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, 243 insertions, 0 deletions

diff --git a/drivers/nvmem/meson-mx-efuse.c b/drivers/nvmem/meson-mx-efuse.c
new file mode 100644
index 000000000..8a16f5f02
--- /dev/null
+++ b/drivers/nvmem/meson-mx-efuse.c
@@ -0,0 +1,243 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Amlogic Meson6, Meson8 and Meson8b eFuse Driver
+ *
+ * Copyright (c) 2017 Martin Blumenstingl <martin.blumenstingl@googlemail.com>
+ */
+
+#include <linux/bitfield.h>
+#include <linux/bitops.h>
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+#include <linux/iopoll.h>
+#include <linux/module.h>
+#include <linux/nvmem-provider.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/sizes.h>
+#include <linux/slab.h>
+
+#define MESON_MX_EFUSE_CNTL1					0x04
+#define MESON_MX_EFUSE_CNTL1_PD_ENABLE				BIT(27)
+#define MESON_MX_EFUSE_CNTL1_AUTO_RD_BUSY			BIT(26)
+#define MESON_MX_EFUSE_CNTL1_AUTO_RD_START			BIT(25)
+#define MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE			BIT(24)
+#define MESON_MX_EFUSE_CNTL1_BYTE_WR_DATA			GENMASK(23, 16)
+#define MESON_MX_EFUSE_CNTL1_AUTO_WR_BUSY			BIT(14)
+#define MESON_MX_EFUSE_CNTL1_AUTO_WR_START			BIT(13)
+#define MESON_MX_EFUSE_CNTL1_AUTO_WR_ENABLE			BIT(12)
+#define MESON_MX_EFUSE_CNTL1_BYTE_ADDR_SET			BIT(11)
+#define MESON_MX_EFUSE_CNTL1_BYTE_ADDR_MASK			GENMASK(10, 0)
+
+#define MESON_MX_EFUSE_CNTL2					0x08
+
+#define MESON_MX_EFUSE_CNTL4					0x10
+#define MESON_MX_EFUSE_CNTL4_ENCRYPT_ENABLE			BIT(10)
+
+struct meson_mx_efuse_platform_data {
+	const char *name;
+	unsigned int word_size;
+};
+
+struct meson_mx_efuse {
+	void __iomem *base;
+	struct clk *core_clk;
+	struct nvmem_config config;
+};
+
+static void meson_mx_efuse_mask_bits(struct meson_mx_efuse *efuse, u32 reg,
+				     u32 mask, u32 set)
+{
+	u32 data;
+
+	data = readl(efuse->base + reg);
+	data &= ~mask;
+	data |= (set & mask);
+
+	writel(data, efuse->base + reg);
+}
+
+static int meson_mx_efuse_hw_enable(struct meson_mx_efuse *efuse)
+{
+	int err;
+
+	err = clk_prepare_enable(efuse->core_clk);
+	if (err)
+		return err;
+
+	/* power up the efuse */
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_PD_ENABLE, 0);
+
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL4,
+				 MESON_MX_EFUSE_CNTL4_ENCRYPT_ENABLE, 0);
+
+	return 0;
+}
+
+static void meson_mx_efuse_hw_disable(struct meson_mx_efuse *efuse)
+{
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_PD_ENABLE,
+				 MESON_MX_EFUSE_CNTL1_PD_ENABLE);
+
+	clk_disable_unprepare(efuse->core_clk);
+}
+
+static int meson_mx_efuse_read_addr(struct meson_mx_efuse *efuse,
+				    unsigned int addr, u32 *value)
+{
+	int err;
+	u32 regval;
+
+	/* write the address to read */
+	regval = FIELD_PREP(MESON_MX_EFUSE_CNTL1_BYTE_ADDR_MASK, addr);
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_BYTE_ADDR_MASK, regval);
+
+	/* inform the hardware that we changed the address */
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_BYTE_ADDR_SET,
+				 MESON_MX_EFUSE_CNTL1_BYTE_ADDR_SET);
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_BYTE_ADDR_SET, 0);
+
+	/* start the read process */
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_AUTO_RD_START,
+				 MESON_MX_EFUSE_CNTL1_AUTO_RD_START);
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_AUTO_RD_START, 0);
+
+	/*
+	 * perform a dummy read to ensure that the HW has the RD_BUSY bit set
+	 * when polling for the status below.
+	 */
+	readl(efuse->base + MESON_MX_EFUSE_CNTL1);
+
+	err = readl_poll_timeout_atomic(efuse->base + MESON_MX_EFUSE_CNTL1,
+			regval,
+			(!(regval & MESON_MX_EFUSE_CNTL1_AUTO_RD_BUSY)),
+			1, 1000);
+	if (err) {
+		dev_err(efuse->config.dev,
+			"Timeout while reading efuse address %u\n", addr);
+		return err;
+	}
+
+	*value = readl(efuse->base + MESON_MX_EFUSE_CNTL2);
+
+	return 0;
+}
+
+static int meson_mx_efuse_read(void *context, unsigned int offset,
+			       void *buf, size_t bytes)
+{
+	struct meson_mx_efuse *efuse = context;
+	u32 tmp;
+	int err, i, addr;
+
+	err = meson_mx_efuse_hw_enable(efuse);
+	if (err)
+		return err;
+
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE,
+				 MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE);
+
+	for (i = 0; i < bytes; i += efuse->config.word_size) {
+		addr = (offset + i) / efuse->config.word_size;
+
+		err = meson_mx_efuse_read_addr(efuse, addr, &tmp);
+		if (err)
+			break;
+
+		memcpy(buf + i, &tmp,
+		       min_t(size_t, bytes - i, efuse->config.word_size));
+	}
+
+	meson_mx_efuse_mask_bits(efuse, MESON_MX_EFUSE_CNTL1,
+				 MESON_MX_EFUSE_CNTL1_AUTO_RD_ENABLE, 0);
+
+	meson_mx_efuse_hw_disable(efuse);
+
+	return err;
+}
+
+static const struct meson_mx_efuse_platform_data meson6_efuse_data = {
+	.name = "meson6-efuse",
+	.word_size = 1,
+};
+
+static const struct meson_mx_efuse_platform_data meson8_efuse_data = {
+	.name = "meson8-efuse",
+	.word_size = 4,
+};
+
+static const struct meson_mx_efuse_platform_data meson8b_efuse_data = {
+	.name = "meson8b-efuse",
+	.word_size = 4,
+};
+
+static const struct of_device_id meson_mx_efuse_match[] = {
+	{ .compatible = "amlogic,meson6-efuse", .data = &meson6_efuse_data },
+	{ .compatible = "amlogic,meson8-efuse", .data = &meson8_efuse_data },
+	{ .compatible = "amlogic,meson8b-efuse", .data = &meson8b_efuse_data },
+	{ /* sentinel */ },
+};
+MODULE_DEVICE_TABLE(of, meson_mx_efuse_match);
+
+static int meson_mx_efuse_probe(struct platform_device *pdev)
+{
+	const struct meson_mx_efuse_platform_data *drvdata;
+	struct meson_mx_efuse *efuse;
+	struct nvmem_device *nvmem;
+
+	drvdata = of_device_get_match_data(&pdev->dev);
+	if (!drvdata)
+		return -EINVAL;
+
+	efuse = devm_kzalloc(&pdev->dev, sizeof(*efuse), GFP_KERNEL);
+	if (!efuse)
+		return -ENOMEM;
+
+	efuse->base = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(efuse->base))
+		return PTR_ERR(efuse->base);
+
+	efuse->config.name = drvdata->name;
+	efuse->config.owner = THIS_MODULE;
+	efuse->config.dev = &pdev->dev;
+	efuse->config.priv = efuse;
+	efuse->config.add_legacy_fixed_of_cells = true;
+	efuse->config.stride = drvdata->word_size;
+	efuse->config.word_size = drvdata->word_size;
+	efuse->config.size = SZ_512;
+	efuse->config.read_only = true;
+	efuse->config.reg_read = meson_mx_efuse_read;
+
+	efuse->core_clk = devm_clk_get(&pdev->dev, "core");
+	if (IS_ERR(efuse->core_clk)) {
+		dev_err(&pdev->dev, "Failed to get core clock\n");
+		return PTR_ERR(efuse->core_clk);
+	}
+
+	nvmem = devm_nvmem_register(&pdev->dev, &efuse->config);
+
+	return PTR_ERR_OR_ZERO(nvmem);
+}
+
+static struct platform_driver meson_mx_efuse_driver = {
+	.probe = meson_mx_efuse_probe,
+	.driver = {
+		.name = "meson-mx-efuse",
+		.of_match_table = meson_mx_efuse_match,
+	},
+};
+
+module_platform_driver(meson_mx_efuse_driver);
+
+MODULE_AUTHOR("Martin Blumenstingl <martin.blumenstingl@googlemail.com>");
+MODULE_DESCRIPTION("Amlogic Meson MX eFuse NVMEM driver");
+MODULE_LICENSE("GPL v2");