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, 371 insertions, 0 deletions

diff --git a/samples/configfs/Makefile b/samples/configfs/Makefile
new file mode 100644
index 000000000..92d661fcb
--- /dev/null
+++ b/samples/configfs/Makefile
@@ -0,0 +1,3 @@
+# SPDX-License-Identifier: GPL-2.0-only
+
+obj-$(CONFIG_SAMPLE_CONFIGFS) += configfs_sample.o
diff --git a/samples/configfs/configfs_sample.c b/samples/configfs/configfs_sample.c
new file mode 100644
index 000000000..c1b108ec4
--- /dev/null
+++ b/samples/configfs/configfs_sample.c
@@ -0,0 +1,368 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * configfs_example_macros.c - This file is a demonstration module
+ *      containing a number of configfs subsystems.  It uses the helper
+ *      macros defined by configfs.h
+ *
+ * Based on sysfs:
+ *      sysfs is Copyright (C) 2001, 2002, 2003 Patrick Mochel
+ *
+ * configfs Copyright (C) 2005 Oracle.  All rights reserved.
+ */
+
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/configfs.h>
+
+/*
+ * 01-childless
+ *
+ * This first example is a childless subsystem.  It cannot create
+ * any config_items.  It just has attributes.
+ *
+ * Note that we are enclosing the configfs_subsystem inside a container.
+ * This is not necessary if a subsystem has no attributes directly
+ * on the subsystem.  See the next example, 02-simple-children, for
+ * such a subsystem.
+ */
+
+struct childless {
+	struct configfs_subsystem subsys;
+	int showme;
+	int storeme;
+};
+
+static inline struct childless *to_childless(struct config_item *item)
+{
+	return container_of(to_configfs_subsystem(to_config_group(item)),
+			    struct childless, subsys);
+}
+
+static ssize_t childless_showme_show(struct config_item *item, char *page)
+{
+	struct childless *childless = to_childless(item);
+	ssize_t pos;
+
+	pos = sprintf(page, "%d\n", childless->showme);
+	childless->showme++;
+
+	return pos;
+}
+
+static ssize_t childless_storeme_show(struct config_item *item, char *page)
+{
+	return sprintf(page, "%d\n", to_childless(item)->storeme);
+}
+
+static ssize_t childless_storeme_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct childless *childless = to_childless(item);
+	int ret;
+
+	ret = kstrtoint(page, 10, &childless->storeme);
+	if (ret)
+		return ret;
+
+	return count;
+}
+
+static ssize_t childless_description_show(struct config_item *item, char *page)
+{
+	return sprintf(page,
+"[01-childless]\n"
+"\n"
+"The childless subsystem is the simplest possible subsystem in\n"
+"configfs.  It does not support the creation of child config_items.\n"
+"It only has a few attributes.  In fact, it isn't much different\n"
+"than a directory in /proc.\n");
+}
+
+CONFIGFS_ATTR_RO(childless_, showme);
+CONFIGFS_ATTR(childless_, storeme);
+CONFIGFS_ATTR_RO(childless_, description);
+
+static struct configfs_attribute *childless_attrs[] = {
+	&childless_attr_showme,
+	&childless_attr_storeme,
+	&childless_attr_description,
+	NULL,
+};
+
+static const struct config_item_type childless_type = {
+	.ct_attrs	= childless_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+static struct childless childless_subsys = {
+	.subsys = {
+		.su_group = {
+			.cg_item = {
+				.ci_namebuf = "01-childless",
+				.ci_type = &childless_type,
+			},
+		},
+	},
+};
+
+/* ----------------------------------------------------------------- */
+
+/*
+ * 02-simple-children
+ *
+ * This example merely has a simple one-attribute child.  Note that
+ * there is no extra attribute structure, as the child's attribute is
+ * known from the get-go.  Also, there is no container for the
+ * subsystem, as it has no attributes of its own.
+ */
+
+struct simple_child {
+	struct config_item item;
+	int storeme;
+};
+
+static inline struct simple_child *to_simple_child(struct config_item *item)
+{
+	return container_of(item, struct simple_child, item);
+}
+
+static ssize_t simple_child_storeme_show(struct config_item *item, char *page)
+{
+	return sprintf(page, "%d\n", to_simple_child(item)->storeme);
+}
+
+static ssize_t simple_child_storeme_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct simple_child *simple_child = to_simple_child(item);
+	int ret;
+
+	ret = kstrtoint(page, 10, &simple_child->storeme);
+	if (ret)
+		return ret;
+
+	return count;
+}
+
+CONFIGFS_ATTR(simple_child_, storeme);
+
+static struct configfs_attribute *simple_child_attrs[] = {
+	&simple_child_attr_storeme,
+	NULL,
+};
+
+static void simple_child_release(struct config_item *item)
+{
+	kfree(to_simple_child(item));
+}
+
+static const struct configfs_item_operations simple_child_item_ops = {
+	.release	= simple_child_release,
+};
+
+static const struct config_item_type simple_child_type = {
+	.ct_item_ops	= &simple_child_item_ops,
+	.ct_attrs	= simple_child_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+struct simple_children {
+	struct config_group group;
+};
+
+static inline struct simple_children *to_simple_children(struct config_item *item)
+{
+	return container_of(to_config_group(item),
+			    struct simple_children, group);
+}
+
+static struct config_item *simple_children_make_item(struct config_group *group,
+		const char *name)
+{
+	struct simple_child *simple_child;
+
+	simple_child = kzalloc(sizeof(struct simple_child), GFP_KERNEL);
+	if (!simple_child)
+		return ERR_PTR(-ENOMEM);
+
+	config_item_init_type_name(&simple_child->item, name,
+				   &simple_child_type);
+
+	return &simple_child->item;
+}
+
+static ssize_t simple_children_description_show(struct config_item *item,
+		char *page)
+{
+	return sprintf(page,
+"[02-simple-children]\n"
+"\n"
+"This subsystem allows the creation of child config_items.  These\n"
+"items have only one attribute that is readable and writeable.\n");
+}
+
+CONFIGFS_ATTR_RO(simple_children_, description);
+
+static struct configfs_attribute *simple_children_attrs[] = {
+	&simple_children_attr_description,
+	NULL,
+};
+
+static void simple_children_release(struct config_item *item)
+{
+	kfree(to_simple_children(item));
+}
+
+static const struct configfs_item_operations simple_children_item_ops = {
+	.release	= simple_children_release,
+};
+
+/*
+ * Note that, since no extra work is required on ->drop_item(),
+ * no ->drop_item() is provided.
+ */
+static const struct configfs_group_operations simple_children_group_ops = {
+	.make_item	= simple_children_make_item,
+};
+
+static const struct config_item_type simple_children_type = {
+	.ct_item_ops	= &simple_children_item_ops,
+	.ct_group_ops	= &simple_children_group_ops,
+	.ct_attrs	= simple_children_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+static struct configfs_subsystem simple_children_subsys = {
+	.su_group = {
+		.cg_item = {
+			.ci_namebuf = "02-simple-children",
+			.ci_type = &simple_children_type,
+		},
+	},
+};
+
+/* ----------------------------------------------------------------- */
+
+/*
+ * 03-group-children
+ *
+ * This example reuses the simple_children group from above.  However,
+ * the simple_children group is not the subsystem itself, it is a
+ * child of the subsystem.  Creation of a group in the subsystem creates
+ * a new simple_children group.  That group can then have simple_child
+ * children of its own.
+ */
+
+static struct config_group *group_children_make_group(
+		struct config_group *group, const char *name)
+{
+	struct simple_children *simple_children;
+
+	simple_children = kzalloc(sizeof(struct simple_children),
+				  GFP_KERNEL);
+	if (!simple_children)
+		return ERR_PTR(-ENOMEM);
+
+	config_group_init_type_name(&simple_children->group, name,
+				    &simple_children_type);
+
+	return &simple_children->group;
+}
+
+static ssize_t group_children_description_show(struct config_item *item,
+		char *page)
+{
+	return sprintf(page,
+"[03-group-children]\n"
+"\n"
+"This subsystem allows the creation of child config_groups.  These\n"
+"groups are like the subsystem simple-children.\n");
+}
+
+CONFIGFS_ATTR_RO(group_children_, description);
+
+static struct configfs_attribute *group_children_attrs[] = {
+	&group_children_attr_description,
+	NULL,
+};
+
+/*
+ * Note that, since no extra work is required on ->drop_item(),
+ * no ->drop_item() is provided.
+ */
+static const struct configfs_group_operations group_children_group_ops = {
+	.make_group	= group_children_make_group,
+};
+
+static const struct config_item_type group_children_type = {
+	.ct_group_ops	= &group_children_group_ops,
+	.ct_attrs	= group_children_attrs,
+	.ct_owner	= THIS_MODULE,
+};
+
+static struct configfs_subsystem group_children_subsys = {
+	.su_group = {
+		.cg_item = {
+			.ci_namebuf = "03-group-children",
+			.ci_type = &group_children_type,
+		},
+	},
+};
+
+/* ----------------------------------------------------------------- */
+
+/*
+ * We're now done with our subsystem definitions.
+ * For convenience in this module, here's a list of them all.  It
+ * allows the init function to easily register them.  Most modules
+ * will only have one subsystem, and will only call register_subsystem
+ * on it directly.
+ */
+static struct configfs_subsystem *example_subsys[] = {
+	&childless_subsys.subsys,
+	&simple_children_subsys,
+	&group_children_subsys,
+	NULL,
+};
+
+static int __init configfs_example_init(void)
+{
+	struct configfs_subsystem *subsys;
+	int ret, i;
+
+	for (i = 0; example_subsys[i]; i++) {
+		subsys = example_subsys[i];
+
+		config_group_init(&subsys->su_group);
+		mutex_init(&subsys->su_mutex);
+		ret = configfs_register_subsystem(subsys);
+		if (ret) {
+			pr_err("Error %d while registering subsystem %s\n",
+			       ret, subsys->su_group.cg_item.ci_namebuf);
+			goto out_unregister;
+		}
+	}
+
+	return 0;
+
+out_unregister:
+	for (i--; i >= 0; i--)
+		configfs_unregister_subsystem(example_subsys[i]);
+
+	return ret;
+}
+
+static void __exit configfs_example_exit(void)
+{
+	int i;
+
+	for (i = 0; example_subsys[i]; i++)
+		configfs_unregister_subsystem(example_subsys[i]);
+}
+
+module_init(configfs_example_init);
+module_exit(configfs_example_exit);
+MODULE_DESCRIPTION("Sample configfs module");
+MODULE_LICENSE("GPL");