powerpc/xive: Native exploitation of the XIVE interrupt controller

The XIVE interrupt controller is the new interrupt controller
found in POWER9. It supports advanced virtualization capabilities
among other things.

Currently we use a set of firmware calls that simulate the old
"XICS" interrupt controller but this is fairly inefficient.

This adds the framework for using XIVE along with a native
backend which OPAL for configuration. Later, a backend allowing
the use in a KVM or PowerVM guest will also be provided.

This disables some fast path for interrupts in KVM when XIVE is
enabled as these rely on the firmware emulation code which is no
longer available when the XIVE is used natively by Linux.

A latter patch will make KVM also directly exploit the XIVE, thus
recovering the lost performance (and more).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[mpe: Fixup pr_xxx("XIVE:"...), don't split pr_xxx() strings,
 tweak Kconfig so XIVE_NATIVE selects XIVE and depends on POWERNV,
 fix build errors when SMP=n, fold in fixes from Ben:
   Don't call cpu_online() on an invalid CPU number
   Fix irq target selection returning out of bounds cpu#
   Extra sanity checks on cpu numbers
 ]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

1 files changed, 639 insertions, 0 deletions

diff --git a/arch/powerpc/sysdev/xive/native.c b/arch/powerpc/sysdev/xive/native.c
new file mode 100644
index 000000000000..5fae59186cb2
--- /dev/null
+++ b/arch/powerpc/sysdev/xive/native.c
@@ -0,0 +1,639 @@
+/*
+ * Copyright 2016,2017 IBM Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ */
+
+#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/slab.h>
+#include <linux/spinlock.h>
+#include <linux/delay.h>
+#include <linux/cpumask.h>
+#include <linux/mm.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/opal.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;
+
+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_LSI)
+		data->flags |= XIVE_IRQ_FLAG_LSI;
+	if (opal_flags & OPAL_XIVE_IRQ_SHIFT_BUG)
+		data->flags |= XIVE_IRQ_FLAG_SHIFT_BUG;
+	if (opal_flags & OPAL_XIVE_IRQ_MASK_VIA_FW)
+		data->flags |= XIVE_IRQ_FLAG_MASK_FW;
+	if (opal_flags & OPAL_XIVE_IRQ_EOI_VIA_FW)
+		data->flags |= XIVE_IRQ_FLAG_EOI_FW;
+	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;
+	}
+
+	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;
+}
+
+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(1);
+	}
+	return rc == 0 ? 0 : -ENXIO;
+}
+
+/* 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) {
+		pr_err("Error %lld getting queue info prio %d\n", rc, prio);
+		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(1);
+	}
+	if (rc) {
+		pr_err("Error %lld setting queue for prio %d\n", rc, prio);
+		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;
+}
+
+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);
+			break;
+		msleep(1);
+	}
+	if (rc)
+		pr_err("Error %lld disabling queue for prio %d\n", rc, prio);
+}
+
+void xive_native_disable_queue(u32 vp_id, struct xive_q *q, u8 prio)
+{
+	__xive_native_disable_queue(vp_id, q, prio);
+}
+
+static int xive_native_setup_queue(unsigned int cpu, struct xive_cpu *xc, u8 prio)
+{
+	struct xive_q *q = &xc->queue[prio];
+	unsigned int alloc_order;
+	struct page *pages;
+	__be32 *qpage;
+
+	alloc_order = (xive_queue_shift > PAGE_SHIFT) ?
+		(xive_queue_shift - PAGE_SHIFT) : 0;
+	pages = alloc_pages_node(cpu_to_node(cpu), GFP_KERNEL, alloc_order);
+	if (!pages)
+		return -ENOMEM;
+	qpage = (__be32 *)page_address(pages);
+	memset(qpage, 0, 1 << xive_queue_shift);
+	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_queue_shift > PAGE_SHIFT) ?
+		(xive_queue_shift - PAGE_SHIFT) : 0;
+	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");
+}
+
+#ifdef CONFIG_SMP
+static int xive_native_get_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	struct device_node *np;
+	unsigned int chip_id;
+	s64 irq;
+
+	/* Find the chip ID */
+	np = of_get_cpu_node(cpu, NULL);
+	if (np) {
+		if (of_property_read_u32(np, "ibm,chip-id", &chip_id) < 0)
+			chip_id = 0;
+	}
+
+	/* Allocate an IPI and populate info about it */
+	for (;;) {
+		irq = opal_xive_allocate_irq(chip_id);
+		if (irq == OPAL_BUSY) {
+			msleep(1);
+			continue;
+		}
+		if (irq < 0) {
+			pr_err("Failed to allocate IPI on CPU %d\n", cpu);
+			return -ENXIO;
+		}
+		xc->hw_ipi = irq;
+		break;
+	}
+	return 0;
+}
+
+u32 xive_native_alloc_irq(void)
+{
+	s64 rc;
+
+	for (;;) {
+		rc = opal_xive_allocate_irq(OPAL_XIVE_ANY_CHIP);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(1);
+	}
+	if (rc < 0)
+		return 0;
+	return rc;
+}
+
+void xive_native_free_irq(u32 irq)
+{
+	for (;;) {
+		s64 rc = opal_xive_free_irq(irq);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(1);
+	}
+}
+
+static void xive_native_put_ipi(unsigned int cpu, struct xive_cpu *xc)
+{
+	s64 rc;
+
+	/* Free the IPI */
+	if (!xc->hw_ipi)
+		return;
+	for (;;) {
+		rc = opal_xive_free_irq(xc->hw_ipi);
+		if (rc == OPAL_BUSY) {
+			msleep(1);
+			continue;
+		}
+		xc->hw_ipi = 0;
+		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 = GETFIELD(TM_QW3_NSR_HE, (ack >> 8));
+	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_eoi(u32 hw_irq)
+{
+	/*
+	 * Not normally used except if specific interrupts need
+	 * a workaround on EOI.
+	 */
+	opal_int_eoi(hw_irq);
+}
+
+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;
+
+	/* Enable the pool VP */
+	vp = xive_pool_vps + get_hard_smp_processor_id(cpu);
+	pr_debug("CPU %d setting up pool VP 0x%x\n", cpu, vp);
+	for (;;) {
+		rc = opal_xive_set_vp_info(vp, OPAL_XIVE_VP_ENABLED, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(1);
+	}
+	if (rc) {
+		pr_err("Failed to enable pool VP on CPU %d\n", cpu);
+		return;
+	}
+
+	/* Grab it's 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);
+
+	pr_debug("VP CAM = %llx\n", vp_cam);
+
+	/* Push it on the CPU (set LSMFB to 0xff to skip backlog scan) */
+	pr_debug("(Old HW value: %08x)\n",
+		 in_be32(xive_tima + TM_QW2_HV_POOL + TM_WORD2));
+	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);
+	pr_debug("(New HW value: %08x)\n",
+		 in_be32(xive_tima + TM_QW2_HV_POOL + TM_WORD2));
+}
+
+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 + get_hard_smp_processor_id(cpu);
+	for (;;) {
+		rc = opal_xive_set_vp_info(vp, 0, 0);
+		if (rc != OPAL_BUSY)
+			break;
+		msleep(1);
+	}
+}
+
+static void xive_native_sync_source(u32 hw_irq)
+{
+	opal_xive_sync(XIVE_SYNC_EAS, hw_irq);
+}
+
+static const struct xive_ops xive_native_ops = {
+	.populate_irq_data	= xive_native_populate_irq_data,
+	.configure_irq		= xive_native_configure_irq,
+	.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,
+	.eoi			= xive_native_eoi,
+	.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 */
+	.name			= "native",
+};
+
+static bool 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 = kzalloc(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;
+}
+
+u32 xive_native_default_eq_shift(void)
+{
+	return xive_queue_shift;
+}
+
+bool xive_native_init(void)
+{
+	struct device_node *np;
+	struct resource r;
+	void __iomem *tima;
+	struct property *prop;
+	u8 max_prio = 7;
+	const __be32 *p;
+	u32 val;
+	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 %s\n", np->full_name);
+
+	/* Resource 1 is HV window */
+	if (of_address_to_resource(np, 1, &r)) {
+		pr_err("Failed to get thread mgmnt area resource\n");
+		return false;
+	}
+	tima = ioremap(r.start, resource_size(&r));
+	if (!tima) {
+		pr_err("Failed to map thread mgmnt area\n");
+		return false;
+	}
+
+	/* 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", prop, p, val) {
+		xive_queue_shift = val;
+		if (val == PAGE_SHIFT)
+			break;
+	}
+
+	/* 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);
+		return false;
+	}
+
+	/* Initialize XIVE core with our backend */
+	if (!xive_core_init(&xive_native_ops, tima, TM_QW3_HV_PHYS,
+			    max_prio)) {
+		opal_xive_reset(OPAL_XIVE_MODE_EMU);
+		return false;
+	}
+	pr_info("Using %dkB queues\n", 1 << (xive_queue_shift - 10));
+	return true;
+}
+
+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;
+		}
+		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_info("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(1);
+			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);