9) & 0x7ULL)
 
 #define IOMMU_PTE_P  (1ULL << 0)
 #define IOMMU_PTE_TV (1ULL << 1)
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index addf6588c366..002cf9cab9e9 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -62,7 +62,7 @@ static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 				      unsigned long start_page,
 				      unsigned int pages);
 static u64 *fetch_pte(struct protection_domain *domain,
-		      unsigned long address);
+		      unsigned long address, int map_size);
 static void update_domain(struct protection_domain *domain);
 
 #ifndef BUS_NOTIFY_UNBOUND_DRIVER
@@ -552,9 +552,9 @@ static int iommu_map_page(struct protection_domain *dom,
 }
 
 static void iommu_unmap_page(struct protection_domain *dom,
-			     unsigned long bus_addr)
+			     unsigned long bus_addr, int map_size)
 {
-	u64 *pte = fetch_pte(dom, bus_addr);
+	u64 *pte = fetch_pte(dom, bus_addr, map_size);
 
 	if (pte)
 		*pte = 0;
@@ -668,7 +668,7 @@ static int init_unity_mappings_for_device(struct dma_ops_domain *dma_dom,
  * there is one, it returns the pointer to it.
  */
 static u64 *fetch_pte(struct protection_domain *domain,
-		      unsigned long address)
+		      unsigned long address, int map_size)
 {
 	int level;
 	u64 *pte;
@@ -676,7 +676,7 @@ static u64 *fetch_pte(struct protection_domain *domain,
 	level =  domain->mode - 1;
 	pte   = &domain->pt_root[PM_LEVEL_INDEX(level, address)];
 
-	while (level > 0) {
+	while (level > map_size) {
 		if (!IOMMU_PTE_PRESENT(*pte))
 			return NULL;
 
@@ -684,6 +684,11 @@ static u64 *fetch_pte(struct protection_domain *domain,
 
 		pte = IOMMU_PTE_PAGE(*pte);
 		pte = &pte[PM_LEVEL_INDEX(level, address)];
+
+		if ((PM_PTE_LEVEL(*pte) == 0) && level != map_size) {
+			pte = NULL;
+			break;
+		}
 	}
 
 	return pte;
@@ -757,7 +762,7 @@ static int alloc_new_range(struct amd_iommu *iommu,
 	for (i = dma_dom->aperture[index]->offset;
 	     i < dma_dom->aperture_size;
 	     i += PAGE_SIZE) {
-		u64 *pte = fetch_pte(&dma_dom->domain, i);
+		u64 *pte = fetch_pte(&dma_dom->domain, i, PM_MAP_4k);
 		if (!pte || !IOMMU_PTE_PRESENT(*pte))
 			continue;
 
@@ -2192,7 +2197,7 @@ static void amd_iommu_unmap_range(struct iommu_domain *dom,
 	iova  &= PAGE_MASK;
 
 	for (i = 0; i < npages; ++i) {
-		iommu_unmap_page(domain, iova);
+		iommu_unmap_page(domain, iova, PM_MAP_4k);
 		iova  += PAGE_SIZE;
 	}
 
@@ -2207,7 +2212,7 @@ static phys_addr_t amd_iommu_iova_to_phys(struct iommu_domain *dom,
 	phys_addr_t paddr;
 	u64 *pte;
 
-	pte = fetch_pte(domain, iova);
+	pte = fetch_pte(domain, iova, PM_MAP_4k);
 
 	if (!pte || !IOMMU_PTE_PRESENT(*pte))
 		return 0;
-- 
cgit v1.2.3-59-g8ed1b


From abdc5eb3d69279039ba6cb89719913d08013ab14 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Thu, 3 Sep 2009 11:33:51 +0200
Subject: x86/amd-iommu: Change iommu_map_page to support multiple page sizes

This patch adds a map_size parameter to the iommu_map_page
function which makes it generic enough to handle multiple
page sizes. This also requires a change to alloc_pte which
is also done in this patch.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |  6 ++++++
 arch/x86/kernel/amd_iommu.c            | 31 ++++++++++++++++++++-----------
 2 files changed, 26 insertions(+), 11 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 351ca39ece05..9e8bb9746dcf 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -160,6 +160,12 @@
 				 IOMMU_PTE_P | IOMMU_PTE_IR | IOMMU_PTE_IW)
 #define PM_PTE_LEVEL(pte)	(((pte) >> 9) & 0x7ULL)
 
+#define PM_MAP_4k		0
+#define PM_ADDR_MASK		0x000ffffffffff000ULL
+#define PM_MAP_MASK(lvl)	(PM_ADDR_MASK & \
+				(~((1ULL << (12 + ((lvl) * 9))) - 1)))
+#define PM_ALIGNED(lvl, addr)	((PM_MAP_MASK(lvl) & (addr)) == (addr))
+
 #define IOMMU_PTE_P  (1ULL << 0)
 #define IOMMU_PTE_TV (1ULL << 1)
 #define IOMMU_PTE_U  (1ULL << 59)
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 002cf9cab9e9..45be9499c973 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -56,8 +56,8 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 			     struct unity_map_entry *e);
 static struct dma_ops_domain *find_protection_domain(u16 devid);
 static u64 *alloc_pte(struct protection_domain *domain,
-		      unsigned long address, u64
-		      **pte_page, gfp_t gfp);
+		      unsigned long address, int end_lvl,
+		      u64 **pte_page, gfp_t gfp);
 static void dma_ops_reserve_addresses(struct dma_ops_domain *dom,
 				      unsigned long start_page,
 				      unsigned int pages);
@@ -523,17 +523,21 @@ void amd_iommu_flush_all_devices(void)
 static int iommu_map_page(struct protection_domain *dom,
 			  unsigned long bus_addr,
 			  unsigned long phys_addr,
-			  int prot)
+			  int prot,
+			  int map_size)
 {
 	u64 __pte, *pte;
 
 	bus_addr  = PAGE_ALIGN(bus_addr);
 	phys_addr = PAGE_ALIGN(phys_addr);
 
+	BUG_ON(!PM_ALIGNED(map_size, bus_addr));
+	BUG_ON(!PM_ALIGNED(map_size, phys_addr));
+
 	if (!(prot & IOMMU_PROT_MASK))
 		return -EINVAL;
 
-	pte = alloc_pte(dom, bus_addr, NULL, GFP_KERNEL);
+	pte = alloc_pte(dom, bus_addr, map_size, NULL, GFP_KERNEL);
 
 	if (IOMMU_PTE_PRESENT(*pte))
 		return -EBUSY;
@@ -612,7 +616,8 @@ static int dma_ops_unity_map(struct dma_ops_domain *dma_dom,
 
 	for (addr = e->address_start; addr < e->address_end;
 	     addr += PAGE_SIZE) {
-		ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot);
+		ret = iommu_map_page(&dma_dom->domain, addr, addr, e->prot,
+				     PM_MAP_4k);
 		if (ret)
 			return ret;
 		/*
@@ -729,7 +734,7 @@ static int alloc_new_range(struct amd_iommu *iommu,
 		u64 *pte, *pte_page;
 
 		for (i = 0; i < num_ptes; ++i) {
-			pte = alloc_pte(&dma_dom->domain, address,
+			pte = alloc_pte(&dma_dom->domain, address, PM_MAP_4k,
 					&pte_page, gfp);
 			if (!pte)
 				goto out_free;
@@ -1356,7 +1361,10 @@ static bool increase_address_space(struct protection_domain *domain,
 }
 
 static u64 *alloc_pte(struct protection_domain *domain,
-		      unsigned long address, u64 **pte_page, gfp_t gfp)
+		      unsigned long address,
+		      int end_lvl,
+		      u64 **pte_page,
+		      gfp_t gfp)
 {
 	u64 *pte, *page;
 	int level;
@@ -1367,7 +1375,7 @@ static u64 *alloc_pte(struct protection_domain *domain,
 	level =  domain->mode - 1;
 	pte   = &domain->pt_root[PM_LEVEL_INDEX(level, address)];
 
-	while (level > 0) {
+	while (level > end_lvl) {
 		if (!IOMMU_PTE_PRESENT(*pte)) {
 			page = (u64 *)get_zeroed_page(gfp);
 			if (!page)
@@ -1379,7 +1387,7 @@ static u64 *alloc_pte(struct protection_domain *domain,
 
 		pte = IOMMU_PTE_PAGE(*pte);
 
-		if (pte_page && level == 0)
+		if (pte_page && level == end_lvl)
 			*pte_page = pte;
 
 		pte = &pte[PM_LEVEL_INDEX(level, address)];
@@ -1403,7 +1411,8 @@ static u64* dma_ops_get_pte(struct dma_ops_domain *dom,
 
 	pte = aperture->pte_pages[APERTURE_PAGE_INDEX(address)];
 	if (!pte) {
-		pte = alloc_pte(&dom->domain, address, &pte_page, GFP_ATOMIC);
+		pte = alloc_pte(&dom->domain, address, PM_MAP_4k, &pte_page,
+				GFP_ATOMIC);
 		aperture->pte_pages[APERTURE_PAGE_INDEX(address)] = pte_page;
 	} else
 		pte += PM_LEVEL_INDEX(0, address);
@@ -2176,7 +2185,7 @@ static int amd_iommu_map_range(struct iommu_domain *dom,
 	paddr &= PAGE_MASK;
 
 	for (i = 0; i < npages; ++i) {
-		ret = iommu_map_page(domain, iova, paddr, prot);
+		ret = iommu_map_page(domain, iova, paddr, prot, PM_MAP_4k);
 		if (ret)
 			return ret;
 
-- 
cgit v1.2.3-59-g8ed1b


From ac0101d396fee24994632f71b55b9f7f9ee16eff Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 16:00:35 +0200
Subject: x86/dma: Mark iommu_pass_through as __read_mostly

This variable is read most of the time. This patch marks it
as such. It also documents the meaning the this variable
while at it.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/pci-dma.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/pci-dma.c b/arch/x86/kernel/pci-dma.c
index 1a041bcf506b..873aa079d166 100644
--- a/arch/x86/kernel/pci-dma.c
+++ b/arch/x86/kernel/pci-dma.c
@@ -32,7 +32,14 @@ int no_iommu __read_mostly;
 /* Set this to 1 if there is a HW IOMMU in the system */
 int iommu_detected __read_mostly = 0;
 
-int iommu_pass_through;
+/*
+ * This variable becomes 1 if iommu=pt is passed on the kernel command line.
+ * If this variable is 1, IOMMU implementations do no DMA ranslation for
+ * devices and allow every device to access to whole physical memory. This is
+ * useful if a user want to use an IOMMU only for KVM device assignment to
+ * guests and not for driver dma translation.
+ */
+int iommu_pass_through __read_mostly;
 
 dma_addr_t bad_dma_address __read_mostly = 0;
 EXPORT_SYMBOL(bad_dma_address);
-- 
cgit v1.2.3-59-g8ed1b


From 2650815fb03fe2bf1e6701584087ba669dcf92cd Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 26 Aug 2009 16:52:40 +0200
Subject: x86/amd-iommu: Add core functions for pd allocation/freeing

This patch factors some code of protection domain allocation
into seperate functions. This way the logic can be used to
allocate the passthrough domain later. As a side effect this
patch fixes an unlikely domain id leakage bug.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 36 ++++++++++++++++++++++++++++++++----
 1 file changed, 32 insertions(+), 4 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 6c99f5037801..0934348abfad 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1988,19 +1988,47 @@ static void cleanup_domain(struct protection_domain *domain)
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
 }
 
-static int amd_iommu_domain_init(struct iommu_domain *dom)
+static void protection_domain_free(struct protection_domain *domain)
+{
+	if (!domain)
+		return;
+
+	if (domain->id)
+		domain_id_free(domain->id);
+
+	kfree(domain);
+}
+
+static struct protection_domain *protection_domain_alloc(void)
 {
 	struct protection_domain *domain;
 
 	domain = kzalloc(sizeof(*domain), GFP_KERNEL);
 	if (!domain)
-		return -ENOMEM;
+		return NULL;
 
 	spin_lock_init(&domain->lock);
-	domain->mode = PAGE_MODE_3_LEVEL;
 	domain->id = domain_id_alloc();
 	if (!domain->id)
+		goto out_err;
+
+	return domain;
+
+out_err:
+	kfree(domain);
+
+	return NULL;
+}
+
+static int amd_iommu_domain_init(struct iommu_domain *dom)
+{
+	struct protection_domain *domain;
+
+	domain = protection_domain_alloc();
+	if (!domain)
 		goto out_free;
+
+	domain->mode    = PAGE_MODE_3_LEVEL;
 	domain->pt_root = (void *)get_zeroed_page(GFP_KERNEL);
 	if (!domain->pt_root)
 		goto out_free;
@@ -2010,7 +2038,7 @@ static int amd_iommu_domain_init(struct iommu_domain *dom)
 	return 0;
 
 out_free:
-	kfree(domain);
+	protection_domain_free(domain);
 
 	return -ENOMEM;
 }
-- 
cgit v1.2.3-59-g8ed1b


From 0feae533ddebe02cda6ccce5cac7349b446776a8 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Wed, 26 Aug 2009 15:26:30 +0200
Subject: x86/amd-iommu: Add passthrough mode initialization functions

When iommu=pt is passed on kernel command line the devices
should run untranslated. This requires the allocation of a
special domain for that purpose. This patch implements the
allocation and initialization path for iommu=pt.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu.h       |  1 +
 arch/x86/include/asm/amd_iommu_types.h |  4 ++
 arch/x86/kernel/amd_iommu.c            | 72 ++++++++++++++++++++++++++++++----
 3 files changed, 69 insertions(+), 8 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/include/asm/amd_iommu.h b/arch/x86/include/asm/amd_iommu.h
index bdf96f119f06..ac95995b7bad 100644
--- a/arch/x86/include/asm/amd_iommu.h
+++ b/arch/x86/include/asm/amd_iommu.h
@@ -25,6 +25,7 @@
 #ifdef CONFIG_AMD_IOMMU
 extern int amd_iommu_init(void);
 extern int amd_iommu_init_dma_ops(void);
+extern int amd_iommu_init_passthrough(void);
 extern void amd_iommu_detect(void);
 extern irqreturn_t amd_iommu_int_handler(int irq, void *data);
 extern void amd_iommu_flush_all_domains(void);
diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 0c878caaa0a2..49f7453bff76 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -143,6 +143,7 @@
 #define EVT_BUFFER_SIZE		8192 /* 512 entries */
 #define EVT_LEN_MASK		(0x9ULL << 56)
 
+#define PAGE_MODE_NONE    0x00
 #define PAGE_MODE_1_LEVEL 0x01
 #define PAGE_MODE_2_LEVEL 0x02
 #define PAGE_MODE_3_LEVEL 0x03
@@ -194,6 +195,9 @@
 #define PD_DMA_OPS_MASK		(1UL << 0) /* domain used for dma_ops */
 #define PD_DEFAULT_MASK		(1UL << 1) /* domain is a default dma_ops
 					      domain for an IOMMU */
+#define PD_PASSTHROUGH_MASK	(1UL << 2) /* domain has no page
+					      translation */
+
 extern bool amd_iommu_dump;
 #define DUMP_printk(format, arg...)					\
 	do {								\
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 0934348abfad..7987f20499ad 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -41,6 +41,12 @@ static DEFINE_RWLOCK(amd_iommu_devtable_lock);
 static LIST_HEAD(iommu_pd_list);
 static DEFINE_SPINLOCK(iommu_pd_list_lock);
 
+/*
+ * Domain for untranslated devices - only allocated
+ * if iommu=pt passed on kernel cmd line.
+ */
+static struct protection_domain *pt_domain;
+
 #ifdef CONFIG_IOMMU_API
 static struct iommu_ops amd_iommu_ops;
 #endif
@@ -1067,9 +1073,9 @@ static struct protection_domain *domain_for_device(u16 devid)
  * If a device is not yet associated with a domain, this function does
  * assigns it visible for the hardware
  */
-static void attach_device(struct amd_iommu *iommu,
-			  struct protection_domain *domain,
-			  u16 devid)
+static void __attach_device(struct amd_iommu *iommu,
+			    struct protection_domain *domain,
+			    u16 devid)
 {
 	unsigned long flags;
 	u64 pte_root = virt_to_phys(domain->pt_root);
@@ -1087,12 +1093,19 @@ static void attach_device(struct amd_iommu *iommu,
 
 	amd_iommu_pd_table[devid] = domain;
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+}
 
-       /*
-        * We might boot into a crash-kernel here. The crashed kernel
-        * left the caches in the IOMMU dirty. So we have to flush
-        * here to evict all dirty stuff.
-        */
+static void attach_device(struct amd_iommu *iommu,
+			  struct protection_domain *domain,
+			  u16 devid)
+{
+	__attach_device(iommu, domain, devid);
+
+	/*
+	 * We might boot into a crash-kernel here. The crashed kernel
+	 * left the caches in the IOMMU dirty. So we have to flush
+	 * here to evict all dirty stuff.
+	 */
 	iommu_queue_inv_dev_entry(iommu, devid);
 	iommu_flush_tlb_pde(iommu, domain->id);
 }
@@ -2219,3 +2232,46 @@ static struct iommu_ops amd_iommu_ops = {
 	.domain_has_cap = amd_iommu_domain_has_cap,
 };
 
+/*****************************************************************************
+ *
+ * The next functions do a basic initialization of IOMMU for pass through
+ * mode
+ *
+ * In passthrough mode the IOMMU is initialized and enabled but not used for
+ * DMA-API translation.
+ *
+ *****************************************************************************/
+
+int __init amd_iommu_init_passthrough(void)
+{
+	struct pci_dev *dev = NULL;
+	u16 devid, devid2;
+
+	/* allocate passthroug domain */
+	pt_domain = protection_domain_alloc();
+	if (!pt_domain)
+		return -ENOMEM;
+
+	pt_domain->mode |= PAGE_MODE_NONE;
+
+	while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+		struct amd_iommu *iommu;
+
+		devid = calc_devid(dev->bus->number, dev->devfn);
+		if (devid > amd_iommu_last_bdf)
+			continue;
+
+		devid2 = amd_iommu_alias_table[devid];
+
+		iommu = amd_iommu_rlookup_table[devid2];
+		if (!iommu)
+			continue;
+
+		__attach_device(iommu, pt_domain, devid);
+		__attach_device(iommu, pt_domain, devid2);
+	}
+
+	pr_info("AMD-Vi: Initialized for Passthrough Mode\n");
+
+	return 0;
+}
-- 
cgit v1.2.3-59-g8ed1b


From aa879fff5d12318259816aa35023e941a1e4d3d9 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Mon, 31 Aug 2009 16:01:48 +0200
Subject: x86/amd-iommu: Fix device table write order

The V bit of the device table entry has to be set after the
rest of the entry is written to not confuse the hardware.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 7987f20499ad..2b1e77c714ff 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1087,9 +1087,9 @@ static void __attach_device(struct amd_iommu *iommu,
 	pte_root |= IOMMU_PTE_IR | IOMMU_PTE_IW | IOMMU_PTE_P | IOMMU_PTE_TV;
 
 	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
-	amd_iommu_dev_table[devid].data[0] = lower_32_bits(pte_root);
-	amd_iommu_dev_table[devid].data[1] = upper_32_bits(pte_root);
 	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);
 
 	amd_iommu_pd_table[devid] = domain;
 	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
-- 
cgit v1.2.3-59-g8ed1b


From eba6ac60ba66c6bf6858938442204feaa67dea31 Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 12:07:08 +0200
Subject: x86/amd-iommu: Align locking between attach_device and detach_device

This patch makes the locking behavior between the functions
attach_device and __attach_device consistent with the
locking behavior between detach_device and __detach_device.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 2b1e77c714ff..9aa135d4453f 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1077,29 +1077,38 @@ static void __attach_device(struct amd_iommu *iommu,
 			    struct protection_domain *domain,
 			    u16 devid)
 {
-	unsigned long flags;
-	u64 pte_root = virt_to_phys(domain->pt_root);
+	u64 pte_root;
 
-	domain->dev_cnt += 1;
+	/* lock domain */
+	spin_lock(&domain->lock);
+
+	pte_root = virt_to_phys(domain->pt_root);
 
 	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;
 
-	write_lock_irqsave(&amd_iommu_devtable_lock, 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);
 
 	amd_iommu_pd_table[devid] = domain;
-	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
+
+	domain->dev_cnt += 1;
+
+	/* ready */
+	spin_unlock(&domain->lock);
 }
 
 static void attach_device(struct amd_iommu *iommu,
 			  struct protection_domain *domain,
 			  u16 devid)
 {
+	unsigned long flags;
+
+	write_lock_irqsave(&amd_iommu_devtable_lock, flags);
 	__attach_device(iommu, domain, devid);
+	write_unlock_irqrestore(&amd_iommu_devtable_lock, flags);
 
 	/*
 	 * We might boot into a crash-kernel here. The crashed kernel
-- 
cgit v1.2.3-59-g8ed1b


From 21129f786f231f7a9dce5b504617b893f50a435f Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 11:59:42 +0200
Subject: x86/amd-iommu: Make sure a device is assigned in passthrough mode

When the IOMMU driver runs in passthrough mode it has to
make sure that every device not assigned to an IOMMU-API
domain must be put into the passthrough domain instead of
keeping it unassigned.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 9 +++++++++
 1 file changed, 9 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 9aa135d4453f..a8e74c34dd27 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1141,6 +1141,15 @@ static void __detach_device(struct protection_domain *domain, u16 devid)
 
 	/* ready */
 	spin_unlock(&domain->lock);
+
+	/*
+	 * If we run in passthrough mode the device must be assigned to the
+	 * passthrough domain if it is detached from any other domain
+	 */
+	if (iommu_pass_through) {
+		struct amd_iommu *iommu = amd_iommu_rlookup_table[devid];
+		__attach_device(iommu, pt_domain, devid);
+	}
 }
 
 /*
-- 
cgit v1.2.3-59-g8ed1b


From a1ca331c8aa75cd58fdf685e2e8745e1d3ec5c8f Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 12:22:22 +0200
Subject: x86/amd-iommu: Don't detach device from pt domain on driver unbind

This patch makes sure a device is not detached from the
passthrough domain when the device driver is unloaded or
does otherwise release the device.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu.c | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index a8e74c34dd27..12a541deae5e 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -1195,6 +1195,8 @@ static int device_change_notifier(struct notifier_block *nb,
 	case BUS_NOTIFY_UNBOUND_DRIVER:
 		if (!domain)
 			goto out;
+		if (iommu_pass_through)
+			break;
 		detach_device(domain, devid);
 		break;
 	case BUS_NOTIFY_ADD_DEVICE:
-- 
cgit v1.2.3-59-g8ed1b


From 4751a95134e05f1172131d2001c6991d671fa58c Mon Sep 17 00:00:00 2001
From: Joerg Roedel <joerg.roedel@amd.com>
Date: Tue, 1 Sep 2009 15:53:54 +0200
Subject: x86/amd-iommu: Initialize passthrough mode when requested

This patch enables the passthrough mode for AMD IOMMU by
running the initialization function when iommu=pt is passed
on the kernel command line.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/amd_iommu_init.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

(limited to 'arch/x86/kernel')

diff --git a/arch/x86/kernel/amd_iommu_init.c b/arch/x86/kernel/amd_iommu_init.c
index c1b17e97252e..f00f489ab150 100644
--- a/arch/x86/kernel/amd_iommu_init.c
+++ b/arch/x86/kernel/amd_iommu_init.c
@@ -1242,12 +1242,18 @@ int __init amd_iommu_init(void)
 	if (ret)
 		goto free;
 
-	ret = amd_iommu_init_dma_ops();
+	if (iommu_pass_through)
+		ret = amd_iommu_init_passthrough();
+	else
+		ret = amd_iommu_init_dma_ops();
 	if (ret)
 		goto free;
 
 	enable_iommus();
 
+	if (iommu_pass_through)
+		goto out;
+
 	printk(KERN_INFO "AMD IOMMU: device isolation ");
 	if (amd_iommu_isolate)
 		printk("enabled\n");
-- 
cgit v1.2.3-59-g8ed1b