Skip to contentSkip to conversation

PCI: brcmstb: enable PCIe for STB chips

PATCH Superseded Jim Quinlan 6 participants 35 messages Jun 3, 2020 – Jun 5, 2020

Subsystems (30)

v3 of 12 Superseded Latest is v12 →

lore
Share & tools
lore
#0 lore raw

v3 cover PCI: brcmstb: enable PCIe for STB chips

v3:
Commit "device core: Introduce multiple dma pfn offsets" Commit "arm: dma-mapping: Invoke dma offset func if needed" -- The above two commits have been squashed. More importantly,
the code has been modified so that the functionality for multiple pfn offsets subsumes the use of dev->dma_pfn_offset. In fact, dma_pfn_offset is removed and supplanted by dma_pfn_offset_map, which is a pointer to an array. The more common case of a uniform offset is now handled as a map with a single entry, while cases requiring multiple pfn offsets use a map with multiple entries. Code paths that used to do this:

dev->dma_pfn_offset = mydrivers_pfn_offset;

have been changed to do this:

attach_uniform_dma_pfn_offset(dev, pfn_offset);

Commit "dt-bindings: PCI: Add bindings for more Brcmstb chips" -- Add if/then clause for required props: resets, reset-names (RobH) -- Change compatible list from const to enum (RobH)
-- Change list of u32-tuples to u64 (RobH)

Commit "of: Include a dev param in of_dma_get_range()"
-- modify of/unittests.c to add NULL param in of_dma_get_range() call.

Commit "device core: Add ability to handle multiple dma offsets" -- align comment in device.h (AndyS).
-- s/cpu_beg/cpu_start/ and s/dma_beg/dma_start/ in struct
dma_pfn_offset_region (AndyS).

v2:
Commit: "device core: Add ability to handle multiple dma offsets"
o Added helper func attach_dma_pfn_offset_map() in address.c (Chistoph)
o Helpers funcs added to __phys_to_dma() & __dma_to_phys() (Christoph)
o Added warning when multiple offsets are needed and !DMA_PFN_OFFSET_MAP
o dev->dma_pfn_map => dev->dma_pfn_offset_map
o s/frm/from/ for dma_pfn_offset_frm_{phys,dma}_addr() (Christoph)
o In device.h: s/const void */const struct dma_pfn_offset_region */
o removed 'unlikely' from unlikely(dev->dma_pfn_offset_map) since guarded by CONFIG_DMA_PFN_OFFSET_MAP (Christoph)
o Since dev->dma_pfn_offset is copied in usb/core/{usb,message}.c, now

    dev->dma_pfn_offset_map is copied as well.

o Merged two of the DMA commits into one (Christoph).

Commit "arm: dma-mapping: Invoke dma offset func if needed":
o Use helper functions instead of #if CONFIG_DMA_PFN_OFFSET

Other commits' changes:
o Removed need for carrying of_id var in priv (Nicolas)
o Commit message rewordings (Bjorn)
o Commit log messages filled to 75 chars (Bjorn)
o devm_reset_control_get_shared())
=> devm_reset_control_get_optional_shared (Philipp)
o Add call to reset_control_assert() in PCIe remove routines (Philipp)

v1:
This patchset expands the usefulness of the Broadcom Settop Box PCIe controller by building upon the PCIe driver used currently by the Raspbery Pi. Other forms of this patchset were submitted by me years ago and not accepted; the major sticking point was the code required for the DMA remapping needed for the PCIe driver to work [1].

There have been many changes to the DMA and OF subsystems since that time, making a cleaner and less intrusive patchset possible. This patchset implements a generalization of "dev->dma_pfn_offset", except that instead of a single scalar offset it provides for multiple offsets via a function which depends upon the "dma-ranges" property of the PCIe host controller. This is required for proper functionality of the BrcmSTB PCIe controller and possibly some other devices.

[1] https://lore.kernel.org/linux-arm-kernel/[email protected]/

Jim Quinlan (13):
PCI: brcmstb: PCIE_BRCMSTB depends on ARCH_BRCMSTB
ata: ahci_brcm: Fix use of BCM7216 reset controller
dt-bindings: PCI: Add bindings for more Brcmstb chips
PCI: brcmstb: Add bcm7278 reigister info
PCI: brcmstb: Add suspend and resume pm_ops
PCI: brcmstb: Add bcm7278 PERST support
PCI: brcmstb: Add control of rescal reset
of: Include a dev param in of_dma_get_range()
device core: Introduce multiple dma pfn offsets
PCI: brcmstb: Set internal memory viewport sizes
PCI: brcmstb: Accommodate MSI for older chips
PCI: brcmstb: Set bus max burst size by chip type
PCI: brcmstb: Add bcm7211, bcm7216, bcm7445, bcm7278 to match list

 .../bindings/pci/brcm,stb-pcie.yaml           |  58 ++-
 arch/arm/include/asm/dma-mapping.h            |   9 +-
 arch/arm/mach-keystone/keystone.c             |   9 +-
 arch/sh/drivers/pci/pcie-sh7786.c             |   3 +-
 arch/sh/kernel/dma-coherent.c                 |  17 +-
 arch/x86/pci/sta2x11-fixup.c                  |   7 +-
 drivers/acpi/arm64/iort.c                     |   5 +-
 drivers/ata/ahci_brcm.c                       |  14 +-
 drivers/gpu/drm/sun4i/sun4i_backend.c         |   7 +-
 drivers/iommu/io-pgtable-arm.c                |   2 +-
 .../platform/sunxi/sun4i-csi/sun4i_csi.c      |   5 +-
 .../platform/sunxi/sun6i-csi/sun6i_csi.c      |   5 +-
 drivers/of/address.c                          |  97 ++++-
 drivers/of/device.c                           |  10 +-
 drivers/of/of_private.h                       |   8 +-
 drivers/of/unittest.c                         |   2 +-
 drivers/pci/controller/Kconfig                |   3 +-
 drivers/pci/controller/pcie-brcmstb.c         | 408 +++++++++++++++---
 drivers/remoteproc/remoteproc_core.c          |   2 +-
 .../staging/media/sunxi/cedrus/cedrus_hw.c    |   7 +-
 drivers/usb/core/message.c                    |   4 +-
 drivers/usb/core/usb.c                        |   2 +-
 include/linux/device.h                        |   4 +-
 include/linux/dma-direct.h                    |  16 +-
 include/linux/dma-mapping.h                   |  45 ++
 kernel/dma/coherent.c                         |  11 +-
 26 files changed, 631 insertions(+), 129 deletions(-)
Signature · 1 line
2.17.1
#1 lore raw

PATCH v3 1/13 PCI: brcmstb: PCIE_BRCMSTB depends on ARCH_BRCMSTB

From: Jim Quinlan <[email protected]>

Have PCIE_BRCMSTB depend on ARCH_BRCMSTB. Also set the default value to ARCH_BRCMSTB.

---
 drivers/pci/controller/Kconfig | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/controller/Kconfig +2 -1
index 91bfdb784829..c0f3d4d10047 100644
@@ -244,9 +244,10 @@ config VMD
 
 config PCIE_BRCMSTB
 	tristate "Broadcom Brcmstb PCIe host controller"
-	depends on ARCH_BCM2835 || COMPILE_TEST
+	depends on ARCH_BRCMSTB || ARCH_BCM2835 || COMPILE_TEST
 	depends on OF
 	depends on PCI_MSI_IRQ_DOMAIN
+	default ARCH_BRCMSTB
 	help
 	  Say Y here to enable PCIe host controller support for
 	  Broadcom STB based SoCs, like the Raspberry Pi 4.
Signature · 1 line
2.17.1
#2 lore raw

PATCH v3 2/13 ata: ahci_brcm: Fix use of BCM7216 reset controller

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

A reset controller "rescal" is shared between the AHCI driver and the PCIe driver for the BrcmSTB 7216 chip. The code is modified to allow this sharing and to deassert() properly.

Fixes 272ecd60a636 ("ata: ahci_brcm: BCM7216 reset is self de-asserting")
Fixes c345ec6a50e9 ("ata: ahci_brcm: Support BCM7216 reset controller

name")

---
 drivers/ata/ahci_brcm.c | 14 +++++---------
 1 file changed, 5 insertions(+), 9 deletions(-)
drivers/ata/ahci_brcm.c +5 -9
index 6853dbb4131d..c4ea555573dd 100644
@@ -428,7 +428,6 @@ static int brcm_ahci_probe(struct platform_device *pdev)
 {
 	const struct of_device_id *of_id;
 	struct device *dev = &pdev->dev;
-	const char *reset_name = NULL;
 	struct brcm_ahci_priv *priv;
 	struct ahci_host_priv *hpriv;
 	struct resource *res;
@@ -452,11 +451,11 @@ static int brcm_ahci_probe(struct platform_device *pdev)
 
 	/* Reset is optional depending on platform and named differently */
 	if (priv->version == BRCM_SATA_BCM7216)
-		reset_name = "rescal";
+		priv->rcdev = devm_reset_control_get_optional_shared(&pdev->dev,
+								     "rescal");
 	else
-		reset_name = "ahci";
-
-	priv->rcdev = devm_reset_control_get_optional(&pdev->dev, reset_name);
+		priv->rcdev = devm_reset_control_get_optional(&pdev->dev,
+							      "ahci");
 	if (IS_ERR(priv->rcdev))
 		return PTR_ERR(priv->rcdev);
 
@@ -479,10 +478,7 @@ static int brcm_ahci_probe(struct platform_device *pdev)
 		break;
 	}
 
-	if (priv->version == BRCM_SATA_BCM7216)
-		ret = reset_control_reset(priv->rcdev);
-	else
-		ret = reset_control_deassert(priv->rcdev);
+	ret = reset_control_deassert(priv->rcdev);
 	if (ret)
 		return ret;
 
Signature · 1 line
2.17.1
#3 lore raw

Re: [PATCH v3 02/13] ata: ahci_brcm: Fix use of BCM7216 reset controller

Quoted 12 lines · Jim Quinlan

From: Jim Quinlan <[email protected]>

A reset controller "rescal" is shared between the AHCI driver and the PCIe driver for the BrcmSTB 7216 chip. The code is modified to allow this sharing and to deassert() properly.

Fixes 272ecd60a636 ("ata: ahci_brcm: BCM7216 reset is self de-asserting")
Fixes c345ec6a50e9 ("ata: ahci_brcm: Support BCM7216 reset controller

name")
---

[snip]

drivers/ata/ahci_brcm.c static int brcm_ahci_probe(struct platform_device *pdev) · line 483

Quoted 11 lines
drivers/ata/ahci_brcm.c +1 -4
@@ -479,10 +478,7 @@ static int brcm_ahci_probe(struct platform_device *pdev)
 		break;
 	}
 
-	if (priv->version == BRCM_SATA_BCM7216)
-		ret = reset_control_reset(priv->rcdev);
-	else
-		ret = reset_control_deassert(priv->rcdev);
+	ret = reset_control_deassert(priv->rcdev);
 	if (ret)
 		return ret;

Do we need a similar change for brcm_ahci_resume()?

Signature · 1 line
Florian
#4 lore raw

PATCH v3 3/13 dt-bindings: PCI: Add bindings for more Brcmstb chips

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

- Add compatible strings for three more Broadcom STB chips: 7278, 7216, 7211 (STB version of RPi4).
- add new property 'brcm,scb-sizes'
- add new property 'resets'
- add new property 'reset-names' for 7216 only
- allow 'ranges' and 'dma-ranges' to have more than one item and update the example to show this.

---
 .../bindings/pci/brcm,stb-pcie.yaml           | 58 ++++++++++++++++---
 1 file changed, 51 insertions(+), 7 deletions(-)
Documentation/devicetree/bindings/pci/brcm,stb-pcie.yaml +51 -7
index 8680a0f86c5a..4a012d77513f 100644
@@ -9,12 +9,15 @@ title: Brcmstb PCIe Host Controller Device Tree Bindings
 maintainers:
   - Nicolas Saenz Julienne <[email protected]>
 
-allOf:
-  - $ref: /schemas/pci/pci-bus.yaml#
-
 properties:
   compatible:
-    const: brcm,bcm2711-pcie # The Raspberry Pi 4
+    items:
+      - enum:
+          - brcm,bcm2711-pcie # The Raspberry Pi 4
+          - brcm,bcm7211-pcie # Broadcom STB version of RPi4
+          - brcm,bcm7278-pcie # Broadcom 7278 Arm
+          - brcm,bcm7216-pcie # Broadcom 7216 Arm
+          - brcm,bcm7445-pcie # Broadcom 7445 Arm
 
   reg:
     maxItems: 1
@@ -34,10 +37,12 @@ properties:
       - const: msi
 
   ranges:
-    maxItems: 1
+    minItems: 1
+    maxItems: 4
 
   dma-ranges:
-    maxItems: 1
+    minItems: 1
+    maxItems: 6
 
   clocks:
     maxItems: 1
@@ -58,8 +63,33 @@ properties:
 
   aspm-no-l0s: true
 
+  resets:
+    description: for "brcm,bcm7216-pcie", must be a valid reset
+      phandle pointing to the RESCAL reset controller provider node.
+    $ref: "/schemas/types.yaml#/definitions/phandle"
+
+  reset-names:
+    items:
+      - const: rescal
+
+  brcm,scb-sizes:
+    description: u64 giving the 64bit PCIe memory
+      viewport size of a memory controller.  There may be up to
+      three controllers, and each size must be a power of two
+      with a size greater or equal to the amount of memory the
+      controller supports.  Note that each memory controller
+      may have two component regions -- base and extended -- so
+      this information cannot be deduced from the dma-ranges.
+
+    allOf:
+      - $ref: /schemas/types.yaml#/definitions/uint64-array
+      - items:
+          minItems: 1
+          maxItems: 3
+
 required:
   - reg
+  - ranges
   - dma-ranges
   - "#interrupt-cells"
   - interrupts
@@ -68,6 +98,18 @@ required:
   - interrupt-map
   - msi-controller
 
+allOf:
+  - $ref: /schemas/pci/pci-bus.yaml#
+  - if:
+      properties:
+        compatible:
+          contains:
+            const: brcm,bcm7216-pcie
+    then:
+      required:
+        - resets
+        - reset-names
+
 unevaluatedProperties: false
 
 examples:
@@ -93,7 +135,9 @@ examples:
                     msi-parent = <&pcie0>;
                     msi-controller;
                     ranges = <0x02000000 0x0 0xf8000000 0x6 0x00000000 0x0 0x04000000>;
-                    dma-ranges = <0x02000000 0x0 0x00000000 0x0 0x00000000 0x0 0x80000000>;
+                    dma-ranges = <0x42000000 0x1 0x00000000 0x0 0x40000000 0x0 0x80000000>,
+                                 <0x42000000 0x1 0x80000000 0x3 0x00000000 0x0 0x80000000>;
                     brcm,enable-ssc;
+                    brcm,scb-sizes =  <0x0000000080000000 0x0000000080000000>;
             };
     };
Signature · 7 lines
2.17.1


_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
#5 lore raw

PATCH v3 4/13 PCI: brcmstb: Add bcm7278 reigister info

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

Add in compatibility strings and code for three Broadcom STB chips. Some of the register locations, shifts, and masks are different for certain chips, requiring the use of different constants based on of_id.

We would like to add the following at this time to the match list but we need to wait until the end of this patchset so that everything works.

    { .compatible = "brcm,bcm7211-pcie", .data = &generic_cfg },
    { .compatible = "brcm,bcm7278-pcie", .data = &bcm7278_cfg },
    { .compatible = "brcm,bcm7216-pcie", .data = &bcm7278_cfg },
    { .compatible = "brcm,bcm7445-pcie", .data = &generic_cfg },
---
 drivers/pci/controller/pcie-brcmstb.c | 108 +++++++++++++++++++++++---
 1 file changed, 96 insertions(+), 12 deletions(-)
drivers/pci/controller/pcie-brcmstb.c +96 -12
index 73020b4ff090..7c707e483181 100644
@@ -120,9 +120,8 @@
 #define  PCIE_EXT_SLOT_SHIFT				15
 #define  PCIE_EXT_FUNC_SHIFT				12
 
-#define PCIE_RGR1_SW_INIT_1				0x9210
 #define  PCIE_RGR1_SW_INIT_1_PERST_MASK			0x1
-#define  PCIE_RGR1_SW_INIT_1_INIT_MASK			0x2
+#define  PCIE_RGR1_SW_INIT_1_PERST_SHIFT		0x0
 
 /* PCIe parameters */
 #define BRCM_NUM_PCIE_OUT_WINS		0x4
@@ -152,6 +151,76 @@
 #define SSC_STATUS_SSC_MASK		0x400
 #define SSC_STATUS_PLL_LOCK_MASK	0x800
 
+#define IDX_ADDR(pcie)	\
+	(pcie->reg_offsets[EXT_CFG_INDEX])
+#define DATA_ADDR(pcie)	\
+	(pcie->reg_offsets[EXT_CFG_DATA])
+#define PCIE_RGR1_SW_INIT_1(pcie) \
+	(pcie->reg_offsets[RGR1_SW_INIT_1])
+
+enum {
+	RGR1_SW_INIT_1,
+	EXT_CFG_INDEX,
+	EXT_CFG_DATA,
+};
+
+enum {
+	RGR1_SW_INIT_1_INIT_MASK,
+	RGR1_SW_INIT_1_INIT_SHIFT,
+};
+
+enum pcie_type {
+	GENERIC,
+	BCM7278,
+	BCM2711,
+};
+
+struct pcie_cfg_data {
+	const int *reg_field_info;
+	const int *offsets;
+	const enum pcie_type type;
+};
+
+static const int pcie_reg_field_info[] = {
+	[RGR1_SW_INIT_1_INIT_MASK] = 0x2,
+	[RGR1_SW_INIT_1_INIT_SHIFT] = 0x1,
+};
+
+static const int pcie_reg_field_info_bcm7278[] = {
+	[RGR1_SW_INIT_1_INIT_MASK] = 0x1,
+	[RGR1_SW_INIT_1_INIT_SHIFT] = 0x0,
+};
+
+static const int pcie_offsets[] = {
+	[RGR1_SW_INIT_1] = 0x9210,
+	[EXT_CFG_INDEX]  = 0x9000,
+	[EXT_CFG_DATA]   = 0x9004,
+};
+
+static const struct pcie_cfg_data generic_cfg = {
+	.reg_field_info	= pcie_reg_field_info,
+	.offsets	= pcie_offsets,
+	.type		= GENERIC,
+};
+
+static const int pcie_offset_bcm7278[] = {
+	[RGR1_SW_INIT_1] = 0xc010,
+	[EXT_CFG_INDEX] = 0x9000,
+	[EXT_CFG_DATA] = 0x9004,
+};
+
+static const struct pcie_cfg_data bcm7278_cfg = {
+	.reg_field_info = pcie_reg_field_info_bcm7278,
+	.offsets	= pcie_offset_bcm7278,
+	.type		= BCM7278,
+};
+
+static const struct pcie_cfg_data bcm2711_cfg = {
+	.reg_field_info	= pcie_reg_field_info,
+	.offsets	= pcie_offsets,
+	.type		= BCM2711,
+};
+
 struct brcm_msi {
 	struct device		*dev;
 	void __iomem		*base;
@@ -176,6 +245,9 @@ struct brcm_pcie {
 	int			gen;
 	u64			msi_target_addr;
 	struct brcm_msi		*msi;
+	const int		*reg_offsets;
+	const int		*reg_field_info;
+	enum pcie_type		type;
 };
 
 /*
@@ -602,20 +674,21 @@ static struct pci_ops brcm_pcie_ops = {
 
 static inline void brcm_pcie_bridge_sw_init_set(struct brcm_pcie *pcie, u32 val)
 {
-	u32 tmp;
+	u32 tmp, mask =  pcie->reg_field_info[RGR1_SW_INIT_1_INIT_MASK];
+	u32 shift = pcie->reg_field_info[RGR1_SW_INIT_1_INIT_SHIFT];
 
-	tmp = readl(pcie->base + PCIE_RGR1_SW_INIT_1);
-	u32p_replace_bits(&tmp, val, PCIE_RGR1_SW_INIT_1_INIT_MASK);
-	writel(tmp, pcie->base + PCIE_RGR1_SW_INIT_1);
+	tmp = readl(pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
+	tmp = (tmp & ~mask) | ((val << shift) & mask);
+	writel(tmp, pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
 }
 
 static inline void brcm_pcie_perst_set(struct brcm_pcie *pcie, u32 val)
 {
 	u32 tmp;
 
-	tmp = readl(pcie->base + PCIE_RGR1_SW_INIT_1);
+	tmp = readl(pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
 	u32p_replace_bits(&tmp, val, PCIE_RGR1_SW_INIT_1_PERST_MASK);
-	writel(tmp, pcie->base + PCIE_RGR1_SW_INIT_1);
+	writel(tmp, pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
 }
 
 static inline int brcm_pcie_get_rc_bar2_size_and_offset(struct brcm_pcie *pcie,
@@ -924,10 +997,16 @@ static int brcm_pcie_remove(struct platform_device *pdev)
 	return 0;
 }
 
+static const struct of_device_id brcm_pcie_match[] = {
+	{ .compatible = "brcm,bcm2711-pcie", .data = &bcm2711_cfg },
+	{},
+};
+
 static int brcm_pcie_probe(struct platform_device *pdev)
 {
 	struct device_node *np = pdev->dev.of_node, *msi_np;
 	struct pci_host_bridge *bridge;
+	const struct pcie_cfg_data *data;
 	struct brcm_pcie *pcie;
 	struct pci_bus *child;
 	struct resource *res;
@@ -937,9 +1016,18 @@ static int brcm_pcie_probe(struct platform_device *pdev)
 	if (!bridge)
 		return -ENOMEM;
 
+	data = of_device_get_match_data(&pdev->dev);
+	if (!data) {
+		pr_err("failed to look up compatible string\n");
+		return -EINVAL;
+	}
+
 	pcie = pci_host_bridge_priv(bridge);
 	pcie->dev = &pdev->dev;
 	pcie->np = np;
+	pcie->reg_offsets = data->offsets;
+	pcie->reg_field_info = data->reg_field_info;
+	pcie->type = data->type;
 
 	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 	pcie->base = devm_ioremap_resource(&pdev->dev, res);
@@ -1005,10 +1093,6 @@ static int brcm_pcie_probe(struct platform_device *pdev)
 	return ret;
 }
 
-static const struct of_device_id brcm_pcie_match[] = {
-	{ .compatible = "brcm,bcm2711-pcie" },
-	{},
-};
 MODULE_DEVICE_TABLE(of, brcm_pcie_match);
 
 static struct platform_driver brcm_pcie_driver = {
Signature · 1 line
2.17.1
Acked-by on v3 4/13
#6 lore raw

Re: [PATCH v3 04/13] PCI: brcmstb: Add bcm7278 reigister info

Quoted 6 lines · Jim Quinlan

From: Jim Quinlan <[email protected]>

Add in compatibility strings and code for three Broadcom STB chips. Some of the register locations, shifts, and masks are different for certain chips, requiring the use of different constants based on of_id.

Quoted 7 lines

We would like to add the following at this time to the match list but we need to wait until the end of this patchset so that everything works.

    { .compatible = "brcm,bcm7211-pcie", .data = &generic_cfg },
    { .compatible = "brcm,bcm7278-pcie", .data = &bcm7278_cfg },
    { .compatible = "brcm,bcm7216-pcie", .data = &bcm7278_cfg },
    { .compatible = "brcm,bcm7445-pcie", .data = &generic_cfg },

If you need to resubmit, there is a typo in the subject: reigister vs. register. Other than that:

Signature · 1 line
Florian
#7 lore raw

PATCH v3 5/13 PCI: brcmstb: Add suspend and resume pm_ops

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

Broadcom Set-top (BrcmSTB) boards typically support S2, S3, and S5 suspend and resume. Now the PCIe driver may do so as well.

---
 drivers/pci/controller/pcie-brcmstb.c | 49 +++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)
drivers/pci/controller/pcie-brcmstb.c +49 -0
index 7c707e483181..f444751e247c 100644
@@ -979,6 +979,49 @@ static void brcm_pcie_turn_off(struct brcm_pcie *pcie)
 	brcm_pcie_bridge_sw_init_set(pcie, 1);
 }
 
+static int brcm_pcie_suspend(struct device *dev)
+{
+	struct brcm_pcie *pcie = dev_get_drvdata(dev);
+	int ret = 0;
+
+	brcm_pcie_turn_off(pcie);
+	clk_disable_unprepare(pcie->clk);
+
+	return ret;
+}
+
+static int brcm_pcie_resume(struct device *dev)
+{
+	struct brcm_pcie *pcie = dev_get_drvdata(dev);
+	void __iomem *base;
+	u32 tmp;
+	int ret;
+
+	base = pcie->base;
+	clk_prepare_enable(pcie->clk);
+
+	/* Take bridge out of reset so we can access the SERDES reg */
+	brcm_pcie_bridge_sw_init_set(pcie, 0);
+
+	/* SERDES_IDDQ = 0 */
+	tmp = readl(base + PCIE_MISC_HARD_PCIE_HARD_DEBUG);
+	u32p_replace_bits(&tmp, 0,
+			  PCIE_MISC_HARD_PCIE_HARD_DEBUG_SERDES_IDDQ_MASK);
+	writel(tmp, base + PCIE_MISC_HARD_PCIE_HARD_DEBUG);
+
+	/* wait for serdes to be stable */
+	udelay(100);
+
+	ret = brcm_pcie_setup(pcie);
+	if (ret)
+		return ret;
+
+	if (pcie->msi)
+		brcm_msi_set_regs(pcie->msi);
+
+	return 0;
+}
+
 static void __brcm_pcie_remove(struct brcm_pcie *pcie)
 {
 	brcm_msi_remove(pcie);
@@ -1095,12 +1138,18 @@ static int brcm_pcie_probe(struct platform_device *pdev)
 
 MODULE_DEVICE_TABLE(of, brcm_pcie_match);
 
+static const struct dev_pm_ops brcm_pcie_pm_ops = {
+	.suspend_noirq = brcm_pcie_suspend,
+	.resume_noirq = brcm_pcie_resume,
+};
+
 static struct platform_driver brcm_pcie_driver = {
 	.probe = brcm_pcie_probe,
 	.remove = brcm_pcie_remove,
 	.driver = {
 		.name = "brcm-pcie",
 		.of_match_table = brcm_pcie_match,
+		.pm = &brcm_pcie_pm_ops,
 	},
 };
 module_platform_driver(brcm_pcie_driver);
Signature · 1 line
2.17.1
#8 lore raw

Re: [PATCH v3 05/13] PCI: brcmstb: Add suspend and resume pm_ops

Quoted 17 lines · Jim Quinlan

From: Jim Quinlan <[email protected]>

Broadcom Set-top (BrcmSTB) boards typically support S2, S3, and S5 suspend and resume. Now the PCIe driver may do so as well.

---
 drivers/pci/controller/pcie-brcmstb.c | 49 +++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)
drivers/pci/controller/pcie-brcmstb.c +0 -0
index 7c707e483181..f444751e247c 100644
@@ -979,6 +979,49 @@ static void brcm_pcie_turn_off(struct brcm_pcie *pcie)
 	brcm_pcie_bridge_sw_init_set(pcie, 1);
 }

drivers/pci/controller/pcie-brcmstb.c static void brcm_pcie_turn_off(struct brcm_pcie *pcie) · line 990

Quoted 10 lines
drivers/pci/controller/pcie-brcmstb.c +9 -0
 
+static int brcm_pcie_suspend(struct device *dev)
+{
+	struct brcm_pcie *pcie = dev_get_drvdata(dev);
+	int ret = 0;
+
+	brcm_pcie_turn_off(pcie);
+	clk_disable_unprepare(pcie->clk);
+
+	return ret;

No need for "ret".

Quoted 1 line
drivers/pci/controller/pcie-brcmstb.c +1 -0
+}
#9 lore raw

PATCH v3 6/13 PCI: brcmstb: Add bcm7278 PERST support

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

The PERST bit was moved to a different register in 7278-type STB chips. In addition, the polarity of the bit was also changed; for other chips writing a 1 specified assert; for 7278-type chips, writing a 0 specifies assert.

Signal-wise, PERST is an asserted-low signal.

---
 drivers/pci/controller/pcie-brcmstb.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)
drivers/pci/controller/pcie-brcmstb.c +12 -3
index f444751e247c..0bcae9eba048 100644
@@ -81,6 +81,7 @@
 
 #define PCIE_MISC_PCIE_CTRL				0x4064
 #define  PCIE_MISC_PCIE_CTRL_PCIE_L23_REQUEST_MASK	0x1
+#define PCIE_MISC_PCIE_CTRL_PCIE_PERSTB_MASK		0x4
 
 #define PCIE_MISC_PCIE_STATUS				0x4068
 #define  PCIE_MISC_PCIE_STATUS_PCIE_PORT_MASK		0x80
@@ -686,9 +687,17 @@ static inline void brcm_pcie_perst_set(struct brcm_pcie *pcie, u32 val)
 {
 	u32 tmp;
 
-	tmp = readl(pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
-	u32p_replace_bits(&tmp, val, PCIE_RGR1_SW_INIT_1_PERST_MASK);
-	writel(tmp, pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
+	if (pcie->type == BCM7278) {
+		/* Perst bit has moved and assert value is 0 */
+		tmp = readl(pcie->base + PCIE_MISC_PCIE_CTRL);
+		u32p_replace_bits(&tmp,
+				  !val, PCIE_MISC_PCIE_CTRL_PCIE_PERSTB_MASK);
+		writel(tmp, pcie->base +  PCIE_MISC_PCIE_CTRL);
+	} else {
+		tmp = readl(pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
+		u32p_replace_bits(&tmp, val, PCIE_RGR1_SW_INIT_1_PERST_MASK);
+		writel(tmp, pcie->base + PCIE_RGR1_SW_INIT_1(pcie));
+	}
 }
 
 static inline int brcm_pcie_get_rc_bar2_size_and_offset(struct brcm_pcie *pcie,
Signature · 1 line
2.17.1
Acked-by on v3 6/13
#10 lore raw

Re: [PATCH v3 06/13] PCI: brcmstb: Add bcm7278 PERST support

Quoted 9 lines · Jim Quinlan

From: Jim Quinlan <[email protected]>

The PERST bit was moved to a different register in 7278-type STB chips. In addition, the polarity of the bit was also changed; for other chips writing a 1 specified assert; for 7278-type chips, writing a 0 specifies assert.

Signal-wise, PERST is an asserted-low signal.

Signature · 6 lines
Florian

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
#11 lore raw

PATCH v3 7/13 PCI: brcmstb: Add control of rescal reset

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

Some STB chips have a special purpose reset controller named RESCAL (reset calibration). The PCIe HW can now control RESCAL to start and stop its operation.

---
 drivers/pci/controller/pcie-brcmstb.c | 81 ++++++++++++++++++++++++++-
 1 file changed, 80 insertions(+), 1 deletion(-)
drivers/pci/controller/pcie-brcmstb.c +80 -1
index 0bcae9eba048..fa356bc149c3 100644
@@ -23,6 +23,7 @@
 #include <linux/of_platform.h>
 #include <linux/pci.h>
 #include <linux/printk.h>
+#include <linux/reset.h>
 #include <linux/sizes.h>
 #include <linux/slab.h>
 #include <linux/string.h>
@@ -152,7 +153,17 @@
 #define SSC_STATUS_SSC_MASK		0x400
 #define SSC_STATUS_PLL_LOCK_MASK	0x800
 
-#define IDX_ADDR(pcie)	\
+/* Rescal registers */
+#define PCIE_DVT_PMU_PCIE_PHY_CTRL				0xc700
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_NFLDS			0x3
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_DIG_RESET_MASK		0x4
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_DIG_RESET_SHIFT	0x2
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_RESET_MASK		0x2
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_RESET_SHIFT		0x1
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_PWRDN_MASK		0x1
+#define  PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_PWRDN_SHIFT		0x0
+
+#define IDX_ADDR(pcie) \
 	(pcie->reg_offsets[EXT_CFG_INDEX])
 #define DATA_ADDR(pcie)	\
 	(pcie->reg_offsets[EXT_CFG_DATA])
@@ -249,6 +260,7 @@ struct brcm_pcie {
 	const int		*reg_offsets;
 	const int		*reg_field_info;
 	enum pcie_type		type;
+	struct reset_control	*rescal;
 };
 
 /*
@@ -964,6 +976,47 @@ static void brcm_pcie_enter_l23(struct brcm_pcie *pcie)
 		dev_err(pcie->dev, "failed to enter low-power link state\n");
 }
 
+static int brcm_phy_cntl(struct brcm_pcie *pcie, const int start)
+{
+	static const u32 shifts[PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_NFLDS] = {
+		PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_PWRDN_SHIFT,
+		PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_RESET_SHIFT,
+		PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_DIG_RESET_SHIFT,};
+	static const u32 masks[PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_NFLDS] = {
+		PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_PWRDN_MASK,
+		PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_RESET_MASK,
+		PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_DIG_RESET_MASK,};
+	const int beg = start ? 0 : PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_NFLDS - 1;
+	const int end = start ? PCIE_DVT_PMU_PCIE_PHY_CTRL_DAST_NFLDS : -1;
+	u32 tmp, combined_mask = 0;
+	u32 val = !!start;
+	void __iomem *base = pcie->base;
+	int i;
+
+	for (i = beg; i != end; start ? i++ : i--) {
+		tmp = readl(base + PCIE_DVT_PMU_PCIE_PHY_CTRL);
+		tmp = (tmp & ~masks[i]) | ((val << shifts[i]) & masks[i]);
+		writel(tmp, base + PCIE_DVT_PMU_PCIE_PHY_CTRL);
+		usleep_range(50, 200);
+		combined_mask |= masks[i];
+	}
+
+	tmp = readl(base + PCIE_DVT_PMU_PCIE_PHY_CTRL);
+	val = start ? combined_mask : 0;
+
+	return (tmp & combined_mask) == val ? 0 : -EIO;
+}
+
+static inline int brcm_phy_start(struct brcm_pcie *pcie)
+{
+	return pcie->rescal ? brcm_phy_cntl(pcie, 1) : 0;
+}
+
+static inline int brcm_phy_stop(struct brcm_pcie *pcie)
+{
+	return pcie->rescal ? brcm_phy_cntl(pcie, 0) : 0;
+}
+
 static void brcm_pcie_turn_off(struct brcm_pcie *pcie)
 {
 	void __iomem *base = pcie->base;
@@ -994,6 +1047,9 @@ static int brcm_pcie_suspend(struct device *dev)
 	int ret = 0;
 
 	brcm_pcie_turn_off(pcie);
+	ret = brcm_phy_stop(pcie);
+	if (ret)
+		dev_err(pcie->dev, "failed to stop phy\n");
 	clk_disable_unprepare(pcie->clk);
 
 	return ret;
@@ -1009,6 +1065,12 @@ static int brcm_pcie_resume(struct device *dev)
 	base = pcie->base;
 	clk_prepare_enable(pcie->clk);
 
+	ret = brcm_phy_start(pcie);
+	if (ret) {
+		dev_err(pcie->dev, "failed to start phy\n");
+		return ret;
+	}
+
 	/* Take bridge out of reset so we can access the SERDES reg */
 	brcm_pcie_bridge_sw_init_set(pcie, 0);
 
@@ -1035,6 +1097,9 @@ static void __brcm_pcie_remove(struct brcm_pcie *pcie)
 {
 	brcm_msi_remove(pcie);
 	brcm_pcie_turn_off(pcie);
+	if (brcm_phy_stop(pcie))
+		dev_err(pcie->dev, "failed to stop phy\n");
+	reset_control_assert(pcie->rescal);
 	clk_disable_unprepare(pcie->clk);
 }
 
@@ -1105,6 +1170,20 @@ static int brcm_pcie_probe(struct platform_device *pdev)
 		dev_err(&pdev->dev, "could not enable clock\n");
 		return ret;
 	}
+	pcie->rescal = devm_reset_control_get_optional_shared(&pdev->dev,
+							      "rescal");
+	if (IS_ERR(pcie->rescal))
+		return PTR_ERR(pcie->rescal);
+
+	ret = reset_control_deassert(pcie->rescal);
+	if (ret)
+		dev_err(&pdev->dev, "failed to deassert 'rescal'\n");
+
+	ret = brcm_phy_start(pcie);
+	if (ret) {
+		dev_err(pcie->dev, "failed to start phy\n");
+		return ret;
+	}
 
 	ret = brcm_pcie_setup(pcie);
 	if (ret)
Signature · 1 line
2.17.1
#13 lore raw

PATCH v3 8/13 of: Include a dev param in of_dma_get_range()

In reply to Jim Quinlan

Currently there is only one caller of of_dma_get_range(). A struct device *dev param is needed For implementing multiple dma offsets. This function will still work if dev == NULL.

---
 drivers/of/address.c    | 4 +++-
 drivers/of/device.c     | 2 +-
 drivers/of/of_private.h | 8 ++++----
 drivers/of/unittest.c   | 2 +-
 4 files changed, 9 insertions(+), 7 deletions(-)
drivers/of/address.c +3 -1
index 8eea3f6e29a4..96d8cfb14a60 100644
@@ -920,6 +920,7 @@ EXPORT_SYMBOL(of_io_request_and_map);
 
 /**
  * of_dma_get_range - Get DMA range info
+ * @dev:	device pointer; only needed for a corner case.
  * @np:		device node to get DMA range info
  * @dma_addr:	pointer to store initial DMA address of DMA range
  * @paddr:	pointer to store initial CPU address of DMA range
@@ -935,7 +936,8 @@ EXPORT_SYMBOL(of_io_request_and_map);
  * It returns -ENODEV if "dma-ranges" property was not found
  * for this device in DT.
  */
-int of_dma_get_range(struct device_node *np, u64 *dma_addr, u64 *paddr, u64 *size)
+int of_dma_get_range(struct device *dev, struct device_node *np, u64 *dma_addr,
+		     u64 *paddr, u64 *size)
 {
 	struct device_node *node = of_node_get(np);
 	const __be32 *ranges = NULL;
drivers/of/device.c +1 -1
index 27203bfd0b22..ef6a741f9f0b 100644
@@ -95,7 +95,7 @@ int of_dma_configure(struct device *dev, struct device_node *np, bool force_dma)
 	const struct iommu_ops *iommu;
 	u64 mask, end;
 
-	ret = of_dma_get_range(np, &dma_addr, &paddr, &size);
+	ret = of_dma_get_range(dev, np, &dma_addr, &paddr, &size);
 	if (ret < 0) {
 		/*
 		 * For legacy reasons, we have to assume some devices need
drivers/of/of_private.h +4 -4
index edc682249c00..7a83d3a81ab6 100644
@@ -158,11 +158,11 @@ extern int of_bus_n_addr_cells(struct device_node *np);
 extern int of_bus_n_size_cells(struct device_node *np);
 
 #ifdef CONFIG_OF_ADDRESS
-extern int of_dma_get_range(struct device_node *np, u64 *dma_addr,
-			    u64 *paddr, u64 *size);
+extern int of_dma_get_range(struct device *dev, struct device_node *np,
+			    u64 *dma_addr, u64 *paddr, u64 *size);
 #else
-static inline int of_dma_get_range(struct device_node *np, u64 *dma_addr,
-				   u64 *paddr, u64 *size)
+static inline int of_dma_get_range(struct device *dev, struct device_node *np,
+				   u64 *dma_addr, u64 *paddr, u64 *size)
 {
 	return -ENODEV;
 }
drivers/of/unittest.c +1 -1
index 398de04fd19c..57a22517c428 100644
@@ -881,7 +881,7 @@ static void __init of_unittest_dma_ranges_one(const char *path,
 		return;
 	}
 
-	rc = of_dma_get_range(np, &dma_addr, &paddr, &size);
+	rc = of_dma_get_range(NULL, np, &dma_addr, &paddr, &size);
 
 	unittest(!rc, "of_dma_get_range failed on node %pOF rc=%i\n", np, rc);
 	if (!rc) {
Signature · 1 line
2.17.1
#14 lore raw

PATCH v3 9/13 device core: Introduce multiple dma pfn offsets

In reply to Jim Quinlan

The new field in struct device 'dma_pfn_offset_map' is used to facilitate the use of multiple pfn offsets between cpu addrs and dma addrs. It subsumes the role of dev->dma_pfn_offset -- a uniform offset -- and designates the single offset a special case.

of_dma_configure() is the typical manner to set pfn offsets but there are a number of ad hoc assignments to dev->dma_pfn_offset in the kernel code. These cases now invoke the function attach_uniform_dma_pfn_offset(dev, pfn_offset).

---
 arch/arm/include/asm/dma-mapping.h            |  9 +-
 arch/arm/mach-keystone/keystone.c             |  9 +-
 arch/sh/drivers/pci/pcie-sh7786.c             |  3 +-
 arch/sh/kernel/dma-coherent.c                 | 17 ++--
 arch/x86/pci/sta2x11-fixup.c                  |  7 +-
 drivers/acpi/arm64/iort.c                     |  5 +-
 drivers/gpu/drm/sun4i/sun4i_backend.c         |  7 +-
 drivers/iommu/io-pgtable-arm.c                |  2 +-
 .../platform/sunxi/sun4i-csi/sun4i_csi.c      |  5 +-
 .../platform/sunxi/sun6i-csi/sun6i_csi.c      |  5 +-
 drivers/of/address.c                          | 93 +++++++++++++++++--
 drivers/of/device.c                           |  8 +-
 drivers/remoteproc/remoteproc_core.c          |  2 +-
 .../staging/media/sunxi/cedrus/cedrus_hw.c    |  7 +-
 drivers/usb/core/message.c                    |  4 +-
 drivers/usb/core/usb.c                        |  2 +-
 include/linux/device.h                        |  4 +-
 include/linux/dma-direct.h                    | 16 +++-
 include/linux/dma-mapping.h                   | 45 +++++++++
 kernel/dma/coherent.c                         | 11 ++-
 20 files changed, 210 insertions(+), 51 deletions(-)
arch/arm/include/asm/dma-mapping.h +5 -4
index bdd80ddbca34..f1e72f99468b 100644
@@ -35,8 +35,9 @@ static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus)
 #ifndef __arch_pfn_to_dma
 static inline dma_addr_t pfn_to_dma(struct device *dev, unsigned long pfn)
 {
-	if (dev)
-		pfn -= dev->dma_pfn_offset;
+	if (dev && dev->dma_pfn_offset_map)
+		pfn -= dma_pfn_offset_from_phys_addr(dev, PFN_PHYS(pfn));
+
 	return (dma_addr_t)__pfn_to_bus(pfn);
 }
 
@@ -44,8 +45,8 @@ static inline unsigned long dma_to_pfn(struct device *dev, dma_addr_t addr)
 {
 	unsigned long pfn = __bus_to_pfn(addr);
 
-	if (dev)
-		pfn += dev->dma_pfn_offset;
+	if (dev && dev->dma_pfn_offset_map)
+		pfn += dma_pfn_offset_from_dma_addr(dev, addr);
 
 	return pfn;
 }
arch/arm/mach-keystone/keystone.c +6 -3
index 638808c4e122..e7d3ee6e9cb5 100644
@@ -8,6 +8,7 @@
  */
 #include <linux/io.h>
 #include <linux/of.h>
+#include <linux/dma-mapping.h>
 #include <linux/init.h>
 #include <linux/of_platform.h>
 #include <linux/of_address.h>
@@ -38,9 +39,11 @@ static int keystone_platform_notifier(struct notifier_block *nb,
 		return NOTIFY_BAD;
 
 	if (!dev->of_node) {
-		dev->dma_pfn_offset = keystone_dma_pfn_offset;
-		dev_err(dev, "set dma_pfn_offset%08lx\n",
-			dev->dma_pfn_offset);
+		int ret = attach_uniform_dma_pfn_offset
+				(dev, keystone_dma_pfn_offset);
+
+		dev_err(dev, "set dma_pfn_offset%08lx%s\n",
+			dev->dma_pfn_offset, ret ? " failed" : "");
 	}
 	return NOTIFY_OK;
 }
arch/sh/drivers/pci/pcie-sh7786.c +2 -1
index e0b568aaa701..2e832a5c58c1 100644
@@ -12,6 +12,7 @@
 #include <linux/io.h>
 #include <linux/async.h>
 #include <linux/delay.h>
+#include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/clk.h>
 #include <linux/sh_clk.h>
@@ -487,7 +488,7 @@ int pcibios_map_platform_irq(const struct pci_dev *pdev, u8 slot, u8 pin)
 
 void pcibios_bus_add_device(struct pci_dev *pdev)
 {
-	pdev->dev.dma_pfn_offset = dma_pfn_offset;
+	attach_uniform_dma_pfn_offset(&pdev->dev, dma_pfn_offset);
 }
 
 static int __init sh7786_pcie_core_init(void)
arch/sh/kernel/dma-coherent.c +11 -6
index d4811691b93c..5fc9e358b6c7 100644
@@ -14,6 +14,8 @@ void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
 {
 	void *ret, *ret_nocache;
 	int order = get_order(size);
+	unsigned long pfn;
+	phys_addr_t phys;
 
 	gfp |= __GFP_ZERO;
 
@@ -34,11 +36,14 @@ void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
 		return NULL;
 	}
 
-	split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
+	phys = virt_to_phys(ret);
+	pfn =  phys >> PAGE_SHIFT;
+	split_page(pfn_to_page(pfn), order);
 
-	*dma_handle = virt_to_phys(ret);
-	if (!WARN_ON(!dev))
-		*dma_handle -= PFN_PHYS(dev->dma_pfn_offset);
+	*dma_handle = (dma_addr_t)phys;
+	if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
+		*dma_handle -= PFN_PHYS(
+			dma_pfn_offset_from_phys_addr(dev, phys));
 
 	return ret_nocache;
 }
@@ -50,8 +55,8 @@ void arch_dma_free(struct device *dev, size_t size, void *vaddr,
 	unsigned long pfn = (dma_handle >> PAGE_SHIFT);
 	int k;
 
-	if (!WARN_ON(!dev))
-		pfn += dev->dma_pfn_offset;
+	if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
+		pfn += dma_pfn_offset_from_dma_addr(dev, dma_handle);
 
 	for (k = 0; k < (1 << order); k++)
 		__free_pages(pfn_to_page(pfn + k), 0);
arch/x86/pci/sta2x11-fixup.c +5 -2
index c313d784efab..4cdeca9f69b6 100644
@@ -12,6 +12,7 @@
 #include <linux/export.h>
 #include <linux/list.h>
 #include <linux/dma-direct.h>
+#include <linux/dma-mapping.h>
 #include <asm/iommu.h>
 
 #define STA2X11_SWIOTLB_SIZE (4*1024*1024)
@@ -133,7 +134,7 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
 	struct sta2x11_instance *instance = sta2x11_pdev_to_instance(pdev);
 	struct device *dev = &pdev->dev;
 	u32 amba_base, max_amba_addr;
-	int i;
+	int i, ret;
 
 	if (!instance)
 		return;
@@ -141,7 +142,9 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
 	pci_read_config_dword(pdev, AHB_BASE(0), &amba_base);
 	max_amba_addr = amba_base + STA2X11_AMBA_SIZE - 1;
 
-	dev->dma_pfn_offset = PFN_DOWN(-amba_base);
+	ret = attach_uniform_dma_pfn_offset(dev, PFN_DOWN(-amba_base));
+	if (ret)
+		dev_err(dev, "sta2x11: could not set PFN offset\n");
 
 	dev->bus_dma_limit = max_amba_addr;
 	pci_set_consistent_dma_mask(pdev, max_amba_addr);
drivers/acpi/arm64/iort.c +3 -2
index 28a6b387e80e..153661ddc74b 100644
@@ -1142,8 +1142,9 @@ void iort_dma_setup(struct device *dev, u64 *dma_addr, u64 *dma_size)
 	*dma_addr = dmaaddr;
 	*dma_size = size;
 
-	dev->dma_pfn_offset = PFN_DOWN(offset);
-	dev_dbg(dev, "dma_pfn_offset(%#08llx)\n", offset);
+	ret = attach_uniform_dma_pfn_offset(dev, PFN_DOWN(offset));
+	dev_dbg(dev, "dma_pfn_offset(%#08llx)%s\n",
+		offset, ret ? " failed!" : "");
 }
 
 static void __init acpi_iort_register_irq(int hwirq, const char *name,
drivers/gpu/drm/sun4i/sun4i_backend.c +5 -2
index 072ea113e6be..3d41dfc7d178 100644
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/of_device.h>
 #include <linux/of_graph.h>
+#include <linux/dma-mapping.h>
 #include <linux/platform_device.h>
 #include <linux/reset.h>
 
@@ -786,7 +787,7 @@ static int sun4i_backend_bind(struct device *dev, struct device *master,
 	const struct sun4i_backend_quirks *quirks;
 	struct resource *res;
 	void __iomem *regs;
-	int i, ret;
+	int i, ret = 0;
 
 	backend = devm_kzalloc(dev, sizeof(*backend), GFP_KERNEL);
 	if (!backend)
@@ -812,7 +813,9 @@ static int sun4i_backend_bind(struct device *dev, struct device *master,
 		 * on our device since the RAM mapping is at 0 for the DMA bus,
 		 * unlike the CPU.
 		 */
-		drm->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
 	}
 
 	backend->engine.node = dev->of_node;
drivers/iommu/io-pgtable-arm.c +1 -1
index 04fbd4bf0ff9..e9cc1c2d47cd 100644
@@ -754,7 +754,7 @@ arm_lpae_alloc_pgtable(struct io_pgtable_cfg *cfg)
 	if (cfg->oas > ARM_LPAE_MAX_ADDR_BITS)
 		return NULL;
 
-	if (!selftest_running && cfg->iommu_dev->dma_pfn_offset) {
+	if (!selftest_running && cfg->iommu_dev->dma_pfn_offset_map) {
 		dev_err(cfg->iommu_dev, "Cannot accommodate DMA offset for IOMMU page tables\n");
 		return NULL;
 	}
drivers/media/platform/sunxi/sun4i-csi/sun4i_csi.c +4 -1
index eff34ded6305..7212da5e1076 100644
@@ -7,6 +7,7 @@
  */
 
 #include <linux/clk.h>
+#include <linux/dma-mapping.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
@@ -183,7 +184,9 @@ static int sun4i_csi_probe(struct platform_device *pdev)
 			return ret;
 	} else {
 #ifdef PHYS_PFN_OFFSET
-		csi->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
 #endif
 	}
 
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device *pdev)
 
 	sdev->dev = &pdev->dev;
 	/* The DMA bus has the memory mapped at 0 */
-	sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+	ret = attach_uniform_dma_pfn_offset(sdev->dev,
+					    PHYS_OFFSET >> PAGE_SHIFT);
+	if (ret)
+		return ret;
 
 	ret = sun6i_csi_resource_request(sdev, pdev);
 	if (ret)
drivers/of/address.c +86 -7

132 lines not on this page · raw message

drivers/of/device.c +1 -7
index ef6a741f9f0b..91c50f40a82e 100644
@@ -91,7 +91,6 @@ int of_dma_configure(struct device *dev, struct device_node *np, bool force_dma)
 	u64 dma_addr, paddr, size = 0;
 	int ret;
 	bool coherent;
-	unsigned long offset;
 	const struct iommu_ops *iommu;
 	u64 mask, end;
 
@@ -105,10 +104,8 @@ int of_dma_configure(struct device *dev, struct device_node *np, bool force_dma)
 		if (!force_dma)
 			return ret == -ENODEV ? 0 : ret;
 
-		dma_addr = offset = 0;
+		dma_addr = 0;
 	} else {
-		offset = PFN_DOWN(paddr - dma_addr);
-
 		/*
 		 * Add a work around to treat the size as mask + 1 in case
 		 * it is defined in DT as a mask.
@@ -123,7 +120,6 @@ int of_dma_configure(struct device *dev, struct device_node *np, bool force_dma)
 			dev_err(dev, "Adjusted size 0x%llx invalid\n", size);
 			return -EINVAL;
 		}
-		dev_dbg(dev, "dma_pfn_offset(%#08lx)\n", offset);
 	}
 
 	/*
@@ -142,8 +138,6 @@ int of_dma_configure(struct device *dev, struct device_node *np, bool force_dma)
 	else if (!size)
 		size = 1ULL << 32;
 
-	dev->dma_pfn_offset = offset;
-
 	/*
 	 * Limit coherent and dma mask based on size and default mask
 	 * set by the driver.
drivers/remoteproc/remoteproc_core.c +1 -1
index e12a54e67588..e49648f77261 100644
@@ -518,7 +518,7 @@ static int rproc_handle_vdev(struct rproc *rproc, struct fw_rsc_vdev *rsc,
 	/* Initialise vdev subdevice */
 	snprintf(name, sizeof(name), "vdev%dbuffer", rvdev->index);
 	rvdev->dev.parent = rproc->dev.parent;
-	rvdev->dev.dma_pfn_offset = rproc->dev.parent->dma_pfn_offset;
+	rvdev->dev.dma_pfn_offset_map = rproc->dev.parent->dma_pfn_offset_map;
 	rvdev->dev.release = rproc_rvdev_release;
 	dev_set_name(&rvdev->dev, "%s#%s", dev_name(rvdev->dev.parent), name);
 	dev_set_drvdata(&rvdev->dev, rvdev);
drivers/staging/media/sunxi/cedrus/cedrus_hw.c +5 -2
index daf5f244f93b..29217e6e4153 100644
@@ -171,8 +171,11 @@ int cedrus_hw_probe(struct cedrus_dev *dev)
 	 */
 
 #ifdef PHYS_PFN_OFFSET
-	if (!(variant->quirks & CEDRUS_QUIRK_NO_DMA_OFFSET))
-		dev->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+	if (!(variant->quirks & CEDRUS_QUIRK_NO_DMA_OFFSET)) {
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
+	}
 #endif
 
 	ret = of_reserved_mem_device_init(dev->dev);
drivers/usb/core/message.c +2 -2
index 6197938dcc2d..071856000428 100644
@@ -1956,10 +1956,10 @@ int usb_set_configuration(struct usb_device *dev, int configuration)
 		intf->dev.groups = usb_interface_groups;
 		/*
 		 * Please refer to usb_alloc_dev() to see why we set
-		 * dma_mask and dma_pfn_offset.
+		 * dma_mask and dma_pfn_offset_map.
 		 */
 		intf->dev.dma_mask = dev->dev.dma_mask;
-		intf->dev.dma_pfn_offset = dev->dev.dma_pfn_offset;
+		intf->dev.dma_pfn_offset_map = dev->dev.dma_pfn_offset_map;
 		INIT_WORK(&intf->reset_ws, __usb_queue_reset_device);
 		intf->minor = -1;
 		device_initialize(&intf->dev);
drivers/usb/core/usb.c +1 -1
index f16c26dc079d..3fbc0c06ce9c 100644
@@ -611,7 +611,7 @@ struct usb_device *usb_alloc_dev(struct usb_device *parent,
 	 * mask for the entire HCD, so don't do that.
 	 */
 	dev->dev.dma_mask = bus->sysdev->dma_mask;
-	dev->dev.dma_pfn_offset = bus->sysdev->dma_pfn_offset;
+	dev->dev.dma_pfn_offset_map = bus->sysdev->dma_pfn_offset_map;
 	set_dev_node(&dev->dev, dev_to_node(bus->sysdev));
 	dev->state = USB_STATE_ATTACHED;
 	dev->lpm_disable_count = 1;
include/linux/device.h +2 -2
index ac8e37cd716a..9c079643ff54 100644
@@ -492,7 +492,7 @@ struct dev_links_info {
  * 		such descriptors.
  * @bus_dma_limit: Limit of an upstream bridge or bus which imposes a smaller
  *		DMA limit than the device itself supports.
- * @dma_pfn_offset: offset of DMA memory range relatively of RAM
+ * @dma_pfn_offset_map: offset map for DMA memory range relatively of RAM
  * @dma_parms:	A low level driver may set these to teach IOMMU code about
  * 		segment limitations.
  * @dma_pools:	Dma pools (if dma'ble device).
@@ -577,7 +577,7 @@ struct device {
 					     64 bit addresses for consistent
 					     allocations such descriptors. */
 	u64		bus_dma_limit;	/* upstream dma constraint */
-	unsigned long	dma_pfn_offset;
+	struct dma_pfn_offset_region *dma_pfn_offset_map;
 
 	struct device_dma_parameters *dma_parms;
 
include/linux/dma-direct.h +14 -2
index 24b8684aa21d..3c3363a3925e 100644
@@ -15,14 +15,26 @@ static inline dma_addr_t __phys_to_dma(struct device *dev, phys_addr_t paddr)
 {
 	dma_addr_t dev_addr = (dma_addr_t)paddr;
 
-	return dev_addr - ((dma_addr_t)dev->dma_pfn_offset << PAGE_SHIFT);
+	if (dev->dma_pfn_offset_map) {
+		unsigned long dma_pfn_offset
+			= dma_pfn_offset_from_phys_addr(dev, paddr);
+
+		dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
+	}
+	return dev_addr;
 }
 
 static inline phys_addr_t __dma_to_phys(struct device *dev, dma_addr_t dev_addr)
 {
 	phys_addr_t paddr = (phys_addr_t)dev_addr;
 
-	return paddr + ((phys_addr_t)dev->dma_pfn_offset << PAGE_SHIFT);
+	if (dev->dma_pfn_offset_map) {
+		unsigned long dma_pfn_offset
+			= dma_pfn_offset_from_dma_addr(dev, dev_addr);
+
+		paddr += ((phys_addr_t)dma_pfn_offset << PAGE_SHIFT);
+	}
+	return paddr;
 }
 #endif /* !CONFIG_ARCH_HAS_PHYS_TO_DMA */
 
include/linux/dma-mapping.h +45 -0

57 lines not on this page · raw message

kernel/dma/coherent.c +7 -4
index 2a0c4985f38e..20fc48d0b9d7 100644
@@ -31,10 +31,13 @@ static inline struct dma_coherent_mem *dev_get_coherent_memory(struct device *de
 static inline dma_addr_t dma_get_device_base(struct device *dev,
 					     struct dma_coherent_mem * mem)
 {
-	if (mem->use_dev_dma_pfn_offset)
-		return (mem->pfn_base - dev->dma_pfn_offset) << PAGE_SHIFT;
-	else
-		return mem->device_base;
+	if (mem->use_dev_dma_pfn_offset && dev->dma_pfn_offset_map) {
+		unsigned long dma_pfn_offset = dma_pfn_offset_from_phys_addr
+			(dev, PFN_PHYS(mem->pfn_base));
+
+		return (mem->pfn_base - dma_pfn_offset) << PAGE_SHIFT;
+	}
+	return mem->device_base;
 }
 
 static int dma_init_coherent_memory(phys_addr_t phys_addr,
Signature · 6 lines
2.17.1

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

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

drivers/gpu/drm/sun4i/sun4i_backend.c static int sun4i_backend_bind(struct device *dev, struct device *master, · line 790

Quoted 6 lines · Jim Quinlan
drivers/gpu/drm/sun4i/sun4i_backend.c +1 -1
@@ -786,7 +787,7 @@ static int sun4i_backend_bind(struct device *dev, struct device *master,
 	const struct sun4i_backend_quirks *quirks;
 	struct resource *res;
 	void __iomem *regs;
-	int i, ret;
+	int i, ret = 0;

No need for this.

Quoted 102 lines
drivers/gpu/drm/sun4i/sun4i_backend.c +3 -1
 	backend = devm_kzalloc(dev, sizeof(*backend), GFP_KERNEL);
 	if (!backend)
@@ -812,7 +813,9 @@ static int sun4i_backend_bind(struct device *dev, struct device *master,
 		 * on our device since the RAM mapping is at 0 for the DMA bus,
 		 * unlike the CPU.
 		 */
-		drm->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
 	}
 
 	backend->engine.node = dev->of_node;
drivers/iommu/io-pgtable-arm.c +1 -1
index 04fbd4bf0ff9..e9cc1c2d47cd 100644
@@ -754,7 +754,7 @@ arm_lpae_alloc_pgtable(struct io_pgtable_cfg *cfg)
 	if (cfg->oas > ARM_LPAE_MAX_ADDR_BITS)
 		return NULL;
 
-	if (!selftest_running && cfg->iommu_dev->dma_pfn_offset) {
+	if (!selftest_running && cfg->iommu_dev->dma_pfn_offset_map) {
 		dev_err(cfg->iommu_dev, "Cannot accommodate DMA offset for IOMMU page tables\n");
 		return NULL;
 	}
drivers/media/platform/sunxi/sun4i-csi/sun4i_csi.c +4 -1
index eff34ded6305..7212da5e1076 100644
@@ -7,6 +7,7 @@
  */
 
 #include <linux/clk.h>
+#include <linux/dma-mapping.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
@@ -183,7 +184,9 @@ static int sun4i_csi_probe(struct platform_device *pdev)
 			return ret;
 	} else {
 #ifdef PHYS_PFN_OFFSET
-		csi->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
 #endif
 	}
 
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device *pdev)
 
 	sdev->dev = &pdev->dev;
 	/* The DMA bus has the memory mapped at 0 */
-	sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+	ret = attach_uniform_dma_pfn_offset(sdev->dev,
+					    PHYS_OFFSET >> PAGE_SHIFT);
+	if (ret)
+		return ret;
 
 	ret = sun6i_csi_resource_request(sdev, pdev);
 	if (ret)
drivers/of/address.c +28 -0
index 96d8cfb14a60..c89333b0a5fb 100644
@@ -918,6 +918,70 @@ void __iomem *of_io_request_and_map(struct device_node *np, int index,
 }
 EXPORT_SYMBOL(of_io_request_and_map);
 
+static int attach_dma_pfn_offset_map(struct device *dev,
+				     struct device_node *node, int num_ranges)
+{
+	struct of_range_parser parser;
+	struct of_range range;
+	struct dma_pfn_offset_region *r;
+
+	r = devm_kcalloc(dev, num_ranges + 1,
+			 sizeof(struct dma_pfn_offset_region), GFP_KERNEL);
+	if (!r)
+		return -ENOMEM;
+	dev->dma_pfn_offset_map = r;
+	of_dma_range_parser_init(&parser, node);
+
+	/*
+	 * Record all info for DMA ranges array.  We could
+	 * just use the of_range struct, but if we did that it
+	 * would require more calculations for phys_to_dma and
+	 * dma_to_phys conversions.
+	 */
+	for_each_of_range(&parser, &range) {
+		r->cpu_start = range.cpu_addr;
+		r->cpu_end = r->cpu_start + range.size - 1;
+		r->dma_start = range.bus_addr;
+		r->dma_end = r->dma_start + range.size - 1;
+		r->pfn_offset = PFN_DOWN(range.cpu_addr)
+			- PFN_DOWN(range.bus_addr);
+		r++;

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 958

Quoted 10 lines
drivers/of/address.c +10 -0
+	}
+	return 0;
+}
+
+
+
+/**
+ * attach_dma_pfn_offset - Assign scalar offset for all addresses.
+ * @dev:	device pointer; only needed for a corner case.
+ * @dma_pfn_offset:	offset to apply when converting from phys addr
      ^^^^^^^^^^^^^^^

This parameter name does not match.

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 961

Quoted 3 lines
drivers/of/address.c +3 -0
+ *			to dma addr and vice versa.
+ *
+ * It returns -ENOMEM if out of memory, otherwise 0.

It can also return -ENODEV. Why are we passing NULL dev pointers to all these functions anyway?

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 974

Quoted 13 lines
drivers/of/address.c +13 -0
+ */
+int attach_uniform_dma_pfn_offset(struct device *dev, unsigned long pfn_offset)
+{
+	struct dma_pfn_offset_region *r;
+
+	if (!dev)
+		return -ENODEV;
+
+	if (!pfn_offset)
+		return 0;
+
+	r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+			 GFP_KERNEL);

Use: r = devm_kzalloc(dev, sizeof(*r), GFP_KERNEL);

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 982

Quoted 8 lines
drivers/of/address.c +8 -0
+	if (!r)
+		return -ENOMEM;
+
+	r->uniform_offset = true;
+	r->pfn_offset = pfn_offset;
+
+	return 0;
+}

This function doesn't seem to do anything useful. Is part of it missing?

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 984

Quoted 2 lines
drivers/of/address.c +2 -0
+EXPORT_SYMBOL_GPL(attach_uniform_dma_pfn_offset);
+

regards,
dan carpenter

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#16 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Hi Dan,

Quoted 10 lines · Dan Carpenter

drivers/gpu/drm/sun4i/sun4i_backend.c static int sun4i_backend_bind(struct device *dev, struct device *master, · line 790

Quoted 6 lines · Jim Quinlan
drivers/gpu/drm/sun4i/sun4i_backend.c +1 -1
@@ -786,7 +787,7 @@ static int sun4i_backend_bind(struct device *dev, struct device *master,
      const struct sun4i_backend_quirks *quirks;
      struct resource *res;
      void __iomem *regs;
-     int i, ret;
+     int i, ret = 0;

No need for this.

Will fix.

Quoted 114 lines
Quoted 103 lines
drivers/gpu/drm/sun4i/sun4i_backend.c +3 -1
      backend = devm_kzalloc(dev, sizeof(*backend), GFP_KERNEL);
      if (!backend)
@@ -812,7 +813,9 @@ static int sun4i_backend_bind(struct device *dev, struct device *master,
               * on our device since the RAM mapping is at 0 for the DMA bus,
               * unlike the CPU.
               */
-             drm->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+             ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+             if (ret)
+                     return ret;
      }
 
      backend->engine.node = dev->of_node;
drivers/iommu/io-pgtable-arm.c +1 -1
index 04fbd4bf0ff9..e9cc1c2d47cd 100644
@@ -754,7 +754,7 @@ arm_lpae_alloc_pgtable(struct io_pgtable_cfg *cfg)
      if (cfg->oas > ARM_LPAE_MAX_ADDR_BITS)
              return NULL;
 
-     if (!selftest_running && cfg->iommu_dev->dma_pfn_offset) {
+     if (!selftest_running && cfg->iommu_dev->dma_pfn_offset_map) {
              dev_err(cfg->iommu_dev, "Cannot accommodate DMA offset for IOMMU page tables\n");
              return NULL;
      }
drivers/media/platform/sunxi/sun4i-csi/sun4i_csi.c +4 -1
index eff34ded6305..7212da5e1076 100644
@@ -7,6 +7,7 @@
  */
 
 #include <linux/clk.h>
+#include <linux/dma-mapping.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
@@ -183,7 +184,9 @@ static int sun4i_csi_probe(struct platform_device *pdev)
                      return ret;
      } else {
 #ifdef PHYS_PFN_OFFSET
-             csi->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+             ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+             if (ret)
+                     return ret;
 #endif
      }
 
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device *pdev)
 
      sdev->dev = &pdev->dev;
      /* The DMA bus has the memory mapped at 0 */
-     sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+     ret = attach_uniform_dma_pfn_offset(sdev->dev,
+                                         PHYS_OFFSET >> PAGE_SHIFT);
+     if (ret)
+             return ret;
 
      ret = sun6i_csi_resource_request(sdev, pdev);
      if (ret)
drivers/of/address.c +28 -0
index 96d8cfb14a60..c89333b0a5fb 100644
@@ -918,6 +918,70 @@ void __iomem *of_io_request_and_map(struct device_node *np, int index,
 }
 EXPORT_SYMBOL(of_io_request_and_map);
 
+static int attach_dma_pfn_offset_map(struct device *dev,
+                                  struct device_node *node, int num_ranges)
+{
+     struct of_range_parser parser;
+     struct of_range range;
+     struct dma_pfn_offset_region *r;
+
+     r = devm_kcalloc(dev, num_ranges + 1,
+                      sizeof(struct dma_pfn_offset_region), GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;
+     dev->dma_pfn_offset_map = r;
+     of_dma_range_parser_init(&parser, node);
+
+     /*
+      * Record all info for DMA ranges array.  We could
+      * just use the of_range struct, but if we did that it
+      * would require more calculations for phys_to_dma and
+      * dma_to_phys conversions.
+      */
+     for_each_of_range(&parser, &range) {
+             r->cpu_start = range.cpu_addr;
+             r->cpu_end = r->cpu_start + range.size - 1;
+             r->dma_start = range.bus_addr;
+             r->dma_end = r->dma_start + range.size - 1;
+             r->pfn_offset = PFN_DOWN(range.cpu_addr)
+                     - PFN_DOWN(range.bus_addr);
+             r++;

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 958

Quoted 10 lines
drivers/of/address.c +10 -0
+     }
+     return 0;
+}
+
+
+
+/**
+ * attach_dma_pfn_offset - Assign scalar offset for all addresses.
+ * @dev:     device pointer; only needed for a corner case.
+ * @dma_pfn_offset:  offset to apply when converting from phys addr
      ^^^^^^^^^^^^^^^
Quoted 1 line

This parameter name does not match.

Will fix.

Quoted 7 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 961

Quoted 4 lines
drivers/of/address.c +3 -0
+ *                   to dma addr and vice versa.
+ *
+ * It returns -ENOMEM if out of memory, otherwise 0.

It can also return -ENODEV. Why are we passing NULL dev pointers to all these functions anyway?

No one should be passing dev==NULL to these functions -- that is why I have the error out for this case. Actually, there is a case of dev==NULL being passed -- drivers/of/unittest.c makes a call with a NULL device because it has no dev in its context. I wasn't sure how to fix this.

Quoted 16 lines
Quoted 4 lines
drivers/of/address.c +3 -0
+ */
+int attach_uniform_dma_pfn_offset(struct device *dev, unsigned long pfn_offset)
+{

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 974

Quoted 10 lines
drivers/of/address.c +10 -0
+     struct dma_pfn_offset_region *r;
+
+     if (!dev)
+             return -ENODEV;
+
+     if (!pfn_offset)
+             return 0;
+
+     r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+                      GFP_KERNEL);

Use: r = devm_kzalloc(dev, sizeof(*r), GFP_KERNEL);

Will fix.

Quoted 13 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 982

Quoted 10 lines
drivers/of/address.c +8 -0
+     if (!r)
+             return -ENOMEM;
+
+     r->uniform_offset = true;
+     r->pfn_offset = pfn_offset;
+
+     return 0;
+}

This function doesn't seem to do anything useful. Is part of it missing?

No, the uniform pfn offset is a special case. With it, there is only one entry in the "map" and 'uniform_offset' is set to true and the pfn_offset is also set. When the offset is to be computed, there are no region bounds calculations needed -- the offset of the first entry is returned immediately. This special case was intentional to (a) preserve backwards compatibility to those using dev->dma_pfn_offset and (b) it is faster than treating it as the general case -- I want to minimize the execution cost of this case. If this is deemed unnecessary, I can remove the boolean and threat all cases in the same way.

The other case is where multiple pfn offsets are needed for multiple regions. In this case the code checks the bounds for each successive region, returning an offset if there is a match.

Quoted 7 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 984

Quoted 3 lines
drivers/of/address.c +2 -0
+EXPORT_SYMBOL_GPL(attach_uniform_dma_pfn_offset);
+

regards,
dan carpenter

Thanks!
Jim Quinlan
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

#17 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Quoted 19 lines · Jim Quinlan
Quoted 4 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 974

Quoted 2 lines
drivers/of/address.c +2 -0
+     r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+                      GFP_KERNEL);

Use: r = devm_kzalloc(dev, sizeof(*r), GFP_KERNEL);

Will fix.

Quoted 13 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 982

Quoted 10 lines
drivers/of/address.c +8 -0
+     if (!r)
+             return -ENOMEM;
+
+     r->uniform_offset = true;
+     r->pfn_offset = pfn_offset;
+
+     return 0;
+}

This function doesn't seem to do anything useful. Is part of it missing?

Quoted 1 line

No, the uniform pfn offset is a special case.

Sorry, I wasn't clear. We're talking about different things. The code does:

	r = devm_kzalloc(dev, sizeof(*r), GFP_KERNEL);
	if (!r)
		return -ENOMEM;

	r->uniform_offset = true;
	r->pfn_offset = pfn_offset;

	return 0;

The code allocates "r" and then doesn't save it anywhere so there is no point.

regards,
dan carpenter

_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel

#18 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Quoted 35 lines · Dan Carpenter
Quoted 19 lines · Jim Quinlan
Quoted 4 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 974

Quoted 2 lines
drivers/of/address.c +2 -0
+     r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+                      GFP_KERNEL);

Use: r = devm_kzalloc(dev, sizeof(*r), GFP_KERNEL);

Will fix.

Quoted 13 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 982

Quoted 10 lines
drivers/of/address.c +8 -0
+     if (!r)
+             return -ENOMEM;
+
+     r->uniform_offset = true;
+     r->pfn_offset = pfn_offset;
+
+     return 0;
+}

This function doesn't seem to do anything useful. Is part of it missing?

Quoted 1 line

No, the uniform pfn offset is a special case.

Sorry, I wasn't clear. We're talking about different things. The code does:

        r = devm_kzalloc(dev, sizeof(*r), GFP_KERNEL);
        if (!r)
                return -ENOMEM;

        r->uniform_offset = true;
        r->pfn_offset = pfn_offset;

        return 0;
Quoted 2 lines

The code allocates "r" and then doesn't save it anywhere so there is no point.

You are absolutely right, sorry I missed your point. Will fix.

Thanks,
Jim Quinlan

Quoted 4 lines

regards,
dan carpenter

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

#19 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

In reply to Jim Quinlan

Hi Jim,

Quoted 40 lines · Jim Quinlan

The new field in struct device 'dma_pfn_offset_map' is used to facilitate the use of multiple pfn offsets between cpu addrs and dma addrs. It subsumes the role of dev->dma_pfn_offset -- a uniform offset -- and designates the single offset a special case.

of_dma_configure() is the typical manner to set pfn offsets but there are a number of ad hoc assignments to dev->dma_pfn_offset in the kernel code. These cases now invoke the function attach_uniform_dma_pfn_offset(dev, pfn_offset).

---
 arch/arm/include/asm/dma-mapping.h            |  9 +-
 arch/arm/mach-keystone/keystone.c             |  9 +-
 arch/sh/drivers/pci/pcie-sh7786.c             |  3 +-
 arch/sh/kernel/dma-coherent.c                 | 17 ++--
 arch/x86/pci/sta2x11-fixup.c                  |  7 +-
 drivers/acpi/arm64/iort.c                     |  5 +-
 drivers/gpu/drm/sun4i/sun4i_backend.c         |  7 +-
 drivers/iommu/io-pgtable-arm.c                |  2 +-
 .../platform/sunxi/sun4i-csi/sun4i_csi.c      |  5 +-
 .../platform/sunxi/sun6i-csi/sun6i_csi.c      |  5 +-
 drivers/of/address.c                          | 93 +++++++++++++++++--
 drivers/of/device.c                           |  8 +-
 drivers/remoteproc/remoteproc_core.c          |  2 +-
 .../staging/media/sunxi/cedrus/cedrus_hw.c    |  7 +-
 drivers/usb/core/message.c                    |  4 +-
 drivers/usb/core/usb.c                        |  2 +-
 include/linux/device.h                        |  4 +-
 include/linux/dma-direct.h                    | 16 +++-
 include/linux/dma-mapping.h                   | 45 +++++++++
 kernel/dma/coherent.c                         | 11 ++-
 20 files changed, 210 insertions(+), 51 deletions(-)
arch/arm/include/asm/dma- renamed from arch/arm/include/asm/dma-mapping.h +0 -0

mapping.h
index bdd80ddbca34..f1e72f99468b 100644

arch/arm/include/asm/dma-mapping.h static inline const struct dma_map_ops · line 39

Quoted 7 lines
arch/arm/include/asm/dma-mapping.h +1 -2
@@ -35,8 +35,9 @@ static inline const struct dma_map_ops
 *get_arch_dma_ops(struct bus_type *bus)
 #ifndef __arch_pfn_to_dma
 static inline dma_addr_t pfn_to_dma(struct device *dev, unsigned long pfn)
 {
-	if (dev)
-		pfn -= dev->dma_pfn_offset;
+	if (dev && dev->dma_pfn_offset_map)

Would it make sense to move the dev->dma_pfn_offset_map check into dma_pfn_offset_from_phys_addr() and return 0 if not available? Same for the opposite variant of the function. I think it'd make the code a little simpler on some of the use cases, and overall less error prone if anyone starts using the function elsewhere.

Quoted 84 lines
arch/arm/include/asm/dma-mapping.h +4 -2
+		pfn -= dma_pfn_offset_from_phys_addr(dev, PFN_PHYS(pfn));
+
 	return (dma_addr_t)__pfn_to_bus(pfn);
 }
@@ -44,8 +45,8 @@ static inline unsigned long dma_to_pfn(struct device *dev,
 dma_addr_t addr)
 {
 	unsigned long pfn = __bus_to_pfn(addr);
 
-	if (dev)
-		pfn += dev->dma_pfn_offset;
+	if (dev && dev->dma_pfn_offset_map)
+		pfn += dma_pfn_offset_from_dma_addr(dev, addr);
 
 	return pfn;

}

arch/arm/mach- renamed from arch/arm/mach-keystone/keystone.c +0 -0

keystone/keystone.c
index 638808c4e122..e7d3ee6e9cb5 100644

arch/arm/mach-keystone/keystone.c +6 -3
@@ -8,6 +8,7 @@
  */
 #include <linux/io.h>
 #include <linux/of.h>
+#include <linux/dma-mapping.h>
 #include <linux/init.h>
 #include <linux/of_platform.h>
 #include <linux/of_address.h>
@@ -38,9 +39,11 @@ static int keystone_platform_notifier(struct notifier_block
 *nb,
 		return NOTIFY_BAD;
 
 	if (!dev->of_node) {
-		dev->dma_pfn_offset = keystone_dma_pfn_offset;
-		dev_err(dev, "set dma_pfn_offset%08lx\n",
-			dev->dma_pfn_offset);
+		int ret = attach_uniform_dma_pfn_offset
+				(dev, keystone_dma_pfn_offset);
+
+		dev_err(dev, "set dma_pfn_offset%08lx%s\n",
+			dev->dma_pfn_offset, ret ? " failed" : "");
 	}
 	return NOTIFY_OK;

}

arch/sh/drivers/pci/pcie- renamed from arch/sh/drivers/pci/pcie-sh7786.c +0 -0

sh7786.c
index e0b568aaa701..2e832a5c58c1 100644

arch/sh/drivers/pci/pcie-sh7786.c +2 -1
@@ -12,6 +12,7 @@
 #include <linux/io.h>
 #include <linux/async.h>
 #include <linux/delay.h>
+#include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/clk.h>
 #include <linux/sh_clk.h>
@@ -487,7 +488,7 @@ int pcibios_map_platform_irq(const struct pci_dev *pdev,
 u8 slot, u8 pin)
 
 void pcibios_bus_add_device(struct pci_dev *pdev)
 {
-	pdev->dev.dma_pfn_offset = dma_pfn_offset;
+	attach_uniform_dma_pfn_offset(&pdev->dev, dma_pfn_offset);
 }
 

static int __init sh7786_pcie_core_init(void)

arch/sh/kernel/dma-coherent.c +2 -0
index d4811691b93c..5fc9e358b6c7 100644
@@ -14,6 +14,8 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
 {
 	void *ret, *ret_nocache;
 	int order = get_order(size);
+	unsigned long pfn;
+	phys_addr_t phys;
 
 	gfp |= __GFP_ZERO;

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, · line 41

Quoted 7 lines
arch/sh/kernel/dma-coherent.c +2 -1
@@ -34,11 +36,14 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
 		return NULL;
 	}
 
-	split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
+	phys = virt_to_phys(ret);
+	pfn =  phys >> PAGE_SHIFT;

nit: not sure it really pays off to have a pfn variable here.

Quoted 169 lines
arch/sh/kernel/dma-coherent.c +5 -3
+	split_page(pfn_to_page(pfn), order);
 
-	*dma_handle = virt_to_phys(ret);
-	if (!WARN_ON(!dev))
-		*dma_handle -= PFN_PHYS(dev->dma_pfn_offset);
+	*dma_handle = (dma_addr_t)phys;
+	if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
+		*dma_handle -= PFN_PHYS(
+			dma_pfn_offset_from_phys_addr(dev, phys));
 
 
arch/sh/kernel/dma-coherent.c +2 -2
 	return ret_nocache;
 }
@@ -50,8 +55,8 @@ void arch_dma_free(struct device *dev, size_t size, void
 *vaddr,
 	unsigned long pfn = (dma_handle >> PAGE_SHIFT);
 	int k;
 
-	if (!WARN_ON(!dev))
-		pfn += dev->dma_pfn_offset;
+	if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
+		pfn += dma_pfn_offset_from_dma_addr(dev, dma_handle);
 
 	for (k = 0; k < (1 << order); k++)
arch/sh/kernel/dma-coherent.c +0 -0
 		__free_pages(pfn_to_page(pfn + k), 0);
arch/x86/pci/sta2x11-fixup.c +5 -2
index c313d784efab..4cdeca9f69b6 100644
@@ -12,6 +12,7 @@
 #include <linux/export.h>
 #include <linux/list.h>
 #include <linux/dma-direct.h>
+#include <linux/dma-mapping.h>
 #include <asm/iommu.h>
 
 #define STA2X11_SWIOTLB_SIZE (4*1024*1024)
@@ -133,7 +134,7 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
 	struct sta2x11_instance *instance = sta2x11_pdev_to_instance(pdev);
 	struct device *dev = &pdev->dev;
 	u32 amba_base, max_amba_addr;
-	int i;
+	int i, ret;
 
 	if (!instance)
 		return;
@@ -141,7 +142,9 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
 	pci_read_config_dword(pdev, AHB_BASE(0), &amba_base);
 	max_amba_addr = amba_base + STA2X11_AMBA_SIZE - 1;
 
-	dev->dma_pfn_offset = PFN_DOWN(-amba_base);
+	ret = attach_uniform_dma_pfn_offset(dev, PFN_DOWN(-amba_base));
+	if (ret)
+		dev_err(dev, "sta2x11: could not set PFN offset\n");
 
 	dev->bus_dma_limit = max_amba_addr;
 	pci_set_consistent_dma_mask(pdev, max_amba_addr);
drivers/acpi/arm64/iort.c +3 -2
index 28a6b387e80e..153661ddc74b 100644
@@ -1142,8 +1142,9 @@ void iort_dma_setup(struct device *dev, u64 *dma_addr,
 u64 *dma_size)
 	*dma_addr = dmaaddr;
 	*dma_size = size;
 
-	dev->dma_pfn_offset = PFN_DOWN(offset);
-	dev_dbg(dev, "dma_pfn_offset(%#08llx)\n", offset);
+	ret = attach_uniform_dma_pfn_offset(dev, PFN_DOWN(offset));
+	dev_dbg(dev, "dma_pfn_offset(%#08llx)%s\n",
+		offset, ret ? " failed!" : "");
 }
 

static void __init acpi_iort_register_irq(int hwirq, const char *name,

drivers/gpu/drm/sun4i/sun4i_backend.c +2 -1
index 072ea113e6be..3d41dfc7d178 100644
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/of_device.h>
 #include <linux/of_graph.h>
+#include <linux/dma-mapping.h>
 #include <linux/platform_device.h>
 #include <linux/reset.h>
 
@@ -786,7 +787,7 @@ static int sun4i_backend_bind(struct device *dev, struct
 device *master,
 	const struct sun4i_backend_quirks *quirks;
 	struct resource *res;
 	void __iomem *regs;
-	int i, ret;
+	int i, ret = 0;
 
 	backend = devm_kzalloc(dev, sizeof(*backend), GFP_KERNEL);
drivers/gpu/drm/sun4i/sun4i_backend.c +3 -1
 	if (!backend)
@@ -812,7 +813,9 @@ static int sun4i_backend_bind(struct device *dev, struct
 device *master,
 		 * on our device since the RAM mapping is at 0 for the DMA bus,
 		 * unlike the CPU.
 		 */
-		drm->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
 	}
 
drivers/gpu/drm/sun4i/sun4i_backend.c +0 -0
 	backend->engine.node = dev->of_node;
drivers/iommu/io-pgtable-arm.c +1 -1
index 04fbd4bf0ff9..e9cc1c2d47cd 100644
@@ -754,7 +754,7 @@ arm_lpae_alloc_pgtable(struct io_pgtable_cfg *cfg)
 	if (cfg->oas > ARM_LPAE_MAX_ADDR_BITS)
 		return NULL;
 
-	if (!selftest_running && cfg->iommu_dev->dma_pfn_offset) {
+	if (!selftest_running && cfg->iommu_dev->dma_pfn_offset_map) {
 		dev_err(cfg->iommu_dev, "Cannot accommodate DMA offset for IOMMU
 page tables\n");
 		return NULL;
drivers/iommu/io-pgtable-arm.c +0 -0
 	}
drivers/media/platform/sunxi/sun4i-csi/sun4i_csi.c +4 -1
index eff34ded6305..7212da5e1076 100644
@@ -7,6 +7,7 @@
  */
 
 #include <linux/clk.h>
+#include <linux/dma-mapping.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
@@ -183,7 +184,9 @@ static int sun4i_csi_probe(struct platform_device *pdev)
 			return ret;
 	} else {
 #ifdef PHYS_PFN_OFFSET
-		csi->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+		ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+		if (ret)
+			return ret;
 #endif
 	}
 
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device *pdev)
 
 	sdev->dev = &pdev->dev;
 	/* The DMA bus has the memory mapped at 0 */
-	sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+	ret = attach_uniform_dma_pfn_offset(sdev->dev,
+					    PHYS_OFFSET >> PAGE_SHIFT);
+	if (ret)
+		return ret;
 
 	ret = sun6i_csi_resource_request(sdev, pdev);
 	if (ret)
drivers/of/address.c +0 -0
index 96d8cfb14a60..c89333b0a5fb 100644

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node · line 923

Quoted 6 lines
drivers/of/address.c +2 -0
@@ -918,6 +918,70 @@ void __iomem *of_io_request_and_map(struct device_node
 *np, int index,
 }
 EXPORT_SYMBOL(of_io_request_and_map);
 
+static int attach_dma_pfn_offset_map(struct device *dev,
+				     struct device_node *node, int num_ranges)

As with the previous review, please take this comment with a grain of salt.

I think there should be a clear split between what belongs to OF and what belongs to the core device infrastructure.

OF's job should be to parse DT and provide a list/array of ranges, whereas the core device infrastructure should provide an API to assign a list of ranges/offset to a device.

As a concrete example, you're forcing devices like the sta2x11 to build with OF support, which, being an Intel device, it's pretty odd. But I'm also thinking of how will all this fit once an ACPI device wants to use it.

Expanding on this idea, once you have a split between the OF's and device core roles, it transpires that of_dma_get_range()'s job should only be to provide the ranges in a device understandable structure and of_dma_configre()'s to actually assign the device's parameters. This would obsolete patch #7.

Quoted 25 lines
drivers/of/address.c +25 -0
+{
+	struct of_range_parser parser;
+	struct of_range range;
+	struct dma_pfn_offset_region *r;
+
+	r = devm_kcalloc(dev, num_ranges + 1,
+			 sizeof(struct dma_pfn_offset_region), GFP_KERNEL);
+	if (!r)
+		return -ENOMEM;
+	dev->dma_pfn_offset_map = r;
+	of_dma_range_parser_init(&parser, node);
+
+	/*
+	 * Record all info for DMA ranges array.  We could
+	 * just use the of_range struct, but if we did that it
+	 * would require more calculations for phys_to_dma and
+	 * dma_to_phys conversions.
+	 */
+	for_each_of_range(&parser, &range) {
+		r->cpu_start = range.cpu_addr;
+		r->cpu_end = r->cpu_start + range.size - 1;
+		r->dma_start = range.bus_addr;
+		r->dma_end = r->dma_start + range.size - 1;
+		r->pfn_offset = PFN_DOWN(range.cpu_addr)
+			- PFN_DOWN(range.bus_addr);

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 957

Quoted 10 lines
drivers/of/address.c +10 -0
+		r++;
+	}
+	return 0;
+}
+
+
+
+/**
+ * attach_dma_pfn_offset - Assign scalar offset for all addresses.
+ * @dev:	device pointer; only needed for a corner case.

That's a huge corner :P

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index,

Quoted 7 lines
drivers/of/address.c +6 -0
+ * @dma_pfn_offset:	offset to apply when converting from phys addr
+ *			to dma addr and vice versa.
+ *
+ * It returns -ENOMEM if out of memory, otherwise 0.
+ */
+int attach_uniform_dma_pfn_offset(struct device *dev, unsigned long
 pfn_offset)

As I say above, does this really belong to of/address.c?

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 976

Quoted 13 lines
drivers/of/address.c +13 -0
+{
+	struct dma_pfn_offset_region *r;
+
+	if (!dev)
+		return -ENODEV;
+
+	if (!pfn_offset)
+		return 0;
+
+	r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+			 GFP_KERNEL);
+	if (!r)
+		return -ENOMEM;

I think you're missing this:

	dev->dma_pfn_offset_map = r;
Quoted 23 lines
drivers/of/address.c +8 -0
+
+	r->uniform_offset = true;
+	r->pfn_offset = pfn_offset;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(attach_uniform_dma_pfn_offset);
+
 /**

* of_dma_get_range - Get DMA range info
* @dev: device pointer; only needed for a corner case.

drivers/of/address.c +1 -1
@@ -933,7 +997,7 @@ EXPORT_SYMBOL(of_io_request_and_map);
  *	CPU addr (phys_addr_t)	: pna cells
  *	size			: nsize cells
  *
- * It returns -ENODEV if "dma-ranges" property was not found
+ * It returns -ENODEV if !dev or "dma-ranges" property was not found
  * for this device in DT.
  */
 int of_dma_get_range(struct device *dev, struct device_node *np, u64

*dma_addr,

drivers/of/address.c +0 -0
@@ -946,7 +1010,13 @@ int of_dma_get_range(struct device *dev, struct
 device_node *np, u64 *dma_addr,

drivers/of/address.c int of_dma_get_range(struct device *dev, struct · line 1020

Quoted 10 lines
drivers/of/address.c +6 -0
 	bool found_dma_ranges = false;
 	struct of_range_parser parser;
 	struct of_range range;
+	phys_addr_t cpu_start = ~(phys_addr_t)0;
 	u64 dma_start = U64_MAX, dma_end = 0, dma_offset = 0;
+	bool dma_multi_pfn_offset = false;
+	int num_ranges = 0;
+
+	if (!dev)
+		return -ENODEV;

Shouldn't this be part of patch #7?

Regards,
Nicolas

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

#20 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Quoted 49 lines · Nicolas Saenz Julienne

Hi Jim,

Quoted 45 lines · Jim Quinlan

The new field in struct device 'dma_pfn_offset_map' is used to facilitate the use of multiple pfn offsets between cpu addrs and dma addrs. It subsumes the role of dev->dma_pfn_offset -- a uniform offset -- and designates the single offset a special case.

of_dma_configure() is the typical manner to set pfn offsets but there are a number of ad hoc assignments to dev->dma_pfn_offset in the kernel code. These cases now invoke the function attach_uniform_dma_pfn_offset(dev, pfn_offset).

---
 arch/arm/include/asm/dma-mapping.h            |  9 +-
 arch/arm/mach-keystone/keystone.c             |  9 +-
 arch/sh/drivers/pci/pcie-sh7786.c             |  3 +-
 arch/sh/kernel/dma-coherent.c                 | 17 ++--
 arch/x86/pci/sta2x11-fixup.c                  |  7 +-
 drivers/acpi/arm64/iort.c                     |  5 +-
 drivers/gpu/drm/sun4i/sun4i_backend.c         |  7 +-
 drivers/iommu/io-pgtable-arm.c                |  2 +-
 .../platform/sunxi/sun4i-csi/sun4i_csi.c      |  5 +-
 .../platform/sunxi/sun6i-csi/sun6i_csi.c      |  5 +-
 drivers/of/address.c                          | 93 +++++++++++++++++--
 drivers/of/device.c                           |  8 +-
 drivers/remoteproc/remoteproc_core.c          |  2 +-
 .../staging/media/sunxi/cedrus/cedrus_hw.c    |  7 +-
 drivers/usb/core/message.c                    |  4 +-
 drivers/usb/core/usb.c                        |  2 +-
 include/linux/device.h                        |  4 +-
 include/linux/dma-direct.h                    | 16 +++-
 include/linux/dma-mapping.h                   | 45 +++++++++
 kernel/dma/coherent.c                         | 11 ++-
 20 files changed, 210 insertions(+), 51 deletions(-)
arch/arm/include/asm/dma- renamed from arch/arm/include/asm/dma-mapping.h +0 -0

mapping.h
index bdd80ddbca34..f1e72f99468b 100644

arch/arm/include/asm/dma-mapping.h +0 -1
@@ -35,8 +35,9 @@ static inline const struct dma_map_ops
 *get_arch_dma_ops(struct bus_type *bus)
 #ifndef __arch_pfn_to_dma
 static inline dma_addr_t pfn_to_dma(struct device *dev, unsigned long pfn)
 {
-     if (dev)
Quoted 8 lines

arch/arm/include/asm/dma-mapping.h static inline const struct dma_map_ops *get_arch_dma_ops(struct bus_type *bus) · line 38

Quoted 2 lines
arch/arm/include/asm/dma-mapping.h +1 -1
-             pfn -= dev->dma_pfn_offset;
+     if (dev && dev->dma_pfn_offset_map)

Would it make sense to move the dev->dma_pfn_offset_map check into dma_pfn_offset_from_phys_addr() and return 0 if not available? Same for the opposite variant of the function. I think it'd make the code a little simpler on some of the use cases, and overall less error prone if anyone starts using the function elsewhere.

Yes it makes sense and I was debating doing it but I just wanted to make it explicit that there was not much cost for this change for the fastpath -- no dma_pfn_offset whatsoever -- as the cost goes from a "pfn += dev->dma_pfn_offset" to a "if (dev->dma_pfn_offset_map)". I will do what you suggest.

Quoted 89 lines
Quoted 89 lines
arch/arm/include/asm/dma-mapping.h +4 -2
+             pfn -= dma_pfn_offset_from_phys_addr(dev, PFN_PHYS(pfn));
+
      return (dma_addr_t)__pfn_to_bus(pfn);
 }
@@ -44,8 +45,8 @@ static inline unsigned long dma_to_pfn(struct device *dev,
 dma_addr_t addr)
 {
      unsigned long pfn = __bus_to_pfn(addr);
 
-     if (dev)
-             pfn += dev->dma_pfn_offset;
+     if (dev && dev->dma_pfn_offset_map)
+             pfn += dma_pfn_offset_from_dma_addr(dev, addr);
 
      return pfn;

}

arch/arm/mach- renamed from arch/arm/mach-keystone/keystone.c +0 -0

keystone/keystone.c
index 638808c4e122..e7d3ee6e9cb5 100644

arch/arm/mach-keystone/keystone.c +6 -3
@@ -8,6 +8,7 @@
  */
 #include <linux/io.h>
 #include <linux/of.h>
+#include <linux/dma-mapping.h>
 #include <linux/init.h>
 #include <linux/of_platform.h>
 #include <linux/of_address.h>
@@ -38,9 +39,11 @@ static int keystone_platform_notifier(struct notifier_block
 *nb,
              return NOTIFY_BAD;
 
      if (!dev->of_node) {
-             dev->dma_pfn_offset = keystone_dma_pfn_offset;
-             dev_err(dev, "set dma_pfn_offset%08lx\n",
-                     dev->dma_pfn_offset);
+             int ret = attach_uniform_dma_pfn_offset
+                             (dev, keystone_dma_pfn_offset);
+
+             dev_err(dev, "set dma_pfn_offset%08lx%s\n",
+                     dev->dma_pfn_offset, ret ? " failed" : "");
      }
      return NOTIFY_OK;

}

arch/sh/drivers/pci/pcie- renamed from arch/sh/drivers/pci/pcie-sh7786.c +0 -0

sh7786.c
index e0b568aaa701..2e832a5c58c1 100644

arch/sh/drivers/pci/pcie-sh7786.c +2 -1
@@ -12,6 +12,7 @@
 #include <linux/io.h>
 #include <linux/async.h>
 #include <linux/delay.h>
+#include <linux/dma-mapping.h>
 #include <linux/slab.h>
 #include <linux/clk.h>
 #include <linux/sh_clk.h>
@@ -487,7 +488,7 @@ int pcibios_map_platform_irq(const struct pci_dev *pdev,
 u8 slot, u8 pin)
 
 void pcibios_bus_add_device(struct pci_dev *pdev)
 {
-     pdev->dev.dma_pfn_offset = dma_pfn_offset;
+     attach_uniform_dma_pfn_offset(&pdev->dev, dma_pfn_offset);
 }
 

static int __init sh7786_pcie_core_init(void)

arch/sh/kernel/dma-coherent.c +2 -0
index d4811691b93c..5fc9e358b6c7 100644
@@ -14,6 +14,8 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
 {
      void *ret, *ret_nocache;
      int order = get_order(size);
+     unsigned long pfn;
+     phys_addr_t phys;
 
      gfp |= __GFP_ZERO;
@@ -34,11 +36,14 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
              return NULL;
      }
 
Quoted 5 lines

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle, · line 40

Quoted 3 lines
arch/sh/kernel/dma-coherent.c +2 -1
-     split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
+     phys = virt_to_phys(ret);
+     pfn =  phys >> PAGE_SHIFT;

nit: not sure it really pays off to have a pfn variable here.

Did it for readability; the compiler's optimization should take care of any extra variables. But I can switch if you insist.

Quoted 182 lines
Quoted 170 lines
arch/sh/kernel/dma-coherent.c +5 -3
+     split_page(pfn_to_page(pfn), order);
 
-     *dma_handle = virt_to_phys(ret);
-     if (!WARN_ON(!dev))
-             *dma_handle -= PFN_PHYS(dev->dma_pfn_offset);
+     *dma_handle = (dma_addr_t)phys;
+     if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
+             *dma_handle -= PFN_PHYS(
+                     dma_pfn_offset_from_phys_addr(dev, phys));
 
 
      return ret_nocache;

}

arch/sh/kernel/dma-coherent.c +2 -2
@@ -50,8 +55,8 @@ void arch_dma_free(struct device *dev, size_t size, void
 *vaddr,
      unsigned long pfn = (dma_handle >> PAGE_SHIFT);
      int k;
 
-     if (!WARN_ON(!dev))
-             pfn += dev->dma_pfn_offset;
+     if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
+             pfn += dma_pfn_offset_from_dma_addr(dev, dma_handle);
 
      for (k = 0; k < (1 << order); k++)
              __free_pages(pfn_to_page(pfn + k), 0);
arch/x86/pci/sta2x11-fixup.c +5 -2
index c313d784efab..4cdeca9f69b6 100644
@@ -12,6 +12,7 @@
 #include <linux/export.h>
 #include <linux/list.h>
 #include <linux/dma-direct.h>
+#include <linux/dma-mapping.h>
 #include <asm/iommu.h>
 
 #define STA2X11_SWIOTLB_SIZE (4*1024*1024)
@@ -133,7 +134,7 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
      struct sta2x11_instance *instance = sta2x11_pdev_to_instance(pdev);
      struct device *dev = &pdev->dev;
      u32 amba_base, max_amba_addr;
-     int i;
+     int i, ret;
 
      if (!instance)
              return;
@@ -141,7 +142,9 @@ static void sta2x11_map_ep(struct pci_dev *pdev)
      pci_read_config_dword(pdev, AHB_BASE(0), &amba_base);
      max_amba_addr = amba_base + STA2X11_AMBA_SIZE - 1;
 
-     dev->dma_pfn_offset = PFN_DOWN(-amba_base);
+     ret = attach_uniform_dma_pfn_offset(dev, PFN_DOWN(-amba_base));
+     if (ret)
+             dev_err(dev, "sta2x11: could not set PFN offset\n");
 
      dev->bus_dma_limit = max_amba_addr;
      pci_set_consistent_dma_mask(pdev, max_amba_addr);
drivers/acpi/arm64/iort.c +3 -2
index 28a6b387e80e..153661ddc74b 100644
@@ -1142,8 +1142,9 @@ void iort_dma_setup(struct device *dev, u64 *dma_addr,
 u64 *dma_size)
      *dma_addr = dmaaddr;
      *dma_size = size;
 
-     dev->dma_pfn_offset = PFN_DOWN(offset);
-     dev_dbg(dev, "dma_pfn_offset(%#08llx)\n", offset);
+     ret = attach_uniform_dma_pfn_offset(dev, PFN_DOWN(offset));
+     dev_dbg(dev, "dma_pfn_offset(%#08llx)%s\n",
+             offset, ret ? " failed!" : "");
 }
 

static void __init acpi_iort_register_irq(int hwirq, const char *name,

drivers/gpu/drm/sun4i/sun4i_backend.c +2 -1
index 072ea113e6be..3d41dfc7d178 100644
@@ -11,6 +11,7 @@
 #include <linux/module.h>
 #include <linux/of_device.h>
 #include <linux/of_graph.h>
+#include <linux/dma-mapping.h>
 #include <linux/platform_device.h>
 #include <linux/reset.h>
 
@@ -786,7 +787,7 @@ static int sun4i_backend_bind(struct device *dev, struct
 device *master,
      const struct sun4i_backend_quirks *quirks;
      struct resource *res;
      void __iomem *regs;
-     int i, ret;
+     int i, ret = 0;
 
      backend = devm_kzalloc(dev, sizeof(*backend), GFP_KERNEL);

if (!backend)

drivers/gpu/drm/sun4i/sun4i_backend.c +3 -1
@@ -812,7 +813,9 @@ static int sun4i_backend_bind(struct device *dev, struct
 device *master,
               * on our device since the RAM mapping is at 0 for the DMA bus,
               * unlike the CPU.
               */
-             drm->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+             ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+             if (ret)
+                     return ret;
      }
 
      backend->engine.node = dev->of_node;
drivers/iommu/io-pgtable-arm.c +1 -1
index 04fbd4bf0ff9..e9cc1c2d47cd 100644
@@ -754,7 +754,7 @@ arm_lpae_alloc_pgtable(struct io_pgtable_cfg *cfg)
      if (cfg->oas > ARM_LPAE_MAX_ADDR_BITS)
              return NULL;
 
-     if (!selftest_running && cfg->iommu_dev->dma_pfn_offset) {
+     if (!selftest_running && cfg->iommu_dev->dma_pfn_offset_map) {
              dev_err(cfg->iommu_dev, "Cannot accommodate DMA offset for IOMMU
 page tables\n");
              return NULL;
      }
drivers/media/platform/sunxi/sun4i-csi/sun4i_csi.c +4 -1
index eff34ded6305..7212da5e1076 100644
@@ -7,6 +7,7 @@
  */
 
 #include <linux/clk.h>
+#include <linux/dma-mapping.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
@@ -183,7 +184,9 @@ static int sun4i_csi_probe(struct platform_device *pdev)
                      return ret;
      } else {
 #ifdef PHYS_PFN_OFFSET
-             csi->dev->dma_pfn_offset = PHYS_PFN_OFFSET;
+             ret = attach_uniform_dma_pfn_offset(dev, PHYS_PFN_OFFSET);
+             if (ret)
+                     return ret;
 #endif
      }
 
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device *pdev)
 
      sdev->dev = &pdev->dev;
      /* The DMA bus has the memory mapped at 0 */
-     sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+     ret = attach_uniform_dma_pfn_offset(sdev->dev,
+                                         PHYS_OFFSET >> PAGE_SHIFT);
+     if (ret)
+             return ret;
 
      ret = sun6i_csi_resource_request(sdev, pdev);
      if (ret)
drivers/of/address.c +0 -0
index 96d8cfb14a60..c89333b0a5fb 100644

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node · line 923

Quoted 6 lines
drivers/of/address.c +2 -0
@@ -918,6 +918,70 @@ void __iomem *of_io_request_and_map(struct device_node
 *np, int index,
 }
 EXPORT_SYMBOL(of_io_request_and_map);
 
+static int attach_dma_pfn_offset_map(struct device *dev,
+                                  struct device_node *node, int num_ranges)

As with the previous review, please take this comment with a grain of salt.

I think there should be a clear split between what belongs to OF and what belongs to the core device infrastructure.

Quoted 7 lines

OF's job should be to parse DT and provide a list/array of ranges, whereas the core device infrastructure should provide an API to assign a list of ranges/offset to a device.

As a concrete example, you're forcing devices like the sta2x11 to build with OF support, which, being an Intel device, it's pretty odd. But I'm also thinking of how will all this fit once an ACPI device wants to use it.

To fix this I only have to move attach_uniform_dma_pfn_offset() from of/address.c to say include/linux/dma-mapping.h. It has no dependencies on OF. Do you agree?

Quoted 5 lines

Expanding on this idea, once you have a split between the OF's and device core roles, it transpires that of_dma_get_range()'s job should only be to provide the ranges in a device understandable structure and of_dma_configre()'s to actually assign the device's parameters. This would obsolete patch #7.

I think you mean patch #8. I agree with you. The reason I needed a "struct device *" in the call is because I wanted to make sure the memory that is alloc'd belongs to the device that needs it. If I do a regular kzalloc(), this memory will become a leak once someone starts unbinding/binding their device. Also, in all uses of of_dma_rtange() -- there is only one -- a dev is required as one can't attach an offset map to NULL.

I do see that there are a number of functions in drivers/of/*.c that take 'struct device *dev' as an argument so there is precedent for something like this. Regardless, I need an owner to the memory I alloc().

Quoted 30 lines
Quoted 30 lines
drivers/of/address.c +29 -0
+{
+     struct of_range_parser parser;
+     struct of_range range;
+     struct dma_pfn_offset_region *r;
+
+     r = devm_kcalloc(dev, num_ranges + 1,
+                      sizeof(struct dma_pfn_offset_region), GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;
+     dev->dma_pfn_offset_map = r;
+     of_dma_range_parser_init(&parser, node);
+
+     /*
+      * Record all info for DMA ranges array.  We could
+      * just use the of_range struct, but if we did that it
+      * would require more calculations for phys_to_dma and
+      * dma_to_phys conversions.
+      */
+     for_each_of_range(&parser, &range) {
+             r->cpu_start = range.cpu_addr;
+             r->cpu_end = r->cpu_start + range.size - 1;
+             r->dma_start = range.bus_addr;
+             r->dma_end = r->dma_start + range.size - 1;
+             r->pfn_offset = PFN_DOWN(range.cpu_addr)
+                     - PFN_DOWN(range.bus_addr);
+             r++;
+     }
+     return 0;
+}
Quoted 8 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 957

Quoted 6 lines
drivers/of/address.c +6 -0
+
+
+
+/**
+ * attach_dma_pfn_offset - Assign scalar offset for all addresses.
+ * @dev:     device pointer; only needed for a corner case.

That's a huge corner :P

Good point; I'm not really sure what percent of Linux configurations require a dma_pfn_offset. I'll drop the "corner".

Quoted 10 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index,

Quoted 8 lines
drivers/of/address.c +6 -0
+ * @dma_pfn_offset:  offset to apply when converting from phys addr
+ *                   to dma addr and vice versa.
+ *
+ * It returns -ENOMEM if out of memory, otherwise 0.
+ */
+int attach_uniform_dma_pfn_offset(struct device *dev, unsigned long
 pfn_offset)

As I say above, does this really belong to of/address.c?

No it does not. Will fix.

Quoted 17 lines
Quoted 4 lines
drivers/of/address.c +3 -0
+{
+     struct dma_pfn_offset_region *r;
+

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 976

Quoted 10 lines
drivers/of/address.c +10 -0
+     if (!dev)
+             return -ENODEV;
+
+     if (!pfn_offset)
+             return 0;
+
+     r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+                      GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;

I think you're missing this:

Quoted 2 lines
        dev->dma_pfn_offset_map = r;

That's a showstopper! DanC also pointed it out but I still didn't see it. Thanks!

Quoted 27 lines
Quoted 27 lines
drivers/of/address.c +8 -0
+
+     r->uniform_offset = true;
+     r->pfn_offset = pfn_offset;
+
+     return 0;
+}
+EXPORT_SYMBOL_GPL(attach_uniform_dma_pfn_offset);
+
 /**

* of_dma_get_range - Get DMA range info
* @dev: device pointer; only needed for a corner case.

drivers/of/address.c +1 -1
@@ -933,7 +997,7 @@ EXPORT_SYMBOL(of_io_request_and_map);
  *   CPU addr (phys_addr_t)  : pna cells
  *   size                    : nsize cells
  *
- * It returns -ENODEV if "dma-ranges" property was not found
+ * It returns -ENODEV if !dev or "dma-ranges" property was not found
  * for this device in DT.
  */
 int of_dma_get_range(struct device *dev, struct device_node *np, u64

*dma_addr,

drivers/of/address.c +1 -0
@@ -946,7 +1010,13 @@ int of_dma_get_range(struct device *dev, struct
 device_node *np, u64 *dma_addr,
      bool found_dma_ranges = false;
      struct of_range_parser parser;
      struct of_range range;
+     phys_addr_t cpu_start = ~(phys_addr_t)0;
Quoted 8 lines

drivers/of/address.c int of_dma_get_range(struct device *dev, struct device_node *np, u64 *dma_addr, · line 1019

Quoted 6 lines
drivers/of/address.c +5 -0
      u64 dma_start = U64_MAX, dma_end = 0, dma_offset = 0;
+     bool dma_multi_pfn_offset = false;
+     int num_ranges = 0;
+
+     if (!dev)
+             return -ENODEV;

Shouldn't this be part of patch #7?

Do you mean #8? Do you mean the test for !dev? It is not required
for #8 so I thought I'd keep these two changes separate. I could squash them.

Quoted 4 lines

Regards,
Nicolas

Many thanks!

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

#21 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Quoted 3 lines · Jim Quinlan
Quoted 1 line · Nicolas Saenz Julienne

On Wed, 2020-06-03 at 15:20 -0400, Jim Quinlan wrote:

...

Quoted 6 lines
Quoted 4 lines

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle, · line 40

Quoted 2 lines
arch/sh/kernel/dma-coherent.c +2 -0
+     phys = virt_to_phys(ret);
+     pfn =  phys >> PAGE_SHIFT;

nit: not sure it really pays off to have a pfn variable here.

Did it for readability; the compiler's optimization should take care of any extra variables. But I can switch if you insist.

One side note: please, try to get familiar with existing helpers in the kernel. For example, above line is like

	pfn = PFN_DOWN(phys);

...

Quoted 4 lines
Quoted 4 lines

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle, · line 46

Quoted 4 lines
arch/sh/kernel/dma-coherent.c +3 -0
+     if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
 
+             *dma_handle -= PFN_PHYS(
+                     dma_pfn_offset_from_phys_addr(dev, phys));

Don't do such indentation, esp. we have now 100! :-)

Signature · 8 lines
With Best Regards,
Andy Shevchenko


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

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Hi Andy,

Quoted 11 lines · Andy Shevchenko
Quoted 3 lines · Jim Quinlan
Quoted 1 line · Nicolas Saenz Julienne

On Wed, 2020-06-03 at 15:20 -0400, Jim Quinlan wrote:

...

Quoted 3 lines
Quoted 3 lines
Quoted 2 lines
arch/sh/kernel/dma-coherent.c +2 -0
+     phys = virt_to_phys(ret);
+     pfn =  phys >> PAGE_SHIFT;
Quoted 8 lines
Quoted 3 lines
Quoted 1 line

nit: not sure it really pays off to have a pfn variable here.

Did it for readability; the compiler's optimization should take care of any extra variables. But I can switch if you insist.

One side note: please, try to get familiar with existing helpers in the kernel. For example, above line is like

        pfn = PFN_DOWN(phys);

I just used the term in the original code; will change to PFN_DOWN().

Quoted 9 lines

...

Quoted 4 lines
Quoted 4 lines

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle, · line 46

Quoted 4 lines
arch/sh/kernel/dma-coherent.c +3 -0
+     if (!WARN_ON(!dev) && dev->dma_pfn_offset_map)
 
+             *dma_handle -= PFN_PHYS(
+                     dma_pfn_offset_from_phys_addr(dev, phys));

Don't do such indentation, esp. we have now 100! :-)

Got it. Thanks,
Jim Quinlan

Quoted 6 lines

--
With Best Regards,
Andy Shevchenko

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

#23 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

In reply to Jim Quinlan

Hi Jim,

On Thu, 2020-06-04 at 10:35 -0400, Jim Quinlan wrote:

[...]

Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
arch/sh/kernel/dma-coherent.c +2 -0
@@ -14,6 +14,8 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
 {
      void *ret, *ret_nocache;
      int order = get_order(size);
+     unsigned long pfn;
+     phys_addr_t phys;
 
      gfp |= __GFP_ZERO;
@@ -34,11 +36,14 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
              return NULL;
      }
 
Quoted 7 lines
Quoted 5 lines

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle, · line 40

Quoted 3 lines
arch/sh/kernel/dma-coherent.c +2 -1
-     split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
+     phys = virt_to_phys(ret);
+     pfn =  phys >> PAGE_SHIFT;

nit: not sure it really pays off to have a pfn variable here.

Did it for readability; the compiler's optimization should take care of any extra variables. But I can switch if you insist.

No need.

[...]

Quoted 42 lines
Quoted 42 lines
Quoted 30 lines
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device
 *pdev)
 
      sdev->dev = &pdev->dev;
      /* The DMA bus has the memory mapped at 0 */
-     sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+     ret = attach_uniform_dma_pfn_offset(sdev->dev,
+                                         PHYS_OFFSET >> PAGE_SHIFT);
+     if (ret)
+             return ret;
 
      ret = sun6i_csi_resource_request(sdev, pdev);

if (ret)

drivers/of/address.c +0 -0
index 96d8cfb14a60..c89333b0a5fb 100644
@@ -918,6 +918,70 @@ void __iomem *of_io_request_and_map(struct

device_node
*np, int index,
}
EXPORT_SYMBOL(of_io_request_and_map);

drivers/of/address.c +2 -0
+static int attach_dma_pfn_offset_map(struct device *dev,
+                                  struct device_node *node, int
Quoted 1 line

num_ranges)

As with the previous review, please take this comment with a grain of salt.

I think there should be a clear split between what belongs to OF and what belongs to the core device infrastructure.

OF's job should be to parse DT and provide a list/array of ranges, whereas the
core device infrastructure should provide an API to assign a list of ranges/offset to a device.

Quoted 8 lines
Quoted 5 lines

As a concrete example, you're forcing devices like the sta2x11 to build with OF
support, which, being an Intel device, it's pretty odd. But I'm also thinking
of how will all this fit once an ACPI device wants to use it.

To fix this I only have to move attach_uniform_dma_pfn_offset() from of/address.c to say include/linux/dma-mapping.h. It has no dependencies on OF. Do you agree?

Yes that seems nicer. In case you didn't had it in mind already, I'd change the function name to match the naming scheme they use there.

On the other hand, I'd also move the non OF parts of the non unifom dma_offset version of the function there.

Quoted 7 lines
Quoted 5 lines

Expanding on this idea, once you have a split between the OF's and device core
roles, it transpires that of_dma_get_range()'s job should only be to provide the ranges in a device understandable structure and of_dma_configre()'s to actually assign the device's parameters. This would obsolete patch #7.

I think you mean patch #8.

Yes, my bad.

Quoted 11 lines

I agree with you. The reason I needed a "struct device *" in the call is because I wanted to make sure the memory that is alloc'd belongs to the device that needs it. If I do a regular kzalloc(), this memory will become a leak once someone starts unbinding/binding their device. Also, in all uses of of_dma_rtange() -- there is only one -- a dev is required as one can't attach an offset map to NULL.

I do see that there are a number of functions in drivers/of/*.c that take 'struct device *dev' as an argument so there is precedent for something like this. Regardless, I need an owner to the memory I alloc().

I understand the need for dev to be around, devm_*() is key. But also it's important to keep the functions on purpose. And if of_dma_get_range() starts setting ranges it calls, for the very least, for a function rename. Although I'd rather split the parsing and setting of ranges as mentioned earlier. That said, I get that's a more drastic move.

Talking about drastic moves. How about getting rid of the concept of dma_pfn_offset for drivers altogether. Let them provide dma_pfn_offset_regions (even when there is only one). I feel it's conceptually nicer, as you'd be dealing only in one currency, so to speak, and you'd centralize the bus DMA ranges setter function which is always easier to maintain.

I'd go as far as not creating a special case for uniform offsets. Let just set cpu_end and dma_end to -1 so we always get a match. It's slightly more compute heavy, but I don't think it's worth the optimization.

Just my two cents :)

Quoted 107 lines
Quoted 29 lines
Quoted 29 lines
drivers/of/address.c +29 -0
+{
+     struct of_range_parser parser;
+     struct of_range range;
+     struct dma_pfn_offset_region *r;
+
+     r = devm_kcalloc(dev, num_ranges + 1,
+                      sizeof(struct dma_pfn_offset_region), GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;
+     dev->dma_pfn_offset_map = r;
+     of_dma_range_parser_init(&parser, node);
+
+     /*
+      * Record all info for DMA ranges array.  We could
+      * just use the of_range struct, but if we did that it
+      * would require more calculations for phys_to_dma and
+      * dma_to_phys conversions.
+      */
+     for_each_of_range(&parser, &range) {
+             r->cpu_start = range.cpu_addr;
+             r->cpu_end = r->cpu_start + range.size - 1;
+             r->dma_start = range.bus_addr;
+             r->dma_end = r->dma_start + range.size - 1;
+             r->pfn_offset = PFN_DOWN(range.cpu_addr)
+                     - PFN_DOWN(range.bus_addr);
+             r++;
+     }
+     return 0;
+}
Quoted 8 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 957

Quoted 6 lines
drivers/of/address.c +6 -0
+
+
+
+/**
+ * attach_dma_pfn_offset - Assign scalar offset for all addresses.
+ * @dev:     device pointer; only needed for a corner case.

That's a huge corner :P

Good point; I'm not really sure what percent of Linux configurations require a dma_pfn_offset. I'll drop the "corner".

Quoted 9 lines
Quoted 7 lines

+ * @dma_pfn_offset: offset to apply when converting from phys addr
+ * to dma addr and vice versa.
+ *
+ * It returns -ENOMEM if out of memory, otherwise 0.
+ */
+int attach_uniform_dma_pfn_offset(struct device *dev, unsigned long pfn_offset)

As I say above, does this really belong to of/address.c?

No it does not. Will fix.

Quoted 16 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 976

Quoted 13 lines
drivers/of/address.c +13 -0
+{
+     struct dma_pfn_offset_region *r;
+
+     if (!dev)
+             return -ENODEV;
+
+     if (!pfn_offset)
+             return 0;
+
+     r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+                      GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;

I think you're missing this:

Quoted 2 lines
        dev->dma_pfn_offset_map = r;

That's a showstopper! DanC also pointed it out but I still didn't see it. Thanks!

Quoted 27 lines
Quoted 27 lines
drivers/of/address.c +9 -1
+
+     r->uniform_offset = true;
+     r->pfn_offset = pfn_offset;
+
+     return 0;
+}
+EXPORT_SYMBOL_GPL(attach_uniform_dma_pfn_offset);
+
 /**
  * of_dma_get_range - Get DMA range info
  * @dev:     device pointer; only needed for a corner case.
@@ -933,7 +997,7 @@ EXPORT_SYMBOL(of_io_request_and_map);
  *   CPU addr (phys_addr_t)  : pna cells
  *   size                    : nsize cells
  *
- * It returns -ENODEV if "dma-ranges" property was not found
+ * It returns -ENODEV if !dev or "dma-ranges" property was not found
  * for this device in DT.
  */
 int of_dma_get_range(struct device *dev, struct device_node *np, u64

*dma_addr,

drivers/of/address.c +1 -0
@@ -946,7 +1010,13 @@ int of_dma_get_range(struct device *dev, struct
 device_node *np, u64 *dma_addr,
      bool found_dma_ranges = false;
      struct of_range_parser parser;
      struct of_range range;
+     phys_addr_t cpu_start = ~(phys_addr_t)0;
Quoted 8 lines

drivers/of/address.c int of_dma_get_range(struct device *dev, struct device_node *np, u64 *dma_addr, · line 1019

Quoted 6 lines
drivers/of/address.c +5 -0
      u64 dma_start = U64_MAX, dma_end = 0, dma_offset = 0;
+     bool dma_multi_pfn_offset = false;
+     int num_ranges = 0;
+
+     if (!dev)
+             return -ENODEV;

Shouldn't this be part of patch #7?

Quoted 3 lines

Do you mean #8? Do you mean the test for !dev? It is not required
for #8 so I thought I'd keep these two changes separate. I could squash them.

#8, of course.

It's more of a subjective matter, but to me it fits better #8's description and keeps this one more focused. That said, it's just a comment, do as you please.

Regads,
Nicolas

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

#24 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Hi Nicolas,

Quoted 115 lines · Nicolas Saenz Julienne

Hi Jim,

On Thu, 2020-06-04 at 10:35 -0400, Jim Quinlan wrote:

[...]

Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
arch/sh/kernel/dma-coherent.c +2 -0
@@ -14,6 +14,8 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
 {
      void *ret, *ret_nocache;
      int order = get_order(size);
+     unsigned long pfn;
+     phys_addr_t phys;
 
      gfp |= __GFP_ZERO;
@@ -34,11 +36,14 @@ void *arch_dma_alloc(struct device *dev, size_t size,
 dma_addr_t *dma_handle,
              return NULL;
      }
 
Quoted 7 lines
Quoted 5 lines

arch/sh/kernel/dma-coherent.c void *arch_dma_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle, · line 40

Quoted 3 lines
arch/sh/kernel/dma-coherent.c +2 -1
-     split_page(pfn_to_page(virt_to_phys(ret) >> PAGE_SHIFT), order);
+     phys = virt_to_phys(ret);
+     pfn =  phys >> PAGE_SHIFT;

nit: not sure it really pays off to have a pfn variable here.

Did it for readability; the compiler's optimization should take care of any extra variables. But I can switch if you insist.

No need.

[...]

Quoted 42 lines
Quoted 42 lines
Quoted 30 lines
drivers/media/platform/sunxi/sun6i-csi/sun6i_csi.c +4 -1
index 055eb0b8e396..2d66d415b6c3 100644
@@ -898,7 +898,10 @@ static int sun6i_csi_probe(struct platform_device
 *pdev)
 
      sdev->dev = &pdev->dev;
      /* The DMA bus has the memory mapped at 0 */
-     sdev->dev->dma_pfn_offset = PHYS_OFFSET >> PAGE_SHIFT;
+     ret = attach_uniform_dma_pfn_offset(sdev->dev,
+                                         PHYS_OFFSET >> PAGE_SHIFT);
+     if (ret)
+             return ret;
 
      ret = sun6i_csi_resource_request(sdev, pdev);

if (ret)

drivers/of/address.c +0 -0
index 96d8cfb14a60..c89333b0a5fb 100644
@@ -918,6 +918,70 @@ void __iomem *of_io_request_and_map(struct

device_node
*np, int index,
}
EXPORT_SYMBOL(of_io_request_and_map);

drivers/of/address.c +2 -0
+static int attach_dma_pfn_offset_map(struct device *dev,
+                                  struct device_node *node, int
Quoted 1 line

num_ranges)

As with the previous review, please take this comment with a grain of salt.

I think there should be a clear split between what belongs to OF and what belongs to the core device infrastructure.

OF's job should be to parse DT and provide a list/array of ranges, whereas the
core device infrastructure should provide an API to assign a list of ranges/offset to a device.

Quoted 8 lines
Quoted 5 lines

As a concrete example, you're forcing devices like the sta2x11 to build with OF
support, which, being an Intel device, it's pretty odd. But I'm also thinking
of how will all this fit once an ACPI device wants to use it.

To fix this I only have to move attach_uniform_dma_pfn_offset() from of/address.c to say include/linux/dma-mapping.h. It has no dependencies on OF. Do you agree?

Yes that seems nicer. In case you didn't had it in mind already, I'd change the function name to match the naming scheme they use there.

On the other hand, I'd also move the non OF parts of the non unifom dma_offset version of the function there.

Quoted 7 lines
Quoted 5 lines

Expanding on this idea, once you have a split between the OF's and device core
roles, it transpires that of_dma_get_range()'s job should only be to provide the ranges in a device understandable structure and of_dma_configre()'s to actually assign the device's parameters. This would obsolete patch #7.

I think you mean patch #8.

Yes, my bad.

Quoted 11 lines

I agree with you. The reason I needed a "struct device *" in the call is because I wanted to make sure the memory that is alloc'd belongs to the device that needs it. If I do a regular kzalloc(), this memory will become a leak once someone starts unbinding/binding their device. Also, in all uses of of_dma_rtange() -- there is only one -- a dev is required as one can't attach an offset map to NULL.

I do see that there are a number of functions in drivers/of/*.c that take 'struct device *dev' as an argument so there is precedent for something like this. Regardless, I need an owner to the memory I alloc().

Quoted 5 lines

I understand the need for dev to be around, devm_*() is key. But also it's important to keep the functions on purpose. And if of_dma_get_range() starts setting ranges it calls, for the very least, for a function rename. Although I'd rather split the parsing and setting of ranges as mentioned earlier. That said, I get that's a more drastic move.

I agree with you. I could do this from device.c:

        of_dma_get_num_ranges(..., &num_ranges); /* new function */
        r = devm_kcalloc(dev, num_ranges + 1, sizeof(*r), GFP_KERNEL);
        of_dma_get_range(np, &dma_addr, &paddr, &size, r, num_ranges);

The problem here is that there could be four ranges, all with offset=0. My current code would optimize this case out but the above would have us holding useless memory and looping through the four ranges on every dma <=> phys conversion only to add 0.

Quoted 6 lines

Talking about drastic moves. How about getting rid of the concept of dma_pfn_offset for drivers altogether. Let them provide dma_pfn_offset_regions (even when there is only one). I feel it's conceptually nicer, as you'd be dealing only in one currency, so to speak, and you'd centralize the bus DMA ranges setter function which is always easier to maintain.

Do you agree that we have to somehow hang this info on the struct device structure? Because in the dma2phys() and phys2dma() all you have is the dev parameter. I don't see how this can be done w/o involving dev.

Quoted 4 lines

I'd go as far as not creating a special case for uniform offsets. Let just set cpu_end and dma_end to -1 so we always get a match. It's slightly more compute heavy, but I don't think it's worth the optimization.

Well, there are two subcases here. One where we do know the bounds and one where we do not. I suppose for the latter I could have the drivers calling it with begin=0 and end=~(dma_addr_t)0. Let me give this some thought...

Quoted 2 lines

Just my two cents :)

Worth much more than $0.02 IMO :-)

BTW, I tried putting the "if (dev->dma_pfn_offset_map)" clause inside the inline functions but the problem is that it slows the fastpath; consider the following code from dma-direct.h

if (dev->dma_pfn_offset_map) {
unsigned long dma_pfn_offset =
dma_pfn_offset_from_phys_addr(dev, paddr);

                dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
        }
        return dev_addr;

becomes

unsigned long dma_pfn_offset =
dma_pfn_offset_from_phys_addr(dev, paddr);

        dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
        return dev_addr;

So those configurations that have no dma_pfn_offsets are doing an unnecessary shift and add.

Thanks much,
Jim Quinlan

Quoted 112 lines
Quoted 108 lines
Quoted 30 lines
Quoted 30 lines
drivers/of/address.c +29 -0
+{
+     struct of_range_parser parser;
+     struct of_range range;
+     struct dma_pfn_offset_region *r;
+
+     r = devm_kcalloc(dev, num_ranges + 1,
+                      sizeof(struct dma_pfn_offset_region), GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;
+     dev->dma_pfn_offset_map = r;
+     of_dma_range_parser_init(&parser, node);
+
+     /*
+      * Record all info for DMA ranges array.  We could
+      * just use the of_range struct, but if we did that it
+      * would require more calculations for phys_to_dma and
+      * dma_to_phys conversions.
+      */
+     for_each_of_range(&parser, &range) {
+             r->cpu_start = range.cpu_addr;
+             r->cpu_end = r->cpu_start + range.size - 1;
+             r->dma_start = range.bus_addr;
+             r->dma_end = r->dma_start + range.size - 1;
+             r->pfn_offset = PFN_DOWN(range.cpu_addr)
+                     - PFN_DOWN(range.bus_addr);
+             r++;
+     }
+     return 0;
+}
Quoted 8 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 957

Quoted 6 lines
drivers/of/address.c +6 -0
+
+
+
+/**
+ * attach_dma_pfn_offset - Assign scalar offset for all addresses.
+ * @dev:     device pointer; only needed for a corner case.

That's a huge corner :P

Good point; I'm not really sure what percent of Linux configurations require a dma_pfn_offset. I'll drop the "corner".

Quoted 9 lines
Quoted 7 lines

+ * @dma_pfn_offset: offset to apply when converting from phys addr
+ * to dma addr and vice versa.
+ *
+ * It returns -ENOMEM if out of memory, otherwise 0.
+ */
+int attach_uniform_dma_pfn_offset(struct device *dev, unsigned long pfn_offset)

As I say above, does this really belong to of/address.c?

No it does not. Will fix.

Quoted 16 lines

drivers/of/address.c void __iomem *of_io_request_and_map(struct device_node *np, int index, · line 976

Quoted 13 lines
drivers/of/address.c +13 -0
+{
+     struct dma_pfn_offset_region *r;
+
+     if (!dev)
+             return -ENODEV;
+
+     if (!pfn_offset)
+             return 0;
+
+     r = devm_kcalloc(dev, 1, sizeof(struct dma_pfn_offset_region),
+                      GFP_KERNEL);
+     if (!r)
+             return -ENOMEM;

I think you're missing this:

Quoted 2 lines
        dev->dma_pfn_offset_map = r;

That's a showstopper! DanC also pointed it out but I still didn't see it. Thanks!

Quoted 27 lines
Quoted 27 lines
drivers/of/address.c +9 -1
+
+     r->uniform_offset = true;
+     r->pfn_offset = pfn_offset;
+
+     return 0;
+}
+EXPORT_SYMBOL_GPL(attach_uniform_dma_pfn_offset);
+
 /**
  * of_dma_get_range - Get DMA range info
  * @dev:     device pointer; only needed for a corner case.
@@ -933,7 +997,7 @@ EXPORT_SYMBOL(of_io_request_and_map);
  *   CPU addr (phys_addr_t)  : pna cells
  *   size                    : nsize cells
  *
- * It returns -ENODEV if "dma-ranges" property was not found
+ * It returns -ENODEV if !dev or "dma-ranges" property was not found
  * for this device in DT.
  */
 int of_dma_get_range(struct device *dev, struct device_node *np, u64

*dma_addr,

drivers/of/address.c +1 -0
@@ -946,7 +1010,13 @@ int of_dma_get_range(struct device *dev, struct
 device_node *np, u64 *dma_addr,
      bool found_dma_ranges = false;
      struct of_range_parser parser;
      struct of_range range;
+     phys_addr_t cpu_start = ~(phys_addr_t)0;
Quoted 8 lines

drivers/of/address.c int of_dma_get_range(struct device *dev, struct device_node *np, u64 *dma_addr, · line 1019

Quoted 6 lines
drivers/of/address.c +5 -0
      u64 dma_start = U64_MAX, dma_end = 0, dma_offset = 0;
+     bool dma_multi_pfn_offset = false;
+     int num_ranges = 0;
+
+     if (!dev)
+             return -ENODEV;

Shouldn't this be part of patch #7?

Quoted 3 lines

Do you mean #8? Do you mean the test for !dev? It is not required
for #8 so I thought I'd keep these two changes separate. I could squash them.

Quoted 8 lines

#8, of course.

It's more of a subjective matter, but to me it fits better #8's description and keeps this one more focused. That said, it's just a comment, do as you please.

Regads,
Nicolas

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

#25 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Hi Christoph,
a question arouse, is there a real value to dealing with PFNs (as opposed to real addresses) in the core DMA code/structures? I see that in some cases it eases interacting with mm, but the overwhelming usage of say, dev->dma_pfn_offset, involves shifting it.

Hi Jim,

Quoted 1 line · Jim Quinlan

Hi Nicolas,

[...]

Quoted 9 lines
Quoted 6 lines

I understand the need for dev to be around, devm_*() is key. But also it's important to keep the functions on purpose. And if of_dma_get_range() starts setting ranges it calls, for the very least, for a function rename. Although I'd rather split the parsing and setting of ranges as mentioned earlier. That
said, I get that's a more drastic move.

I agree with you. I could do this from device.c:

Quoted 8 lines
        of_dma_get_num_ranges(..., &num_ranges); /* new function */
        r = devm_kcalloc(dev, num_ranges + 1, sizeof(*r), GFP_KERNEL);
        of_dma_get_range(np, &dma_addr, &paddr, &size, r, num_ranges);

The problem here is that there could be four ranges, all with offset=0. My current code would optimize this case out but the above would have us holding useless memory and looping through the four ranges on every dma <=> phys conversion only to add 0.

Point taken. Ultimately it's setting the device's dma ranges in of_dma_get_range() that was really bothering me, so if we have to pass the device pointer for allocations, be it.

Quoted 10 lines
Quoted 6 lines

Talking about drastic moves. How about getting rid of the concept of dma_pfn_offset for drivers altogether. Let them provide dma_pfn_offset_regions
(even when there is only one). I feel it's conceptually nicer, as you'd be dealing only in one currency, so to speak, and you'd centralize the bus DMA ranges setter function which is always easier to maintain.

Do you agree that we have to somehow hang this info on the struct device structure? Because in the dma2phys() and phys2dma() all you have is the dev parameter. I don't see how this can be done w/o involving dev.

Sorry I didn't make myself clear here. What bothers me is having two functions setting the same device parameter trough different means, I'd be happy to get rid of attach_uniform_dma_pfn_offset(), and always use the same function to set a device's bus dma regions. Something the likes of this comes to mind:

dma_attach_pfn_offset_region(struct device *dev, struct dma_pfn_offset_regions *r)

We could maybe use some helper macros for the linear case. But that's the gist of it.

Also, it goes hand in hand with the comment below. Why having a special case for non sparse DMA offsets in struct dma_pfn_offset_regions? The way I see it, in this case, code simplicity is more interesting than a small optimization.

Quoted 5 lines
Quoted 5 lines

I'd go as far as not creating a special case for uniform offsets. Let just set
cpu_end and dma_end to -1 so we always get a match. It's slightly more compute
heavy, but I don't think it's worth the optimization.

Quoted 8 lines

Well, there are two subcases here. One where we do know the bounds and one where we do not. I suppose for the latter I could have the drivers calling it with begin=0 and end=~(dma_addr_t)0. Let me give this some thought...

Quoted 1 line

Just my two cents :)

Worth much more than $0.02 IMO :-)

BTW, would you consider renaming the DMA offset struct to something simpler like, struct bus_dma_region? It complements 'dev->bus_dma_limit' better IMO.

Quoted 6 lines

BTW, I tried putting the "if (dev->dma_pfn_offset_map)" clause inside the inline functions but the problem is that it slows the fastpath; consider the following code from dma-direct.h

if (dev->dma_pfn_offset_map) {
unsigned long dma_pfn_offset =

dma_pfn_offset_from_phys_addr(dev, paddr);

Quoted 8 lines
                dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
        }
        return dev_addr;

becomes

unsigned long dma_pfn_offset = dma_pfn_offset_from_phys_addr(dev,

paddr);

Quoted 6 lines
        dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
        return dev_addr;

So those configurations that have no dma_pfn_offsets are doing an unnecessary shift and add.

Fair enough. Still not a huge difference, but I see the value being the most common case.

Regards,
Nicolas

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

#26 lore raw

Re: [PATCH v3 09/13] device core: Introduce multiple dma pfn offsets

Quoted 51 lines · Nicolas Saenz Julienne

Hi Christoph,
a question arouse, is there a real value to dealing with PFNs (as opposed to real addresses) in the core DMA code/structures? I see that in some cases it eases interacting with mm, but the overwhelming usage of say, dev->dma_pfn_offset, involves shifting it.

Hi Jim,

Quoted 1 line · Jim Quinlan

Hi Nicolas,

[...]

Quoted 9 lines
Quoted 6 lines

I understand the need for dev to be around, devm_*() is key. But also it's important to keep the functions on purpose. And if of_dma_get_range() starts setting ranges it calls, for the very least, for a function rename. Although I'd rather split the parsing and setting of ranges as mentioned earlier. That
said, I get that's a more drastic move.

I agree with you. I could do this from device.c:

Quoted 8 lines
        of_dma_get_num_ranges(..., &num_ranges); /* new function */
        r = devm_kcalloc(dev, num_ranges + 1, sizeof(*r), GFP_KERNEL);
        of_dma_get_range(np, &dma_addr, &paddr, &size, r, num_ranges);

The problem here is that there could be four ranges, all with offset=0. My current code would optimize this case out but the above would have us holding useless memory and looping through the four ranges on every dma <=> phys conversion only to add 0.

Point taken. Ultimately it's setting the device's dma ranges in of_dma_get_range() that was really bothering me, so if we have to pass the device pointer for allocations, be it.

Quoted 10 lines
Quoted 6 lines

Talking about drastic moves. How about getting rid of the concept of dma_pfn_offset for drivers altogether. Let them provide dma_pfn_offset_regions
(even when there is only one). I feel it's conceptually nicer, as you'd be dealing only in one currency, so to speak, and you'd centralize the bus DMA ranges setter function which is always easier to maintain.

Do you agree that we have to somehow hang this info on the struct device structure? Because in the dma2phys() and phys2dma() all you have is the dev parameter. I don't see how this can be done w/o involving dev.

Sorry I didn't make myself clear here. What bothers me is having two functions setting the same device parameter trough different means, I'd be happy to get rid of attach_uniform_dma_pfn_offset(), and always use the same function to set a device's bus dma regions. Something the likes of this comes to mind:

Quoted 8 lines

dma_attach_pfn_offset_region(struct device *dev, struct dma_pfn_offset_regions *r)

We could maybe use some helper macros for the linear case. But that's the gist of it.

Also, it goes hand in hand with the comment below. Why having a special case for non sparse DMA offsets in struct dma_pfn_offset_regions? The way I see it, in this case, code simplicity is more interesting than a small optimization.

I've removed the special case and also need for 'dev' in of_dma_get_range(). v4 is comming...

Quoted 11 lines
Quoted 11 lines
Quoted 6 lines

I'd go as far as not creating a special case for uniform offsets. Let just set
cpu_end and dma_end to -1 so we always get a match. It's slightly more compute
heavy, but I don't think it's worth the optimization.

Well, there are two subcases here. One where we do know the bounds and one where we do not. I suppose for the latter I could have the drivers calling it with begin=0 and end=~(dma_addr_t)0. Let me give this some thought...

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

Just my two cents :)

Worth much more than $0.02 IMO :-)

BTW, would you consider renaming the DMA offset struct to something simpler like, struct bus_dma_region? It complements 'dev->bus_dma_limit' better IMO.

Will do

Thanks,
Jim

Quoted 21 lines
Quoted 7 lines

BTW, I tried putting the "if (dev->dma_pfn_offset_map)" clause inside the inline functions but the problem is that it slows the fastpath; consider the following code from dma-direct.h

if (dev->dma_pfn_offset_map) {
unsigned long dma_pfn_offset =

dma_pfn_offset_from_phys_addr(dev, paddr);

Quoted 8 lines
                dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
        }
        return dev_addr;

becomes

unsigned long dma_pfn_offset = dma_pfn_offset_from_phys_addr(dev,

paddr);

Quoted 4 lines
        dev_addr -= ((dma_addr_t)dma_pfn_offset << PAGE_SHIFT);
        return dev_addr;
Quoted 9 lines
Quoted 2 lines

So those configurations that have no dma_pfn_offsets are doing an unnecessary shift and add.

Fair enough. Still not a huge difference, but I see the value being the most common case.

Regards,
Nicolas

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

#27 lore raw

PATCH v3 10/13 PCI: brcmstb: Set internal memory viewport sizes

In reply to Jim Quinlan

BrcmSTB PCIe controllers are intimately connected to the memory controller(s) on the SOC. There is a "viewport" for each memory controller that allows inbound accesses to CPU memory. Each viewport's size must be set to a power of two, and that size must be equal to or larger than the amount of memory each controller supports.

---
 drivers/pci/controller/pcie-brcmstb.c | 67 ++++++++++++++++++++-------
 1 file changed, 49 insertions(+), 18 deletions(-)
drivers/pci/controller/pcie-brcmstb.c +49 -18
index fa356bc149c3..338e9ed44230 100644
@@ -55,6 +55,8 @@
 #define  PCIE_MISC_MISC_CTRL_MAX_BURST_SIZE_MASK	0x300000
 #define  PCIE_MISC_MISC_CTRL_MAX_BURST_SIZE_128		0x0
 #define  PCIE_MISC_MISC_CTRL_SCB0_SIZE_MASK		0xf8000000
+#define  PCIE_MISC_MISC_CTRL_SCB1_SIZE_MASK		0x07c00000
+#define  PCIE_MISC_MISC_CTRL_SCB2_SIZE_MASK		0x0000001f
 
 #define PCIE_MISC_CPU_2_PCIE_MEM_WIN0_LO		0x400c
 #define PCIE_MEM_WIN0_LO(win)	\
@@ -152,6 +154,7 @@
 #define SSC_STATUS_OFFSET		0x1
 #define SSC_STATUS_SSC_MASK		0x400
 #define SSC_STATUS_PLL_LOCK_MASK	0x800
+#define PCIE_BRCM_MAX_MEMC		3
 
 /* Rescal registers */
 #define PCIE_DVT_PMU_PCIE_PHY_CTRL				0xc700
@@ -261,6 +264,8 @@ struct brcm_pcie {
 	const int		*reg_field_info;
 	enum pcie_type		type;
 	struct reset_control	*rescal;
+	int			num_memc;
+	u64			memc_size[PCIE_BRCM_MAX_MEMC];
 };
 
 /*
@@ -717,22 +722,40 @@ static inline int brcm_pcie_get_rc_bar2_size_and_offset(struct brcm_pcie *pcie,
 							u64 *rc_bar2_offset)
 {
 	struct pci_host_bridge *bridge = pci_host_bridge_from_priv(pcie);
-	struct device *dev = pcie->dev;
 	struct resource_entry *entry;
+	struct device *dev = pcie->dev;
+	u64 lowest_pcie_addr = ~(u64)0;
+	int ret, i = 0;
+	u64 size = 0;
 
-	entry = resource_list_first_type(&bridge->dma_ranges, IORESOURCE_MEM);
-	if (!entry)
-		return -ENODEV;
+	resource_list_for_each_entry(entry, &bridge->dma_ranges) {
+		u64 pcie_beg = entry->res->start - entry->offset;
 
+		size += entry->res->end - entry->res->start + 1;
+		if (pcie_beg < lowest_pcie_addr)
+			lowest_pcie_addr = pcie_beg;
+	}
 
-	/*
-	 * The controller expects the inbound window offset to be calculated as
-	 * the difference between PCIe's address space and CPU's. The offset
-	 * provided by the firmware is calculated the opposite way, so we
-	 * negate it.
-	 */
-	*rc_bar2_offset = -entry->offset;
-	*rc_bar2_size = 1ULL << fls64(entry->res->end - entry->res->start);
+	ret = of_property_read_variable_u64_array(
+		pcie->np, "brcm,scb-sizes", pcie->memc_size, 1,
+		PCIE_BRCM_MAX_MEMC);
+
+	if (ret <= 0) {
+		/* Make an educated guess */
+		pcie->num_memc = 1;
+		pcie->memc_size[0] = 1 << fls64(size - 1);
+	} else {
+		pcie->num_memc = ret;
+	}
+
+	/* Each memc is viewed through a "port" that is a power of 2 */
+	for (i = 0, size = 0; i < pcie->num_memc; i++)
+		size += pcie->memc_size[i];
+
+	/* System memory starts at this address in PCIe-space */
+	*rc_bar2_offset = lowest_pcie_addr;
+	/* The sum of all memc views must also be a power of 2 */
+	*rc_bar2_size = 1ULL << fls64(size - 1);
 
 	/*
 	 * We validate the inbound memory view even though we should trust
@@ -784,12 +807,11 @@ static int brcm_pcie_setup(struct brcm_pcie *pcie)
 	void __iomem *base = pcie->base;
 	struct device *dev = pcie->dev;
 	struct resource_entry *entry;
-	unsigned int scb_size_val;
 	bool ssc_good = false;
 	struct resource *res;
 	int num_out_wins = 0;
 	u16 nlw, cls, lnksta;
-	int i, ret;
+	int i, ret, memc;
 	u32 tmp, aspm_support;
 
 	/* Reset the bridge */
@@ -825,11 +847,20 @@ static int brcm_pcie_setup(struct brcm_pcie *pcie)
 	writel(upper_32_bits(rc_bar2_offset),
 	       base + PCIE_MISC_RC_BAR2_CONFIG_HI);
 
-	scb_size_val = rc_bar2_size ?
-		       ilog2(rc_bar2_size) - 15 : 0xf; /* 0xf is 1GB */
 	tmp = readl(base + PCIE_MISC_MISC_CTRL);
-	u32p_replace_bits(&tmp, scb_size_val,
-			  PCIE_MISC_MISC_CTRL_SCB0_SIZE_MASK);
+	for (memc = 0; memc < pcie->num_memc; memc++) {
+		u32 scb_size_val = ilog2(pcie->memc_size[memc]) - 15;
+
+		if (memc == 0)
+			u32p_replace_bits(&tmp, scb_size_val,
+					  PCIE_MISC_MISC_CTRL_SCB0_SIZE_MASK);
+		else if (memc == 1)
+			u32p_replace_bits(&tmp, scb_size_val,
+					  PCIE_MISC_MISC_CTRL_SCB1_SIZE_MASK);
+		else if (memc == 2)
+			u32p_replace_bits(&tmp, scb_size_val,
+					  PCIE_MISC_MISC_CTRL_SCB2_SIZE_MASK);
+	}
 	writel(tmp, base + PCIE_MISC_MISC_CTRL);
 
 	/*
Signature · 1 line
2.17.1
Acked-by on v3 10/13
#28 lore raw

Re: [PATCH v3 10/13] PCI: brcmstb: Set internal memory viewport sizes

Quoted 7 lines · Jim Quinlan

BrcmSTB PCIe controllers are intimately connected to the memory controller(s) on the SOC. There is a "viewport" for each memory controller that allows inbound accesses to CPU memory. Each viewport's size must be set to a power of two, and that size must be equal to or larger than the amount of memory each controller supports.

Signature · 6 lines
Florian

_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
#29 lore raw

PATCH v3 11/13 PCI: brcmstb: Accommodate MSI for older chips

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

Older BrcmSTB chips do not have a separate register for MSI interrupts; the MSIs are in a register that also contains unrelated interrupts. In addition, the interrupts lie in bits [31..24] for these legacy chips. This commit provides common code for both legacy and non-legacy MSI interrupt registers.

---
 drivers/pci/controller/pcie-brcmstb.c | 72 +++++++++++++++++++--------
 1 file changed, 52 insertions(+), 20 deletions(-)
drivers/pci/controller/pcie-brcmstb.c +52 -20
index 338e9ed44230..9930419e3ac2 100644
@@ -80,7 +80,8 @@
 #define PCIE_MISC_MSI_BAR_CONFIG_HI			0x4048
 
 #define PCIE_MISC_MSI_DATA_CONFIG			0x404c
-#define  PCIE_MISC_MSI_DATA_CONFIG_VAL			0xffe06540
+#define  PCIE_MISC_MSI_DATA_CONFIG_VAL_32		0xffe06540
+#define  PCIE_MISC_MSI_DATA_CONFIG_VAL_8		0xfff86540
 
 #define PCIE_MISC_PCIE_CTRL				0x4064
 #define  PCIE_MISC_PCIE_CTRL_PCIE_L23_REQUEST_MASK	0x1
@@ -92,6 +93,9 @@
 #define  PCIE_MISC_PCIE_STATUS_PCIE_PHYLINKUP_MASK	0x10
 #define  PCIE_MISC_PCIE_STATUS_PCIE_LINK_IN_L23_MASK	0x40
 
+#define PCIE_MISC_REVISION				0x406c
+#define  BRCM_PCIE_HW_REV_33				0x0303
+
 #define PCIE_MISC_CPU_2_PCIE_MEM_WIN0_BASE_LIMIT		0x4070
 #define  PCIE_MISC_CPU_2_PCIE_MEM_WIN0_BASE_LIMIT_LIMIT_MASK	0xfff00000
 #define  PCIE_MISC_CPU_2_PCIE_MEM_WIN0_BASE_LIMIT_BASE_MASK	0xfff0
@@ -112,10 +116,14 @@
 #define  PCIE_MISC_HARD_PCIE_HARD_DEBUG_CLKREQ_DEBUG_ENABLE_MASK	0x2
 #define  PCIE_MISC_HARD_PCIE_HARD_DEBUG_SERDES_IDDQ_MASK		0x08000000
 
-#define PCIE_MSI_INTR2_STATUS				0x4500
-#define PCIE_MSI_INTR2_CLR				0x4508
-#define PCIE_MSI_INTR2_MASK_SET				0x4510
-#define PCIE_MSI_INTR2_MASK_CLR				0x4514
+
+#define PCIE_INTR2_CPU_BASE		0x4300
+#define PCIE_MSI_INTR2_BASE		0x4500
+/* Offsets from PCIE_INTR2_CPU_BASE and PCIE_MSI_INTR2_BASE */
+#define  MSI_INT_STATUS			0x0
+#define  MSI_INT_CLR			0x8
+#define  MSI_INT_MASK_SET		0x10
+#define  MSI_INT_MASK_CLR		0x14
 
 #define PCIE_EXT_CFG_DATA				0x8000
 
@@ -130,6 +138,8 @@
 /* PCIe parameters */
 #define BRCM_NUM_PCIE_OUT_WINS		0x4
 #define BRCM_INT_PCI_MSI_NR		32
+#define BRCM_INT_PCI_MSI_LEGACY_NR	8
+#define BRCM_INT_PCI_MSI_SHIFT		0
 
 /* MSI target adresses */
 #define BRCM_MSI_TARGET_ADDR_LT_4GB	0x0fffffffcULL
@@ -247,6 +257,12 @@ struct brcm_msi {
 	int			irq;
 	/* used indicates which MSI interrupts have been alloc'd */
 	unsigned long		used;
+	bool			legacy;
+	/* Some chips have MSIs in bits [31..24] of a shared register. */
+	int			legacy_shift;
+	int			nr; /* No. of MSI available, depends on chip */
+	/* This is the base pointer for interrupt status/set/clr regs */
+	void __iomem		*intr_base;
 };
 
 /* Internal PCIe Host Controller Information.*/
@@ -266,6 +282,7 @@ struct brcm_pcie {
 	struct reset_control	*rescal;
 	int			num_memc;
 	u64			memc_size[PCIE_BRCM_MAX_MEMC];
+	u32			hw_rev;
 };
 
 /*
@@ -456,8 +473,10 @@ static void brcm_pcie_msi_isr(struct irq_desc *desc)
 	msi = irq_desc_get_handler_data(desc);
 	dev = msi->dev;
 
-	status = readl(msi->base + PCIE_MSI_INTR2_STATUS);
-	for_each_set_bit(bit, &status, BRCM_INT_PCI_MSI_NR) {
+	status = readl(msi->intr_base + MSI_INT_STATUS);
+	status >>= msi->legacy_shift;
+
+	for_each_set_bit(bit, &status, msi->nr) {
 		virq = irq_find_mapping(msi->inner_domain, bit);
 		if (virq)
 			generic_handle_irq(virq);
@@ -474,7 +493,7 @@ static void brcm_msi_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
 
 	msg->address_lo = lower_32_bits(msi->target_addr);
 	msg->address_hi = upper_32_bits(msi->target_addr);
-	msg->data = (0xffff & PCIE_MISC_MSI_DATA_CONFIG_VAL) | data->hwirq;
+	msg->data = (0xffff & PCIE_MISC_MSI_DATA_CONFIG_VAL_32) | data->hwirq;
 }
 
 static int brcm_msi_set_affinity(struct irq_data *irq_data,
@@ -486,8 +505,9 @@ static int brcm_msi_set_affinity(struct irq_data *irq_data,
 static void brcm_msi_ack_irq(struct irq_data *data)
 {
 	struct brcm_msi *msi = irq_data_get_irq_chip_data(data);
+	const int shift_amt = data->hwirq + msi->legacy_shift;
 
-	writel(1 << data->hwirq, msi->base + PCIE_MSI_INTR2_CLR);
+	writel(1 << shift_amt, msi->intr_base + MSI_INT_CLR);
 }
 
 
@@ -503,7 +523,7 @@ static int brcm_msi_alloc(struct brcm_msi *msi)
 	int hwirq;
 
 	mutex_lock(&msi->lock);
-	hwirq = bitmap_find_free_region(&msi->used, BRCM_INT_PCI_MSI_NR, 0);
+	hwirq = bitmap_find_free_region(&msi->used, msi->nr, 0);
 	mutex_unlock(&msi->lock);
 
 	return hwirq;
@@ -552,7 +572,7 @@ static int brcm_allocate_domains(struct brcm_msi *msi)
 	struct fwnode_handle *fwnode = of_node_to_fwnode(msi->np);
 	struct device *dev = msi->dev;
 
-	msi->inner_domain = irq_domain_add_linear(NULL, BRCM_INT_PCI_MSI_NR,
+	msi->inner_domain = irq_domain_add_linear(NULL, msi->nr,
 						  &msi_domain_ops, msi);
 	if (!msi->inner_domain) {
 		dev_err(dev, "failed to create IRQ domain\n");
@@ -590,7 +610,10 @@ static void brcm_msi_remove(struct brcm_pcie *pcie)
 
 static void brcm_msi_set_regs(struct brcm_msi *msi)
 {
-	writel(0xffffffff, msi->base + PCIE_MSI_INTR2_MASK_CLR);
+	u32 val = __GENMASK(31, msi->legacy_shift);
+
+	writel(val, msi->intr_base + MSI_INT_MASK_CLR);
+	writel(val, msi->intr_base + MSI_INT_CLR);
 
 	/*
 	 * The 0 bit of PCIE_MISC_MSI_BAR_CONFIG_LO is repurposed to MSI
@@ -601,8 +624,10 @@ static void brcm_msi_set_regs(struct brcm_msi *msi)
 	writel(upper_32_bits(msi->target_addr),
 	       msi->base + PCIE_MISC_MSI_BAR_CONFIG_HI);
 
-	writel(PCIE_MISC_MSI_DATA_CONFIG_VAL,
-	       msi->base + PCIE_MISC_MSI_DATA_CONFIG);
+	val = msi->legacy ? PCIE_MISC_MSI_DATA_CONFIG_VAL_8 :
+		PCIE_MISC_MSI_DATA_CONFIG_VAL_32;
+
+	writel(val, msi->base + PCIE_MISC_MSI_DATA_CONFIG);
 }
 
 static int brcm_pcie_enable_msi(struct brcm_pcie *pcie)
@@ -627,6 +652,17 @@ static int brcm_pcie_enable_msi(struct brcm_pcie *pcie)
 	msi->np = pcie->np;
 	msi->target_addr = pcie->msi_target_addr;
 	msi->irq = irq;
+	msi->legacy = pcie->hw_rev < BRCM_PCIE_HW_REV_33;
+
+	if (msi->legacy) {
+		msi->intr_base = msi->base + PCIE_INTR2_CPU_BASE;
+		msi->nr = BRCM_INT_PCI_MSI_LEGACY_NR;
+		msi->legacy_shift = 24;
+	} else {
+		msi->intr_base = msi->base + PCIE_MSI_INTR2_BASE;
+		msi->nr = BRCM_INT_PCI_MSI_NR;
+		msi->legacy_shift = 0;
+	}
 
 	ret = brcm_allocate_domains(msi);
 	if (ret)
@@ -885,12 +921,6 @@ static int brcm_pcie_setup(struct brcm_pcie *pcie)
 	tmp &= ~PCIE_MISC_RC_BAR3_CONFIG_LO_SIZE_MASK;
 	writel(tmp, base + PCIE_MISC_RC_BAR3_CONFIG_LO);
 
-	/* Mask all interrupts since we are not handling any yet */
-	writel(0xffffffff, pcie->base + PCIE_MSI_INTR2_MASK_SET);
-
-	/* clear any interrupts we find on boot */
-	writel(0xffffffff, pcie->base + PCIE_MSI_INTR2_CLR);
-
 	if (pcie->gen)
 		brcm_pcie_set_gen(pcie, pcie->gen);
 
@@ -1220,6 +1250,8 @@ static int brcm_pcie_probe(struct platform_device *pdev)
 	if (ret)
 		goto fail;
 
+	pcie->hw_rev = readl(pcie->base + PCIE_MISC_REVISION);
+
 	msi_np = of_parse_phandle(pcie->np, "msi-parent", 0);
 	if (pci_msi_enabled() && msi_np == pcie->np) {
 		ret = brcm_pcie_enable_msi(pcie);
Signature · 1 line
2.17.1
Acked-by on v3 11/13
#30 lore raw

Re: [PATCH v3 11/13] PCI: brcmstb: Accommodate MSI for older chips

Quoted 9 lines · Jim Quinlan

From: Jim Quinlan <[email protected]>

Older BrcmSTB chips do not have a separate register for MSI interrupts; the MSIs are in a register that also contains unrelated interrupts. In addition, the interrupts lie in bits [31..24] for these legacy chips. This commit provides common code for both legacy and non-legacy MSI interrupt registers.

Signature · 1 line
Florian
#31 lore raw

PATCH v3 12/13 PCI: brcmstb: Set bus max burst size by chip type

In reply to Jim Quinlan

From: Jim Quinlan <[email protected]>

The proper value of the parameter SCB_MAX_BURST_SIZE varies per chip. The 2711 family requires 128B whereas other devices can employ 512. The assignment is complicated by the fact that the values for this two-bit field have different meanings;

Value Type_Generic Type_7278

     00       Reserved         128B
     01           128B         256B
     10           256B         512B

11 512B Reserved

---
 drivers/pci/controller/pcie-brcmstb.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)
drivers/pci/controller/pcie-brcmstb.c +15 -3
index 9930419e3ac2..131cf0a51398 100644
@@ -53,7 +53,7 @@
 #define  PCIE_MISC_MISC_CTRL_SCB_ACCESS_EN_MASK		0x1000
 #define  PCIE_MISC_MISC_CTRL_CFG_READ_UR_MODE_MASK	0x2000
 #define  PCIE_MISC_MISC_CTRL_MAX_BURST_SIZE_MASK	0x300000
-#define  PCIE_MISC_MISC_CTRL_MAX_BURST_SIZE_128		0x0
+
 #define  PCIE_MISC_MISC_CTRL_SCB0_SIZE_MASK		0xf8000000
 #define  PCIE_MISC_MISC_CTRL_SCB1_SIZE_MASK		0x07c00000
 #define  PCIE_MISC_MISC_CTRL_SCB2_SIZE_MASK		0x0000001f
@@ -848,7 +848,7 @@ static int brcm_pcie_setup(struct brcm_pcie *pcie)
 	int num_out_wins = 0;
 	u16 nlw, cls, lnksta;
 	int i, ret, memc;
-	u32 tmp, aspm_support;
+	u32 tmp, burst, aspm_support;
 
 	/* Reset the bridge */
 	brcm_pcie_bridge_sw_init_set(pcie, 1);
@@ -864,10 +864,22 @@ static int brcm_pcie_setup(struct brcm_pcie *pcie)
 	/* Wait for SerDes to be stable */
 	usleep_range(100, 200);
 
+	/*
+	 * SCB_MAX_BURST_SIZE is a two bit field.  For GENERIC chips it
+	 * is encoded as 0=128, 1=256, 2=512, 3=Rsvd, for BCM7278 it
+	 * is encoded as 0=Rsvd, 1=128, 2=256, 3=512.
+	 */
+	if (pcie->type == BCM2711)
+		burst = 0x0; /* 128B */
+	else if (pcie->type == BCM7278)
+		burst = 0x3; /* 512 bytes */
+	else
+		burst = 0x2; /* 512 bytes */
+
 	/* Set SCB_MAX_BURST_SIZE, CFG_READ_UR_MODE, SCB_ACCESS_EN */
 	u32p_replace_bits(&tmp, 1, PCIE_MISC_MISC_CTRL_SCB_ACCESS_EN_MASK);
 	u32p_replace_bits(&tmp, 1, PCIE_MISC_MISC_CTRL_CFG_READ_UR_MODE_MASK);
-	u32p_replace_bits(&tmp, PCIE_MISC_MISC_CTRL_MAX_BURST_SIZE_128,
+	u32p_replace_bits(&tmp, burst,
 			  PCIE_MISC_MISC_CTRL_MAX_BURST_SIZE_MASK);
 	writel(tmp, base + PCIE_MISC_MISC_CTRL);
 
Signature · 1 line
2.17.1
Acked-by on v3 12/13
#32 lore raw

Re: [PATCH v3 12/13] PCI: brcmstb: Set bus max burst size by chip type

Quoted 8 lines · Jim Quinlan

From: Jim Quinlan <[email protected]>

The proper value of the parameter SCB_MAX_BURST_SIZE varies per chip. The 2711 family requires 128B whereas other devices can employ 512. The assignment is complicated by the fact that the values for this two-bit field have different meanings;

Value Type_Generic Type_7278

It looks like Type_Generic and Type_7278 should be swapped in this description. Other than that:

Quoted 5 lines
     00       Reserved         128B
     01           128B         256B
     10           256B         512B

11 512B Reserved

Signature · 1 line
Florian
#33 lore raw

PATCH v3 13/13 PCI: brcmstb: Add bcm7211, bcm7216, bcm7445, bcm7278 to match list

In reply to Jim Quinlan

Now that the support is in place with previous commits, we add several chips that use the BrcmSTB driver.

---
 drivers/pci/controller/pcie-brcmstb.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/pci/controller/pcie-brcmstb.c +4 -0
index 131cf0a51398..22dbecb5403c 100644
@@ -1189,6 +1189,10 @@ static int brcm_pcie_remove(struct platform_device *pdev)
 
 static const struct of_device_id brcm_pcie_match[] = {
 	{ .compatible = "brcm,bcm2711-pcie", .data = &bcm2711_cfg },
+	{ .compatible = "brcm,bcm7211-pcie", .data = &generic_cfg },
+	{ .compatible = "brcm,bcm7278-pcie", .data = &bcm7278_cfg },
+	{ .compatible = "brcm,bcm7216-pcie", .data = &bcm7278_cfg },
+	{ .compatible = "brcm,bcm7445-pcie", .data = &generic_cfg },
 	{},
 };
 
Signature · 1 line
2.17.1