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, 414 insertions, 0 deletions

diff --git a/drivers/acpi/glue.c b/drivers/acpi/glue.c
new file mode 100644
index 000000000..b17768092
--- /dev/null
+++ b/drivers/acpi/glue.c
@@ -0,0 +1,414 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Link physical devices with ACPI devices support
+ *
+ * Copyright (c) 2005 David Shaohua Li <shaohua.li@intel.com>
+ * Copyright (c) 2005 Intel Corp.
+ */
+
+#define pr_fmt(fmt) "ACPI: " fmt
+
+#include <linux/acpi_iort.h>
+#include <linux/export.h>
+#include <linux/init.h>
+#include <linux/list.h>
+#include <linux/device.h>
+#include <linux/slab.h>
+#include <linux/rwsem.h>
+#include <linux/acpi.h>
+#include <linux/dma-mapping.h>
+#include <linux/pci.h>
+#include <linux/pci-acpi.h>
+#include <linux/platform_device.h>
+
+#include "internal.h"
+
+static LIST_HEAD(bus_type_list);
+static DECLARE_RWSEM(bus_type_sem);
+
+#define PHYSICAL_NODE_STRING "physical_node"
+#define PHYSICAL_NODE_NAME_SIZE (sizeof(PHYSICAL_NODE_STRING) + 10)
+
+int register_acpi_bus_type(struct acpi_bus_type *type)
+{
+	if (acpi_disabled)
+		return -ENODEV;
+	if (type && type->match && type->find_companion) {
+		down_write(&bus_type_sem);
+		list_add_tail(&type->list, &bus_type_list);
+		up_write(&bus_type_sem);
+		pr_info("bus type %s registered\n", type->name);
+		return 0;
+	}
+	return -ENODEV;
+}
+EXPORT_SYMBOL_GPL(register_acpi_bus_type);
+
+int unregister_acpi_bus_type(struct acpi_bus_type *type)
+{
+	if (acpi_disabled)
+		return 0;
+	if (type) {
+		down_write(&bus_type_sem);
+		list_del_init(&type->list);
+		up_write(&bus_type_sem);
+		pr_info("bus type %s unregistered\n", type->name);
+		return 0;
+	}
+	return -ENODEV;
+}
+EXPORT_SYMBOL_GPL(unregister_acpi_bus_type);
+
+static struct acpi_bus_type *acpi_get_bus_type(struct device *dev)
+{
+	struct acpi_bus_type *tmp, *ret = NULL;
+
+	down_read(&bus_type_sem);
+	list_for_each_entry(tmp, &bus_type_list, list) {
+		if (tmp->match(dev)) {
+			ret = tmp;
+			break;
+		}
+	}
+	up_read(&bus_type_sem);
+	return ret;
+}
+
+#define FIND_CHILD_MIN_SCORE	1
+#define FIND_CHILD_MID_SCORE	2
+#define FIND_CHILD_MAX_SCORE	3
+
+static int match_any(struct acpi_device *adev, void *not_used)
+{
+	return 1;
+}
+
+static bool acpi_dev_has_children(struct acpi_device *adev)
+{
+	return acpi_dev_for_each_child(adev, match_any, NULL) > 0;
+}
+
+static int find_child_checks(struct acpi_device *adev, bool check_children)
+{
+	unsigned long long sta;
+	acpi_status status;
+
+	if (check_children && !acpi_dev_has_children(adev))
+		return -ENODEV;
+
+	status = acpi_evaluate_integer(adev->handle, "_STA", NULL, &sta);
+	if (status == AE_NOT_FOUND) {
+		/*
+		 * Special case: backlight device objects without _STA are
+		 * preferred to other objects with the same _ADR value, because
+		 * it is more likely that they are actually useful.
+		 */
+		if (adev->pnp.type.backlight)
+			return FIND_CHILD_MID_SCORE;
+
+		return FIND_CHILD_MIN_SCORE;
+	}
+
+	if (ACPI_FAILURE(status) || !(sta & ACPI_STA_DEVICE_ENABLED))
+		return -ENODEV;
+
+	/*
+	 * If the device has a _HID returning a valid ACPI/PNP device ID, it is
+	 * better to make it look less attractive here, so that the other device
+	 * with the same _ADR value (that may not have a valid device ID) can be
+	 * matched going forward.  [This means a second spec violation in a row,
+	 * so whatever we do here is best effort anyway.]
+	 */
+	if (adev->pnp.type.platform_id)
+		return FIND_CHILD_MIN_SCORE;
+
+	return FIND_CHILD_MAX_SCORE;
+}
+
+struct find_child_walk_data {
+	struct acpi_device *adev;
+	u64 address;
+	int score;
+	bool check_sta;
+	bool check_children;
+};
+
+static int check_one_child(struct acpi_device *adev, void *data)
+{
+	struct find_child_walk_data *wd = data;
+	int score;
+
+	if (!adev->pnp.type.bus_address || acpi_device_adr(adev) != wd->address)
+		return 0;
+
+	if (!wd->adev) {
+		/*
+		 * This is the first matching object, so save it.  If it is not
+		 * necessary to look for any other matching objects, stop the
+		 * search.
+		 */
+		wd->adev = adev;
+		return !(wd->check_sta || wd->check_children);
+	}
+
+	/*
+	 * There is more than one matching device object with the same _ADR
+	 * value.  That really is unexpected, so we are kind of beyond the scope
+	 * of the spec here.  We have to choose which one to return, though.
+	 *
+	 * First, get the score for the previously found object and terminate
+	 * the walk if it is maximum.
+	*/
+	if (!wd->score) {
+		score = find_child_checks(wd->adev, wd->check_children);
+		if (score == FIND_CHILD_MAX_SCORE)
+			return 1;
+
+		wd->score = score;
+	}
+	/*
+	 * Second, if the object that has just been found has a better score,
+	 * replace the previously found one with it and terminate the walk if
+	 * the new score is maximum.
+	 */
+	score = find_child_checks(adev, wd->check_children);
+	if (score > wd->score) {
+		wd->adev = adev;
+		if (score == FIND_CHILD_MAX_SCORE)
+			return 1;
+
+		wd->score = score;
+	}
+
+	/* Continue, because there may be better matches. */
+	return 0;
+}
+
+static struct acpi_device *acpi_find_child(struct acpi_device *parent,
+					   u64 address, bool check_children,
+					   bool check_sta)
+{
+	struct find_child_walk_data wd = {
+		.address = address,
+		.check_children = check_children,
+		.check_sta = check_sta,
+		.adev = NULL,
+		.score = 0,
+	};
+
+	if (parent)
+		acpi_dev_for_each_child(parent, check_one_child, &wd);
+
+	return wd.adev;
+}
+
+struct acpi_device *acpi_find_child_device(struct acpi_device *parent,
+					   u64 address, bool check_children)
+{
+	return acpi_find_child(parent, address, check_children, true);
+}
+EXPORT_SYMBOL_GPL(acpi_find_child_device);
+
+struct acpi_device *acpi_find_child_by_adr(struct acpi_device *adev,
+					   acpi_bus_address adr)
+{
+	return acpi_find_child(adev, adr, false, false);
+}
+EXPORT_SYMBOL_GPL(acpi_find_child_by_adr);
+
+static void acpi_physnode_link_name(char *buf, unsigned int node_id)
+{
+	if (node_id > 0)
+		snprintf(buf, PHYSICAL_NODE_NAME_SIZE,
+			 PHYSICAL_NODE_STRING "%u", node_id);
+	else
+		strcpy(buf, PHYSICAL_NODE_STRING);
+}
+
+int acpi_bind_one(struct device *dev, struct acpi_device *acpi_dev)
+{
+	struct acpi_device_physical_node *physical_node, *pn;
+	char physical_node_name[PHYSICAL_NODE_NAME_SIZE];
+	struct list_head *physnode_list;
+	unsigned int node_id;
+	int retval = -EINVAL;
+
+	if (has_acpi_companion(dev)) {
+		if (acpi_dev) {
+			dev_warn(dev, "ACPI companion already set\n");
+			return -EINVAL;
+		} else {
+			acpi_dev = ACPI_COMPANION(dev);
+		}
+	}
+	if (!acpi_dev)
+		return -EINVAL;
+
+	acpi_dev_get(acpi_dev);
+	get_device(dev);
+	physical_node = kzalloc_obj(*physical_node);
+	if (!physical_node) {
+		retval = -ENOMEM;
+		goto err;
+	}
+
+	mutex_lock(&acpi_dev->physical_node_lock);
+
+	/*
+	 * Keep the list sorted by node_id so that the IDs of removed nodes can
+	 * be recycled easily.
+	 */
+	physnode_list = &acpi_dev->physical_node_list;
+	node_id = 0;
+	list_for_each_entry(pn, &acpi_dev->physical_node_list, node) {
+		/* Sanity check. */
+		if (pn->dev == dev) {
+			mutex_unlock(&acpi_dev->physical_node_lock);
+
+			dev_warn(dev, "Already associated with ACPI node\n");
+			kfree(physical_node);
+			if (ACPI_COMPANION(dev) != acpi_dev)
+				goto err;
+
+			put_device(dev);
+			acpi_dev_put(acpi_dev);
+			return 0;
+		}
+		if (pn->node_id == node_id) {
+			physnode_list = &pn->node;
+			node_id++;
+		}
+	}
+
+	physical_node->node_id = node_id;
+	physical_node->dev = dev;
+	list_add(&physical_node->node, physnode_list);
+	acpi_dev->physical_node_count++;
+
+	if (!has_acpi_companion(dev))
+		ACPI_COMPANION_SET(dev, acpi_dev);
+
+	acpi_physnode_link_name(physical_node_name, node_id);
+	retval = sysfs_create_link(&acpi_dev->dev.kobj, &dev->kobj,
+				   physical_node_name);
+	if (retval)
+		dev_err(&acpi_dev->dev, "Failed to create link %s (%d)\n",
+			physical_node_name, retval);
+
+	retval = sysfs_create_link(&dev->kobj, &acpi_dev->dev.kobj,
+				   "firmware_node");
+	if (retval)
+		dev_err(dev, "Failed to create link firmware_node (%d)\n",
+			retval);
+
+	mutex_unlock(&acpi_dev->physical_node_lock);
+
+	if (acpi_dev->wakeup.flags.valid)
+		device_set_wakeup_capable(dev, true);
+
+	return 0;
+
+ err:
+	ACPI_COMPANION_SET(dev, NULL);
+	put_device(dev);
+	acpi_dev_put(acpi_dev);
+	return retval;
+}
+EXPORT_SYMBOL_GPL(acpi_bind_one);
+
+int acpi_unbind_one(struct device *dev)
+{
+	struct acpi_device *acpi_dev = ACPI_COMPANION(dev);
+	struct acpi_device_physical_node *entry;
+
+	if (!acpi_dev)
+		return 0;
+
+	mutex_lock(&acpi_dev->physical_node_lock);
+
+	list_for_each_entry(entry, &acpi_dev->physical_node_list, node)
+		if (entry->dev == dev) {
+			char physnode_name[PHYSICAL_NODE_NAME_SIZE];
+
+			list_del(&entry->node);
+			acpi_dev->physical_node_count--;
+
+			acpi_physnode_link_name(physnode_name, entry->node_id);
+			sysfs_remove_link(&acpi_dev->dev.kobj, physnode_name);
+			sysfs_remove_link(&dev->kobj, "firmware_node");
+			ACPI_COMPANION_SET(dev, NULL);
+			/* Drop references taken by acpi_bind_one(). */
+			put_device(dev);
+			acpi_dev_put(acpi_dev);
+			kfree(entry);
+			break;
+		}
+
+	mutex_unlock(&acpi_dev->physical_node_lock);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(acpi_unbind_one);
+
+void acpi_device_notify(struct device *dev)
+{
+	struct acpi_device *adev;
+	int ret;
+
+	ret = acpi_bind_one(dev, NULL);
+	if (ret) {
+		struct acpi_bus_type *type = acpi_get_bus_type(dev);
+
+		if (!type)
+			goto err;
+
+		adev = type->find_companion(dev);
+		if (!adev) {
+			dev_dbg(dev, "ACPI companion not found\n");
+			goto err;
+		}
+		ret = acpi_bind_one(dev, adev);
+		if (ret)
+			goto err;
+
+		if (type->setup) {
+			type->setup(dev);
+			goto done;
+		}
+	} else {
+		adev = ACPI_COMPANION(dev);
+
+		if (dev_is_pci(dev)) {
+			pci_acpi_setup(dev, adev);
+			goto done;
+		} else if (dev_is_platform(dev)) {
+			acpi_configure_pmsi_domain(dev);
+		}
+	}
+
+	if (adev->handler && adev->handler->bind)
+		adev->handler->bind(dev);
+
+done:
+	acpi_handle_debug(ACPI_HANDLE(dev), "Bound to device %s\n",
+			  dev_name(dev));
+
+	return;
+
+err:
+	dev_dbg(dev, "No ACPI support\n");
+}
+
+void acpi_device_notify_remove(struct device *dev)
+{
+	struct acpi_device *adev = ACPI_COMPANION(dev);
+
+	if (!adev)
+		return;
+
+	if (dev_is_pci(dev))
+		pci_acpi_cleanup(dev, adev);
+	else if (adev->handler && adev->handler->unbind)
+		adev->handler->unbind(dev);
+
+	acpi_unbind_one(dev);
+}