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 --- arch/alpha/mm/init.c | 266 +++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 266 insertions(+) create mode 100644 arch/alpha/mm/init.c (limited to 'arch/alpha/mm/init.c') diff --git a/arch/alpha/mm/init.c b/arch/alpha/mm/init.c new file mode 100644 index 000000000..f4d65a60c --- /dev/null +++ b/arch/alpha/mm/init.c @@ -0,0 +1,266 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * linux/arch/alpha/mm/init.c + * + * Copyright (C) 1995 Linus Torvalds + */ + +/* 2.3.x zone allocator, 1999 Andrea Arcangeli */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include /* max_low_pfn */ +#include +#include + +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "../kernel/proto.h" + +static struct pcb_struct original_pcb; + +pgd_t * +pgd_alloc(struct mm_struct *mm) +{ + pgd_t *ret, *init; + + ret = __pgd_alloc(mm, 0); + init = pgd_offset(&init_mm, 0UL); + if (ret) { +#ifdef CONFIG_ALPHA_LARGE_VMALLOC + memcpy (ret + USER_PTRS_PER_PGD, init + USER_PTRS_PER_PGD, + (PTRS_PER_PGD - USER_PTRS_PER_PGD - 1)*sizeof(pgd_t)); +#else + pgd_val(ret[PTRS_PER_PGD-2]) = pgd_val(init[PTRS_PER_PGD-2]); +#endif + + /* The last PGD entry is the VPTB self-map. */ + pgd_val(ret[PTRS_PER_PGD-1]) + = pte_val(mk_pte(virt_to_page(ret), PAGE_KERNEL)); + } + return ret; +} + + +static inline unsigned long +load_PCB(struct pcb_struct *pcb) +{ + pcb->ksp = (unsigned long)current_stack_pointer; + return __reload_thread(pcb); +} + +/* Set up initial PCB, VPTB, and other such nicities. */ + +static inline void +switch_to_system_map(void) +{ + unsigned long newptbr; + unsigned long original_pcb_ptr; + + /* Initialize the kernel's page tables. Linux puts the vptb in + the last slot of the L1 page table. */ + memset(swapper_pg_dir, 0, PAGE_SIZE); + newptbr = ((unsigned long) swapper_pg_dir - PAGE_OFFSET) >> PAGE_SHIFT; + pgd_val(swapper_pg_dir[1023]) = + (newptbr << 32) | pgprot_val(PAGE_KERNEL); + + /* Set the vptb. This is often done by the bootloader, but + shouldn't be required. */ + if (hwrpb->vptb != 0xfffffffe00000000UL) { + wrvptptr(0xfffffffe00000000UL); + hwrpb->vptb = 0xfffffffe00000000UL; + hwrpb_update_checksum(hwrpb); + } + + /* Also set up the real kernel PCB while we're at it. */ + init_thread_info.pcb.ptbr = newptbr; + init_thread_info.pcb.flags = 1; /* set FEN, clear everything else */ + original_pcb_ptr = load_PCB(&init_thread_info.pcb); + tbia(); + + /* Save off the contents of the original PCB so that we can + restore the original console's page tables for a clean reboot. + + Note that the PCB is supposed to be a physical address, but + since KSEG values also happen to work, folks get confused. + Check this here. */ + + if (original_pcb_ptr < PAGE_OFFSET) { + original_pcb_ptr = (unsigned long) + phys_to_virt(original_pcb_ptr); + } + original_pcb = *(struct pcb_struct *) original_pcb_ptr; +} + +int callback_init_done; + +void * __init +callback_init(void * kernel_end) +{ + struct crb_struct * crb; + pgd_t *pgd; + p4d_t *p4d; + pud_t *pud; + pmd_t *pmd; + void *two_pages; + + /* Starting at the HWRPB, locate the CRB. */ + crb = (struct crb_struct *)((char *)hwrpb + hwrpb->crb_offset); + + if (alpha_using_srm) { + /* Tell the console whither it is to be remapped. */ + if (srm_fixup(VMALLOC_START, (unsigned long)hwrpb)) + __halt(); /* "We're boned." --Bender */ + + /* Edit the procedure descriptors for DISPATCH and FIXUP. */ + crb->dispatch_va = (struct procdesc_struct *) + (VMALLOC_START + (unsigned long)crb->dispatch_va + - crb->map[0].va); + crb->fixup_va = (struct procdesc_struct *) + (VMALLOC_START + (unsigned long)crb->fixup_va + - crb->map[0].va); + } + + switch_to_system_map(); + + /* Allocate one PGD and one PMD. In the case of SRM, we'll need + these to actually remap the console. There is an assumption + here that only one of each is needed, and this allows for 8MB. + On systems with larger consoles, additional pages will be + allocated as needed during the mapping process. + + In the case of not SRM, but not CONFIG_ALPHA_LARGE_VMALLOC, + we need to allocate the PGD we use for vmalloc before we start + forking other tasks. */ + + two_pages = (void *) + (((unsigned long)kernel_end + ~PAGE_MASK) & PAGE_MASK); + kernel_end = two_pages + 2*PAGE_SIZE; + memset(two_pages, 0, 2*PAGE_SIZE); + + pgd = pgd_offset_k(VMALLOC_START); + p4d = p4d_offset(pgd, VMALLOC_START); + pud = pud_offset(p4d, VMALLOC_START); + pud_set(pud, (pmd_t *)two_pages); + pmd = pmd_offset(pud, VMALLOC_START); + pmd_set(pmd, (pte_t *)(two_pages + PAGE_SIZE)); + + if (alpha_using_srm) { + static struct vm_struct console_remap_vm; + unsigned long nr_pages = 0; + unsigned long vaddr; + unsigned long i, j; + + /* calculate needed size */ + for (i = 0; i < crb->map_entries; ++i) + nr_pages += crb->map[i].count; + + /* register the vm area */ + console_remap_vm.flags = VM_ALLOC; + console_remap_vm.size = nr_pages << PAGE_SHIFT; + vm_area_register_early(&console_remap_vm, PAGE_SIZE); + + vaddr = (unsigned long)console_remap_vm.addr; + + /* Set up the third level PTEs and update the virtual + addresses of the CRB entries. */ + for (i = 0; i < crb->map_entries; ++i) { + unsigned long pfn = crb->map[i].pa >> PAGE_SHIFT; + crb->map[i].va = vaddr; + for (j = 0; j < crb->map[i].count; ++j) { + /* Newer consoles (especially on larger + systems) may require more pages of + PTEs. Grab additional pages as needed. */ + if (pmd != pmd_offset(pud, vaddr)) { + memset(kernel_end, 0, PAGE_SIZE); + pmd = pmd_offset(pud, vaddr); + pmd_set(pmd, (pte_t *)kernel_end); + kernel_end += PAGE_SIZE; + } + set_pte(pte_offset_kernel(pmd, vaddr), + pfn_pte(pfn, PAGE_KERNEL)); + pfn++; + vaddr += PAGE_SIZE; + } + } + } + + callback_init_done = 1; + return kernel_end; +} + +void __init arch_zone_limits_init(unsigned long *max_zone_pfn) +{ + unsigned long dma_pfn; + + dma_pfn = virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; + max_pfn = max_low_pfn; + + max_zone_pfn[ZONE_DMA] = dma_pfn; + max_zone_pfn[ZONE_NORMAL] = max_pfn; +} + +/* + * paging_init() initializes the kernel's ZERO_PGE. + */ +void __init paging_init(void) +{ + memset(absolute_pointer(ZERO_PGE), 0, PAGE_SIZE); +} + +#if defined(CONFIG_ALPHA_GENERIC) || defined(CONFIG_ALPHA_SRM) +void +srm_paging_stop (void) +{ + /* Move the vptb back to where the SRM console expects it. */ + swapper_pg_dir[1] = swapper_pg_dir[1023]; + tbia(); + wrvptptr(0x200000000UL); + hwrpb->vptb = 0x200000000UL; + hwrpb_update_checksum(hwrpb); + + /* Reload the page tables that the console had in use. */ + load_PCB(&original_pcb); + tbia(); +} +#endif + +static const pgprot_t protection_map[16] = { + [VM_NONE] = _PAGE_P(_PAGE_FOE | _PAGE_FOW | + _PAGE_FOR), + [VM_READ] = _PAGE_P(_PAGE_FOE | _PAGE_FOW), + [VM_WRITE] = _PAGE_P(_PAGE_FOE), + [VM_WRITE | VM_READ] = _PAGE_P(_PAGE_FOE), + [VM_EXEC] = _PAGE_P(_PAGE_FOW | _PAGE_FOR), + [VM_EXEC | VM_READ] = _PAGE_P(_PAGE_FOW), + [VM_EXEC | VM_WRITE] = _PAGE_P(0), + [VM_EXEC | VM_WRITE | VM_READ] = _PAGE_P(0), + [VM_SHARED] = _PAGE_S(_PAGE_FOE | _PAGE_FOW | + _PAGE_FOR), + [VM_SHARED | VM_READ] = _PAGE_S(_PAGE_FOE | _PAGE_FOW), + [VM_SHARED | VM_WRITE] = _PAGE_S(_PAGE_FOE), + [VM_SHARED | VM_WRITE | VM_READ] = _PAGE_S(_PAGE_FOE), + [VM_SHARED | VM_EXEC] = _PAGE_S(_PAGE_FOW | _PAGE_FOR), + [VM_SHARED | VM_EXEC | VM_READ] = _PAGE_S(_PAGE_FOW), + [VM_SHARED | VM_EXEC | VM_WRITE] = _PAGE_S(0), + [VM_SHARED | VM_EXEC | VM_WRITE | VM_READ] = _PAGE_S(0) +}; +DECLARE_VM_GET_PAGE_PROT -- cgit v1.3.1