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, 300 insertions, 0 deletions

diff --git a/drivers/memory/pl172.c b/drivers/memory/pl172.c
new file mode 100644
index 000000000..be7ba599c
--- /dev/null
+++ b/drivers/memory/pl172.c
@@ -0,0 +1,300 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Memory controller driver for ARM PrimeCell PL172
+ * PrimeCell MultiPort Memory Controller (PL172)
+ *
+ * Copyright (C) 2015 Joachim Eastwood <manabian@gmail.com>
+ *
+ * Based on:
+ * TI AEMIF driver, Copyright (C) 2010 - 2013 Texas Instruments Inc.
+ */
+
+#include <linux/amba/bus.h>
+#include <linux/clk.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/io.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_platform.h>
+#include <linux/time.h>
+
+#define MPMC_STATIC_CFG(n)		(0x200 + 0x20 * (n))
+#define  MPMC_STATIC_CFG_MW_8BIT	0x0
+#define  MPMC_STATIC_CFG_MW_16BIT	0x1
+#define  MPMC_STATIC_CFG_MW_32BIT	0x2
+#define  MPMC_STATIC_CFG_PM		BIT(3)
+#define  MPMC_STATIC_CFG_PC		BIT(6)
+#define  MPMC_STATIC_CFG_PB		BIT(7)
+#define  MPMC_STATIC_CFG_EW		BIT(8)
+#define  MPMC_STATIC_CFG_B		BIT(19)
+#define  MPMC_STATIC_CFG_P		BIT(20)
+#define MPMC_STATIC_WAIT_WEN(n)		(0x204 + 0x20 * (n))
+#define  MPMC_STATIC_WAIT_WEN_MAX	0x0f
+#define MPMC_STATIC_WAIT_OEN(n)		(0x208 + 0x20 * (n))
+#define  MPMC_STATIC_WAIT_OEN_MAX	0x0f
+#define MPMC_STATIC_WAIT_RD(n)		(0x20c + 0x20 * (n))
+#define  MPMC_STATIC_WAIT_RD_MAX	0x1f
+#define MPMC_STATIC_WAIT_PAGE(n)	(0x210 + 0x20 * (n))
+#define  MPMC_STATIC_WAIT_PAGE_MAX	0x1f
+#define MPMC_STATIC_WAIT_WR(n)		(0x214 + 0x20 * (n))
+#define  MPMC_STATIC_WAIT_WR_MAX	0x1f
+#define MPMC_STATIC_WAIT_TURN(n)	(0x218 + 0x20 * (n))
+#define  MPMC_STATIC_WAIT_TURN_MAX	0x0f
+
+/* Maximum number of static chip selects */
+#define PL172_MAX_CS		4
+
+struct pl172_data {
+	void __iomem *base;
+	unsigned long rate;
+	struct clk *clk;
+};
+
+static int pl172_timing_prop(struct amba_device *adev,
+			     const struct device_node *np, const char *name,
+			     u32 reg_offset, u32 max, int start)
+{
+	struct pl172_data *pl172 = amba_get_drvdata(adev);
+	int cycles;
+	u32 val;
+
+	if (!of_property_read_u32(np, name, &val)) {
+		cycles = DIV_ROUND_UP(val * pl172->rate, NSEC_PER_MSEC) - start;
+		if (cycles < 0) {
+			cycles = 0;
+		} else if (cycles > max) {
+			dev_err(&adev->dev, "%s timing too tight\n", name);
+			return -EINVAL;
+		}
+
+		writel(cycles, pl172->base + reg_offset);
+	}
+
+	dev_dbg(&adev->dev, "%s: %u cycle(s)\n", name, start +
+				readl(pl172->base + reg_offset));
+
+	return 0;
+}
+
+static int pl172_setup_static(struct amba_device *adev,
+			      struct device_node *np, u32 cs)
+{
+	struct pl172_data *pl172 = amba_get_drvdata(adev);
+	u32 cfg;
+	int ret;
+
+	/* MPMC static memory configuration */
+	if (!of_property_read_u32(np, "mpmc,memory-width", &cfg)) {
+		if (cfg == 8) {
+			cfg = MPMC_STATIC_CFG_MW_8BIT;
+		} else if (cfg == 16) {
+			cfg = MPMC_STATIC_CFG_MW_16BIT;
+		} else if (cfg == 32) {
+			cfg = MPMC_STATIC_CFG_MW_32BIT;
+		} else {
+			dev_err(&adev->dev, "invalid memory width cs%u\n", cs);
+			return -EINVAL;
+		}
+	} else {
+		dev_err(&adev->dev, "memory-width property required\n");
+		return -EINVAL;
+	}
+
+	if (of_property_read_bool(np, "mpmc,async-page-mode"))
+		cfg |= MPMC_STATIC_CFG_PM;
+
+	if (of_property_read_bool(np, "mpmc,cs-active-high"))
+		cfg |= MPMC_STATIC_CFG_PC;
+
+	if (of_property_read_bool(np, "mpmc,byte-lane-low"))
+		cfg |= MPMC_STATIC_CFG_PB;
+
+	if (of_property_read_bool(np, "mpmc,extended-wait"))
+		cfg |= MPMC_STATIC_CFG_EW;
+
+	if (amba_part(adev) == 0x172 &&
+	    of_property_read_bool(np, "mpmc,buffer-enable"))
+		cfg |= MPMC_STATIC_CFG_B;
+
+	if (of_property_read_bool(np, "mpmc,write-protect"))
+		cfg |= MPMC_STATIC_CFG_P;
+
+	writel(cfg, pl172->base + MPMC_STATIC_CFG(cs));
+	dev_dbg(&adev->dev, "mpmc static config cs%u: 0x%08x\n", cs, cfg);
+
+	/* MPMC static memory timing */
+	ret = pl172_timing_prop(adev, np, "mpmc,write-enable-delay",
+				MPMC_STATIC_WAIT_WEN(cs),
+				MPMC_STATIC_WAIT_WEN_MAX, 1);
+	if (ret)
+		goto fail;
+
+	ret = pl172_timing_prop(adev, np, "mpmc,output-enable-delay",
+				MPMC_STATIC_WAIT_OEN(cs),
+				MPMC_STATIC_WAIT_OEN_MAX, 0);
+	if (ret)
+		goto fail;
+
+	ret = pl172_timing_prop(adev, np, "mpmc,read-access-delay",
+				MPMC_STATIC_WAIT_RD(cs),
+				MPMC_STATIC_WAIT_RD_MAX, 1);
+	if (ret)
+		goto fail;
+
+	ret = pl172_timing_prop(adev, np, "mpmc,page-mode-read-delay",
+				MPMC_STATIC_WAIT_PAGE(cs),
+				MPMC_STATIC_WAIT_PAGE_MAX, 1);
+	if (ret)
+		goto fail;
+
+	ret = pl172_timing_prop(adev, np, "mpmc,write-access-delay",
+				MPMC_STATIC_WAIT_WR(cs),
+				MPMC_STATIC_WAIT_WR_MAX, 2);
+	if (ret)
+		goto fail;
+
+	ret = pl172_timing_prop(adev, np, "mpmc,turn-round-delay",
+				MPMC_STATIC_WAIT_TURN(cs),
+				MPMC_STATIC_WAIT_TURN_MAX, 1);
+	if (ret)
+		goto fail;
+
+	return 0;
+fail:
+	dev_err(&adev->dev, "failed to configure cs%u\n", cs);
+	return ret;
+}
+
+static int pl172_parse_cs_config(struct amba_device *adev,
+				 struct device_node *np)
+{
+	u32 cs;
+
+	if (!of_property_read_u32(np, "mpmc,cs", &cs)) {
+		if (cs >= PL172_MAX_CS) {
+			dev_err(&adev->dev, "cs%u invalid\n", cs);
+			return -EINVAL;
+		}
+
+		return pl172_setup_static(adev, np, cs);
+	}
+
+	dev_err(&adev->dev, "cs property required\n");
+
+	return -EINVAL;
+}
+
+static void pl172_amba_release_regions(void *data)
+{
+	struct amba_device *adev = data;
+
+	amba_release_regions(adev);
+}
+
+static const char * const pl172_revisions[] = {"r1", "r2", "r2p3", "r2p4"};
+static const char * const pl175_revisions[] = {"r1"};
+static const char * const pl176_revisions[] = {"r0"};
+
+static int pl172_probe(struct amba_device *adev, const struct amba_id *id)
+{
+	struct device_node *child_np, *np = adev->dev.of_node;
+	struct device *dev = &adev->dev;
+	static const char *rev = "?";
+	struct pl172_data *pl172;
+	int ret;
+
+	if (amba_part(adev) == 0x172) {
+		if (amba_rev(adev) < ARRAY_SIZE(pl172_revisions))
+			rev = pl172_revisions[amba_rev(adev)];
+	} else if (amba_part(adev) == 0x175) {
+		if (amba_rev(adev) < ARRAY_SIZE(pl175_revisions))
+			rev = pl175_revisions[amba_rev(adev)];
+	} else if (amba_part(adev) == 0x176) {
+		if (amba_rev(adev) < ARRAY_SIZE(pl176_revisions))
+			rev = pl176_revisions[amba_rev(adev)];
+	}
+
+	dev_info(dev, "ARM PL%x revision %s\n", amba_part(adev), rev);
+
+	pl172 = devm_kzalloc(dev, sizeof(*pl172), GFP_KERNEL);
+	if (!pl172)
+		return -ENOMEM;
+
+	pl172->clk = devm_clk_get_enabled(dev, "mpmcclk");
+	if (IS_ERR(pl172->clk))
+		return dev_err_probe(dev, PTR_ERR(pl172->clk),
+				     "no mpmcclk provided clock\n");
+
+	pl172->rate = clk_get_rate(pl172->clk) / MSEC_PER_SEC;
+	if (!pl172->rate)
+		return dev_err_probe(dev, -EINVAL,
+				     "unable to get mpmcclk clock rate\n");
+
+	ret = amba_request_regions(adev, NULL);
+	if (ret) {
+		dev_err(dev, "unable to request AMBA regions\n");
+		return ret;
+	}
+
+	ret = devm_add_action_or_reset(dev, pl172_amba_release_regions, adev);
+	if (ret)
+		return ret;
+
+	pl172->base = devm_ioremap(dev, adev->res.start,
+				   resource_size(&adev->res));
+	if (!pl172->base)
+		return dev_err_probe(dev, -ENOMEM, "ioremap failed\n");
+
+	amba_set_drvdata(adev, pl172);
+
+	/*
+	 * Loop through each child node, which represent a chip select, and
+	 * configure parameters and timing. If successful; populate devices
+	 * under that node.
+	 */
+	for_each_available_child_of_node(np, child_np) {
+		ret = pl172_parse_cs_config(adev, child_np);
+		if (ret)
+			continue;
+
+		of_platform_populate(child_np, NULL, NULL, dev);
+	}
+
+	return 0;
+}
+
+static const struct amba_id pl172_ids[] = {
+	/*  PrimeCell MPMC PL172, EMC found on NXP LPC18xx and LPC43xx */
+	{
+		.id	= 0x07041172,
+		.mask	= 0x3f0fffff,
+	},
+	/* PrimeCell MPMC PL175, EMC found on NXP LPC32xx */
+	{
+		.id	= 0x07041175,
+		.mask	= 0x3f0fffff,
+	},
+	/* PrimeCell MPMC PL176 */
+	{
+		.id	= 0x89041176,
+		.mask	= 0xff0fffff,
+	},
+	{ 0, 0 },
+};
+MODULE_DEVICE_TABLE(amba, pl172_ids);
+
+static struct amba_driver pl172_driver = {
+	.drv = {
+		.name	= "memory-pl172",
+	},
+	.probe		= pl172_probe,
+	.id_table	= pl172_ids,
+};
+module_amba_driver(pl172_driver);
+
+MODULE_AUTHOR("Joachim Eastwood <manabian@gmail.com>");
+MODULE_DESCRIPTION("PL172 Memory Controller Driver");
+MODULE_LICENSE("GPL v2");