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, 277 insertions, 0 deletions

diff --git a/drivers/clocksource/timer-clint.c b/drivers/clocksource/timer-clint.c
new file mode 100644
index 000000000..0bdd9d7ec
--- /dev/null
+++ b/drivers/clocksource/timer-clint.c
@@ -0,0 +1,277 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2020 Western Digital Corporation or its affiliates.
+ *
+ * Most of the M-mode (i.e. NoMMU) RISC-V systems usually have a
+ * CLINT MMIO timer device.
+ */
+
+#define pr_fmt(fmt) "clint: " fmt
+#include <linux/bitops.h>
+#include <linux/clocksource.h>
+#include <linux/clockchips.h>
+#include <linux/cpu.h>
+#include <linux/delay.h>
+#include <linux/module.h>
+#include <linux/of_address.h>
+#include <linux/sched_clock.h>
+#include <linux/io-64-nonatomic-lo-hi.h>
+#include <linux/interrupt.h>
+#include <linux/irq.h>
+#include <linux/irqchip/chained_irq.h>
+#include <linux/irqdomain.h>
+#include <linux/of_irq.h>
+#include <linux/smp.h>
+#include <linux/timex.h>
+
+#ifndef CONFIG_RISCV_M_MODE
+#include <asm/clint.h>
+#endif
+
+#define CLINT_IPI_OFF		0
+#define CLINT_TIMER_CMP_OFF	0x4000
+#define CLINT_TIMER_VAL_OFF	0xbff8
+
+/* CLINT manages IPI and Timer for RISC-V M-mode  */
+static u32 __iomem *clint_ipi_base;
+static unsigned int clint_ipi_irq;
+static u64 __iomem *clint_timer_cmp;
+static u64 __iomem *clint_timer_val;
+static unsigned long clint_timer_freq;
+static unsigned int clint_timer_irq;
+
+#ifdef CONFIG_RISCV_M_MODE
+u64 __iomem *clint_time_val;
+EXPORT_SYMBOL(clint_time_val);
+#endif
+
+#ifdef CONFIG_SMP
+static void clint_send_ipi(unsigned int cpu)
+{
+	writel(1, clint_ipi_base + cpuid_to_hartid_map(cpu));
+}
+
+static void clint_clear_ipi(void)
+{
+	writel(0, clint_ipi_base + cpuid_to_hartid_map(smp_processor_id()));
+}
+
+static void clint_ipi_interrupt(struct irq_desc *desc)
+{
+	struct irq_chip *chip = irq_desc_get_chip(desc);
+
+	chained_irq_enter(chip, desc);
+
+	clint_clear_ipi();
+	ipi_mux_process();
+
+	chained_irq_exit(chip, desc);
+}
+#endif
+
+#ifdef CONFIG_64BIT
+#define clint_get_cycles()	readq_relaxed(clint_timer_val)
+#else
+#define clint_get_cycles()	readl_relaxed(clint_timer_val)
+#define clint_get_cycles_hi()	readl_relaxed(((u32 *)clint_timer_val) + 1)
+#endif
+
+#ifdef CONFIG_64BIT
+static u64 notrace clint_get_cycles64(void)
+{
+	return clint_get_cycles();
+}
+#else /* CONFIG_64BIT */
+static u64 notrace clint_get_cycles64(void)
+{
+	u32 hi, lo;
+
+	do {
+		hi = clint_get_cycles_hi();
+		lo = clint_get_cycles();
+	} while (hi != clint_get_cycles_hi());
+
+	return ((u64)hi << 32) | lo;
+}
+#endif /* CONFIG_64BIT */
+
+static u64 clint_rdtime(struct clocksource *cs)
+{
+	return clint_get_cycles64();
+}
+
+static struct clocksource clint_clocksource = {
+	.name		= "clint_clocksource",
+	.rating		= 300,
+	.mask		= CLOCKSOURCE_MASK(64),
+	.flags		= CLOCK_SOURCE_IS_CONTINUOUS,
+	.read		= clint_rdtime,
+};
+
+static int clint_clock_next_event(unsigned long delta,
+				   struct clock_event_device *ce)
+{
+	void __iomem *r = clint_timer_cmp +
+			  cpuid_to_hartid_map(smp_processor_id());
+
+	csr_set(CSR_IE, IE_TIE);
+	writeq_relaxed(clint_get_cycles64() + delta, r);
+	return 0;
+}
+
+static DEFINE_PER_CPU(struct clock_event_device, clint_clock_event) = {
+	.name		= "clint_clockevent",
+	.features	= CLOCK_EVT_FEAT_ONESHOT,
+	.rating		= 100,
+	.set_next_event	= clint_clock_next_event,
+};
+
+static int clint_timer_starting_cpu(unsigned int cpu)
+{
+	struct clock_event_device *ce = per_cpu_ptr(&clint_clock_event, cpu);
+
+	ce->cpumask = cpumask_of(cpu);
+	clockevents_config_and_register(ce, clint_timer_freq, 100, ULONG_MAX);
+
+	enable_percpu_irq(clint_timer_irq,
+			  irq_get_trigger_type(clint_timer_irq));
+	enable_percpu_irq(clint_ipi_irq,
+			  irq_get_trigger_type(clint_ipi_irq));
+	return 0;
+}
+
+static int clint_timer_dying_cpu(unsigned int cpu)
+{
+	disable_percpu_irq(clint_timer_irq);
+	/*
+	 * Don't disable IPI when CPU goes offline because
+	 * the masking/unmasking of virtual IPIs is done
+	 * via generic IPI-Mux
+	 */
+	return 0;
+}
+
+static irqreturn_t clint_timer_interrupt(int irq, void *dev_id)
+{
+	struct clock_event_device *evdev = this_cpu_ptr(&clint_clock_event);
+
+	csr_clear(CSR_IE, IE_TIE);
+	evdev->event_handler(evdev);
+
+	return IRQ_HANDLED;
+}
+
+static int __init clint_timer_init_dt(struct device_node *np)
+{
+	int rc;
+	u32 i, nr_irqs;
+	void __iomem *base;
+	struct of_phandle_args oirq;
+
+	/*
+	 * Ensure that CLINT device interrupts are either RV_IRQ_TIMER or
+	 * RV_IRQ_SOFT. If it's anything else then we ignore the device.
+	 */
+	nr_irqs = of_irq_count(np);
+	for (i = 0; i < nr_irqs; i++) {
+		if (of_irq_parse_one(np, i, &oirq)) {
+			pr_err("%pOFP: failed to parse irq %d.\n", np, i);
+			continue;
+		}
+
+		if ((oirq.args_count != 1) ||
+		    (oirq.args[0] != RV_IRQ_TIMER &&
+		     oirq.args[0] != RV_IRQ_SOFT)) {
+			pr_err("%pOFP: invalid irq %d (hwirq %d)\n",
+			       np, i, oirq.args[0]);
+			return -ENODEV;
+		}
+
+		/* Find parent irq domain and map ipi irq */
+		if (!clint_ipi_irq &&
+		    oirq.args[0] == RV_IRQ_SOFT &&
+		    irq_find_host(oirq.np))
+			clint_ipi_irq = irq_of_parse_and_map(np, i);
+
+		/* Find parent irq domain and map timer irq */
+		if (!clint_timer_irq &&
+		    oirq.args[0] == RV_IRQ_TIMER &&
+		    irq_find_host(oirq.np))
+			clint_timer_irq = irq_of_parse_and_map(np, i);
+	}
+
+	/* If CLINT ipi or timer irq not found then fail */
+	if (!clint_ipi_irq || !clint_timer_irq) {
+		pr_err("%pOFP: ipi/timer irq not found\n", np);
+		return -ENODEV;
+	}
+
+	base = of_iomap(np, 0);
+	if (!base) {
+		pr_err("%pOFP: could not map registers\n", np);
+		return -ENODEV;
+	}
+
+	clint_ipi_base = base + CLINT_IPI_OFF;
+	clint_timer_cmp = base + CLINT_TIMER_CMP_OFF;
+	clint_timer_val = base + CLINT_TIMER_VAL_OFF;
+	clint_timer_freq = riscv_timebase;
+
+#ifdef CONFIG_RISCV_M_MODE
+	/*
+	 * Yes, that's an odd naming scheme.  time_val is public, but hopefully
+	 * will die in favor of something cleaner.
+	 */
+	clint_time_val = clint_timer_val;
+#endif
+
+	pr_info("%pOFP: timer running at %ld Hz\n", np, clint_timer_freq);
+
+	rc = clocksource_register_hz(&clint_clocksource, clint_timer_freq);
+	if (rc) {
+		pr_err("%pOFP: clocksource register failed [%d]\n", np, rc);
+		goto fail_iounmap;
+	}
+
+	sched_clock_register(clint_get_cycles64, 64, clint_timer_freq);
+
+	rc = request_percpu_irq(clint_timer_irq, clint_timer_interrupt,
+				 "clint-timer", &clint_clock_event);
+	if (rc) {
+		pr_err("registering percpu irq failed [%d]\n", rc);
+		goto fail_iounmap;
+	}
+
+#ifdef CONFIG_SMP
+	rc = ipi_mux_create(BITS_PER_BYTE, clint_send_ipi);
+	if (rc <= 0) {
+		pr_err("unable to create muxed IPIs\n");
+		rc = (rc < 0) ? rc : -ENODEV;
+		goto fail_free_irq;
+	}
+
+	irq_set_chained_handler(clint_ipi_irq, clint_ipi_interrupt);
+	riscv_ipi_set_virq_range(rc, BITS_PER_BYTE);
+	clint_clear_ipi();
+#endif
+
+	rc = cpuhp_setup_state(CPUHP_AP_CLINT_TIMER_STARTING,
+				"clockevents/clint/timer:starting",
+				clint_timer_starting_cpu,
+				clint_timer_dying_cpu);
+	if (rc) {
+		pr_err("%pOFP: cpuhp setup state failed [%d]\n", np, rc);
+		goto fail_free_irq;
+	}
+
+	return 0;
+
+fail_free_irq:
+	free_percpu_irq(clint_timer_irq, &clint_clock_event);
+fail_iounmap:
+	iounmap(base);
+	return rc;
+}
+
+TIMER_OF_DECLARE(clint_timer, "riscv,clint0", clint_timer_init_dt);
+TIMER_OF_DECLARE(clint_timer1, "sifive,clint0", clint_timer_init_dt);