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, 205 insertions, 0 deletions

diff --git a/drivers/nvmem/layouts.c b/drivers/nvmem/layouts.c
new file mode 100644
index 000000000..07a34be96
--- /dev/null
+++ b/drivers/nvmem/layouts.c
@@ -0,0 +1,205 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * NVMEM layout bus handling
+ *
+ * Copyright (C) 2023 Bootlin
+ * Author: Miquel Raynal <miquel.raynal@bootlin.com
+ */
+
+#include <linux/device.h>
+#include <linux/dma-mapping.h>
+#include <linux/nvmem-consumer.h>
+#include <linux/nvmem-provider.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/of_irq.h>
+
+#include "internals.h"
+
+#define to_nvmem_layout_driver(drv) \
+	(container_of_const((drv), struct nvmem_layout_driver, driver))
+#define to_nvmem_layout_device(_dev) \
+	container_of((_dev), struct nvmem_layout, dev)
+
+static int nvmem_layout_bus_match(struct device *dev, const struct device_driver *drv)
+{
+	return of_driver_match_device(dev, drv);
+}
+
+static int nvmem_layout_bus_probe(struct device *dev)
+{
+	struct nvmem_layout_driver *drv = to_nvmem_layout_driver(dev->driver);
+	struct nvmem_layout *layout = to_nvmem_layout_device(dev);
+
+	if (!drv->probe || !drv->remove)
+		return -EINVAL;
+
+	return drv->probe(layout);
+}
+
+static void nvmem_layout_bus_remove(struct device *dev)
+{
+	struct nvmem_layout_driver *drv = to_nvmem_layout_driver(dev->driver);
+	struct nvmem_layout *layout = to_nvmem_layout_device(dev);
+
+	return drv->remove(layout);
+}
+
+static int nvmem_layout_bus_uevent(const struct device *dev,
+				   struct kobj_uevent_env *env)
+{
+	int ret;
+
+	ret = of_device_uevent_modalias(dev, env);
+	if (ret != -ENODEV)
+		return ret;
+
+	return 0;
+}
+
+static const struct bus_type nvmem_layout_bus_type = {
+	.name		= "nvmem-layout",
+	.match		= nvmem_layout_bus_match,
+	.probe		= nvmem_layout_bus_probe,
+	.remove		= nvmem_layout_bus_remove,
+	.uevent		= nvmem_layout_bus_uevent,
+};
+
+int __nvmem_layout_driver_register(struct nvmem_layout_driver *drv,
+				   struct module *owner)
+{
+	drv->driver.bus = &nvmem_layout_bus_type;
+	drv->driver.owner = owner;
+
+	return driver_register(&drv->driver);
+}
+EXPORT_SYMBOL_GPL(__nvmem_layout_driver_register);
+
+void nvmem_layout_driver_unregister(struct nvmem_layout_driver *drv)
+{
+	driver_unregister(&drv->driver);
+}
+EXPORT_SYMBOL_GPL(nvmem_layout_driver_unregister);
+
+static void nvmem_layout_release_device(struct device *dev)
+{
+	struct nvmem_layout *layout = to_nvmem_layout_device(dev);
+
+	of_node_put(layout->dev.of_node);
+	kfree(layout);
+}
+
+static int nvmem_layout_create_device(struct nvmem_device *nvmem,
+				      struct device_node *np)
+{
+	struct nvmem_layout *layout;
+	struct device *dev;
+	int ret;
+
+	layout = kzalloc_obj(*layout);
+	if (!layout)
+		return -ENOMEM;
+
+	/* Create a bidirectional link */
+	layout->nvmem = nvmem;
+	nvmem->layout = layout;
+
+	/* Device model registration */
+	dev = &layout->dev;
+	device_initialize(dev);
+	dev->parent = &nvmem->dev;
+	dev->bus = &nvmem_layout_bus_type;
+	dev->release = nvmem_layout_release_device;
+	dev->coherent_dma_mask = DMA_BIT_MASK(32);
+	dev->dma_mask = &dev->coherent_dma_mask;
+	device_set_node(dev, of_fwnode_handle(of_node_get(np)));
+	of_device_make_bus_id(dev);
+	of_msi_configure(dev, dev->of_node);
+
+	ret = device_add(dev);
+	if (ret) {
+		put_device(dev);
+		return ret;
+	}
+
+	return 0;
+}
+
+static int nvmem_layout_bus_populate(struct nvmem_device *nvmem,
+				     struct device_node *layout_dn)
+{
+	int ret;
+
+	/* Make sure it has a compatible property */
+	if (!of_property_present(layout_dn, "compatible")) {
+		pr_debug("%s() - skipping %pOF, no compatible prop\n",
+			 __func__, layout_dn);
+		return 0;
+	}
+
+	if (of_node_check_flag(layout_dn, OF_POPULATED_BUS)) {
+		pr_debug("%s() - skipping %pOF, already populated\n",
+			 __func__, layout_dn);
+
+		return 0;
+	}
+
+	/* NVMEM layout buses expect only a single device representing the layout */
+	ret = nvmem_layout_create_device(nvmem, layout_dn);
+	if (ret)
+		return ret;
+
+	of_node_set_flag(layout_dn, OF_POPULATED_BUS);
+
+	return 0;
+}
+
+struct device_node *of_nvmem_layout_get_container(struct nvmem_device *nvmem)
+{
+	return of_get_child_by_name(nvmem->dev.of_node, "nvmem-layout");
+}
+EXPORT_SYMBOL_GPL(of_nvmem_layout_get_container);
+
+/*
+ * Returns the number of devices populated, 0 if the operation was not relevant
+ * for this nvmem device, an error code otherwise.
+ */
+int nvmem_populate_layout(struct nvmem_device *nvmem)
+{
+	struct device_node *layout_dn;
+	int ret;
+
+	layout_dn = of_nvmem_layout_get_container(nvmem);
+	if (!layout_dn)
+		return 0;
+
+	/* Populate the layout device */
+	device_links_supplier_sync_state_pause();
+	ret = nvmem_layout_bus_populate(nvmem, layout_dn);
+	device_links_supplier_sync_state_resume();
+
+	of_node_put(layout_dn);
+	return ret;
+}
+
+void nvmem_destroy_layout(struct nvmem_device *nvmem)
+{
+	struct device *dev;
+
+	if (!nvmem->layout)
+		return;
+
+	dev = &nvmem->layout->dev;
+	of_node_clear_flag(dev->of_node, OF_POPULATED_BUS);
+	device_unregister(dev);
+}
+
+int nvmem_layout_bus_register(void)
+{
+	return bus_register(&nvmem_layout_bus_type);
+}
+
+void nvmem_layout_bus_unregister(void)
+{
+	bus_unregister(&nvmem_layout_bus_type);
+}