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/acpi/evged.c | 195 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 195 insertions(+) create mode 100644 drivers/acpi/evged.c (limited to 'drivers/acpi/evged.c') diff --git a/drivers/acpi/evged.c b/drivers/acpi/evged.c new file mode 100644 index 000000000..5c35cbc7f --- /dev/null +++ b/drivers/acpi/evged.c @@ -0,0 +1,195 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Generic Event Device for ACPI. + * + * Copyright (c) 2016, The Linux Foundation. All rights reserved. + * + * Generic Event Device allows platforms to handle interrupts in ACPI + * ASL statements. It follows very similar to _EVT method approach + * from GPIO events. All interrupts are listed in _CRS and the handler + * is written in _EVT method. Here is an example. + * + * Device (GED0) + * { + * + * Name (_HID, "ACPI0013") + * Name (_UID, 0) + * Method (_CRS, 0x0, Serialized) + * { + * Name (RBUF, ResourceTemplate () + * { + * Interrupt(ResourceConsumer, Edge, ActiveHigh, Shared, , , ) + * {123} + * } + * }) + * + * Method (_EVT, 1) { + * if (Lequal(123, Arg0)) + * { + * } + * } + * } + */ + +#include +#include +#include +#include +#include +#include + +#define MODULE_NAME "acpi-ged" + +struct acpi_ged_device { + struct device *dev; + struct list_head event_list; +}; + +struct acpi_ged_event { + struct list_head node; + struct device *dev; + unsigned int gsi; + unsigned int irq; + acpi_handle handle; +}; + +static irqreturn_t acpi_ged_irq_handler(int irq, void *data) +{ + struct acpi_ged_event *event = data; + acpi_status acpi_ret; + + acpi_ret = acpi_execute_simple_method(event->handle, NULL, event->gsi); + if (ACPI_FAILURE(acpi_ret)) + dev_err_once(event->dev, "IRQ method execution failed\n"); + + return IRQ_HANDLED; +} + +static acpi_status acpi_ged_request_interrupt(struct acpi_resource *ares, + void *context) +{ + struct acpi_ged_event *event; + unsigned int irq; + unsigned int gsi; + unsigned int irqflags = IRQF_ONESHOT; + struct acpi_ged_device *geddev = context; + struct device *dev = geddev->dev; + acpi_handle handle = ACPI_HANDLE(dev); + acpi_handle evt_handle; + struct resource r; + struct acpi_resource_irq *p = &ares->data.irq; + struct acpi_resource_extended_irq *pext = &ares->data.extended_irq; + char ev_name[5]; + u8 trigger; + + if (ares->type == ACPI_RESOURCE_TYPE_END_TAG) + return AE_OK; + + if (!acpi_dev_resource_interrupt(ares, 0, &r)) { + dev_err(dev, "unable to parse IRQ resource\n"); + return AE_ERROR; + } + if (ares->type == ACPI_RESOURCE_TYPE_IRQ) { + gsi = p->interrupts[0]; + trigger = p->triggering; + } else { + gsi = pext->interrupts[0]; + trigger = pext->triggering; + } + + irq = r.start; + + switch (gsi) { + case 0 ... 255: + sprintf(ev_name, "_%c%02X", + trigger == ACPI_EDGE_SENSITIVE ? 'E' : 'L', gsi); + + if (ACPI_SUCCESS(acpi_get_handle(handle, ev_name, &evt_handle))) + break; + fallthrough; + default: + if (ACPI_SUCCESS(acpi_get_handle(handle, "_EVT", &evt_handle))) + break; + + dev_err(dev, "cannot locate _EVT method\n"); + return AE_ERROR; + } + + event = devm_kzalloc(dev, sizeof(*event), GFP_KERNEL); + if (!event) + return AE_ERROR; + + event->gsi = gsi; + event->dev = dev; + event->irq = irq; + event->handle = evt_handle; + + if (r.flags & IORESOURCE_IRQ_SHAREABLE) + irqflags |= IRQF_SHARED; + + if (request_threaded_irq(irq, NULL, acpi_ged_irq_handler, + irqflags, "ACPI:Ged", event)) { + dev_err(dev, "failed to setup event handler for irq %u\n", irq); + return AE_ERROR; + } + + dev_dbg(dev, "GED listening GSI %u @ IRQ %u\n", gsi, irq); + list_add_tail(&event->node, &geddev->event_list); + return AE_OK; +} + +static int ged_probe(struct platform_device *pdev) +{ + struct acpi_ged_device *geddev; + acpi_status acpi_ret; + + geddev = devm_kzalloc(&pdev->dev, sizeof(*geddev), GFP_KERNEL); + if (!geddev) + return -ENOMEM; + + geddev->dev = &pdev->dev; + INIT_LIST_HEAD(&geddev->event_list); + acpi_ret = acpi_walk_resources(ACPI_HANDLE(&pdev->dev), "_CRS", + acpi_ged_request_interrupt, geddev); + if (ACPI_FAILURE(acpi_ret)) { + dev_err(&pdev->dev, "unable to parse the _CRS record\n"); + return -EINVAL; + } + platform_set_drvdata(pdev, geddev); + + return 0; +} + +static void ged_shutdown(struct platform_device *pdev) +{ + struct acpi_ged_device *geddev = platform_get_drvdata(pdev); + struct acpi_ged_event *event, *next; + + list_for_each_entry_safe(event, next, &geddev->event_list, node) { + free_irq(event->irq, event); + list_del(&event->node); + dev_dbg(geddev->dev, "GED releasing GSI %u @ IRQ %u\n", + event->gsi, event->irq); + } +} + +static void ged_remove(struct platform_device *pdev) +{ + ged_shutdown(pdev); +} + +static const struct acpi_device_id ged_acpi_ids[] = { + {"ACPI0013"}, + {}, +}; + +static struct platform_driver ged_driver = { + .probe = ged_probe, + .remove = ged_remove, + .shutdown = ged_shutdown, + .driver = { + .name = MODULE_NAME, + .acpi_match_table = ACPI_PTR(ged_acpi_ids), + }, +}; +builtin_platform_driver(ged_driver); -- cgit v1.3.1