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, 426 insertions, 0 deletions

diff --git a/drivers/clk/clk-gpio.c b/drivers/clk/clk-gpio.c
new file mode 100644
index 000000000..9099c57e2
--- /dev/null
+++ b/drivers/clk/clk-gpio.c
@@ -0,0 +1,426 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2013 - 2014 Texas Instruments Incorporated - https://www.ti.com
+ *
+ * Authors:
+ *    Jyri Sarha <jsarha@ti.com>
+ *    Sergej Sawazki <ce3a@gmx.de>
+ *
+ * Gpio controlled clock implementation
+ */
+
+#include <linux/clk-provider.h>
+#include <linux/export.h>
+#include <linux/slab.h>
+#include <linux/gpio/consumer.h>
+#include <linux/err.h>
+#include <linux/device.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/regulator/consumer.h>
+
+/**
+ * DOC: basic gpio gated clock which can be enabled and disabled
+ *      with gpio output
+ * Traits of this clock:
+ * prepare - clk_(un)prepare are functional and control a gpio that can sleep
+ * enable - clk_enable and clk_disable are functional & control
+ *          non-sleeping gpio
+ * rate - inherits rate from parent.  No clk_set_rate support
+ * parent - fixed parent.  No clk_set_parent support
+ */
+
+/**
+ * struct clk_gpio - gpio gated clock
+ *
+ * @hw:		handle between common and hardware-specific interfaces
+ * @gpiod:	gpio descriptor
+ *
+ * Clock with a gpio control for enabling and disabling the parent clock
+ * or switching between two parents by asserting or deasserting the gpio.
+ *
+ * Implements .enable, .disable and .is_enabled or
+ * .get_parent, .set_parent and .determine_rate depending on which clk_ops
+ * is used.
+ */
+struct clk_gpio {
+	struct clk_hw	hw;
+	struct gpio_desc *gpiod;
+};
+
+#define to_clk_gpio(_hw) container_of(_hw, struct clk_gpio, hw)
+
+static int clk_gpio_gate_enable(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	gpiod_set_value(clk->gpiod, 1);
+
+	return 0;
+}
+
+static void clk_gpio_gate_disable(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	gpiod_set_value(clk->gpiod, 0);
+}
+
+static int clk_gpio_gate_is_enabled(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	return gpiod_get_value(clk->gpiod);
+}
+
+static const struct clk_ops clk_gpio_gate_ops = {
+	.enable = clk_gpio_gate_enable,
+	.disable = clk_gpio_gate_disable,
+	.is_enabled = clk_gpio_gate_is_enabled,
+};
+
+static int clk_sleeping_gpio_gate_prepare(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	gpiod_set_value_cansleep(clk->gpiod, 1);
+
+	return 0;
+}
+
+static void clk_sleeping_gpio_gate_unprepare(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	gpiod_set_value_cansleep(clk->gpiod, 0);
+}
+
+static int clk_sleeping_gpio_gate_is_prepared(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	return gpiod_get_value_cansleep(clk->gpiod);
+}
+
+static const struct clk_ops clk_sleeping_gpio_gate_ops = {
+	.prepare = clk_sleeping_gpio_gate_prepare,
+	.unprepare = clk_sleeping_gpio_gate_unprepare,
+	.is_prepared = clk_sleeping_gpio_gate_is_prepared,
+};
+
+/**
+ * DOC: basic clock multiplexer which can be controlled with a gpio output
+ * Traits of this clock:
+ * prepare - clk_prepare only ensures that parents are prepared
+ * rate - rate is only affected by parent switching.  No clk_set_rate support
+ * parent - parent is adjustable through clk_set_parent
+ */
+
+static u8 clk_gpio_mux_get_parent(struct clk_hw *hw)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	return gpiod_get_value_cansleep(clk->gpiod);
+}
+
+static int clk_gpio_mux_set_parent(struct clk_hw *hw, u8 index)
+{
+	struct clk_gpio *clk = to_clk_gpio(hw);
+
+	gpiod_set_value_cansleep(clk->gpiod, index);
+
+	return 0;
+}
+
+static const struct clk_ops clk_gpio_mux_ops = {
+	.get_parent = clk_gpio_mux_get_parent,
+	.set_parent = clk_gpio_mux_set_parent,
+	.determine_rate = __clk_mux_determine_rate,
+};
+
+static struct clk_hw *clk_register_gpio(struct device *dev, u8 num_parents,
+					struct gpio_desc *gpiod,
+					const struct clk_ops *clk_gpio_ops)
+{
+	struct clk_gpio *clk_gpio;
+	struct clk_hw *hw;
+	struct clk_init_data init = {};
+	int err;
+	const struct clk_parent_data gpio_parent_data[] = {
+		{ .index = 0 },
+		{ .index = 1 },
+	};
+
+	clk_gpio = devm_kzalloc(dev, sizeof(*clk_gpio),	GFP_KERNEL);
+	if (!clk_gpio)
+		return ERR_PTR(-ENOMEM);
+
+	init.name = dev->of_node->name;
+	init.ops = clk_gpio_ops;
+	init.parent_data = gpio_parent_data;
+	init.num_parents = num_parents;
+	init.flags = CLK_SET_RATE_PARENT;
+
+	clk_gpio->gpiod = gpiod;
+	clk_gpio->hw.init = &init;
+
+	hw = &clk_gpio->hw;
+	err = devm_clk_hw_register(dev, hw);
+	if (err)
+		return ERR_PTR(err);
+
+	return hw;
+}
+
+static struct clk_hw *clk_hw_register_gpio_gate(struct device *dev,
+						int num_parents,
+						struct gpio_desc *gpiod)
+{
+	const struct clk_ops *ops;
+
+	if (gpiod_cansleep(gpiod))
+		ops = &clk_sleeping_gpio_gate_ops;
+	else
+		ops = &clk_gpio_gate_ops;
+
+	return clk_register_gpio(dev, num_parents, gpiod, ops);
+}
+
+static struct clk_hw *clk_hw_register_gpio_mux(struct device *dev,
+					       struct gpio_desc *gpiod)
+{
+	return clk_register_gpio(dev, 2, gpiod, &clk_gpio_mux_ops);
+}
+
+static int gpio_clk_driver_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *node = dev->of_node;
+	const char *gpio_name;
+	unsigned int num_parents;
+	struct gpio_desc *gpiod;
+	struct clk_hw *hw;
+	bool is_mux;
+
+	is_mux = of_device_is_compatible(node, "gpio-mux-clock");
+
+	num_parents = of_clk_get_parent_count(node);
+	if (is_mux && num_parents != 2) {
+		dev_err(dev, "mux-clock must have 2 parents\n");
+		return -EINVAL;
+	}
+
+	gpio_name = is_mux ? "select" : "enable";
+	gpiod = devm_gpiod_get(dev, gpio_name, GPIOD_OUT_LOW);
+	if (IS_ERR(gpiod))
+		return dev_err_probe(dev, PTR_ERR(gpiod),
+				     "Can't get '%s' named GPIO property\n", gpio_name);
+
+	if (is_mux)
+		hw = clk_hw_register_gpio_mux(dev, gpiod);
+	else
+		hw = clk_hw_register_gpio_gate(dev, num_parents, gpiod);
+	if (IS_ERR(hw))
+		return PTR_ERR(hw);
+
+	return devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get, hw);
+}
+
+static const struct of_device_id gpio_clk_match_table[] = {
+	{ .compatible = "gpio-mux-clock" },
+	{ .compatible = "gpio-gate-clock" },
+	{ }
+};
+
+static struct platform_driver gpio_clk_driver = {
+	.probe		= gpio_clk_driver_probe,
+	.driver		= {
+		.name	= "gpio-clk",
+		.of_match_table = gpio_clk_match_table,
+	},
+};
+builtin_platform_driver(gpio_clk_driver);
+
+/**
+ * DOC: gated fixed clock, controlled with a gpio output and a regulator
+ * Traits of this clock:
+ * prepare - clk_prepare and clk_unprepare are function & control regulator
+ *           optionally a gpio that can sleep
+ * enable - clk_enable and clk_disable are functional & control gpio
+ * rate - rate is fixed and set on clock registration
+ * parent - fixed clock is a root clock and has no parent
+ */
+
+/**
+ * struct clk_gated_fixed - Gateable fixed rate clock
+ * @clk_gpio:	instance of clk_gpio for gate-gpio
+ * @supply:	supply regulator
+ * @rate:	fixed rate
+ */
+struct clk_gated_fixed {
+	struct clk_gpio clk_gpio;
+	struct regulator *supply;
+	unsigned long rate;
+};
+
+#define to_clk_gated_fixed(_clk_gpio) container_of(_clk_gpio, struct clk_gated_fixed, clk_gpio)
+
+static unsigned long clk_gated_fixed_recalc_rate(struct clk_hw *hw,
+						 unsigned long parent_rate)
+{
+	return to_clk_gated_fixed(to_clk_gpio(hw))->rate;
+}
+
+static int clk_gated_fixed_prepare(struct clk_hw *hw)
+{
+	struct clk_gated_fixed *clk = to_clk_gated_fixed(to_clk_gpio(hw));
+
+	if (!clk->supply)
+		return 0;
+
+	return regulator_enable(clk->supply);
+}
+
+static void clk_gated_fixed_unprepare(struct clk_hw *hw)
+{
+	struct clk_gated_fixed *clk = to_clk_gated_fixed(to_clk_gpio(hw));
+
+	if (!clk->supply)
+		return;
+
+	regulator_disable(clk->supply);
+}
+
+static int clk_gated_fixed_is_prepared(struct clk_hw *hw)
+{
+	struct clk_gated_fixed *clk = to_clk_gated_fixed(to_clk_gpio(hw));
+
+	if (!clk->supply)
+		return true;
+
+	return regulator_is_enabled(clk->supply);
+}
+
+/*
+ * Fixed gated clock with non-sleeping gpio.
+ *
+ * Prepare operation turns on the supply regulator
+ * and the enable operation switches the enable-gpio.
+ */
+static const struct clk_ops clk_gated_fixed_ops = {
+	.prepare = clk_gated_fixed_prepare,
+	.unprepare = clk_gated_fixed_unprepare,
+	.is_prepared = clk_gated_fixed_is_prepared,
+	.enable = clk_gpio_gate_enable,
+	.disable = clk_gpio_gate_disable,
+	.is_enabled = clk_gpio_gate_is_enabled,
+	.recalc_rate = clk_gated_fixed_recalc_rate,
+};
+
+static int clk_sleeping_gated_fixed_prepare(struct clk_hw *hw)
+{
+	int ret;
+
+	ret = clk_gated_fixed_prepare(hw);
+	if (ret)
+		return ret;
+
+	ret = clk_sleeping_gpio_gate_prepare(hw);
+	if (ret)
+		clk_gated_fixed_unprepare(hw);
+
+	return ret;
+}
+
+static void clk_sleeping_gated_fixed_unprepare(struct clk_hw *hw)
+{
+	clk_gated_fixed_unprepare(hw);
+	clk_sleeping_gpio_gate_unprepare(hw);
+}
+
+/*
+ * Fixed gated clock with non-sleeping gpio.
+ *
+ * Enabling the supply regulator and switching the enable-gpio happens
+ * both in the prepare step.
+ * is_prepared only needs to check the gpio state, as toggling the
+ * gpio is the last step when preparing.
+ */
+static const struct clk_ops clk_sleeping_gated_fixed_ops = {
+	.prepare = clk_sleeping_gated_fixed_prepare,
+	.unprepare = clk_sleeping_gated_fixed_unprepare,
+	.is_prepared = clk_sleeping_gpio_gate_is_prepared,
+	.recalc_rate = clk_gated_fixed_recalc_rate,
+};
+
+static int clk_gated_fixed_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct clk_gated_fixed *clk;
+	const struct clk_ops *ops;
+	const char *clk_name;
+	u32 rate;
+	int ret;
+
+	clk = devm_kzalloc(dev, sizeof(*clk), GFP_KERNEL);
+	if (!clk)
+		return -ENOMEM;
+
+	ret = device_property_read_u32(dev, "clock-frequency", &rate);
+	if (ret)
+		return dev_err_probe(dev, ret, "Failed to get clock-frequency\n");
+	clk->rate = rate;
+
+	ret = device_property_read_string(dev, "clock-output-names", &clk_name);
+	if (ret)
+		clk_name = fwnode_get_name(dev->fwnode);
+
+	clk->supply = devm_regulator_get_optional(dev, "vdd");
+	if (IS_ERR(clk->supply)) {
+		if (PTR_ERR(clk->supply) != -ENODEV)
+			return dev_err_probe(dev, PTR_ERR(clk->supply),
+					     "Failed to get regulator\n");
+		clk->supply = NULL;
+	}
+
+	clk->clk_gpio.gpiod = devm_gpiod_get_optional(dev, "enable",
+						      GPIOD_OUT_LOW);
+	if (IS_ERR(clk->clk_gpio.gpiod))
+		return dev_err_probe(dev, PTR_ERR(clk->clk_gpio.gpiod),
+				     "Failed to get gpio\n");
+
+	if (gpiod_cansleep(clk->clk_gpio.gpiod))
+		ops = &clk_sleeping_gated_fixed_ops;
+	else
+		ops = &clk_gated_fixed_ops;
+
+	clk->clk_gpio.hw.init = CLK_HW_INIT_NO_PARENT(clk_name, ops, 0);
+
+	/* register the clock */
+	ret = devm_clk_hw_register(dev, &clk->clk_gpio.hw);
+	if (ret)
+		return dev_err_probe(dev, ret,
+				     "Failed to register clock\n");
+
+	ret = devm_of_clk_add_hw_provider(dev, of_clk_hw_simple_get,
+					  &clk->clk_gpio.hw);
+	if (ret)
+		return dev_err_probe(dev, ret,
+				     "Failed to register clock provider\n");
+
+	return 0;
+}
+
+static const struct of_device_id gated_fixed_clk_match_table[] = {
+	{ .compatible = "gated-fixed-clock" },
+	{ /* sentinel */ }
+};
+
+static struct platform_driver gated_fixed_clk_driver = {
+	.probe		= clk_gated_fixed_probe,
+	.driver		= {
+		.name	= "gated-fixed-clk",
+		.of_match_table = gated_fixed_clk_match_table,
+	},
+};
+builtin_platform_driver(gated_fixed_clk_driver);