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

2 files changed, 157 insertions, 0 deletions

diff --git a/samples/fprobe/Makefile b/samples/fprobe/Makefile
new file mode 100644
index 000000000..ecccbfa6e
--- /dev/null
+++ b/samples/fprobe/Makefile
@@ -0,0 +1,3 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-$(CONFIG_SAMPLE_FPROBE) += fprobe_example.o
diff --git a/samples/fprobe/fprobe_example.c b/samples/fprobe/fprobe_example.c
new file mode 100644
index 000000000..bfe98ce82
--- /dev/null
+++ b/samples/fprobe/fprobe_example.c
@@ -0,0 +1,154 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Here's a sample kernel module showing the use of fprobe to dump a
+ * stack trace and selected registers when kernel_clone() is called.
+ *
+ * For more information on theory of operation of kprobes, see
+ * Documentation/trace/kprobes.rst
+ *
+ * You will see the trace data in /var/log/messages and on the console
+ * whenever kernel_clone() is invoked to create a new process.
+ */
+
+#define pr_fmt(fmt) "%s: " fmt, __func__
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/fprobe.h>
+#include <linux/sched/debug.h>
+#include <linux/slab.h>
+
+#define BACKTRACE_DEPTH 16
+#define MAX_SYMBOL_LEN 4096
+static struct fprobe sample_probe;
+static unsigned long nhit;
+
+static char symbol[MAX_SYMBOL_LEN] = "kernel_clone";
+module_param_string(symbol, symbol, sizeof(symbol), 0644);
+MODULE_PARM_DESC(symbol, "Probed symbol(s), given by comma separated symbols or a wildcard pattern.");
+
+static char nosymbol[MAX_SYMBOL_LEN] = "";
+module_param_string(nosymbol, nosymbol, sizeof(nosymbol), 0644);
+MODULE_PARM_DESC(nosymbol, "Not-probed symbols, given by a wildcard pattern.");
+
+static bool stackdump = true;
+module_param(stackdump, bool, 0644);
+MODULE_PARM_DESC(stackdump, "Enable stackdump.");
+
+static bool use_trace = false;
+module_param(use_trace, bool, 0644);
+MODULE_PARM_DESC(use_trace, "Use trace_printk instead of printk. This is only for debugging.");
+
+static void show_backtrace(void)
+{
+	unsigned long stacks[BACKTRACE_DEPTH];
+	unsigned int len;
+
+	len = stack_trace_save(stacks, BACKTRACE_DEPTH, 2);
+	stack_trace_print(stacks, len, 24);
+}
+
+static int sample_entry_handler(struct fprobe *fp, unsigned long ip,
+				unsigned long ret_ip,
+				struct ftrace_regs *fregs, void *data)
+{
+	if (use_trace)
+		/*
+		 * This is just an example, no kernel code should call
+		 * trace_printk() except when actively debugging.
+		 */
+		trace_printk("Enter <%pS> ip = 0x%p\n", (void *)ip, (void *)ip);
+	else
+		pr_info("Enter <%pS> ip = 0x%p\n", (void *)ip, (void *)ip);
+	nhit++;
+	if (stackdump)
+		show_backtrace();
+	return 0;
+}
+
+static void sample_exit_handler(struct fprobe *fp, unsigned long ip,
+				unsigned long ret_ip, struct ftrace_regs *regs,
+				void *data)
+{
+	unsigned long rip = ret_ip;
+
+	if (use_trace)
+		/*
+		 * This is just an example, no kernel code should call
+		 * trace_printk() except when actively debugging.
+		 */
+		trace_printk("Return from <%pS> ip = 0x%p to rip = 0x%p (%pS)\n",
+			(void *)ip, (void *)ip, (void *)rip, (void *)rip);
+	else
+		pr_info("Return from <%pS> ip = 0x%p to rip = 0x%p (%pS)\n",
+			(void *)ip, (void *)ip, (void *)rip, (void *)rip);
+	nhit++;
+	if (stackdump)
+		show_backtrace();
+}
+
+static int __init fprobe_init(void)
+{
+	char *p, *symbuf = NULL;
+	const char **syms;
+	int ret, count, i;
+
+	sample_probe.entry_handler = sample_entry_handler;
+	sample_probe.exit_handler = sample_exit_handler;
+
+	if (strchr(symbol, '*')) {
+		/* filter based fprobe */
+		ret = register_fprobe(&sample_probe, symbol,
+				      nosymbol[0] == '\0' ? NULL : nosymbol);
+		goto out;
+	} else if (!strchr(symbol, ',')) {
+		symbuf = symbol;
+		ret = register_fprobe_syms(&sample_probe, (const char **)&symbuf, 1);
+		goto out;
+	}
+
+	/* Comma separated symbols */
+	symbuf = kstrdup(symbol, GFP_KERNEL);
+	if (!symbuf)
+		return -ENOMEM;
+	p = symbuf;
+	count = 1;
+	while ((p = strchr(++p, ',')) != NULL)
+		count++;
+
+	pr_info("%d symbols found\n", count);
+
+	syms = kcalloc(count, sizeof(char *), GFP_KERNEL);
+	if (!syms) {
+		kfree(symbuf);
+		return -ENOMEM;
+	}
+
+	p = symbuf;
+	for (i = 0; i < count; i++)
+		syms[i] = strsep(&p, ",");
+
+	ret = register_fprobe_syms(&sample_probe, syms, count);
+	kfree(syms);
+	kfree(symbuf);
+out:
+	if (ret < 0)
+		pr_err("register_fprobe failed, returned %d\n", ret);
+	else
+		pr_info("Planted fprobe at %s\n", symbol);
+
+	return ret;
+}
+
+static void __exit fprobe_exit(void)
+{
+	unregister_fprobe(&sample_probe);
+
+	pr_info("fprobe at %s unregistered. %ld times hit, %ld times missed\n",
+		symbol, nhit, sample_probe.nmissed);
+}
+
+module_init(fprobe_init)
+module_exit(fprobe_exit)
+MODULE_DESCRIPTION("sample kernel module showing the use of fprobe");
+MODULE_LICENSE("GPL");