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, 576 insertions, 0 deletions

diff --git a/drivers/rtc/dev.c b/drivers/rtc/dev.c
new file mode 100644
index 000000000..8ba7c25d2
--- /dev/null
+++ b/drivers/rtc/dev.c
@@ -0,0 +1,576 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * RTC subsystem, dev interface
+ *
+ * Copyright (C) 2005 Tower Technologies
+ * Author: Alessandro Zummo <a.zummo@towertech.it>
+ *
+ * based on arch/arm/common/rtctime.c
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include <linux/compat.h>
+#include <linux/module.h>
+#include <linux/rtc.h>
+#include <linux/sched/signal.h>
+#include "rtc-core.h"
+
+static dev_t rtc_devt;
+
+#define RTC_DEV_MAX 16 /* 16 RTCs should be enough for everyone... */
+
+static int rtc_dev_open(struct inode *inode, struct file *file)
+{
+	struct rtc_device *rtc = container_of(inode->i_cdev,
+					struct rtc_device, char_dev);
+
+	if (test_and_set_bit_lock(RTC_DEV_BUSY, &rtc->flags))
+		return -EBUSY;
+
+	file->private_data = rtc;
+
+	spin_lock_irq(&rtc->irq_lock);
+	rtc->irq_data = 0;
+	spin_unlock_irq(&rtc->irq_lock);
+
+	return 0;
+}
+
+#ifdef CONFIG_RTC_INTF_DEV_UIE_EMUL
+/*
+ * Routine to poll RTC seconds field for change as often as possible,
+ * after first RTC_UIE use timer to reduce polling
+ */
+static void rtc_uie_task(struct work_struct *work)
+{
+	struct rtc_device *rtc =
+		container_of(work, struct rtc_device, uie_task);
+	struct rtc_time tm;
+	int num = 0;
+	int err;
+
+	err = rtc_read_time(rtc, &tm);
+
+	spin_lock_irq(&rtc->irq_lock);
+	if (rtc->stop_uie_polling || err) {
+		rtc->uie_task_active = 0;
+	} else if (rtc->oldsecs != tm.tm_sec) {
+		num = (tm.tm_sec + 60 - rtc->oldsecs) % 60;
+		rtc->oldsecs = tm.tm_sec;
+		rtc->uie_timer.expires = jiffies + HZ - (HZ / 10);
+		rtc->uie_timer_active = 1;
+		rtc->uie_task_active = 0;
+		add_timer(&rtc->uie_timer);
+	} else if (schedule_work(&rtc->uie_task) == 0) {
+		rtc->uie_task_active = 0;
+	}
+	spin_unlock_irq(&rtc->irq_lock);
+	if (num)
+		rtc_handle_legacy_irq(rtc, num, RTC_UF);
+}
+
+static void rtc_uie_timer(struct timer_list *t)
+{
+	struct rtc_device *rtc = timer_container_of(rtc, t, uie_timer);
+	unsigned long flags;
+
+	spin_lock_irqsave(&rtc->irq_lock, flags);
+	rtc->uie_timer_active = 0;
+	rtc->uie_task_active = 1;
+	if ((schedule_work(&rtc->uie_task) == 0))
+		rtc->uie_task_active = 0;
+	spin_unlock_irqrestore(&rtc->irq_lock, flags);
+}
+
+static int clear_uie(struct rtc_device *rtc)
+{
+	spin_lock_irq(&rtc->irq_lock);
+	if (rtc->uie_irq_active) {
+		rtc->stop_uie_polling = 1;
+		if (rtc->uie_timer_active) {
+			spin_unlock_irq(&rtc->irq_lock);
+			timer_delete_sync(&rtc->uie_timer);
+			spin_lock_irq(&rtc->irq_lock);
+			rtc->uie_timer_active = 0;
+		}
+		if (rtc->uie_task_active) {
+			spin_unlock_irq(&rtc->irq_lock);
+			flush_work(&rtc->uie_task);
+			spin_lock_irq(&rtc->irq_lock);
+		}
+		rtc->uie_irq_active = 0;
+	}
+	spin_unlock_irq(&rtc->irq_lock);
+	return 0;
+}
+
+static int set_uie(struct rtc_device *rtc)
+{
+	struct rtc_time tm;
+	int err;
+
+	err = rtc_read_time(rtc, &tm);
+	if (err)
+		return err;
+	spin_lock_irq(&rtc->irq_lock);
+	if (!rtc->uie_irq_active) {
+		rtc->uie_irq_active = 1;
+		rtc->stop_uie_polling = 0;
+		rtc->oldsecs = tm.tm_sec;
+		rtc->uie_task_active = 1;
+		if (schedule_work(&rtc->uie_task) == 0)
+			rtc->uie_task_active = 0;
+	}
+	rtc->irq_data = 0;
+	spin_unlock_irq(&rtc->irq_lock);
+	return 0;
+}
+
+int rtc_dev_update_irq_enable_emul(struct rtc_device *rtc, unsigned int enabled)
+{
+	if (enabled)
+		return set_uie(rtc);
+	else
+		return clear_uie(rtc);
+}
+EXPORT_SYMBOL(rtc_dev_update_irq_enable_emul);
+
+#endif /* CONFIG_RTC_INTF_DEV_UIE_EMUL */
+
+static ssize_t
+rtc_dev_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
+{
+	struct rtc_device *rtc = file->private_data;
+
+	DECLARE_WAITQUEUE(wait, current);
+	unsigned long data;
+	ssize_t ret;
+
+	if (count != sizeof(unsigned int) && count < sizeof(unsigned long))
+		return -EINVAL;
+
+	add_wait_queue(&rtc->irq_queue, &wait);
+	do {
+		__set_current_state(TASK_INTERRUPTIBLE);
+
+		spin_lock_irq(&rtc->irq_lock);
+		data = rtc->irq_data;
+		rtc->irq_data = 0;
+		spin_unlock_irq(&rtc->irq_lock);
+
+		if (data != 0) {
+			ret = 0;
+			break;
+		}
+		if (file->f_flags & O_NONBLOCK) {
+			ret = -EAGAIN;
+			break;
+		}
+		if (signal_pending(current)) {
+			ret = -ERESTARTSYS;
+			break;
+		}
+		schedule();
+	} while (1);
+	set_current_state(TASK_RUNNING);
+	remove_wait_queue(&rtc->irq_queue, &wait);
+
+	if (ret == 0) {
+		if (sizeof(int) != sizeof(long) &&
+		    count == sizeof(unsigned int))
+			ret = put_user(data, (unsigned int __user *)buf) ?:
+				sizeof(unsigned int);
+		else
+			ret = put_user(data, (unsigned long __user *)buf) ?:
+				sizeof(unsigned long);
+	}
+	return ret;
+}
+
+static __poll_t rtc_dev_poll(struct file *file, poll_table *wait)
+{
+	struct rtc_device *rtc = file->private_data;
+	unsigned long data;
+
+	poll_wait(file, &rtc->irq_queue, wait);
+
+	/*
+	 * This read can race with the write in rtc_handle_legacy_irq().
+	 *
+	 * - If this check misses a zero to non-zero transition the next check
+	 *   will pick it up (rtc_handle_legacy_irq() wakes up rtc->irq_queue).
+	 * - Non-zero to non-zero transition misses do not change return value.
+	 * - And a non-zero to zero transition is unlikely to be missed, since
+	 *   it occurs on rtc_dev_read(), during which polling is not expected.
+	 */
+	data = data_race(rtc->irq_data);
+
+	return (data != 0) ? (EPOLLIN | EPOLLRDNORM) : 0;
+}
+
+static long rtc_dev_ioctl(struct file *file,
+			  unsigned int cmd, unsigned long arg)
+{
+	int err = 0;
+	struct rtc_device *rtc = file->private_data;
+	const struct rtc_class_ops *ops = rtc->ops;
+	struct rtc_time tm;
+	struct rtc_wkalrm alarm;
+	struct rtc_param param;
+	void __user *uarg = (void __user *)arg;
+
+	err = mutex_lock_interruptible(&rtc->ops_lock);
+	if (err)
+		return err;
+
+	/* check that the calling task has appropriate permissions
+	 * for certain ioctls. doing this check here is useful
+	 * to avoid duplicate code in each driver.
+	 */
+	switch (cmd) {
+	case RTC_EPOCH_SET:
+	case RTC_SET_TIME:
+	case RTC_PARAM_SET:
+		if (!capable(CAP_SYS_TIME))
+			err = -EACCES;
+		break;
+
+	case RTC_IRQP_SET:
+		if (arg > rtc->max_user_freq && !capable(CAP_SYS_RESOURCE))
+			err = -EACCES;
+		break;
+
+	case RTC_PIE_ON:
+		if (rtc->irq_freq > rtc->max_user_freq &&
+		    !capable(CAP_SYS_RESOURCE))
+			err = -EACCES;
+		break;
+	}
+
+	if (err)
+		goto done;
+
+	/*
+	 * Drivers *SHOULD NOT* provide ioctl implementations
+	 * for these requests.  Instead, provide methods to
+	 * support the following code, so that the RTC's main
+	 * features are accessible without using ioctls.
+	 *
+	 * RTC and alarm times will be in UTC, by preference,
+	 * but dual-booting with MS-Windows implies RTCs must
+	 * use the local wall clock time.
+	 */
+
+	switch (cmd) {
+	case RTC_ALM_READ:
+		mutex_unlock(&rtc->ops_lock);
+
+		err = rtc_read_alarm(rtc, &alarm);
+		if (err < 0)
+			return err;
+
+		if (copy_to_user(uarg, &alarm.time, sizeof(tm)))
+			err = -EFAULT;
+		return err;
+
+	case RTC_ALM_SET:
+		mutex_unlock(&rtc->ops_lock);
+
+		if (copy_from_user(&alarm.time, uarg, sizeof(tm)))
+			return -EFAULT;
+
+		alarm.enabled = 0;
+		alarm.pending = 0;
+		alarm.time.tm_wday = -1;
+		alarm.time.tm_yday = -1;
+		alarm.time.tm_isdst = -1;
+
+		/* RTC_ALM_SET alarms may be up to 24 hours in the future.
+		 * Rather than expecting every RTC to implement "don't care"
+		 * for day/month/year fields, just force the alarm to have
+		 * the right values for those fields.
+		 *
+		 * RTC_WKALM_SET should be used instead.  Not only does it
+		 * eliminate the need for a separate RTC_AIE_ON call, it
+		 * doesn't have the "alarm 23:59:59 in the future" race.
+		 *
+		 * NOTE:  some legacy code may have used invalid fields as
+		 * wildcards, exposing hardware "periodic alarm" capabilities.
+		 * Not supported here.
+		 */
+		{
+			time64_t now, then;
+
+			err = rtc_read_time(rtc, &tm);
+			if (err < 0)
+				return err;
+			now = rtc_tm_to_time64(&tm);
+
+			alarm.time.tm_mday = tm.tm_mday;
+			alarm.time.tm_mon = tm.tm_mon;
+			alarm.time.tm_year = tm.tm_year;
+			err  = rtc_valid_tm(&alarm.time);
+			if (err < 0)
+				return err;
+			then = rtc_tm_to_time64(&alarm.time);
+
+			/* alarm may need to wrap into tomorrow */
+			if (then < now) {
+				rtc_time64_to_tm(now + 24 * 60 * 60, &tm);
+				alarm.time.tm_mday = tm.tm_mday;
+				alarm.time.tm_mon = tm.tm_mon;
+				alarm.time.tm_year = tm.tm_year;
+			}
+		}
+
+		return rtc_set_alarm(rtc, &alarm);
+
+	case RTC_RD_TIME:
+		mutex_unlock(&rtc->ops_lock);
+
+		err = rtc_read_time(rtc, &tm);
+		if (err < 0)
+			return err;
+
+		if (copy_to_user(uarg, &tm, sizeof(tm)))
+			err = -EFAULT;
+		return err;
+
+	case RTC_SET_TIME:
+		mutex_unlock(&rtc->ops_lock);
+
+		if (copy_from_user(&tm, uarg, sizeof(tm)))
+			return -EFAULT;
+
+		return rtc_set_time(rtc, &tm);
+
+	case RTC_PIE_ON:
+		err = rtc_irq_set_state(rtc, 1);
+		break;
+
+	case RTC_PIE_OFF:
+		err = rtc_irq_set_state(rtc, 0);
+		break;
+
+	case RTC_AIE_ON:
+		mutex_unlock(&rtc->ops_lock);
+		return rtc_alarm_irq_enable(rtc, 1);
+
+	case RTC_AIE_OFF:
+		mutex_unlock(&rtc->ops_lock);
+		return rtc_alarm_irq_enable(rtc, 0);
+
+	case RTC_UIE_ON:
+		mutex_unlock(&rtc->ops_lock);
+		return rtc_update_irq_enable(rtc, 1);
+
+	case RTC_UIE_OFF:
+		mutex_unlock(&rtc->ops_lock);
+		return rtc_update_irq_enable(rtc, 0);
+
+	case RTC_IRQP_SET:
+		err = rtc_irq_set_freq(rtc, arg);
+		break;
+	case RTC_IRQP_READ:
+		err = put_user(rtc->irq_freq, (unsigned long __user *)uarg);
+		break;
+
+	case RTC_WKALM_SET:
+		mutex_unlock(&rtc->ops_lock);
+		if (copy_from_user(&alarm, uarg, sizeof(alarm)))
+			return -EFAULT;
+
+		return rtc_set_alarm(rtc, &alarm);
+
+	case RTC_WKALM_RD:
+		mutex_unlock(&rtc->ops_lock);
+		err = rtc_read_alarm(rtc, &alarm);
+		if (err < 0)
+			return err;
+
+		if (copy_to_user(uarg, &alarm, sizeof(alarm)))
+			err = -EFAULT;
+		return err;
+
+	case RTC_PARAM_GET:
+		if (copy_from_user(&param, uarg, sizeof(param))) {
+			mutex_unlock(&rtc->ops_lock);
+			return -EFAULT;
+		}
+
+		switch(param.param) {
+		case RTC_PARAM_FEATURES:
+			if (param.index != 0)
+				err = -EINVAL;
+			param.uvalue = rtc->features[0];
+			break;
+
+		case RTC_PARAM_CORRECTION: {
+			long offset;
+			mutex_unlock(&rtc->ops_lock);
+			if (param.index != 0)
+				return -EINVAL;
+			err = rtc_read_offset(rtc, &offset);
+			mutex_lock(&rtc->ops_lock);
+			if (err == 0)
+				param.svalue = offset;
+			break;
+		}
+		default:
+			if (rtc->ops->param_get)
+				err = rtc->ops->param_get(rtc->dev.parent, &param);
+			else
+				err = -EINVAL;
+		}
+
+		if (!err)
+			if (copy_to_user(uarg, &param, sizeof(param)))
+				err = -EFAULT;
+
+		break;
+
+	case RTC_PARAM_SET:
+		if (copy_from_user(&param, uarg, sizeof(param))) {
+			mutex_unlock(&rtc->ops_lock);
+			return -EFAULT;
+		}
+
+		switch(param.param) {
+		case RTC_PARAM_FEATURES:
+			err = -EINVAL;
+			break;
+
+		case RTC_PARAM_CORRECTION:
+			mutex_unlock(&rtc->ops_lock);
+			if (param.index != 0)
+				return -EINVAL;
+			return rtc_set_offset(rtc, param.svalue);
+
+		default:
+			if (rtc->ops->param_set)
+				err = rtc->ops->param_set(rtc->dev.parent, &param);
+			else
+				err = -EINVAL;
+		}
+
+		break;
+
+	default:
+		/* Finally try the driver's ioctl interface */
+		if (ops->ioctl) {
+			err = ops->ioctl(rtc->dev.parent, cmd, arg);
+			if (err == -ENOIOCTLCMD)
+				err = -ENOTTY;
+		} else {
+			err = -ENOTTY;
+		}
+		break;
+	}
+
+done:
+	mutex_unlock(&rtc->ops_lock);
+	return err;
+}
+
+#ifdef CONFIG_COMPAT
+#define RTC_IRQP_SET32		_IOW('p', 0x0c, __u32)
+#define RTC_IRQP_READ32		_IOR('p', 0x0b, __u32)
+#define RTC_EPOCH_SET32		_IOW('p', 0x0e, __u32)
+
+static long rtc_dev_compat_ioctl(struct file *file,
+				 unsigned int cmd, unsigned long arg)
+{
+	struct rtc_device *rtc = file->private_data;
+	void __user *uarg = compat_ptr(arg);
+
+	switch (cmd) {
+	case RTC_IRQP_READ32:
+		return put_user(rtc->irq_freq, (__u32 __user *)uarg);
+
+	case RTC_IRQP_SET32:
+		/* arg is a plain integer, not pointer */
+		return rtc_dev_ioctl(file, RTC_IRQP_SET, arg);
+
+	case RTC_EPOCH_SET32:
+		/* arg is a plain integer, not pointer */
+		return rtc_dev_ioctl(file, RTC_EPOCH_SET, arg);
+	}
+
+	return rtc_dev_ioctl(file, cmd, (unsigned long)uarg);
+}
+#endif
+
+static int rtc_dev_fasync(int fd, struct file *file, int on)
+{
+	struct rtc_device *rtc = file->private_data;
+
+	return fasync_helper(fd, file, on, &rtc->async_queue);
+}
+
+static int rtc_dev_release(struct inode *inode, struct file *file)
+{
+	struct rtc_device *rtc = file->private_data;
+
+	/* We shut down the repeating IRQs that userspace enabled,
+	 * since nothing is listening to them.
+	 *  - Update (UIE) ... currently only managed through ioctls
+	 *  - Periodic (PIE) ... also used through rtc_*() interface calls
+	 *
+	 * Leave the alarm alone; it may be set to trigger a system wakeup
+	 * later, or be used by kernel code, and is a one-shot event anyway.
+	 */
+
+	/* Keep ioctl until all drivers are converted */
+	rtc_dev_ioctl(file, RTC_UIE_OFF, 0);
+	rtc_update_irq_enable(rtc, 0);
+	rtc_irq_set_state(rtc, 0);
+
+	clear_bit_unlock(RTC_DEV_BUSY, &rtc->flags);
+	return 0;
+}
+
+static const struct file_operations rtc_dev_fops = {
+	.owner		= THIS_MODULE,
+	.read		= rtc_dev_read,
+	.poll		= rtc_dev_poll,
+	.unlocked_ioctl	= rtc_dev_ioctl,
+#ifdef CONFIG_COMPAT
+	.compat_ioctl	= rtc_dev_compat_ioctl,
+#endif
+	.open		= rtc_dev_open,
+	.release	= rtc_dev_release,
+	.fasync		= rtc_dev_fasync,
+};
+
+/* insertion/removal hooks */
+
+void rtc_dev_prepare(struct rtc_device *rtc)
+{
+	if (!rtc_devt)
+		return;
+
+	if (rtc->id >= RTC_DEV_MAX) {
+		dev_dbg(&rtc->dev, "too many RTC devices\n");
+		return;
+	}
+
+	rtc->dev.devt = MKDEV(MAJOR(rtc_devt), rtc->id);
+
+#ifdef CONFIG_RTC_INTF_DEV_UIE_EMUL
+	INIT_WORK(&rtc->uie_task, rtc_uie_task);
+	timer_setup(&rtc->uie_timer, rtc_uie_timer, 0);
+#endif
+
+	cdev_init(&rtc->char_dev, &rtc_dev_fops);
+	rtc->char_dev.owner = rtc->owner;
+}
+
+void __init rtc_dev_init(void)
+{
+	int err;
+
+	err = alloc_chrdev_region(&rtc_devt, 0, RTC_DEV_MAX, "rtc");
+	if (err < 0)
+		pr_err("failed to allocate char dev region\n");
+}