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, 179 insertions, 0 deletions

diff --git a/drivers/char/hangcheck-timer.c b/drivers/char/hangcheck-timer.c
new file mode 100644
index 000000000..231cbf7b3
--- /dev/null
+++ b/drivers/char/hangcheck-timer.c
@@ -0,0 +1,179 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * hangcheck-timer.c
+ *
+ * Driver for a little io fencing timer.
+ *
+ * Copyright (C) 2002, 2003 Oracle.  All rights reserved.
+ *
+ * Author: Joel Becker <joel.becker@oracle.com>
+ */
+
+/*
+ * The hangcheck-timer driver uses the TSC to catch delays that
+ * jiffies does not notice.  A timer is set.  When the timer fires, it
+ * checks whether it was delayed and if that delay exceeds a given
+ * margin of error.  The hangcheck_tick module parameter takes the timer
+ * duration in seconds.  The hangcheck_margin parameter defines the
+ * margin of error, in seconds.  The defaults are 60 seconds for the
+ * timer and 180 seconds for the margin of error.  IOW, a timer is set
+ * for 60 seconds.  When the timer fires, the callback checks the
+ * actual duration that the timer waited.  If the duration exceeds the
+ * allotted time and margin (here 60 + 180, or 240 seconds), the machine
+ * is restarted.  A healthy machine will have the duration match the
+ * expected timeout very closely.
+ */
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/fs.h>
+#include <linux/mm.h>
+#include <linux/reboot.h>
+#include <linux/init.h>
+#include <linux/delay.h>
+#include <linux/uaccess.h>
+#include <linux/sysrq.h>
+#include <linux/timer.h>
+#include <linux/hrtimer.h>
+
+#define VERSION_STR "0.9.1"
+
+#define DEFAULT_IOFENCE_MARGIN 60	/* Default fudge factor, in seconds */
+#define DEFAULT_IOFENCE_TICK 180	/* Default timer timeout, in seconds */
+
+static int hangcheck_tick = DEFAULT_IOFENCE_TICK;
+static int hangcheck_margin = DEFAULT_IOFENCE_MARGIN;
+static int hangcheck_reboot;  /* Defaults to not reboot */
+static int hangcheck_dump_tasks;  /* Defaults to not dumping SysRQ T */
+
+/* options - modular */
+module_param(hangcheck_tick, int, 0);
+MODULE_PARM_DESC(hangcheck_tick, "Timer delay.");
+module_param(hangcheck_margin, int, 0);
+MODULE_PARM_DESC(hangcheck_margin, "If the hangcheck timer has been delayed more than hangcheck_margin seconds, the driver will fire.");
+module_param(hangcheck_reboot, int, 0);
+MODULE_PARM_DESC(hangcheck_reboot, "If nonzero, the machine will reboot when the timer margin is exceeded.");
+module_param(hangcheck_dump_tasks, int, 0);
+MODULE_PARM_DESC(hangcheck_dump_tasks, "If nonzero, the machine will dump the system task state when the timer margin is exceeded.");
+
+MODULE_AUTHOR("Oracle");
+MODULE_DESCRIPTION("Hangcheck-timer detects when the system has gone out to lunch past a certain margin.");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(VERSION_STR);
+
+/* options - nonmodular */
+#ifndef MODULE
+
+static int __init hangcheck_parse_tick(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_tick = par;
+	return 1;
+}
+
+static int __init hangcheck_parse_margin(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_margin = par;
+	return 1;
+}
+
+static int __init hangcheck_parse_reboot(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_reboot = par;
+	return 1;
+}
+
+static int __init hangcheck_parse_dump_tasks(char *str)
+{
+	int par;
+
+	if (get_option(&str, &par))
+		hangcheck_dump_tasks = par;
+	return 1;
+}
+
+__setup("hcheck_tick", hangcheck_parse_tick);
+__setup("hcheck_margin", hangcheck_parse_margin);
+__setup("hcheck_reboot", hangcheck_parse_reboot);
+__setup("hcheck_dump_tasks", hangcheck_parse_dump_tasks);
+#endif /* not MODULE */
+
+#define TIMER_FREQ 1000000000ULL
+
+/* Last time scheduled */
+static unsigned long long hangcheck_tsc, hangcheck_tsc_margin;
+
+static void hangcheck_fire(struct timer_list *);
+
+static DEFINE_TIMER(hangcheck_ticktock, hangcheck_fire);
+
+static void hangcheck_fire(struct timer_list *unused)
+{
+	unsigned long long cur_tsc, tsc_diff;
+
+	cur_tsc = ktime_get_ns();
+
+	if (cur_tsc > hangcheck_tsc)
+		tsc_diff = cur_tsc - hangcheck_tsc;
+	else
+		tsc_diff = (cur_tsc + (~0ULL - hangcheck_tsc)); /* or something */
+
+	if (tsc_diff > hangcheck_tsc_margin) {
+		if (hangcheck_dump_tasks) {
+			pr_crit("Hangcheck: Task state:\n");
+#ifdef CONFIG_MAGIC_SYSRQ
+			handle_sysrq('t');
+#endif  /* CONFIG_MAGIC_SYSRQ */
+		}
+		if (hangcheck_reboot) {
+			pr_crit("Hangcheck: hangcheck is restarting the machine.\n");
+			emergency_restart();
+		} else {
+			pr_crit("Hangcheck: hangcheck value past margin!\n");
+		}
+	}
+#if 0
+	/*
+	 * Enable to investigate delays in detail
+	 */
+	pr_debug("Hangcheck: called %lld ns since last time (%lld ns overshoot)\n",
+			tsc_diff, tsc_diff - hangcheck_tick*TIMER_FREQ);
+#endif
+	mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ));
+	hangcheck_tsc = ktime_get_ns();
+}
+
+
+static int __init hangcheck_init(void)
+{
+	pr_debug("Hangcheck: starting hangcheck timer %s (tick is %d seconds, margin is %d seconds).\n",
+	       VERSION_STR, hangcheck_tick, hangcheck_margin);
+	hangcheck_tsc_margin =
+		(unsigned long long)hangcheck_margin + hangcheck_tick;
+	hangcheck_tsc_margin *= TIMER_FREQ;
+
+	hangcheck_tsc = ktime_get_ns();
+	mod_timer(&hangcheck_ticktock, jiffies + (hangcheck_tick*HZ));
+
+	return 0;
+}
+
+
+static void __exit hangcheck_exit(void)
+{
+	timer_delete_sync(&hangcheck_ticktock);
+	pr_debug("Hangcheck: Stopped hangcheck timer.\n");
+}
+
+module_init(hangcheck_init);
+module_exit(hangcheck_exit);