Merge branch 'pci/host/mobiveil'
authorBjorn Helgaas <bhelgaas@google.com>
Wed, 6 Jun 2018 21:10:38 +0000 (16:10 -0500)
committerBjorn Helgaas <bhelgaas@google.com>
Wed, 6 Jun 2018 21:10:38 +0000 (16:10 -0500)
  - add Mobiveil PCIe host controller driver (Subrahmanya Lingappa)

  - add Mobiveil MSI support (Subrahmanya Lingappa)

* pci/host/mobiveil:
  PCI: mobiveil: Add MSI support
  PCI: mobiveil: Add Mobiveil PCIe Host Bridge IP driver
  PCI: mobiveil: Add Mobiveil PCIe Host Bridge IP driver DT bindings

1  2 
MAINTAINERS
drivers/pci/host/pcie-mobiveil.c

diff --cc MAINTAINERS
Simple merge
index 0000000000000000000000000000000000000000,dce0f27ead658f30afcea07c1ea99e2f905c7a2f..4d6c20e47bed4f8c19ea4e44c0bfee78b40d28a6
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,867 +1,866 @@@
 -      struct device_node *node = dev->of_node;
+ // SPDX-License-Identifier: GPL-2.0
+ /*
+  * PCIe host controller driver for Mobiveil PCIe Host controller
+  *
+  * Copyright (c) 2018 Mobiveil Inc.
+  * Author: Subrahmanya Lingappa <l.subrahmanya@mobiveil.co.in>
+  */
+ #include <linux/delay.h>
+ #include <linux/init.h>
+ #include <linux/interrupt.h>
+ #include <linux/irq.h>
+ #include <linux/irqchip/chained_irq.h>
+ #include <linux/irqdomain.h>
+ #include <linux/kernel.h>
+ #include <linux/module.h>
+ #include <linux/msi.h>
+ #include <linux/of_address.h>
+ #include <linux/of_irq.h>
+ #include <linux/of_platform.h>
+ #include <linux/of_pci.h>
+ #include <linux/pci.h>
+ #include <linux/platform_device.h>
+ #include <linux/slab.h>
+ /* register offsets and bit positions */
+ /*
+  * translation tables are grouped into windows, each window registers are
+  * grouped into blocks of 4 or 16 registers each
+  */
+ #define PAB_REG_BLOCK_SIZE    16
+ #define PAB_EXT_REG_BLOCK_SIZE        4
+ #define PAB_REG_ADDR(offset, win) (offset + (win * PAB_REG_BLOCK_SIZE))
+ #define PAB_EXT_REG_ADDR(offset, win) (offset + (win * PAB_EXT_REG_BLOCK_SIZE))
+ #define LTSSM_STATUS          0x0404
+ #define  LTSSM_STATUS_L0_MASK 0x3f
+ #define  LTSSM_STATUS_L0      0x2d
+ #define PAB_CTRL              0x0808
+ #define  AMBA_PIO_ENABLE_SHIFT        0
+ #define  PEX_PIO_ENABLE_SHIFT 1
+ #define  PAGE_SEL_SHIFT       13
+ #define  PAGE_SEL_MASK                0x3f
+ #define  PAGE_LO_MASK         0x3ff
+ #define  PAGE_SEL_EN          0xc00
+ #define  PAGE_SEL_OFFSET_SHIFT        10
+ #define PAB_AXI_PIO_CTRL      0x0840
+ #define  APIO_EN_MASK         0xf
+ #define PAB_PEX_PIO_CTRL      0x08c0
+ #define  PIO_ENABLE_SHIFT     0
+ #define PAB_INTP_AMBA_MISC_ENB                0x0b0c
+ #define PAB_INTP_AMBA_MISC_STAT       0x0b1c
+ #define  PAB_INTP_INTX_MASK           0x01e0
+ #define  PAB_INTP_MSI_MASK            0x8
+ #define PAB_AXI_AMAP_CTRL(win)        PAB_REG_ADDR(0x0ba0, win)
+ #define  WIN_ENABLE_SHIFT     0
+ #define  WIN_TYPE_SHIFT       1
+ #define PAB_EXT_AXI_AMAP_SIZE(win)    PAB_EXT_REG_ADDR(0xbaf0, win)
+ #define PAB_AXI_AMAP_AXI_WIN(win)     PAB_REG_ADDR(0x0ba4, win)
+ #define  AXI_WINDOW_ALIGN_MASK                3
+ #define PAB_AXI_AMAP_PEX_WIN_L(win)   PAB_REG_ADDR(0x0ba8, win)
+ #define  PAB_BUS_SHIFT                24
+ #define  PAB_DEVICE_SHIFT     19
+ #define  PAB_FUNCTION_SHIFT   16
+ #define PAB_AXI_AMAP_PEX_WIN_H(win)   PAB_REG_ADDR(0x0bac, win)
+ #define PAB_INTP_AXI_PIO_CLASS                0x474
+ #define PAB_PEX_AMAP_CTRL(win)        PAB_REG_ADDR(0x4ba0, win)
+ #define  AMAP_CTRL_EN_SHIFT   0
+ #define  AMAP_CTRL_TYPE_SHIFT 1
+ #define PAB_EXT_PEX_AMAP_SIZEN(win)   PAB_EXT_REG_ADDR(0xbef0, win)
+ #define PAB_PEX_AMAP_AXI_WIN(win)     PAB_REG_ADDR(0x4ba4, win)
+ #define PAB_PEX_AMAP_PEX_WIN_L(win)   PAB_REG_ADDR(0x4ba8, win)
+ #define PAB_PEX_AMAP_PEX_WIN_H(win)   PAB_REG_ADDR(0x4bac, win)
+ /* starting offset of INTX bits in status register */
+ #define PAB_INTX_START        5
+ /* supported number of MSI interrupts */
+ #define PCI_NUM_MSI   16
+ /* MSI registers */
+ #define MSI_BASE_LO_OFFSET    0x04
+ #define MSI_BASE_HI_OFFSET    0x08
+ #define MSI_SIZE_OFFSET       0x0c
+ #define MSI_ENABLE_OFFSET     0x14
+ #define MSI_STATUS_OFFSET     0x18
+ #define MSI_DATA_OFFSET       0x20
+ #define MSI_ADDR_L_OFFSET     0x24
+ #define MSI_ADDR_H_OFFSET     0x28
+ /* outbound and inbound window definitions */
+ #define WIN_NUM_0             0
+ #define WIN_NUM_1             1
+ #define CFG_WINDOW_TYPE       0
+ #define IO_WINDOW_TYPE                1
+ #define MEM_WINDOW_TYPE       2
+ #define IB_WIN_SIZE           (256 * 1024 * 1024 * 1024)
+ #define MAX_PIO_WINDOWS       8
+ /* Parameters for the waiting for link up routine */
+ #define LINK_WAIT_MAX_RETRIES 10
+ #define LINK_WAIT_MIN 90000
+ #define LINK_WAIT_MAX 100000
+ struct mobiveil_msi {                 /* MSI information */
+       struct mutex lock;              /* protect bitmap variable */
+       struct irq_domain *msi_domain;
+       struct irq_domain *dev_domain;
+       phys_addr_t msi_pages_phys;
+       int num_of_vectors;
+       DECLARE_BITMAP(msi_irq_in_use, PCI_NUM_MSI);
+ };
+ struct mobiveil_pcie {
+       struct platform_device *pdev;
+       struct list_head resources;
+       void __iomem *config_axi_slave_base;    /* endpoint config base */
+       void __iomem *csr_axi_slave_base;       /* root port config base */
+       void __iomem *apb_csr_base;     /* MSI register base */
+       void __iomem *pcie_reg_base;    /* Physical PCIe Controller Base */
+       struct irq_domain *intx_domain;
+       raw_spinlock_t intx_mask_lock;
+       int irq;
+       int apio_wins;
+       int ppio_wins;
+       int ob_wins_configured;         /* configured outbound windows */
+       int ib_wins_configured;         /* configured inbound windows */
+       struct resource *ob_io_res;
+       char root_bus_nr;
+       struct mobiveil_msi msi;
+ };
+ static inline void csr_writel(struct mobiveil_pcie *pcie, const u32 value,
+               const u32 reg)
+ {
+       writel_relaxed(value, pcie->csr_axi_slave_base + reg);
+ }
+ static inline u32 csr_readl(struct mobiveil_pcie *pcie, const u32 reg)
+ {
+       return readl_relaxed(pcie->csr_axi_slave_base + reg);
+ }
+ static bool mobiveil_pcie_link_up(struct mobiveil_pcie *pcie)
+ {
+       return (csr_readl(pcie, LTSSM_STATUS) &
+               LTSSM_STATUS_L0_MASK) == LTSSM_STATUS_L0;
+ }
+ static bool mobiveil_pcie_valid_device(struct pci_bus *bus, unsigned int devfn)
+ {
+       struct mobiveil_pcie *pcie = bus->sysdata;
+       /* Only one device down on each root port */
+       if ((bus->number == pcie->root_bus_nr) && (devfn > 0))
+               return false;
+       /*
+        * Do not read more than one device on the bus directly
+        * attached to RC
+        */
+       if ((bus->primary == pcie->root_bus_nr) && (devfn > 0))
+               return false;
+       return true;
+ }
+ /*
+  * mobiveil_pcie_map_bus - routine to get the configuration base of either
+  * root port or endpoint
+  */
+ static void __iomem *mobiveil_pcie_map_bus(struct pci_bus *bus,
+                                       unsigned int devfn, int where)
+ {
+       struct mobiveil_pcie *pcie = bus->sysdata;
+       if (!mobiveil_pcie_valid_device(bus, devfn))
+               return NULL;
+       if (bus->number == pcie->root_bus_nr) {
+               /* RC config access */
+               return pcie->csr_axi_slave_base + where;
+       }
+       /*
+        * EP config access (in Config/APIO space)
+        * Program PEX Address base (31..16 bits) with appropriate value
+        * (BDF) in PAB_AXI_AMAP_PEX_WIN_L0 Register.
+        * Relies on pci_lock serialization
+        */
+       csr_writel(pcie, bus->number << PAB_BUS_SHIFT |
+                       PCI_SLOT(devfn) << PAB_DEVICE_SHIFT |
+                       PCI_FUNC(devfn) << PAB_FUNCTION_SHIFT,
+                       PAB_AXI_AMAP_PEX_WIN_L(WIN_NUM_0));
+       return pcie->config_axi_slave_base + where;
+ }
+ static struct pci_ops mobiveil_pcie_ops = {
+       .map_bus = mobiveil_pcie_map_bus,
+       .read = pci_generic_config_read,
+       .write = pci_generic_config_write,
+ };
+ static void mobiveil_pcie_isr(struct irq_desc *desc)
+ {
+       struct irq_chip *chip = irq_desc_get_chip(desc);
+       struct mobiveil_pcie *pcie = irq_desc_get_handler_data(desc);
+       struct device *dev = &pcie->pdev->dev;
+       struct mobiveil_msi *msi = &pcie->msi;
+       u32 msi_data, msi_addr_lo, msi_addr_hi;
+       u32 intr_status, msi_status;
+       unsigned long shifted_status;
+       u32 bit, virq, val, mask;
+       /*
+        * The core provides a single interrupt for both INTx/MSI messages.
+        * So we'll read both INTx and MSI status
+        */
+       chained_irq_enter(chip, desc);
+       /* read INTx status */
+       val = csr_readl(pcie, PAB_INTP_AMBA_MISC_STAT);
+       mask = csr_readl(pcie, PAB_INTP_AMBA_MISC_ENB);
+       intr_status = val & mask;
+       /* Handle INTx */
+       if (intr_status & PAB_INTP_INTX_MASK) {
+               shifted_status = csr_readl(pcie, PAB_INTP_AMBA_MISC_STAT) >>
+                       PAB_INTX_START;
+               do {
+                       for_each_set_bit(bit, &shifted_status, PCI_NUM_INTX) {
+                               virq = irq_find_mapping(pcie->intx_domain,
+                                               bit + 1);
+                               if (virq)
+                                       generic_handle_irq(virq);
+                               else
+                                       dev_err_ratelimited(dev,
+                                               "unexpected IRQ, INT%d\n", bit);
+                               /* clear interrupt */
+                               csr_writel(pcie,
+                                       shifted_status << PAB_INTX_START,
+                                       PAB_INTP_AMBA_MISC_STAT);
+                       }
+               } while ((shifted_status >> PAB_INTX_START) != 0);
+       }
+       /* read extra MSI status register */
+       msi_status = readl_relaxed(pcie->apb_csr_base + MSI_STATUS_OFFSET);
+       /* handle MSI interrupts */
+       while (msi_status & 1) {
+               msi_data = readl_relaxed(pcie->apb_csr_base
+                               + MSI_DATA_OFFSET);
+               /*
+                * MSI_STATUS_OFFSET register gets updated to zero
+                * once we pop not only the MSI data but also address
+                * from MSI hardware FIFO. So keeping these following
+                * two dummy reads.
+                */
+               msi_addr_lo = readl_relaxed(pcie->apb_csr_base +
+                               MSI_ADDR_L_OFFSET);
+               msi_addr_hi = readl_relaxed(pcie->apb_csr_base +
+                               MSI_ADDR_H_OFFSET);
+               dev_dbg(dev, "MSI registers, data: %08x, addr: %08x:%08x\n",
+                               msi_data, msi_addr_hi, msi_addr_lo);
+               virq = irq_find_mapping(msi->dev_domain, msi_data);
+               if (virq)
+                       generic_handle_irq(virq);
+               msi_status = readl_relaxed(pcie->apb_csr_base +
+                               MSI_STATUS_OFFSET);
+       }
+       /* Clear the interrupt status */
+       csr_writel(pcie, intr_status, PAB_INTP_AMBA_MISC_STAT);
+       chained_irq_exit(chip, desc);
+ }
+ static int mobiveil_pcie_parse_dt(struct mobiveil_pcie *pcie)
+ {
+       struct device *dev = &pcie->pdev->dev;
+       struct platform_device *pdev = pcie->pdev;
+       struct device_node *node = dev->of_node;
+       struct resource *res;
+       const char *type;
+       type = of_get_property(node, "device_type", NULL);
+       if (!type || strcmp(type, "pci")) {
+               dev_err(dev, "invalid \"device_type\" %s\n", type);
+               return -EINVAL;
+       }
+       /* map config resource */
+       res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+                       "config_axi_slave");
+       pcie->config_axi_slave_base = devm_pci_remap_cfg_resource(dev, res);
+       if (IS_ERR(pcie->config_axi_slave_base))
+               return PTR_ERR(pcie->config_axi_slave_base);
+       pcie->ob_io_res = res;
+       /* map csr resource */
+       res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+                       "csr_axi_slave");
+       pcie->csr_axi_slave_base = devm_pci_remap_cfg_resource(dev, res);
+       if (IS_ERR(pcie->csr_axi_slave_base))
+               return PTR_ERR(pcie->csr_axi_slave_base);
+       pcie->pcie_reg_base = res->start;
+       /* map MSI config resource */
+       res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "apb_csr");
+       pcie->apb_csr_base = devm_pci_remap_cfg_resource(dev, res);
+       if (IS_ERR(pcie->apb_csr_base))
+               return PTR_ERR(pcie->apb_csr_base);
+       /* read the number of windows requested */
+       if (of_property_read_u32(node, "apio-wins", &pcie->apio_wins))
+               pcie->apio_wins = MAX_PIO_WINDOWS;
+       if (of_property_read_u32(node, "ppio-wins", &pcie->ppio_wins))
+               pcie->ppio_wins = MAX_PIO_WINDOWS;
+       pcie->irq = platform_get_irq(pdev, 0);
+       if (pcie->irq <= 0) {
+               dev_err(dev, "failed to map IRQ: %d\n", pcie->irq);
+               return -ENODEV;
+       }
+       irq_set_chained_handler_and_data(pcie->irq, mobiveil_pcie_isr, pcie);
+       return 0;
+ }
+ /*
+  * select_paged_register - routine to access paged register of root complex
+  *
+  * registers of RC are paged, for this scheme to work
+  * extracted higher 6 bits of the offset will be written to pg_sel
+  * field of PAB_CTRL register and rest of the lower 10 bits enabled with
+  * PAGE_SEL_EN are used as offset of the register.
+  */
+ static void select_paged_register(struct mobiveil_pcie *pcie, u32 offset)
+ {
+       int pab_ctrl_dw, pg_sel;
+       /* clear pg_sel field */
+       pab_ctrl_dw = csr_readl(pcie, PAB_CTRL);
+       pab_ctrl_dw = (pab_ctrl_dw & ~(PAGE_SEL_MASK << PAGE_SEL_SHIFT));
+       /* set pg_sel field */
+       pg_sel = (offset >> PAGE_SEL_OFFSET_SHIFT) & PAGE_SEL_MASK;
+       pab_ctrl_dw |= ((pg_sel << PAGE_SEL_SHIFT));
+       csr_writel(pcie, pab_ctrl_dw, PAB_CTRL);
+ }
+ static void write_paged_register(struct mobiveil_pcie *pcie,
+               u32 val, u32 offset)
+ {
+       u32 off = (offset & PAGE_LO_MASK) | PAGE_SEL_EN;
+       select_paged_register(pcie, offset);
+       csr_writel(pcie, val, off);
+ }
+ static u32 read_paged_register(struct mobiveil_pcie *pcie, u32 offset)
+ {
+       u32 off = (offset & PAGE_LO_MASK) | PAGE_SEL_EN;
+       select_paged_register(pcie, offset);
+       return csr_readl(pcie, off);
+ }
+ static void program_ib_windows(struct mobiveil_pcie *pcie, int win_num,
+               int pci_addr, u32 type, u64 size)
+ {
+       int pio_ctrl_val;
+       int amap_ctrl_dw;
+       u64 size64 = ~(size - 1);
+       if ((pcie->ib_wins_configured + 1) > pcie->ppio_wins) {
+               dev_err(&pcie->pdev->dev,
+                       "ERROR: max inbound windows reached !\n");
+               return;
+       }
+       pio_ctrl_val = csr_readl(pcie, PAB_PEX_PIO_CTRL);
+       csr_writel(pcie,
+               pio_ctrl_val | (1 << PIO_ENABLE_SHIFT), PAB_PEX_PIO_CTRL);
+       amap_ctrl_dw = read_paged_register(pcie, PAB_PEX_AMAP_CTRL(win_num));
+       amap_ctrl_dw = (amap_ctrl_dw | (type << AMAP_CTRL_TYPE_SHIFT));
+       amap_ctrl_dw = (amap_ctrl_dw | (1 << AMAP_CTRL_EN_SHIFT));
+       write_paged_register(pcie, amap_ctrl_dw | lower_32_bits(size64),
+                               PAB_PEX_AMAP_CTRL(win_num));
+       write_paged_register(pcie, upper_32_bits(size64),
+                               PAB_EXT_PEX_AMAP_SIZEN(win_num));
+       write_paged_register(pcie, pci_addr, PAB_PEX_AMAP_AXI_WIN(win_num));
+       write_paged_register(pcie, pci_addr, PAB_PEX_AMAP_PEX_WIN_L(win_num));
+       write_paged_register(pcie, 0, PAB_PEX_AMAP_PEX_WIN_H(win_num));
+ }
+ /*
+  * routine to program the outbound windows
+  */
+ static void program_ob_windows(struct mobiveil_pcie *pcie, int win_num,
+               u64 cpu_addr, u64 pci_addr, u32 config_io_bit, u64 size)
+ {
+       u32 value, type;
+       u64 size64 = ~(size - 1);
+       if ((pcie->ob_wins_configured + 1) > pcie->apio_wins) {
+               dev_err(&pcie->pdev->dev,
+                       "ERROR: max outbound windows reached !\n");
+               return;
+       }
+       /*
+        * program Enable Bit to 1, Type Bit to (00) base 2, AXI Window Size Bit
+        * to 4 KB in PAB_AXI_AMAP_CTRL register
+        */
+       type = config_io_bit;
+       value = csr_readl(pcie, PAB_AXI_AMAP_CTRL(win_num));
+       csr_writel(pcie, 1 << WIN_ENABLE_SHIFT | type << WIN_TYPE_SHIFT |
+                       lower_32_bits(size64), PAB_AXI_AMAP_CTRL(win_num));
+       write_paged_register(pcie, upper_32_bits(size64),
+                               PAB_EXT_AXI_AMAP_SIZE(win_num));
+       /*
+        * program AXI window base with appropriate value in
+        * PAB_AXI_AMAP_AXI_WIN0 register
+        */
+       value = csr_readl(pcie, PAB_AXI_AMAP_AXI_WIN(win_num));
+       csr_writel(pcie, cpu_addr & (~AXI_WINDOW_ALIGN_MASK),
+                       PAB_AXI_AMAP_AXI_WIN(win_num));
+       value = csr_readl(pcie, PAB_AXI_AMAP_PEX_WIN_H(win_num));
+       csr_writel(pcie, lower_32_bits(pci_addr),
+                       PAB_AXI_AMAP_PEX_WIN_L(win_num));
+       csr_writel(pcie, upper_32_bits(pci_addr),
+                       PAB_AXI_AMAP_PEX_WIN_H(win_num));
+       pcie->ob_wins_configured++;
+ }
+ static int mobiveil_bringup_link(struct mobiveil_pcie *pcie)
+ {
+       int retries;
+       /* check if the link is up or not */
+       for (retries = 0; retries < LINK_WAIT_MAX_RETRIES; retries++) {
+               if (mobiveil_pcie_link_up(pcie))
+                       return 0;
+               usleep_range(LINK_WAIT_MIN, LINK_WAIT_MAX);
+       }
+       dev_err(&pcie->pdev->dev, "link never came up\n");
+       return -ETIMEDOUT;
+ }
+ static void mobiveil_pcie_enable_msi(struct mobiveil_pcie *pcie)
+ {
+       phys_addr_t msg_addr = pcie->pcie_reg_base;
+       struct mobiveil_msi *msi = &pcie->msi;
+       pcie->msi.num_of_vectors = PCI_NUM_MSI;
+       msi->msi_pages_phys = (phys_addr_t)msg_addr;
+       writel_relaxed(lower_32_bits(msg_addr),
+               pcie->apb_csr_base + MSI_BASE_LO_OFFSET);
+       writel_relaxed(upper_32_bits(msg_addr),
+               pcie->apb_csr_base + MSI_BASE_HI_OFFSET);
+       writel_relaxed(4096, pcie->apb_csr_base + MSI_SIZE_OFFSET);
+       writel_relaxed(1, pcie->apb_csr_base + MSI_ENABLE_OFFSET);
+ }
+ static int mobiveil_host_init(struct mobiveil_pcie *pcie)
+ {
+       u32 value, pab_ctrl, type = 0;
+       int err;
+       struct resource_entry *win, *tmp;
+       err = mobiveil_bringup_link(pcie);
+       if (err) {
+               dev_info(&pcie->pdev->dev, "link bring-up failed\n");
+               return err;
+       }
+       /*
+        * program Bus Master Enable Bit in Command Register in PAB Config
+        * Space
+        */
+       value = csr_readl(pcie, PCI_COMMAND);
+       csr_writel(pcie, value | PCI_COMMAND_IO | PCI_COMMAND_MEMORY |
+               PCI_COMMAND_MASTER, PCI_COMMAND);
+       /*
+        * program PIO Enable Bit to 1 (and PEX PIO Enable to 1) in PAB_CTRL
+        * register
+        */
+       pab_ctrl = csr_readl(pcie, PAB_CTRL);
+       csr_writel(pcie, pab_ctrl | (1 << AMBA_PIO_ENABLE_SHIFT) |
+               (1 << PEX_PIO_ENABLE_SHIFT), PAB_CTRL);
+       csr_writel(pcie, (PAB_INTP_INTX_MASK | PAB_INTP_MSI_MASK),
+               PAB_INTP_AMBA_MISC_ENB);
+       /*
+        * program PIO Enable Bit to 1 and Config Window Enable Bit to 1 in
+        * PAB_AXI_PIO_CTRL Register
+        */
+       value = csr_readl(pcie, PAB_AXI_PIO_CTRL);
+       csr_writel(pcie, value | APIO_EN_MASK, PAB_AXI_PIO_CTRL);
+       /*
+        * we'll program one outbound window for config reads and
+        * another default inbound window for all the upstream traffic
+        * rest of the outbound windows will be configured according to
+        * the "ranges" field defined in device tree
+        */
+       /* config outbound translation window */
+       program_ob_windows(pcie, pcie->ob_wins_configured,
+                       pcie->ob_io_res->start, 0, CFG_WINDOW_TYPE,
+                       resource_size(pcie->ob_io_res));
+       /* memory inbound translation window */
+       program_ib_windows(pcie, WIN_NUM_1, 0, MEM_WINDOW_TYPE, IB_WIN_SIZE);
+       /* Get the I/O and memory ranges from DT */
+       resource_list_for_each_entry_safe(win, tmp, &pcie->resources) {
+               type = 0;
+               if (resource_type(win->res) == IORESOURCE_MEM)
+                       type = MEM_WINDOW_TYPE;
+               if (resource_type(win->res) == IORESOURCE_IO)
+                       type = IO_WINDOW_TYPE;
+               if (type) {
+                       /* configure outbound translation window */
+                       program_ob_windows(pcie, pcie->ob_wins_configured,
+                               win->res->start, 0, type,
+                               resource_size(win->res));
+               }
+       }
+       /* setup MSI hardware registers */
+       mobiveil_pcie_enable_msi(pcie);
+       return err;
+ }
+ static void mobiveil_mask_intx_irq(struct irq_data *data)
+ {
+       struct irq_desc *desc = irq_to_desc(data->irq);
+       struct mobiveil_pcie *pcie;
+       unsigned long flags;
+       u32 mask, shifted_val;
+       pcie = irq_desc_get_chip_data(desc);
+       mask = 1 << ((data->hwirq + PAB_INTX_START) - 1);
+       raw_spin_lock_irqsave(&pcie->intx_mask_lock, flags);
+       shifted_val = csr_readl(pcie, PAB_INTP_AMBA_MISC_ENB);
+       csr_writel(pcie, (shifted_val & (~mask)), PAB_INTP_AMBA_MISC_ENB);
+       raw_spin_unlock_irqrestore(&pcie->intx_mask_lock, flags);
+ }
+ static void mobiveil_unmask_intx_irq(struct irq_data *data)
+ {
+       struct irq_desc *desc = irq_to_desc(data->irq);
+       struct mobiveil_pcie *pcie;
+       unsigned long flags;
+       u32 shifted_val, mask;
+       pcie = irq_desc_get_chip_data(desc);
+       mask = 1 << ((data->hwirq + PAB_INTX_START) - 1);
+       raw_spin_lock_irqsave(&pcie->intx_mask_lock, flags);
+       shifted_val = csr_readl(pcie, PAB_INTP_AMBA_MISC_ENB);
+       csr_writel(pcie, (shifted_val | mask), PAB_INTP_AMBA_MISC_ENB);
+       raw_spin_unlock_irqrestore(&pcie->intx_mask_lock, flags);
+ }
+ static struct irq_chip intx_irq_chip = {
+       .name = "mobiveil_pcie:intx",
+       .irq_enable = mobiveil_unmask_intx_irq,
+       .irq_disable = mobiveil_mask_intx_irq,
+       .irq_mask = mobiveil_mask_intx_irq,
+       .irq_unmask = mobiveil_unmask_intx_irq,
+ };
+ /* routine to setup the INTx related data */
+ static int mobiveil_pcie_intx_map(struct irq_domain *domain, unsigned int irq,
+               irq_hw_number_t hwirq)
+ {
+       irq_set_chip_and_handler(irq, &intx_irq_chip, handle_level_irq);
+       irq_set_chip_data(irq, domain->host_data);
+       return 0;
+ }
+ /* INTx domain operations structure */
+ static const struct irq_domain_ops intx_domain_ops = {
+       .map = mobiveil_pcie_intx_map,
+ };
+ static struct irq_chip mobiveil_msi_irq_chip = {
+       .name = "Mobiveil PCIe MSI",
+       .irq_mask = pci_msi_mask_irq,
+       .irq_unmask = pci_msi_unmask_irq,
+ };
+ static struct msi_domain_info mobiveil_msi_domain_info = {
+       .flags  = (MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS |
+               MSI_FLAG_MULTI_PCI_MSI | MSI_FLAG_PCI_MSIX),
+       .chip   = &mobiveil_msi_irq_chip,
+ };
+ static void mobiveil_compose_msi_msg(struct irq_data *data, struct msi_msg *msg)
+ {
+       struct mobiveil_pcie *pcie = irq_data_get_irq_chip_data(data);
+       phys_addr_t addr = pcie->pcie_reg_base + (data->hwirq * sizeof(int));
+       msg->address_lo = lower_32_bits(addr);
+       msg->address_hi = upper_32_bits(addr);
+       msg->data = data->hwirq;
+       dev_dbg(&pcie->pdev->dev, "msi#%d address_hi %#x address_lo %#x\n",
+               (int)data->hwirq, msg->address_hi, msg->address_lo);
+ }
+ static int mobiveil_msi_set_affinity(struct irq_data *irq_data,
+               const struct cpumask *mask, bool force)
+ {
+       return -EINVAL;
+ }
+ static struct irq_chip mobiveil_msi_bottom_irq_chip = {
+       .name                   = "Mobiveil MSI",
+       .irq_compose_msi_msg    = mobiveil_compose_msi_msg,
+       .irq_set_affinity       = mobiveil_msi_set_affinity,
+ };
+ static int mobiveil_irq_msi_domain_alloc(struct irq_domain *domain,
+               unsigned int virq, unsigned int nr_irqs, void *args)
+ {
+       struct mobiveil_pcie *pcie = domain->host_data;
+       struct mobiveil_msi *msi = &pcie->msi;
+       unsigned long bit;
+       WARN_ON(nr_irqs != 1);
+       mutex_lock(&msi->lock);
+       bit = find_first_zero_bit(msi->msi_irq_in_use, msi->num_of_vectors);
+       if (bit >= msi->num_of_vectors) {
+               mutex_unlock(&msi->lock);
+               return -ENOSPC;
+       }
+       set_bit(bit, msi->msi_irq_in_use);
+       mutex_unlock(&msi->lock);
+       irq_domain_set_info(domain, virq, bit, &mobiveil_msi_bottom_irq_chip,
+                               domain->host_data, handle_level_irq,
+                               NULL, NULL);
+       return 0;
+ }
+ static void mobiveil_irq_msi_domain_free(struct irq_domain *domain,
+               unsigned int virq, unsigned int nr_irqs)
+ {
+       struct irq_data *d = irq_domain_get_irq_data(domain, virq);
+       struct mobiveil_pcie *pcie = irq_data_get_irq_chip_data(d);
+       struct mobiveil_msi *msi = &pcie->msi;
+       mutex_lock(&msi->lock);
+       if (!test_bit(d->hwirq, msi->msi_irq_in_use)) {
+               dev_err(&pcie->pdev->dev, "trying to free unused MSI#%lu\n",
+                       d->hwirq);
+       } else {
+               __clear_bit(d->hwirq, msi->msi_irq_in_use);
+       }
+       mutex_unlock(&msi->lock);
+ }
+ static const struct irq_domain_ops msi_domain_ops = {
+       .alloc  = mobiveil_irq_msi_domain_alloc,
+       .free   = mobiveil_irq_msi_domain_free,
+ };
+ static int mobiveil_allocate_msi_domains(struct mobiveil_pcie *pcie)
+ {
+       struct device *dev = &pcie->pdev->dev;
+       struct fwnode_handle *fwnode = of_node_to_fwnode(dev->of_node);
+       struct mobiveil_msi *msi = &pcie->msi;
+       mutex_init(&pcie->msi.lock);
+       msi->dev_domain = irq_domain_add_linear(NULL, msi->num_of_vectors,
+                                               &msi_domain_ops, pcie);
+       if (!msi->dev_domain) {
+               dev_err(dev, "failed to create IRQ domain\n");
+               return -ENOMEM;
+       }
+       msi->msi_domain = pci_msi_create_irq_domain(fwnode,
+                               &mobiveil_msi_domain_info, msi->dev_domain);
+       if (!msi->msi_domain) {
+               dev_err(dev, "failed to create MSI domain\n");
+               irq_domain_remove(msi->dev_domain);
+               return -ENOMEM;
+       }
+       return 0;
+ }
+ static int mobiveil_pcie_init_irq_domain(struct mobiveil_pcie *pcie)
+ {
+       struct device *dev = &pcie->pdev->dev;
+       struct device_node *node = dev->of_node;
+       int ret;
+       /* setup INTx */
+       pcie->intx_domain = irq_domain_add_linear(node,
+                               PCI_NUM_INTX, &intx_domain_ops, pcie);
+       if (!pcie->intx_domain) {
+               dev_err(dev, "Failed to get a INTx IRQ domain\n");
+               return -ENODEV;
+       }
+       raw_spin_lock_init(&pcie->intx_mask_lock);
+       /* setup MSI */
+       ret = mobiveil_allocate_msi_domains(pcie);
+       if (ret)
+               return ret;
+       return 0;
+ }
+ static int mobiveil_pcie_probe(struct platform_device *pdev)
+ {
+       struct mobiveil_pcie *pcie;
+       struct pci_bus *bus;
+       struct pci_bus *child;
+       struct pci_host_bridge *bridge;
+       struct device *dev = &pdev->dev;
 -      ret = of_pci_get_host_bridge_resources(node, 0, 0xff,
 -                      &pcie->resources, &iobase);
+       resource_size_t iobase;
+       int ret;
+       /* allocate the PCIe port */
+       bridge = devm_pci_alloc_host_bridge(dev, sizeof(*pcie));
+       if (!bridge)
+               return -ENODEV;
+       pcie = pci_host_bridge_priv(bridge);
+       if (!pcie)
+               return -ENOMEM;
+       pcie->pdev = pdev;
+       ret = mobiveil_pcie_parse_dt(pcie);
+       if (ret) {
+               dev_err(dev, "Parsing DT failed, ret: %x\n", ret);
+               return ret;
+       }
+       INIT_LIST_HEAD(&pcie->resources);
+       /* parse the host bridge base addresses from the device tree file */
++      ret = devm_of_pci_get_host_bridge_resources(dev, 0, 0xff,
++                                                  &pcie->resources, &iobase);
+       if (ret) {
+               dev_err(dev, "Getting bridge resources failed\n");
+               return -ENOMEM;
+       }
+       /*
+        * configure all inbound and outbound windows and prepare the RC for
+        * config access
+        */
+       ret = mobiveil_host_init(pcie);
+       if (ret) {
+               dev_err(dev, "Failed to initialize host\n");
+               goto error;
+       }
+       /* fixup for PCIe class register */
+       csr_writel(pcie, 0x060402ab, PAB_INTP_AXI_PIO_CLASS);
+       /* initialize the IRQ domains */
+       ret = mobiveil_pcie_init_irq_domain(pcie);
+       if (ret) {
+               dev_err(dev, "Failed creating IRQ Domain\n");
+               goto error;
+       }
+       ret = devm_request_pci_bus_resources(dev, &pcie->resources);
+       if (ret)
+               goto error;
+       /* Initialize bridge */
+       list_splice_init(&pcie->resources, &bridge->windows);
+       bridge->dev.parent = dev;
+       bridge->sysdata = pcie;
+       bridge->busnr = pcie->root_bus_nr;
+       bridge->ops = &mobiveil_pcie_ops;
+       bridge->map_irq = of_irq_parse_and_map_pci;
+       bridge->swizzle_irq = pci_common_swizzle;
+       /* setup the kernel resources for the newly added PCIe root bus */
+       ret = pci_scan_root_bus_bridge(bridge);
+       if (ret)
+               goto error;
+       bus = bridge->bus;
+       pci_assign_unassigned_bus_resources(bus);
+       list_for_each_entry(child, &bus->children, node)
+               pcie_bus_configure_settings(child);
+       pci_bus_add_devices(bus);
+       return 0;
+ error:
+       pci_free_resource_list(&pcie->resources);
+       return ret;
+ }
+ static const struct of_device_id mobiveil_pcie_of_match[] = {
+       {.compatible = "mbvl,gpex40-pcie",},
+       {},
+ };
+ MODULE_DEVICE_TABLE(of, mobiveil_pcie_of_match);
+ static struct platform_driver mobiveil_pcie_driver = {
+       .probe = mobiveil_pcie_probe,
+       .driver = {
+                       .name = "mobiveil-pcie",
+                       .of_match_table = mobiveil_pcie_of_match,
+                       .suppress_bind_attrs = true,
+               },
+ };
+ builtin_platform_driver(mobiveil_pcie_driver);
+ MODULE_LICENSE("GPL v2");
+ MODULE_DESCRIPTION("Mobiveil PCIe host controller driver");
+ MODULE_AUTHOR("Subrahmanya Lingappa <l.subrahmanya@mobiveil.co.in>");