Skip to contentSkip to conversation

PCI: add DT bindings for Cortina Gemini PCI Host Bridge

PATCH Superseded Linus Walleij 5 participants 27 messages Jan 28, 2017 – Feb 27, 2017

Subsystems (6)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

This adds device tree bindings for the Cortina Systems Gemini PCI Host Bridge.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc devicetree at vger.kernel.org

---
This can be merged to the PCI tree whenever it is considered fine for inclusion.

---
 .../devicetree/bindings/pci/cortina,gemini-pci.txt | 64 ++++++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +64 -0
new file mode 100644
index 000000000000..e3090d995e1e
@@ -0,0 +1,64 @@
+* Cortina Systems Gemini PCI Host Bridge
+
+Mandatory properties:
+
+- compatible: should be "cortina,gemini-pci"
+- reg: memory base and size for the host bridge
+- interrupts: the four PCI interrupts
+- #address-cells: set to <3>
+- #size-cells: set to <2>
+- #interrupt-cells: set to <1>
+- bus-range: set to <0x00 0x00> (only root bus)
+- device_type, set to "pci"
+- ranges: see pci.txt
+- interrupt-map-mask: see pci.txt
+- interrupt-map: see pci.txt
+
+Mandatory subnodes:
+- One node reprenting the interrupt-controller inside the host bridge
+  with the following mandatory properties:
+  - interrupt-controller: see interrupt-controller/interrupts.txt
+  - #address-cells: set to <0>
+  - #interrupt-cells: set to <1>
+
+Example:
+
+pci at 50000000 {
+	compatible = "cortina,gemini-pci";
+	reg = <0x50000000 0x100>;
+	interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
+			<26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
+			<27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
+			<28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+	#address-cells = <3>;
+	#size-cells = <2>;
+	#interrupt-cells = <1>;
+
+	bus-range = <0x00 0x00>; /* Only root bus */
+	ranges = /* 1MiB I/O space 0x50000000-0x500fffff */
+		 <0x01000000 0 0          0x50000000 0 0x00100000>,
+		 /* 128MiB non-prefetchable memory 0x58000000-0x5fffffff */
+		 <0x02000000 0 0x58000000 0x58000000 0 0x08000000>;
+	interrupt-map-mask = <0xff00 0 0 7>;
+	interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+			<0x4900 0 0 2 &pci_intc 1>,
+			<0x4a00 0 0 3 &pci_intc 2>,
+			<0x4b00 0 0 4 &pci_intc 3>,
+			<0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+			<0x5100 0 0 2 &pci_intc 1>,
+			<0x5200 0 0 3 &pci_intc 2>,
+			<0x5300 0 0 4 &pci_intc 3>,
+			<0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+			<0x5900 0 0 2 &pci_intc 1>,
+			<0x5a00 0 0 3 &pci_intc 2>,
+			<0x5b00 0 0 4 &pci_intc 3>,
+			<0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+			<0x6100 0 0 2 &pci_intc 1>,
+			<0x6200 0 0 3 &pci_intc 2>,
+			<0x6300 0 0 4 &pci_intc 3>;
+	pci_intc: interrupt-controller {
+		interrupt-controller;
+		#address-cells = <0>;
+		#interrupt-cells = <1>;
+	};
+};
Signature · 1 line
2.9.3
#1 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

This adds a host bridge driver for the Cortina Systems Gemini SoC (SL3516) PCI Host Bridge.

This code is inspired by the out-of-tree OpenWRT patch and then extensively rewritten for device tree and using the modern helpers to cut down and modernize the code to all new PCI frameworks.

Tested on the ITian Square One SQ201 NAS with the following result in the boot log (trimmed to relevant parts):

OF: PCI: host bridge /pci at 50000000 ranges:
OF: PCI: IO 0x50000000..0x500fffff -> 0x00000000
OF: PCI: MEM 0x58000000..0x5fffffff -> 0x58000000
gemini-pci 50000000.pci: PCI host bridge to bus 0000:00
pci_bus 0000:00: root bus resource [bus 00]
pci_bus 0000:00: root bus resource [io 0x0000-0xfffff]
pci_bus 0000:00: root bus resource [mem 0x58000000-0x5fffffff]
pci 0000:00:00.0: [159b:4321] type 00 class 0x060000
pci 0000:00:09.0: [1106:3038] type 00 class 0x0c0300
pci 0000:00:09.0: reg 0x20: [io 0xfce0-0xfcff]
pci 0000:00:09.0: supports D1 D2
pci 0000:00:09.0: PME# supported from D0 D1 D2 D3hot D3cold
pci 0000:00:09.1: [1106:3038] type 00 class 0x0c0300
pci 0000:00:09.1: reg 0x20: [io 0xfce0-0xfcff]
pci 0000:00:09.1: supports D1 D2
pci 0000:00:09.1: PME# supported from D0 D1 D2 D3hot D3cold
pci 0000:00:09.2: [1106:3104] type 00 class 0x0c0320
pci 0000:00:09.2: reg 0x10: [mem 0x00000000-0x000000ff]
pci 0000:00:09.2: supports D1 D2
pci 0000:00:09.2: PME# supported from D0 D1 D2 D3hot D3cold
pci 0000:00:0c.0: [1814:0301] type 00 class 0x028000
pci 0000:00:0c.0: reg 0x10: [mem 0x58000000-0x58007fff]
PCI: bus0: Fast back to back transfers disabled
gemini-pci 50000000.pci: clear all IRQs
gemini-pci 50000000.pci: setting up PCI DMA
pci 0000:00:00.0: of_irq_parse_pci() failed with rc=-22
pci 0000:00:0c.0: BAR 0: assigned [mem 0x58000000-0x58007fff]
pci 0000:00:09.2: BAR 0: assigned [mem 0x58008000-0x580080ff]
pci 0000:00:09.0: BAR 4: assigned [io 0x0400-0x041f]
pci 0000:00:09.1: BAR 4: assigned [io 0x0420-0x043f]
pci 0000:00:09.0: enabling device (0140 -> 0141)
pci 0000:00:09.0: HCRESET not completed yet!
pci 0000:00:09.1: enabling device (0140 -> 0141)
pci 0000:00:09.1: HCRESET not completed yet!
pci 0000:00:09.2: enabling device (0140 -> 0142)
ieee80211 phy0: rt2x00_set_chip: Info - Chipset detected - rt: 2561, rf: 0003, rev: 000c ieee80211 phy0: Selected rate control algorithm 'minstrel_ht'
ehci_hcd: USB 2.0 'Enhanced' Host Controller (EHCI) Driver
ehci-pci: EHCI PCI platform driver
ehci-pci 0000:00:09.2: EHCI Host Controller
ehci-pci 0000:00:09.2: new USB bus registered, assigned bus number 1 ehci-pci 0000:00:09.2: irq 146, io mem 0x58008000
ehci-pci 0000:00:09.2: USB 2.0 started, EHCI 1.00
hub 1-0:1.0: USB hub found
hub 1-0:1.0: 4 ports detected
uhci_hcd: USB Universal Host Controller Interface driver
uhci_hcd 0000:00:09.0: UHCI Host Controller
uhci_hcd 0000:00:09.0: new USB bus registered, assigned bus number 2 uhci_hcd 0000:00:09.0: HCRESET not completed yet!
uhci_hcd 0000:00:09.0: irq 144, io base 0x00000400
hub 2-0:1.0: USB hub found
hub 2-0:1.0: config failed, hub doesn't have any ports! (err -19)
uhci_hcd 0000:00:09.1: UHCI Host Controller
uhci_hcd 0000:00:09.1: new USB bus registered, assigned bus number 3 uhci_hcd 0000:00:09.1: HCRESET not completed yet!
uhci_hcd 0000:00:09.1: irq 145, io base 0x00000420
hub 3-0:1.0: USB hub found
hub 3-0:1.0: config failed, hub doesn't have any ports! (err -19)
usb 1-1: new high-speed USB device number 2 using ehci-pci
usb-storage 1-1:1.0: USB Mass Storage device detected
scsi host0: usb-storage 1-1:1.0
scsi 0:0:0:0: Direct-Access USB Flash Disk 1.00 PQ: 0 ANSI: 2
sd 0:0:0:0: [sda] 7900336 512-byte logical blocks: (4.04 GB/3.77 GiB)
sd 0:0:0:0: [sda] Write Protect is off
sd 0:0:0:0: [sda] Mode Sense: 0b 00 00 08
sd 0:0:0:0: [sda] No Caching mode page found
sd 0:0:0:0: [sda] Assuming drive cache: write through
sda: sda1 sda2 sda3
sd 0:0:0:0: [sda] Attached SCSI removable disk

$ lspci
00:00.0 Class 0600: 159b:4321
00:09.2 Class 0c03: 1106:3104
00:09.0 Class 0c03: 1106:3038
00:09.1 Class 0c03: 1106:3038
00:0c.0 Class 0280: 1814:0301

cat /proc/interrupts
CPU0

 19:          0    GEMINI   3 Level     watchdog bark
 30:       4943    GEMINI  14 Edge      Gemini Timer Tick
 33:          0    GEMINI  17 Level     45000000.rtc
 34:        651    GEMINI  18 Level     serial
 66:          0      GPIO  18 Edge      factory reset
144:          0       PCI   0 Edge      uhci_hcd:usb2
145:          0       PCI   1 Edge      uhci_hcd:usb3
146:        160       PCI   2 Edge      ehci_hcd:usb1

Well the EHCI USB hub works fine, I can mount and manage files and the IRQs just keep ticking up.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>

---
This can be merged to the PCI tree whenever it is considered fine for inclusion.

---
 drivers/pci/host/Kconfig      |   7 +
 drivers/pci/host/Makefile     |   1 +
 drivers/pci/host/pci-gemini.c | 375 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 383 insertions(+)
 create mode 100644 drivers/pci/host/pci-gemini.c
drivers/pci/host/Kconfig +7 -0
index 898d2c48239c..e29c2caf3492 100644
@@ -60,6 +60,13 @@ config PCI_EXYNOS
 	select PCIEPORTBUS
 	select PCIE_DW
 
+config PCI_GEMINI
+	bool "Cortina Gemini SL351x PCI roller"
+	depends on ARCH_GEMINI
+	depends on ARM
+	depends on OF
+	default ARCH_GEMINI
+
 config PCI_IMX6
 	bool "Freescale i.MX6 PCIe controller"
 	depends on SOC_IMX6Q
drivers/pci/host/Makefile +1 -0
index bfe3179ae74c..8f007fe7a19d 100644
@@ -2,6 +2,7 @@ obj-$(CONFIG_PCIE_DW) += pcie-designware.o
 obj-$(CONFIG_PCIE_DW_PLAT) += pcie-designware-plat.o
 obj-$(CONFIG_PCI_DRA7XX) += pci-dra7xx.o
 obj-$(CONFIG_PCI_EXYNOS) += pci-exynos.o
+obj-$(CONFIG_PCI_GEMINI) += pci-gemini.o
 obj-$(CONFIG_PCI_IMX6) += pci-imx6.o
 obj-$(CONFIG_PCI_HYPERV) += pci-hyperv.o
 obj-$(CONFIG_PCI_MVEBU) += pci-mvebu.o
drivers/pci/host/pci-gemini.c new file +375 -0

375 lines not on this page · raw message

Signature · 1 line
2.9.3
#2 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Hi Linus,

Looks nice; a couple unused definitions below. It looks like /proc/iomem should make sense, too, since you used devm_request_pci_bus_resources().

Quoted 175 lines · Linus Walleij

This adds a host bridge driver for the Cortina Systems Gemini SoC (SL3516) PCI Host Bridge.

This code is inspired by the out-of-tree OpenWRT patch and then extensively rewritten for device tree and using the modern helpers to cut down and modernize the code to all new PCI frameworks.

Tested on the ITian Square One SQ201 NAS with the following result in the boot log (trimmed to relevant parts):

OF: PCI: host bridge /pci at 50000000 ranges:
OF: PCI: IO 0x50000000..0x500fffff -> 0x00000000
OF: PCI: MEM 0x58000000..0x5fffffff -> 0x58000000
gemini-pci 50000000.pci: PCI host bridge to bus 0000:00
pci_bus 0000:00: root bus resource [bus 00]
pci_bus 0000:00: root bus resource [io 0x0000-0xfffff]
pci_bus 0000:00: root bus resource [mem 0x58000000-0x5fffffff]
pci 0000:00:00.0: [159b:4321] type 00 class 0x060000
pci 0000:00:09.0: [1106:3038] type 00 class 0x0c0300
pci 0000:00:09.0: reg 0x20: [io 0xfce0-0xfcff]
pci 0000:00:09.0: supports D1 D2
pci 0000:00:09.0: PME# supported from D0 D1 D2 D3hot D3cold
pci 0000:00:09.1: [1106:3038] type 00 class 0x0c0300
pci 0000:00:09.1: reg 0x20: [io 0xfce0-0xfcff]
pci 0000:00:09.1: supports D1 D2
pci 0000:00:09.1: PME# supported from D0 D1 D2 D3hot D3cold
pci 0000:00:09.2: [1106:3104] type 00 class 0x0c0320
pci 0000:00:09.2: reg 0x10: [mem 0x00000000-0x000000ff]
pci 0000:00:09.2: supports D1 D2
pci 0000:00:09.2: PME# supported from D0 D1 D2 D3hot D3cold
pci 0000:00:0c.0: [1814:0301] type 00 class 0x028000
pci 0000:00:0c.0: reg 0x10: [mem 0x58000000-0x58007fff]
PCI: bus0: Fast back to back transfers disabled
gemini-pci 50000000.pci: clear all IRQs
gemini-pci 50000000.pci: setting up PCI DMA
pci 0000:00:00.0: of_irq_parse_pci() failed with rc=-22
pci 0000:00:0c.0: BAR 0: assigned [mem 0x58000000-0x58007fff]
pci 0000:00:09.2: BAR 0: assigned [mem 0x58008000-0x580080ff]
pci 0000:00:09.0: BAR 4: assigned [io 0x0400-0x041f]
pci 0000:00:09.1: BAR 4: assigned [io 0x0420-0x043f]
pci 0000:00:09.0: enabling device (0140 -> 0141)
pci 0000:00:09.0: HCRESET not completed yet!
pci 0000:00:09.1: enabling device (0140 -> 0141)
pci 0000:00:09.1: HCRESET not completed yet!
pci 0000:00:09.2: enabling device (0140 -> 0142)
ieee80211 phy0: rt2x00_set_chip: Info - Chipset detected - rt: 2561, rf: 0003, rev: 000c ieee80211 phy0: Selected rate control algorithm 'minstrel_ht'
ehci_hcd: USB 2.0 'Enhanced' Host Controller (EHCI) Driver
ehci-pci: EHCI PCI platform driver
ehci-pci 0000:00:09.2: EHCI Host Controller
ehci-pci 0000:00:09.2: new USB bus registered, assigned bus number 1 ehci-pci 0000:00:09.2: irq 146, io mem 0x58008000
ehci-pci 0000:00:09.2: USB 2.0 started, EHCI 1.00
hub 1-0:1.0: USB hub found
hub 1-0:1.0: 4 ports detected
uhci_hcd: USB Universal Host Controller Interface driver
uhci_hcd 0000:00:09.0: UHCI Host Controller
uhci_hcd 0000:00:09.0: new USB bus registered, assigned bus number 2 uhci_hcd 0000:00:09.0: HCRESET not completed yet!
uhci_hcd 0000:00:09.0: irq 144, io base 0x00000400
hub 2-0:1.0: USB hub found
hub 2-0:1.0: config failed, hub doesn't have any ports! (err -19)
uhci_hcd 0000:00:09.1: UHCI Host Controller
uhci_hcd 0000:00:09.1: new USB bus registered, assigned bus number 3 uhci_hcd 0000:00:09.1: HCRESET not completed yet!
uhci_hcd 0000:00:09.1: irq 145, io base 0x00000420
hub 3-0:1.0: USB hub found
hub 3-0:1.0: config failed, hub doesn't have any ports! (err -19)
usb 1-1: new high-speed USB device number 2 using ehci-pci
usb-storage 1-1:1.0: USB Mass Storage device detected
scsi host0: usb-storage 1-1:1.0
scsi 0:0:0:0: Direct-Access USB Flash Disk 1.00 PQ: 0 ANSI: 2
sd 0:0:0:0: [sda] 7900336 512-byte logical blocks: (4.04 GB/3.77 GiB)
sd 0:0:0:0: [sda] Write Protect is off
sd 0:0:0:0: [sda] Mode Sense: 0b 00 00 08
sd 0:0:0:0: [sda] No Caching mode page found
sd 0:0:0:0: [sda] Assuming drive cache: write through
sda: sda1 sda2 sda3
sd 0:0:0:0: [sda] Attached SCSI removable disk

$ lspci
00:00.0 Class 0600: 159b:4321
00:09.2 Class 0c03: 1106:3104
00:09.0 Class 0c03: 1106:3038
00:09.1 Class 0c03: 1106:3038
00:0c.0 Class 0280: 1814:0301

cat /proc/interrupts
CPU0

 19:          0    GEMINI   3 Level     watchdog bark
 30:       4943    GEMINI  14 Edge      Gemini Timer Tick
 33:          0    GEMINI  17 Level     45000000.rtc
 34:        651    GEMINI  18 Level     serial
 66:          0      GPIO  18 Edge      factory reset
144:          0       PCI   0 Edge      uhci_hcd:usb2
145:          0       PCI   1 Edge      uhci_hcd:usb3
146:        160       PCI   2 Edge      ehci_hcd:usb1

Well the EHCI USB hub works fine, I can mount and manage files and the IRQs just keep ticking up.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>

---
This can be merged to the PCI tree whenever it is considered fine for inclusion.

---
 drivers/pci/host/Kconfig      |   7 +
 drivers/pci/host/Makefile     |   1 +
 drivers/pci/host/pci-gemini.c | 375 ++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 383 insertions(+)
 create mode 100644 drivers/pci/host/pci-gemini.c
drivers/pci/host/Kconfig +7 -0
index 898d2c48239c..e29c2caf3492 100644
@@ -60,6 +60,13 @@ config PCI_EXYNOS
 	select PCIEPORTBUS
 	select PCIE_DW
 
+config PCI_GEMINI
+	bool "Cortina Gemini SL351x PCI roller"
+	depends on ARCH_GEMINI
+	depends on ARM
+	depends on OF
+	default ARCH_GEMINI
+
 config PCI_IMX6
 	bool "Freescale i.MX6 PCIe controller"
 	depends on SOC_IMX6Q
drivers/pci/host/Makefile +1 -0
index bfe3179ae74c..8f007fe7a19d 100644
@@ -2,6 +2,7 @@ obj-$(CONFIG_PCIE_DW) += pcie-designware.o
 obj-$(CONFIG_PCIE_DW_PLAT) += pcie-designware-plat.o
 obj-$(CONFIG_PCI_DRA7XX) += pci-dra7xx.o
 obj-$(CONFIG_PCI_EXYNOS) += pci-exynos.o
+obj-$(CONFIG_PCI_GEMINI) += pci-gemini.o
 obj-$(CONFIG_PCI_IMX6) += pci-imx6.o
 obj-$(CONFIG_PCI_HYPERV) += pci-hyperv.o
 obj-$(CONFIG_PCI_MVEBU) += pci-mvebu.o
drivers/pci/host/pci-gemini.c new file +22 -0
new file mode 100644
index 000000000000..7051dd992114
@@ -0,0 +1,375 @@
+/*
+ * Support for Gemini PCI Controller
+ *
+ * Copyright (C) 2017 Linus Walleij <[email protected]>
+ *
+ * Based on the out-of-tree OpenWRT patch:
+ * Copyright (C) 2009 Janos Laube <[email protected]>
+ * Copyright (C) 2009 Paulius Zaleckas <[email protected]>
+ * Based on SL2312 PCI controller code
+ * Storlink (C) 2003
+ */
+
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/of_address.h>
+#include <linux/of_pci.h>
+#include <linux/pci.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/irqdomain.h>

drivers/pci/host/pci-gemini.c line 32

Quoted 10 lines
drivers/pci/host/pci-gemini.c new file +10 -0
+#include <linux/irqchip/chained_irq.h>
+#include <linux/bitops.h>
+#include <linux/irq.h>
+#include <linux/spinlock.h>
+
+#define GEMINI_PCI_IOSIZE_1M		0x0000
+
+#define GEMINI_PCI_PMC			0x40
+#define GEMINI_PCI_PMCSR		0x44
+#define GEMINI_PCI_CTRL1		0x48

Above three definitions unused.

Quoted 14 lines
drivers/pci/host/pci-gemini.c new file +14 -0
+#define GEMINI_PCI_CTRL2		0x4C
+#define GEMINI_PCI_MEM1_BASE_SIZE	0x50
+#define GEMINI_PCI_MEM2_BASE_SIZE	0x54
+#define GEMINI_PCI_MEM3_BASE_SIZE	0x58
+
+#define PCI_CTRL2_INTSTS_SHIFT		28
+#define PCI_CTRL2_INTMASK_SHIFT		22
+
+#define GEMINI_PCI_DMA_MASK		0xFFF00000
+#define GEMINI_PCI_DMA_MEM1_BASE	0x00000000
+#define GEMINI_PCI_DMA_MEM2_BASE	0x00000000
+#define GEMINI_PCI_DMA_MEM3_BASE	0x00000000
+#define GEMINI_PCI_DMA_MEM1_SIZE	7
+#define GEMINI_PCI_DMA_MEM2_SIZE	6

drivers/pci/host/pci-gemini.c line 56

Quoted 10 lines
drivers/pci/host/pci-gemini.c new file +10 -0
+#define GEMINI_PCI_DMA_MEM3_SIZE	6
+
+#define PCI_CONF_ENABLE		BIT(31)
+#define PCI_CONF_WHERE(r)	((r) & 0xFC)
+#define PCI_CONF_BUS(b)		(((b) & 0xFF) << 16)
+#define PCI_CONF_DEVICE(d)	(((d) & 0x1F) << 11)
+#define PCI_CONF_FUNCTION(f)	(((f) & 0x07) << 8)
+
+#define PCI_IOSIZE	0x00
+#define PCI_PROT	0x04

Unused.

drivers/pci/host/pci-gemini.c line 58

Quoted 2 lines
drivers/pci/host/pci-gemini.c new file +2 -0
+#define PCI_CTRL	0x08
+#define PCI_SOFTRST	0x10

Unused.

Quoted 324 lines
drivers/pci/host/pci-gemini.c new file +317 -0
+#define PCI_CONFIG	0x28
+#define PCI_DATA	0x2C
+
+struct gemini_pci {
+	struct device *dev;
+	void __iomem *base;
+	struct irq_domain *irqdomain;
+	spinlock_t lock;
+	struct pci_bus *bus;
+};
+
+static int gemini_pci_read_config(struct pci_bus *bus, unsigned int fn,
+				  int config, int size, u32 *value)
+{
+	struct gemini_pci *p = bus->sysdata;
+	unsigned long irq_flags;
+
+	spin_lock_irqsave(&p->lock, irq_flags);
+
+	writel(PCI_CONF_BUS(bus->number) |
+			PCI_CONF_DEVICE(PCI_SLOT(fn)) |
+			PCI_CONF_FUNCTION(PCI_FUNC(fn)) |
+			PCI_CONF_WHERE(config) |
+			PCI_CONF_ENABLE,
+			p->base + PCI_CONFIG);
+
+	*value = readl(p->base + PCI_DATA);
+
+	if (size == 1)
+		*value = (*value >> (8 * (config & 3))) & 0xFF;
+	else if (size == 2)
+		*value = (*value >> (8 * (config & 3))) & 0xFFFF;
+
+	spin_unlock_irqrestore(&p->lock, irq_flags);
+
+	dev_dbg(&bus->dev,
+		"[read]  slt: %.2d, fnc: %d, cnf: 0x%.2X, val (%d bytes): 0x%.8X\n",
+		PCI_SLOT(fn), PCI_FUNC(fn), config, size, *value);
+
+	return PCIBIOS_SUCCESSFUL;
+}
+
+static int gemini_pci_write_config(struct pci_bus *bus, unsigned int fn,
+				   int config, int size, u32 value)
+{
+	struct gemini_pci *p = bus->sysdata;
+	unsigned long irq_flags = 0;
+	int ret = PCIBIOS_SUCCESSFUL;
+
+	dev_dbg(&bus->dev,
+		"[write] slt: %.2d, fnc: %d, cnf: 0x%.2X, val (%d bytes): 0x%.8X\n",
+		PCI_SLOT(fn), PCI_FUNC(fn), config, size, value);
+
+	spin_lock_irqsave(&p->lock, irq_flags);
+
+	writel(PCI_CONF_BUS(bus->number) |
+			PCI_CONF_DEVICE(PCI_SLOT(fn)) |
+			PCI_CONF_FUNCTION(PCI_FUNC(fn)) |
+			PCI_CONF_WHERE(config) |
+			PCI_CONF_ENABLE,
+			p->base + PCI_CONFIG);
+
+	switch (size) {
+	case 4:
+		writel(value, p->base + PCI_DATA);
+		break;
+	case 2:
+		writew(value, p->base + PCI_DATA + (config & 3));
+		break;
+	case 1:
+		writeb(value, p->base + PCI_DATA + (config & 3));
+		break;
+	default:
+		ret = PCIBIOS_BAD_REGISTER_NUMBER;
+	}
+
+	spin_unlock_irqrestore(&p->lock, irq_flags);
+
+	return ret;
+}
+
+static struct pci_ops gemini_pci_ops = {
+	.read	= gemini_pci_read_config,
+	.write	= gemini_pci_write_config,
+};
+
+static void gemini_pci_ack_irq(struct irq_data *d)
+{
+	struct gemini_pci *p = irq_data_get_irq_chip_data(d);
+	unsigned int reg;
+
+	gemini_pci_read_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, &reg);
+	reg &= ~(0xF << PCI_CTRL2_INTSTS_SHIFT);
+	reg |= BIT(irqd_to_hwirq(d) + PCI_CTRL2_INTSTS_SHIFT);
+	gemini_pci_write_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, reg);
+}
+
+static void gemini_pci_mask_irq(struct irq_data *d)
+{
+	struct gemini_pci *p = irq_data_get_irq_chip_data(d);
+	unsigned int reg;
+
+	gemini_pci_read_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, &reg);
+	reg &= ~((0xF << PCI_CTRL2_INTSTS_SHIFT)
+		 | BIT(irqd_to_hwirq(d) + PCI_CTRL2_INTMASK_SHIFT));
+	gemini_pci_write_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, reg);
+}
+
+static void gemini_pci_unmask_irq(struct irq_data *d)
+{
+	struct gemini_pci *p = irq_data_get_irq_chip_data(d);
+	unsigned int reg;
+
+	gemini_pci_read_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, &reg);
+	reg &= ~(0xF << PCI_CTRL2_INTSTS_SHIFT);
+	reg |= BIT(irqd_to_hwirq(d) + PCI_CTRL2_INTMASK_SHIFT);
+	gemini_pci_write_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, reg);
+}
+
+static void gemini_pci_irq_handler(struct irq_desc *desc)
+{
+	struct gemini_pci *p = irq_desc_get_handler_data(desc);
+	struct irq_chip *irqchip = irq_desc_get_chip(desc);
+	unsigned int irq_stat, reg, i;
+
+	gemini_pci_read_config(p->bus, 0, GEMINI_PCI_CTRL2, 4, &reg);
+	irq_stat = reg >> PCI_CTRL2_INTSTS_SHIFT;
+
+	chained_irq_enter(irqchip, desc);
+
+	for (i = 0; i < 4; i++) {
+		if ((irq_stat & BIT(i)) == 0)
+			continue;
+		generic_handle_irq(irq_find_mapping(p->irqdomain, i));
+	}
+
+	chained_irq_exit(irqchip, desc);
+}
+
+static struct irq_chip gemini_pci_irq_chip = {
+	.name = "PCI",
+	.irq_ack = gemini_pci_ack_irq,
+	.irq_mask = gemini_pci_mask_irq,
+	.irq_unmask = gemini_pci_unmask_irq,
+};
+
+static int gemini_pci_irq_map(struct irq_domain *domain, unsigned int irq,
+			      irq_hw_number_t hwirq)
+{
+	irq_set_chip_and_handler(irq, &gemini_pci_irq_chip, handle_level_irq);
+	irq_set_chip_data(irq, domain->host_data);
+
+	return 0;
+}
+
+static const struct irq_domain_ops gemini_pci_irqdomain_ops = {
+	.map = gemini_pci_irq_map,
+};
+
+static int gemini_pci_setup_irq(struct gemini_pci *p, int irq)
+{
+	struct device_node *intc = of_get_next_child(p->dev->of_node, NULL);
+	int i;
+
+	if (!intc) {
+		dev_err(p->dev, "missing child interrupt-controller node\n");
+		return -EINVAL;
+	}
+
+	p->irqdomain = irq_domain_add_linear(intc, 4,
+					     &gemini_pci_irqdomain_ops,
+					     p);
+	if (!p->irqdomain) {
+		dev_err(p->dev, "failed to create Gemini PCI IRQ domain\n");
+		return -EINVAL;
+	}
+
+	irq_set_chained_handler_and_data(irq, gemini_pci_irq_handler, p);
+
+	for (i = 0; i < 4; i++)
+		irq_create_mapping(p->irqdomain, i);
+
+	return 0;
+}
+
+static int gemini_pci_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct resource *regs;
+	resource_size_t io_base;
+	struct resource_entry *win;
+	struct gemini_pci *p;
+	struct resource *mem;
+	struct resource *io;
+	struct pci_bus *bus;
+	int irq;
+	int ret;
+	u32 val;
+	LIST_HEAD(res);
+
+	p = devm_kzalloc(dev, sizeof(*p), GFP_KERNEL);
+	if (!p)
+		return -ENOMEM;
+
+	p->dev = dev;
+	spin_lock_init(&p->lock);
+
+	regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	p->base = devm_ioremap_resource(dev, regs);
+	if (IS_ERR(p->base))
+		return PTR_ERR(p->base);
+
+	ret = of_pci_get_host_bridge_resources(dev->of_node, 0, 0xff,
+					       &res, &io_base);
+	if (ret)
+		return ret;
+
+	ret = devm_request_pci_bus_resources(dev, &res);
+	if (ret)
+		return ret;
+
+	/* No clue what these do */
+	pcibios_min_io = 0x100;
+	pcibios_min_mem = 0;
+
+	/* setup I/O space to 1MB size */
+	writel(GEMINI_PCI_IOSIZE_1M, p->base + PCI_IOSIZE);
+
+	/* setup hostbridge */
+	val = readl(p->base + PCI_CTRL);
+	val |= PCI_COMMAND_IO;
+	val |= PCI_COMMAND_MEMORY;
+	val |= PCI_COMMAND_MASTER;
+	writel(val, p->base + PCI_CTRL);
+
+	/* Get the I/O and memory ranges from DT */
+	resource_list_for_each_entry(win, &res) {
+		switch (resource_type(win->res)) {
+		case IORESOURCE_IO:
+			io = win->res;
+			io->name = "Gemini PCI I/O";
+			ret = pci_remap_iospace(io, io_base);
+			if (ret) {
+				dev_warn(dev, "error %d: failed to map resource %pR\n",
+					 ret, io);
+				continue;
+			}
+			break;
+		case IORESOURCE_MEM:
+			mem = win->res;
+			mem->name = "Gemini PCI MEM";
+			break;
+		case IORESOURCE_BUS:
+			break;
+		default:
+			break;
+		}
+	}
+
+	bus = pci_scan_root_bus(&pdev->dev, 0, &gemini_pci_ops, p, &res);
+	if (!bus)
+		return -ENOMEM;
+	p->bus = bus;
+
+	dev_info(dev, "clear all IRQs\n");
+
+	/* Mask and clear all interrupts */
+	gemini_pci_write_config(bus, 0, GEMINI_PCI_CTRL2 + 2, 2, 0xF000);
+
+	/* IRQ - all PCI IRQs cascade off this one */
+	irq = platform_get_irq(pdev, 0);
+	if (!irq) {
+		dev_err(dev, "failed to get IRQ\n");
+		return -EINVAL;
+	}
+
+	ret = gemini_pci_setup_irq(p, irq);
+	if (ret) {
+		dev_err(dev, "failed to setup IRQ\n");
+		return ret;
+	}
+
+	dev_info(dev, "setting up PCI DMA\n");
+	val = (GEMINI_PCI_DMA_MEM1_BASE & GEMINI_PCI_DMA_MASK)
+		| (GEMINI_PCI_DMA_MEM1_SIZE << 16);
+	gemini_pci_write_config(bus, 0, GEMINI_PCI_MEM1_BASE_SIZE, 4, val);
+	val = (GEMINI_PCI_DMA_MEM2_BASE & GEMINI_PCI_DMA_MASK)
+		| (GEMINI_PCI_DMA_MEM2_SIZE << 16);
+	gemini_pci_write_config(bus, 0, GEMINI_PCI_MEM2_BASE_SIZE, 4, val);
+	val = (GEMINI_PCI_DMA_MEM3_BASE & GEMINI_PCI_DMA_MASK)
+		| (GEMINI_PCI_DMA_MEM3_SIZE << 16);
+	gemini_pci_write_config(bus, 0, GEMINI_PCI_MEM3_BASE_SIZE, 4, val);
+
+	pci_fixup_irqs(pci_common_swizzle, of_irq_parse_and_map_pci);
+	pci_bus_assign_resources(bus);
+	pci_assign_unassigned_bus_resources(bus);
+	pci_bus_add_devices(bus);
+	pci_free_resource_list(&res);
+
+	return 0;
+}
+
+static const struct of_device_id gemini_pci_of_match[] = {
+	{
+		.compatible = "cortina,gemini-pci",
+	},
+	{},
+};
+
+static struct platform_driver gemini_pci_driver = {
+	.driver = {
+		.name = "gemini-pci",
+		.of_match_table = of_match_ptr(gemini_pci_of_match),
+	},
+	.probe  = gemini_pci_probe,
+};
+builtin_platform_driver(gemini_pci_driver);
Signature · 6 lines
2.9.3

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#3 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Quoted 1 line · Bjorn Helgaas

Looks nice; a couple unused definitions below.

Just so I do things right: what is your policy on unused defines? I'm asking because the defines in this case is pretty much the documentation... there is no datasheet for this IP core. I usually like to keep the defines around so people can have them as a hint if they want to tinker and hack the driver.

Would you be OK with keeping them if I comment them out?

I'll repost the driver soon, including all things asked to be fixed up by you & Arnd.

Yours,
Linus Walleij

#4 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Quoted 11 lines · Linus Walleij
Quoted 1 line · Bjorn Helgaas

Looks nice; a couple unused definitions below.

Just so I do things right: what is your policy on unused defines? I'm asking because the defines in this case is pretty much the documentation... there is no datasheet for this IP core. I usually like to keep the defines around so people can have them as a hint if they want to tinker and hack the driver.

Would you be OK with keeping them if I comment them out?

It's fine if you keep them; don't even bother commenting them out.

If there's a datasheet, especially a public one like the PCI specs (I know even those aren't completely free), I have been removing unused defines simply because they can't be tested and there may be transcription errors. But I'm starting to think that might be too aggressive.

Bjorn

#5 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

In reply to Linus Walleij
Quoted 13 lines · CET Linus Walleij
drivers/pci/host/pci-gemini.c new file +13 -0
+static int gemini_pci_read_config(struct pci_bus *bus, unsigned int fn,
+				  int config, int size, u32 *value)
+{
+	struct gemini_pci *p = bus->sysdata;
+	unsigned long irq_flags;
+
+	spin_lock_irqsave(&p->lock, irq_flags);
+
+	writel(PCI_CONF_BUS(bus->number) |
+			PCI_CONF_DEVICE(PCI_SLOT(fn)) |
+			PCI_CONF_FUNCTION(PCI_FUNC(fn)) |
+			PCI_CONF_WHERE(config) |
+			PCI_CONF_ENABLE,

drivers/pci/host/pci-gemini.c line 92

Quoted 10 lines
drivers/pci/host/pci-gemini.c new file +10 -0
+			p->base + PCI_CONFIG);
+
+	*value = readl(p->base + PCI_DATA);
+
+	if (size == 1)
+		*value = (*value >> (8 * (config & 3))) & 0xFF;
+	else if (size == 2)
+		*value = (*value >> (8 * (config & 3))) & 0xFFFF;
+
+	spin_unlock_irqrestore(&p->lock, irq_flags);

The read_config/write_config functions are called under a spinlock, no need for another one.

Quoted 7 lines
drivers/pci/host/pci-gemini.c new file +7 -0
+	regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	p->base = devm_ioremap_resource(dev, regs);
+	if (IS_ERR(p->base))
+		return PTR_ERR(p->base);
+
+	ret = of_pci_get_host_bridge_resources(dev->of_node, 0, 0xff,
+					       &res, &io_base);

drivers/pci/host/pci-gemini.c line 282

Quoted 10 lines
drivers/pci/host/pci-gemini.c new file +10 -0
+	if (ret)
+		return ret;
+
+	ret = devm_request_pci_bus_resources(dev, &res);
+	if (ret)
+		return ret;
+
+	/* No clue what these do */
+	pcibios_min_io = 0x100;
+	pcibios_min_mem = 0;

Don't touch these

drivers/pci/host/pci-gemini.c line 285

Quoted 2 lines
drivers/pci/host/pci-gemini.c new file +2 -0
+	/* setup I/O space to 1MB size */
+	writel(GEMINI_PCI_IOSIZE_1M, p->base + PCI_IOSIZE);

Maybe get the size from the resource instead? Note that pci_remap_iospace only registers 64K of I/O space, so you could also use that.

Quoted 22 lines
drivers/pci/host/pci-gemini.c new file +22 -0
+	/* setup hostbridge */
+	val = readl(p->base + PCI_CTRL);
+	val |= PCI_COMMAND_IO;
+	val |= PCI_COMMAND_MEMORY;
+	val |= PCI_COMMAND_MASTER;
+	writel(val, p->base + PCI_CTRL);
+
+	/* Get the I/O and memory ranges from DT */
+	resource_list_for_each_entry(win, &res) {
+		switch (resource_type(win->res)) {
+		case IORESOURCE_IO:
+			io = win->res;
+			io->name = "Gemini PCI I/O";
+			ret = pci_remap_iospace(io, io_base);
+			if (ret) {
+				dev_warn(dev, "error %d: failed to map resource %pR\n",
+					 ret, io);
+				continue;
+			}
+			break;
+		case IORESOURCE_MEM:
+			mem = win->res;

drivers/pci/host/pci-gemini.c line 318

Quoted 10 lines
drivers/pci/host/pci-gemini.c new file +10 -0
+			mem->name = "Gemini PCI MEM";
+			break;
+		case IORESOURCE_BUS:
+			break;
+		default:
+			break;
+		}
+	}
+
+	bus = pci_scan_root_bus(&pdev->dev, 0, &gemini_pci_ops, p, &res);

Can you try using the new pci_register_host_bridge() API?

Arnd

#6 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Quoted 7 lines · Arnd Bergmann

drivers/pci/host/pci-gemini.c line 282

Quoted 3 lines · CET Linus Walleij
drivers/pci/host/pci-gemini.c new file +3 -0
+     /* No clue what these do */
+     pcibios_min_io = 0x100;
+     pcibios_min_mem = 0;

Don't touch these

OK I have a clue why this is there now, atleast the first one.

The first 0x100 in the IOspace is actually configuration registers for the bridge. That is why we have this:

reg = <0x50000000 0x100>;
(...)
/* PCI ranges mappings */

                        ranges = /* 1MiB I/O space 0x50000000-0x500fffff */
                         <0x01000000 0 0          0x50000000 0 0x00100000>,

This is in all the vendor code I have located too.

So the pcibios_min_io is manipulated to avoid touching that sensitive area. But I also see that arch/arm/mm/iomap.c sets it to 0x1000, according to the commit because you said it's the only valid value :D

I tried setting the IO range to <0x50000100 0x000FFF00> instead of <0x50000000 0x00100000>
but predictably that doesn't work. Maybe it should, I don't know really.

Yours,
Linus Walleij

#7 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Quoted 7 lines · CET Linus Walleij
Quoted 6 lines · Arnd Bergmann
Quoted 3 lines · CET Linus Walleij
drivers/pci/host/pci-gemini.c new file +3 -0
+     /* No clue what these do */
+     pcibios_min_io = 0x100;
+     pcibios_min_mem = 0;
Quoted 6 lines
Quoted 1 line

Don't touch these

OK I have a clue why this is there now, atleast the first one.

The first 0x100 in the IOspace is actually configuration registers for the bridge. That is why we have this:

I see. It's normal to have the PCI config space done through ports 0cf8-0cff, but apparently this one uses other ports in the same range.

Quoted 13 lines

reg = <0x50000000 0x100>;
(...)
/* PCI ranges mappings */

                        ranges = /* 1MiB I/O space 0x50000000-0x500fffff */
                         <0x01000000 0 0          0x50000000 0 0x00100000>,

This is in all the vendor code I have located too.

So the pcibios_min_io is manipulated to avoid touching that sensitive area. But I also see that arch/arm/mm/iomap.c sets it to 0x1000, according to the commit because you said it's the only valid value :D

Quoted 4 lines

I tried setting the IO range to <0x50000100 0x000FFF00> instead of <0x50000000 0x00100000>
but predictably that doesn't work. Maybe it should, I don't know really.

It should work in theory, but then you'd have to update io_offset accordingly, and it would be a bit confusing.

Did you notice my other comment below (quoting from my own message)?

Quoted 4 lines

drivers/pci/host/pci-gemini.c line 318

Quoted 2 lines
drivers/pci/host/pci-gemini.c new file +2 -0
+
+     bus = pci_scan_root_bus(&pdev->dev, 0, &gemini_pci_ops, p, &res);

Can you try using the new pci_register_host_bridge() API?

Arnd

#8 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Quoted 9 lines · Arnd Bergmann

On Saturday, February 4, 2017 7:43:15 PM CET Linus Walleij wrote:

Did you notice my other comment below (quoting from my own message)?

Quoted 4 lines

drivers/pci/host/pci-gemini.c line 318

Quoted 2 lines
drivers/pci/host/pci-gemini.c new file +2 -0
+
+     bus = pci_scan_root_bus(&pdev->dev, 0, &gemini_pci_ops, p, &res);

Can you try using the new pci_register_host_bridge() API?

Yes. It crashes badly when I try it so I'm still trying to figure this out... If I do I will send an iteration.

Yours,
Linus Walleij

maintainer
#9 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

In reply to Linus Walleij

Hi Linus

while debugging my boot issues on my NAS 4220, I saw this warning

irq: type mismatch, failed to map hwirq-26 for /soc/interrupt-controller at 48000000!

We need to use the use the use the PCI IRQ from gemini as root for the irq domain, not the next one in DT.

drivers/pci/host/pci-gemini.c +1 -2
index 7051dd992114..bb564a8e6379 100644
@@ -217,7 +217,6 @@ static const struct irq_domain_ops gemini_pci_irqdomain_ops = {
 
 static int gemini_pci_setup_irq(struct gemini_pci *p, int irq)
 {
-	struct device_node *intc = of_get_next_child(p->dev->of_node, NULL);
 	int i;
 
 	if (!intc) {
@@ -225,7 +224,7 @@ static int gemini_pci_setup_irq(struct gemini_pci *p, int irq)
 		return -EINVAL;
 	}
 
-	p->irqdomain = irq_domain_add_linear(intc, 4,
+	p->irqdomain = irq_domain_add_linear(p->dev->of_node, 4,
 					     &gemini_pci_irqdomain_ops,
 					     p);
 	if (!p->irqdomain) {
Signature · 1 line
2.11.0
#10 lore raw

PATCH 2/4 PCI: add driver for Cortina Gemini Host Bridge

Quoted 4 lines · Hans Ulli Kroll

irq: type mismatch, failed to map hwirq-26 for /soc/interrupt-controller at 48000000!

We need to use the use the use the PCI IRQ from gemini as root for the irq domain, not the next one in DT.

Hm I wonder if that is because I need to set the parent for the intcon in the device tree as Arnd pointed out in the bindings patch?

Yours,
Linus Walleij

#11 lore raw

PATCH 3/4 ARM: gemini: select MIGHT_HAVE_PCI

In reply to Linus Walleij

As we have a PCI driver for the Gemini, we should select MIGHT_HAVE_PCI.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>

---
PCI maintainers: this is FYI only, I will funnel this to the ARM SoC tree once we are done with the PCI driver.

---
 arch/arm/mach-gemini/Kconfig | 1 +
 1 file changed, 1 insertion(+)
arch/arm/mach-gemini/Kconfig +1 -0
index 798abb16bb88..91ea52ccf13c 100644
@@ -5,6 +5,7 @@ menuconfig ARCH_GEMINI
 	select GEMINI_TIMER
 	select GPIO_GEMINI
 	select GPIOLIB
+	select MIGHT_HAVE_PCI
 	select POWER_RESET
 	select POWER_RESET_SYSCON
 	select SERIAL_OF_PLATFORM
Signature · 1 line
2.9.3
#12 lore raw

PATCH 4/4 ARM: dts: add PCI to the Gemini DTSI

In reply to Linus Walleij

The Cortina Gemini has an internal PCI root bus, add this to the device tree.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>

---
PCI maintainers: this is FYI only, I will funnel this to the ARM SoC tree once we are done with the PCI driver.

---
 arch/arm/boot/dts/gemini.dtsi | 45 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)
arch/arm/boot/dts/gemini.dtsi +45 -0
index 405d6cedf409..df5630958038 100644
@@ -99,4 +99,49 @@
 		interrupt-controller;
 		#interrupt-cells = <2>;
 	};
+
+	pci at 50000000 {
+		compatible = "cortina,gemini-pci";
+		reg = <0x50000000 0x100>;
+		interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
+				<26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
+				<27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
+				<28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+		#address-cells = <3>;
+		#size-cells = <2>;
+		#interrupt-cells = <1>;
+
+		bus-range = <0x00 0x00>; /* Only root bus */
+		/* PCI ranges mappings */
+		ranges = /* 1MiB I/O space 0x50000000-0x500fffff */
+			 <0x01000000 0 0          0x50000000 0 0x00100000>,
+			 /* 128MiB non-prefetchable memory 0x58000000-0x5fffffff */
+			 <0x02000000 0 0x58000000 0x58000000 0 0x08000000>;
+
+		interrupt-map-mask = <0xff00 0 0 7>;
+		/*
+		 * The interrupt map is done by sub-device and per-slot.
+		 */
+		interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+				<0x4900 0 0 2 &pci_intc 1>,
+				<0x4a00 0 0 3 &pci_intc 2>,
+				<0x4b00 0 0 4 &pci_intc 3>,
+				<0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+				<0x5100 0 0 2 &pci_intc 1>,
+				<0x5200 0 0 3 &pci_intc 2>,
+				<0x5300 0 0 4 &pci_intc 3>,
+				<0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+				<0x5900 0 0 2 &pci_intc 1>,
+				<0x5a00 0 0 3 &pci_intc 2>,
+				<0x5b00 0 0 4 &pci_intc 3>,
+				<0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+				<0x6100 0 0 2 &pci_intc 1>,
+				<0x6200 0 0 3 &pci_intc 2>,
+				<0x6300 0 0 4 &pci_intc 3>;
+		pci_intc: interrupt-controller {
+			interrupt-controller;
+			#address-cells = <0>;
+			#interrupt-cells = <1>;
+		};
+	};
 };
Signature · 1 line
2.9.3
maintainer
#13 lore raw

PATCH 4/4 ARM: dts: add PCI to the Gemini DTSI

Hi Linus

We need to the remove hwirq 26-28 from DT.
First one will print this warning while boot.

irq: type mismatch, failed to map hwirq-26 for /soc/interrupt-controller at 48000000!

 arch/arm/boot/dts/gemini.dtsi | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)
arch/arm/boot/dts/gemini.dtsi +1 -4
index 11441e577058..0e2c021cd8c5 100644
@@ -114,10 +114,7 @@
 		pci at 50000000 {
 			compatible = "cortina,gemini-pci";
 			reg = <0x50000000 0x100>;
-			interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
-				   	<26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
-					<27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
-					<28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+			interrupts = <8 IRQ_TYPE_LEVEL_HIGH>; /* chained irq PCI A-D */
 			#address-cells = <3>;
 			#size-cells = <2>;
 			#interrupt-cells = <1>;
Signature · 1 line
2.11.0
#14 lore raw

PATCH 4/4 ARM: dts: add PCI to the Gemini DTSI

Quoted 4 lines · Hans Ulli Kroll

We need to the remove hwirq 26-28 from DT.
First one will print this warning while boot.

irq: type mismatch, failed to map hwirq-26 for /soc/interrupt-controller at 48000000!

(...)

arch/arm/boot/dts/gemini.dtsi line 117

Quoted 5 lines
arch/arm/boot/dts/gemini.dtsi +1 -4
-                       interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
-                                       <26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
-                                       <27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
-                                       <28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+                       interrupts = <8 IRQ_TYPE_LEVEL_HIGH>; /* chained irq PCI A-D */

Sure I can remove them ... just found them in the irqs.h file and thought it made sense to add them. I'll just cut them.

Since there is actually an internal IRQ controller in the host controller cascading the four PCI child IRQs that we model as an irqchip, I don't really see why they have these "PCI B-D" IRQs... anyone has a guess?

Yours,
Linus Walleij

maintainer
#15 lore raw

PATCH 4/4 ARM: dts: add PCI to the Gemini DTSI

Hi Linus,

Quoted 14 lines · Linus Walleij
Quoted 4 lines · Hans Ulli Kroll

We need to the remove hwirq 26-28 from DT.
First one will print this warning while boot.

irq: type mismatch, failed to map hwirq-26 for /soc/interrupt-controller at 48000000!

(...)

arch/arm/boot/dts/gemini.dtsi line 117

Quoted 5 lines
arch/arm/boot/dts/gemini.dtsi +1 -4
-                       interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
-                                       <26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
-                                       <27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
-                                       <28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+                       interrupts = <8 IRQ_TYPE_LEVEL_HIGH>; /* chained irq PCI A-D */
Quoted 6 lines

Sure I can remove them ... just found them in the irqs.h file and thought it made sense to add them. I'll just cut them.

Since there is actually an internal IRQ controller in the host controller cascading the four PCI child IRQs that we model as an irqchip, I don't really see why they have these "PCI B-D" IRQs... anyone has a guess?

I think they used some other IP vendor.

from my IB 4220 sources.

#define IRQ_PCI_INTA PCI_IRQ_OFFSET + 0
#ifndef CONFIG_DUAL_PCI

#define IRQ_PCI_INTB				       PCI_IRQ_OFFSET + 1
#define IRQ_PCI_INTC				       PCI_IRQ_OFFSET + 2
#define IRQ_PCI_INTD				       PCI_IRQ_OFFSET + 3
#else
#define IRQ_PCI_INTB				       27
#define IRQ_PCI_INTC				       28
#define IRQ_PCI_INTD				       29

#endif

CONFIG_DUAL_PCI is never used
IRQ_PCIB - IRQ_PCID or IRQ_PCI_INTB - IRQ_PCI_INTD are also never used.

You can download my original NAS 4220 here
http://ulli-kroll.de/gemini/kernel.tgz

Greetings
Hans Ulli Kroll

#16 lore raw

PATCH 4/4 ARM: dts: add PCI to the Gemini DTSI

Quoted 12 lines · CET Hans Ulli Kroll

from my IB 4220 sources.

#define IRQ_PCI_INTA PCI_IRQ_OFFSET + 0
#ifndef CONFIG_DUAL_PCI

#define IRQ_PCI_INTB                                   PCI_IRQ_OFFSET + 1
#define IRQ_PCI_INTC                                   PCI_IRQ_OFFSET + 2
#define IRQ_PCI_INTD                                   PCI_IRQ_OFFSET + 3
#else
#define IRQ_PCI_INTB                                   27
#define IRQ_PCI_INTC                                   28
#define IRQ_PCI_INTD                                   29
Quoted 6 lines

#endif

CONFIG_DUAL_PCI is never used
IRQ_PCIB - IRQ_PCID or IRQ_PCI_INTB - IRQ_PCI_INTD are also never used.

The source code that Linus quoted earlier had references to 'IRQ_PCI_INTA +n' though, which is basically the same thing.

Arnd

#17 lore raw

PATCH 4/4 ARM: dts: add PCI to the Gemini DTSI

Quoted 15 lines · Arnd Bergmann
Quoted 14 lines · CET Hans Ulli Kroll

from my IB 4220 sources.

#define IRQ_PCI_INTA PCI_IRQ_OFFSET + 0
#ifndef CONFIG_DUAL_PCI

#define IRQ_PCI_INTB                                   PCI_IRQ_OFFSET + 1
#define IRQ_PCI_INTC                                   PCI_IRQ_OFFSET + 2
#define IRQ_PCI_INTD                                   PCI_IRQ_OFFSET + 3
#else
#define IRQ_PCI_INTB                                   27
#define IRQ_PCI_INTC                                   28
#define IRQ_PCI_INTD                                   29

#endif

Quoted 5 lines
Quoted 2 lines

CONFIG_DUAL_PCI is never used
IRQ_PCIB - IRQ_PCID or IRQ_PCI_INTB - IRQ_PCI_INTD are also never used.

The source code that Linus quoted earlier had references to 'IRQ_PCI_INTA +n' though, which is basically the same thing.

There is essentially two versions of the IP block, one which has a cascaded interrupt controller in the PCI host bridge, and one named "dual PCI" that has dedicated IRQs for PCIA, B, C, D on the primary interrupt controller.

I'll try to codify it into the driver so it's clear how this works on the two variants.

I'm also pretty sure this is a faraday IP block and not something from Storlink/Storm/Cortina, so I will rename the compatible strings etc reflecting that.

Yours,
Linus Walleij

#18 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

In reply to Linus Walleij

Hi Linus,

Quoted 9 lines · Linus Walleij

This adds device tree bindings for the Cortina Systems Gemini PCI Host Bridge.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc devicetree at vger.kernel.org

I'm hoping one of the DT folks will take a quick look at this.

Quoted 15 lines

---
This can be merged to the PCI tree whenever it is considered fine for inclusion.

---
 .../devicetree/bindings/pci/cortina,gemini-pci.txt | 64 ++++++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +1 -0
new file mode 100644
index 000000000000..e3090d995e1e
@@ -0,0 +1,64 @@
+* Cortina Systems Gemini PCI Host Bridge

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 11

Quoted 10 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +10 -0
+
+Mandatory properties:
+
+- compatible: should be "cortina,gemini-pci"
+- reg: memory base and size for the host bridge
+- interrupts: the four PCI interrupts
+- #address-cells: set to <3>
+- #size-cells: set to <2>
+- #interrupt-cells: set to <1>
+- bus-range: set to <0x00 0x00> (only root bus)

Why is this limited to bus 0? Is everything completely soldered down with no possibility of adding or replacing PCI devices? The interrupt-map below suggests slots, though. If there's a slot, we could plug in a card with a bridge, which would mean more than just bus 0.

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 18

Quoted 7 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +7 -0
+- device_type, set to "pci"
+- ranges: see pci.txt
+- interrupt-map-mask: see pci.txt
+- interrupt-map: see pci.txt
+
+Mandatory subnodes:
+- One node reprenting the interrupt-controller inside the host bridge

s/reprenting/representing/

Quoted 53 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +46 -0
+  with the following mandatory properties:
+  - interrupt-controller: see interrupt-controller/interrupts.txt
+  - #address-cells: set to <0>
+  - #interrupt-cells: set to <1>
+
+Example:
+
+pci at 50000000 {
+	compatible = "cortina,gemini-pci";
+	reg = <0x50000000 0x100>;
+	interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
+			<26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
+			<27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
+			<28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+	#address-cells = <3>;
+	#size-cells = <2>;
+	#interrupt-cells = <1>;
+
+	bus-range = <0x00 0x00>; /* Only root bus */
+	ranges = /* 1MiB I/O space 0x50000000-0x500fffff */
+		 <0x01000000 0 0          0x50000000 0 0x00100000>,
+		 /* 128MiB non-prefetchable memory 0x58000000-0x5fffffff */
+		 <0x02000000 0 0x58000000 0x58000000 0 0x08000000>;
+	interrupt-map-mask = <0xff00 0 0 7>;
+	interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+			<0x4900 0 0 2 &pci_intc 1>,
+			<0x4a00 0 0 3 &pci_intc 2>,
+			<0x4b00 0 0 4 &pci_intc 3>,
+			<0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+			<0x5100 0 0 2 &pci_intc 1>,
+			<0x5200 0 0 3 &pci_intc 2>,
+			<0x5300 0 0 4 &pci_intc 3>,
+			<0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+			<0x5900 0 0 2 &pci_intc 1>,
+			<0x5a00 0 0 3 &pci_intc 2>,
+			<0x5b00 0 0 4 &pci_intc 3>,
+			<0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+			<0x6100 0 0 2 &pci_intc 1>,
+			<0x6200 0 0 3 &pci_intc 2>,
+			<0x6300 0 0 4 &pci_intc 3>;
+	pci_intc: interrupt-controller {
+		interrupt-controller;
+		#address-cells = <0>;
+		#interrupt-cells = <1>;
+	};
+};
Signature · 6 lines
2.9.3

--
To unsubscribe from this list: send the line "unsubscribe linux-pci" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#19 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

Quoted 6 lines · Bjorn Helgaas
Quoted 1 line

+- bus-range: set to <0x00 0x00> (only root bus)

Why is this limited to bus 0? Is everything completely soldered down with no possibility of adding or replacing PCI devices? The interrupt-map below suggests slots, though. If there's a slot, we could plug in a card with a bridge, which would mean more than just bus 0.

I thought so because it's a NAS box (the typical usecase) but when I dismantled it and turned it over, wow, there is indeed a MiniPCI slot!

OK I will augment this to 0x00 0xff.

Yours,
Linus Walleij

#20 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

In reply to Linus Walleij

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 63

Quoted 6 lines · CET Linus Walleij
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +5 -0
+       pci_intc: interrupt-controller {
+               interrupt-controller;
+               #address-cells = <0>;
+               #interrupt-cells = <1>;
+       };

I think this one needs at least an interrupt-parent property, otherwise you would recursively translate the numbers back through the interrupt-map property of its parent.

Arnd

#21 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

Quoted 10 lines · Arnd Bergmann

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 63

Quoted 5 lines · CET Linus Walleij
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +5 -0
+       pci_intc: interrupt-controller {
+               interrupt-controller;
+               #address-cells = <0>;
+               #interrupt-cells = <1>;
+       };

I think this one needs at least an interrupt-parent property, otherwise you would recursively translate the numbers back through the interrupt-map property of its parent.

OK I added this, I think it was the cause of the error Hand Uli Kroll was seeing. I guess I didn't run into it because I was only testing the device in slot 9 so everything was defaulting to zero.

Yours,
Linus Walleij

#22 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

In reply to Linus Walleij
Quoted 8 lines · CET Linus Walleij
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +7 -0
+       interrupt-map-mask = <0xff00 0 0 7>;
+       interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+                       <0x4900 0 0 2 &pci_intc 1>,
+                       <0x4a00 0 0 3 &pci_intc 2>,
+                       <0x4b00 0 0 4 &pci_intc 3>,
+                       <0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+                       <0x5100 0 0 2 &pci_intc 1>,

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 58

Quoted 10 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +10 -0
+                       <0x5200 0 0 3 &pci_intc 2>,
+                       <0x5300 0 0 4 &pci_intc 3>,
+                       <0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+                       <0x5900 0 0 2 &pci_intc 1>,
+                       <0x5a00 0 0 3 &pci_intc 2>,
+                       <0x5b00 0 0 4 &pci_intc 3>,
+                       <0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+                       <0x6100 0 0 2 &pci_intc 1>,
+                       <0x6200 0 0 3 &pci_intc 2>,
+                       <0x6300 0 0 4 &pci_intc 3>;

The mapping looks wrong here, we normally don't list interrupts per function so the mask should be 0xf800.

Note that the interrupt map is board specific, so this should probably go in the board.dts file rather than platform.dtsi.

For this particular board, the interrupt lines appear to have been badly configured so all slots use the same interrupt 0 for IntA. IIRC This also means you can probably use <0 0 0 7> as the mask and just specify each of the four interrupts once. A properly wired board would swizzle the interrupts so that each slot has a different IRQ for its IntA line.

Arnd

#23 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

Quoted 20 lines · Arnd Bergmann
Quoted 8 lines · CET Linus Walleij
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +7 -0
+       interrupt-map-mask = <0xff00 0 0 7>;
+       interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+                       <0x4900 0 0 2 &pci_intc 1>,
+                       <0x4a00 0 0 3 &pci_intc 2>,
+                       <0x4b00 0 0 4 &pci_intc 3>,
+                       <0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+                       <0x5100 0 0 2 &pci_intc 1>,

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 58

Quoted 10 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +10 -0
+                       <0x5200 0 0 3 &pci_intc 2>,
+                       <0x5300 0 0 4 &pci_intc 3>,
+                       <0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+                       <0x5900 0 0 2 &pci_intc 1>,
+                       <0x5a00 0 0 3 &pci_intc 2>,
+                       <0x5b00 0 0 4 &pci_intc 3>,
+                       <0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+                       <0x6100 0 0 2 &pci_intc 1>,
+                       <0x6200 0 0 3 &pci_intc 2>,
+                       <0x6300 0 0 4 &pci_intc 3>;
Quoted 2 lines

The mapping looks wrong here, we normally don't list interrupts per function so the mask should be 0xf800.

Yup it works that way too, and indeed the USB hub on function 2 was requesting the right pin and everything.

Quoted 2 lines

Note that the interrupt map is board specific, so this should probably go in the board.dts file rather than platform.dtsi.

OK moving it down there.

Quoted 5 lines

For this particular board, the interrupt lines appear to have been badly configured so all slots use the same interrupt 0 for IntA. IIRC This also means you can probably use <0 0 0 7> as the mask and just specify each of the four interrupts once. A properly wired board would swizzle the interrupts so that each slot has a different IRQ for its IntA line.

They are swizzled, I just have very bad hardware docs. (Only code.) It turns out when I'm browsing through old board support code that there is a comment followed by a piece of code like this:

/*
* No swizzle on SL2312
*/
static u8 __init sl2312_pci_swizzle(struct pci_dev *dev, u8 *pinp)

{
        return PCI_SLOT(dev->devfn);
}

/*
* map the specified device/slot/pin to an IRQ. This works out such
* that slot 9 pin 1 is INT0, pin 2 is INT1, and slot 10 pin 1 is INT1. */
static int __init sl2312_pci_map_irq(struct pci_dev *dev, u8 slot, u8 pin)

{
        int intnr = ((slot  + (pin - 1)) & 3) + 4;  /* the IRQ number

of PCI bridge */

        printk("%s : slot = %d  pin = %d \n",__func__,slot,pin);
    switch (slot)
    {
        case 12:
                if (pin==1)
                {
                        intnr = 3;
                    }
                    else
                    {
                        intnr = 0;
                    }
#ifdef CONFIG_DUAL_PCI
                    return IRQ_PCI_INTD;
#endif
            break;
        case 11:
                    intnr = (2 + (pin - 1)) & 3;
#ifdef CONFIG_DUAL_PCI
                    return IRQ_PCI_INTC;
#endif
            break;
        case 10:
                    intnr = (1 + (pin - 1)) & 3;
#ifdef CONFIG_DUAL_PCI
                    return IRQ_PCI_INTB;
#endif
            break;
        case  9:
                    intnr = (pin - 1) & 3;
            break;
    }
//      if (slot == 10)
//              intnr = (1 + (pin - 1)) & 3;
//      else if (slot == 9)
//              intnr = (pin - 1) & 3;
        return (IRQ_PCI_INTA + intnr);
}

If I understand correctly they say that the IRQs are not swizzled on the PCI bridge side but on the IRQ handler side.

So if I put it in the device tree like so:

+ interrupt-map =

+                               <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+                               <0x4800 0 0 2 &pci_intc 1>,
+                               <0x4800 0 0 3 &pci_intc 2>,
+                               <0x4800 0 0 4 &pci_intc 3>,
+                               <0x5000 0 0 1 &pci_intc 1>, /* Slot 10 */
+                               <0x5000 0 0 2 &pci_intc 2>,
+                               <0x5000 0 0 3 &pci_intc 3>,
+                               <0x5000 0 0 4 &pci_intc 0>,
+                               <0x5800 0 0 1 &pci_intc 2>, /* Slot 11 */
+                               <0x5800 0 0 2 &pci_intc 3>,
+                               <0x5800 0 0 3 &pci_intc 0>,
+                               <0x5800 0 0 4 &pci_intc 1>,
+                               <0x6000 0 0 1 &pci_intc 3>, /* Slot 12 */
+                               <0x6000 0 0 2 &pci_intc 0>,
+                               <0x6000 0 0 3 &pci_intc 1>,
+                               <0x6000 0 0 4 &pci_intc 2>;

So the IRQs on the right side are swizzled instead of the pins being swizzled.

...but that looks a bit insane.

Isn't that exactly the same thing just exposed in some inverse way?

I'll try to get the RALink MiniPCI I have ?n the slot going and see if it works the same with just good old vanilla swizzling.

Yours,
Linus Walleij

#24 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

Quoted 88 lines · Linus Walleij
Quoted 21 lines · Arnd Bergmann
Quoted 8 lines · CET Linus Walleij
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +7 -0
+       interrupt-map-mask = <0xff00 0 0 7>;
+       interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+                       <0x4900 0 0 2 &pci_intc 1>,
+                       <0x4a00 0 0 3 &pci_intc 2>,
+                       <0x4b00 0 0 4 &pci_intc 3>,
+                       <0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+                       <0x5100 0 0 2 &pci_intc 1>,

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 58

Quoted 10 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +10 -0
+                       <0x5200 0 0 3 &pci_intc 2>,
+                       <0x5300 0 0 4 &pci_intc 3>,
+                       <0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+                       <0x5900 0 0 2 &pci_intc 1>,
+                       <0x5a00 0 0 3 &pci_intc 2>,
+                       <0x5b00 0 0 4 &pci_intc 3>,
+                       <0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+                       <0x6100 0 0 2 &pci_intc 1>,
+                       <0x6200 0 0 3 &pci_intc 2>,
+                       <0x6300 0 0 4 &pci_intc 3>;
Quoted 5 lines

For this particular board, the interrupt lines appear to have been badly configured so all slots use the same interrupt 0 for IntA. IIRC This also means you can probably use <0 0 0 7> as the mask and just specify each of the four interrupts once. A properly wired board would swizzle the interrupts so that each slot has a different IRQ for its IntA line.

They are swizzled, I just have very bad hardware docs. (Only code.) It turns out when I'm browsing through old board support code that there is a comment followed by a piece of code like this:

/*
* No swizzle on SL2312
*/
static u8 __init sl2312_pci_swizzle(struct pci_dev *dev, u8 *pinp)

{
        return PCI_SLOT(dev->devfn);
}

/*
* map the specified device/slot/pin to an IRQ. This works out such
* that slot 9 pin 1 is INT0, pin 2 is INT1, and slot 10 pin 1 is INT1. */
static int __init sl2312_pci_map_irq(struct pci_dev *dev, u8 slot, u8 pin)

{
        int intnr = ((slot  + (pin - 1)) & 3) + 4;  /* the IRQ number

of PCI bridge */

        printk("%s : slot = %d  pin = %d \n",__func__,slot,pin);
    switch (slot)
    {
        case 12:
                if (pin==1)
                {
                        intnr = 3;
                    }
                    else
                    {
                        intnr = 0;
                    }
#ifdef CONFIG_DUAL_PCI
                    return IRQ_PCI_INTD;
#endif
            break;
        case 11:
                    intnr = (2 + (pin - 1)) & 3;
#ifdef CONFIG_DUAL_PCI
                    return IRQ_PCI_INTC;
#endif
            break;
        case 10:
                    intnr = (1 + (pin - 1)) & 3;
#ifdef CONFIG_DUAL_PCI
                    return IRQ_PCI_INTB;
#endif
            break;
        case  9:
                    intnr = (pin - 1) & 3;
            break;
    }
//      if (slot == 10)
//              intnr = (1 + (pin - 1)) & 3;
//      else if (slot == 9)
//              intnr = (pin - 1) & 3;
        return (IRQ_PCI_INTA + intnr);
}
Quoted 2 lines

If I understand correctly they say that the IRQs are not swizzled on the PCI bridge side but on the IRQ handler side.

That depends on whether CONFIG_DUAL_PCI is set or not:

It seems that slot 12 is hardwired to have hwirq 3 on IntA and hwirq 0 on all others by default, but have hwirq on IntA through IntD on boards that use CONFIG_DUAL_PCI.

Slot 10/11 use the default swizzling without CONFIG_DUAL_PCI but uses hwirq 1 or 2 for all four pins with CONFIG_DUAL_PCI.

Slot 9 seems to always use the default swizzling, so it has all four IRQ pins connected to the four hwirqs of the SoC.

This is still an awful wiring, but not unthinkable, as board designers may have had a specific usage for the four interrupts in mind, to spread the devices that are actually present across the available hwirqs.

Quoted 5 lines

So if I put it in the device tree like so:

(quoted diff) +3 -0
+                         interrupt-map =
+                               <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+                               <0x4800 0 0 2 &pci_intc 1>,

Review

Quoted 10 lines
(quoted diff) +10 -0
+                               <0x4800 0 0 3 &pci_intc 2>,
+                               <0x4800 0 0 4 &pci_intc 3>,
+                               <0x5000 0 0 1 &pci_intc 1>, /* Slot 10 */
+                               <0x5000 0 0 2 &pci_intc 2>,
+                               <0x5000 0 0 3 &pci_intc 3>,
+                               <0x5000 0 0 4 &pci_intc 0>,
+                               <0x5800 0 0 1 &pci_intc 2>, /* Slot 11 */
+                               <0x5800 0 0 2 &pci_intc 3>,
+                               <0x5800 0 0 3 &pci_intc 0>,
+                               <0x5800 0 0 4 &pci_intc 1>,

For boards that require CONFIG_DUAL_PCI, this would be

                               <0x5000 0 0 1 &pci_intc 1>, /* Slot 10 */
                               <0x5000 0 0 2 &pci_intc 1>,
                               <0x5000 0 0 3 &pci_intc 1>,
                               <0x5000 0 0 4 &pci_intc 1>,
                               <0x5800 0 0 1 &pci_intc 2>, /* Slot 11 */
                               <0x5800 0 0 2 &pci_intc 2>,
                               <0x5800 0 0 3 &pci_intc 2>,
                               <0x5800 0 0 4 &pci_intc 2>,

It probably works either way if you put just one device in there, as almost everything uses just IntA, but you'd notice the difference if there is ever a PCI bridge behind that slot.

Review

Quoted 4 lines
(quoted diff) +4 -0
+                               <0x6000 0 0 1 &pci_intc 3>, /* Slot 12 */
+                               <0x6000 0 0 2 &pci_intc 0>,
+                               <0x6000 0 0 3 &pci_intc 1>,
+                               <0x6000 0 0 4 &pci_intc 2>;

The last two have to be

                               <0x6000 0 0 3 &pci_intc 0>, /* [sic] */
                               <0x6000 0 0 4 &pci_intc 0>; /* [sic] */
Quoted 7 lines

So the IRQs on the right side are swizzled instead of the pins being swizzled.

...but that looks a bit insane.

Isn't that exactly the same thing just exposed in some inverse way?

I don't understand your question. The mapping that you list is the default swizzling, right? I think if you have a board with sane wiring, you could write that as

                      interrupt-map-mask = <0x0 0x0 0x0 0x7>;
                      interrupt-map =
                              <0x0 0x0 0x0 0x1 &pci_intc 0>, /* int A */
                              <0x0 0x0 0x0 0x2 &pci_intc 1>, /* int B */
                              <0x0 0x0 0x0 0x3 &pci_intc 2>, /* int C */
                              <0x0 0x0 0x0 0x4 &pci_intc 3>; /* int D */

meaning that the IRQs are wired correctly to the root bus, and all slots are connected one would normally do with the standard swizzling.

Arnd

maintainer
#25 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

In reply to Linus Walleij
Quoted 26 lines · Linus Walleij

This adds device tree bindings for the Cortina Systems Gemini PCI Host Bridge.

Cc Janos Laube <[email protected]>
Cc Paulius Zaleckas <[email protected]>
Cc Hans Ulli Kroll <[email protected]>
Cc Florian Fainelli <[email protected]>
Cc devicetree at vger.kernel.org

---
This can be merged to the PCI tree whenever it is considered fine for inclusion.

---
 .../devicetree/bindings/pci/cortina,gemini-pci.txt | 64 ++++++++++++++++++++++
 1 file changed, 64 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +3 -0
new file mode 100644
index 000000000000..e3090d995e1e
@@ -0,0 +1,64 @@
+* Cortina Systems Gemini PCI Host Bridge
+
+Mandatory properties:

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 13

Quoted 10 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +10 -0
+
+- compatible: should be "cortina,gemini-pci"
+- reg: memory base and size for the host bridge
+- interrupts: the four PCI interrupts
+- #address-cells: set to <3>
+- #size-cells: set to <2>
+- #interrupt-cells: set to <1>
+- bus-range: set to <0x00 0x00> (only root bus)
+- device_type, set to "pci"
+- ranges: see pci.txt

You need to specify what ranges should have for this bridge.

Quoted 5 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +5 -0
+- interrupt-map-mask: see pci.txt
+- interrupt-map: see pci.txt
+
+Mandatory subnodes:
+- One node reprenting the interrupt-controller inside the host bridge

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 28

Quoted 10 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +10 -0
+  with the following mandatory properties:
+  - interrupt-controller: see interrupt-controller/interrupts.txt
+  - #address-cells: set to <0>
+  - #interrupt-cells: set to <1>
+
+Example:
+
+pci at 50000000 {
+	compatible = "cortina,gemini-pci";
+	reg = <0x50000000 0x100>;

Config space is indirectly accessed?

Quoted 43 lines
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +36 -0
+	interrupts = <8 IRQ_TYPE_LEVEL_HIGH>, /* PCI A */
+			<26 IRQ_TYPE_LEVEL_HIGH>, /* PCI B */
+			<27 IRQ_TYPE_LEVEL_HIGH>, /* PCI C */
+			<28 IRQ_TYPE_LEVEL_HIGH>; /* PCI D */
+	#address-cells = <3>;
+	#size-cells = <2>;
+	#interrupt-cells = <1>;
+
+	bus-range = <0x00 0x00>; /* Only root bus */
+	ranges = /* 1MiB I/O space 0x50000000-0x500fffff */
+		 <0x01000000 0 0          0x50000000 0 0x00100000>,
+		 /* 128MiB non-prefetchable memory 0x58000000-0x5fffffff */
+		 <0x02000000 0 0x58000000 0x58000000 0 0x08000000>;
+	interrupt-map-mask = <0xff00 0 0 7>;
+	interrupt-map = <0x4800 0 0 1 &pci_intc 0>, /* Slot 9 */
+			<0x4900 0 0 2 &pci_intc 1>,
+			<0x4a00 0 0 3 &pci_intc 2>,
+			<0x4b00 0 0 4 &pci_intc 3>,
+			<0x5000 0 0 1 &pci_intc 0>, /* Slot 10 */
+			<0x5100 0 0 2 &pci_intc 1>,
+			<0x5200 0 0 3 &pci_intc 2>,
+			<0x5300 0 0 4 &pci_intc 3>,
+			<0x5800 0 0 1 &pci_intc 0>, /* Slot 11 */
+			<0x5900 0 0 2 &pci_intc 1>,
+			<0x5a00 0 0 3 &pci_intc 2>,
+			<0x5b00 0 0 4 &pci_intc 3>,
+			<0x6000 0 0 1 &pci_intc 0>, /* Slot 12 */
+			<0x6100 0 0 2 &pci_intc 1>,
+			<0x6200 0 0 3 &pci_intc 2>,
+			<0x6300 0 0 4 &pci_intc 3>;
+	pci_intc: interrupt-controller {
+		interrupt-controller;
+		#address-cells = <0>;
+		#interrupt-cells = <1>;
+	};
+};
Signature · 6 lines
2.9.3

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo at vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#26 lore raw

PATCH 1/4 PCI: add DT bindings for Cortina Gemini PCI Host Bridge

Quoted 9 lines · Rob Herring

Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt line 28

Quoted 5 lines · Linus Walleij
Documentation/devicetree/bindings/pci/cortina,gemini-pci.txt new file +5 -0
+Example:
+
+pci at 50000000 {
+     compatible = "cortina,gemini-pci";
+     reg = <0x50000000 0x100>;

Config space is indirectly accessed?

Yes it is a really annoying construction. The first device, the Faraday roob hub is at that address, the init code uses that to set up the bus.

As soon as the bus is up, we use bus accessors to talk to the root hub on slot 0 (this manages interrupts etc).

I guess I could even unmap this reg range at that point ...

I guess I could also get the address directly from the IO range and start poking around. I don't know if that is any better.

Yours,
Linus Walleij