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, 227 insertions, 0 deletions

diff --git a/drivers/reset/reset-microchip-sparx5.c b/drivers/reset/reset-microchip-sparx5.c
new file mode 100644
index 000000000..6d3e75b33
--- /dev/null
+++ b/drivers/reset/reset-microchip-sparx5.c
@@ -0,0 +1,227 @@
+// SPDX-License-Identifier: GPL-2.0+
+/* Microchip Sparx5 Switch Reset driver
+ *
+ * Copyright (c) 2020 Microchip Technology Inc. and its subsidiaries.
+ *
+ * The Sparx5 Chip Register Model can be browsed at this location:
+ * https://github.com/microchip-ung/sparx-5_reginfo
+ */
+#include <linux/mfd/syscon.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/property.h>
+#include <linux/regmap.h>
+#include <linux/reset-controller.h>
+
+struct reset_props {
+	u32 protect_reg;
+	u32 protect_bit;
+	u32 reset_reg;
+	u32 reset_bit;
+};
+
+struct mchp_reset_context {
+	struct regmap *cpu_ctrl;
+	struct regmap *gcb_ctrl;
+	struct reset_controller_dev rcdev;
+	const struct reset_props *props;
+};
+
+static struct regmap_config sparx5_reset_regmap_config = {
+	.reg_bits	= 32,
+	.val_bits	= 32,
+	.reg_stride	= 4,
+};
+
+static int sparx5_switch_reset(struct mchp_reset_context *ctx)
+{
+	u32 val;
+
+	/* Make sure the core is PROTECTED from reset */
+	regmap_update_bits(ctx->cpu_ctrl, ctx->props->protect_reg,
+			   ctx->props->protect_bit, ctx->props->protect_bit);
+
+	/* Start soft reset */
+	regmap_write(ctx->gcb_ctrl, ctx->props->reset_reg,
+		     ctx->props->reset_bit);
+
+	/* Wait for soft reset done */
+	return regmap_read_poll_timeout(ctx->gcb_ctrl, ctx->props->reset_reg, val,
+					(val & ctx->props->reset_bit) == 0,
+					1, 100);
+}
+
+static int sparx5_reset_noop(struct reset_controller_dev *rcdev,
+			     unsigned long id)
+{
+	return 0;
+}
+
+static const struct reset_control_ops sparx5_reset_ops = {
+	.reset = sparx5_reset_noop,
+};
+
+static const struct regmap_config mchp_lan966x_syscon_regmap_config = {
+	.reg_bits = 32,
+	.val_bits = 32,
+	.reg_stride = 4,
+};
+
+static struct regmap *mchp_lan966x_syscon_to_regmap(struct device *dev,
+						    struct device_node *syscon_np)
+{
+	struct regmap_config regmap_config = mchp_lan966x_syscon_regmap_config;
+	struct resource res;
+	void __iomem *base;
+	int err;
+
+	err = of_address_to_resource(syscon_np, 0, &res);
+	if (err)
+		return ERR_PTR(err);
+
+	/* It is not possible to use devm_of_iomap because this resource is
+	 * shared with other drivers.
+	 */
+	base = devm_ioremap(dev, res.start, resource_size(&res));
+	if (!base)
+		return ERR_PTR(-ENOMEM);
+
+	regmap_config.max_register =  resource_size(&res) - 4;
+
+	return devm_regmap_init_mmio(dev, base, &regmap_config);
+}
+
+static int mchp_sparx5_map_syscon(struct platform_device *pdev, char *name,
+				  struct regmap **target)
+{
+	struct device_node *syscon_np;
+	struct regmap *regmap;
+	int err;
+
+	syscon_np = of_parse_phandle(pdev->dev.of_node, name, 0);
+	if (!syscon_np)
+		return -ENODEV;
+
+	/*
+	 * The syscon API doesn't support syscon device removal.
+	 * When used in LAN966x PCI device, the cpu-syscon device needs to be
+	 * removed when the PCI device is removed.
+	 * In case of LAN966x, map the syscon device locally to support the
+	 * device removal.
+	 */
+	if (of_device_is_compatible(pdev->dev.of_node, "microchip,lan966x-switch-reset"))
+		regmap = mchp_lan966x_syscon_to_regmap(&pdev->dev, syscon_np);
+	else
+		regmap = syscon_node_to_regmap(syscon_np);
+	of_node_put(syscon_np);
+	if (IS_ERR(regmap)) {
+		err = PTR_ERR(regmap);
+		dev_err(&pdev->dev, "No '%s' map: %d\n", name, err);
+		return err;
+	}
+	*target = regmap;
+	return 0;
+}
+
+static int mchp_sparx5_map_io(struct platform_device *pdev, int index,
+			      struct regmap **target)
+{
+	struct resource *res;
+	struct regmap *map;
+	void __iomem *mem;
+
+	mem = devm_platform_get_and_ioremap_resource(pdev, index, &res);
+	if (IS_ERR(mem)) {
+		dev_err(&pdev->dev, "Could not map resource %d\n", index);
+		return PTR_ERR(mem);
+	}
+	sparx5_reset_regmap_config.name = res->name;
+	map = devm_regmap_init_mmio(&pdev->dev, mem, &sparx5_reset_regmap_config);
+	if (IS_ERR(map))
+		return PTR_ERR(map);
+	*target = map;
+	return 0;
+}
+
+static int mchp_sparx5_reset_probe(struct platform_device *pdev)
+{
+	struct device_node *dn = pdev->dev.of_node;
+	struct mchp_reset_context *ctx;
+	int err;
+
+	ctx = devm_kzalloc(&pdev->dev, sizeof(*ctx), GFP_KERNEL);
+	if (!ctx)
+		return -ENOMEM;
+
+	err = mchp_sparx5_map_syscon(pdev, "cpu-syscon", &ctx->cpu_ctrl);
+	if (err)
+		return err;
+	err = mchp_sparx5_map_io(pdev, 0, &ctx->gcb_ctrl);
+	if (err)
+		return err;
+
+	ctx->rcdev.owner = THIS_MODULE;
+	ctx->rcdev.dev = &pdev->dev;
+	ctx->rcdev.nr_resets = 1;
+	ctx->rcdev.ops = &sparx5_reset_ops;
+	ctx->rcdev.of_node = dn;
+	ctx->props = device_get_match_data(&pdev->dev);
+
+	/* Issue the reset very early, our actual reset callback is a noop. */
+	err = sparx5_switch_reset(ctx);
+	if (err)
+		return err;
+
+	return devm_reset_controller_register(&pdev->dev, &ctx->rcdev);
+}
+
+static const struct reset_props reset_props_sparx5 = {
+	.protect_reg    = 0x84,
+	.protect_bit    = BIT(10),
+	.reset_reg      = 0x0,
+	.reset_bit      = BIT(1),
+};
+
+static const struct reset_props reset_props_lan966x = {
+	.protect_reg    = 0x88,
+	.protect_bit    = BIT(5),
+	.reset_reg      = 0x0,
+	.reset_bit      = BIT(1),
+};
+
+static const struct of_device_id mchp_sparx5_reset_of_match[] = {
+	{
+		.compatible = "microchip,sparx5-switch-reset",
+		.data = &reset_props_sparx5,
+	}, {
+		.compatible = "microchip,lan966x-switch-reset",
+		.data = &reset_props_lan966x,
+	},
+	{ }
+};
+MODULE_DEVICE_TABLE(of, mchp_sparx5_reset_of_match);
+
+static struct platform_driver mchp_sparx5_reset_driver = {
+	.probe = mchp_sparx5_reset_probe,
+	.driver = {
+		.name = "sparx5-switch-reset",
+		.of_match_table = mchp_sparx5_reset_of_match,
+	},
+};
+
+static int __init mchp_sparx5_reset_init(void)
+{
+	return platform_driver_register(&mchp_sparx5_reset_driver);
+}
+
+/*
+ * Because this is a global reset, keep this postcore_initcall() to issue the
+ * reset as early as possible during the kernel startup.
+ */
+postcore_initcall(mchp_sparx5_reset_init);
+
+MODULE_DESCRIPTION("Microchip Sparx5 switch reset driver");
+MODULE_AUTHOR("Steen Hegelund <steen.hegelund@microchip.com>");
+MODULE_LICENSE("GPL");