drm/amd/display: Try RGB before YCbCr 4:4:4 in stream validationgrafted

amdgpu_dm_create_validate_stream_for_sink() walks encoding_order[] and
uses the first encoding that validates. YCbCr 4:4:4 is listed before
RGB, so an HDMI sink that advertises 4:4:4 gets YCbCr 4:4:4 whenever the
"color format" property is left at AUTO, even though RGB fits the same
link.

That contradicts the documented AUTO behaviour for HDMI in enum
drm_connector_color_format (RGB, falling back to YCbCr 4:2:0 only when
the bandwidth is not available or the mode is 4:2:0-only), which the
amdgpu implementation of the property also describes. It also leaves
the "Broadcast RGB" property without effect on such sinks, since the
quantization range it selects only applies to RGB output.

Try RGB first. The mask still holds every encoding the sink supports,
so a mode that cannot carry RGB falls back exactly as before.

For reference, v7.2 picked RGB here unless YCbCr 4:4:4 was forced
through debugfs, while earlier kernels picked YCbCr 4:4:4 for any HDMI
sink that advertised it.

Fixes: 0b0ff65d3ca1 ("drm/amd/display: Refactor stream validation")
Suggested-by: Adolfo Rodrigues <adolfotregosa@gmail.com>
Assisted-by: Claude Code:claude-fable-5-1
Signed-off-by: Adrian Betschart <adrian.betschart@cinemaone.ch>
Reviewed-by: Fangzhi Zuo <jerry.zuo@amd.com>
Tested-by: Adolfo Rodrigues <adolfotregosa@gmail.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit 7de432bc753133764dc40d37f9388da56de251e7)

6 files changed, 3724 insertions, 0 deletions

diff --git a/arch/powerpc/sysdev/xive/Kconfig b/arch/powerpc/sysdev/xive/Kconfig
new file mode 100644
index 000000000..785c292d1
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/Kconfig
@@ -0,0 +1,14 @@
+# SPDX-License-Identifier: GPL-2.0
+config PPC_XIVE
+	bool
+	select PPC_SMP_MUXED_IPI
+	select HARDIRQS_SW_RESEND
+
+config PPC_XIVE_NATIVE
+	bool
+	select PPC_XIVE
+	depends on PPC_POWERNV
+
+config PPC_XIVE_SPAPR
+	bool
+	select PPC_XIVE
diff --git a/arch/powerpc/sysdev/xive/Makefile b/arch/powerpc/sysdev/xive/Makefile
new file mode 100644
index 000000000..e51088838
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/Makefile
@@ -0,0 +1,5 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-y				+= common.o
+obj-$(CONFIG_PPC_XIVE_NATIVE)	+= native.o
+obj-$(CONFIG_PPC_XIVE_SPAPR)	+= spapr.o
diff --git a/arch/powerpc/sysdev/xive/common.c b/arch/powerpc/sysdev/xive/common.c
new file mode 100644
index 000000000..ed3d3e26c
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/common.c
@@ -0,0 +1,1868 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016,2017 IBM Corporation.
+ */
+
+#define pr_fmt(fmt) "xive: " fmt
+
+#include <linux/types.h>
+#include <linux/threads.h>
+#include <linux/kernel.h>
+#include <linux/irq.h>
+#include <linux/irqdomain.h>
+#include <linux/debugfs.h>
+#include <linux/smp.h>
+#include <linux/interrupt.h>
+#include <linux/seq_file.h>
+#include <linux/init.h>
+#include <linux/cpu.h>
+#include <linux/of.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/msi.h>
+#include <linux/vmalloc.h>
+
+#include <asm/io.h>
+#include <asm/smp.h>
+#include <asm/machdep.h>
+#include <asm/irq.h>
+#include <asm/errno.h>
+#include <asm/xive.h>
+#include <asm/xive-regs.h>
+#include <asm/xmon.h>
+
+#include "xive-internal.h"
+
+#undef DEBUG_FLUSH
+#undef DEBUG_ALL
+
+#ifdef DEBUG_ALL
+#define DBG_VERBOSE(fmt, ...)	pr_devel("cpu %d - " fmt, \
+					 smp_processor_id(), ## __VA_ARGS__)
+#else
+#define DBG_VERBOSE(fmt...)	do { } while(0)
+#endif
+
+bool __xive_enabled;
+EXPORT_SYMBOL_GPL(__xive_enabled);
+bool xive_cmdline_disabled;
+
+/* We use only one priority for now */
+static u8 xive_irq_priority;
+
+/* TIMA exported to KVM */
+void __iomem *xive_tima;
+EXPORT_SYMBOL_GPL(xive_tima);
+u32 xive_tima_offset;
+
+/* Backend ops */
+static const struct xive_ops *xive_ops;
+
+/* Our global interrupt domain */
+static struct irq_domain *xive_irq_domain;
+
+#ifdef CONFIG_SMP
+/* The IPIs use the same logical irq number when on the same chip */
+static struct xive_ipi_desc {
+	unsigned int irq;
+	char name[16];
+	atomic_t started;
+} *xive_ipis;
+
+/*
+ * Use early_cpu_to_node() for hot-plugged CPUs
+ */
+static unsigned int xive_ipi_cpu_to_irq(unsigned int cpu)
+{
+	if (!xive_ipis)
+		return XIVE_BAD_IRQ;
+	return xive_ipis[early_cpu_to_node(cpu)].irq;
+}
+#endif
+
+/* Xive state for each CPU */
+static DEFINE_PER_CPU(struct xive_cpu *, xive_cpu);
+
+/* An invalid CPU target */
+#define XIVE_INVALID_TARGET	(-1)
+
+/*
+ * Global toggle to switch on/off StoreEOI
+ */
+static bool xive_store_eoi = true;
+
+static bool xive_is_store_eoi(struct xive_irq_data *xd)
+{
+	return xd->flags & XIVE_IRQ_FLAG_STORE_EOI && xive_store_eoi;
+}
+
+/*
+ * Read the next entry in a queue, return its content if it's valid
+ * or 0 if there is no new entry.
+ *
+ * The queue pointer is moved forward unless "just_peek" is set
+ */
+static u32 xive_read_eq(struct xive_q *q, bool just_peek)
+{
+	u32 cur;
+
+	if (!q->qpage)
+		return 0;
+	cur = be32_to_cpup(q->qpage + q->idx);
+
+	/* Check valid bit (31) vs current toggle polarity */
+	if ((cur >> 31) == q->toggle)
+		return 0;
+
+	/* If consuming from the queue ... */
+	if (!just_peek) {
+		/* Next entry */
+		q->idx = (q->idx + 1) & q->msk;
+
+		/* Wrap around: flip valid toggle */
+		if (q->idx == 0)
+			q->toggle ^= 1;
+	}
+	/* Mask out the valid bit (31) */
+	return cur & 0x7fffffff;
+}
+
+/*
+ * Scans all the queue that may have interrupts in them
+ * (based on "pending_prio") in priority order until an
+ * interrupt is found or all the queues are empty.
+ *
+ * Then updates the CPPR (Current Processor Priority
+ * Register) based on the most favored interrupt found
+ * (0xff if none) and return what was found (0 if none).
+ *
+ * If just_peek is set, return the most favored pending
+ * interrupt if any but don't update the queue pointers.
+ *
+ * Note: This function can operate generically on any number
+ * of queues (up to 8). The current implementation of the XIVE
+ * driver only uses a single queue however.
+ *
+ * Note2: This will also "flush" "the pending_count" of a queue
+ * into the "count" when that queue is observed to be empty.
+ * This is used to keep track of the amount of interrupts
+ * targetting a queue. When an interrupt is moved away from
+ * a queue, we only decrement that queue count once the queue
+ * has been observed empty to avoid races.
+ */
+static u32 xive_scan_interrupts(struct xive_cpu *xc, bool just_peek)
+{
+	u32 irq = 0;
+	u8 prio = 0;
+
+	/* Find highest pending priority */
+	while (xc->pending_prio != 0) {
+		struct xive_q *q;
+
+		prio = ffs(xc->pending_prio) - 1;
+		DBG_VERBOSE("scan_irq: trying prio %d\n", prio);
+
+		/* Try to fetch */
+		irq = xive_read_eq(&xc->queue[prio], just_peek);
+
+		/* Found something ? That's it */
+		if (irq) {
+			if (just_peek || irq_to_desc(irq))
+				break;
+			/*
+			 * We should never get here; if we do then we must
+			 * have failed to synchronize the interrupt properly
+			 * when shutting it down.
+			 */
+			pr_crit("xive: got interrupt %d without descriptor, dropping\n",
+				irq);
+			WARN_ON(1);
+			continue;
+		}
+
+		/* Clear pending bits */
+		xc->pending_prio &= ~(1 << prio);
+
+		/*
+		 * Check if the queue count needs adjusting due to
+		 * interrupts being moved away. See description of
+		 * xive_dec_target_count()
+		 */
+		q = &xc->queue[prio];
+		if (atomic_read(&q->pending_count)) {
+			int p = atomic_xchg(&q->pending_count, 0);
+			if (p) {
+				WARN_ON(p > atomic_read(&q->count));
+				atomic_sub(p, &q->count);
+			}
+		}
+	}
+
+	/* If nothing was found, set CPPR to 0xff */
+	if (irq == 0)
+		prio = 0xff;
+
+	/* Update HW CPPR to match if necessary */
+	if (prio != xc->cppr) {
+		DBG_VERBOSE("scan_irq: adjusting CPPR to %d\n", prio);
+		xc->cppr = prio;
+		out_8(xive_tima + xive_tima_offset + TM_CPPR, prio);
+	}
+
+	return irq;
+}
+
+/*
+ * This is used to perform the magic loads from an ESB
+ * described in xive-regs.h
+ */
+static notrace u8 xive_esb_read(struct xive_irq_data *xd, u32 offset)
+{
+	u64 val;
+
+	if (offset == XIVE_ESB_SET_PQ_10 && xive_is_store_eoi(xd))
+		offset |= XIVE_ESB_LD_ST_MO;
+
+	if ((xd->flags & XIVE_IRQ_FLAG_H_INT_ESB) && xive_ops->esb_rw)
+		val = xive_ops->esb_rw(xd->hw_irq, offset, 0, 0);
+	else
+		val = in_be64(xd->eoi_mmio + offset);
+
+	return (u8)val;
+}
+
+static void xive_esb_write(struct xive_irq_data *xd, u32 offset, u64 data)
+{
+	if ((xd->flags & XIVE_IRQ_FLAG_H_INT_ESB) && xive_ops->esb_rw)
+		xive_ops->esb_rw(xd->hw_irq, offset, data, 1);
+	else
+		out_be64(xd->eoi_mmio + offset, data);
+}
+
+#if defined(CONFIG_XMON) || defined(CONFIG_DEBUG_FS)
+static void xive_irq_data_dump(struct xive_irq_data *xd, char *buffer, size_t size)
+{
+	u64 val = xive_esb_read(xd, XIVE_ESB_GET);
+
+	snprintf(buffer, size, "flags=%c%c%c PQ=%c%c 0x%016llx 0x%016llx",
+		 xive_is_store_eoi(xd) ? 'S' : ' ',
+		 xd->flags & XIVE_IRQ_FLAG_LSI ? 'L' : ' ',
+		 xd->flags & XIVE_IRQ_FLAG_H_INT_ESB ? 'H' : ' ',
+		 val & XIVE_ESB_VAL_P ? 'P' : '-',
+		 val & XIVE_ESB_VAL_Q ? 'Q' : '-',
+		 xd->trig_page, xd->eoi_page);
+}
+#endif
+
+#ifdef CONFIG_XMON
+static notrace void xive_dump_eq(const char *name, struct xive_q *q)
+{
+	u32 i0, i1, idx;
+
+	if (!q->qpage)
+		return;
+	idx = q->idx;
+	i0 = be32_to_cpup(q->qpage + idx);
+	idx = (idx + 1) & q->msk;
+	i1 = be32_to_cpup(q->qpage + idx);
+	xmon_printf("%s idx=%d T=%d %08x %08x ...", name,
+		     q->idx, q->toggle, i0, i1);
+}
+
+notrace void xmon_xive_do_dump(int cpu)
+{
+	struct xive_cpu *xc = per_cpu(xive_cpu, cpu);
+
+	xmon_printf("CPU %d:", cpu);
+	if (xc) {
+		xmon_printf("pp=%02x CPPR=%02x ", xc->pending_prio, xc->cppr);
+
+#ifdef CONFIG_SMP
+		{
+			char buffer[128];
+
+			xive_irq_data_dump(&xc->ipi_data, buffer, sizeof(buffer));
+			xmon_printf("IPI=0x%08x %s", xc->hw_ipi, buffer);
+		}
+#endif
+		xive_dump_eq("EQ", &xc->queue[xive_irq_priority]);
+	}
+	xmon_printf("\n");
+}
+
+static struct irq_data *xive_get_irq_data(u32 hw_irq)
+{
+	unsigned int irq = irq_find_mapping(xive_irq_domain, hw_irq);
+
+	return irq ? irq_get_irq_data(irq) : NULL;
+}
+
+int xmon_xive_get_irq_config(u32 hw_irq, struct irq_data *d)
+{
+	int rc;
+	u32 target;
+	u8 prio;
+	u32 lirq;
+
+	rc = xive_ops->get_irq_config(hw_irq, &target, &prio, &lirq);
+	if (rc) {
+		xmon_printf("IRQ 0x%08x : no config rc=%d\n", hw_irq, rc);
+		return rc;
+	}
+
+	xmon_printf("IRQ 0x%08x : target=0x%x prio=%02x lirq=0x%x ",
+		    hw_irq, target, prio, lirq);
+
+	if (!d)
+		d = xive_get_irq_data(hw_irq);
+
+	if (d) {
+		char buffer[128];
+
+		xive_irq_data_dump(irq_data_get_irq_chip_data(d),
+				   buffer, sizeof(buffer));
+		xmon_printf("%s", buffer);
+	}
+
+	xmon_printf("\n");
+	return 0;
+}
+
+void xmon_xive_get_irq_all(void)
+{
+	unsigned int i;
+	struct irq_desc *desc;
+
+	for_each_irq_desc(i, desc) {
+		struct irq_data *d = irq_domain_get_irq_data(xive_irq_domain, i);
+
+		if (d)
+			xmon_xive_get_irq_config(irqd_to_hwirq(d), d);
+	}
+}
+
+#endif /* CONFIG_XMON */
+
+static unsigned int xive_get_irq(void)
+{
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+	u32 irq;
+
+	/*
+	 * This can be called either as a result of a HW interrupt or
+	 * as a "replay" because EOI decided there was still something
+	 * in one of the queues.
+	 *
+	 * First we perform an ACK cycle in order to update our mask
+	 * of pending priorities. This will also have the effect of
+	 * updating the CPPR to the most favored pending interrupts.
+	 *
+	 * In the future, if we have a way to differentiate a first
+	 * entry (on HW interrupt) from a replay triggered by EOI,
+	 * we could skip this on replays unless we soft-mask tells us
+	 * that a new HW interrupt occurred.
+	 */
+	xive_ops->update_pending(xc);
+
+	DBG_VERBOSE("get_irq: pending=%02x\n", xc->pending_prio);
+
+	/* Scan our queue(s) for interrupts */
+	irq = xive_scan_interrupts(xc, false);
+
+	DBG_VERBOSE("get_irq: got irq 0x%x, new pending=0x%02x\n",
+	    irq, xc->pending_prio);
+
+	/* Return pending interrupt if any */
+	if (irq == XIVE_BAD_IRQ)
+		return 0;
+	return irq;
+}
+
+/*
+ * After EOI'ing an interrupt, we need to re-check the queue
+ * to see if another interrupt is pending since multiple
+ * interrupts can coalesce into a single notification to the
+ * CPU.
+ *
+ * If we find that there is indeed more in there, we call
+ * force_external_irq_replay() to make Linux synthesize an
+ * external interrupt on the next call to local_irq_restore().
+ */
+static void xive_do_queue_eoi(struct xive_cpu *xc)
+{
+	if (xive_scan_interrupts(xc, true) != 0) {
+		DBG_VERBOSE("eoi: pending=0x%02x\n", xc->pending_prio);
+		force_external_irq_replay();
+	}
+}
+
+/*
+ * EOI an interrupt at the source. There are several methods
+ * to do this depending on the HW version and source type
+ */
+static void xive_do_source_eoi(struct xive_irq_data *xd)
+{
+	u8 eoi_val;
+
+	xd->stale_p = false;
+
+	/* If the XIVE supports the new "store EOI facility, use it */
+	if (xive_is_store_eoi(xd)) {
+		xive_esb_write(xd, XIVE_ESB_STORE_EOI, 0);
+		return;
+	}
+
+	/*
+	 * For LSIs, we use the "EOI cycle" special load rather than
+	 * PQ bits, as they are automatically re-triggered in HW when
+	 * still pending.
+	 */
+	if (xd->flags & XIVE_IRQ_FLAG_LSI) {
+		xive_esb_read(xd, XIVE_ESB_LOAD_EOI);
+		return;
+	}
+
+	/*
+	 * Otherwise, we use the special MMIO that does a clear of
+	 * both P and Q and returns the old Q. This allows us to then
+	 * do a re-trigger if Q was set rather than synthesizing an
+	 * interrupt in software
+	 */
+	eoi_val = xive_esb_read(xd, XIVE_ESB_SET_PQ_00);
+	DBG_VERBOSE("eoi_val=%x\n", eoi_val);
+
+	/* Re-trigger if needed */
+	if ((eoi_val & XIVE_ESB_VAL_Q) && xd->trig_mmio)
+		out_be64(xd->trig_mmio, 0);
+}
+
+/* irq_chip eoi callback, called with irq descriptor lock held */
+static void xive_irq_eoi(struct irq_data *d)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+
+	DBG_VERBOSE("eoi_irq: irq=%d [0x%lx] pending=%02x\n",
+		    d->irq, irqd_to_hwirq(d), xc->pending_prio);
+
+	/*
+	 * EOI the source if it hasn't been disabled and hasn't
+	 * been passed-through to a KVM guest
+	 */
+	if (!irqd_irq_disabled(d) && !irqd_is_forwarded_to_vcpu(d) &&
+	    !(xd->flags & XIVE_IRQ_FLAG_NO_EOI))
+		xive_do_source_eoi(xd);
+	else
+		xd->stale_p = true;
+
+	/*
+	 * Clear saved_p to indicate that it's no longer occupying
+	 * a queue slot on the target queue
+	 */
+	xd->saved_p = false;
+
+	/* Check for more work in the queue */
+	xive_do_queue_eoi(xc);
+}
+
+/*
+ * Helper used to mask and unmask an interrupt source.
+ */
+static void xive_do_source_set_mask(struct xive_irq_data *xd,
+				    bool mask)
+{
+	u64 val;
+
+	pr_debug("%s: HW 0x%x %smask\n", __func__, xd->hw_irq, mask ? "" : "un");
+
+	/*
+	 * If the interrupt had P set, it may be in a queue.
+	 *
+	 * We need to make sure we don't re-enable it until it
+	 * has been fetched from that queue and EOId. We keep
+	 * a copy of that P state and use it to restore the
+	 * ESB accordingly on unmask.
+	 */
+	if (mask) {
+		val = xive_esb_read(xd, XIVE_ESB_SET_PQ_01);
+		if (!xd->stale_p && !!(val & XIVE_ESB_VAL_P))
+			xd->saved_p = true;
+		xd->stale_p = false;
+	} else if (xd->saved_p) {
+		xive_esb_read(xd, XIVE_ESB_SET_PQ_10);
+		xd->saved_p = false;
+	} else {
+		xive_esb_read(xd, XIVE_ESB_SET_PQ_00);
+		xd->stale_p = false;
+	}
+}
+
+/*
+ * Try to chose "cpu" as a new interrupt target. Increments
+ * the queue accounting for that target if it's not already
+ * full.
+ */
+static bool xive_try_pick_target(int cpu)
+{
+	struct xive_cpu *xc = per_cpu(xive_cpu, cpu);
+	struct xive_q *q = &xc->queue[xive_irq_priority];
+	int max;
+
+	/*
+	 * Calculate max number of interrupts in that queue.
+	 *
+	 * We leave a gap of 1 just in case...
+	 */
+	max = (q->msk + 1) - 1;
+	return !!atomic_add_unless(&q->count, 1, max);
+}
+
+/*
+ * Un-account an interrupt for a target CPU. We don't directly
+ * decrement q->count since the interrupt might still be present
+ * in the queue.
+ *
+ * Instead increment a separate counter "pending_count" which
+ * will be substracted from "count" later when that CPU observes
+ * the queue to be empty.
+ */
+static void xive_dec_target_count(int cpu)
+{
+	struct xive_cpu *xc = per_cpu(xive_cpu, cpu);
+	struct xive_q *q = &xc->queue[xive_irq_priority];
+
+	if (WARN_ON(cpu < 0 || !xc)) {
+		pr_err("%s: cpu=%d xc=%p\n", __func__, cpu, xc);
+		return;
+	}
+
+	/*
+	 * We increment the "pending count" which will be used
+	 * to decrement the target queue count whenever it's next
+	 * processed and found empty. This ensure that we don't
+	 * decrement while we still have the interrupt there
+	 * occupying a slot.
+	 */
+	atomic_inc(&q->pending_count);
+}
+
+/* Find a tentative CPU target in a CPU mask */
+static int xive_find_target_in_mask(const struct cpumask *mask,
+				    unsigned int fuzz)
+{
+	int cpu, first;
+
+	/* Pick up a starting point CPU in the mask based on  fuzz */
+	fuzz %= cpumask_weight(mask);
+	first = cpumask_nth(fuzz, mask);
+	WARN_ON(first >= nr_cpu_ids);
+
+	/*
+	 * Now go through the entire mask until we find a valid
+	 * target.
+	 */
+	for_each_cpu_wrap(cpu, mask, first) {
+		if (cpu_online(cpu) && xive_try_pick_target(cpu))
+			return cpu;
+	}
+
+	WARN_ONCE(1, "target CPU not found in mask: %*pbl\n", cpumask_pr_args(mask));
+	return -1;
+}
+
+/*
+ * Pick a target CPU for an interrupt. This is done at
+ * startup or if the affinity is changed in a way that
+ * invalidates the current target.
+ */
+static int xive_pick_irq_target(struct irq_data *d,
+				const struct cpumask *affinity)
+{
+	static unsigned int fuzz;
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+	cpumask_var_t mask;
+	int cpu = -1;
+
+	/*
+	 * If we have chip IDs, first we try to build a mask of
+	 * CPUs matching the CPU and find a target in there
+	 */
+	if (xd->src_chip != XIVE_INVALID_CHIP_ID &&
+		zalloc_cpumask_var(&mask, GFP_ATOMIC)) {
+		/* Build a mask of matching chip IDs */
+		for_each_cpu_and(cpu, affinity, cpu_online_mask) {
+			struct xive_cpu *xc = per_cpu(xive_cpu, cpu);
+			if (xc->chip_id == xd->src_chip)
+				cpumask_set_cpu(cpu, mask);
+		}
+		/* Try to find a target */
+		if (cpumask_empty(mask))
+			cpu = -1;
+		else
+			cpu = xive_find_target_in_mask(mask, fuzz++);
+		free_cpumask_var(mask);
+		if (cpu >= 0)
+			return cpu;
+		fuzz--;
+	}
+
+	/* No chip IDs, fallback to using the affinity mask */
+	return xive_find_target_in_mask(affinity, fuzz++);
+}
+
+static unsigned int xive_irq_startup(struct irq_data *d)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+	unsigned int hw_irq = (unsigned int)irqd_to_hwirq(d);
+	int target, rc;
+
+	xd->saved_p = false;
+	xd->stale_p = false;
+
+	pr_debug("%s: irq %d [0x%x] data @%p\n", __func__, d->irq, hw_irq, d);
+
+	/* Pick a target */
+	target = xive_pick_irq_target(d, irq_data_get_affinity_mask(d));
+	if (target == XIVE_INVALID_TARGET) {
+		/* Try again breaking affinity */
+		target = xive_pick_irq_target(d, cpu_online_mask);
+		if (target == XIVE_INVALID_TARGET)
+			return -ENXIO;
+		pr_warn("irq %d started with broken affinity\n", d->irq);
+	}
+
+	/* Sanity check */
+	if (WARN_ON(target == XIVE_INVALID_TARGET ||
+		    target >= nr_cpu_ids))
+		target = smp_processor_id();
+
+	xd->target = target;
+
+	/*
+	 * Configure the logical number to be the Linux IRQ number
+	 * and set the target queue
+	 */
+	rc = xive_ops->configure_irq(hw_irq,
+				     get_hard_smp_processor_id(target),
+				     xive_irq_priority, d->irq);
+	if (rc)
+		return rc;
+
+	/* Unmask the ESB */
+	xive_do_source_set_mask(xd, false);
+
+	return 0;
+}
+
+/* called with irq descriptor lock held */
+static void xive_irq_shutdown(struct irq_data *d)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+	unsigned int hw_irq = (unsigned int)irqd_to_hwirq(d);
+
+	pr_debug("%s: irq %d [0x%x] data @%p\n", __func__, d->irq, hw_irq, d);
+
+	if (WARN_ON(xd->target == XIVE_INVALID_TARGET))
+		return;
+
+	/* Mask the interrupt at the source */
+	xive_do_source_set_mask(xd, true);
+
+	/*
+	 * Mask the interrupt in HW in the IVT/EAS and set the number
+	 * to be the "bad" IRQ number
+	 */
+	xive_ops->configure_irq(hw_irq,
+				get_hard_smp_processor_id(xd->target),
+				0xff, XIVE_BAD_IRQ);
+
+	xive_dec_target_count(xd->target);
+	xd->target = XIVE_INVALID_TARGET;
+}
+
+static void xive_irq_unmask(struct irq_data *d)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+
+	pr_debug("%s: irq %d data @%p\n", __func__, d->irq, xd);
+
+	xive_do_source_set_mask(xd, false);
+}
+
+static void xive_irq_mask(struct irq_data *d)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+
+	pr_debug("%s: irq %d data @%p\n", __func__, d->irq, xd);
+
+	xive_do_source_set_mask(xd, true);
+}
+
+static int xive_irq_set_affinity(struct irq_data *d,
+				 const struct cpumask *cpumask,
+				 bool force)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+	unsigned int hw_irq = (unsigned int)irqd_to_hwirq(d);
+	u32 target, old_target;
+	int rc = 0;
+
+	pr_debug("%s: irq %d/0x%x\n", __func__, d->irq, hw_irq);
+
+	/* Is this valid ? */
+	if (!cpumask_intersects(cpumask, cpu_online_mask))
+		return -EINVAL;
+
+	/*
+	 * If existing target is already in the new mask, and is
+	 * online then do nothing.
+	 */
+	if (xd->target != XIVE_INVALID_TARGET &&
+	    cpu_online(xd->target) &&
+	    cpumask_test_cpu(xd->target, cpumask))
+		return IRQ_SET_MASK_OK;
+
+	/* Pick a new target */
+	target = xive_pick_irq_target(d, cpumask);
+
+	/* No target found */
+	if (target == XIVE_INVALID_TARGET)
+		return -ENXIO;
+
+	/* Sanity check */
+	if (WARN_ON(target >= nr_cpu_ids))
+		target = smp_processor_id();
+
+	old_target = xd->target;
+
+	/*
+	 * Only configure the irq if it's not currently passed-through to
+	 * a KVM guest
+	 */
+	if (!irqd_is_forwarded_to_vcpu(d))
+		rc = xive_ops->configure_irq(hw_irq,
+					     get_hard_smp_processor_id(target),
+					     xive_irq_priority, d->irq);
+	if (rc < 0) {
+		pr_err("Error %d reconfiguring irq %d\n", rc, d->irq);
+		return rc;
+	}
+
+	pr_debug("  target: 0x%x\n", target);
+	xd->target = target;
+
+	/* Give up previous target */
+	if (old_target != XIVE_INVALID_TARGET)
+	    xive_dec_target_count(old_target);
+
+	return IRQ_SET_MASK_OK;
+}
+
+static int xive_irq_set_type(struct irq_data *d, unsigned int flow_type)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+
+	/*
+	 * We only support these. This has really no effect other than setting
+	 * the corresponding descriptor bits mind you but those will in turn
+	 * affect the resend function when re-enabling an edge interrupt.
+	 *
+	 * Set the default to edge as explained in map().
+	 */
+	if (flow_type == IRQ_TYPE_DEFAULT || flow_type == IRQ_TYPE_NONE)
+		flow_type = IRQ_TYPE_EDGE_RISING;
+
+	if (flow_type != IRQ_TYPE_EDGE_RISING &&
+	    flow_type != IRQ_TYPE_LEVEL_LOW)
+		return -EINVAL;
+
+	irqd_set_trigger_type(d, flow_type);
+
+	/*
+	 * Double check it matches what the FW thinks
+	 *
+	 * NOTE: We don't know yet if the PAPR interface will provide
+	 * the LSI vs MSI information apart from the device-tree so
+	 * this check might have to move into an optional backend call
+	 * that is specific to the native backend
+	 */
+	if ((flow_type == IRQ_TYPE_LEVEL_LOW) !=
+	    !!(xd->flags & XIVE_IRQ_FLAG_LSI)) {
+		pr_warn("Interrupt %d (HW 0x%x) type mismatch, Linux says %s, FW says %s\n",
+			d->irq, (u32)irqd_to_hwirq(d),
+			(flow_type == IRQ_TYPE_LEVEL_LOW) ? "Level" : "Edge",
+			(xd->flags & XIVE_IRQ_FLAG_LSI) ? "Level" : "Edge");
+	}
+
+	return IRQ_SET_MASK_OK_NOCOPY;
+}
+
+static int xive_irq_retrigger(struct irq_data *d)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+
+	/* This should be only for MSIs */
+	if (WARN_ON(xd->flags & XIVE_IRQ_FLAG_LSI))
+		return 0;
+
+	/*
+	 * To perform a retrigger, we first set the PQ bits to
+	 * 11, then perform an EOI.
+	 */
+	xive_esb_read(xd, XIVE_ESB_SET_PQ_11);
+	xive_do_source_eoi(xd);
+
+	return 1;
+}
+
+/*
+ * Caller holds the irq descriptor lock, so this won't be called
+ * concurrently with xive_get_irqchip_state on the same interrupt.
+ */
+static int xive_irq_set_vcpu_affinity(struct irq_data *d, void *state)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(d);
+	unsigned int hw_irq = (unsigned int)irqd_to_hwirq(d);
+	int rc;
+	u8 pq;
+
+	/*
+	 * This is called by KVM with state non-NULL for enabling
+	 * pass-through or NULL for disabling it
+	 */
+	if (state) {
+		irqd_set_forwarded_to_vcpu(d);
+
+		/* Set it to PQ=10 state to prevent further sends */
+		pq = xive_esb_read(xd, XIVE_ESB_SET_PQ_10);
+		if (!xd->stale_p) {
+			xd->saved_p = !!(pq & XIVE_ESB_VAL_P);
+			xd->stale_p = !xd->saved_p;
+		}
+
+		/* No target ? nothing to do */
+		if (xd->target == XIVE_INVALID_TARGET) {
+			/*
+			 * An untargetted interrupt should have been
+			 * also masked at the source
+			 */
+			WARN_ON(xd->saved_p);
+
+			return 0;
+		}
+
+		/*
+		 * If P was set, adjust state to PQ=11 to indicate
+		 * that a resend is needed for the interrupt to reach
+		 * the guest. Also remember the value of P.
+		 *
+		 * This also tells us that it's in flight to a host queue
+		 * or has already been fetched but hasn't been EOIed yet
+		 * by the host. Thus it's potentially using up a host
+		 * queue slot. This is important to know because as long
+		 * as this is the case, we must not hard-unmask it when
+		 * "returning" that interrupt to the host.
+		 *
+		 * This saved_p is cleared by the host EOI, when we know
+		 * for sure the queue slot is no longer in use.
+		 */
+		if (xd->saved_p) {
+			xive_esb_read(xd, XIVE_ESB_SET_PQ_11);
+
+			/*
+			 * Sync the XIVE source HW to ensure the interrupt
+			 * has gone through the EAS before we change its
+			 * target to the guest. That should guarantee us
+			 * that we *will* eventually get an EOI for it on
+			 * the host. Otherwise there would be a small window
+			 * for P to be seen here but the interrupt going
+			 * to the guest queue.
+			 */
+			if (xive_ops->sync_source)
+				xive_ops->sync_source(hw_irq);
+		}
+	} else {
+		irqd_clr_forwarded_to_vcpu(d);
+
+		/* No host target ? hard mask and return */
+		if (xd->target == XIVE_INVALID_TARGET) {
+			xive_do_source_set_mask(xd, true);
+			return 0;
+		}
+
+		/*
+		 * Sync the XIVE source HW to ensure the interrupt
+		 * has gone through the EAS before we change its
+		 * target to the host.
+		 */
+		if (xive_ops->sync_source)
+			xive_ops->sync_source(hw_irq);
+
+		/*
+		 * By convention we are called with the interrupt in
+		 * a PQ=10 or PQ=11 state, ie, it won't fire and will
+		 * have latched in Q whether there's a pending HW
+		 * interrupt or not.
+		 *
+		 * First reconfigure the target.
+		 */
+		rc = xive_ops->configure_irq(hw_irq,
+					     get_hard_smp_processor_id(xd->target),
+					     xive_irq_priority, d->irq);
+		if (rc)
+			return rc;
+
+		/*
+		 * Then if saved_p is not set, effectively re-enable the
+		 * interrupt with an EOI. If it is set, we know there is
+		 * still a message in a host queue somewhere that will be
+		 * EOId eventually.
+		 *
+		 * Note: We don't check irqd_irq_disabled(). Effectively,
+		 * we *will* let the irq get through even if masked if the
+		 * HW is still firing it in order to deal with the whole
+		 * saved_p business properly. If the interrupt triggers
+		 * while masked, the generic code will re-mask it anyway.
+		 */
+		if (!xd->saved_p)
+			xive_do_source_eoi(xd);
+
+	}
+	return 0;
+}
+
+/* Called with irq descriptor lock held. */
+static int xive_get_irqchip_state(struct irq_data *data,
+				  enum irqchip_irq_state which, bool *state)
+{
+	struct xive_irq_data *xd = irq_data_get_irq_chip_data(data);
+	u8 pq;
+
+	switch (which) {
+	case IRQCHIP_STATE_ACTIVE:
+		pq = xive_esb_read(xd, XIVE_ESB_GET);
+
+		/*
+		 * The esb value being all 1's means we couldn't get
+		 * the PQ state of the interrupt through mmio. It may
+		 * happen, for example when querying a PHB interrupt
+		 * while the PHB is in an error state. We consider the
+		 * interrupt to be inactive in that case.
+		 */
+		*state = (pq != XIVE_ESB_INVALID) && !xd->stale_p &&
+			(xd->saved_p || (!!(pq & XIVE_ESB_VAL_P) &&
+			 !irqd_irq_disabled(data)));
+		return 0;
+	default:
+		return -EINVAL;
+	}
+}
+
+static struct irq_chip xive_irq_chip = {
+	.name = "XIVE-IRQ",
+	.irq_startup = xive_irq_startup,
+	.irq_shutdown = xive_irq_shutdown,
+	.irq_eoi = xive_irq_eoi,
+	.irq_mask = xive_irq_mask,
+	.irq_unmask = xive_irq_unmask,
+	.irq_set_affinity = xive_irq_set_affinity,
+	.irq_set_type = xive_irq_set_type,
+	.irq_retrigger = xive_irq_retrigger,
+	.irq_set_vcpu_affinity = xive_irq_set_vcpu_affinity,
+	.irq_get_irqchip_state = xive_get_irqchip_state,
+};
+
+bool is_xive_irq(struct irq_chip *chip)
+{
+	return chip == &xive_irq_chip;
+}
+EXPORT_SYMBOL_GPL(is_xive_irq);
+
+void xive_cleanup_irq_data(struct xive_irq_data *xd)
+{
+	pr_debug("%s for HW 0x%x\n", __func__, xd->hw_irq);
+
+	if (xd->eoi_mmio) {
+		iounmap(xd->eoi_mmio);
+		if (xd->eoi_mmio == xd->trig_mmio)
+			xd->trig_mmio = NULL;
+		xd->eoi_mmio = NULL;
+	}
+	if (xd->trig_mmio) {
+		iounmap(xd->trig_mmio);
+		xd->trig_mmio = NULL;
+	}
+}
+EXPORT_SYMBOL_GPL(xive_cleanup_irq_data);
+
+static struct xive_irq_data *xive_irq_alloc_data(unsigned int virq, irq_hw_number_t hw)
+{
+	struct xive_irq_data *xd;
+	int rc;
+
+	xd = kzalloc_obj(struct xive_irq_data);
+	if (!xd)
+		return ERR_PTR(-ENOMEM);
+	rc = xive_ops->populate_irq_data(hw, xd);
+	if (rc) {
+		kfree(xd);
+		return ERR_PTR(rc);
+	}
+	xd->target = XIVE_INVALID_TARGET;
+
+	/*
+	 * Turn OFF by default the interrupt being mapped. A side
+	 * effect of this check is the mapping the ESB page of the
+	 * interrupt in the Linux address space. This prevents page
+	 * fault issues in the crash handler which masks all
+	 * interrupts.
+	 */
+	xive_esb_read(xd, XIVE_ESB_SET_PQ_01);
+
+	return xd;
+}
+
+static void xive_irq_free_data(struct irq_domain *domain, unsigned int virq)
+{
+	struct xive_irq_data *xd;
+	struct irq_data *data = irq_domain_get_irq_data(domain, virq);
+
+	if (!data)
+		return;
+
+	xd = irq_data_get_irq_chip_data(data);
+	if (!xd)
+		return;
+
+	irq_domain_reset_irq_data(data);
+	xive_cleanup_irq_data(xd);
+	kfree(xd);
+}
+
+#ifdef CONFIG_SMP
+
+static void xive_cause_ipi(int cpu)
+{
+	struct xive_cpu *xc;
+	struct xive_irq_data *xd;
+
+	xc = per_cpu(xive_cpu, cpu);
+
+	DBG_VERBOSE("IPI CPU %d -> %d (HW IRQ 0x%x)\n",
+		    smp_processor_id(), cpu, xc->hw_ipi);
+
+	xd = &xc->ipi_data;
+	if (WARN_ON(!xd->trig_mmio))
+		return;
+	out_be64(xd->trig_mmio, 0);
+}
+
+static irqreturn_t xive_muxed_ipi_action(int irq, void *dev_id)
+{
+	return smp_ipi_demux();
+}
+
+static void xive_ipi_eoi(struct irq_data *d)
+{
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+
+	/* Handle possible race with unplug and drop stale IPIs */
+	if (!xc)
+		return;
+
+	DBG_VERBOSE("IPI eoi: irq=%d [0x%lx] (HW IRQ 0x%x) pending=%02x\n",
+		    d->irq, irqd_to_hwirq(d), xc->hw_ipi, xc->pending_prio);
+
+	xive_do_source_eoi(&xc->ipi_data);
+	xive_do_queue_eoi(xc);
+}
+
+static void xive_ipi_do_nothing(struct irq_data *d)
+{
+	/*
+	 * Nothing to do, we never mask/unmask IPIs, but the callback
+	 * has to exist for the struct irq_chip.
+	 */
+}
+
+static struct irq_chip xive_ipi_chip = {
+	.name = "XIVE-IPI",
+	.irq_eoi = xive_ipi_eoi,
+	.irq_mask = xive_ipi_do_nothing,
+	.irq_unmask = xive_ipi_do_nothing,
+};
+
+/*
+ * IPIs are marked per-cpu. We use separate HW interrupts under the
+ * hood but associated with the same "linux" interrupt
+ */
+struct xive_ipi_alloc_info {
+	irq_hw_number_t hwirq;
+};
+
+static int xive_ipi_irq_domain_alloc(struct irq_domain *domain, unsigned int virq,
+				     unsigned int nr_irqs, void *arg)
+{
+	struct xive_ipi_alloc_info *info = arg;
+	int i;
+
+	for (i = 0; i < nr_irqs; i++) {
+		irq_domain_set_info(domain, virq + i, info->hwirq + i, &xive_ipi_chip,
+				    domain->host_data, handle_percpu_irq,
+				    NULL, NULL);
+	}
+	return 0;
+}
+
+static const struct irq_domain_ops xive_ipi_irq_domain_ops = {
+	.alloc  = xive_ipi_irq_domain_alloc,
+};
+
+static int __init xive_init_ipis(void)
+{
+	struct fwnode_handle *fwnode;
+	struct irq_domain *ipi_domain;
+	unsigned int node;
+	int ret = -ENOMEM;
+
+	fwnode = irq_domain_alloc_named_fwnode("XIVE-IPI");
+	if (!fwnode)
+		goto out;
+
+	ipi_domain = irq_domain_create_linear(fwnode, nr_node_ids,
+					      &xive_ipi_irq_domain_ops, NULL);
+	if (!ipi_domain)
+		goto out_free_fwnode;
+
+	xive_ipis = kzalloc_objs(*xive_ipis, nr_node_ids);
+	if (!xive_ipis)
+		goto out_free_domain;
+
+	for_each_node(node) {
+		struct xive_ipi_desc *xid = &xive_ipis[node];
+		struct xive_ipi_alloc_info info = { node };
+
+		/*
+		 * Map one IPI interrupt per node for all cpus of that node.
+		 * Since the HW interrupt number doesn't have any meaning,
+		 * simply use the node number.
+		 */
+		ret = irq_domain_alloc_irqs(ipi_domain, 1, node, &info);
+		if (ret < 0)
+			goto out_free_xive_ipis;
+		xid->irq = ret;
+
+		snprintf(xid->name, sizeof(xid->name), "IPI-%d", node);
+	}
+
+	return ret;
+
+out_free_xive_ipis:
+	kfree(xive_ipis);
+	xive_ipis = NULL;
+out_free_domain:
+	irq_domain_remove(ipi_domain);
+out_free_fwnode:
+	irq_domain_free_fwnode(fwnode);
+out:
+	return ret;
+}
+
+static int xive_request_ipi(unsigned int cpu)
+{
+	struct xive_ipi_desc *xid = &xive_ipis[early_cpu_to_node(cpu)];
+	int ret;
+
+	if (atomic_inc_return(&xid->started) > 1)
+		return 0;
+
+	ret = request_irq(xid->irq, xive_muxed_ipi_action,
+			  IRQF_NO_DEBUG | IRQF_PERCPU | IRQF_NO_THREAD,
+			  xid->name, NULL);
+
+	WARN(ret < 0, "Failed to request IPI %d: %d\n", xid->irq, ret);
+	return ret;
+}
+
+static int xive_setup_cpu_ipi(unsigned int cpu)
+{
+	unsigned int xive_ipi_irq = xive_ipi_cpu_to_irq(cpu);
+	struct xive_cpu *xc;
+	int rc;
+
+	pr_debug("Setting up IPI for CPU %d\n", cpu);
+
+	if (xive_ipi_irq == XIVE_BAD_IRQ)
+		return -EIO;
+
+	xc = per_cpu(xive_cpu, cpu);
+
+	/* Check if we are already setup */
+	if (xc->hw_ipi != XIVE_BAD_IRQ)
+		return 0;
+
+	/* Register the IPI */
+	xive_request_ipi(cpu);
+
+	/* Grab an IPI from the backend, this will populate xc->hw_ipi */
+	if (xive_ops->get_ipi(cpu, xc))
+		return -EIO;
+
+	/*
+	 * Populate the IRQ data in the xive_cpu structure and
+	 * configure the HW / enable the IPIs.
+	 */
+	rc = xive_ops->populate_irq_data(xc->hw_ipi, &xc->ipi_data);
+	if (rc) {
+		pr_err("Failed to populate IPI data on CPU %d\n", cpu);
+		return -EIO;
+	}
+	rc = xive_ops->configure_irq(xc->hw_ipi,
+				     get_hard_smp_processor_id(cpu),
+				     xive_irq_priority, xive_ipi_irq);
+	if (rc) {
+		pr_err("Failed to map IPI CPU %d\n", cpu);
+		return -EIO;
+	}
+	pr_debug("CPU %d HW IPI 0x%x, virq %d, trig_mmio=%p\n", cpu,
+		 xc->hw_ipi, xive_ipi_irq, xc->ipi_data.trig_mmio);
+
+	/* Unmask it */
+	xive_do_source_set_mask(&xc->ipi_data, false);
+
+	return 0;
+}
+
+noinstr static void xive_cleanup_cpu_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	unsigned int xive_ipi_irq = xive_ipi_cpu_to_irq(cpu);
+
+	/* Disable the IPI and free the IRQ data */
+
+	if (xive_ipi_irq == XIVE_BAD_IRQ)
+		return;
+
+	/* Already cleaned up ? */
+	if (xc->hw_ipi == XIVE_BAD_IRQ)
+		return;
+
+	/* TODO: clear IPI mapping */
+
+	/* Mask the IPI */
+	xive_do_source_set_mask(&xc->ipi_data, true);
+
+	/*
+	 * Note: We don't call xive_cleanup_irq_data() to free
+	 * the mappings as this is called from an IPI on kexec
+	 * which is not a safe environment to call iounmap()
+	 */
+
+	/* Deconfigure/mask in the backend */
+	xive_ops->configure_irq(xc->hw_ipi, hard_smp_processor_id(),
+				0xff, xive_ipi_irq);
+
+	/* Free the IPIs in the backend */
+	xive_ops->put_ipi(cpu, xc);
+}
+
+int __init xive_smp_probe(void)
+{
+	int ret;
+
+	/* Register the IPI */
+	ret = xive_init_ipis();
+	if (ret < 0)
+		return ret;
+
+	/* Allocate and setup IPI for the boot CPU */
+	ret = xive_setup_cpu_ipi(smp_processor_id());
+	if (ret < 0)
+		return ret;
+
+	smp_ops->cause_ipi = xive_cause_ipi;
+
+	return 0;
+}
+
+#endif /* CONFIG_SMP */
+
+static int xive_irq_domain_map(struct irq_domain *h, unsigned int virq,
+			       irq_hw_number_t hw)
+{
+	struct xive_irq_data *xd;
+
+	/*
+	 * Mark interrupts as edge sensitive by default so that resend
+	 * actually works. Will fix that up below if needed.
+	 */
+	irq_clear_status_flags(virq, IRQ_LEVEL);
+
+	xd = xive_irq_alloc_data(virq, hw);
+	if (IS_ERR(xd))
+		return PTR_ERR(xd);
+
+	irq_set_chip_and_handler(virq, &xive_irq_chip, handle_fasteoi_irq);
+	irq_set_chip_data(virq, xd);
+
+	return 0;
+}
+
+static void xive_irq_domain_unmap(struct irq_domain *d, unsigned int virq)
+{
+	xive_irq_free_data(d, virq);
+}
+
+static int xive_irq_domain_xlate(struct irq_domain *h, struct device_node *ct,
+				 const u32 *intspec, unsigned int intsize,
+				 irq_hw_number_t *out_hwirq, unsigned int *out_flags)
+
+{
+	*out_hwirq = intspec[0];
+
+	/*
+	 * If intsize is at least 2, we look for the type in the second cell,
+	 * we assume the LSB indicates a level interrupt.
+	 */
+	if (intsize > 1) {
+		if (intspec[1] & 1)
+			*out_flags = IRQ_TYPE_LEVEL_LOW;
+		else
+			*out_flags = IRQ_TYPE_EDGE_RISING;
+	} else
+		*out_flags = IRQ_TYPE_LEVEL_LOW;
+
+	return 0;
+}
+
+static int xive_irq_domain_match(struct irq_domain *h, struct device_node *node,
+				 enum irq_domain_bus_token bus_token)
+{
+	return xive_ops->match(node);
+}
+
+#ifdef CONFIG_GENERIC_IRQ_DEBUGFS
+static const char * const esb_names[] = { "RESET", "OFF", "PENDING", "QUEUED" };
+
+static const struct {
+	u64  mask;
+	char *name;
+} xive_irq_flags[] = {
+	{ XIVE_IRQ_FLAG_STORE_EOI, "STORE_EOI" },
+	{ XIVE_IRQ_FLAG_LSI,       "LSI"       },
+	{ XIVE_IRQ_FLAG_H_INT_ESB, "H_INT_ESB" },
+	{ XIVE_IRQ_FLAG_NO_EOI,    "NO_EOI"    },
+};
+
+static void xive_irq_domain_debug_show(struct seq_file *m, struct irq_domain *d,
+				       struct irq_data *irqd, int ind)
+{
+	struct xive_irq_data *xd;
+	u64 val;
+	int i;
+
+	/* No IRQ domain level information. To be done */
+	if (!irqd)
+		return;
+
+	if (!is_xive_irq(irq_data_get_irq_chip(irqd)))
+		return;
+
+	seq_printf(m, "%*sXIVE:\n", ind, "");
+	ind++;
+
+	xd = irq_data_get_irq_chip_data(irqd);
+	if (!xd) {
+		seq_printf(m, "%*snot assigned\n", ind, "");
+		return;
+	}
+
+	val = xive_esb_read(xd, XIVE_ESB_GET);
+	seq_printf(m, "%*sESB:      %s\n", ind, "", esb_names[val & 0x3]);
+	seq_printf(m, "%*sPstate:   %s %s\n", ind, "", xd->stale_p ? "stale" : "",
+		   xd->saved_p ? "saved" : "");
+	seq_printf(m, "%*sTarget:   %d\n", ind, "", xd->target);
+	seq_printf(m, "%*sChip:     %d\n", ind, "", xd->src_chip);
+	seq_printf(m, "%*sTrigger:  0x%016llx\n", ind, "", xd->trig_page);
+	seq_printf(m, "%*sEOI:      0x%016llx\n", ind, "", xd->eoi_page);
+	seq_printf(m, "%*sFlags:    0x%llx\n", ind, "", xd->flags);
+	for (i = 0; i < ARRAY_SIZE(xive_irq_flags); i++) {
+		if (xd->flags & xive_irq_flags[i].mask)
+			seq_printf(m, "%*s%s\n", ind + 12, "", xive_irq_flags[i].name);
+	}
+}
+#endif
+
+#ifdef	CONFIG_IRQ_DOMAIN_HIERARCHY
+static int xive_irq_domain_translate(struct irq_domain *d,
+				     struct irq_fwspec *fwspec,
+				     unsigned long *hwirq,
+				     unsigned int *type)
+{
+	return xive_irq_domain_xlate(d, to_of_node(fwspec->fwnode),
+				     fwspec->param, fwspec->param_count,
+				     hwirq, type);
+}
+
+static int xive_irq_domain_alloc(struct irq_domain *domain, unsigned int virq,
+				 unsigned int nr_irqs, void *arg)
+{
+	struct irq_fwspec *fwspec = arg;
+	struct xive_irq_data *xd;
+	irq_hw_number_t hwirq;
+	unsigned int type = IRQ_TYPE_NONE;
+	int i, rc;
+
+	rc = xive_irq_domain_translate(domain, fwspec, &hwirq, &type);
+	if (rc)
+		return rc;
+
+	pr_debug("%s %d/0x%lx #%d\n", __func__, virq, hwirq, nr_irqs);
+
+	for (i = 0; i < nr_irqs; i++) {
+		/* TODO: call xive_irq_domain_map() */
+
+		/*
+		 * Mark interrupts as edge sensitive by default so that resend
+		 * actually works. Will fix that up below if needed.
+		 */
+		irq_clear_status_flags(virq, IRQ_LEVEL);
+
+		/* allocates and sets handler data */
+		xd = xive_irq_alloc_data(virq + i, hwirq + i);
+		if (IS_ERR(xd))
+			return PTR_ERR(xd);
+
+		irq_domain_set_hwirq_and_chip(domain, virq + i, hwirq + i, &xive_irq_chip, xd);
+		irq_set_handler(virq + i, handle_fasteoi_irq);
+	}
+
+	return 0;
+}
+
+static void xive_irq_domain_free(struct irq_domain *domain,
+				 unsigned int virq, unsigned int nr_irqs)
+{
+	int i;
+
+	pr_debug("%s %d #%d\n", __func__, virq, nr_irqs);
+
+	for (i = 0; i < nr_irqs; i++)
+		xive_irq_free_data(domain, virq + i);
+}
+#endif
+
+static const struct irq_domain_ops xive_irq_domain_ops = {
+#ifdef	CONFIG_IRQ_DOMAIN_HIERARCHY
+	.alloc	= xive_irq_domain_alloc,
+	.free	= xive_irq_domain_free,
+	.translate = xive_irq_domain_translate,
+#endif
+	.match = xive_irq_domain_match,
+	.map = xive_irq_domain_map,
+	.unmap = xive_irq_domain_unmap,
+	.xlate = xive_irq_domain_xlate,
+#ifdef CONFIG_GENERIC_IRQ_DEBUGFS
+	.debug_show = xive_irq_domain_debug_show,
+#endif
+};
+
+static void __init xive_init_host(struct device_node *np)
+{
+	xive_irq_domain = irq_domain_create_tree(of_fwnode_handle(np), &xive_irq_domain_ops, NULL);
+	if (WARN_ON(xive_irq_domain == NULL))
+		return;
+	irq_set_default_domain(xive_irq_domain);
+}
+
+static void xive_cleanup_cpu_queues(unsigned int cpu, struct xive_cpu *xc)
+{
+	if (xc->queue[xive_irq_priority].qpage)
+		xive_ops->cleanup_queue(cpu, xc, xive_irq_priority);
+}
+
+static int xive_setup_cpu_queues(unsigned int cpu, struct xive_cpu *xc)
+{
+	int rc = 0;
+
+	/* We setup 1 queues for now with a 64k page */
+	if (!xc->queue[xive_irq_priority].qpage)
+		rc = xive_ops->setup_queue(cpu, xc, xive_irq_priority);
+
+	return rc;
+}
+
+static int xive_prepare_cpu(unsigned int cpu)
+{
+	struct xive_cpu *xc;
+
+	xc = per_cpu(xive_cpu, cpu);
+	if (!xc) {
+		xc = kzalloc_node(sizeof(struct xive_cpu),
+				  GFP_KERNEL, cpu_to_node(cpu));
+		if (!xc)
+			return -ENOMEM;
+		xc->hw_ipi = XIVE_BAD_IRQ;
+		xc->chip_id = XIVE_INVALID_CHIP_ID;
+		if (xive_ops->prepare_cpu)
+			xive_ops->prepare_cpu(cpu, xc);
+
+		per_cpu(xive_cpu, cpu) = xc;
+	}
+
+	/* Setup EQs if not already */
+	return xive_setup_cpu_queues(cpu, xc);
+}
+
+static void xive_setup_cpu(void)
+{
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+
+	/* The backend might have additional things to do */
+	if (xive_ops->setup_cpu)
+		xive_ops->setup_cpu(smp_processor_id(), xc);
+
+	/* Set CPPR to 0xff to enable flow of interrupts */
+	xc->cppr = 0xff;
+	out_8(xive_tima + xive_tima_offset + TM_CPPR, 0xff);
+}
+
+#ifdef CONFIG_SMP
+void xive_smp_setup_cpu(void)
+{
+	pr_debug("SMP setup CPU %d\n", smp_processor_id());
+
+	/* This will have already been done on the boot CPU */
+	if (smp_processor_id() != boot_cpuid)
+		xive_setup_cpu();
+
+}
+
+int xive_smp_prepare_cpu(unsigned int cpu)
+{
+	int rc;
+
+	/* Allocate per-CPU data and queues */
+	rc = xive_prepare_cpu(cpu);
+	if (rc)
+		return rc;
+
+	/* Allocate and setup IPI for the new CPU */
+	return xive_setup_cpu_ipi(cpu);
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+static void xive_flush_cpu_queue(unsigned int cpu, struct xive_cpu *xc)
+{
+	u32 irq;
+
+	/* We assume local irqs are disabled */
+	WARN_ON(!irqs_disabled());
+
+	/* Check what's already in the CPU queue */
+	while ((irq = xive_scan_interrupts(xc, false)) != 0) {
+		/*
+		 * We need to re-route that interrupt to its new destination.
+		 * First get and lock the descriptor
+		 */
+		struct irq_desc *desc = irq_to_desc(irq);
+		struct irq_data *d = irq_desc_get_irq_data(desc);
+		struct xive_irq_data *xd;
+
+		/*
+		 * Ignore anything that isn't a XIVE irq and ignore
+		 * IPIs, so can just be dropped.
+		 */
+		if (d->domain != xive_irq_domain)
+			continue;
+
+		/*
+		 * The IRQ should have already been re-routed, it's just a
+		 * stale in the old queue, so re-trigger it in order to make
+		 * it reach is new destination.
+		 */
+#ifdef DEBUG_FLUSH
+		pr_info("CPU %d: Got irq %d while offline, re-sending...\n",
+			cpu, irq);
+#endif
+		raw_spin_lock(&desc->lock);
+		xd = irq_desc_get_chip_data(desc);
+
+		/*
+		 * Clear saved_p to indicate that it's no longer pending
+		 */
+		xd->saved_p = false;
+
+		/*
+		 * For LSIs, we EOI, this will cause a resend if it's
+		 * still asserted. Otherwise do an MSI retrigger.
+		 */
+		if (xd->flags & XIVE_IRQ_FLAG_LSI)
+			xive_do_source_eoi(xd);
+		else
+			xive_irq_retrigger(d);
+
+		raw_spin_unlock(&desc->lock);
+	}
+}
+
+void xive_smp_disable_cpu(void)
+{
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+	unsigned int cpu = smp_processor_id();
+
+	/* Migrate interrupts away from the CPU */
+	irq_migrate_all_off_this_cpu();
+
+	/* Set CPPR to 0 to disable flow of interrupts */
+	xc->cppr = 0;
+	out_8(xive_tima + xive_tima_offset + TM_CPPR, 0);
+
+	/* Flush everything still in the queue */
+	xive_flush_cpu_queue(cpu, xc);
+
+	/* Re-enable CPPR  */
+	xc->cppr = 0xff;
+	out_8(xive_tima + xive_tima_offset + TM_CPPR, 0xff);
+}
+
+void xive_flush_interrupt(void)
+{
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+	unsigned int cpu = smp_processor_id();
+
+	/* Called if an interrupt occurs while the CPU is hot unplugged */
+	xive_flush_cpu_queue(cpu, xc);
+}
+
+#endif /* CONFIG_HOTPLUG_CPU */
+
+#endif /* CONFIG_SMP */
+
+noinstr void xive_teardown_cpu(void)
+{
+	struct xive_cpu *xc = __this_cpu_read(xive_cpu);
+	unsigned int cpu = smp_processor_id();
+
+	/* Set CPPR to 0 to disable flow of interrupts */
+	xc->cppr = 0;
+	out_8(xive_tima + xive_tima_offset + TM_CPPR, 0);
+
+	if (xive_ops->teardown_cpu)
+		xive_ops->teardown_cpu(cpu, xc);
+
+#ifdef CONFIG_SMP
+	/* Get rid of IPI */
+	xive_cleanup_cpu_ipi(cpu, xc);
+#endif
+
+	/* Disable and free the queues */
+	xive_cleanup_cpu_queues(cpu, xc);
+}
+
+void xive_shutdown(void)
+{
+	xive_ops->shutdown();
+}
+
+bool __init xive_core_init(struct device_node *np, const struct xive_ops *ops,
+			   void __iomem *area, u32 offset, u8 max_prio)
+{
+	xive_tima = area;
+	xive_tima_offset = offset;
+	xive_ops = ops;
+	xive_irq_priority = max_prio;
+
+	ppc_md.get_irq = xive_get_irq;
+	__xive_enabled = true;
+
+	pr_debug("Initializing host..\n");
+	xive_init_host(np);
+
+	pr_debug("Initializing boot CPU..\n");
+
+	/* Allocate per-CPU data and queues */
+	xive_prepare_cpu(smp_processor_id());
+
+	/* Get ready for interrupts */
+	xive_setup_cpu();
+
+	pr_info("Interrupt handling initialized with %s backend\n",
+		xive_ops->name);
+	pr_info("Using priority %d for all interrupts\n", max_prio);
+
+	return true;
+}
+
+__be32 *xive_queue_page_alloc(unsigned int cpu, u32 queue_shift)
+{
+	unsigned int alloc_order;
+	struct page *pages;
+	__be32 *qpage;
+
+	alloc_order = xive_alloc_order(queue_shift);
+	pages = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, alloc_order);
+	if (!pages)
+		return ERR_PTR(-ENOMEM);
+	qpage = (__be32 *)page_address(pages);
+	memset(qpage, 0, 1 << queue_shift);
+
+	return qpage;
+}
+
+static int __init xive_off(char *arg)
+{
+	xive_cmdline_disabled = true;
+	return 1;
+}
+__setup("xive=off", xive_off);
+
+static int __init xive_store_eoi_cmdline(char *arg)
+{
+	if (!arg)
+		return 1;
+
+	if (strncmp(arg, "off", 3) == 0) {
+		pr_info("StoreEOI disabled on kernel command line\n");
+		xive_store_eoi = false;
+	}
+	return 1;
+}
+__setup("xive.store-eoi=", xive_store_eoi_cmdline);
+
+#ifdef CONFIG_DEBUG_FS
+static void xive_debug_show_ipi(struct seq_file *m, int cpu)
+{
+	struct xive_cpu *xc = per_cpu(xive_cpu, cpu);
+
+	seq_printf(m, "CPU %d: ", cpu);
+	if (xc) {
+		seq_printf(m, "pp=%02x CPPR=%02x ", xc->pending_prio, xc->cppr);
+
+#ifdef CONFIG_SMP
+		{
+			char buffer[128];
+
+			xive_irq_data_dump(&xc->ipi_data, buffer, sizeof(buffer));
+			seq_printf(m, "IPI=0x%08x %s", xc->hw_ipi, buffer);
+		}
+#endif
+	}
+	seq_puts(m, "\n");
+}
+
+static void xive_debug_show_irq(struct seq_file *m, struct irq_data *d)
+{
+	unsigned int hw_irq = (unsigned int)irqd_to_hwirq(d);
+	int rc;
+	u32 target;
+	u8 prio;
+	u32 lirq;
+	char buffer[128];
+
+	rc = xive_ops->get_irq_config(hw_irq, &target, &prio, &lirq);
+	if (rc) {
+		seq_printf(m, "IRQ 0x%08x : no config rc=%d\n", hw_irq, rc);
+		return;
+	}
+
+	seq_printf(m, "IRQ 0x%08x : target=0x%x prio=%02x lirq=0x%x ",
+		   hw_irq, target, prio, lirq);
+
+	xive_irq_data_dump(irq_data_get_irq_chip_data(d), buffer, sizeof(buffer));
+	seq_puts(m, buffer);
+	seq_puts(m, "\n");
+}
+
+static int xive_irq_debug_show(struct seq_file *m, void *private)
+{
+	unsigned int i;
+	struct irq_desc *desc;
+
+	for_each_irq_desc(i, desc) {
+		struct irq_data *d = irq_domain_get_irq_data(xive_irq_domain, i);
+
+		if (d)
+			xive_debug_show_irq(m, d);
+	}
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(xive_irq_debug);
+
+static int xive_ipi_debug_show(struct seq_file *m, void *private)
+{
+	int cpu;
+
+	if (xive_ops->debug_show)
+		xive_ops->debug_show(m, private);
+
+	for_each_online_cpu(cpu)
+		xive_debug_show_ipi(m, cpu);
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(xive_ipi_debug);
+
+static void xive_eq_debug_show_one(struct seq_file *m, struct xive_q *q, u8 prio)
+{
+	int i;
+
+	seq_printf(m, "EQ%d idx=%d T=%d\n", prio, q->idx, q->toggle);
+	if (q->qpage) {
+		for (i = 0; i < q->msk + 1; i++) {
+			if (!(i % 8))
+				seq_printf(m, "%05d ", i);
+			seq_printf(m, "%08x%s", be32_to_cpup(q->qpage + i),
+				   (i + 1) % 8 ? " " : "\n");
+		}
+	}
+	seq_puts(m, "\n");
+}
+
+static int xive_eq_debug_show(struct seq_file *m, void *private)
+{
+	int cpu = (long)m->private;
+	struct xive_cpu *xc = per_cpu(xive_cpu, cpu);
+
+	if (xc)
+		xive_eq_debug_show_one(m, &xc->queue[xive_irq_priority],
+				       xive_irq_priority);
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(xive_eq_debug);
+
+static void xive_core_debugfs_create(void)
+{
+	struct dentry *xive_dir;
+	struct dentry *xive_eq_dir;
+	long cpu;
+	char name[16];
+
+	xive_dir = debugfs_create_dir("xive", arch_debugfs_dir);
+	if (IS_ERR(xive_dir))
+		return;
+
+	debugfs_create_file("ipis", 0400, xive_dir,
+			    NULL, &xive_ipi_debug_fops);
+	debugfs_create_file("interrupts", 0400, xive_dir,
+			    NULL, &xive_irq_debug_fops);
+	xive_eq_dir = debugfs_create_dir("eqs", xive_dir);
+	for_each_possible_cpu(cpu) {
+		snprintf(name, sizeof(name), "cpu%ld", cpu);
+		debugfs_create_file(name, 0400, xive_eq_dir, (void *)cpu,
+				    &xive_eq_debug_fops);
+	}
+	debugfs_create_bool("store-eoi", 0600, xive_dir, &xive_store_eoi);
+
+	if (xive_ops->debug_create)
+		xive_ops->debug_create(xive_dir);
+}
+#else
+static inline void xive_core_debugfs_create(void) { }
+#endif /* CONFIG_DEBUG_FS */
+
+int xive_core_debug_init(void)
+{
+	if (xive_enabled() && IS_ENABLED(CONFIG_DEBUG_FS))
+		xive_core_debugfs_create();
+
+	return 0;
+}
diff --git a/arch/powerpc/sysdev/xive/native.c b/arch/powerpc/sysdev/xive/native.c
new file mode 100644
index 000000000..a0934b516
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/native.c
@@ -0,0 +1,875 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016,2017 IBM Corporation.
+ */
+
+#define pr_fmt(fmt) "xive: " fmt
+
+#include <linux/types.h>
+#include <linux/irq.h>
+#include <linux/debugfs.h>
+#include <linux/smp.h>
+#include <linux/interrupt.h>
+#include <linux/seq_file.h>
+#include <linux/init.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+#include <linux/cpumask.h>
+#include <linux/mm.h>
+#include <linux/kmemleak.h>
+
+#include <asm/machdep.h>
+#include <asm/io.h>
+#include <asm/smp.h>
+#include <asm/irq.h>
+#include <asm/errno.h>
+#include <asm/xive.h>
+#include <asm/xive-regs.h>
+#include <asm/opal.h>
+#include <asm/kvm_ppc.h>
+
+#include "xive-internal.h"
+
+
+static u32 xive_provision_size;
+static u32 *xive_provision_chips;
+static u32 xive_provision_chip_count;
+static u32 xive_queue_shift;
+static u32 xive_pool_vps = XIVE_INVALID_VP;
+static struct kmem_cache *xive_provision_cache;
+static bool xive_has_single_esc;
+bool xive_has_save_restore;
+
+int xive_native_populate_irq_data(u32 hw_irq, struct xive_irq_data *data)
+{
+	__be64 flags, eoi_page, trig_page;
+	__be32 esb_shift, src_chip;
+	u64 opal_flags;
+	s64 rc;
+
+	memset(data, 0, sizeof(*data));
+
+	rc = opal_xive_get_irq_info(hw_irq, &flags, &eoi_page, &trig_page,
+				    &esb_shift, &src_chip);
+	if (rc) {
+		pr_err("opal_xive_get_irq_info(0x%x) returned %lld\n",
+		       hw_irq, rc);
+		return -EINVAL;
+	}
+
+	opal_flags = be64_to_cpu(flags);
+	if (opal_flags & OPAL_XIVE_IRQ_STORE_EOI)
+		data->flags |= XIVE_IRQ_FLAG_STORE_EOI;
+	if (opal_flags & OPAL_XIVE_IRQ_STORE_EOI2)
+		data->flags |= XIVE_IRQ_FLAG_STORE_EOI;
+	if (opal_flags & OPAL_XIVE_IRQ_LSI)
+		data->flags |= XIVE_IRQ_FLAG_LSI;
+	data->eoi_page = be64_to_cpu(eoi_page);
+	data->trig_page = be64_to_cpu(trig_page);
+	data->esb_shift = be32_to_cpu(esb_shift);
+	data->src_chip = be32_to_cpu(src_chip);
+
+	data->eoi_mmio = ioremap(data->eoi_page, 1u << data->esb_shift);
+	if (!data->eoi_mmio) {
+		pr_err("Failed to map EOI page for irq 0x%x\n", hw_irq);
+		return -ENOMEM;
+	}
+
+	data->hw_irq = hw_irq;
+
+	if (!data->trig_page)
+		return 0;
+	if (data->trig_page == data->eoi_page) {
+		data->trig_mmio = data->eoi_mmio;
+		return 0;
+	}
+
+	data->trig_mmio = ioremap(data->trig_page, 1u << data->esb_shift);
+	if (!data->trig_mmio) {
+		pr_err("Failed to map trigger page for irq 0x%x\n", hw_irq);
+		return -ENOMEM;
+	}
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_populate_irq_data);
+
+int xive_native_configure_irq(u32 hw_irq, u32 target, u8 prio, u32 sw_irq)
+{
+	s64 rc;
+
+	for (;;) {
+		rc = opal_xive_set_irq_config(hw_irq, target, prio, sw_irq);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	return rc == 0 ? 0 : -ENXIO;
+}
+EXPORT_SYMBOL_GPL(xive_native_configure_irq);
+
+static int xive_native_get_irq_config(u32 hw_irq, u32 *target, u8 *prio,
+				      u32 *sw_irq)
+{
+	s64 rc;
+	__be64 vp;
+	__be32 lirq;
+
+	rc = opal_xive_get_irq_config(hw_irq, &vp, prio, &lirq);
+
+	*target = be64_to_cpu(vp);
+	*sw_irq = be32_to_cpu(lirq);
+
+	return rc == 0 ? 0 : -ENXIO;
+}
+
+#define vp_err(vp, fmt, ...) pr_err("VP[0x%x]: " fmt, vp, ##__VA_ARGS__)
+
+/* This can be called multiple time to change a queue configuration */
+int xive_native_configure_queue(u32 vp_id, struct xive_q *q, u8 prio,
+				__be32 *qpage, u32 order, bool can_escalate)
+{
+	s64 rc = 0;
+	__be64 qeoi_page_be;
+	__be32 esc_irq_be;
+	u64 flags, qpage_phys;
+
+	/* If there's an actual queue page, clean it */
+	if (order) {
+		if (WARN_ON(!qpage))
+			return -EINVAL;
+		qpage_phys = __pa(qpage);
+	} else
+		qpage_phys = 0;
+
+	/* Initialize the rest of the fields */
+	q->msk = order ? ((1u << (order - 2)) - 1) : 0;
+	q->idx = 0;
+	q->toggle = 0;
+
+	rc = opal_xive_get_queue_info(vp_id, prio, NULL, NULL,
+				      &qeoi_page_be,
+				      &esc_irq_be,
+				      NULL);
+	if (rc) {
+		vp_err(vp_id, "Failed to get queue %d info : %lld\n", prio, rc);
+		rc = -EIO;
+		goto fail;
+	}
+	q->eoi_phys = be64_to_cpu(qeoi_page_be);
+
+	/* Default flags */
+	flags = OPAL_XIVE_EQ_ALWAYS_NOTIFY | OPAL_XIVE_EQ_ENABLED;
+
+	/* Escalation needed ? */
+	if (can_escalate) {
+		q->esc_irq = be32_to_cpu(esc_irq_be);
+		flags |= OPAL_XIVE_EQ_ESCALATE;
+	}
+
+	/* Configure and enable the queue in HW */
+	for (;;) {
+		rc = opal_xive_set_queue_info(vp_id, prio, qpage_phys, order, flags);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	if (rc) {
+		vp_err(vp_id, "Failed to set queue %d info: %lld\n", prio, rc);
+		rc = -EIO;
+	} else {
+		/*
+		 * KVM code requires all of the above to be visible before
+		 * q->qpage is set due to how it manages IPI EOIs
+		 */
+		wmb();
+		q->qpage = qpage;
+	}
+fail:
+	return rc;
+}
+EXPORT_SYMBOL_GPL(xive_native_configure_queue);
+
+static void __xive_native_disable_queue(u32 vp_id, struct xive_q *q, u8 prio)
+{
+	s64 rc;
+
+	/* Disable the queue in HW */
+	for (;;) {
+		rc = opal_xive_set_queue_info(vp_id, prio, 0, 0, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	if (rc)
+		vp_err(vp_id, "Failed to disable queue %d : %lld\n", prio, rc);
+}
+
+void xive_native_disable_queue(u32 vp_id, struct xive_q *q, u8 prio)
+{
+	__xive_native_disable_queue(vp_id, q, prio);
+}
+EXPORT_SYMBOL_GPL(xive_native_disable_queue);
+
+static int xive_native_setup_queue(unsigned int cpu, struct xive_cpu *xc, u8 prio)
+{
+	struct xive_q *q = &xc->queue[prio];
+	__be32 *qpage;
+
+	qpage = xive_queue_page_alloc(cpu, xive_queue_shift);
+	if (IS_ERR(qpage))
+		return PTR_ERR(qpage);
+
+	return xive_native_configure_queue(get_hard_smp_processor_id(cpu),
+					   q, prio, qpage, xive_queue_shift, false);
+}
+
+static void xive_native_cleanup_queue(unsigned int cpu, struct xive_cpu *xc, u8 prio)
+{
+	struct xive_q *q = &xc->queue[prio];
+	unsigned int alloc_order;
+
+	/*
+	 * We use the variant with no iounmap as this is called on exec
+	 * from an IPI and iounmap isn't safe
+	 */
+	__xive_native_disable_queue(get_hard_smp_processor_id(cpu), q, prio);
+	alloc_order = xive_alloc_order(xive_queue_shift);
+	free_pages((unsigned long)q->qpage, alloc_order);
+	q->qpage = NULL;
+}
+
+static bool xive_native_match(struct device_node *node)
+{
+	return of_device_is_compatible(node, "ibm,opal-xive-vc");
+}
+
+static s64 opal_xive_allocate_irq(u32 chip_id)
+{
+	s64 irq = opal_xive_allocate_irq_raw(chip_id);
+
+	/*
+	 * Old versions of skiboot can incorrectly return 0xffffffff to
+	 * indicate no space, fix it up here.
+	 */
+	return irq == 0xffffffff ? OPAL_RESOURCE : irq;
+}
+
+#ifdef CONFIG_SMP
+static int xive_native_get_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	s64 irq;
+
+	/* Allocate an IPI and populate info about it */
+	for (;;) {
+		irq = opal_xive_allocate_irq(xc->chip_id);
+		if (irq == OPAL_BUSY) {
+			msleep(OPAL_BUSY_DELAY_MS);
+			continue;
+		}
+		if (irq < 0) {
+			pr_err("Failed to allocate IPI on CPU %d\n", cpu);
+			return -ENXIO;
+		}
+		xc->hw_ipi = irq;
+		break;
+	}
+	return 0;
+}
+#endif /* CONFIG_SMP */
+
+u32 xive_native_alloc_irq_on_chip(u32 chip_id)
+{
+	s64 rc;
+
+	for (;;) {
+		rc = opal_xive_allocate_irq(chip_id);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	if (rc < 0)
+		return 0;
+	return rc;
+}
+EXPORT_SYMBOL_GPL(xive_native_alloc_irq_on_chip);
+
+void xive_native_free_irq(u32 irq)
+{
+	for (;;) {
+		s64 rc = opal_xive_free_irq(irq);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+}
+EXPORT_SYMBOL_GPL(xive_native_free_irq);
+
+#ifdef CONFIG_SMP
+static void xive_native_put_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	s64 rc;
+
+	/* Free the IPI */
+	if (xc->hw_ipi == XIVE_BAD_IRQ)
+		return;
+	for (;;) {
+		rc = opal_xive_free_irq(xc->hw_ipi);
+		if (rc == OPAL_BUSY) {
+			msleep(OPAL_BUSY_DELAY_MS);
+			continue;
+		}
+		xc->hw_ipi = XIVE_BAD_IRQ;
+		break;
+	}
+}
+#endif /* CONFIG_SMP */
+
+static void xive_native_shutdown(void)
+{
+	/* Switch the XIVE to emulation mode */
+	opal_xive_reset(OPAL_XIVE_MODE_EMU);
+}
+
+/*
+ * Perform an "ack" cycle on the current thread, thus
+ * grabbing the pending active priorities and updating
+ * the CPPR to the most favored one.
+ */
+static void xive_native_update_pending(struct xive_cpu *xc)
+{
+	u8 he, cppr;
+	u16 ack;
+
+	/* Perform the acknowledge hypervisor to register cycle */
+	ack = be16_to_cpu(__raw_readw(xive_tima + TM_SPC_ACK_HV_REG));
+
+	/* Synchronize subsequent queue accesses */
+	mb();
+
+	/*
+	 * Grab the CPPR and the "HE" field which indicates the source
+	 * of the hypervisor interrupt (if any)
+	 */
+	cppr = ack & 0xff;
+	he = (ack >> 8) >> 6;
+	switch(he) {
+	case TM_QW3_NSR_HE_NONE: /* Nothing to see here */
+		break;
+	case TM_QW3_NSR_HE_PHYS: /* Physical thread interrupt */
+		if (cppr == 0xff)
+			return;
+		/* Mark the priority pending */
+		xc->pending_prio |= 1 << cppr;
+
+		/*
+		 * A new interrupt should never have a CPPR less favored
+		 * than our current one.
+		 */
+		if (cppr >= xc->cppr)
+			pr_err("CPU %d odd ack CPPR, got %d at %d\n",
+			       smp_processor_id(), cppr, xc->cppr);
+
+		/* Update our idea of what the CPPR is */
+		xc->cppr = cppr;
+		break;
+	case TM_QW3_NSR_HE_POOL: /* HV Pool interrupt (unused) */
+	case TM_QW3_NSR_HE_LSI:  /* Legacy FW LSI (unused) */
+		pr_err("CPU %d got unexpected interrupt type HE=%d\n",
+		       smp_processor_id(), he);
+		return;
+	}
+}
+
+static void xive_native_prepare_cpu(unsigned int cpu, struct xive_cpu *xc)
+{
+	xc->chip_id = cpu_to_chip_id(cpu);
+}
+
+static void xive_native_setup_cpu(unsigned int cpu, struct xive_cpu *xc)
+{
+	s64 rc;
+	u32 vp;
+	__be64 vp_cam_be;
+	u64 vp_cam;
+
+	if (xive_pool_vps == XIVE_INVALID_VP)
+		return;
+
+	/* Check if pool VP already active, if it is, pull it */
+	if (in_be32(xive_tima + TM_QW2_HV_POOL + TM_WORD2) & TM_QW2W2_VP)
+		in_be64(xive_tima + TM_SPC_PULL_POOL_CTX);
+
+	/* Enable the pool VP */
+	vp = xive_pool_vps + cpu;
+	for (;;) {
+		rc = opal_xive_set_vp_info(vp, OPAL_XIVE_VP_ENABLED, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	if (rc) {
+		pr_err("Failed to enable pool VP on CPU %d\n", cpu);
+		return;
+	}
+
+	/* Grab its CAM value */
+	rc = opal_xive_get_vp_info(vp, NULL, &vp_cam_be, NULL, NULL);
+	if (rc) {
+		pr_err("Failed to get pool VP info CPU %d\n", cpu);
+		return;
+	}
+	vp_cam = be64_to_cpu(vp_cam_be);
+
+	/* Push it on the CPU (set LSMFB to 0xff to skip backlog scan) */
+	out_be32(xive_tima + TM_QW2_HV_POOL + TM_WORD0, 0xff);
+	out_be32(xive_tima + TM_QW2_HV_POOL + TM_WORD2, TM_QW2W2_VP | vp_cam);
+}
+
+static void xive_native_teardown_cpu(unsigned int cpu, struct xive_cpu *xc)
+{
+	s64 rc;
+	u32 vp;
+
+	if (xive_pool_vps == XIVE_INVALID_VP)
+		return;
+
+	/* Pull the pool VP from the CPU */
+	in_be64(xive_tima + TM_SPC_PULL_POOL_CTX);
+
+	/* Disable it */
+	vp = xive_pool_vps + cpu;
+	for (;;) {
+		rc = opal_xive_set_vp_info(vp, 0, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+}
+
+void xive_native_sync_source(u32 hw_irq)
+{
+	opal_xive_sync(XIVE_SYNC_EAS, hw_irq);
+}
+EXPORT_SYMBOL_GPL(xive_native_sync_source);
+
+void xive_native_sync_queue(u32 hw_irq)
+{
+	opal_xive_sync(XIVE_SYNC_QUEUE, hw_irq);
+}
+EXPORT_SYMBOL_GPL(xive_native_sync_queue);
+
+#ifdef CONFIG_DEBUG_FS
+static int xive_native_debug_create(struct dentry *xive_dir)
+{
+	debugfs_create_bool("save-restore", 0600, xive_dir, &xive_has_save_restore);
+	return 0;
+}
+#endif
+
+static const struct xive_ops xive_native_ops = {
+	.populate_irq_data	= xive_native_populate_irq_data,
+	.configure_irq		= xive_native_configure_irq,
+	.get_irq_config		= xive_native_get_irq_config,
+	.setup_queue		= xive_native_setup_queue,
+	.cleanup_queue		= xive_native_cleanup_queue,
+	.match			= xive_native_match,
+	.shutdown		= xive_native_shutdown,
+	.update_pending		= xive_native_update_pending,
+	.prepare_cpu		= xive_native_prepare_cpu,
+	.setup_cpu		= xive_native_setup_cpu,
+	.teardown_cpu		= xive_native_teardown_cpu,
+	.sync_source		= xive_native_sync_source,
+#ifdef CONFIG_SMP
+	.get_ipi		= xive_native_get_ipi,
+	.put_ipi		= xive_native_put_ipi,
+#endif /* CONFIG_SMP */
+#ifdef CONFIG_DEBUG_FS
+	.debug_create		= xive_native_debug_create,
+#endif /* CONFIG_DEBUG_FS */
+	.name			= "native",
+};
+
+static bool __init xive_parse_provisioning(struct device_node *np)
+{
+	int rc;
+
+	if (of_property_read_u32(np, "ibm,xive-provision-page-size",
+				 &xive_provision_size) < 0)
+		return true;
+	rc = of_property_count_elems_of_size(np, "ibm,xive-provision-chips", 4);
+	if (rc < 0) {
+		pr_err("Error %d getting provision chips array\n", rc);
+		return false;
+	}
+	xive_provision_chip_count = rc;
+	if (rc == 0)
+		return true;
+
+	xive_provision_chips = kcalloc(4, xive_provision_chip_count,
+				       GFP_KERNEL);
+	if (WARN_ON(!xive_provision_chips))
+		return false;
+
+	rc = of_property_read_u32_array(np, "ibm,xive-provision-chips",
+					xive_provision_chips,
+					xive_provision_chip_count);
+	if (rc < 0) {
+		pr_err("Error %d reading provision chips array\n", rc);
+		return false;
+	}
+
+	xive_provision_cache = kmem_cache_create("xive-provision",
+						 xive_provision_size,
+						 xive_provision_size,
+						 0, NULL);
+	if (!xive_provision_cache) {
+		pr_err("Failed to allocate provision cache\n");
+		return false;
+	}
+	return true;
+}
+
+static void __init xive_native_setup_pools(void)
+{
+	/* Allocate a pool big enough */
+	pr_debug("Allocating VP block for pool size %u\n", nr_cpu_ids);
+
+	xive_pool_vps = xive_native_alloc_vp_block(nr_cpu_ids);
+	if (WARN_ON(xive_pool_vps == XIVE_INVALID_VP))
+		pr_err("Failed to allocate pool VP, KVM might not function\n");
+
+	pr_debug("Pool VPs allocated at 0x%x for %u max CPUs\n",
+		 xive_pool_vps, nr_cpu_ids);
+}
+
+u32 xive_native_default_eq_shift(void)
+{
+	return xive_queue_shift;
+}
+EXPORT_SYMBOL_GPL(xive_native_default_eq_shift);
+
+unsigned long xive_tima_os;
+EXPORT_SYMBOL_GPL(xive_tima_os);
+
+bool __init xive_native_init(void)
+{
+	struct device_node *np;
+	struct resource r;
+	void __iomem *tima;
+	u8 max_prio = 7;
+	u32 val, cpu;
+	s64 rc;
+
+	if (xive_cmdline_disabled)
+		return false;
+
+	pr_devel("xive_native_init()\n");
+	np = of_find_compatible_node(NULL, NULL, "ibm,opal-xive-pe");
+	if (!np) {
+		pr_devel("not found !\n");
+		return false;
+	}
+	pr_devel("Found %pOF\n", np);
+
+	/* Resource 1 is HV window */
+	if (of_address_to_resource(np, 1, &r)) {
+		pr_err("Failed to get thread mgmnt area resource\n");
+		goto err_put;
+	}
+	tima = ioremap(r.start, resource_size(&r));
+	if (!tima) {
+		pr_err("Failed to map thread mgmnt area\n");
+		goto err_put;
+	}
+
+	/* Read number of priorities */
+	if (of_property_read_u32(np, "ibm,xive-#priorities", &val) == 0)
+		max_prio = val - 1;
+
+	/* Iterate the EQ sizes and pick one */
+	of_property_for_each_u32(np, "ibm,xive-eq-sizes", val) {
+		xive_queue_shift = val;
+		if (val == PAGE_SHIFT)
+			break;
+	}
+
+	/* Do we support single escalation */
+	xive_has_single_esc = of_property_read_bool(np, "single-escalation-support");
+
+	xive_has_save_restore = of_property_read_bool(np, "vp-save-restore");
+
+	/* Configure Thread Management areas for KVM */
+	for_each_possible_cpu(cpu)
+		kvmppc_set_xive_tima(cpu, r.start, tima);
+
+	/* Resource 2 is OS window */
+	if (of_address_to_resource(np, 2, &r)) {
+		pr_err("Failed to get thread mgmnt area resource\n");
+		goto err_put;
+	}
+
+	xive_tima_os = r.start;
+
+	/* Grab size of provisioning pages */
+	xive_parse_provisioning(np);
+
+	/* Switch the XIVE to exploitation mode */
+	rc = opal_xive_reset(OPAL_XIVE_MODE_EXPL);
+	if (rc) {
+		pr_err("Switch to exploitation mode failed with error %lld\n", rc);
+		goto err_put;
+	}
+
+	/* Setup some dummy HV pool VPs */
+	xive_native_setup_pools();
+
+	/* Initialize XIVE core with our backend */
+	if (!xive_core_init(np, &xive_native_ops, tima, TM_QW3_HV_PHYS,
+			    max_prio)) {
+		opal_xive_reset(OPAL_XIVE_MODE_EMU);
+		goto err_put;
+	}
+	of_node_put(np);
+	pr_info("Using %dkB queues\n", 1 << (xive_queue_shift - 10));
+	return true;
+
+err_put:
+	of_node_put(np);
+	return false;
+}
+
+static bool xive_native_provision_pages(void)
+{
+	u32 i;
+	void *p;
+
+	for (i = 0; i < xive_provision_chip_count; i++) {
+		u32 chip = xive_provision_chips[i];
+
+		/*
+		 * XXX TODO: Try to make the allocation local to the node where
+		 * the chip resides.
+		 */
+		p = kmem_cache_alloc(xive_provision_cache, GFP_KERNEL);
+		if (!p) {
+			pr_err("Failed to allocate provisioning page\n");
+			return false;
+		}
+		kmemleak_ignore(p);
+		opal_xive_donate_page(chip, __pa(p));
+	}
+	return true;
+}
+
+u32 xive_native_alloc_vp_block(u32 max_vcpus)
+{
+	s64 rc;
+	u32 order;
+
+	order = fls(max_vcpus) - 1;
+	if (max_vcpus > (1 << order))
+		order++;
+
+	pr_debug("VP block alloc, for max VCPUs %d use order %d\n",
+		 max_vcpus, order);
+
+	for (;;) {
+		rc = opal_xive_alloc_vp_block(order);
+		switch (rc) {
+		case OPAL_BUSY:
+			msleep(OPAL_BUSY_DELAY_MS);
+			break;
+		case OPAL_XIVE_PROVISIONING:
+			if (!xive_native_provision_pages())
+				return XIVE_INVALID_VP;
+			break;
+		default:
+			if (rc < 0) {
+				pr_err("OPAL failed to allocate VCPUs order %d, err %lld\n",
+				       order, rc);
+				return XIVE_INVALID_VP;
+			}
+			return rc;
+		}
+	}
+}
+EXPORT_SYMBOL_GPL(xive_native_alloc_vp_block);
+
+void xive_native_free_vp_block(u32 vp_base)
+{
+	s64 rc;
+
+	if (vp_base == XIVE_INVALID_VP)
+		return;
+
+	rc = opal_xive_free_vp_block(vp_base);
+	if (rc < 0)
+		pr_warn("OPAL error %lld freeing VP block\n", rc);
+}
+EXPORT_SYMBOL_GPL(xive_native_free_vp_block);
+
+int xive_native_enable_vp(u32 vp_id, bool single_escalation)
+{
+	s64 rc;
+	u64 flags = OPAL_XIVE_VP_ENABLED;
+
+	if (single_escalation)
+		flags |= OPAL_XIVE_VP_SINGLE_ESCALATION;
+	for (;;) {
+		rc = opal_xive_set_vp_info(vp_id, flags, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	if (rc)
+		vp_err(vp_id, "Failed to enable VP : %lld\n", rc);
+	return rc ? -EIO : 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_enable_vp);
+
+int xive_native_disable_vp(u32 vp_id)
+{
+	s64 rc;
+
+	for (;;) {
+		rc = opal_xive_set_vp_info(vp_id, 0, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(OPAL_BUSY_DELAY_MS);
+	}
+	if (rc)
+		vp_err(vp_id, "Failed to disable VP : %lld\n", rc);
+	return rc ? -EIO : 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_disable_vp);
+
+int xive_native_get_vp_info(u32 vp_id, u32 *out_cam_id, u32 *out_chip_id)
+{
+	__be64 vp_cam_be;
+	__be32 vp_chip_id_be;
+	s64 rc;
+
+	rc = opal_xive_get_vp_info(vp_id, NULL, &vp_cam_be, NULL, &vp_chip_id_be);
+	if (rc) {
+		vp_err(vp_id, "Failed to get VP info : %lld\n", rc);
+		return -EIO;
+	}
+	*out_cam_id = be64_to_cpu(vp_cam_be) & 0xffffffffu;
+	*out_chip_id = be32_to_cpu(vp_chip_id_be);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_get_vp_info);
+
+bool xive_native_has_single_escalation(void)
+{
+	return xive_has_single_esc;
+}
+EXPORT_SYMBOL_GPL(xive_native_has_single_escalation);
+
+bool xive_native_has_save_restore(void)
+{
+	return xive_has_save_restore;
+}
+EXPORT_SYMBOL_GPL(xive_native_has_save_restore);
+
+int xive_native_get_queue_info(u32 vp_id, u32 prio,
+			       u64 *out_qpage,
+			       u64 *out_qsize,
+			       u64 *out_qeoi_page,
+			       u32 *out_escalate_irq,
+			       u64 *out_qflags)
+{
+	__be64 qpage;
+	__be64 qsize;
+	__be64 qeoi_page;
+	__be32 escalate_irq;
+	__be64 qflags;
+	s64 rc;
+
+	rc = opal_xive_get_queue_info(vp_id, prio, &qpage, &qsize,
+				      &qeoi_page, &escalate_irq, &qflags);
+	if (rc) {
+		vp_err(vp_id, "failed to get queue %d info : %lld\n", prio, rc);
+		return -EIO;
+	}
+
+	if (out_qpage)
+		*out_qpage = be64_to_cpu(qpage);
+	if (out_qsize)
+		*out_qsize = be64_to_cpu(qsize);
+	if (out_qeoi_page)
+		*out_qeoi_page = be64_to_cpu(qeoi_page);
+	if (out_escalate_irq)
+		*out_escalate_irq = be32_to_cpu(escalate_irq);
+	if (out_qflags)
+		*out_qflags = be64_to_cpu(qflags);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_get_queue_info);
+
+int xive_native_get_queue_state(u32 vp_id, u32 prio, u32 *qtoggle, u32 *qindex)
+{
+	__be32 opal_qtoggle;
+	__be32 opal_qindex;
+	s64 rc;
+
+	rc = opal_xive_get_queue_state(vp_id, prio, &opal_qtoggle,
+				       &opal_qindex);
+	if (rc) {
+		vp_err(vp_id, "failed to get queue %d state : %lld\n", prio, rc);
+		return -EIO;
+	}
+
+	if (qtoggle)
+		*qtoggle = be32_to_cpu(opal_qtoggle);
+	if (qindex)
+		*qindex = be32_to_cpu(opal_qindex);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_get_queue_state);
+
+int xive_native_set_queue_state(u32 vp_id, u32 prio, u32 qtoggle, u32 qindex)
+{
+	s64 rc;
+
+	rc = opal_xive_set_queue_state(vp_id, prio, qtoggle, qindex);
+	if (rc) {
+		vp_err(vp_id, "failed to set queue %d state : %lld\n", prio, rc);
+		return -EIO;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_set_queue_state);
+
+bool xive_native_has_queue_state_support(void)
+{
+	return opal_check_token(OPAL_XIVE_GET_QUEUE_STATE) &&
+		opal_check_token(OPAL_XIVE_SET_QUEUE_STATE);
+}
+EXPORT_SYMBOL_GPL(xive_native_has_queue_state_support);
+
+int xive_native_get_vp_state(u32 vp_id, u64 *out_state)
+{
+	__be64 state;
+	s64 rc;
+
+	rc = opal_xive_get_vp_state(vp_id, &state);
+	if (rc) {
+		vp_err(vp_id, "failed to get vp state : %lld\n", rc);
+		return -EIO;
+	}
+
+	if (out_state)
+		*out_state = be64_to_cpu(state);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(xive_native_get_vp_state);
+
+machine_arch_initcall(powernv, xive_core_debug_init);
diff --git a/arch/powerpc/sysdev/xive/spapr.c b/arch/powerpc/sysdev/xive/spapr.c
new file mode 100644
index 000000000..fdf52c78a
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/spapr.c
@@ -0,0 +1,884 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright 2016,2017 IBM Corporation.
+ */
+
+#define pr_fmt(fmt) "xive: " fmt
+
+#include <linux/types.h>
+#include <linux/irq.h>
+#include <linux/seq_file.h>
+#include <linux/smp.h>
+#include <linux/interrupt.h>
+#include <linux/init.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_fdt.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/bitmap.h>
+#include <linux/cpumask.h>
+#include <linux/mm.h>
+#include <linux/delay.h>
+#include <linux/libfdt.h>
+
+#include <asm/machdep.h>
+#include <asm/prom.h>
+#include <asm/io.h>
+#include <asm/smp.h>
+#include <asm/irq.h>
+#include <asm/errno.h>
+#include <asm/xive.h>
+#include <asm/xive-regs.h>
+#include <asm/hvcall.h>
+#include <asm/svm.h>
+#include <asm/ultravisor.h>
+
+#include "xive-internal.h"
+
+static u32 xive_queue_shift;
+
+struct xive_irq_bitmap {
+	unsigned long		*bitmap;
+	unsigned int		base;
+	unsigned int		count;
+	spinlock_t		lock;
+	struct list_head	list;
+};
+
+static LIST_HEAD(xive_irq_bitmaps);
+
+static int __init xive_irq_bitmap_add(int base, int count)
+{
+	struct xive_irq_bitmap *xibm;
+
+	xibm = kzalloc_obj(*xibm);
+	if (!xibm)
+		return -ENOMEM;
+
+	spin_lock_init(&xibm->lock);
+	xibm->base = base;
+	xibm->count = count;
+	xibm->bitmap = bitmap_zalloc(xibm->count, GFP_KERNEL);
+	if (!xibm->bitmap) {
+		kfree(xibm);
+		return -ENOMEM;
+	}
+	list_add(&xibm->list, &xive_irq_bitmaps);
+
+	pr_info("Using IRQ range [%x-%x]", xibm->base,
+		xibm->base + xibm->count - 1);
+	return 0;
+}
+
+static void xive_irq_bitmap_remove_all(void)
+{
+	struct xive_irq_bitmap *xibm, *tmp;
+
+	list_for_each_entry_safe(xibm, tmp, &xive_irq_bitmaps, list) {
+		list_del(&xibm->list);
+		bitmap_free(xibm->bitmap);
+		kfree(xibm);
+	}
+}
+
+static int __xive_irq_bitmap_alloc(struct xive_irq_bitmap *xibm)
+{
+	int irq;
+
+	irq = find_first_zero_bit(xibm->bitmap, xibm->count);
+	if (irq != xibm->count) {
+		set_bit(irq, xibm->bitmap);
+		irq += xibm->base;
+	} else {
+		irq = -ENOMEM;
+	}
+
+	return irq;
+}
+
+static int xive_irq_bitmap_alloc(void)
+{
+	struct xive_irq_bitmap *xibm;
+	unsigned long flags;
+	int irq = -ENOENT;
+
+	list_for_each_entry(xibm, &xive_irq_bitmaps, list) {
+		spin_lock_irqsave(&xibm->lock, flags);
+		irq = __xive_irq_bitmap_alloc(xibm);
+		spin_unlock_irqrestore(&xibm->lock, flags);
+		if (irq >= 0)
+			break;
+	}
+	return irq;
+}
+
+static void xive_irq_bitmap_free(int irq)
+{
+	unsigned long flags;
+	struct xive_irq_bitmap *xibm;
+
+	list_for_each_entry(xibm, &xive_irq_bitmaps, list) {
+		if ((irq >= xibm->base) && (irq < xibm->base + xibm->count)) {
+			spin_lock_irqsave(&xibm->lock, flags);
+			clear_bit(irq - xibm->base, xibm->bitmap);
+			spin_unlock_irqrestore(&xibm->lock, flags);
+			break;
+		}
+	}
+}
+
+
+/* Based on the similar routines in RTAS */
+static unsigned int plpar_busy_delay_time(long rc)
+{
+	unsigned int ms = 0;
+
+	if (H_IS_LONG_BUSY(rc)) {
+		ms = get_longbusy_msecs(rc);
+	} else if (rc == H_BUSY) {
+		ms = 10; /* seems appropriate for XIVE hcalls */
+	}
+
+	return ms;
+}
+
+static unsigned int plpar_busy_delay(int rc)
+{
+	unsigned int ms;
+
+	ms = plpar_busy_delay_time(rc);
+	if (ms)
+		mdelay(ms);
+
+	return ms;
+}
+
+/*
+ * Note: this call has a partition wide scope and can take a while to
+ * complete. If it returns H_LONG_BUSY_* it should be retried
+ * periodically.
+ */
+static long plpar_int_reset(unsigned long flags)
+{
+	long rc;
+
+	do {
+		rc = plpar_hcall_norets(H_INT_RESET, flags);
+	} while (plpar_busy_delay(rc));
+
+	if (rc)
+		pr_err("H_INT_RESET failed %ld\n", rc);
+
+	return rc;
+}
+
+static long plpar_int_get_source_info(unsigned long flags,
+				      unsigned long lisn,
+				      unsigned long *src_flags,
+				      unsigned long *eoi_page,
+				      unsigned long *trig_page,
+				      unsigned long *esb_shift)
+{
+	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
+	long rc;
+
+	do {
+		rc = plpar_hcall(H_INT_GET_SOURCE_INFO, retbuf, flags, lisn);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_GET_SOURCE_INFO lisn=0x%lx failed %ld\n", lisn, rc);
+		return rc;
+	}
+
+	*src_flags = retbuf[0];
+	*eoi_page  = retbuf[1];
+	*trig_page = retbuf[2];
+	*esb_shift = retbuf[3];
+
+	pr_debug("H_INT_GET_SOURCE_INFO lisn=0x%lx flags=0x%lx eoi=0x%lx trig=0x%lx shift=0x%lx\n",
+		 lisn, retbuf[0], retbuf[1], retbuf[2], retbuf[3]);
+
+	return 0;
+}
+
+#define XIVE_SRC_SET_EISN (1ull << (63 - 62))
+#define XIVE_SRC_MASK     (1ull << (63 - 63)) /* unused */
+
+static long plpar_int_set_source_config(unsigned long flags,
+					unsigned long lisn,
+					unsigned long target,
+					unsigned long prio,
+					unsigned long sw_irq)
+{
+	long rc;
+
+
+	pr_debug("H_INT_SET_SOURCE_CONFIG flags=0x%lx lisn=0x%lx target=%ld prio=%ld sw_irq=%ld\n",
+		 flags, lisn, target, prio, sw_irq);
+
+
+	do {
+		rc = plpar_hcall_norets(H_INT_SET_SOURCE_CONFIG, flags, lisn,
+					target, prio, sw_irq);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_SET_SOURCE_CONFIG lisn=0x%lx target=%ld prio=%ld failed %ld\n",
+		       lisn, target, prio, rc);
+		return rc;
+	}
+
+	return 0;
+}
+
+static long plpar_int_get_source_config(unsigned long flags,
+					unsigned long lisn,
+					unsigned long *target,
+					unsigned long *prio,
+					unsigned long *sw_irq)
+{
+	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
+	long rc;
+
+	pr_debug("H_INT_GET_SOURCE_CONFIG flags=0x%lx lisn=0x%lx\n", flags, lisn);
+
+	do {
+		rc = plpar_hcall(H_INT_GET_SOURCE_CONFIG, retbuf, flags, lisn,
+				 target, prio, sw_irq);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_GET_SOURCE_CONFIG lisn=0x%lx failed %ld\n",
+		       lisn, rc);
+		return rc;
+	}
+
+	*target = retbuf[0];
+	*prio   = retbuf[1];
+	*sw_irq = retbuf[2];
+
+	pr_debug("H_INT_GET_SOURCE_CONFIG target=%ld prio=%ld sw_irq=%ld\n",
+		 retbuf[0], retbuf[1], retbuf[2]);
+
+	return 0;
+}
+
+static long plpar_int_get_queue_info(unsigned long flags,
+				     unsigned long target,
+				     unsigned long priority,
+				     unsigned long *esn_page,
+				     unsigned long *esn_size)
+{
+	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
+	long rc;
+
+	do {
+		rc = plpar_hcall(H_INT_GET_QUEUE_INFO, retbuf, flags, target,
+				 priority);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_GET_QUEUE_INFO cpu=%ld prio=%ld failed %ld\n",
+		       target, priority, rc);
+		return rc;
+	}
+
+	*esn_page = retbuf[0];
+	*esn_size = retbuf[1];
+
+	pr_debug("H_INT_GET_QUEUE_INFO cpu=%ld prio=%ld page=0x%lx size=0x%lx\n",
+		 target, priority, retbuf[0], retbuf[1]);
+
+	return 0;
+}
+
+#define XIVE_EQ_ALWAYS_NOTIFY (1ull << (63 - 63))
+
+static long plpar_int_set_queue_config(unsigned long flags,
+				       unsigned long target,
+				       unsigned long priority,
+				       unsigned long qpage,
+				       unsigned long qsize)
+{
+	long rc;
+
+	pr_debug("H_INT_SET_QUEUE_CONFIG flags=0x%lx target=%ld priority=0x%lx qpage=0x%lx qsize=0x%lx\n",
+		 flags,  target, priority, qpage, qsize);
+
+	do {
+		rc = plpar_hcall_norets(H_INT_SET_QUEUE_CONFIG, flags, target,
+					priority, qpage, qsize);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_SET_QUEUE_CONFIG cpu=%ld prio=%ld qpage=0x%lx returned %ld\n",
+		       target, priority, qpage, rc);
+		return  rc;
+	}
+
+	return 0;
+}
+
+static long plpar_int_sync(unsigned long flags, unsigned long lisn)
+{
+	long rc;
+
+	do {
+		rc = plpar_hcall_norets(H_INT_SYNC, flags, lisn);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_SYNC lisn=0x%lx returned %ld\n", lisn, rc);
+		return  rc;
+	}
+
+	return 0;
+}
+
+#define XIVE_ESB_FLAG_STORE (1ull << (63 - 63))
+
+static long plpar_int_esb(unsigned long flags,
+			  unsigned long lisn,
+			  unsigned long offset,
+			  unsigned long in_data,
+			  unsigned long *out_data)
+{
+	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
+	long rc;
+
+	pr_debug("H_INT_ESB flags=0x%lx lisn=0x%lx offset=0x%lx in=0x%lx\n",
+		 flags,  lisn, offset, in_data);
+
+	do {
+		rc = plpar_hcall(H_INT_ESB, retbuf, flags, lisn, offset,
+				 in_data);
+	} while (plpar_busy_delay(rc));
+
+	if (rc) {
+		pr_err("H_INT_ESB lisn=0x%lx offset=0x%lx returned %ld\n",
+		       lisn, offset, rc);
+		return  rc;
+	}
+
+	*out_data = retbuf[0];
+
+	return 0;
+}
+
+static u64 xive_spapr_esb_rw(u32 lisn, u32 offset, u64 data, bool write)
+{
+	unsigned long read_data;
+	long rc;
+
+	rc = plpar_int_esb(write ? XIVE_ESB_FLAG_STORE : 0,
+			   lisn, offset, data, &read_data);
+	if (rc)
+		return -1;
+
+	return write ? 0 : read_data;
+}
+
+#define XIVE_SRC_H_INT_ESB     (1ull << (63 - 60))
+#define XIVE_SRC_LSI           (1ull << (63 - 61))
+#define XIVE_SRC_TRIGGER       (1ull << (63 - 62))
+#define XIVE_SRC_STORE_EOI     (1ull << (63 - 63))
+
+static int xive_spapr_populate_irq_data(u32 hw_irq, struct xive_irq_data *data)
+{
+	long rc;
+	unsigned long flags;
+	unsigned long eoi_page;
+	unsigned long trig_page;
+	unsigned long esb_shift;
+
+	memset(data, 0, sizeof(*data));
+
+	rc = plpar_int_get_source_info(0, hw_irq, &flags, &eoi_page, &trig_page,
+				       &esb_shift);
+	if (rc)
+		return  -EINVAL;
+
+	if (flags & XIVE_SRC_H_INT_ESB)
+		data->flags  |= XIVE_IRQ_FLAG_H_INT_ESB;
+	if (flags & XIVE_SRC_STORE_EOI)
+		data->flags  |= XIVE_IRQ_FLAG_STORE_EOI;
+	if (flags & XIVE_SRC_LSI)
+		data->flags  |= XIVE_IRQ_FLAG_LSI;
+	data->eoi_page  = eoi_page;
+	data->esb_shift = esb_shift;
+	data->trig_page = trig_page;
+
+	data->hw_irq = hw_irq;
+
+	/*
+	 * No chip-id for the sPAPR backend. This has an impact how we
+	 * pick a target. See xive_pick_irq_target().
+	 */
+	data->src_chip = XIVE_INVALID_CHIP_ID;
+
+	/*
+	 * When the H_INT_ESB flag is set, the H_INT_ESB hcall should
+	 * be used for interrupt management. Skip the remapping of the
+	 * ESB pages which are not available.
+	 */
+	if (data->flags & XIVE_IRQ_FLAG_H_INT_ESB)
+		return 0;
+
+	data->eoi_mmio = ioremap(data->eoi_page, 1u << data->esb_shift);
+	if (!data->eoi_mmio) {
+		pr_err("Failed to map EOI page for irq 0x%x\n", hw_irq);
+		return -ENOMEM;
+	}
+
+	/* Full function page supports trigger */
+	if (flags & XIVE_SRC_TRIGGER) {
+		data->trig_mmio = data->eoi_mmio;
+		return 0;
+	}
+
+	data->trig_mmio = ioremap(data->trig_page, 1u << data->esb_shift);
+	if (!data->trig_mmio) {
+		iounmap(data->eoi_mmio);
+		pr_err("Failed to map trigger page for irq 0x%x\n", hw_irq);
+		return -ENOMEM;
+	}
+	return 0;
+}
+
+static int xive_spapr_configure_irq(u32 hw_irq, u32 target, u8 prio, u32 sw_irq)
+{
+	long rc;
+
+	rc = plpar_int_set_source_config(XIVE_SRC_SET_EISN, hw_irq, target,
+					 prio, sw_irq);
+
+	return rc == 0 ? 0 : -ENXIO;
+}
+
+static int xive_spapr_get_irq_config(u32 hw_irq, u32 *target, u8 *prio,
+				     u32 *sw_irq)
+{
+	long rc;
+	unsigned long h_target;
+	unsigned long h_prio;
+	unsigned long h_sw_irq;
+
+	rc = plpar_int_get_source_config(0, hw_irq, &h_target, &h_prio,
+					 &h_sw_irq);
+
+	*target = h_target;
+	*prio = h_prio;
+	*sw_irq = h_sw_irq;
+
+	return rc == 0 ? 0 : -ENXIO;
+}
+
+/* This can be called multiple time to change a queue configuration */
+static int xive_spapr_configure_queue(u32 target, struct xive_q *q, u8 prio,
+				   __be32 *qpage, u32 order)
+{
+	s64 rc = 0;
+	unsigned long esn_page;
+	unsigned long esn_size;
+	u64 flags, qpage_phys;
+
+	/* If there's an actual queue page, clean it */
+	if (order) {
+		if (WARN_ON(!qpage))
+			return -EINVAL;
+		qpage_phys = __pa(qpage);
+	} else {
+		qpage_phys = 0;
+	}
+
+	/* Initialize the rest of the fields */
+	q->msk = order ? ((1u << (order - 2)) - 1) : 0;
+	q->idx = 0;
+	q->toggle = 0;
+
+	rc = plpar_int_get_queue_info(0, target, prio, &esn_page, &esn_size);
+	if (rc) {
+		pr_err("Error %lld getting queue info CPU %d prio %d\n", rc,
+		       target, prio);
+		rc = -EIO;
+		goto fail;
+	}
+
+	/* TODO: add support for the notification page */
+	q->eoi_phys = esn_page;
+
+	/* Default is to always notify */
+	flags = XIVE_EQ_ALWAYS_NOTIFY;
+
+	/* Configure and enable the queue in HW */
+	rc = plpar_int_set_queue_config(flags, target, prio, qpage_phys, order);
+	if (rc) {
+		pr_err("Error %lld setting queue for CPU %d prio %d\n", rc,
+		       target, prio);
+		rc = -EIO;
+	} else {
+		q->qpage = qpage;
+		if (is_secure_guest())
+			uv_share_page(PHYS_PFN(qpage_phys),
+					1 << xive_alloc_order(order));
+	}
+fail:
+	return rc;
+}
+
+static int xive_spapr_setup_queue(unsigned int cpu, struct xive_cpu *xc,
+				  u8 prio)
+{
+	struct xive_q *q = &xc->queue[prio];
+	__be32 *qpage;
+
+	qpage = xive_queue_page_alloc(cpu, xive_queue_shift);
+	if (IS_ERR(qpage))
+		return PTR_ERR(qpage);
+
+	return xive_spapr_configure_queue(get_hard_smp_processor_id(cpu),
+					  q, prio, qpage, xive_queue_shift);
+}
+
+static void xive_spapr_cleanup_queue(unsigned int cpu, struct xive_cpu *xc,
+				  u8 prio)
+{
+	struct xive_q *q = &xc->queue[prio];
+	unsigned int alloc_order;
+	long rc;
+	int hw_cpu = get_hard_smp_processor_id(cpu);
+
+	rc = plpar_int_set_queue_config(0, hw_cpu, prio, 0, 0);
+	if (rc)
+		pr_err("Error %ld setting queue for CPU %d prio %d\n", rc,
+		       hw_cpu, prio);
+
+	alloc_order = xive_alloc_order(xive_queue_shift);
+	if (is_secure_guest())
+		uv_unshare_page(PHYS_PFN(__pa(q->qpage)), 1 << alloc_order);
+	free_pages((unsigned long)q->qpage, alloc_order);
+	q->qpage = NULL;
+}
+
+static bool xive_spapr_match(struct device_node *node)
+{
+	/* Ignore cascaded controllers for the moment */
+	return true;
+}
+
+#ifdef CONFIG_SMP
+static int xive_spapr_get_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	int irq = xive_irq_bitmap_alloc();
+
+	if (irq < 0) {
+		pr_err("Failed to allocate IPI on CPU %d\n", cpu);
+		return -ENXIO;
+	}
+
+	xc->hw_ipi = irq;
+	return 0;
+}
+
+static void xive_spapr_put_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	if (xc->hw_ipi == XIVE_BAD_IRQ)
+		return;
+
+	xive_irq_bitmap_free(xc->hw_ipi);
+	xc->hw_ipi = XIVE_BAD_IRQ;
+}
+#endif /* CONFIG_SMP */
+
+static void xive_spapr_shutdown(void)
+{
+	plpar_int_reset(0);
+}
+
+/*
+ * Perform an "ack" cycle on the current thread. Grab the pending
+ * active priorities and update the CPPR to the most favored one.
+ */
+static void xive_spapr_update_pending(struct xive_cpu *xc)
+{
+	u8 nsr, cppr;
+	u16 ack;
+
+	/*
+	 * Perform the "Acknowledge O/S to Register" cycle.
+	 *
+	 * Let's speedup the access to the TIMA using the raw I/O
+	 * accessor as we don't need the synchronisation routine of
+	 * the higher level ones
+	 */
+	ack = be16_to_cpu(__raw_readw(xive_tima + TM_SPC_ACK_OS_REG));
+
+	/* Synchronize subsequent queue accesses */
+	mb();
+
+	/*
+	 * Grab the CPPR and the "NSR" field which indicates the source
+	 * of the interrupt (if any)
+	 */
+	cppr = ack & 0xff;
+	nsr = ack >> 8;
+
+	if (nsr & TM_QW1_NSR_EO) {
+		if (cppr == 0xff)
+			return;
+		/* Mark the priority pending */
+		xc->pending_prio |= 1 << cppr;
+
+		/*
+		 * A new interrupt should never have a CPPR less favored
+		 * than our current one.
+		 */
+		if (cppr >= xc->cppr)
+			pr_err("CPU %d odd ack CPPR, got %d at %d\n",
+			       smp_processor_id(), cppr, xc->cppr);
+
+		/* Update our idea of what the CPPR is */
+		xc->cppr = cppr;
+	}
+}
+
+static void xive_spapr_setup_cpu(unsigned int cpu, struct xive_cpu *xc)
+{
+	/* Only some debug on the TIMA settings */
+	pr_debug("(HW value: %08x %08x %08x)\n",
+		 in_be32(xive_tima + TM_QW1_OS + TM_WORD0),
+		 in_be32(xive_tima + TM_QW1_OS + TM_WORD1),
+		 in_be32(xive_tima + TM_QW1_OS + TM_WORD2));
+}
+
+static void xive_spapr_teardown_cpu(unsigned int cpu, struct xive_cpu *xc)
+{
+	/* Nothing to do */;
+}
+
+static void xive_spapr_sync_source(u32 hw_irq)
+{
+	/* Specs are unclear on what this is doing */
+	plpar_int_sync(0, hw_irq);
+}
+
+static int xive_spapr_debug_show(struct seq_file *m, void *private)
+{
+	struct xive_irq_bitmap *xibm;
+
+	list_for_each_entry(xibm, &xive_irq_bitmaps, list)
+		seq_printf(m, "bitmap #%d: %*pbl\n", xibm->count, xibm->count, xibm->bitmap);
+
+	return 0;
+}
+
+static const struct xive_ops xive_spapr_ops = {
+	.populate_irq_data	= xive_spapr_populate_irq_data,
+	.configure_irq		= xive_spapr_configure_irq,
+	.get_irq_config		= xive_spapr_get_irq_config,
+	.setup_queue		= xive_spapr_setup_queue,
+	.cleanup_queue		= xive_spapr_cleanup_queue,
+	.match			= xive_spapr_match,
+	.shutdown		= xive_spapr_shutdown,
+	.update_pending		= xive_spapr_update_pending,
+	.setup_cpu		= xive_spapr_setup_cpu,
+	.teardown_cpu		= xive_spapr_teardown_cpu,
+	.sync_source		= xive_spapr_sync_source,
+	.esb_rw			= xive_spapr_esb_rw,
+#ifdef CONFIG_SMP
+	.get_ipi		= xive_spapr_get_ipi,
+	.put_ipi		= xive_spapr_put_ipi,
+	.debug_show		= xive_spapr_debug_show,
+#endif /* CONFIG_SMP */
+	.name			= "spapr",
+};
+
+/*
+ * get max priority from "/ibm,plat-res-int-priorities"
+ */
+static bool __init xive_get_max_prio(u8 *max_prio)
+{
+	struct device_node *rootdn;
+	const __be32 *reg;
+	u32 len;
+	int prio, found;
+
+	rootdn = of_find_node_by_path("/");
+	if (!rootdn) {
+		pr_err("not root node found !\n");
+		return false;
+	}
+
+	reg = of_get_property(rootdn, "ibm,plat-res-int-priorities", &len);
+	of_node_put(rootdn);
+	if (!reg) {
+		pr_err("Failed to read 'ibm,plat-res-int-priorities' property\n");
+		return false;
+	}
+
+	if (len % (2 * sizeof(u32)) != 0) {
+		pr_err("invalid 'ibm,plat-res-int-priorities' property\n");
+		return false;
+	}
+
+	/* HW supports priorities in the range [0-7] and 0xFF is a
+	 * wildcard priority used to mask. We scan the ranges reserved
+	 * by the hypervisor to find the lowest priority we can use.
+	 */
+	found = 0xFF;
+	for (prio = 0; prio < 8; prio++) {
+		int reserved = 0;
+		int i;
+
+		for (i = 0; i < len / (2 * sizeof(u32)); i++) {
+			int base  = be32_to_cpu(reg[2 * i]);
+			int range = be32_to_cpu(reg[2 * i + 1]);
+
+			if (prio >= base && prio < base + range)
+				reserved++;
+		}
+
+		if (!reserved)
+			found = prio;
+	}
+
+	if (found == 0xFF) {
+		pr_err("no valid priority found in 'ibm,plat-res-int-priorities'\n");
+		return false;
+	}
+
+	*max_prio = found;
+	return true;
+}
+
+static const u8 *__init get_vec5_feature(unsigned int index)
+{
+	unsigned long root, chosen;
+	int size;
+	const u8 *vec5;
+
+	root = of_get_flat_dt_root();
+	chosen = of_get_flat_dt_subnode_by_name(root, "chosen");
+	if (chosen == -FDT_ERR_NOTFOUND)
+		return NULL;
+
+	vec5 = of_get_flat_dt_prop(chosen, "ibm,architecture-vec-5", &size);
+	if (!vec5)
+		return NULL;
+
+	if (size <= index)
+		return NULL;
+
+	return vec5 + index;
+}
+
+static bool __init xive_spapr_disabled(void)
+{
+	const u8 *vec5_xive;
+
+	vec5_xive = get_vec5_feature(OV5_INDX(OV5_XIVE_SUPPORT));
+	if (vec5_xive) {
+		u8 val;
+
+		val = *vec5_xive & OV5_FEAT(OV5_XIVE_SUPPORT);
+		switch (val) {
+		case OV5_FEAT(OV5_XIVE_EITHER):
+		case OV5_FEAT(OV5_XIVE_LEGACY):
+			break;
+		case OV5_FEAT(OV5_XIVE_EXPLOIT):
+			/* Hypervisor only supports XIVE */
+			if (xive_cmdline_disabled)
+				pr_warn("WARNING: Ignoring cmdline option xive=off\n");
+			return false;
+		default:
+			pr_warn("%s: Unknown xive support option: 0x%x\n",
+				__func__, val);
+			break;
+		}
+	}
+
+	return xive_cmdline_disabled;
+}
+
+bool __init xive_spapr_init(void)
+{
+	struct device_node *np;
+	struct resource r;
+	void __iomem *tima;
+	u8 max_prio;
+	u32 val;
+	u32 len;
+	const __be32 *reg;
+	int i, err;
+
+	if (xive_spapr_disabled())
+		return false;
+
+	pr_devel("%s()\n", __func__);
+	np = of_find_compatible_node(NULL, NULL, "ibm,power-ivpe");
+	if (!np) {
+		pr_devel("not found !\n");
+		return false;
+	}
+	pr_devel("Found %s\n", np->full_name);
+
+	/* Resource 1 is the OS ring TIMA */
+	if (of_address_to_resource(np, 1, &r)) {
+		pr_err("Failed to get thread mgmnt area resource\n");
+		goto err_put;
+	}
+	tima = ioremap(r.start, resource_size(&r));
+	if (!tima) {
+		pr_err("Failed to map thread mgmnt area\n");
+		goto err_put;
+	}
+
+	if (!xive_get_max_prio(&max_prio))
+		goto err_unmap;
+
+	/* Feed the IRQ number allocator with the ranges given in the DT */
+	reg = of_get_property(np, "ibm,xive-lisn-ranges", &len);
+	if (!reg) {
+		pr_err("Failed to read 'ibm,xive-lisn-ranges' property\n");
+		goto err_unmap;
+	}
+
+	if (len % (2 * sizeof(u32)) != 0) {
+		pr_err("invalid 'ibm,xive-lisn-ranges' property\n");
+		goto err_unmap;
+	}
+
+	for (i = 0; i < len / (2 * sizeof(u32)); i++, reg += 2) {
+		err = xive_irq_bitmap_add(be32_to_cpu(reg[0]),
+					  be32_to_cpu(reg[1]));
+		if (err < 0)
+			goto err_mem_free;
+	}
+
+	/* Iterate the EQ sizes and pick one */
+	of_property_for_each_u32(np, "ibm,xive-eq-sizes", val) {
+		xive_queue_shift = val;
+		if (val == PAGE_SHIFT)
+			break;
+	}
+
+	/* Initialize XIVE core with our backend */
+	if (!xive_core_init(np, &xive_spapr_ops, tima, TM_QW1_OS, max_prio))
+		goto err_mem_free;
+
+	of_node_put(np);
+	pr_info("Using %dkB queues\n", 1 << (xive_queue_shift - 10));
+	return true;
+
+err_mem_free:
+	xive_irq_bitmap_remove_all();
+err_unmap:
+	iounmap(tima);
+err_put:
+	of_node_put(np);
+	return false;
+}
+
+machine_arch_initcall(pseries, xive_core_debug_init);
diff --git a/arch/powerpc/sysdev/xive/xive-internal.h b/arch/powerpc/sysdev/xive/xive-internal.h
new file mode 100644
index 000000000..fe6d95d54
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/xive-internal.h
@@ -0,0 +1,78 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright 2016,2017 IBM Corporation.
+ */
+#ifndef __XIVE_INTERNAL_H
+#define __XIVE_INTERNAL_H
+
+/*
+ * A "disabled" interrupt should never fire, to catch problems
+ * we set its logical number to this
+ */
+#define XIVE_BAD_IRQ		0x7fffffff
+#define XIVE_MAX_IRQ		(XIVE_BAD_IRQ - 1)
+
+/* Each CPU carry one of these with various per-CPU state */
+struct xive_cpu {
+#ifdef CONFIG_SMP
+	/* HW irq number and data of IPI */
+	u32 hw_ipi;
+	struct xive_irq_data ipi_data;
+#endif /* CONFIG_SMP */
+
+	int chip_id;
+
+	/* Queue datas. Only one is populated */
+#define XIVE_MAX_QUEUES	8
+	struct xive_q queue[XIVE_MAX_QUEUES];
+
+	/*
+	 * Pending mask. Each bit corresponds to a priority that
+	 * potentially has pending interrupts.
+	 */
+	u8 pending_prio;
+
+	/* Cache of HW CPPR */
+	u8 cppr;
+};
+
+/* Backend ops */
+struct xive_ops {
+	int	(*populate_irq_data)(u32 hw_irq, struct xive_irq_data *data);
+	int 	(*configure_irq)(u32 hw_irq, u32 target, u8 prio, u32 sw_irq);
+	int	(*get_irq_config)(u32 hw_irq, u32 *target, u8 *prio,
+				  u32 *sw_irq);
+	int	(*setup_queue)(unsigned int cpu, struct xive_cpu *xc, u8 prio);
+	void	(*cleanup_queue)(unsigned int cpu, struct xive_cpu *xc, u8 prio);
+	void	(*prepare_cpu)(unsigned int cpu, struct xive_cpu *xc);
+	void	(*setup_cpu)(unsigned int cpu, struct xive_cpu *xc);
+	void	(*teardown_cpu)(unsigned int cpu, struct xive_cpu *xc);
+	bool	(*match)(struct device_node *np);
+	void	(*shutdown)(void);
+
+	void	(*update_pending)(struct xive_cpu *xc);
+	void	(*sync_source)(u32 hw_irq);
+	u64	(*esb_rw)(u32 hw_irq, u32 offset, u64 data, bool write);
+#ifdef CONFIG_SMP
+	int	(*get_ipi)(unsigned int cpu, struct xive_cpu *xc);
+	void	(*put_ipi)(unsigned int cpu, struct xive_cpu *xc);
+#endif
+	int	(*debug_show)(struct seq_file *m, void *private);
+	int	(*debug_create)(struct dentry *xive_dir);
+	const char *name;
+};
+
+bool xive_core_init(struct device_node *np, const struct xive_ops *ops,
+		    void __iomem *area, u32 offset, u8 max_prio);
+__be32 *xive_queue_page_alloc(unsigned int cpu, u32 queue_shift);
+int xive_core_debug_init(void);
+
+static inline u32 xive_alloc_order(u32 queue_shift)
+{
+	return (queue_shift > PAGE_SHIFT) ? (queue_shift - PAGE_SHIFT) : 0;
+}
+
+extern bool xive_cmdline_disabled;
+extern bool xive_has_save_restore;
+
+#endif /*  __XIVE_INTERNAL_H */