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, 248 insertions, 0 deletions

diff --git a/drivers/char/nwbutton.c b/drivers/char/nwbutton.c
new file mode 100644
index 000000000..92cee5717
--- /dev/null
+++ b/drivers/char/nwbutton.c
@@ -0,0 +1,248 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * 	NetWinder Button Driver-
+ *	Copyright (C) Alex Holden <alex@linuxhacker.org> 1998, 1999.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/sched/signal.h>
+#include <linux/interrupt.h>
+#include <linux/time.h>
+#include <linux/timer.h>
+#include <linux/fs.h>
+#include <linux/miscdevice.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/init.h>
+
+#include <linux/uaccess.h>
+#include <asm/irq.h>
+#include <asm/mach-types.h>
+
+#define __NWBUTTON_C		/* Tell the header file who we are */
+#include "nwbutton.h"
+
+static void button_sequence_finished(struct timer_list *unused);
+
+static int button_press_count;		/* The count of button presses */
+/* Times for the end of a sequence */
+static DEFINE_TIMER(button_timer, button_sequence_finished);
+static DECLARE_WAIT_QUEUE_HEAD(button_wait_queue); /* Used for blocking read */
+static char button_output_buffer[32];	/* Stores data to write out of device */
+static int bcount;			/* The number of bytes in the buffer */
+static int bdelay = BUTTON_DELAY;	/* The delay, in jiffies */
+static struct button_callback button_callback_list[32]; /* The callback list */
+static int callback_count;		/* The number of callbacks registered */
+static int reboot_count = NUM_PRESSES_REBOOT; /* Number of presses to reboot */
+
+/*
+ * This function is called by other drivers to register a callback function
+ * to be called when a particular number of button presses occurs.
+ * The callback list is a static array of 32 entries (I somehow doubt many
+ * people are ever going to want to register more than 32 different actions
+ * to be performed by the kernel on different numbers of button presses ;).
+ * However, if an attempt to register a 33rd entry (perhaps a stuck loop
+ * somewhere registering the same entry over and over?) it will fail to
+ * do so and return -ENOMEM. If an attempt is made to register a null pointer,
+ * it will fail to do so and return -EINVAL.
+ * Because callbacks can be unregistered at random the list can become
+ * fragmented, so we need to search through the list until we find the first
+ * free entry.
+ *
+ * FIXME: Has anyone spotted any locking functions int his code recently ??
+ */
+
+int button_add_callback (void (*callback) (void), int count)
+{
+	int lp = 0;
+	if (callback_count == 32) {
+		return -ENOMEM;
+	}
+	if (!callback) {
+		return -EINVAL;
+	}
+	callback_count++;
+	for (; (button_callback_list [lp].callback); lp++);
+	button_callback_list [lp].callback = callback;
+	button_callback_list [lp].count = count;
+	return 0;
+}
+
+/*
+ * This function is called by other drivers to deregister a callback function.
+ * If you attempt to unregister a callback which does not exist, it will fail
+ * with -EINVAL. If there is more than one entry with the same address,
+ * because it searches the list from end to beginning, it will unregister the
+ * last one to be registered first (FILO- First In Last Out).
+ * Note that this is not necessarily true if the entries are not submitted
+ * at the same time, because another driver could have unregistered a callback
+ * between the submissions creating a gap earlier in the list, which would
+ * be filled first at submission time.
+ */
+
+int button_del_callback (void (*callback) (void))
+{
+	int lp = 31;
+	if (!callback) {
+		return -EINVAL;
+	}
+	while (lp >= 0) {
+		if ((button_callback_list [lp].callback) == callback) {
+			button_callback_list [lp].callback = NULL;
+			button_callback_list [lp].count = 0;
+			callback_count--;
+			return 0;
+		}
+		lp--;
+	}
+	return -EINVAL;
+}
+
+/*
+ * This function is called by button_sequence_finished to search through the
+ * list of callback functions, and call any of them whose count argument
+ * matches the current count of button presses. It starts at the beginning
+ * of the list and works up to the end. It will refuse to follow a null
+ * pointer (which should never happen anyway).
+ */
+
+static void button_consume_callbacks (int bpcount)
+{
+	int lp = 0;
+	for (; lp <= 31; lp++) {
+		if ((button_callback_list [lp].count) == bpcount) {
+			if (button_callback_list [lp].callback) {
+				button_callback_list[lp].callback();
+			}
+		}
+	}
+}
+
+/* 
+ * This function is called when the button_timer times out.
+ * ie. When you don't press the button for bdelay jiffies, this is taken to
+ * mean you have ended the sequence of key presses, and this function is
+ * called to wind things up (write the press_count out to /dev/button, call
+ * any matching registered function callbacks, initiate reboot, etc.).
+ */
+
+static void button_sequence_finished(struct timer_list *unused)
+{
+	if (IS_ENABLED(CONFIG_NWBUTTON_REBOOT) &&
+	    button_press_count == reboot_count)
+		kill_cad_pid(SIGINT, 1);	/* Ask init to reboot us */
+	button_consume_callbacks (button_press_count);
+	bcount = sprintf (button_output_buffer, "%d\n", button_press_count);
+	button_press_count = 0;		/* Reset the button press counter */
+	wake_up_interruptible (&button_wait_queue);
+}
+
+/* 
+ *  This handler is called when the orange button is pressed (GPIO 10 of the
+ *  SuperIO chip, which maps to logical IRQ 26). If the press_count is 0,
+ *  this is the first press, so it starts a timer and increments the counter.
+ *  If it is higher than 0, it deletes the old timer, starts a new one, and
+ *  increments the counter.
+ */ 
+
+static irqreturn_t button_handler (int irq, void *dev_id)
+{
+	button_press_count++;
+	mod_timer(&button_timer, jiffies + bdelay);
+
+	return IRQ_HANDLED;
+}
+
+/*
+ * This function is called when a user space program attempts to read
+ * /dev/nwbutton. It puts the device to sleep on the wait queue until
+ * button_sequence_finished writes some data to the buffer and flushes
+ * the queue, at which point it writes the data out to the device and
+ * returns the number of characters it has written. This function is
+ * reentrant, so that many processes can be attempting to read from the
+ * device at any one time.
+ */
+
+static int button_read (struct file *filp, char __user *buffer,
+			size_t count, loff_t *ppos)
+{
+	DEFINE_WAIT(wait);
+	prepare_to_wait(&button_wait_queue, &wait, TASK_INTERRUPTIBLE);
+	schedule();
+	finish_wait(&button_wait_queue, &wait);
+	return (copy_to_user (buffer, &button_output_buffer, bcount))
+		 ? -EFAULT : bcount;
+}
+
+/* 
+ * This structure is the file operations structure, which specifies what
+ * callbacks functions the kernel should call when a user mode process
+ * attempts to perform these operations on the device.
+ */
+
+static const struct file_operations button_fops = {
+	.owner		= THIS_MODULE,
+	.read		= button_read,
+	.llseek		= noop_llseek,
+};
+
+/* 
+ * This structure is the misc device structure, which specifies the minor
+ * device number (158 in this case), the name of the device (for /proc/misc),
+ * and the address of the above file operations structure.
+ */
+
+static struct miscdevice button_misc_device = {
+	BUTTON_MINOR,
+	"nwbutton",
+	&button_fops,
+};
+
+/*
+ * This function is called to initialise the driver, either from misc.c at
+ * bootup if the driver is compiled into the kernel, or from init_module
+ * below at module insert time. It attempts to register the device node
+ * and the IRQ and fails with a warning message if either fails, though
+ * neither ever should because the device number and IRQ are unique to
+ * this driver.
+ */
+
+static int __init nwbutton_init(void)
+{
+	if (!machine_is_netwinder())
+		return -ENODEV;
+
+	printk (KERN_INFO "NetWinder Button Driver Version %s (C) Alex Holden "
+			"<alex@linuxhacker.org> 1998.\n", VERSION);
+
+	if (misc_register (&button_misc_device)) {
+		printk (KERN_WARNING "nwbutton: Couldn't register device 10, "
+				"%d.\n", BUTTON_MINOR);
+		return -EBUSY;
+	}
+
+	if (request_irq (IRQ_NETWINDER_BUTTON, button_handler, 0,
+			"nwbutton", NULL)) {
+		printk (KERN_WARNING "nwbutton: IRQ %d is not free.\n",
+				IRQ_NETWINDER_BUTTON);
+		misc_deregister (&button_misc_device);
+		return -EIO;
+	}
+	return 0;
+}
+
+static void __exit nwbutton_exit (void) 
+{
+	free_irq (IRQ_NETWINDER_BUTTON, NULL);
+	misc_deregister (&button_misc_device);
+}
+
+
+MODULE_AUTHOR("Alex Holden");
+MODULE_DESCRIPTION("NetWinder button driver");
+MODULE_LICENSE("GPL");
+
+module_init(nwbutton_init);
+module_exit(nwbutton_exit);