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

6 files changed, 408 insertions, 0 deletions

diff --git a/drivers/misc/pvpanic/Kconfig b/drivers/misc/pvpanic/Kconfig
new file mode 100644
index 000000000..12d40a21f
--- /dev/null
+++ b/drivers/misc/pvpanic/Kconfig
@@ -0,0 +1,27 @@
+# SPDX-License-Identifier: GPL-2.0+
+#
+# Pvpanic Kconfig
+#
+# Copyright (C) 2021 Oracle.
+#
+
+config PVPANIC
+	bool "pvpanic device support"
+	help
+	  This option allows to select a specific pvpanic device driver.
+	  pvpanic is a paravirtualized device provided by QEMU; it lets
+	  a virtual machine (guest) communicate panic events to the host.
+
+config PVPANIC_MMIO
+	tristate "pvpanic MMIO device support"
+	depends on HAS_IOMEM && (ACPI || OF) && PVPANIC
+	help
+	  This driver provides support for the MMIO pvpanic device.
+
+config PVPANIC_PCI
+	tristate "pvpanic PCI device support"
+	depends on PCI && PVPANIC
+	help
+	  This driver provides support for the PCI pvpanic device.
+	  pvpanic is a paravirtualized device provided by QEMU which
+	  forwards the panic events from the guest to the host.
diff --git a/drivers/misc/pvpanic/Makefile b/drivers/misc/pvpanic/Makefile
new file mode 100644
index 000000000..9471df7d4
--- /dev/null
+++ b/drivers/misc/pvpanic/Makefile
@@ -0,0 +1,8 @@
+# SPDX-License-Identifier: GPL-2.0+
+#
+# Pvpanic Makefile
+#
+# Copyright (C) 2021 Oracle.
+#
+obj-$(CONFIG_PVPANIC_MMIO)	+= pvpanic.o pvpanic-mmio.o
+obj-$(CONFIG_PVPANIC_PCI)	+= pvpanic.o pvpanic-pci.o
diff --git a/drivers/misc/pvpanic/pvpanic-mmio.c b/drivers/misc/pvpanic/pvpanic-mmio.c
new file mode 100644
index 000000000..724dfddf6
--- /dev/null
+++ b/drivers/misc/pvpanic/pvpanic-mmio.c
@@ -0,0 +1,74 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ *  Pvpanic MMIO Device Support
+ *
+ *  Copyright (C) 2013 Fujitsu.
+ *  Copyright (C) 2018 ZTE.
+ *  Copyright (C) 2021 Oracle.
+ */
+
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+#include <linux/kexec.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/types.h>
+
+#include "pvpanic.h"
+
+MODULE_AUTHOR("Hu Tao <hutao@cn.fujitsu.com>");
+MODULE_DESCRIPTION("pvpanic-mmio device driver");
+MODULE_LICENSE("GPL");
+
+static int pvpanic_mmio_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct resource *res;
+	void __iomem *base;
+
+	res = platform_get_mem_or_io(pdev, 0);
+	if (!res)
+		return -EINVAL;
+
+	switch (resource_type(res)) {
+	case IORESOURCE_IO:
+		base = devm_ioport_map(dev, res->start, resource_size(res));
+		if (!base)
+			return -ENOMEM;
+		break;
+	case IORESOURCE_MEM:
+		base = devm_ioremap_resource(dev, res);
+		if (IS_ERR(base))
+			return PTR_ERR(base);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return devm_pvpanic_probe(dev, base);
+}
+
+static const struct of_device_id pvpanic_mmio_match[] = {
+	{ .compatible = "qemu,pvpanic-mmio" },
+	{ }
+};
+MODULE_DEVICE_TABLE(of, pvpanic_mmio_match);
+
+static const struct acpi_device_id pvpanic_device_ids[] = {
+	{ .id = "QEMU0001" },
+	{ }
+};
+MODULE_DEVICE_TABLE(acpi, pvpanic_device_ids);
+
+static struct platform_driver pvpanic_mmio_driver = {
+	.driver = {
+		.name = "pvpanic-mmio",
+		.of_match_table = pvpanic_mmio_match,
+		.acpi_match_table = pvpanic_device_ids,
+		.dev_groups = pvpanic_dev_groups,
+	},
+	.probe = pvpanic_mmio_probe,
+};
+module_platform_driver(pvpanic_mmio_driver);
diff --git a/drivers/misc/pvpanic/pvpanic-pci.c b/drivers/misc/pvpanic/pvpanic-pci.c
new file mode 100644
index 000000000..572c6148a
--- /dev/null
+++ b/drivers/misc/pvpanic/pvpanic-pci.c
@@ -0,0 +1,49 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ *  Pvpanic PCI Device Support
+ *
+ *  Copyright (C) 2021 Oracle.
+ */
+
+#include <linux/errno.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/types.h>
+
+#include "pvpanic.h"
+
+#define PCI_DEVICE_ID_REDHAT_PVPANIC     0x0011
+
+MODULE_AUTHOR("Mihai Carabas <mihai.carabas@oracle.com>");
+MODULE_DESCRIPTION("pvpanic device driver");
+MODULE_LICENSE("GPL");
+
+static int pvpanic_pci_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	void __iomem *base;
+	int ret;
+
+	ret = pcim_enable_device(pdev);
+	if (ret < 0)
+		return ret;
+
+	base = pcim_iomap(pdev, 0, 0);
+	if (!base)
+		return -ENOMEM;
+
+	return devm_pvpanic_probe(&pdev->dev, base);
+}
+
+static const struct pci_device_id pvpanic_pci_id_tbl[]  = {
+	{ PCI_DEVICE(PCI_VENDOR_ID_REDHAT, PCI_DEVICE_ID_REDHAT_PVPANIC) },
+	{ }
+};
+MODULE_DEVICE_TABLE(pci, pvpanic_pci_id_tbl);
+
+static struct pci_driver pvpanic_pci_driver = {
+	.name =         "pvpanic-pci",
+	.id_table =     pvpanic_pci_id_tbl,
+	.probe =        pvpanic_pci_probe,
+	.dev_groups =   pvpanic_dev_groups,
+};
+module_pci_driver(pvpanic_pci_driver);
diff --git a/drivers/misc/pvpanic/pvpanic.c b/drivers/misc/pvpanic/pvpanic.c
new file mode 100644
index 000000000..b57d773f6
--- /dev/null
+++ b/drivers/misc/pvpanic/pvpanic.c
@@ -0,0 +1,231 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ *  Pvpanic Device Support
+ *
+ *  Copyright (C) 2013 Fujitsu.
+ *  Copyright (C) 2018 ZTE.
+ *  Copyright (C) 2021 Oracle.
+ */
+
+#include <linux/device.h>
+#include <linux/errno.h>
+#include <linux/gfp_types.h>
+#include <linux/io.h>
+#include <linux/kexec.h>
+#include <linux/kstrtox.h>
+#include <linux/limits.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/panic_notifier.h>
+#include <linux/platform_device.h>
+#include <linux/reboot.h>
+#include <linux/spinlock.h>
+#include <linux/sysfs.h>
+#include <linux/types.h>
+
+#include <uapi/misc/pvpanic.h>
+
+#include "pvpanic.h"
+
+MODULE_AUTHOR("Mihai Carabas <mihai.carabas@oracle.com>");
+MODULE_DESCRIPTION("pvpanic device driver");
+MODULE_LICENSE("GPL");
+
+struct pvpanic_instance {
+	void __iomem *base;
+	unsigned int capability;
+	unsigned int events;
+	struct sys_off_handler *sys_off;
+	struct list_head list;
+};
+
+static struct list_head pvpanic_list;
+static spinlock_t pvpanic_lock;
+
+static void
+pvpanic_send_event(unsigned int event)
+{
+	struct pvpanic_instance *pi_cur;
+
+	if (!spin_trylock(&pvpanic_lock))
+		return;
+
+	list_for_each_entry(pi_cur, &pvpanic_list, list) {
+		if (event & pi_cur->capability & pi_cur->events)
+			iowrite8(event, pi_cur->base);
+	}
+	spin_unlock(&pvpanic_lock);
+}
+
+static int
+pvpanic_panic_notify(struct notifier_block *nb, unsigned long code, void *unused)
+{
+	unsigned int event = PVPANIC_PANICKED;
+
+	if (kexec_crash_loaded())
+		event = PVPANIC_CRASH_LOADED;
+
+	pvpanic_send_event(event);
+
+	return NOTIFY_DONE;
+}
+
+/*
+ * Call our notifier very early on panic, deferring the
+ * action taken to the hypervisor.
+ */
+static struct notifier_block pvpanic_panic_nb = {
+	.notifier_call = pvpanic_panic_notify,
+	.priority = INT_MAX,
+};
+
+static int pvpanic_sys_off(struct sys_off_data *data)
+{
+	pvpanic_send_event(PVPANIC_SHUTDOWN);
+
+	return NOTIFY_DONE;
+}
+
+static void pvpanic_synchronize_sys_off_handler(struct device *dev, struct pvpanic_instance *pi)
+{
+	/* The kernel core has logic to fall back to system halt if no
+	 * sys_off_handler is registered.
+	 * When the pvpanic sys_off_handler is disabled via sysfs the kernel
+	 * should use that fallback logic, so the handler needs to be unregistered.
+	 */
+
+	struct sys_off_handler *sys_off;
+
+	if (!(pi->events & PVPANIC_SHUTDOWN) == !pi->sys_off)
+		return;
+
+	if (!pi->sys_off) {
+		sys_off = register_sys_off_handler(SYS_OFF_MODE_POWER_OFF, SYS_OFF_PRIO_LOW,
+						   pvpanic_sys_off, NULL);
+		if (IS_ERR(sys_off))
+			dev_warn(dev, "Could not register sys_off_handler: %pe\n", sys_off);
+		else
+			pi->sys_off = sys_off;
+	} else {
+		unregister_sys_off_handler(pi->sys_off);
+		pi->sys_off = NULL;
+	}
+}
+
+static void pvpanic_remove(void *param)
+{
+	struct pvpanic_instance *pi_cur, *pi_next;
+	struct pvpanic_instance *pi = param;
+
+	spin_lock(&pvpanic_lock);
+	list_for_each_entry_safe(pi_cur, pi_next, &pvpanic_list, list) {
+		if (pi_cur == pi) {
+			list_del(&pi_cur->list);
+			break;
+		}
+	}
+	spin_unlock(&pvpanic_lock);
+
+	unregister_sys_off_handler(pi->sys_off);
+}
+
+static ssize_t capability_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct pvpanic_instance *pi = dev_get_drvdata(dev);
+
+	return sysfs_emit(buf, "%x\n", pi->capability);
+}
+static DEVICE_ATTR_RO(capability);
+
+static ssize_t events_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+	struct pvpanic_instance *pi = dev_get_drvdata(dev);
+
+	return sysfs_emit(buf, "%x\n", pi->events);
+}
+
+static ssize_t events_store(struct device *dev, struct device_attribute *attr,
+			    const char *buf, size_t count)
+{
+	struct pvpanic_instance *pi = dev_get_drvdata(dev);
+	unsigned int tmp;
+	int err;
+
+	err = kstrtouint(buf, 16, &tmp);
+	if (err)
+		return err;
+
+	if ((tmp & pi->capability) != tmp)
+		return -EINVAL;
+
+	pi->events = tmp;
+	pvpanic_synchronize_sys_off_handler(dev, pi);
+
+	return count;
+}
+static DEVICE_ATTR_RW(events);
+
+static struct attribute *pvpanic_dev_attrs[] = {
+	&dev_attr_capability.attr,
+	&dev_attr_events.attr,
+	NULL
+};
+
+static const struct attribute_group pvpanic_dev_group = {
+	.attrs = pvpanic_dev_attrs,
+};
+
+const struct attribute_group *pvpanic_dev_groups[] = {
+	&pvpanic_dev_group,
+	NULL
+};
+EXPORT_SYMBOL_GPL(pvpanic_dev_groups);
+
+int devm_pvpanic_probe(struct device *dev, void __iomem *base)
+{
+	struct pvpanic_instance *pi;
+
+	if (!base)
+		return -EINVAL;
+
+	pi = devm_kmalloc(dev, sizeof(*pi), GFP_KERNEL);
+	if (!pi)
+		return -ENOMEM;
+
+	pi->base = base;
+	pi->capability = PVPANIC_PANICKED | PVPANIC_CRASH_LOADED | PVPANIC_SHUTDOWN;
+
+	/* initlize capability by RDPT */
+	pi->capability &= ioread8(base);
+	pi->events = pi->capability;
+
+	pi->sys_off = NULL;
+	pvpanic_synchronize_sys_off_handler(dev, pi);
+
+	spin_lock(&pvpanic_lock);
+	list_add(&pi->list, &pvpanic_list);
+	spin_unlock(&pvpanic_lock);
+
+	dev_set_drvdata(dev, pi);
+
+	return devm_add_action_or_reset(dev, pvpanic_remove, pi);
+}
+EXPORT_SYMBOL_GPL(devm_pvpanic_probe);
+
+static int pvpanic_init(void)
+{
+	INIT_LIST_HEAD(&pvpanic_list);
+	spin_lock_init(&pvpanic_lock);
+
+	atomic_notifier_chain_register(&panic_notifier_list, &pvpanic_panic_nb);
+
+	return 0;
+}
+module_init(pvpanic_init);
+
+static void pvpanic_exit(void)
+{
+	atomic_notifier_chain_unregister(&panic_notifier_list, &pvpanic_panic_nb);
+
+}
+module_exit(pvpanic_exit);
diff --git a/drivers/misc/pvpanic/pvpanic.h b/drivers/misc/pvpanic/pvpanic.h
new file mode 100644
index 000000000..a42fa760e
--- /dev/null
+++ b/drivers/misc/pvpanic/pvpanic.h
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ *  Pvpanic Device Support
+ *
+ *  Copyright (C) 2021 Oracle.
+ */
+
+#ifndef PVPANIC_H_
+#define PVPANIC_H_
+
+#include <linux/compiler_types.h>
+
+struct attribute_group;
+struct device;
+
+int devm_pvpanic_probe(struct device *dev, void __iomem *base);
+extern const struct attribute_group *pvpanic_dev_groups[];
+
+#endif /* PVPANIC_H_ */