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/firmware/sysfb.c | 225 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 225 insertions(+) create mode 100644 drivers/firmware/sysfb.c (limited to 'drivers/firmware/sysfb.c') diff --git a/drivers/firmware/sysfb.c b/drivers/firmware/sysfb.c new file mode 100644 index 000000000..8833582c1 --- /dev/null +++ b/drivers/firmware/sysfb.c @@ -0,0 +1,225 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Generic System Framebuffers + * Copyright (c) 2012-2013 David Herrmann + */ + +/* + * Simple-Framebuffer support + * Create a platform-device for any available boot framebuffer. The + * simple-framebuffer platform device is already available on DT systems, so + * this module parses the global "screen_info" object and creates a suitable + * platform device compatible with the "simple-framebuffer" DT object. If + * the framebuffer is incompatible, we instead create a legacy + * "vesa-framebuffer", "efi-framebuffer" or "platform-framebuffer" device and + * pass the screen_info as platform_data. This allows legacy drivers + * to pick these devices up without messing with simple-framebuffer drivers. + * The global "screen_info" is still valid at all times. + * + * If CONFIG_SYSFB_SIMPLEFB is not selected, never register "simple-framebuffer" + * platform devices, but only use legacy framebuffer devices for + * backwards compatibility. + * + * TODO: We set the dev_id field of all platform-devices to 0. This allows + * other OF/DT parsers to create such devices, too. However, they must + * start at offset 1 for this to work. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +static struct platform_device *pd; +static DEFINE_MUTEX(disable_lock); +static bool disabled; + +static struct device *sysfb_parent_dev(const struct screen_info *si); + +static bool sysfb_unregister(void) +{ + if (IS_ERR_OR_NULL(pd)) + return false; + + platform_device_unregister(pd); + pd = NULL; + + return true; +} + +/** + * sysfb_disable() - disable the Generic System Framebuffers support + * @dev: the device to check if non-NULL + * + * This disables the registration of system framebuffer devices that match the + * generic drivers that make use of the system framebuffer set up by firmware. + * + * It also unregisters a device if this was already registered by sysfb_init(). + * + * Context: The function can sleep. A @disable_lock mutex is acquired to serialize + * against sysfb_init(), that registers a system framebuffer device. + */ +void sysfb_disable(struct device *dev) +{ + struct screen_info *si = &sysfb_primary_display.screen; + struct device *parent; + + mutex_lock(&disable_lock); + parent = sysfb_parent_dev(si); + if (!dev || !parent || dev == parent) { + sysfb_unregister(); + disabled = true; + } + mutex_unlock(&disable_lock); +} +EXPORT_SYMBOL_GPL(sysfb_disable); + +/** + * sysfb_handles_screen_info() - reports if sysfb handles the global screen_info + * + * Callers can use sysfb_handles_screen_info() to determine whether the Generic + * System Framebuffers (sysfb) can handle the global screen_info data structure + * or not. Drivers might need this information to know if they have to setup the + * system framebuffer, or if they have to delegate this action to sysfb instead. + * + * Returns: + * True if sysfb handles the global screen_info data structure. + */ +bool sysfb_handles_screen_info(void) +{ + const struct screen_info *si = &sysfb_primary_display.screen; + + return !!screen_info_video_type(si); +} +EXPORT_SYMBOL_GPL(sysfb_handles_screen_info); + +#if defined(CONFIG_PCI) +static bool sysfb_pci_dev_is_enabled(struct pci_dev *pdev) +{ + /* + * TODO: Try to integrate this code into the PCI subsystem + */ + int ret; + u16 command; + + ret = pci_read_config_word(pdev, PCI_COMMAND, &command); + if (ret != PCIBIOS_SUCCESSFUL) + return false; + if (!(command & PCI_COMMAND_MEMORY)) + return false; + return true; +} +#else +static bool sysfb_pci_dev_is_enabled(struct pci_dev *pdev) +{ + return false; +} +#endif + +static struct device *sysfb_parent_dev(const struct screen_info *si) +{ + struct pci_dev *pdev; + + pdev = screen_info_pci_dev(si); + if (IS_ERR(pdev)) { + return ERR_CAST(pdev); + } else if (pdev) { + if (!sysfb_pci_dev_is_enabled(pdev)) { + pci_dev_put(pdev); + return ERR_PTR(-ENODEV); + } + return &pdev->dev; + } + + return NULL; +} + +static __init int sysfb_init(void) +{ + struct sysfb_display_info *dpy = &sysfb_primary_display; + struct screen_info *si = &dpy->screen; + struct device *parent; + unsigned int type; + struct simplefb_platform_data mode; + const char *name; + bool compatible; + int ret = 0; + + screen_info_apply_fixups(); + + mutex_lock(&disable_lock); + if (disabled) + goto unlock_mutex; + + sysfb_apply_efi_quirks(si); + + parent = sysfb_parent_dev(si); + if (IS_ERR(parent)) { + ret = PTR_ERR(parent); + goto unlock_mutex; + } + + /* try to create a simple-framebuffer device */ + compatible = sysfb_parse_mode(si, &mode); + if (compatible) { + pd = sysfb_create_simplefb(si, &mode, parent); + if (!IS_ERR(pd)) + goto put_device; + } + + type = screen_info_video_type(si); + + /* if the FB is incompatible, create a legacy framebuffer device */ + switch (type) { + case VIDEO_TYPE_EGAC: + name = "ega-framebuffer"; + break; + case VIDEO_TYPE_VGAC: + name = "vga-framebuffer"; + break; + case VIDEO_TYPE_VLFB: + name = "vesa-framebuffer"; + break; + case VIDEO_TYPE_EFI: + name = "efi-framebuffer"; + break; + default: + name = "platform-framebuffer"; + break; + } + + pd = platform_device_alloc(name, 0); + if (!pd) { + ret = -ENOMEM; + goto put_device; + } + + pd->dev.parent = parent; + + sysfb_set_efifb_fwnode(si, pd); + + ret = platform_device_add_data(pd, dpy, sizeof(*dpy)); + if (ret) + goto err; + + ret = platform_device_add(pd); + if (ret) + goto err; + + goto put_device; +err: + platform_device_put(pd); +put_device: + put_device(parent); +unlock_mutex: + mutex_unlock(&disable_lock); + return ret; +} + +/* must execute after PCI subsystem for EFI quirks */ +device_initcall(sysfb_init); -- cgit v1.3.1