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, 271 insertions, 0 deletions

diff --git a/drivers/gpio/gpio-siox.c b/drivers/gpio/gpio-siox.c
new file mode 100644
index 000000000..958034b9f
--- /dev/null
+++ b/drivers/gpio/gpio-siox.c
@@ -0,0 +1,271 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2015-2018 Pengutronix, Uwe Kleine-König <kernel@pengutronix.de>
+ */
+
+#include <linux/module.h>
+#include <linux/siox.h>
+#include <linux/gpio/driver.h>
+#include <linux/of.h>
+
+struct gpio_siox_ddata {
+	struct gpio_chip gchip;
+	struct mutex lock;
+	u8 setdata[1];
+	u8 getdata[3];
+
+	raw_spinlock_t irqlock;
+	u32 irq_enable;
+	u32 irq_status;
+	u32 irq_type[20];
+};
+
+/*
+ * Note that this callback only sets the value that is clocked out in the next
+ * cycle.
+ */
+static int gpio_siox_set_data(struct siox_device *sdevice, u8 status, u8 buf[])
+{
+	struct gpio_siox_ddata *ddata = dev_get_drvdata(&sdevice->dev);
+
+	mutex_lock(&ddata->lock);
+	buf[0] = ddata->setdata[0];
+	mutex_unlock(&ddata->lock);
+
+	return 0;
+}
+
+static int gpio_siox_get_data(struct siox_device *sdevice, const u8 buf[])
+{
+	struct gpio_siox_ddata *ddata = dev_get_drvdata(&sdevice->dev);
+	size_t offset;
+	u32 trigger;
+
+	mutex_lock(&ddata->lock);
+
+	raw_spin_lock_irq(&ddata->irqlock);
+
+	for (offset = 0; offset < 12; ++offset) {
+		unsigned int bitpos = 11 - offset;
+		unsigned int gpiolevel = buf[bitpos / 8] & (1 << bitpos % 8);
+		unsigned int prev_level =
+			ddata->getdata[bitpos / 8] & (1 << (bitpos % 8));
+		u32 irq_type = ddata->irq_type[offset];
+
+		if (gpiolevel) {
+			if ((irq_type & IRQ_TYPE_LEVEL_HIGH) ||
+			    ((irq_type & IRQ_TYPE_EDGE_RISING) && !prev_level))
+				ddata->irq_status |= 1 << offset;
+		} else {
+			if ((irq_type & IRQ_TYPE_LEVEL_LOW) ||
+			    ((irq_type & IRQ_TYPE_EDGE_FALLING) && prev_level))
+				ddata->irq_status |= 1 << offset;
+		}
+	}
+
+	trigger = ddata->irq_status & ddata->irq_enable;
+
+	raw_spin_unlock_irq(&ddata->irqlock);
+
+	ddata->getdata[0] = buf[0];
+	ddata->getdata[1] = buf[1];
+	ddata->getdata[2] = buf[2];
+
+	mutex_unlock(&ddata->lock);
+
+	for (offset = 0; offset < 12; ++offset) {
+		if (trigger & (1 << offset)) {
+			struct irq_domain *irqdomain = ddata->gchip.irq.domain;
+			unsigned int irq = irq_find_mapping(irqdomain, offset);
+
+			/*
+			 * Conceptually handle_nested_irq should call the flow
+			 * handler of the irq chip. But it doesn't, so we have
+			 * to clean the irq_status here.
+			 */
+			raw_spin_lock_irq(&ddata->irqlock);
+			ddata->irq_status &= ~(1 << offset);
+			raw_spin_unlock_irq(&ddata->irqlock);
+
+			handle_nested_irq(irq);
+		}
+	}
+
+	return 0;
+}
+
+static void gpio_siox_irq_ack(struct irq_data *d)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
+	struct gpio_siox_ddata *ddata = gpiochip_get_data(gc);
+
+	raw_spin_lock(&ddata->irqlock);
+	ddata->irq_status &= ~(1 << d->hwirq);
+	raw_spin_unlock(&ddata->irqlock);
+}
+
+static void gpio_siox_irq_mask(struct irq_data *d)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
+	struct gpio_siox_ddata *ddata = gpiochip_get_data(gc);
+
+	raw_spin_lock(&ddata->irqlock);
+	ddata->irq_enable &= ~(1 << d->hwirq);
+	raw_spin_unlock(&ddata->irqlock);
+	gpiochip_disable_irq(gc, irqd_to_hwirq(d));
+}
+
+static void gpio_siox_irq_unmask(struct irq_data *d)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
+	struct gpio_siox_ddata *ddata = gpiochip_get_data(gc);
+
+	gpiochip_enable_irq(gc, irqd_to_hwirq(d));
+	raw_spin_lock(&ddata->irqlock);
+	ddata->irq_enable |= 1 << d->hwirq;
+	raw_spin_unlock(&ddata->irqlock);
+}
+
+static int gpio_siox_irq_set_type(struct irq_data *d, u32 type)
+{
+	struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
+	struct gpio_siox_ddata *ddata = gpiochip_get_data(gc);
+
+	raw_spin_lock(&ddata->irqlock);
+	ddata->irq_type[d->hwirq] = type;
+	raw_spin_unlock(&ddata->irqlock);
+
+	return 0;
+}
+
+static int gpio_siox_get(struct gpio_chip *chip, unsigned int offset)
+{
+	struct gpio_siox_ddata *ddata = gpiochip_get_data(chip);
+	int ret;
+
+	mutex_lock(&ddata->lock);
+
+	if (offset >= 12) {
+		unsigned int bitpos = 19 - offset;
+
+		ret = ddata->setdata[0] & (1 << bitpos);
+	} else {
+		unsigned int bitpos = 11 - offset;
+
+		ret = ddata->getdata[bitpos / 8] & (1 << (bitpos % 8));
+	}
+
+	mutex_unlock(&ddata->lock);
+
+	return ret;
+}
+
+static int gpio_siox_set(struct gpio_chip *chip,
+			 unsigned int offset, int value)
+{
+	struct gpio_siox_ddata *ddata = gpiochip_get_data(chip);
+	u8 mask = 1 << (19 - offset);
+
+	mutex_lock(&ddata->lock);
+
+	if (value)
+		ddata->setdata[0] |= mask;
+	else
+		ddata->setdata[0] &= ~mask;
+
+	mutex_unlock(&ddata->lock);
+
+	return 0;
+}
+
+static int gpio_siox_direction_input(struct gpio_chip *chip,
+				     unsigned int offset)
+{
+	if (offset >= 12)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int gpio_siox_direction_output(struct gpio_chip *chip,
+				      unsigned int offset, int value)
+{
+	if (offset < 12)
+		return -EINVAL;
+
+	return gpio_siox_set(chip, offset, value);
+}
+
+static int gpio_siox_get_direction(struct gpio_chip *chip, unsigned int offset)
+{
+	if (offset < 12)
+		return GPIO_LINE_DIRECTION_IN;
+	else
+		return GPIO_LINE_DIRECTION_OUT;
+}
+
+static const struct irq_chip gpio_siox_irq_chip = {
+	.name = "siox-gpio",
+	.irq_ack = gpio_siox_irq_ack,
+	.irq_mask = gpio_siox_irq_mask,
+	.irq_unmask = gpio_siox_irq_unmask,
+	.irq_set_type = gpio_siox_irq_set_type,
+	.flags = IRQCHIP_IMMUTABLE,
+	GPIOCHIP_IRQ_RESOURCE_HELPERS,
+};
+
+static int gpio_siox_probe(struct siox_device *sdevice)
+{
+	struct gpio_siox_ddata *ddata;
+	struct gpio_irq_chip *girq;
+	struct device *dev = &sdevice->dev;
+	struct gpio_chip *gc;
+	int ret;
+
+	ddata = devm_kzalloc(dev, sizeof(*ddata), GFP_KERNEL);
+	if (!ddata)
+		return -ENOMEM;
+
+	dev_set_drvdata(dev, ddata);
+
+	mutex_init(&ddata->lock);
+	raw_spin_lock_init(&ddata->irqlock);
+
+	gc = &ddata->gchip;
+	gc->base = -1;
+	gc->can_sleep = 1;
+	gc->parent = dev;
+	gc->owner = THIS_MODULE;
+	gc->get = gpio_siox_get;
+	gc->set = gpio_siox_set;
+	gc->direction_input = gpio_siox_direction_input;
+	gc->direction_output = gpio_siox_direction_output;
+	gc->get_direction = gpio_siox_get_direction;
+	gc->ngpio = 20;
+
+	girq = &gc->irq;
+	gpio_irq_chip_set_chip(girq, &gpio_siox_irq_chip);
+	girq->default_type = IRQ_TYPE_NONE;
+	girq->handler = handle_level_irq;
+	girq->threaded = true;
+
+	ret = devm_gpiochip_add_data(dev, gc, ddata);
+	if (ret)
+		dev_err(dev, "Failed to register gpio chip (%d)\n", ret);
+
+	return ret;
+}
+
+static struct siox_driver gpio_siox_driver = {
+	.probe = gpio_siox_probe,
+	.set_data = gpio_siox_set_data,
+	.get_data = gpio_siox_get_data,
+	.driver = {
+		.name = "gpio-siox",
+	},
+};
+module_siox_driver(gpio_siox_driver);
+
+MODULE_AUTHOR("Uwe Kleine-Koenig <u.kleine-koenig@pengutronix.de>");
+MODULE_DESCRIPTION("SIOX gpio driver");
+MODULE_LICENSE("GPL v2");