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, 216 insertions, 0 deletions

diff --git a/drivers/clocksource/timer-econet-en751221.c b/drivers/clocksource/timer-econet-en751221.c
new file mode 100644
index 000000000..4008076b1
--- /dev/null
+++ b/drivers/clocksource/timer-econet-en751221.c
@@ -0,0 +1,216 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Timer present on EcoNet EN75xx MIPS based SoCs.
+ *
+ * Copyright (C) 2025 by Caleb James DeLisle <cjd@cjdns.fr>
+ */
+
+#include <linux/io.h>
+#include <linux/cpumask.h>
+#include <linux/interrupt.h>
+#include <linux/clockchips.h>
+#include <linux/sched_clock.h>
+#include <linux/of.h>
+#include <linux/of_irq.h>
+#include <linux/of_address.h>
+#include <linux/cpuhotplug.h>
+#include <linux/clk.h>
+
+#define ECONET_BITS			32
+#define ECONET_MIN_DELTA		0x00001000
+#define ECONET_MAX_DELTA		GENMASK(ECONET_BITS - 2, 0)
+/* 34Kc hardware has 1 block and 1004Kc has 2. */
+#define ECONET_NUM_BLOCKS		DIV_ROUND_UP(NR_CPUS, 2)
+
+static struct {
+	void __iomem	*membase[ECONET_NUM_BLOCKS];
+	u32		freq_hz;
+} econet_timer __ro_after_init;
+
+static DEFINE_PER_CPU(struct clock_event_device, econet_timer_pcpu);
+
+/* Each memory block has 2 timers, the order of registers is:
+ * CTL, CMR0, CNT0, CMR1, CNT1
+ */
+static inline void __iomem *reg_ctl(u32 timer_n)
+{
+	return econet_timer.membase[timer_n >> 1];
+}
+
+static inline void __iomem *reg_compare(u32 timer_n)
+{
+	return econet_timer.membase[timer_n >> 1] + (timer_n & 1) * 0x08 + 0x04;
+}
+
+static inline void __iomem *reg_count(u32 timer_n)
+{
+	return econet_timer.membase[timer_n >> 1] + (timer_n & 1) * 0x08 + 0x08;
+}
+
+static inline u32 ctl_bit_enabled(u32 timer_n)
+{
+	return 1U << (timer_n & 1);
+}
+
+static inline u32 ctl_bit_pending(u32 timer_n)
+{
+	return 1U << ((timer_n & 1) + 16);
+}
+
+static bool cevt_is_pending(int cpu_id)
+{
+	return ioread32(reg_ctl(cpu_id)) & ctl_bit_pending(cpu_id);
+}
+
+static irqreturn_t cevt_interrupt(int irq, void *dev_id)
+{
+	struct clock_event_device *dev = this_cpu_ptr(&econet_timer_pcpu);
+	int cpu = cpumask_first(dev->cpumask);
+
+	/* Each VPE has its own events,
+	 * so this will only happen on spurious interrupt.
+	 */
+	if (!cevt_is_pending(cpu))
+		return IRQ_NONE;
+
+	iowrite32(ioread32(reg_count(cpu)), reg_compare(cpu));
+	dev->event_handler(dev);
+	return IRQ_HANDLED;
+}
+
+static int cevt_set_next_event(ulong delta, struct clock_event_device *dev)
+{
+	u32 next;
+	int cpu;
+
+	cpu = cpumask_first(dev->cpumask);
+	next = ioread32(reg_count(cpu)) + delta;
+	iowrite32(next, reg_compare(cpu));
+
+	if ((s32)(next - ioread32(reg_count(cpu))) < ECONET_MIN_DELTA / 2)
+		return -ETIME;
+
+	return 0;
+}
+
+static int cevt_init_cpu(uint cpu)
+{
+	struct clock_event_device *cd = &per_cpu(econet_timer_pcpu, cpu);
+	u32 reg;
+
+	pr_debug("%s: Setting up clockevent for CPU %d\n", cd->name, cpu);
+
+	reg = ioread32(reg_ctl(cpu)) | ctl_bit_enabled(cpu);
+	iowrite32(reg, reg_ctl(cpu));
+
+	enable_percpu_irq(cd->irq, IRQ_TYPE_NONE);
+
+	/* Do this last because it synchronously configures the timer */
+	clockevents_config_and_register(cd, econet_timer.freq_hz,
+					ECONET_MIN_DELTA, ECONET_MAX_DELTA);
+
+	return 0;
+}
+
+static u64 notrace sched_clock_read(void)
+{
+	/* Always read from clock zero no matter the CPU */
+	return (u64)ioread32(reg_count(0));
+}
+
+/* Init */
+
+static void __init cevt_dev_init(uint cpu)
+{
+	iowrite32(0, reg_count(cpu));
+	iowrite32(U32_MAX, reg_compare(cpu));
+}
+
+static int __init cevt_init(struct device_node *np)
+{
+	int i, irq, ret;
+
+	irq = irq_of_parse_and_map(np, 0);
+	if (irq <= 0) {
+		pr_err("%pOFn: irq_of_parse_and_map failed", np);
+		return -EINVAL;
+	}
+
+	ret = request_percpu_irq(irq, cevt_interrupt, np->name, &econet_timer_pcpu);
+
+	if (ret < 0) {
+		pr_err("%pOFn: IRQ %d setup failed (%d)\n", np, irq, ret);
+		goto err_unmap_irq;
+	}
+
+	for_each_possible_cpu(i) {
+		struct clock_event_device *cd = &per_cpu(econet_timer_pcpu, i);
+
+		cd->rating		= 310;
+		cd->features		= CLOCK_EVT_FEAT_ONESHOT |
+					  CLOCK_EVT_FEAT_C3STOP |
+					  CLOCK_EVT_FEAT_PERCPU;
+		cd->set_next_event	= cevt_set_next_event;
+		cd->irq			= irq;
+		cd->cpumask		= cpumask_of(i);
+		cd->name		= np->name;
+
+		cevt_dev_init(i);
+	}
+
+	cpuhp_setup_state(CPUHP_AP_ONLINE_DYN,
+			  "clockevents/econet/timer:starting",
+			  cevt_init_cpu, NULL);
+	return 0;
+
+err_unmap_irq:
+	irq_dispose_mapping(irq);
+	return ret;
+}
+
+static int __init timer_init(struct device_node *np)
+{
+	int num_blocks = DIV_ROUND_UP(num_possible_cpus(), 2);
+	struct clk *clk;
+	int ret;
+
+	clk = of_clk_get(np, 0);
+	if (IS_ERR(clk)) {
+		pr_err("%pOFn: Failed to get CPU clock from DT %ld\n", np, PTR_ERR(clk));
+		return PTR_ERR(clk);
+	}
+
+	econet_timer.freq_hz = clk_get_rate(clk);
+
+	for (int i = 0; i < num_blocks; i++) {
+		econet_timer.membase[i] = of_iomap(np, i);
+		if (!econet_timer.membase[i]) {
+			pr_err("%pOFn: failed to map register [%d]\n", np, i);
+			return -ENXIO;
+		}
+	}
+
+	/* For clocksource purposes always read clock zero, whatever the CPU */
+	ret = clocksource_mmio_init(reg_count(0), np->name,
+				    econet_timer.freq_hz, 301, ECONET_BITS,
+				    clocksource_mmio_readl_up);
+	if (ret) {
+		pr_err("%pOFn: clocksource_mmio_init failed: %d", np, ret);
+		return ret;
+	}
+
+	ret = cevt_init(np);
+	if (ret < 0)
+		return ret;
+
+	sched_clock_register(sched_clock_read, ECONET_BITS,
+			     econet_timer.freq_hz);
+
+	pr_info("%pOFn: using %u.%03u MHz high precision timer\n", np,
+		econet_timer.freq_hz / 1000000,
+		(econet_timer.freq_hz / 1000) % 1000);
+
+	return 0;
+}
+
+TIMER_OF_DECLARE(econet_timer_hpt, "econet,en751221-timer", timer_init);