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, 256 insertions, 0 deletions

diff --git a/drivers/reset/reset-intel-gw.c b/drivers/reset/reset-intel-gw.c
new file mode 100644
index 000000000..0db64cc8a
--- /dev/null
+++ b/drivers/reset/reset-intel-gw.c
@@ -0,0 +1,256 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (c) 2019 Intel Corporation.
+ * Lei Chuanhua <Chuanhua.lei@intel.com>
+ */
+
+#include <linux/bitfield.h>
+#include <linux/init.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/reboot.h>
+#include <linux/regmap.h>
+#include <linux/reset-controller.h>
+
+#define RCU_RST_STAT	0x0024
+#define RCU_RST_REQ	0x0048
+
+#define REG_OFFSET_MASK	GENMASK(31, 16)
+#define BIT_OFFSET_MASK	GENMASK(15, 8)
+#define STAT_BIT_OFFSET_MASK	GENMASK(7, 0)
+
+#define to_reset_data(x)	container_of(x, struct intel_reset_data, rcdev)
+
+struct intel_reset_soc {
+	bool legacy;
+	u32 reset_cell_count;
+};
+
+struct intel_reset_data {
+	struct reset_controller_dev rcdev;
+	const struct intel_reset_soc *soc_data;
+	struct regmap *regmap;
+	struct device *dev;
+	u32 reboot_id;
+};
+
+static const struct regmap_config intel_rcu_regmap_config = {
+	.name =		"intel-reset",
+	.reg_bits =	32,
+	.reg_stride =	4,
+	.val_bits =	32,
+};
+
+/*
+ * Reset status register offset relative to
+ * the reset control register(X) is X + 4
+ */
+static u32 id_to_reg_and_bit_offsets(struct intel_reset_data *data,
+				     unsigned long id, u32 *rst_req,
+				     u32 *req_bit, u32 *stat_bit)
+{
+	*rst_req = FIELD_GET(REG_OFFSET_MASK, id);
+	*req_bit = FIELD_GET(BIT_OFFSET_MASK, id);
+
+	if (data->soc_data->legacy)
+		*stat_bit = FIELD_GET(STAT_BIT_OFFSET_MASK, id);
+	else
+		*stat_bit = *req_bit;
+
+	if (data->soc_data->legacy && *rst_req == RCU_RST_REQ)
+		return RCU_RST_STAT;
+	else
+		return *rst_req + 0x4;
+}
+
+static int intel_set_clr_bits(struct intel_reset_data *data, unsigned long id,
+			      bool set)
+{
+	u32 rst_req, req_bit, rst_stat, stat_bit, val;
+	int ret;
+
+	rst_stat = id_to_reg_and_bit_offsets(data, id, &rst_req,
+					     &req_bit, &stat_bit);
+
+	val = set ? BIT(req_bit) : 0;
+	ret = regmap_update_bits(data->regmap, rst_req,  BIT(req_bit), val);
+	if (ret)
+		return ret;
+
+	return regmap_read_poll_timeout(data->regmap, rst_stat, val,
+					set == !!(val & BIT(stat_bit)), 20,
+					200);
+}
+
+static int intel_assert_device(struct reset_controller_dev *rcdev,
+			       unsigned long id)
+{
+	struct intel_reset_data *data = to_reset_data(rcdev);
+	int ret;
+
+	ret = intel_set_clr_bits(data, id, true);
+	if (ret)
+		dev_err(data->dev, "Reset assert failed %d\n", ret);
+
+	return ret;
+}
+
+static int intel_deassert_device(struct reset_controller_dev *rcdev,
+				 unsigned long id)
+{
+	struct intel_reset_data *data = to_reset_data(rcdev);
+	int ret;
+
+	ret = intel_set_clr_bits(data, id, false);
+	if (ret)
+		dev_err(data->dev, "Reset deassert failed %d\n", ret);
+
+	return ret;
+}
+
+static int intel_reset_status(struct reset_controller_dev *rcdev,
+			      unsigned long id)
+{
+	struct intel_reset_data *data = to_reset_data(rcdev);
+	u32 rst_req, req_bit, rst_stat, stat_bit, val;
+	int ret;
+
+	rst_stat = id_to_reg_and_bit_offsets(data, id, &rst_req,
+					     &req_bit, &stat_bit);
+	ret = regmap_read(data->regmap, rst_stat, &val);
+	if (ret)
+		return ret;
+
+	return !!(val & BIT(stat_bit));
+}
+
+static const struct reset_control_ops intel_reset_ops = {
+	.assert =	intel_assert_device,
+	.deassert =	intel_deassert_device,
+	.status	=	intel_reset_status,
+};
+
+static int intel_reset_xlate(struct reset_controller_dev *rcdev,
+			     const struct of_phandle_args *spec)
+{
+	struct intel_reset_data *data = to_reset_data(rcdev);
+	u32 id;
+
+	if (spec->args[1] > 31)
+		return -EINVAL;
+
+	id = FIELD_PREP(REG_OFFSET_MASK, spec->args[0]);
+	id |= FIELD_PREP(BIT_OFFSET_MASK, spec->args[1]);
+
+	if (data->soc_data->legacy) {
+		if (spec->args[2] > 31)
+			return -EINVAL;
+
+		id |= FIELD_PREP(STAT_BIT_OFFSET_MASK, spec->args[2]);
+	}
+
+	return id;
+}
+
+static int intel_reset_restart_handler(struct sys_off_data *data)
+{
+	struct intel_reset_data *reset_data = data->cb_data;
+
+	intel_assert_device(&reset_data->rcdev, reset_data->reboot_id);
+
+	return NOTIFY_DONE;
+}
+
+static int intel_reset_probe(struct platform_device *pdev)
+{
+	struct device_node *np = pdev->dev.of_node;
+	struct device *dev = &pdev->dev;
+	struct intel_reset_data *data;
+	void __iomem *base;
+	u32 rb_id[3];
+	int ret;
+
+	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
+	if (!data)
+		return -ENOMEM;
+
+	data->soc_data = of_device_get_match_data(dev);
+	if (!data->soc_data)
+		return -ENODEV;
+
+	base = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(base))
+		return PTR_ERR(base);
+
+	data->regmap = devm_regmap_init_mmio(dev, base,
+					     &intel_rcu_regmap_config);
+	if (IS_ERR(data->regmap)) {
+		dev_err(dev, "regmap initialization failed\n");
+		return PTR_ERR(data->regmap);
+	}
+
+	ret = device_property_read_u32_array(dev, "intel,global-reset", rb_id,
+					     data->soc_data->reset_cell_count);
+	if (ret) {
+		dev_err(dev, "Failed to get global reset offset!\n");
+		return ret;
+	}
+
+	data->dev =			dev;
+	data->rcdev.of_node =		np;
+	data->rcdev.owner =		dev->driver->owner;
+	data->rcdev.ops	=		&intel_reset_ops;
+	data->rcdev.of_xlate =		intel_reset_xlate;
+	data->rcdev.of_reset_n_cells =	data->soc_data->reset_cell_count;
+	ret = devm_reset_controller_register(&pdev->dev, &data->rcdev);
+	if (ret)
+		return ret;
+
+	data->reboot_id = FIELD_PREP(REG_OFFSET_MASK, rb_id[0]);
+	data->reboot_id |= FIELD_PREP(BIT_OFFSET_MASK, rb_id[1]);
+
+	if (data->soc_data->legacy)
+		data->reboot_id |= FIELD_PREP(STAT_BIT_OFFSET_MASK, rb_id[2]);
+
+	devm_register_restart_handler(&pdev->dev, intel_reset_restart_handler, data);
+
+	return 0;
+}
+
+static const struct intel_reset_soc xrx200_data = {
+	.legacy =		true,
+	.reset_cell_count =	3,
+};
+
+static const struct intel_reset_soc lgm_data = {
+	.legacy =		false,
+	.reset_cell_count =	2,
+};
+
+static const struct of_device_id intel_reset_match[] = {
+	{ .compatible = "intel,rcu-lgm", .data = &lgm_data },
+	{ .compatible = "intel,rcu-xrx200", .data = &xrx200_data },
+	{}
+};
+
+static struct platform_driver intel_reset_driver = {
+	.probe = intel_reset_probe,
+	.driver = {
+		.name = "intel-reset",
+		.of_match_table = intel_reset_match,
+	},
+};
+
+static int __init intel_reset_init(void)
+{
+	return platform_driver_register(&intel_reset_driver);
+}
+
+/*
+ * RCU is system core entity which is in Always On Domain whose clocks
+ * or resource initialization happens in system core initialization.
+ * Also, it is required for most of the platform or architecture
+ * specific devices to perform reset operation as part of initialization.
+ * So perform RCU as post core initialization.
+ */
+postcore_initcall(intel_reset_init);