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/test_kho.c | 365 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 365 insertions(+) create mode 100644 lib/test_kho.c (limited to 'lib/test_kho.c') diff --git a/lib/test_kho.c b/lib/test_kho.c new file mode 100644 index 000000000..aa6a0956b --- /dev/null +++ b/lib/test_kho.c @@ -0,0 +1,365 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * Test module for KHO + * Copyright (c) 2025 Microsoft Corporation. + * + * Authors: + * Saurabh Sengar + * Mike Rapoport + */ + +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define KHO_TEST_MAGIC 0x4b484f21 /* KHO! */ +#define KHO_TEST_FDT "kho_test" +#define KHO_TEST_COMPAT "kho-test-v1" + +static long max_mem = (PAGE_SIZE << MAX_PAGE_ORDER) * 2; +module_param(max_mem, long, 0644); + +struct kho_test_state { + unsigned int nr_folios; + struct folio **folios; + phys_addr_t *folios_info; + struct kho_vmalloc folios_info_phys; + int nr_folios_preserved; + struct folio *fdt; + __wsum csum; +}; + +static struct kho_test_state kho_test_state; + +static void kho_test_unpreserve_data(struct kho_test_state *state) +{ + for (int i = 0; i < state->nr_folios_preserved; i++) + kho_unpreserve_folio(state->folios[i]); + + kho_unpreserve_vmalloc(&state->folios_info_phys); + vfree(state->folios_info); +} + +static int kho_test_preserve_data(struct kho_test_state *state) +{ + struct kho_vmalloc folios_info_phys; + phys_addr_t *folios_info; + int err; + + folios_info = vmalloc_array(state->nr_folios, sizeof(*folios_info)); + if (!folios_info) + return -ENOMEM; + + err = kho_preserve_vmalloc(folios_info, &folios_info_phys); + if (err) + goto err_free_info; + + state->folios_info_phys = folios_info_phys; + state->folios_info = folios_info; + + for (int i = 0; i < state->nr_folios; i++) { + struct folio *folio = state->folios[i]; + unsigned int order = folio_order(folio); + + folios_info[i] = virt_to_phys(folio_address(folio)) | order; + err = kho_preserve_folio(folio); + if (err) + goto err_unpreserve; + state->nr_folios_preserved++; + } + + return 0; + +err_unpreserve: + /* + * kho_test_unpreserve_data frees folio_info, bail out immediately to + * avoid double free + */ + kho_test_unpreserve_data(state); + return err; + +err_free_info: + vfree(folios_info); + return err; +} + +static int kho_test_prepare_fdt(struct kho_test_state *state, ssize_t fdt_size) +{ + const char compatible[] = KHO_TEST_COMPAT; + unsigned int magic = KHO_TEST_MAGIC; + void *fdt = folio_address(state->fdt); + int err; + + err = fdt_create(fdt, fdt_size); + err |= fdt_finish_reservemap(fdt); + err |= fdt_begin_node(fdt, ""); + err |= fdt_property(fdt, "compatible", compatible, sizeof(compatible)); + err |= fdt_property(fdt, "magic", &magic, sizeof(magic)); + + err |= fdt_begin_node(fdt, "data"); + err |= fdt_property(fdt, "nr_folios", &state->nr_folios, + sizeof(state->nr_folios)); + err |= fdt_property(fdt, "folios_info", &state->folios_info_phys, + sizeof(state->folios_info_phys)); + err |= fdt_property(fdt, "csum", &state->csum, sizeof(state->csum)); + err |= fdt_end_node(fdt); + + err |= fdt_end_node(fdt); + err |= fdt_finish(fdt); + + return err; +} + +static int kho_test_preserve(struct kho_test_state *state) +{ + ssize_t fdt_size; + int err; + + fdt_size = state->nr_folios * sizeof(phys_addr_t) + PAGE_SIZE; + state->fdt = folio_alloc(GFP_KERNEL, get_order(fdt_size)); + if (!state->fdt) + return -ENOMEM; + + err = kho_preserve_folio(state->fdt); + if (err) + goto err_free_fdt; + + err = kho_test_preserve_data(state); + if (err) + goto err_unpreserve_fdt; + + err = kho_test_prepare_fdt(state, fdt_size); + if (err) + goto err_unpreserve_data; + + err = kho_add_subtree(KHO_TEST_FDT, folio_address(state->fdt), + fdt_totalsize(folio_address(state->fdt))); + if (err) + goto err_unpreserve_data; + + return 0; + +err_unpreserve_data: + kho_test_unpreserve_data(state); +err_unpreserve_fdt: + kho_unpreserve_folio(state->fdt); +err_free_fdt: + folio_put(state->fdt); + return err; +} + +static int kho_test_generate_data(struct kho_test_state *state) +{ + size_t alloc_size = 0; + __wsum csum = 0; + + while (alloc_size < max_mem) { + int order = get_random_u32() % NR_PAGE_ORDERS; + struct folio *folio; + unsigned int size; + void *addr; + + /* + * Since get_order() rounds up, make sure that actual + * allocation is smaller so that we won't exceed max_mem + */ + if (alloc_size + (PAGE_SIZE << order) > max_mem) { + order = get_order(max_mem - alloc_size); + if (order) + order--; + } + size = PAGE_SIZE << order; + + folio = folio_alloc(GFP_KERNEL | __GFP_NORETRY, order); + if (!folio) + goto err_free_folios; + + state->folios[state->nr_folios++] = folio; + addr = folio_address(folio); + get_random_bytes(addr, size); + csum = csum_partial(addr, size, csum); + alloc_size += size; + } + + state->csum = csum; + return 0; + +err_free_folios: + for (int i = 0; i < state->nr_folios; i++) + folio_put(state->folios[i]); + state->nr_folios = 0; + return -ENOMEM; +} + +static int kho_test_save(void) +{ + struct kho_test_state *state = &kho_test_state; + struct folio **folios; + unsigned long max_nr; + int err; + + max_mem = PAGE_ALIGN(max_mem); + max_nr = max_mem >> PAGE_SHIFT; + + folios = kvmalloc_objs(*state->folios, max_nr); + if (!folios) + return -ENOMEM; + state->folios = folios; + + err = kho_test_generate_data(state); + if (err) + goto err_free_folios; + + err = kho_test_preserve(state); + if (err) + goto err_free_folios; + + return 0; + +err_free_folios: + kvfree(folios); + return err; +} + +static int kho_test_restore_data(const void *fdt, int node) +{ + const struct kho_vmalloc *folios_info_phys; + const unsigned int *nr_folios; + phys_addr_t *folios_info; + const __wsum *old_csum; + __wsum csum = 0; + int len; + + node = fdt_path_offset(fdt, "/data"); + + nr_folios = fdt_getprop(fdt, node, "nr_folios", &len); + if (!nr_folios || len != sizeof(*nr_folios)) + return -EINVAL; + + old_csum = fdt_getprop(fdt, node, "csum", &len); + if (!old_csum || len != sizeof(*old_csum)) + return -EINVAL; + + folios_info_phys = fdt_getprop(fdt, node, "folios_info", &len); + if (!folios_info_phys || len != sizeof(*folios_info_phys)) + return -EINVAL; + + folios_info = kho_restore_vmalloc(folios_info_phys); + if (!folios_info) + return -EINVAL; + + for (int i = 0; i < *nr_folios; i++) { + unsigned int order = folios_info[i] & ~PAGE_MASK; + phys_addr_t phys = folios_info[i] & PAGE_MASK; + unsigned int size = PAGE_SIZE << order; + struct folio *folio; + + folio = kho_restore_folio(phys); + if (!folio) + break; + + if (folio_order(folio) != order) + break; + + csum = csum_partial(folio_address(folio), size, csum); + folio_put(folio); + } + + vfree(folios_info); + + if (csum != *old_csum) + return -EINVAL; + + return 0; +} + +static int kho_test_restore(phys_addr_t fdt_phys) +{ + void *fdt = phys_to_virt(fdt_phys); + const unsigned int *magic; + int node, len, err; + + node = fdt_path_offset(fdt, "/"); + if (node < 0) + return -EINVAL; + + if (fdt_node_check_compatible(fdt, node, KHO_TEST_COMPAT)) + return -EINVAL; + + magic = fdt_getprop(fdt, node, "magic", &len); + if (!magic || len != sizeof(*magic)) + return -EINVAL; + + if (*magic != KHO_TEST_MAGIC) + return -EINVAL; + + err = kho_test_restore_data(fdt, node); + if (err) + return err; + + return 0; +} + +static int __init kho_test_init(void) +{ + phys_addr_t fdt_phys; + int err; + + if (!kho_is_enabled()) + return 0; + + err = kho_retrieve_subtree(KHO_TEST_FDT, &fdt_phys, NULL); + if (!err) { + err = kho_test_restore(fdt_phys); + if (err) + pr_err("KHO restore failed\n"); + else + pr_info("KHO restore succeeded\n"); + + return err; + } + + if (err != -ENOENT) { + pr_warn("failed to retrieve %s FDT: %d\n", KHO_TEST_FDT, err); + return err; + } + + return kho_test_save(); +} +module_init(kho_test_init); + +static void kho_test_cleanup(void) +{ + /* unpreserve and free the data stored in folios */ + kho_test_unpreserve_data(&kho_test_state); + for (int i = 0; i < kho_test_state.nr_folios; i++) + folio_put(kho_test_state.folios[i]); + + kvfree(kho_test_state.folios); + + /* Unpreserve and release the FDT folio */ + kho_unpreserve_folio(kho_test_state.fdt); + folio_put(kho_test_state.fdt); +} + +static void __exit kho_test_exit(void) +{ + kho_remove_subtree(folio_address(kho_test_state.fdt)); + kho_test_cleanup(); +} +module_exit(kho_test_exit); + +MODULE_AUTHOR("Mike Rapoport "); +MODULE_DESCRIPTION("KHO test module"); +MODULE_LICENSE("GPL"); -- cgit v1.3.1