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/efi/memmap.c | 148 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 148 insertions(+) create mode 100644 drivers/firmware/efi/memmap.c (limited to 'drivers/firmware/efi/memmap.c') diff --git a/drivers/firmware/efi/memmap.c b/drivers/firmware/efi/memmap.c new file mode 100644 index 000000000..f1c04d7cf --- /dev/null +++ b/drivers/firmware/efi/memmap.c @@ -0,0 +1,148 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Common EFI memory map functions. + */ + +#define pr_fmt(fmt) "efi: " fmt + +#include +#include +#include +#include +#include +#include + +#include +#include + +/** + * __efi_memmap_init - Common code for mapping the EFI memory map + * @data: EFI memory map data + * + * This function takes care of figuring out which function to use to + * map the EFI memory map in efi.memmap based on how far into the boot + * we are. + * + * During bootup EFI_MEMMAP_LATE in data->flags should be clear since we + * only have access to the early_memremap*() functions as the vmalloc + * space isn't setup. Once the kernel is fully booted we can fallback + * to the more robust memremap*() API. + * + * Returns: zero on success, a negative error code on failure. + */ +int __init __efi_memmap_init(struct efi_memory_map_data *data) +{ + struct efi_memory_map map; + phys_addr_t phys_map; + + phys_map = data->phys_map; + + if (data->flags & EFI_MEMMAP_LATE) + map.map = memremap(phys_map, data->size, MEMREMAP_WB); + else + map.map = early_memremap(phys_map, data->size); + + if (!map.map) { + pr_err("Could not map the memory map! phys_map=%pa, size=0x%lx\n", + &phys_map, data->size); + return -ENOMEM; + } + + map.phys_map = data->phys_map; + map.nr_map = data->size / data->desc_size; + map.map_end = map.map + data->size; + + map.desc_version = data->desc_version; + map.desc_size = data->desc_size; + map.flags = data->flags; + + set_bit(EFI_MEMMAP, &efi.flags); + + efi.memmap = map; + + return 0; +} + +/** + * efi_memmap_init_early - Map the EFI memory map data structure + * @data: EFI memory map data + * + * Use early_memremap() to map the passed in EFI memory map and assign + * it to efi.memmap. + * + * Returns: zero on success, a negative error code on failure. + */ +int __init efi_memmap_init_early(struct efi_memory_map_data *data) +{ + /* Cannot go backwards */ + WARN_ON(efi.memmap.flags & EFI_MEMMAP_LATE); + + data->flags = 0; + return __efi_memmap_init(data); +} + +void __init efi_memmap_unmap(void) +{ + if (!efi_enabled(EFI_MEMMAP)) + return; + + if (!(efi.memmap.flags & EFI_MEMMAP_LATE)) { + unsigned long size; + + size = efi.memmap.desc_size * efi.memmap.nr_map; + early_memunmap(efi.memmap.map, size); + } else { + memunmap(efi.memmap.map); + } + + efi.memmap.map = NULL; + clear_bit(EFI_MEMMAP, &efi.flags); +} + +/** + * efi_memmap_init_late - Map efi.memmap with memremap() + * @addr: Physical address of the new EFI memory map + * @size: Size in bytes of the new EFI memory map + * + * Setup a mapping of the EFI memory map using ioremap_cache(). This + * function should only be called once the vmalloc space has been + * setup and is therefore not suitable for calling during early EFI + * initialise, e.g. in efi_init(). Additionally, it expects + * efi_memmap_init_early() to have already been called. + * + * The reason there are two EFI memmap initialisation + * (efi_memmap_init_early() and this late version) is because the + * early EFI memmap should be explicitly unmapped once EFI + * initialisation is complete as the fixmap space used to map the EFI + * memmap (via early_memremap()) is a scarce resource. + * + * This late mapping is intended to persist for the duration of + * runtime so that things like efi_mem_desc_lookup() and + * efi_mem_attributes() always work. + * + * Returns: zero on success, a negative error code on failure. + */ +int __init efi_memmap_init_late(phys_addr_t addr, unsigned long size) +{ + struct efi_memory_map_data data = { + .phys_map = addr, + .size = size, + .flags = EFI_MEMMAP_LATE, + }; + + /* Did we forget to unmap the early EFI memmap? */ + WARN_ON(efi.memmap.map); + + /* Were we already called? */ + WARN_ON(efi.memmap.flags & EFI_MEMMAP_LATE); + + /* + * It makes no sense to allow callers to register different + * values for the following fields. Copy them out of the + * existing early EFI memmap. + */ + data.desc_version = efi.memmap.desc_version; + data.desc_size = efi.memmap.desc_size; + + return __efi_memmap_init(&data); +} -- cgit v1.3.1