KVM: SEV: Do cache maintenance on the source VM during intra-host migrationgrafted

Manually perform cache maintenance on the source VM during intra-host
migration to ensure no stale data is left in CPU caches after the VM is
destroyed.  Because the source VM is "converted" to a non-SEV VM, KVM's
memory reclaim flows won't trigger cache maintenance, e.g. when all guest
memory is reclaimed in response to detaching from the mmu_notifier.

Note, relying on the destination VM to do cache maintenance isn't an option
as KVM doesn't require identical guest memory configurations, i.e. the
source VM may have access to memory that the destination VM does not.
Enforcing equivalent memory configurations is infeasible, as it would
require a *deep* comparison of memslots, e.g. to verify that not only are
the memslot identical, but what the memslots point at is also identical.

Fixes: b56639318bb2 ("KVM: SEV: Add support for SEV intra host migration")
Cc: stable@vger.kernel.org
Reported-by: Stefan Teodorescu <fane@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20260923163721.1584779-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 284 insertions, 0 deletions

diff --git a/drivers/gpio/gpio-vx855.c b/drivers/gpio/gpio-vx855.c
new file mode 100644
index 000000000..84b3a973a
--- /dev/null
+++ b/drivers/gpio/gpio-vx855.c
@@ -0,0 +1,284 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Linux GPIOlib driver for the VIA VX855 integrated southbridge GPIO
+ *
+ * Copyright (C) 2009 VIA Technologies, Inc.
+ * Copyright (C) 2010 One Laptop per Child
+ * Author: Harald Welte <HaraldWelte@viatech.com>
+ * All rights reserved.
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/gpio/driver.h>
+#include <linux/slab.h>
+#include <linux/device.h>
+#include <linux/platform_device.h>
+#include <linux/pci.h>
+#include <linux/io.h>
+
+#define MODULE_NAME "vx855_gpio"
+
+/* The VX855 south bridge has the following GPIO pins:
+ *	GPI 0...13	General Purpose Input
+ *	GPO 0...12	General Purpose Output
+ *	GPIO 0...14	General Purpose I/O (Open-Drain)
+ */
+
+#define NR_VX855_GPI	14
+#define NR_VX855_GPO	13
+#define NR_VX855_GPIO	15
+
+#define NR_VX855_GPInO	(NR_VX855_GPI + NR_VX855_GPO)
+#define NR_VX855_GP	(NR_VX855_GPI + NR_VX855_GPO + NR_VX855_GPIO)
+
+struct vx855_gpio {
+	struct gpio_chip gpio;
+	spinlock_t lock;
+	u32 io_gpi;
+	u32 io_gpo;
+};
+
+/* resolve a GPIx into the corresponding bit position */
+static inline u_int32_t gpi_i_bit(int i)
+{
+	if (i < 10)
+		return 1 << i;
+	else
+		return 1 << (i + 14);
+}
+
+static inline u_int32_t gpo_o_bit(int i)
+{
+	if (i < 11)
+		return 1 << i;
+	else
+		return 1 << (i + 14);
+}
+
+static inline u_int32_t gpio_i_bit(int i)
+{
+	if (i < 14)
+		return 1 << (i + 10);
+	else
+		return 1 << (i + 14);
+}
+
+static inline u_int32_t gpio_o_bit(int i)
+{
+	if (i < 14)
+		return 1 << (i + 11);
+	else
+		return 1 << (i + 13);
+}
+
+/* Mapping between numeric GPIO ID and the actual GPIO hardware numbering:
+ * 0..13	GPI 0..13
+ * 14..26	GPO 0..12
+ * 27..41	GPIO 0..14
+ */
+
+static int vx855gpio_direction_input(struct gpio_chip *gpio,
+				     unsigned int nr)
+{
+	struct vx855_gpio *vg = gpiochip_get_data(gpio);
+	unsigned long flags;
+	u_int32_t reg_out;
+
+	/* Real GPI bits are always in input direction */
+	if (nr < NR_VX855_GPI)
+		return 0;
+
+	/* Real GPO bits cannot be put in output direction */
+	if (nr < NR_VX855_GPInO)
+		return -EINVAL;
+
+	/* Open Drain GPIO have to be set to one */
+	spin_lock_irqsave(&vg->lock, flags);
+	reg_out = inl(vg->io_gpo);
+	reg_out |= gpio_o_bit(nr - NR_VX855_GPInO);
+	outl(reg_out, vg->io_gpo);
+	spin_unlock_irqrestore(&vg->lock, flags);
+
+	return 0;
+}
+
+static int vx855gpio_get(struct gpio_chip *gpio, unsigned int nr)
+{
+	struct vx855_gpio *vg = gpiochip_get_data(gpio);
+	u_int32_t reg_in;
+	int ret = 0;
+
+	if (nr < NR_VX855_GPI) {
+		reg_in = inl(vg->io_gpi);
+		if (reg_in & gpi_i_bit(nr))
+			ret = 1;
+	} else if (nr < NR_VX855_GPInO) {
+		/* GPO don't have an input bit, we need to read it
+		 * back from the output register */
+		reg_in = inl(vg->io_gpo);
+		if (reg_in & gpo_o_bit(nr - NR_VX855_GPI))
+			ret = 1;
+	} else {
+		reg_in = inl(vg->io_gpi);
+		if (reg_in & gpio_i_bit(nr - NR_VX855_GPInO))
+			ret = 1;
+	}
+
+	return ret;
+}
+
+static int vx855gpio_set(struct gpio_chip *gpio, unsigned int nr, int val)
+{
+	struct vx855_gpio *vg = gpiochip_get_data(gpio);
+	unsigned long flags;
+	u_int32_t reg_out;
+
+	/* True GPI cannot be switched to output mode */
+	if (nr < NR_VX855_GPI)
+		return -EPERM;
+
+	spin_lock_irqsave(&vg->lock, flags);
+	reg_out = inl(vg->io_gpo);
+	if (nr < NR_VX855_GPInO) {
+		if (val)
+			reg_out |= gpo_o_bit(nr - NR_VX855_GPI);
+		else
+			reg_out &= ~gpo_o_bit(nr - NR_VX855_GPI);
+	} else {
+		if (val)
+			reg_out |= gpio_o_bit(nr - NR_VX855_GPInO);
+		else
+			reg_out &= ~gpio_o_bit(nr - NR_VX855_GPInO);
+	}
+	outl(reg_out, vg->io_gpo);
+	spin_unlock_irqrestore(&vg->lock, flags);
+
+	return 0;
+}
+
+static int vx855gpio_direction_output(struct gpio_chip *gpio,
+				      unsigned int nr, int val)
+{
+	/* True GPI cannot be switched to output mode */
+	if (nr < NR_VX855_GPI)
+		return -EINVAL;
+
+	/* True GPO don't need to be switched to output mode,
+	 * and GPIO are open-drain, i.e. also need no switching,
+	 * so all we do is set the level */
+	vx855gpio_set(gpio, nr, val);
+
+	return 0;
+}
+
+static int vx855gpio_set_config(struct gpio_chip *gpio, unsigned int nr,
+				unsigned long config)
+{
+	enum pin_config_param param = pinconf_to_config_param(config);
+
+	/* The GPI cannot be single-ended */
+	if (nr < NR_VX855_GPI)
+		return -EINVAL;
+
+	/* The GPO's are push-pull */
+	if (nr < NR_VX855_GPInO) {
+		if (param != PIN_CONFIG_DRIVE_PUSH_PULL)
+			return -ENOTSUPP;
+		return 0;
+	}
+
+	/* The GPIO's are open drain */
+	if (param != PIN_CONFIG_DRIVE_OPEN_DRAIN)
+		return -ENOTSUPP;
+
+	return 0;
+}
+
+static const char *vx855gpio_names[NR_VX855_GP] = {
+	"VX855_GPI0", "VX855_GPI1", "VX855_GPI2", "VX855_GPI3", "VX855_GPI4",
+	"VX855_GPI5", "VX855_GPI6", "VX855_GPI7", "VX855_GPI8", "VX855_GPI9",
+	"VX855_GPI10", "VX855_GPI11", "VX855_GPI12", "VX855_GPI13",
+	"VX855_GPO0", "VX855_GPO1", "VX855_GPO2", "VX855_GPO3", "VX855_GPO4",
+	"VX855_GPO5", "VX855_GPO6", "VX855_GPO7", "VX855_GPO8", "VX855_GPO9",
+	"VX855_GPO10", "VX855_GPO11", "VX855_GPO12",
+	"VX855_GPIO0", "VX855_GPIO1", "VX855_GPIO2", "VX855_GPIO3",
+	"VX855_GPIO4", "VX855_GPIO5", "VX855_GPIO6", "VX855_GPIO7",
+	"VX855_GPIO8", "VX855_GPIO9", "VX855_GPIO10", "VX855_GPIO11",
+	"VX855_GPIO12", "VX855_GPIO13", "VX855_GPIO14"
+};
+
+static void vx855gpio_gpio_setup(struct vx855_gpio *vg)
+{
+	struct gpio_chip *c = &vg->gpio;
+
+	c->label = "VX855 South Bridge";
+	c->owner = THIS_MODULE;
+	c->direction_input = vx855gpio_direction_input;
+	c->direction_output = vx855gpio_direction_output;
+	c->get = vx855gpio_get;
+	c->set = vx855gpio_set;
+	c->set_config = vx855gpio_set_config;
+	c->dbg_show = NULL;
+	c->base = 0;
+	c->ngpio = NR_VX855_GP;
+	c->can_sleep = false;
+	c->names = vx855gpio_names;
+}
+
+/* This platform device is ordinarily registered by the vx855 mfd driver */
+static int vx855gpio_probe(struct platform_device *pdev)
+{
+	struct resource *res_gpi;
+	struct resource *res_gpo;
+	struct vx855_gpio *vg;
+
+	res_gpi = platform_get_resource(pdev, IORESOURCE_IO, 0);
+	res_gpo = platform_get_resource(pdev, IORESOURCE_IO, 1);
+	if (!res_gpi || !res_gpo)
+		return -EBUSY;
+
+	vg = devm_kzalloc(&pdev->dev, sizeof(*vg), GFP_KERNEL);
+	if (!vg)
+		return -ENOMEM;
+
+	dev_info(&pdev->dev, "found VX855 GPIO controller\n");
+	vg->io_gpi = res_gpi->start;
+	vg->io_gpo = res_gpo->start;
+	spin_lock_init(&vg->lock);
+
+	/*
+	 * A single byte is used to control various GPIO ports on the VX855,
+	 * and in the case of the OLPC XO-1.5, some of those ports are used
+	 * for switches that are interpreted and exposed through ACPI. ACPI
+	 * will have reserved the region, so our own reservation will not
+	 * succeed. Ignore and continue.
+	 */
+
+	if (!devm_request_region(&pdev->dev, res_gpi->start,
+				 resource_size(res_gpi), MODULE_NAME "_gpi"))
+		dev_warn(&pdev->dev,
+			"GPI I/O resource busy, probably claimed by ACPI\n");
+
+	if (!devm_request_region(&pdev->dev, res_gpo->start,
+				 resource_size(res_gpo), MODULE_NAME "_gpo"))
+		dev_warn(&pdev->dev,
+			"GPO I/O resource busy, probably claimed by ACPI\n");
+
+	vx855gpio_gpio_setup(vg);
+
+	return devm_gpiochip_add_data(&pdev->dev, &vg->gpio, vg);
+}
+
+static struct platform_driver vx855gpio_driver = {
+	.driver = {
+		.name	= MODULE_NAME,
+	},
+	.probe		= vx855gpio_probe,
+};
+
+module_platform_driver(vx855gpio_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Harald Welte <HaraldWelte@viatech.com>");
+MODULE_DESCRIPTION("GPIO driver for the VIA VX855 chipset");
+MODULE_ALIAS("platform:vx855_gpio");