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, 212 insertions, 0 deletions

diff --git a/samples/livepatch/livepatch-shadow-mod.c b/samples/livepatch/livepatch-shadow-mod.c
new file mode 100644
index 000000000..5d83ad5a8
--- /dev/null
+++ b/samples/livepatch/livepatch-shadow-mod.c
@@ -0,0 +1,212 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2017 Joe Lawrence <joe.lawrence@redhat.com>
+ */
+
+/*
+ * livepatch-shadow-mod.c - Shadow variables, buggy module demo
+ *
+ * Purpose
+ * -------
+ *
+ * As a demonstration of livepatch shadow variable API, this module
+ * introduces memory leak behavior that livepatch modules
+ * livepatch-shadow-fix1.ko and livepatch-shadow-fix2.ko correct and
+ * enhance.
+ *
+ * WARNING - even though the livepatch-shadow-fix modules patch the
+ * memory leak, please load these modules at your own risk -- some
+ * amount of memory may leaked before the bug is patched.
+ *
+ *
+ * Usage
+ * -----
+ *
+ * Step 1 - Load the buggy demonstration module:
+ *
+ *   insmod samples/livepatch/livepatch-shadow-mod.ko
+ *
+ * Watch dmesg output for a few moments to see new dummy being allocated
+ * and a periodic cleanup check.  (Note: a small amount of memory is
+ * being leaked.)
+ *
+ *
+ * Step 2 - Load livepatch fix1:
+ *
+ *   insmod samples/livepatch/livepatch-shadow-fix1.ko
+ *
+ * Continue watching dmesg and note that now livepatch_fix1_dummy_free()
+ * and livepatch_fix1_dummy_alloc() are logging messages about leaked
+ * memory and eventually leaks prevented.
+ *
+ *
+ * Step 3 - Load livepatch fix2 (on top of fix1):
+ *
+ *   insmod samples/livepatch/livepatch-shadow-fix2.ko
+ *
+ * This module extends functionality through shadow variables, as a new
+ * "check" counter is added to the dummy structure.  Periodic dmesg
+ * messages will log these as dummies are cleaned up.
+ *
+ *
+ * Step 4 - Cleanup
+ *
+ * Unwind the demonstration by disabling the livepatch fix modules, then
+ * removing them and the demo module:
+ *
+ *   echo 0 > /sys/kernel/livepatch/livepatch_shadow_fix2/enabled
+ *   echo 0 > /sys/kernel/livepatch/livepatch_shadow_fix1/enabled
+ *   rmmod livepatch-shadow-fix2
+ *   rmmod livepatch-shadow-fix1
+ *   rmmod livepatch-shadow-mod
+ */
+
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/stat.h>
+#include <linux/workqueue.h>
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Joe Lawrence <joe.lawrence@redhat.com>");
+MODULE_DESCRIPTION("Buggy module for shadow variable demo");
+
+/* Allocate new dummies every second */
+#define ALLOC_PERIOD	1
+/* Check for expired dummies after a few new ones have been allocated */
+#define CLEANUP_PERIOD	(3 * ALLOC_PERIOD)
+/* Dummies expire after a few cleanup instances */
+#define EXPIRE_PERIOD	(4 * CLEANUP_PERIOD)
+
+/*
+ * Keep a list of all the dummies so we can clean up any residual ones
+ * on module exit
+ */
+static LIST_HEAD(dummy_list);
+static DEFINE_MUTEX(dummy_list_mutex);
+
+struct dummy {
+	struct list_head list;
+	unsigned long jiffies_expire;
+};
+
+static __used noinline struct dummy *dummy_alloc(void)
+{
+	struct dummy *d;
+	int *leak;
+
+	d = kzalloc(sizeof(*d), GFP_KERNEL);
+	if (!d)
+		return NULL;
+
+	d->jiffies_expire = jiffies + secs_to_jiffies(EXPIRE_PERIOD);
+
+	/* Oops, forgot to save leak! */
+	leak = kzalloc(sizeof(*leak), GFP_KERNEL);
+	if (!leak) {
+		kfree(d);
+		return NULL;
+	}
+
+	pr_info("%s: dummy @ %p, expires @ %lx\n",
+		__func__, d, d->jiffies_expire);
+
+	return d;
+}
+
+static __used noinline void dummy_free(struct dummy *d)
+{
+	pr_info("%s: dummy @ %p, expired = %lx\n",
+		__func__, d, d->jiffies_expire);
+
+	kfree(d);
+}
+
+static __used noinline bool dummy_check(struct dummy *d,
+					   unsigned long jiffies)
+{
+	return time_after(jiffies, d->jiffies_expire);
+}
+
+/*
+ * alloc_work_func: allocates new dummy structures, allocates additional
+ *                  memory, aptly named "leak", but doesn't keep
+ *                  permanent record of it.
+ */
+
+static void alloc_work_func(struct work_struct *work);
+static DECLARE_DELAYED_WORK(alloc_dwork, alloc_work_func);
+
+static void alloc_work_func(struct work_struct *work)
+{
+	struct dummy *d;
+
+	d = dummy_alloc();
+	if (!d)
+		return;
+
+	mutex_lock(&dummy_list_mutex);
+	list_add(&d->list, &dummy_list);
+	mutex_unlock(&dummy_list_mutex);
+
+	schedule_delayed_work(&alloc_dwork, secs_to_jiffies(ALLOC_PERIOD));
+}
+
+/*
+ * cleanup_work_func: frees dummy structures.  Without knownledge of
+ *                    "leak", it leaks the additional memory that
+ *                    alloc_work_func created.
+ */
+
+static void cleanup_work_func(struct work_struct *work);
+static DECLARE_DELAYED_WORK(cleanup_dwork, cleanup_work_func);
+
+static void cleanup_work_func(struct work_struct *work)
+{
+	struct dummy *d, *tmp;
+	unsigned long j;
+
+	j = jiffies;
+	pr_info("%s: jiffies = %lx\n", __func__, j);
+
+	mutex_lock(&dummy_list_mutex);
+	list_for_each_entry_safe(d, tmp, &dummy_list, list) {
+
+		/* Kick out and free any expired dummies */
+		if (dummy_check(d, j)) {
+			list_del(&d->list);
+			dummy_free(d);
+		}
+	}
+	mutex_unlock(&dummy_list_mutex);
+
+	schedule_delayed_work(&cleanup_dwork, secs_to_jiffies(CLEANUP_PERIOD));
+}
+
+static int livepatch_shadow_mod_init(void)
+{
+	schedule_delayed_work(&alloc_dwork, secs_to_jiffies(ALLOC_PERIOD));
+	schedule_delayed_work(&cleanup_dwork, secs_to_jiffies(CLEANUP_PERIOD));
+
+	return 0;
+}
+
+static void livepatch_shadow_mod_exit(void)
+{
+	struct dummy *d, *tmp;
+
+	/* Wait for any dummies at work */
+	cancel_delayed_work_sync(&alloc_dwork);
+	cancel_delayed_work_sync(&cleanup_dwork);
+
+	/* Cleanup residual dummies */
+	list_for_each_entry_safe(d, tmp, &dummy_list, list) {
+		list_del(&d->list);
+		dummy_free(d);
+	}
+}
+
+module_init(livepatch_shadow_mod_init);
+module_exit(livepatch_shadow_mod_exit);