Web lists-archives.com

RE: [RFC PATCH] iommu/dma: account pci host bridge dma_mask for IOVA allocation

Hi Robin,

Currently this patch involves multiple framework.
I have coalesced the patch into one to present it as a whole as one RFC.

it involves

1) pcie of framework changes
2) iommu ops
3) pci dma-ranges discussion.
4) also it talks about the bug in device tree framework (dma-ranges) (just
in the commit message)

There are some minor glitches it remains with, which is open for feedback
and improvement.
And this could be split later into separate patches, once you see it as a
potential scope for solving the problem statement end to end.

-----Original Message-----
From: Oza Oza [mailto:oza.oza@xxxxxxxxxxxx]
Sent: Friday, March 17, 2017 11:45 AM
To: 'Joerg Roedel'; 'Robin Murphy'
Cc: 'iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx'; 'linux-kernel@xxxxxxxxxxxxxxx';
'linux-arm-kernel@xxxxxxxxxxxxxxxxxxx'; 'devicetree@xxxxxxxxxxxxxxx';
Subject: RE: [RFC PATCH] iommu/dma: account pci host bridge dma_mask for
IOVA allocation


There are certain areas which requires contemplation.
And this problem requires more attention from Pci of framework and iommu,
and integration of both.


-----Original Message-----
From: Oza Pawandeep [mailto:oza.oza@xxxxxxxxxxxx]
Sent: Friday, March 17, 2017 11:41 AM
To: Joerg Roedel; Robin Murphy
Cc: iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx;
bcm-kernel-feedback-list@xxxxxxxxxxxx; Oza Pawandeep
Subject: [RFC PATCH] iommu/dma: account pci host bridge dma_mask for IOVA

It is possible that PCI device supports 64-bit DMA addressing, and thus
it's driver sets device's dma_mask to DMA_BIT_MASK(64), however PCI host
bridge may have limitations on the inbound transaction addressing. As an
example, consider NVME SSD device connected to iproc-PCIe controller.

Currently, the IOMMU DMA ops only considers PCI device dma_mask when
allocating an IOVA. This is particularly problematic on
ARM/ARM64 SOCs where the IOMMU (i.e. SMMU) translates IOVA to PA for
in-bound transactions only after PCI Host has forwarded these transactions
on SOC IO bus. This means on such ARM/ARM64 SOCs the IOVA of in-bound
transactions has to honor the addressing restrictions of the PCI Host.

this patch is inspired by

but above inspiraiton solves the half of the problem.
the rest of the problem is descrbied below, what we face on iproc based

current pcie frmework and of framework integration assumes dma-ranges in a
way where memory-mapped devices define their dma-ranges.
dma-ranges: (child-bus-address, parent-bus-address, length).

but iproc based SOCs and even Rcar based SOCs has PCI world dma-ranges.
dma-ranges = <0x43000000 0x00 0x00 0x00 0x00 0x80 0x00>;

of_dma_configure is specifically witten to take care of memory mapped
but no implementation exists for pci to take care of pcie based memory
in fact pci world doesnt seem to define standard dma-ranges since there is
an absense of the same, the dma_mask used to remain 32bit because of
0 size return (parsed by of_dma_configure())

this patch also implements of_pci_get_dma_ranges to cater to pci world
so then the returned size get best possible (largest) dma_mask.
for e.g.
dma-ranges = <0x43000000 0x00 0x00 0x00 0x00 0x80 0x00>; we should get

conclusion: there are following problems
1) linux pci and iommu framework integration has glitches with respect to
2) pci linux framework look very uncertain about dma-ranges, rather
binding is not defined
   the way it is defined for memory mapped devices.
   rcar and iproc based SOCs use their custom one dma-ranges
   (rather can be standard)
3) even if in case of default parser of_dma_get_ranges,:
   it throws and erro"
   "no dma-ranges found for node"
   because of the bug which exists.
   following lines should be moved to the end of while(1)
	839                 node = of_get_next_parent(node);
	840                 if (!node)
	841                         break;

Reviewed-by: Anup Patel <anup.patel@xxxxxxxxxxxx>
Reviewed-by: Scott Branden <scott.branden@xxxxxxxxxxxx>
Signed-off-by: Oza Pawandeep <oza.oza@xxxxxxxxxxxx>

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig index
8c7c244..20cfff7 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -217,6 +217,9 @@ config NEED_DMA_MAP_STATE  config NEED_SG_DMA_LENGTH
 	def_bool y

+	def_bool y
 config SMP
 	def_bool y

diff --git a/arch/arm64/include/asm/device.h
b/arch/arm64/include/asm/device.h index 73d5bab..64b4dc3 100644
--- a/arch/arm64/include/asm/device.h
+++ b/arch/arm64/include/asm/device.h
@@ -20,6 +20,7 @@ struct dev_archdata {
 	void *iommu;			/* private IOMMU data */
+	u64 parent_dma_mask;
 	bool dma_coherent;

diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
index 81cdb2e..5845ecd 100644
--- a/arch/arm64/mm/dma-mapping.c
+++ b/arch/arm64/mm/dma-mapping.c
@@ -564,6 +564,7 @@ static void flush_page(struct device *dev, const void
*virt, phys_addr_t phys)
 	__dma_flush_area(virt, PAGE_SIZE);

 static void *__iommu_alloc_attrs(struct device *dev, size_t size,
 				 dma_addr_t *handle, gfp_t gfp,
 				 unsigned long attrs)
@@ -795,6 +796,20 @@ static void __iommu_unmap_sg_attrs(struct device
 	iommu_dma_unmap_sg(dev, sgl, nelems, dir, attrs);  }

+static int __iommu_set_dma_mask(struct device *dev, u64 mask) {
+	/* device is not DMA capable */
+	if (!dev->dma_mask)
+		return -EIO;
+	if (mask > dev->archdata.parent_dma_mask)
+		mask = dev->archdata.parent_dma_mask;
+	*dev->dma_mask = mask;
+	return 0;
 static const struct dma_map_ops iommu_dma_ops = {
 	.alloc = __iommu_alloc_attrs,
 	.free = __iommu_free_attrs,
@@ -811,8 +826,21 @@ static void __iommu_unmap_sg_attrs(struct device
 	.map_resource = iommu_dma_map_resource,
 	.unmap_resource = iommu_dma_unmap_resource,
 	.mapping_error = iommu_dma_mapping_error,
+	.set_dma_mask = __iommu_set_dma_mask,

+int dma_set_coherent_mask(struct device *dev, u64 mask) {
+	if (get_dma_ops(dev) == &iommu_dma_ops &&
+	    mask > dev->archdata.parent_dma_mask)
+		mask = dev->archdata.parent_dma_mask;
+	dev->coherent_dma_mask = mask;
+	return 0;
  * TODO: Right now __iommu_setup_dma_ops() gets called too early to do
  * everything it needs to - the device is only partially created and the
@@ -975,6 +1003,8 @@ void arch_setup_dma_ops(struct device *dev, u64
dma_base, u64 size,
 	if (!dev->dma_ops)
 		dev->dma_ops = &swiotlb_dma_ops;

+	dev->archdata.parent_dma_mask = size - 1;
 	dev->archdata.dma_coherent = coherent;
 	__iommu_setup_dma_ops(dev, dma_base, size, iommu);  } diff --git
a/drivers/of/of_pci.c b/drivers/of/of_pci.c index 0ee42c3..5804717 100644
--- a/drivers/of/of_pci.c
+++ b/drivers/of/of_pci.c
@@ -283,6 +283,51 @@ int of_pci_get_host_bridge_resources(struct
device_node *dev,
 	return err;
+int of_pci_get_dma_ranges(struct device_node *np, u64 *dma_addr, u64
+*paddr, u64 *size) {
+	struct device_node *node = of_node_get(np);
+	int rlen, naddr, nsize, pna;
+	int ret = 0;
+	const int na = 3, ns = 2;
+	struct of_pci_range_parser parser;
+	struct of_pci_range range;
+	if (!node)
+		return -EINVAL;
+	parser.node = node;
+	parser.pna = of_n_addr_cells(node);
+	parser.np = parser.pna + na + ns;
+	parser.range = of_get_property(node, "dma-ranges", &rlen);
+	if (!parser.range) {
+		pr_debug("pcie device has no dma-ranges defined for
node(%s)\n", np->full_name);
+		ret = -ENODEV;
+		goto out;
+	}
+	parser.end = parser.range + rlen / sizeof(__be32);
+	/* how do we take care of multiple dma windows ?. */
+	for_each_of_pci_range(&parser, &range) {
+		*dma_addr = range.pci_addr;
+		*size = range.size;
+		*paddr = range.cpu_addr;
+	}
+	pr_debug("dma_addr(%llx) cpu_addr(%llx) size(%llx)\n",
+		 *dma_addr, *paddr, *size);
+		 *dma_addr = range.pci_addr;
+		 *size = range.size;
+	of_node_put(node);
+	return ret;
 #endif /* CONFIG_OF_ADDRESS */

diff --git a/include/linux/of_pci.h b/include/linux/of_pci.h index
0e0974e..907ace0 100644
--- a/include/linux/of_pci.h
+++ b/include/linux/of_pci.h
@@ -76,6 +76,7 @@ static inline void of_pci_check_probe_only(void) { }
int of_pci_get_host_bridge_resources(struct device_node *dev,
 			unsigned char busno, unsigned char bus_max,
 			struct list_head *resources, resource_size_t
+int of_pci_get_dma_ranges(struct device_node *np, u64 *dma_addr, u64
+*paddr, u64 *size);
 static inline int of_pci_get_host_bridge_resources(struct device_node
 			unsigned char busno, unsigned char bus_max, @@
-83,6 +84,11 @@ static inline int of_pci_get_host_bridge_resources(struct
device_node *dev,  {
 	return -EINVAL;
+static inline int of_pci_get_dma_ranges(struct device_node *np, u64
+*dma_addr, u64 *paddr, u64 *size) {
+	return -EINVAL;

 #if defined(CONFIG_OF) && defined(CONFIG_PCI_MSI)