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

4 files changed, 346 insertions, 0 deletions

diff --git a/arch/m68k/virt/Makefile b/arch/m68k/virt/Makefile
new file mode 100644
index 000000000..54b9b2866
--- /dev/null
+++ b/arch/m68k/virt/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Makefile for Linux arch/m68k/virt source directory
+#
+
+obj-y		:= config.o ints.o platform.o
diff --git a/arch/m68k/virt/config.c b/arch/m68k/virt/config.c
new file mode 100644
index 000000000..b338e2a8d
--- /dev/null
+++ b/arch/m68k/virt/config.c
@@ -0,0 +1,92 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/reboot.h>
+#include <linux/serial_core.h>
+#include <clocksource/timer-goldfish.h>
+
+#include <asm/bootinfo.h>
+#include <asm/bootinfo-virt.h>
+#include <asm/byteorder.h>
+#include <asm/machdep.h>
+#include <asm/virt.h>
+#include <asm/config.h>
+
+struct virt_booter_data virt_bi_data;
+
+static void virt_get_model(char *str)
+{
+	/* str is 80 characters long */
+	sprintf(str, "QEMU Virtual M68K Machine (%u.%u.%u)",
+		(u8)(virt_bi_data.qemu_version >> 24),
+		(u8)(virt_bi_data.qemu_version >> 16),
+		(u8)(virt_bi_data.qemu_version >> 8));
+}
+static void virt_reset(void)
+{
+	do_kernel_restart(NULL);
+}
+
+/*
+ * Parse a virtual-m68k-specific record in the bootinfo
+ */
+
+int __init virt_parse_bootinfo(const struct bi_record *record)
+{
+	int unknown = 0;
+	const void *data = record->data;
+
+	switch (be16_to_cpu(record->tag)) {
+	case BI_VIRT_QEMU_VERSION:
+		virt_bi_data.qemu_version = be32_to_cpup(data);
+		break;
+	case BI_VIRT_GF_PIC_BASE:
+		virt_bi_data.pic.mmio = be32_to_cpup(data);
+		data += 4;
+		virt_bi_data.pic.irq = be32_to_cpup(data);
+		break;
+	case BI_VIRT_GF_RTC_BASE:
+		virt_bi_data.rtc.mmio = be32_to_cpup(data);
+		data += 4;
+		virt_bi_data.rtc.irq = be32_to_cpup(data);
+		break;
+	case BI_VIRT_GF_TTY_BASE:
+		virt_bi_data.tty.mmio = be32_to_cpup(data);
+		data += 4;
+		virt_bi_data.tty.irq = be32_to_cpup(data);
+		break;
+	case BI_VIRT_CTRL_BASE:
+		virt_bi_data.ctrl.mmio = be32_to_cpup(data);
+		data += 4;
+		virt_bi_data.ctrl.irq = be32_to_cpup(data);
+		break;
+	case BI_VIRT_VIRTIO_BASE:
+		virt_bi_data.virtio.mmio = be32_to_cpup(data);
+		data += 4;
+		virt_bi_data.virtio.irq = be32_to_cpup(data);
+		break;
+	default:
+		unknown = 1;
+		break;
+	}
+	return unknown;
+}
+
+static void __init virt_sched_init(void)
+{
+	goldfish_timer_init(virt_bi_data.rtc.irq,
+			    (void __iomem *)virt_bi_data.rtc.mmio);
+}
+
+void __init config_virt(void)
+{
+	char earlycon[24];
+
+	snprintf(earlycon, sizeof(earlycon), "early_gf_tty,0x%08x",
+		 virt_bi_data.tty.mmio);
+	setup_earlycon(earlycon);
+
+	mach_init_IRQ = virt_init_IRQ;
+	mach_sched_init = virt_sched_init;
+	mach_get_model = virt_get_model;
+	mach_reset = virt_reset;
+}
diff --git a/arch/m68k/virt/ints.c b/arch/m68k/virt/ints.c
new file mode 100644
index 000000000..896aa6eb8
--- /dev/null
+++ b/arch/m68k/virt/ints.c
@@ -0,0 +1,154 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/delay.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/kernel.h>
+#include <linux/sched.h>
+#include <linux/sched/debug.h>
+#include <linux/types.h>
+#include <linux/ioport.h>
+
+#include <asm/hwtest.h>
+#include <asm/irq.h>
+#include <asm/irq_regs.h>
+#include <asm/processor.h>
+#include <asm/virt.h>
+
+#define GFPIC_REG_IRQ_PENDING           0x04
+#define GFPIC_REG_IRQ_DISABLE_ALL       0x08
+#define GFPIC_REG_IRQ_DISABLE           0x0c
+#define GFPIC_REG_IRQ_ENABLE            0x10
+
+static struct resource picres[6];
+static const char *picname[6] = {
+	"goldfish_pic.0",
+	"goldfish_pic.1",
+	"goldfish_pic.2",
+	"goldfish_pic.3",
+	"goldfish_pic.4",
+	"goldfish_pic.5"
+};
+
+/*
+ * 6 goldfish-pic for CPU IRQ #1 to IRQ #6
+ * CPU IRQ #1 -> PIC #1
+ *               IRQ #1 to IRQ #31 -> unused
+ *               IRQ #32 -> goldfish-tty
+ * CPU IRQ #2 -> PIC #2
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 1 to 32
+ * CPU IRQ #3 -> PIC #3
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 33 to 64
+ * CPU IRQ #4 -> PIC #4
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 65 to 96
+ * CPU IRQ #5 -> PIC #5
+ *               IRQ #1 to IRQ #32 -> virtio-mmio from 97 to 128
+ * CPU IRQ #6 -> PIC #6
+ *               IRQ #1 -> goldfish-timer
+ *               IRQ #2 -> goldfish-rtc
+ *               IRQ #3 to IRQ #32 -> unused
+ * CPU IRQ #7 -> NMI
+ */
+
+static u32 gfpic_read(int pic, int reg)
+{
+	void __iomem *base = (void __iomem *)(virt_bi_data.pic.mmio +
+					      pic * 0x1000);
+
+	return ioread32be(base + reg);
+}
+
+static void gfpic_write(u32 value, int pic, int reg)
+{
+	void __iomem *base = (void __iomem *)(virt_bi_data.pic.mmio +
+					      pic * 0x1000);
+
+	iowrite32be(value, base + reg);
+}
+
+#define GF_PIC(irq) ((irq - IRQ_USER) / 32)
+#define GF_IRQ(irq) ((irq - IRQ_USER) % 32)
+
+static void virt_irq_enable(struct irq_data *data)
+{
+	gfpic_write(BIT(GF_IRQ(data->irq)), GF_PIC(data->irq),
+		    GFPIC_REG_IRQ_ENABLE);
+}
+
+static void virt_irq_disable(struct irq_data *data)
+{
+	gfpic_write(BIT(GF_IRQ(data->irq)), GF_PIC(data->irq),
+		    GFPIC_REG_IRQ_DISABLE);
+}
+
+static unsigned int virt_irq_startup(struct irq_data *data)
+{
+	virt_irq_enable(data);
+	return 0;
+}
+
+static irqreturn_t virt_nmi_handler(int irq, void *dev_id)
+{
+	static int in_nmi;
+
+	if (READ_ONCE(in_nmi))
+		return IRQ_HANDLED;
+	WRITE_ONCE(in_nmi, 1);
+
+	pr_warn("Non-Maskable Interrupt\n");
+	show_registers(get_irq_regs());
+
+	WRITE_ONCE(in_nmi, 0);
+	return IRQ_HANDLED;
+}
+
+static struct irq_chip virt_irq_chip = {
+	.name		= "virt",
+	.irq_enable	= virt_irq_enable,
+	.irq_disable	= virt_irq_disable,
+	.irq_startup	= virt_irq_startup,
+	.irq_shutdown	= virt_irq_disable,
+};
+
+static void goldfish_pic_irq(struct irq_desc *desc)
+{
+	u32 irq_pending;
+	unsigned int irq_num;
+	unsigned int pic = desc->irq_data.irq - 1;
+
+	irq_pending = gfpic_read(pic, GFPIC_REG_IRQ_PENDING);
+	irq_num = IRQ_USER + pic * 32;
+
+	do {
+		if (irq_pending & 1)
+			generic_handle_irq(irq_num);
+		++irq_num;
+		irq_pending >>= 1;
+	} while (irq_pending);
+}
+
+void __init virt_init_IRQ(void)
+{
+	unsigned int i;
+
+	m68k_setup_irq_controller(&virt_irq_chip, handle_simple_irq, IRQ_USER,
+				  NUM_VIRT_SOURCES - IRQ_USER);
+
+	for (i = 0; i < 6; i++) {
+
+		picres[i] = (struct resource)
+		    DEFINE_RES_MEM_NAMED(virt_bi_data.pic.mmio + i * 0x1000,
+					 0x1000, picname[i]);
+		if (request_resource(&iomem_resource, &picres[i])) {
+			pr_err("Cannot allocate %s resource\n", picname[i]);
+			return;
+		}
+
+		irq_set_chained_handler(virt_bi_data.pic.irq + i,
+					goldfish_pic_irq);
+	}
+
+	if (request_irq(IRQ_AUTO_7, virt_nmi_handler, 0, "NMI",
+			virt_nmi_handler))
+		pr_err("Couldn't register NMI\n");
+}
diff --git a/arch/m68k/virt/platform.c b/arch/m68k/virt/platform.c
new file mode 100644
index 000000000..764f556b4
--- /dev/null
+++ b/arch/m68k/virt/platform.c
@@ -0,0 +1,94 @@
+// SPDX-License-Identifier: GPL-2.0
+
+#include <linux/platform_device.h>
+#include <linux/interrupt.h>
+#include <linux/memblock.h>
+#include <asm/virt.h>
+#include <asm/irq.h>
+
+#define VIRTIO_BUS_NB	128
+
+static struct platform_device * __init virt_virtio_init(unsigned int id)
+{
+	const struct resource res[] = {
+		DEFINE_RES_MEM(virt_bi_data.virtio.mmio + id * 0x200, 0x200),
+		DEFINE_RES_IRQ(virt_bi_data.virtio.irq + id),
+	};
+
+	return platform_device_register_simple("virtio-mmio", id,
+					       res, ARRAY_SIZE(res));
+}
+
+static int __init virt_platform_init(void)
+{
+	const struct resource goldfish_tty_res[] = {
+		DEFINE_RES_MEM(virt_bi_data.tty.mmio, 1),
+		DEFINE_RES_IRQ(virt_bi_data.tty.irq),
+	};
+	/* this is the second gf-rtc, the first one is used by the scheduler */
+	const struct resource goldfish_rtc_res[] = {
+		DEFINE_RES_MEM(virt_bi_data.rtc.mmio + 0x1000, 0x1000),
+		DEFINE_RES_IRQ(virt_bi_data.rtc.irq + 1),
+	};
+	const struct resource virt_ctrl_res[] = {
+		DEFINE_RES_MEM(virt_bi_data.ctrl.mmio, 0x100),
+	};
+	struct platform_device *pdev1, *pdev2, *pdev3;
+	struct platform_device *pdevs[VIRTIO_BUS_NB];
+	unsigned int i;
+	int ret = 0;
+
+	if (!MACH_IS_VIRT)
+		return -ENODEV;
+
+	/* We need this to have DMA'able memory provided to goldfish-tty */
+	min_low_pfn = 0;
+
+	pdev1 = platform_device_register_simple("goldfish_tty",
+						PLATFORM_DEVID_NONE,
+						goldfish_tty_res,
+						ARRAY_SIZE(goldfish_tty_res));
+	if (IS_ERR(pdev1))
+		return PTR_ERR(pdev1);
+
+	pdev2 = platform_device_register_simple("goldfish_rtc",
+						PLATFORM_DEVID_NONE,
+						goldfish_rtc_res,
+						ARRAY_SIZE(goldfish_rtc_res));
+	if (IS_ERR(pdev2)) {
+		ret = PTR_ERR(pdev2);
+		goto err_unregister_tty;
+	}
+
+	pdev3 = platform_device_register_simple("qemu-virt-ctrl",
+						PLATFORM_DEVID_NONE,
+						virt_ctrl_res,
+						ARRAY_SIZE(virt_ctrl_res));
+	if (IS_ERR(pdev3)) {
+		ret = PTR_ERR(pdev3);
+		goto err_unregister_rtc;
+	}
+
+	for (i = 0; i < VIRTIO_BUS_NB; i++) {
+		pdevs[i] = virt_virtio_init(i);
+		if (IS_ERR(pdevs[i])) {
+			ret = PTR_ERR(pdevs[i]);
+			goto err_unregister_virtio;
+		}
+	}
+
+	return 0;
+
+err_unregister_virtio:
+	while (i > 0)
+		platform_device_unregister(pdevs[--i]);
+	platform_device_unregister(pdev3);
+err_unregister_rtc:
+	platform_device_unregister(pdev2);
+err_unregister_tty:
+	platform_device_unregister(pdev1);
+
+	return ret;
+}
+
+arch_initcall(virt_platform_init);