x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

When using eBPF to probe CS CALL instructions within a function,
a crash can be triggered.

The eBPF tool probes offset 257 of the __hrtimer_run_queues()
function:

<__hrtimer_run_queues+249>:  nopl   0x0(%rax,%rax,1)
<__hrtimer_run_queues+254>:  mov    %r14,%rdi
<__hrtimer_run_queues+257>:  cs call <__x86_indirect_thunk_r12>
<__hrtimer_run_queues+263>:  mov    %eax,%r12d
<__hrtimer_run_queues+266>:  xchg   %ax,%ax
<__hrtimer_run_queues+268>:  mov    %r13,%rdi

Which triggers this crash:

  BUG: unable to handle page fault for address: 00000000000f41c9
  #PF: supervisor write access in kernel mode
  #PF: error_code(0x0002) - not-present page
  PGD 0 P4D 0
  Oops: 0002 [#1] SMP NOPTI
  CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P
  RIP: 0010:__hrtimer_run_queues+0x106/0x230

Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is
at the 6th byte of the above CS CALL instruction. Since the CS CALL
instruction occupies 6 bytes, the exception occurred in the middle of that
call instruction.

The root cause is that when using eBPF tools to probe in the middle of a
function, a kprobe with INT3 is used as the underlying implementation.

During single-step emulation of the original CALL instruction,
int3_emulate_call() assumes that the probed CALL instruction is 5 bytes
long. However, the actual CS-prefixed CALL instruction occupies 6 bytes,
so it constructs an incorrect exception return address. When the CPU
returns from the kprobe handler, the next instruction to be executed is at
the address of the last byte of that CS CALL instruction. Coincidentally,
starting from that address, the CPU fetches and decodes a completely
different instruction, which ultimately triggers a kernel crash.

Fix the issue by using the actual instruction length obtained from
the instruction decoder when constructing the exception return
address, rather than relying on the hardcoded CALL_INSN_SIZE macro.

[ mingo: Refined the changelog ]

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Suggested-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 292 insertions, 0 deletions

diff --git a/drivers/gpio/gpio-mlxbf3.c b/drivers/gpio/gpio-mlxbf3.c
new file mode 100644
index 000000000..566326644
--- /dev/null
+++ b/drivers/gpio/gpio-mlxbf3.c
@@ -0,0 +1,292 @@
+// SPDX-License-Identifier: GPL-2.0-only OR BSD-3-Clause
+/* Copyright (C) 2022 NVIDIA CORPORATION & AFFILIATES */
+
+#include <linux/bitfield.h>
+#include <linux/bitops.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/gpio/driver.h>
+#include <linux/gpio/generic.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/spinlock.h>
+#include <linux/types.h>
+
+/*
+ * There are 2 YU GPIO blocks:
+ * gpio[0]: HOST_GPIO0->HOST_GPIO31
+ * gpio[1]: HOST_GPIO32->HOST_GPIO55
+ */
+#define MLXBF3_GPIO_MAX_PINS_PER_BLOCK 32
+#define MLXBF3_GPIO_MAX_PINS_BLOCK0    32
+#define MLXBF3_GPIO_MAX_PINS_BLOCK1    24
+
+/*
+ * fw_gpio[x] block registers and their offset
+ */
+#define MLXBF_GPIO_FW_OUTPUT_ENABLE_SET	  0x00
+#define MLXBF_GPIO_FW_DATA_OUT_SET        0x04
+
+#define MLXBF_GPIO_FW_OUTPUT_ENABLE_CLEAR 0x00
+#define MLXBF_GPIO_FW_DATA_OUT_CLEAR      0x04
+
+#define MLXBF_GPIO_CAUSE_RISE_EN          0x00
+#define MLXBF_GPIO_CAUSE_FALL_EN          0x04
+#define MLXBF_GPIO_READ_DATA_IN           0x08
+
+#define MLXBF_GPIO_CAUSE_OR_CAUSE_EVTEN0  0x00
+#define MLXBF_GPIO_CAUSE_OR_EVTEN0        0x14
+#define MLXBF_GPIO_CAUSE_OR_CLRCAUSE      0x18
+
+#define MLXBF_GPIO_CLR_ALL_INTS           GENMASK(31, 0)
+
+struct mlxbf3_gpio_context {
+	struct gpio_generic_chip chip;
+
+	/* YU GPIO block address */
+	void __iomem *gpio_set_io;
+	void __iomem *gpio_clr_io;
+	void __iomem *gpio_io;
+
+	/* YU GPIO cause block address */
+	void __iomem *gpio_cause_io;
+};
+
+static void mlxbf3_gpio_irq_enable(struct irq_data *irqd)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(irqd);
+	struct mlxbf3_gpio_context *gs = gpiochip_get_data(gc);
+	irq_hw_number_t offset = irqd_to_hwirq(irqd);
+	u32 val;
+
+	gpiochip_enable_irq(gc, offset);
+
+	guard(gpio_generic_lock_irqsave)(&gs->chip);
+
+	writel(BIT(offset), gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_CLRCAUSE);
+
+	val = readl(gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_EVTEN0);
+	val |= BIT(offset);
+	writel(val, gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_EVTEN0);
+}
+
+static void mlxbf3_gpio_irq_disable(struct irq_data *irqd)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(irqd);
+	struct mlxbf3_gpio_context *gs = gpiochip_get_data(gc);
+	irq_hw_number_t offset = irqd_to_hwirq(irqd);
+	u32 val;
+
+	scoped_guard(gpio_generic_lock_irqsave, &gs->chip) {
+		val = readl(gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_EVTEN0);
+		val &= ~BIT(offset);
+		writel(val, gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_EVTEN0);
+
+		writel(BIT(offset), gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_CLRCAUSE);
+	}
+
+	gpiochip_disable_irq(gc, offset);
+}
+
+static irqreturn_t mlxbf3_gpio_irq_handler(int irq, void *ptr)
+{
+	struct mlxbf3_gpio_context *gs = ptr;
+	struct gpio_chip *gc = &gs->chip.gc;
+	unsigned long pending;
+	u32 level;
+
+	pending = readl(gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_CAUSE_EVTEN0);
+	writel(pending, gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_CLRCAUSE);
+
+	for_each_set_bit(level, &pending, gc->ngpio)
+		generic_handle_domain_irq(gc->irq.domain, level);
+
+	return IRQ_RETVAL(pending);
+}
+
+static int
+mlxbf3_gpio_irq_set_type(struct irq_data *irqd, unsigned int type)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(irqd);
+	struct mlxbf3_gpio_context *gs = gpiochip_get_data(gc);
+	irq_hw_number_t offset = irqd_to_hwirq(irqd);
+	u32 val;
+
+	scoped_guard(gpio_generic_lock_irqsave, &gs->chip) {
+		switch (type & IRQ_TYPE_SENSE_MASK) {
+		case IRQ_TYPE_EDGE_BOTH:
+			val = readl(gs->gpio_io + MLXBF_GPIO_CAUSE_FALL_EN);
+			val |= BIT(offset);
+			writel(val, gs->gpio_io + MLXBF_GPIO_CAUSE_FALL_EN);
+			val = readl(gs->gpio_io + MLXBF_GPIO_CAUSE_RISE_EN);
+			val |= BIT(offset);
+			writel(val, gs->gpio_io + MLXBF_GPIO_CAUSE_RISE_EN);
+			break;
+		case IRQ_TYPE_EDGE_RISING:
+			val = readl(gs->gpio_io + MLXBF_GPIO_CAUSE_RISE_EN);
+			val |= BIT(offset);
+			writel(val, gs->gpio_io + MLXBF_GPIO_CAUSE_RISE_EN);
+			break;
+		case IRQ_TYPE_EDGE_FALLING:
+			val = readl(gs->gpio_io + MLXBF_GPIO_CAUSE_FALL_EN);
+			val |= BIT(offset);
+			writel(val, gs->gpio_io + MLXBF_GPIO_CAUSE_FALL_EN);
+			break;
+		default:
+			return -EINVAL;
+		}
+	}
+
+	irq_set_handler_locked(irqd, handle_edge_irq);
+
+	return 0;
+}
+
+/* This function needs to be defined for handle_edge_irq() */
+static void mlxbf3_gpio_irq_ack(struct irq_data *data)
+{
+}
+
+static const struct irq_chip gpio_mlxbf3_irqchip = {
+	.name = "MLNXBF33",
+	.irq_ack = mlxbf3_gpio_irq_ack,
+	.irq_set_type = mlxbf3_gpio_irq_set_type,
+	.irq_enable = mlxbf3_gpio_irq_enable,
+	.irq_disable = mlxbf3_gpio_irq_disable,
+	.flags = IRQCHIP_IMMUTABLE,
+	GPIOCHIP_IRQ_RESOURCE_HELPERS,
+};
+
+static int mlxbf3_gpio_add_pin_ranges(struct gpio_chip *chip)
+{
+	unsigned int id;
+
+	switch(chip->ngpio) {
+	case MLXBF3_GPIO_MAX_PINS_BLOCK0:
+		id = 0;
+		break;
+	case MLXBF3_GPIO_MAX_PINS_BLOCK1:
+		id = 1;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return gpiochip_add_pin_range(chip, "MLNXBF34:00",
+			chip->base, id * MLXBF3_GPIO_MAX_PINS_PER_BLOCK,
+			chip->ngpio);
+}
+
+static int mlxbf3_gpio_probe(struct platform_device *pdev)
+{
+	struct gpio_generic_chip_config config;
+	struct device *dev = &pdev->dev;
+	struct mlxbf3_gpio_context *gs;
+	struct gpio_irq_chip *girq;
+	struct gpio_chip *gc;
+	int ret, irq;
+
+	gs = devm_kzalloc(dev, sizeof(*gs), GFP_KERNEL);
+	if (!gs)
+		return -ENOMEM;
+
+	gs->gpio_io = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(gs->gpio_io))
+		return PTR_ERR(gs->gpio_io);
+
+	gs->gpio_cause_io = devm_platform_ioremap_resource(pdev, 1);
+	if (IS_ERR(gs->gpio_cause_io))
+		return PTR_ERR(gs->gpio_cause_io);
+
+	gs->gpio_set_io = devm_platform_ioremap_resource(pdev, 2);
+	if (IS_ERR(gs->gpio_set_io))
+		return PTR_ERR(gs->gpio_set_io);
+
+	gs->gpio_clr_io = devm_platform_ioremap_resource(pdev, 3);
+	if (IS_ERR(gs->gpio_clr_io))
+		return PTR_ERR(gs->gpio_clr_io);
+	gc = &gs->chip.gc;
+
+	config = (struct gpio_generic_chip_config) {
+		.dev = dev,
+		.sz = 4,
+		.dat = gs->gpio_io + MLXBF_GPIO_READ_DATA_IN,
+		.set = gs->gpio_set_io + MLXBF_GPIO_FW_DATA_OUT_SET,
+		.clr = gs->gpio_clr_io + MLXBF_GPIO_FW_DATA_OUT_CLEAR,
+		.dirout = gs->gpio_set_io + MLXBF_GPIO_FW_OUTPUT_ENABLE_SET,
+		.dirin = gs->gpio_clr_io + MLXBF_GPIO_FW_OUTPUT_ENABLE_CLEAR,
+	};
+
+	ret = gpio_generic_chip_init(&gs->chip, &config);
+	if (ret)
+		return dev_err_probe(dev, ret,
+				     "%s: failed to initialize the generic GPIO chip",
+				     __func__);
+
+	gc->request = gpiochip_generic_request;
+	gc->free = gpiochip_generic_free;
+	gc->owner = THIS_MODULE;
+	gc->add_pin_ranges = mlxbf3_gpio_add_pin_ranges;
+
+	irq = platform_get_irq_optional(pdev, 0);
+	if (irq >= 0) {
+		girq = &gs->chip.gc.irq;
+		gpio_irq_chip_set_chip(girq, &gpio_mlxbf3_irqchip);
+		girq->default_type = IRQ_TYPE_NONE;
+		/* This will let us handle the parent IRQ in the driver */
+		girq->num_parents = 0;
+		girq->parents = NULL;
+		girq->parent_handler = NULL;
+		girq->handler = handle_bad_irq;
+
+		/*
+		 * Directly request the irq here instead of passing
+		 * a flow-handler because the irq is shared.
+		 */
+		ret = devm_request_irq(dev, irq, mlxbf3_gpio_irq_handler,
+				       IRQF_SHARED, dev_name(dev), gs);
+		if (ret)
+			return dev_err_probe(dev, ret, "failed to request IRQ");
+	}
+
+	platform_set_drvdata(pdev, gs);
+
+	ret = devm_gpiochip_add_data(dev, gc, gs);
+	if (ret)
+		return dev_err_probe(dev, ret,
+				     "Failed adding memory mapped gpiochip\n");
+
+	return 0;
+}
+
+static void mlxbf3_gpio_shutdown(struct platform_device *pdev)
+{
+	struct mlxbf3_gpio_context *gs = platform_get_drvdata(pdev);
+
+	/* Disable and clear all interrupts */
+	writel(0, gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_EVTEN0);
+	writel(MLXBF_GPIO_CLR_ALL_INTS, gs->gpio_cause_io + MLXBF_GPIO_CAUSE_OR_CLRCAUSE);
+}
+
+static const struct acpi_device_id mlxbf3_gpio_acpi_match[] = {
+	{ "MLNXBF33", 0 },
+	{}
+};
+MODULE_DEVICE_TABLE(acpi, mlxbf3_gpio_acpi_match);
+
+static struct platform_driver mlxbf3_gpio_driver = {
+	.driver = {
+		.name = "mlxbf3_gpio",
+		.acpi_match_table = mlxbf3_gpio_acpi_match,
+	},
+	.probe    = mlxbf3_gpio_probe,
+	.shutdown = mlxbf3_gpio_shutdown,
+};
+module_platform_driver(mlxbf3_gpio_driver);
+
+MODULE_SOFTDEP("pre: pinctrl-mlxbf3");
+MODULE_DESCRIPTION("NVIDIA BlueField-3 GPIO Driver");
+MODULE_AUTHOR("Asmaa Mnebhi <asmaa@nvidia.com>");
+MODULE_LICENSE("Dual BSD/GPL");