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

3 files changed, 478 insertions, 0 deletions

diff --git a/drivers/pmdomain/starfive/Kconfig b/drivers/pmdomain/starfive/Kconfig
new file mode 100644
index 000000000..1e9b0c414
--- /dev/null
+++ b/drivers/pmdomain/starfive/Kconfig
@@ -0,0 +1,12 @@
+# SPDX-License-Identifier: GPL-2.0
+
+config JH71XX_PMU
+	bool "Support PMU for StarFive JH71XX Soc"
+	depends on PM
+	depends on ARCH_STARFIVE || COMPILE_TEST
+	default ARCH_STARFIVE
+	select PM_GENERIC_DOMAINS
+	help
+	  Say 'y' here to enable support power domain support.
+	  In order to meet low power requirements, a Power Management Unit (PMU)
+	  is designed for controlling power resources in StarFive JH71XX SoCs.
diff --git a/drivers/pmdomain/starfive/Makefile b/drivers/pmdomain/starfive/Makefile
new file mode 100644
index 000000000..975bba2a2
--- /dev/null
+++ b/drivers/pmdomain/starfive/Makefile
@@ -0,0 +1,2 @@
+# SPDX-License-Identifier: GPL-2.0
+obj-$(CONFIG_JH71XX_PMU)		+= jh71xx-pmu.o
diff --git a/drivers/pmdomain/starfive/jh71xx-pmu.c b/drivers/pmdomain/starfive/jh71xx-pmu.c
new file mode 100644
index 000000000..74720c09a
--- /dev/null
+++ b/drivers/pmdomain/starfive/jh71xx-pmu.c
@@ -0,0 +1,464 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * StarFive JH71XX PMU (Power Management Unit) Controller Driver
+ *
+ * Copyright (C) 2022-2023 StarFive Technology Co., Ltd.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/iopoll.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/pm_domain.h>
+#include <dt-bindings/power/starfive,jh7110-pmu.h>
+
+/* register offset */
+#define JH71XX_PMU_SW_TURN_ON_POWER	0x0C
+#define JH71XX_PMU_SW_TURN_OFF_POWER	0x10
+#define JH71XX_PMU_SW_ENCOURAGE		0x44
+#define JH71XX_PMU_TIMER_INT_MASK	0x48
+#define JH71XX_PMU_CURR_POWER_MODE	0x80
+#define JH71XX_PMU_EVENT_STATUS		0x88
+#define JH71XX_PMU_INT_STATUS		0x8C
+
+/* aon pmu register offset */
+#define JH71XX_AON_PMU_SWITCH		0x00
+
+/* sw encourage cfg */
+#define JH71XX_PMU_SW_ENCOURAGE_EN_LO	0x05
+#define JH71XX_PMU_SW_ENCOURAGE_EN_HI	0x50
+#define JH71XX_PMU_SW_ENCOURAGE_DIS_LO	0x0A
+#define JH71XX_PMU_SW_ENCOURAGE_DIS_HI	0xA0
+#define JH71XX_PMU_SW_ENCOURAGE_ON	0xFF
+
+/* pmu int status */
+#define JH71XX_PMU_INT_SEQ_DONE		BIT(0)
+#define JH71XX_PMU_INT_HW_REQ		BIT(1)
+#define JH71XX_PMU_INT_SW_FAIL		GENMASK(3, 2)
+#define JH71XX_PMU_INT_HW_FAIL		GENMASK(5, 4)
+#define JH71XX_PMU_INT_PCH_FAIL		GENMASK(8, 6)
+#define JH71XX_PMU_INT_ALL_MASK		GENMASK(8, 0)
+
+/*
+ * The time required for switching power status is based on the time
+ * to turn on the largest domain's power, which is at microsecond level
+ */
+#define JH71XX_PMU_TIMEOUT_US		100
+
+struct jh71xx_domain_info {
+	const char * const name;
+	unsigned int flags;
+	u8 bit;
+};
+
+struct jh71xx_pmu;
+struct jh71xx_pmu_dev;
+
+struct jh71xx_pmu_match_data {
+	const struct jh71xx_domain_info *domain_info;
+	int num_domains;
+	unsigned int pmu_status;
+	int (*pmu_parse_irq)(struct platform_device *pdev,
+			     struct jh71xx_pmu *pmu);
+	int (*pmu_set_state)(struct jh71xx_pmu_dev *pmd,
+			     u32 mask, bool on);
+};
+
+struct jh71xx_pmu {
+	struct device *dev;
+	const struct jh71xx_pmu_match_data *match_data;
+	void __iomem *base;
+	struct generic_pm_domain **genpd;
+	struct genpd_onecell_data genpd_data;
+	int irq;
+	spinlock_t lock;	/* protects pmu reg */
+};
+
+struct jh71xx_pmu_dev {
+	const struct jh71xx_domain_info *domain_info;
+	struct jh71xx_pmu *pmu;
+	struct generic_pm_domain genpd;
+};
+
+static int jh71xx_pmu_get_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool *is_on)
+{
+	struct jh71xx_pmu *pmu = pmd->pmu;
+
+	if (!mask)
+		return -EINVAL;
+
+	*is_on = readl(pmu->base + pmu->match_data->pmu_status) & mask;
+
+	return 0;
+}
+
+static int jh7110_pmu_set_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool on)
+{
+	struct jh71xx_pmu *pmu = pmd->pmu;
+	unsigned long flags;
+	u32 val;
+	u32 mode;
+	u32 encourage_lo;
+	u32 encourage_hi;
+	int ret;
+
+	spin_lock_irqsave(&pmu->lock, flags);
+
+	/*
+	 * The PMU accepts software encourage to switch power mode in the following 2 steps:
+	 *
+	 * 1.Configure the register SW_TURN_ON_POWER (offset 0x0c) by writing 1 to
+	 *   the bit corresponding to the power domain that will be turned on
+	 *   and writing 0 to the others.
+	 *   Likewise, configure the register SW_TURN_OFF_POWER (offset 0x10) by
+	 *   writing 1 to the bit corresponding to the power domain that will be
+	 *   turned off and writing 0 to the others.
+	 */
+	if (on) {
+		mode = JH71XX_PMU_SW_TURN_ON_POWER;
+		encourage_lo = JH71XX_PMU_SW_ENCOURAGE_EN_LO;
+		encourage_hi = JH71XX_PMU_SW_ENCOURAGE_EN_HI;
+	} else {
+		mode = JH71XX_PMU_SW_TURN_OFF_POWER;
+		encourage_lo = JH71XX_PMU_SW_ENCOURAGE_DIS_LO;
+		encourage_hi = JH71XX_PMU_SW_ENCOURAGE_DIS_HI;
+	}
+
+	writel(mask, pmu->base + mode);
+
+	/*
+	 * 2.Write SW encourage command sequence to the Software Encourage Reg (offset 0x44)
+	 *   First write SW_MODE_ENCOURAGE_ON to JH71XX_PMU_SW_ENCOURAGE. This will reset
+	 *   the state machine which parses the command sequence. This register must be
+	 *   written every time software wants to power on/off a domain.
+	 *   Then write the lower bits of the command sequence, followed by the upper
+	 *   bits. The sequence differs between powering on & off a domain.
+	 */
+	writel(JH71XX_PMU_SW_ENCOURAGE_ON, pmu->base + JH71XX_PMU_SW_ENCOURAGE);
+	writel(encourage_lo, pmu->base + JH71XX_PMU_SW_ENCOURAGE);
+	writel(encourage_hi, pmu->base + JH71XX_PMU_SW_ENCOURAGE);
+
+	spin_unlock_irqrestore(&pmu->lock, flags);
+
+	/* Wait for the power domain bit to be enabled / disabled */
+	if (on) {
+		ret = readl_poll_timeout_atomic(pmu->base + JH71XX_PMU_CURR_POWER_MODE,
+						val, val & mask,
+						1, JH71XX_PMU_TIMEOUT_US);
+	} else {
+		ret = readl_poll_timeout_atomic(pmu->base + JH71XX_PMU_CURR_POWER_MODE,
+						val, !(val & mask),
+						1, JH71XX_PMU_TIMEOUT_US);
+	}
+
+	if (ret) {
+		dev_err(pmu->dev, "%s: failed to power %s\n",
+			pmd->genpd.name, on ? "on" : "off");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static int jh7110_aon_pmu_set_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool on)
+{
+	struct jh71xx_pmu *pmu = pmd->pmu;
+	unsigned long flags;
+	u32 val;
+
+	spin_lock_irqsave(&pmu->lock, flags);
+	val = readl(pmu->base + JH71XX_AON_PMU_SWITCH);
+
+	if (on)
+		val |= mask;
+	else
+		val &= ~mask;
+
+	writel(val, pmu->base + JH71XX_AON_PMU_SWITCH);
+	spin_unlock_irqrestore(&pmu->lock, flags);
+
+	return 0;
+}
+
+static int jh71xx_pmu_set_state(struct jh71xx_pmu_dev *pmd, u32 mask, bool on)
+{
+	struct jh71xx_pmu *pmu = pmd->pmu;
+	const struct jh71xx_pmu_match_data *match_data = pmu->match_data;
+	bool is_on;
+	int ret;
+
+	ret = jh71xx_pmu_get_state(pmd, mask, &is_on);
+	if (ret) {
+		dev_dbg(pmu->dev, "unable to get current state for %s\n",
+			pmd->genpd.name);
+		return ret;
+	}
+
+	if (is_on == on) {
+		dev_dbg(pmu->dev, "pm domain [%s] is already %sable status.\n",
+			pmd->genpd.name, on ? "en" : "dis");
+		return 0;
+	}
+
+	return match_data->pmu_set_state(pmd, mask, on);
+}
+
+static int jh71xx_pmu_on(struct generic_pm_domain *genpd)
+{
+	struct jh71xx_pmu_dev *pmd = container_of(genpd,
+						  struct jh71xx_pmu_dev, genpd);
+	u32 pwr_mask = BIT(pmd->domain_info->bit);
+
+	return jh71xx_pmu_set_state(pmd, pwr_mask, true);
+}
+
+static int jh71xx_pmu_off(struct generic_pm_domain *genpd)
+{
+	struct jh71xx_pmu_dev *pmd = container_of(genpd,
+						  struct jh71xx_pmu_dev, genpd);
+	u32 pwr_mask = BIT(pmd->domain_info->bit);
+
+	return jh71xx_pmu_set_state(pmd, pwr_mask, false);
+}
+
+static void jh71xx_pmu_int_enable(struct jh71xx_pmu *pmu, u32 mask, bool enable)
+{
+	u32 val;
+	unsigned long flags;
+
+	spin_lock_irqsave(&pmu->lock, flags);
+	val = readl(pmu->base + JH71XX_PMU_TIMER_INT_MASK);
+
+	if (enable)
+		val &= ~mask;
+	else
+		val |= mask;
+
+	writel(val, pmu->base + JH71XX_PMU_TIMER_INT_MASK);
+	spin_unlock_irqrestore(&pmu->lock, flags);
+}
+
+static irqreturn_t jh71xx_pmu_interrupt(int irq, void *data)
+{
+	struct jh71xx_pmu *pmu = data;
+	u32 val;
+
+	val = readl(pmu->base + JH71XX_PMU_INT_STATUS);
+
+	if (val & JH71XX_PMU_INT_SEQ_DONE)
+		dev_dbg(pmu->dev, "sequence done.\n");
+	if (val & JH71XX_PMU_INT_HW_REQ)
+		dev_dbg(pmu->dev, "hardware encourage requestion.\n");
+	if (val & JH71XX_PMU_INT_SW_FAIL)
+		dev_err(pmu->dev, "software encourage fail.\n");
+	if (val & JH71XX_PMU_INT_HW_FAIL)
+		dev_err(pmu->dev, "hardware encourage fail.\n");
+	if (val & JH71XX_PMU_INT_PCH_FAIL)
+		dev_err(pmu->dev, "p-channel fail event.\n");
+
+	/* clear interrupts */
+	writel(val, pmu->base + JH71XX_PMU_INT_STATUS);
+	writel(val, pmu->base + JH71XX_PMU_EVENT_STATUS);
+
+	return IRQ_HANDLED;
+}
+
+static int jh7110_pmu_parse_irq(struct platform_device *pdev, struct jh71xx_pmu *pmu)
+{
+	struct device *dev = &pdev->dev;
+	int ret;
+
+	pmu->irq = platform_get_irq(pdev, 0);
+	if (pmu->irq < 0)
+		return pmu->irq;
+
+	ret = devm_request_irq(dev, pmu->irq, jh71xx_pmu_interrupt,
+			       0, pdev->name, pmu);
+	if (ret)
+		dev_err(dev, "failed to request irq\n");
+
+	jh71xx_pmu_int_enable(pmu, JH71XX_PMU_INT_ALL_MASK & ~JH71XX_PMU_INT_PCH_FAIL, true);
+
+	return 0;
+}
+
+static int jh71xx_pmu_init_domain(struct jh71xx_pmu *pmu, int index)
+{
+	struct jh71xx_pmu_dev *pmd;
+	u32 pwr_mask;
+	int ret;
+	bool is_on = false;
+
+	pmd = devm_kzalloc(pmu->dev, sizeof(*pmd), GFP_KERNEL);
+	if (!pmd)
+		return -ENOMEM;
+
+	pmd->domain_info = &pmu->match_data->domain_info[index];
+	pmd->pmu = pmu;
+	pwr_mask = BIT(pmd->domain_info->bit);
+
+	pmd->genpd.name = pmd->domain_info->name;
+	pmd->genpd.flags = pmd->domain_info->flags;
+
+	ret = jh71xx_pmu_get_state(pmd, pwr_mask, &is_on);
+	if (ret)
+		dev_warn(pmu->dev, "unable to get current state for %s\n",
+			 pmd->genpd.name);
+
+	pmd->genpd.power_on = jh71xx_pmu_on;
+	pmd->genpd.power_off = jh71xx_pmu_off;
+	pm_genpd_init(&pmd->genpd, NULL, !is_on);
+
+	pmu->genpd_data.domains[index] = &pmd->genpd;
+
+	return 0;
+}
+
+static int jh71xx_pmu_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *np = dev->of_node;
+	const struct jh71xx_pmu_match_data *match_data;
+	struct jh71xx_pmu *pmu;
+	unsigned int i;
+	int ret;
+
+	pmu = devm_kzalloc(dev, sizeof(*pmu), GFP_KERNEL);
+	if (!pmu)
+		return -ENOMEM;
+
+	pmu->base = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(pmu->base))
+		return PTR_ERR(pmu->base);
+
+	spin_lock_init(&pmu->lock);
+
+	match_data = of_device_get_match_data(dev);
+	if (!match_data)
+		return -EINVAL;
+
+	if (match_data->pmu_parse_irq) {
+		ret = match_data->pmu_parse_irq(pdev, pmu);
+		if (ret) {
+			dev_err(dev, "failed to parse irq\n");
+			return ret;
+		}
+	}
+
+	pmu->genpd = devm_kcalloc(dev, match_data->num_domains,
+				  sizeof(struct generic_pm_domain *),
+				  GFP_KERNEL);
+	if (!pmu->genpd)
+		return -ENOMEM;
+
+	pmu->dev = dev;
+	pmu->match_data = match_data;
+	pmu->genpd_data.domains = pmu->genpd;
+	pmu->genpd_data.num_domains = match_data->num_domains;
+
+	for (i = 0; i < match_data->num_domains; i++) {
+		ret = jh71xx_pmu_init_domain(pmu, i);
+		if (ret) {
+			dev_err(dev, "failed to initialize power domain\n");
+			return ret;
+		}
+	}
+
+	ret = of_genpd_add_provider_onecell(np, &pmu->genpd_data);
+	if (ret) {
+		dev_err(dev, "failed to register genpd driver: %d\n", ret);
+		return ret;
+	}
+
+	dev_dbg(dev, "registered %u power domains\n", i);
+
+	return 0;
+}
+
+static const struct jh71xx_domain_info jh7110_power_domains[] = {
+	[JH7110_PD_SYSTOP] = {
+		.name = "SYSTOP",
+		.bit = 0,
+		.flags = GENPD_FLAG_ALWAYS_ON,
+	},
+	[JH7110_PD_CPU] = {
+		.name = "CPU",
+		.bit = 1,
+		.flags = GENPD_FLAG_ALWAYS_ON,
+	},
+	[JH7110_PD_GPUA] = {
+		.name = "GPUA",
+		.bit = 2,
+	},
+	[JH7110_PD_VDEC] = {
+		.name = "VDEC",
+		.bit = 3,
+	},
+	[JH7110_PD_VOUT] = {
+		.name = "VOUT",
+		.bit = 4,
+	},
+	[JH7110_PD_ISP] = {
+		.name = "ISP",
+		.bit = 5,
+	},
+	[JH7110_PD_VENC] = {
+		.name = "VENC",
+		.bit = 6,
+	},
+};
+
+static const struct jh71xx_pmu_match_data jh7110_pmu = {
+	.num_domains = ARRAY_SIZE(jh7110_power_domains),
+	.domain_info = jh7110_power_domains,
+	.pmu_status = JH71XX_PMU_CURR_POWER_MODE,
+	.pmu_parse_irq = jh7110_pmu_parse_irq,
+	.pmu_set_state = jh7110_pmu_set_state,
+};
+
+static const struct jh71xx_domain_info jh7110_aon_power_domains[] = {
+	[JH7110_AON_PD_DPHY_TX] = {
+		.name = "DPHY-TX",
+		.bit = 30,
+	},
+	[JH7110_AON_PD_DPHY_RX] = {
+		.name = "DPHY-RX",
+		.bit = 31,
+	},
+};
+
+static const struct jh71xx_pmu_match_data jh7110_aon_pmu = {
+	.num_domains = ARRAY_SIZE(jh7110_aon_power_domains),
+	.domain_info = jh7110_aon_power_domains,
+	.pmu_status = JH71XX_AON_PMU_SWITCH,
+	.pmu_set_state = jh7110_aon_pmu_set_state,
+};
+
+static const struct of_device_id jh71xx_pmu_of_match[] = {
+	{
+		.compatible = "starfive,jh7110-pmu",
+		.data = (void *)&jh7110_pmu,
+	}, {
+		.compatible = "starfive,jh7110-aon-syscon",
+		.data = (void *)&jh7110_aon_pmu,
+	}, {
+		/* sentinel */
+	}
+};
+
+static struct platform_driver jh71xx_pmu_driver = {
+	.probe = jh71xx_pmu_probe,
+	.driver = {
+		.name = "jh71xx-pmu",
+		.of_match_table = jh71xx_pmu_of_match,
+		.suppress_bind_attrs = true,
+	},
+};
+builtin_platform_driver(jh71xx_pmu_driver);
+
+MODULE_AUTHOR("Walker Chen <walker.chen@starfivetech.com>");
+MODULE_AUTHOR("Changhuang Liang <changhuang.liang@starfivetech.com>");
+MODULE_DESCRIPTION("StarFive JH71XX PMU Driver");
+MODULE_LICENSE("GPL");