From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions 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) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- drivers/clocksource/timer-ti-32k.c | 165 +++++++++++++++++++++++++++++++++++++ 1 file changed, 165 insertions(+) create mode 100644 drivers/clocksource/timer-ti-32k.c (limited to 'drivers/clocksource/timer-ti-32k.c') diff --git a/drivers/clocksource/timer-ti-32k.c b/drivers/clocksource/timer-ti-32k.c new file mode 100644 index 000000000..a86529a70 --- /dev/null +++ b/drivers/clocksource/timer-ti-32k.c @@ -0,0 +1,165 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * timer-ti-32k.c - OMAP2 32k Timer Support + * + * Copyright (C) 2009 Nokia Corporation + * + * Update to use new clocksource/clockevent layers + * Author: Kevin Hilman, MontaVista Software, Inc. + * Copyright (C) 2007 MontaVista Software, Inc. + * + * Original driver: + * Copyright (C) 2005 Nokia Corporation + * Author: Paul Mundt + * Juha Yrjölä + * OMAP Dual-mode timer framework support by Timo Teras + * + * Some parts based off of TI's 24xx code: + * + * Copyright (C) 2004-2009 Texas Instruments, Inc. + * + * Roughly modelled after the OMAP1 MPU timer code. + * Added OMAP4 support - Santosh Shilimkar + * + * Copyright (C) 2015 Texas Instruments Incorporated - https://www.ti.com + */ + +#include +#include +#include +#include +#include +#include +#include + +/* + * 32KHz clocksource ... always available, on pretty most chips except + * OMAP 730 and 1510. Other timers could be used as clocksources, with + * higher resolution in free-running counter modes (e.g. 12 MHz xtal), + * but systems won't necessarily want to spend resources that way. + */ + +#define OMAP2_32KSYNCNT_REV_OFF 0x0 +#define OMAP2_32KSYNCNT_REV_SCHEME (0x3 << 30) +#define OMAP2_32KSYNCNT_CR_OFF_LOW 0x10 +#define OMAP2_32KSYNCNT_CR_OFF_HIGH 0x30 + +struct ti_32k { + void __iomem *base; + void __iomem *counter; + struct clocksource cs; +}; + +static inline struct ti_32k *to_ti_32k(struct clocksource *cs) +{ + return container_of(cs, struct ti_32k, cs); +} + +static u64 notrace ti_32k_read_cycles(struct clocksource *cs) +{ + struct ti_32k *ti = to_ti_32k(cs); + + return (u64)readl_relaxed(ti->counter); +} + +static struct ti_32k ti_32k_timer = { + .cs = { + .name = "32k_counter", + .rating = 250, + .read = ti_32k_read_cycles, + .mask = CLOCKSOURCE_MASK(32), + .flags = CLOCK_SOURCE_IS_CONTINUOUS, + }, +}; + +static u64 notrace omap_32k_read_sched_clock(void) +{ + return ti_32k_read_cycles(&ti_32k_timer.cs); +} + +static void __init ti_32k_timer_enable_clock(struct device_node *np, + const char *name) +{ + struct clk *clock; + int error; + + clock = of_clk_get_by_name(np->parent, name); + if (IS_ERR(clock)) { + /* Only some SoCs have a separate interface clock */ + if (PTR_ERR(clock) == -EINVAL && !strncmp("ick", name, 3)) + return; + + pr_warn("%s: could not get clock %s %li\n", + __func__, name, PTR_ERR(clock)); + return; + } + + error = clk_prepare_enable(clock); + if (error) { + pr_warn("%s: could not enable %s: %i\n", + __func__, name, error); + return; + } +} + +static void __init ti_32k_timer_module_init(struct device_node *np, + void __iomem *base) +{ + void __iomem *sysc = base + 4; + + if (!of_device_is_compatible(np->parent, "ti,sysc")) + return; + + ti_32k_timer_enable_clock(np, "fck"); + ti_32k_timer_enable_clock(np, "ick"); + + /* + * Force idle module as wkup domain is active with MPU. + * No need to tag the module disabled for ti-sysc probe. + */ + writel_relaxed(0, sysc); +} + +static int __init ti_32k_timer_init(struct device_node *np) +{ + int ret; + + ti_32k_timer.base = of_iomap(np, 0); + if (!ti_32k_timer.base) { + pr_err("Can't ioremap 32k timer base\n"); + return -ENXIO; + } + + if (!of_machine_is_compatible("ti,am43")) + ti_32k_timer.cs.flags |= CLOCK_SOURCE_SUSPEND_NONSTOP; + + ti_32k_timer.counter = ti_32k_timer.base; + ti_32k_timer_module_init(np, ti_32k_timer.base); + + /* + * 32k sync Counter IP register offsets vary between the highlander + * version and the legacy ones. + * + * The 'SCHEME' bits(30-31) of the revision register is used to identify + * the version. + */ + if (readl_relaxed(ti_32k_timer.base + OMAP2_32KSYNCNT_REV_OFF) & + OMAP2_32KSYNCNT_REV_SCHEME) + ti_32k_timer.counter += OMAP2_32KSYNCNT_CR_OFF_HIGH; + else + ti_32k_timer.counter += OMAP2_32KSYNCNT_CR_OFF_LOW; + + pr_info("OMAP clocksource: 32k_counter at 32768 Hz\n"); + + ret = clocksource_register_hz(&ti_32k_timer.cs, 32768); + if (ret) { + pr_err("32k_counter: can't register clocksource\n"); + return ret; + } + + sched_clock_register(omap_32k_read_sched_clock, 32, 32768); + + return 0; +} +TIMER_OF_DECLARE(ti_32k_timer, "ti,omap-counter32k", + ti_32k_timer_init); -- cgit v1.3.1