PCI: iproc: Add iProc PCIe MSI support

Add PCIe MSI support for both PAXB and PAXC interfaces on all iProc-based
platforms.

The iProc PCIe MSI support deploys an event queue-based implementation.
Each event queue is serviced by a GIC interrupt and can support up to 64
MSI vectors.  Host memory is allocated for the event queues, and each event
queue consists of 64 word-sized entries.  MSI data is written to the lower
16-bit of each entry, whereas the upper 16-bit of the entry is reserved for
the controller for internal processing.

Each event queue is tracked by a head pointer and tail pointer.  Head
pointer indicates the next entry in the event queue to be processed by
the driver and is updated by the driver after processing is done.
The controller uses the tail pointer as the next MSI data insertion
point.  The controller ensures MSI data is flushed to host memory before
updating the tail pointer and then triggering the interrupt.

MSI IRQ affinity is supported by evenly distributing the interrupts to each
CPU core.  MSI vector is moved from one GIC interrupt to another in order
to steer to the target CPU.

Therefore, the actual number of supported MSI vectors is:

  M * 64 / N

where M denotes the number of GIC interrupts (event queues), and N denotes
the number of CPU cores.

This iProc event queue-based MSI support should not be used with newer
platforms with integrated MSI support in the GIC (e.g., giv2m or
gicv3-its).

[bhelgaas: fold in Kconfig fixes from Arnd Bergmann <arnd@arndb.de>]
Signed-off-by: Ray Jui <rjui@broadcom.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Anup Patel <anup.patel@broadcom.com>
Reviewed-by: Vikram Prakash <vikramp@broadcom.com>
Reviewed-by: Scott Branden <sbranden@broadcom.com>
Reviewed-by: Marc Zyngier <marc.zyngier@arm.com>

7 files changed, 735 insertions, 2 deletions

diff --git a/drivers/pci/host/Kconfig b/drivers/pci/host/Kconfig
index 767605a5c659..1cb50d21c21d 100644
--- a/drivers/pci/host/Kconfig
+++ b/drivers/pci/host/Kconfig
@@ -145,6 +145,16 @@ config PCIE_IPROC_BCMA
 	  Say Y here if you want to use the Broadcom iProc PCIe controller
 	  through the BCMA bus interface
 
+config PCIE_IPROC_MSI
+	bool "Broadcom iProc PCIe MSI support"
+	depends on PCIE_IPROC_PLATFORM || PCIE_IPROC_BCMA
+	depends on PCI_MSI
+	select PCI_MSI_IRQ_DOMAIN
+	default ARCH_BCM_IPROC
+	help
+	  Say Y here if you want to enable MSI support for Broadcom's iProc
+	  PCIe controller
+
 config PCIE_ALTERA
 	bool "Altera PCIe controller"
 	depends on ARM || NIOS2
diff --git a/drivers/pci/host/Makefile b/drivers/pci/host/Makefile
index 9d4d3c6924a1..0e4e95e39c21 100644
--- a/drivers/pci/host/Makefile
+++ b/drivers/pci/host/Makefile
@@ -15,6 +15,7 @@ obj-$(CONFIG_PCI_XGENE_MSI) += pci-xgene-msi.o
 obj-$(CONFIG_PCI_LAYERSCAPE) += pci-layerscape.o
 obj-$(CONFIG_PCI_VERSATILE) += pci-versatile.o
 obj-$(CONFIG_PCIE_IPROC) += pcie-iproc.o
+obj-$(CONFIG_PCIE_IPROC_MSI) += pcie-iproc-msi.o
 obj-$(CONFIG_PCIE_IPROC_PLATFORM) += pcie-iproc-platform.o
 obj-$(CONFIG_PCIE_IPROC_BCMA) += pcie-iproc-bcma.o
 obj-$(CONFIG_PCIE_ALTERA) += pcie-altera.o
diff --git a/drivers/pci/host/pcie-iproc-bcma.c b/drivers/pci/host/pcie-iproc-bcma.c
index 96a7d999fd5e..0d7bee4a0d26 100644
--- a/drivers/pci/host/pcie-iproc-bcma.c
+++ b/drivers/pci/host/pcie-iproc-bcma.c
@@ -55,6 +55,7 @@ static int iproc_pcie_bcma_probe(struct bcma_device *bdev)
 	bcma_set_drvdata(bdev, pcie);
 
 	pcie->base = bdev->io_addr;
+	pcie->base_addr = bdev->addr;
 
 	res_mem.start = bdev->addr_s[0];
 	res_mem.end = bdev->addr_s[0] + SZ_128M - 1;
diff --git a/drivers/pci/host/pcie-iproc-msi.c b/drivers/pci/host/pcie-iproc-msi.c
new file mode 100644
index 000000000000..9a2973bdc78a
--- /dev/null
+++ b/drivers/pci/host/pcie-iproc-msi.c
@@ -0,0 +1,675 @@
+/*
+ * Copyright (C) 2015 Broadcom Corporation
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation version 2.
+ *
+ * This program is distributed "as is" WITHOUT ANY WARRANTY of any
+ * kind, whether express or implied; without even the implied warranty
+ * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/irqchip/chained_irq.h>
+#include <linux/irqdomain.h>
+#include <linux/msi.h>
+#include <linux/of_irq.h>
+#include <linux/of_pci.h>
+#include <linux/pci.h>
+
+#include "pcie-iproc.h"
+
+#define IPROC_MSI_INTR_EN_SHIFT        11
+#define IPROC_MSI_INTR_EN              BIT(IPROC_MSI_INTR_EN_SHIFT)
+#define IPROC_MSI_INT_N_EVENT_SHIFT    1
+#define IPROC_MSI_INT_N_EVENT          BIT(IPROC_MSI_INT_N_EVENT_SHIFT)
+#define IPROC_MSI_EQ_EN_SHIFT          0
+#define IPROC_MSI_EQ_EN                BIT(IPROC_MSI_EQ_EN_SHIFT)
+
+#define IPROC_MSI_EQ_MASK              0x3f
+
+/* Max number of GIC interrupts */
+#define NR_HW_IRQS                     6
+
+/* Number of entries in each event queue */
+#define EQ_LEN                         64
+
+/* Size of each event queue memory region */
+#define EQ_MEM_REGION_SIZE             SZ_4K
+
+/* Size of each MSI address region */
+#define MSI_MEM_REGION_SIZE            SZ_4K
+
+enum iproc_msi_reg {
+	IPROC_MSI_EQ_PAGE = 0,
+	IPROC_MSI_EQ_PAGE_UPPER,
+	IPROC_MSI_PAGE,
+	IPROC_MSI_PAGE_UPPER,
+	IPROC_MSI_CTRL,
+	IPROC_MSI_EQ_HEAD,
+	IPROC_MSI_EQ_TAIL,
+	IPROC_MSI_INTS_EN,
+	IPROC_MSI_REG_SIZE,
+};
+
+struct iproc_msi;
+
+/**
+ * iProc MSI group
+ *
+ * One MSI group is allocated per GIC interrupt, serviced by one iProc MSI
+ * event queue.
+ *
+ * @msi: pointer to iProc MSI data
+ * @gic_irq: GIC interrupt
+ * @eq: Event queue number
+ */
+struct iproc_msi_grp {
+	struct iproc_msi *msi;
+	int gic_irq;
+	unsigned int eq;
+};
+
+/**
+ * iProc event queue based MSI
+ *
+ * Only meant to be used on platforms without MSI support integrated into the
+ * GIC.
+ *
+ * @pcie: pointer to iProc PCIe data
+ * @reg_offsets: MSI register offsets
+ * @grps: MSI groups
+ * @nr_irqs: number of total interrupts connected to GIC
+ * @nr_cpus: number of toal CPUs
+ * @has_inten_reg: indicates the MSI interrupt enable register needs to be
+ * set explicitly (required for some legacy platforms)
+ * @bitmap: MSI vector bitmap
+ * @bitmap_lock: lock to protect access to the MSI bitmap
+ * @nr_msi_vecs: total number of MSI vectors
+ * @inner_domain: inner IRQ domain
+ * @msi_domain: MSI IRQ domain
+ * @nr_eq_region: required number of 4K aligned memory region for MSI event
+ * queues
+ * @nr_msi_region: required number of 4K aligned address region for MSI posted
+ * writes
+ * @eq_cpu: pointer to allocated memory region for MSI event queues
+ * @eq_dma: DMA address of MSI event queues
+ * @msi_addr: MSI address
+ */
+struct iproc_msi {
+	struct iproc_pcie *pcie;
+	const u16 (*reg_offsets)[IPROC_MSI_REG_SIZE];
+	struct iproc_msi_grp *grps;
+	int nr_irqs;
+	int nr_cpus;
+	bool has_inten_reg;
+	unsigned long *bitmap;
+	struct mutex bitmap_lock;
+	unsigned int nr_msi_vecs;
+	struct irq_domain *inner_domain;
+	struct irq_domain *msi_domain;
+	unsigned int nr_eq_region;
+	unsigned int nr_msi_region;
+	void *eq_cpu;
+	dma_addr_t eq_dma;
+	phys_addr_t msi_addr;
+};
+
+static const u16 iproc_msi_reg_paxb[NR_HW_IRQS][IPROC_MSI_REG_SIZE] = {
+	{ 0x200, 0x2c0, 0x204, 0x2c4, 0x210, 0x250, 0x254, 0x208 },
+	{ 0x200, 0x2c0, 0x204, 0x2c4, 0x214, 0x258, 0x25c, 0x208 },
+	{ 0x200, 0x2c0, 0x204, 0x2c4, 0x218, 0x260, 0x264, 0x208 },
+	{ 0x200, 0x2c0, 0x204, 0x2c4, 0x21c, 0x268, 0x26c, 0x208 },
+	{ 0x200, 0x2c0, 0x204, 0x2c4, 0x220, 0x270, 0x274, 0x208 },
+	{ 0x200, 0x2c0, 0x204, 0x2c4, 0x224, 0x278, 0x27c, 0x208 },
+};
+
+static const u16 iproc_msi_reg_paxc[NR_HW_IRQS][IPROC_MSI_REG_SIZE] = {
+	{ 0xc00, 0xc04, 0xc08, 0xc0c, 0xc40, 0xc50, 0xc60 },
+	{ 0xc10, 0xc14, 0xc18, 0xc1c, 0xc44, 0xc54, 0xc64 },
+	{ 0xc20, 0xc24, 0xc28, 0xc2c, 0xc48, 0xc58, 0xc68 },
+	{ 0xc30, 0xc34, 0xc38, 0xc3c, 0xc4c, 0xc5c, 0xc6c },
+};
+
+static inline u32 iproc_msi_read_reg(struct iproc_msi *msi,
+				     enum iproc_msi_reg reg,
+				     unsigned int eq)
+{
+	struct iproc_pcie *pcie = msi->pcie;
+
+	return readl_relaxed(pcie->base + msi->reg_offsets[eq][reg]);
+}
+
+static inline void iproc_msi_write_reg(struct iproc_msi *msi,
+				       enum iproc_msi_reg reg,
+				       int eq, u32 val)
+{
+	struct iproc_pcie *pcie = msi->pcie;
+
+	writel_relaxed(val, pcie->base + msi->reg_offsets[eq][reg]);
+}
+
+static inline u32 hwirq_to_group(struct iproc_msi *msi, unsigned long hwirq)
+{
+	return (hwirq % msi->nr_irqs);
+}
+
+static inline unsigned int iproc_msi_addr_offset(struct iproc_msi *msi,
+						 unsigned long hwirq)
+{
+	if (msi->nr_msi_region > 1)
+		return hwirq_to_group(msi, hwirq) * MSI_MEM_REGION_SIZE;
+	else
+		return hwirq_to_group(msi, hwirq) * sizeof(u32);
+}
+
+static inline unsigned int iproc_msi_eq_offset(struct iproc_msi *msi, u32 eq)
+{
+	if (msi->nr_eq_region > 1)
+		return eq * EQ_MEM_REGION_SIZE;
+	else
+		return eq * EQ_LEN * sizeof(u32);
+}
+
+static struct irq_chip iproc_msi_irq_chip = {
+	.name = "iProc-MSI",
+};
+
+static struct msi_domain_info iproc_msi_domain_info = {
+	.flags = MSI_FLAG_USE_DEF_DOM_OPS | MSI_FLAG_USE_DEF_CHIP_OPS |
+		MSI_FLAG_PCI_MSIX,
+	.chip = &iproc_msi_irq_chip,
+};
+
+/*
+ * In iProc PCIe core, each MSI group is serviced by a GIC interrupt and a
+ * dedicated event queue.  Each MSI group can support up to 64 MSI vectors.
+ *
+ * The number of MSI groups varies between different iProc SoCs.  The total
+ * number of CPU cores also varies.  To support MSI IRQ affinity, we
+ * distribute GIC interrupts across all available CPUs.  MSI vector is moved
+ * from one GIC interrupt to another to steer to the target CPU.
+ *
+ * Assuming:
+ * - the number of MSI groups is M
+ * - the number of CPU cores is N
+ * - M is always a multiple of N
+ *
+ * Total number of raw MSI vectors = M * 64
+ * Total number of supported MSI vectors = (M * 64) / N
+ */
+static inline int hwirq_to_cpu(struct iproc_msi *msi, unsigned long hwirq)
+{
+	return (hwirq % msi->nr_cpus);
+}
+
+static inline unsigned long hwirq_to_canonical_hwirq(struct iproc_msi *msi,
+						     unsigned long hwirq)
+{
+	return (hwirq - hwirq_to_cpu(msi, hwirq));
+}
+
+static int iproc_msi_irq_set_affinity(struct irq_data *data,
+				      const struct cpumask *mask, bool force)
+{
+	struct iproc_msi *msi = irq_data_get_irq_chip_data(data);
+	int target_cpu = cpumask_first(mask);
+	int curr_cpu;
+
+	curr_cpu = hwirq_to_cpu(msi, data->hwirq);
+	if (curr_cpu == target_cpu)
+		return IRQ_SET_MASK_OK_DONE;
+
+	/* steer MSI to the target CPU */
+	data->hwirq = hwirq_to_canonical_hwirq(msi, data->hwirq) + target_cpu;
+
+	return IRQ_SET_MASK_OK;
+}
+
+static void iproc_msi_irq_compose_msi_msg(struct irq_data *data,
+					  struct msi_msg *msg)
+{
+	struct iproc_msi *msi = irq_data_get_irq_chip_data(data);
+	dma_addr_t addr;
+
+	addr = msi->msi_addr + iproc_msi_addr_offset(msi, data->hwirq);
+	msg->address_lo = lower_32_bits(addr);
+	msg->address_hi = upper_32_bits(addr);
+	msg->data = data->hwirq;
+}
+
+static struct irq_chip iproc_msi_bottom_irq_chip = {
+	.name = "MSI",
+	.irq_set_affinity = iproc_msi_irq_set_affinity,
+	.irq_compose_msi_msg = iproc_msi_irq_compose_msi_msg,
+};
+
+static int iproc_msi_irq_domain_alloc(struct irq_domain *domain,
+				      unsigned int virq, unsigned int nr_irqs,
+				      void *args)
+{
+	struct iproc_msi *msi = domain->host_data;
+	int hwirq;
+
+	mutex_lock(&msi->bitmap_lock);
+
+	/* Allocate 'nr_cpus' number of MSI vectors each time */
+	hwirq = bitmap_find_next_zero_area(msi->bitmap, msi->nr_msi_vecs, 0,
+					   msi->nr_cpus, 0);
+	if (hwirq < msi->nr_msi_vecs) {
+		bitmap_set(msi->bitmap, hwirq, msi->nr_cpus);
+	} else {
+		mutex_unlock(&msi->bitmap_lock);
+		return -ENOSPC;
+	}
+
+	mutex_unlock(&msi->bitmap_lock);
+
+	irq_domain_set_info(domain, virq, hwirq, &iproc_msi_bottom_irq_chip,
+			    domain->host_data, handle_simple_irq, NULL, NULL);
+
+	return 0;
+}
+
+static void iproc_msi_irq_domain_free(struct irq_domain *domain,
+				      unsigned int virq, unsigned int nr_irqs)
+{
+	struct irq_data *data = irq_domain_get_irq_data(domain, virq);
+	struct iproc_msi *msi = irq_data_get_irq_chip_data(data);
+	unsigned int hwirq;
+
+	mutex_lock(&msi->bitmap_lock);
+
+	hwirq = hwirq_to_canonical_hwirq(msi, data->hwirq);
+	bitmap_clear(msi->bitmap, hwirq, msi->nr_cpus);
+
+	mutex_unlock(&msi->bitmap_lock);
+
+	irq_domain_free_irqs_parent(domain, virq, nr_irqs);
+}
+
+static const struct irq_domain_ops msi_domain_ops = {
+	.alloc = iproc_msi_irq_domain_alloc,
+	.free = iproc_msi_irq_domain_free,
+};
+
+static inline u32 decode_msi_hwirq(struct iproc_msi *msi, u32 eq, u32 head)
+{
+	u32 *msg, hwirq;
+	unsigned int offs;
+
+	offs = iproc_msi_eq_offset(msi, eq) + head * sizeof(u32);
+	msg = (u32 *)(msi->eq_cpu + offs);
+	hwirq = *msg & IPROC_MSI_EQ_MASK;
+
+	/*
+	 * Since we have multiple hwirq mapped to a single MSI vector,
+	 * now we need to derive the hwirq at CPU0.  It can then be used to
+	 * mapped back to virq.
+	 */
+	return hwirq_to_canonical_hwirq(msi, hwirq);
+}
+
+static void iproc_msi_handler(struct irq_desc *desc)
+{
+	struct irq_chip *chip = irq_desc_get_chip(desc);
+	struct iproc_msi_grp *grp;
+	struct iproc_msi *msi;
+	struct iproc_pcie *pcie;
+	u32 eq, head, tail, nr_events;
+	unsigned long hwirq;
+	int virq;
+
+	chained_irq_enter(chip, desc);
+
+	grp = irq_desc_get_handler_data(desc);
+	msi = grp->msi;
+	pcie = msi->pcie;
+	eq = grp->eq;
+
+	/*
+	 * iProc MSI event queue is tracked by head and tail pointers.  Head
+	 * pointer indicates the next entry (MSI data) to be consumed by SW in
+	 * the queue and needs to be updated by SW.  iProc MSI core uses the
+	 * tail pointer as the next data insertion point.
+	 *
+	 * Entries between head and tail pointers contain valid MSI data.  MSI
+	 * data is guaranteed to be in the event queue memory before the tail
+	 * pointer is updated by the iProc MSI core.
+	 */
+	head = iproc_msi_read_reg(msi, IPROC_MSI_EQ_HEAD,
+				  eq) & IPROC_MSI_EQ_MASK;
+	do {
+		tail = iproc_msi_read_reg(msi, IPROC_MSI_EQ_TAIL,
+					  eq) & IPROC_MSI_EQ_MASK;
+
+		/*
+		 * Figure out total number of events (MSI data) to be
+		 * processed.
+		 */
+		nr_events = (tail < head) ?
+			(EQ_LEN - (head - tail)) : (tail - head);
+		if (!nr_events)
+			break;
+
+		/* process all outstanding events */
+		while (nr_events--) {
+			hwirq = decode_msi_hwirq(msi, eq, head);
+			virq = irq_find_mapping(msi->inner_domain, hwirq);
+			generic_handle_irq(virq);
+
+			head++;
+			head %= EQ_LEN;
+		}
+
+		/*
+		 * Now all outstanding events have been processed.  Update the
+		 * head pointer.
+		 */
+		iproc_msi_write_reg(msi, IPROC_MSI_EQ_HEAD, eq, head);
+
+		/*
+		 * Now go read the tail pointer again to see if there are new
+		 * oustanding events that came in during the above window.
+		 */
+	} while (true);
+
+	chained_irq_exit(chip, desc);
+}
+
+static void iproc_msi_enable(struct iproc_msi *msi)
+{
+	int i, eq;
+	u32 val;
+
+	/* Program memory region for each event queue */
+	for (i = 0; i < msi->nr_eq_region; i++) {
+		dma_addr_t addr = msi->eq_dma + (i * EQ_MEM_REGION_SIZE);
+
+		iproc_msi_write_reg(msi, IPROC_MSI_EQ_PAGE, i,
+				    lower_32_bits(addr));
+		iproc_msi_write_reg(msi, IPROC_MSI_EQ_PAGE_UPPER, i,
+				    upper_32_bits(addr));
+	}
+
+	/* Program address region for MSI posted writes */
+	for (i = 0; i < msi->nr_msi_region; i++) {
+		phys_addr_t addr = msi->msi_addr + (i * MSI_MEM_REGION_SIZE);
+
+		iproc_msi_write_reg(msi, IPROC_MSI_PAGE, i,
+				    lower_32_bits(addr));
+		iproc_msi_write_reg(msi, IPROC_MSI_PAGE_UPPER, i,
+				    upper_32_bits(addr));
+	}
+
+	for (eq = 0; eq < msi->nr_irqs; eq++) {
+		/* Enable MSI event queue */
+		val = IPROC_MSI_INTR_EN | IPROC_MSI_INT_N_EVENT |
+			IPROC_MSI_EQ_EN;
+		iproc_msi_write_reg(msi, IPROC_MSI_CTRL, eq, val);
+
+		/*
+		 * Some legacy platforms require the MSI interrupt enable
+		 * register to be set explicitly.
+		 */
+		if (msi->has_inten_reg) {
+			val = iproc_msi_read_reg(msi, IPROC_MSI_INTS_EN, eq);
+			val |= BIT(eq);
+			iproc_msi_write_reg(msi, IPROC_MSI_INTS_EN, eq, val);
+		}
+	}
+}
+
+static void iproc_msi_disable(struct iproc_msi *msi)
+{
+	u32 eq, val;
+
+	for (eq = 0; eq < msi->nr_irqs; eq++) {
+		if (msi->has_inten_reg) {
+			val = iproc_msi_read_reg(msi, IPROC_MSI_INTS_EN, eq);
+			val &= ~BIT(eq);
+			iproc_msi_write_reg(msi, IPROC_MSI_INTS_EN, eq, val);
+		}
+
+		val = iproc_msi_read_reg(msi, IPROC_MSI_CTRL, eq);
+		val &= ~(IPROC_MSI_INTR_EN | IPROC_MSI_INT_N_EVENT |
+			 IPROC_MSI_EQ_EN);
+		iproc_msi_write_reg(msi, IPROC_MSI_CTRL, eq, val);
+	}
+}
+
+static int iproc_msi_alloc_domains(struct device_node *node,
+				   struct iproc_msi *msi)
+{
+	msi->inner_domain = irq_domain_add_linear(NULL, msi->nr_msi_vecs,
+						  &msi_domain_ops, msi);
+	if (!msi->inner_domain)
+		return -ENOMEM;
+
+	msi->msi_domain = pci_msi_create_irq_domain(of_node_to_fwnode(node),
+						    &iproc_msi_domain_info,
+						    msi->inner_domain);
+	if (!msi->msi_domain) {
+		irq_domain_remove(msi->inner_domain);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void iproc_msi_free_domains(struct iproc_msi *msi)
+{
+	if (msi->msi_domain)
+		irq_domain_remove(msi->msi_domain);
+
+	if (msi->inner_domain)
+		irq_domain_remove(msi->inner_domain);
+}
+
+static void iproc_msi_irq_free(struct iproc_msi *msi, unsigned int cpu)
+{
+	int i;
+
+	for (i = cpu; i < msi->nr_irqs; i += msi->nr_cpus) {
+		irq_set_chained_handler_and_data(msi->grps[i].gic_irq,
+						 NULL, NULL);
+	}
+}
+
+static int iproc_msi_irq_setup(struct iproc_msi *msi, unsigned int cpu)
+{
+	int i, ret;
+	cpumask_var_t mask;
+	struct iproc_pcie *pcie = msi->pcie;
+
+	for (i = cpu; i < msi->nr_irqs; i += msi->nr_cpus) {
+		irq_set_chained_handler_and_data(msi->grps[i].gic_irq,
+						 iproc_msi_handler,
+						 &msi->grps[i]);
+		/* Dedicate GIC interrupt to each CPU core */
+		if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
+			cpumask_clear(mask);
+			cpumask_set_cpu(cpu, mask);
+			ret = irq_set_affinity(msi->grps[i].gic_irq, mask);
+			if (ret)
+				dev_err(pcie->dev,
+					"failed to set affinity for IRQ%d\n",
+					msi->grps[i].gic_irq);
+			free_cpumask_var(mask);
+		} else {
+			dev_err(pcie->dev, "failed to alloc CPU mask\n");
+			ret = -EINVAL;
+		}
+
+		if (ret) {
+			/* Free all configured/unconfigured IRQs */
+			iproc_msi_irq_free(msi, cpu);
+			return ret;
+		}
+	}
+
+	return 0;
+}
+
+int iproc_msi_init(struct iproc_pcie *pcie, struct device_node *node)
+{
+	struct iproc_msi *msi;
+	int i, ret;
+	unsigned int cpu;
+
+	if (!of_device_is_compatible(node, "brcm,iproc-msi"))
+		return -ENODEV;
+
+	if (!of_find_property(node, "msi-controller", NULL))
+		return -ENODEV;
+
+	if (pcie->msi)
+		return -EBUSY;
+
+	msi = devm_kzalloc(pcie->dev, sizeof(*msi), GFP_KERNEL);
+	if (!msi)
+		return -ENOMEM;
+
+	msi->pcie = pcie;
+	pcie->msi = msi;
+	msi->msi_addr = pcie->base_addr;
+	mutex_init(&msi->bitmap_lock);
+	msi->nr_cpus = num_possible_cpus();
+
+	msi->nr_irqs = of_irq_count(node);
+	if (!msi->nr_irqs) {
+		dev_err(pcie->dev, "found no MSI GIC interrupt\n");
+		return -ENODEV;
+	}
+
+	if (msi->nr_irqs > NR_HW_IRQS) {
+		dev_warn(pcie->dev, "too many MSI GIC interrupts defined %d\n",
+			 msi->nr_irqs);
+		msi->nr_irqs = NR_HW_IRQS;
+	}
+
+	if (msi->nr_irqs < msi->nr_cpus) {
+		dev_err(pcie->dev,
+			"not enough GIC interrupts for MSI affinity\n");
+		return -EINVAL;
+	}
+
+	if (msi->nr_irqs % msi->nr_cpus != 0) {
+		msi->nr_irqs -= msi->nr_irqs % msi->nr_cpus;
+		dev_warn(pcie->dev, "Reducing number of interrupts to %d\n",
+			 msi->nr_irqs);
+	}
+
+	switch (pcie->type) {
+	case IPROC_PCIE_PAXB:
+		msi->reg_offsets = iproc_msi_reg_paxb;
+		msi->nr_eq_region = 1;
+		msi->nr_msi_region = 1;
+		break;
+	case IPROC_PCIE_PAXC:
+		msi->reg_offsets = iproc_msi_reg_paxc;
+		msi->nr_eq_region = msi->nr_irqs;
+		msi->nr_msi_region = msi->nr_irqs;
+		break;
+	default:
+		dev_err(pcie->dev, "incompatible iProc PCIe interface\n");
+		return -EINVAL;
+	}
+
+	if (of_find_property(node, "brcm,pcie-msi-inten", NULL))
+		msi->has_inten_reg = true;
+
+	msi->nr_msi_vecs = msi->nr_irqs * EQ_LEN;
+	msi->bitmap = devm_kcalloc(pcie->dev, BITS_TO_LONGS(msi->nr_msi_vecs),
+				   sizeof(*msi->bitmap), GFP_KERNEL);
+	if (!msi->bitmap)
+		return -ENOMEM;
+
+	msi->grps = devm_kcalloc(pcie->dev, msi->nr_irqs, sizeof(*msi->grps),
+				 GFP_KERNEL);
+	if (!msi->grps)
+		return -ENOMEM;
+
+	for (i = 0; i < msi->nr_irqs; i++) {
+		unsigned int irq = irq_of_parse_and_map(node, i);
+
+		if (!irq) {
+			dev_err(pcie->dev, "unable to parse/map interrupt\n");
+			ret = -ENODEV;
+			goto free_irqs;
+		}
+		msi->grps[i].gic_irq = irq;
+		msi->grps[i].msi = msi;
+		msi->grps[i].eq = i;
+	}
+
+	/* Reserve memory for event queue and make sure memories are zeroed */
+	msi->eq_cpu = dma_zalloc_coherent(pcie->dev,
+					  msi->nr_eq_region * EQ_MEM_REGION_SIZE,
+					  &msi->eq_dma, GFP_KERNEL);
+	if (!msi->eq_cpu) {
+		ret = -ENOMEM;
+		goto free_irqs;
+	}
+
+	ret = iproc_msi_alloc_domains(node, msi);
+	if (ret) {
+		dev_err(pcie->dev, "failed to create MSI domains\n");
+		goto free_eq_dma;
+	}
+
+	for_each_online_cpu(cpu) {
+		ret = iproc_msi_irq_setup(msi, cpu);
+		if (ret)
+			goto free_msi_irq;
+	}
+
+	iproc_msi_enable(msi);
+
+	return 0;
+
+free_msi_irq:
+	for_each_online_cpu(cpu)
+		iproc_msi_irq_free(msi, cpu);
+	iproc_msi_free_domains(msi);
+
+free_eq_dma:
+	dma_free_coherent(pcie->dev, msi->nr_eq_region * EQ_MEM_REGION_SIZE,
+			  msi->eq_cpu, msi->eq_dma);
+
+free_irqs:
+	for (i = 0; i < msi->nr_irqs; i++) {
+		if (msi->grps[i].gic_irq)
+			irq_dispose_mapping(msi->grps[i].gic_irq);
+	}
+	pcie->msi = NULL;
+	return ret;
+}
+EXPORT_SYMBOL(iproc_msi_init);
+
+void iproc_msi_exit(struct iproc_pcie *pcie)
+{
+	struct iproc_msi *msi = pcie->msi;
+	unsigned int i, cpu;
+
+	if (!msi)
+		return;
+
+	iproc_msi_disable(msi);
+
+	for_each_online_cpu(cpu)
+		iproc_msi_irq_free(msi, cpu);
+
+	iproc_msi_free_domains(msi);
+
+	dma_free_coherent(pcie->dev, msi->nr_eq_region * EQ_MEM_REGION_SIZE,
+			  msi->eq_cpu, msi->eq_dma);
+
+	for (i = 0; i < msi->nr_irqs; i++) {
+		if (msi->grps[i].gic_irq)
+			irq_dispose_mapping(msi->grps[i].gic_irq);
+	}
+}
+EXPORT_SYMBOL(iproc_msi_exit);
diff --git a/drivers/pci/host/pcie-iproc-platform.c b/drivers/pci/host/pcie-iproc-platform.c
index e8b32d856bd8..1738c5288eb6 100644
--- a/drivers/pci/host/pcie-iproc-platform.c
+++ b/drivers/pci/host/pcie-iproc-platform.c
@@ -71,6 +71,7 @@ static int iproc_pcie_pltfm_probe(struct platform_device *pdev)
 		dev_err(pcie->dev, "unable to map controller registers\n");
 		return -ENOMEM;
 	}
+	pcie->base_addr = reg.start;
 
 	if (of_property_read_bool(np, "brcm,pcie-ob")) {
 		u32 val;
diff --git a/drivers/pci/host/pcie-iproc.c b/drivers/pci/host/pcie-iproc.c
index 0e11f85ba7b0..5816bceddb65 100644
--- a/drivers/pci/host/pcie-iproc.c
+++ b/drivers/pci/host/pcie-iproc.c
@@ -440,6 +440,26 @@ static int iproc_pcie_map_ranges(struct iproc_pcie *pcie,
 	return 0;
 }
 
+static int iproc_pcie_msi_enable(struct iproc_pcie *pcie)
+{
+	struct device_node *msi_node;
+
+	msi_node = of_parse_phandle(pcie->dev->of_node, "msi-parent", 0);
+	if (!msi_node)
+		return -ENODEV;
+
+	/*
+	 * If another MSI controller is being used, the call below should fail
+	 * but that is okay
+	 */
+	return iproc_msi_init(pcie, msi_node);
+}
+
+static void iproc_pcie_msi_disable(struct iproc_pcie *pcie)
+{
+	iproc_msi_exit(pcie);
+}
+
 int iproc_pcie_setup(struct iproc_pcie *pcie, struct list_head *res)
 {
 	int ret;
@@ -507,6 +527,10 @@ int iproc_pcie_setup(struct iproc_pcie *pcie, struct list_head *res)
 
 	iproc_pcie_enable(pcie);
 
+	if (IS_ENABLED(CONFIG_PCI_MSI))
+		if (iproc_pcie_msi_enable(pcie))
+			dev_info(pcie->dev, "not using iProc MSI\n");
+
 	pci_scan_child_bus(bus);
 	pci_assign_unassigned_bus_resources(bus);
 	pci_fixup_irqs(pci_common_swizzle, pcie->map_irq);
@@ -531,6 +555,8 @@ int iproc_pcie_remove(struct iproc_pcie *pcie)
 	pci_stop_root_bus(pcie->root_bus);
 	pci_remove_root_bus(pcie->root_bus);
 
+	iproc_pcie_msi_disable(pcie);
+
 	phy_power_off(pcie->phy);
 	phy_exit(pcie->phy);
 
diff --git a/drivers/pci/host/pcie-iproc.h b/drivers/pci/host/pcie-iproc.h
index a38e96ae156d..e84d93c53c7b 100644
--- a/drivers/pci/host/pcie-iproc.h
+++ b/drivers/pci/host/pcie-iproc.h
@@ -41,6 +41,8 @@ struct iproc_pcie_ob {
 	resource_size_t window_size;
 };
 
+struct iproc_msi;
+
 /**
  * iProc PCIe device
  *
@@ -48,19 +50,21 @@ struct iproc_pcie_ob {
  * @type: iProc PCIe interface type
  * @reg_offsets: register offsets
  * @base: PCIe host controller I/O register base
+ * @base_addr: PCIe host controller register base physical address
  * @sysdata: Per PCI controller data (ARM-specific)
  * @root_bus: pointer to root bus
  * @phy: optional PHY device that controls the Serdes
- * @irqs: interrupt IDs
  * @map_irq: function callback to map interrupts
- * @need_ob_cfg: indidates SW needs to configure the outbound mapping window
+ * @need_ob_cfg: indicates SW needs to configure the outbound mapping window
  * @ob: outbound mapping parameters
+ * @msi: MSI data
  */
 struct iproc_pcie {
 	struct device *dev;
 	enum iproc_pcie_type type;
 	const u16 *reg_offsets;
 	void __iomem *base;
+	phys_addr_t base_addr;
 #ifdef CONFIG_ARM
 	struct pci_sys_data sysdata;
 #endif
@@ -69,9 +73,24 @@ struct iproc_pcie {
 	int (*map_irq)(const struct pci_dev *, u8, u8);
 	bool need_ob_cfg;
 	struct iproc_pcie_ob ob;
+	struct iproc_msi *msi;
 };
 
 int iproc_pcie_setup(struct iproc_pcie *pcie, struct list_head *res);
 int iproc_pcie_remove(struct iproc_pcie *pcie);
 
+#ifdef CONFIG_PCIE_IPROC_MSI
+int iproc_msi_init(struct iproc_pcie *pcie, struct device_node *node);
+void iproc_msi_exit(struct iproc_pcie *pcie);
+#else
+static inline int iproc_msi_init(struct iproc_pcie *pcie,
+				 struct device_node *node)
+{
+	return -ENODEV;
+}
+static inline void iproc_msi_exit(struct iproc_pcie *pcie)
+{
+}
+#endif
+
 #endif /* _PCIE_IPROC_H */