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 --- lib/vdso/datastore.c | 129 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 129 insertions(+) create mode 100644 lib/vdso/datastore.c (limited to 'lib/vdso/datastore.c') diff --git a/lib/vdso/datastore.c b/lib/vdso/datastore.c new file mode 100644 index 000000000..1426bf4e0 --- /dev/null +++ b/lib/vdso/datastore.c @@ -0,0 +1,129 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include +#include +#include +#include +#include +#include +#include + +static u8 vdso_initdata[VDSO_NR_PAGES * PAGE_SIZE] __aligned(PAGE_SIZE) __initdata = {}; + +#ifdef CONFIG_GENERIC_GETTIMEOFDAY +struct vdso_time_data *vdso_k_time_data __ro_after_init = + (void *)&vdso_initdata[VDSO_TIME_PAGE_OFFSET * PAGE_SIZE]; + +static_assert(sizeof(struct vdso_time_data) <= PAGE_SIZE); +#endif /* CONFIG_GENERIC_GETTIMEOFDAY */ + +#ifdef CONFIG_VDSO_GETRANDOM +struct vdso_rng_data *vdso_k_rng_data __ro_after_init = + (void *)&vdso_initdata[VDSO_RNG_PAGE_OFFSET * PAGE_SIZE]; + +static_assert(sizeof(struct vdso_rng_data) <= PAGE_SIZE); +#endif /* CONFIG_VDSO_GETRANDOM */ + +#ifdef CONFIG_ARCH_HAS_VDSO_ARCH_DATA +struct vdso_arch_data *vdso_k_arch_data __ro_after_init = + (void *)&vdso_initdata[VDSO_ARCH_PAGES_START * PAGE_SIZE]; +#endif /* CONFIG_ARCH_HAS_VDSO_ARCH_DATA */ + +static struct page *vdso_data_pages __ro_after_init; + +void __init vdso_setup_data_pages(void) +{ + unsigned int order = get_order(VDSO_NR_PAGES * PAGE_SIZE); + + /* + * Allocate the data pages dynamically. SPARC does not support mapping + * static pages to be mapped into userspace. + * It is also a requirement for mlockall() support. + * + * Do not use folios. In time namespaces the pages are mapped in a different order + * to userspace, which is not handled by the folio optimizations in finish_fault(). + */ + vdso_data_pages = alloc_pages(GFP_KERNEL, order); + if (!vdso_data_pages) + panic("Unable to allocate VDSO storage pages"); + + /* The pages are mapped one-by-one into userspace and each one needs to be refcounted. */ + split_page(vdso_data_pages, order); + + /* Move the data already written by other subsystems to the new pages */ + memcpy(page_address(vdso_data_pages), vdso_initdata, VDSO_NR_PAGES * PAGE_SIZE); + + if (IS_ENABLED(CONFIG_GENERIC_GETTIMEOFDAY)) + vdso_k_time_data = page_address(vdso_data_pages + VDSO_TIME_PAGE_OFFSET); + + if (IS_ENABLED(CONFIG_VDSO_GETRANDOM)) + vdso_k_rng_data = page_address(vdso_data_pages + VDSO_RNG_PAGE_OFFSET); + + if (IS_ENABLED(CONFIG_ARCH_HAS_VDSO_ARCH_DATA)) + vdso_k_arch_data = page_address(vdso_data_pages + VDSO_ARCH_PAGES_START); +} + +static vm_fault_t vvar_fault(const struct vm_special_mapping *sm, + struct vm_area_struct *vma, struct vm_fault *vmf) +{ + struct page *page, *timens_page; + + if (unlikely(vmf->flags & FAULT_FLAG_REMOTE)) + return VM_FAULT_SIGBUS; + + page = vdso_data_pages + vmf->pgoff; + timens_page = find_timens_vvar_page(vma); + + switch (vmf->pgoff) { + case VDSO_TIME_PAGE_OFFSET: + if (!IS_ENABLED(CONFIG_GENERIC_GETTIMEOFDAY) || !timens_page) + break; + /* + * Fault in VVAR page too, since it will be accessed + * to get clock data anyway. + */ + unsigned long addr; + vm_fault_t err; + + addr = vmf->address + VDSO_TIMENS_PAGE_OFFSET * PAGE_SIZE; + err = vmf_insert_page(vma, addr, page); + if (unlikely(err & VM_FAULT_ERROR)) + return err; + page = timens_page; + break; + case VDSO_TIMENS_PAGE_OFFSET: + /* + * If a task belongs to a time namespace then a namespace + * specific VVAR is mapped with the VVAR_DATA_PAGE_OFFSET and + * the real VVAR page is mapped with the VVAR_TIMENS_PAGE_OFFSET + * offset. + * See also the comment near timens_setup_vdso_data(). + */ + if (!IS_ENABLED(CONFIG_TIME_NS) || !timens_page) + break; + page = vdso_data_pages + VDSO_TIME_PAGE_OFFSET; + break; + case VDSO_RNG_PAGE_OFFSET: + case VDSO_ARCH_PAGES_START ... VDSO_ARCH_PAGES_END: + break; + default: + return VM_FAULT_SIGBUS; + } + + get_page(page); + vmf->page = page; + return 0; +} + +const struct vm_special_mapping vdso_vvar_mapping = { + .name = "[vvar]", + .fault = vvar_fault, +}; + +struct vm_area_struct *vdso_install_vvar_mapping(struct mm_struct *mm, unsigned long addr) +{ + return _install_special_mapping(mm, addr, VDSO_NR_PAGES * PAGE_SIZE, + VM_READ | VM_MAYREAD | VM_DONTDUMP | + VM_MIXEDMAP | VM_SEALED_SYSMAP, + &vdso_vvar_mapping); +} -- cgit v1.3.1