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/error-inject.c | 239 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 239 insertions(+) create mode 100644 lib/error-inject.c (limited to 'lib/error-inject.c') diff --git a/lib/error-inject.c b/lib/error-inject.c new file mode 100644 index 000000000..32f3d1ca9 --- /dev/null +++ b/lib/error-inject.c @@ -0,0 +1,239 @@ +// SPDX-License-Identifier: GPL-2.0 +// error-inject.c: Function-level error injection table +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* Whitelist of symbols that can be overridden for error injection. */ +static LIST_HEAD(error_injection_list); +static DEFINE_MUTEX(ei_mutex); +struct ei_entry { + struct list_head list; + unsigned long start_addr; + unsigned long end_addr; + int etype; + void *priv; +}; + +bool within_error_injection_list(unsigned long addr) +{ + struct ei_entry *ent; + bool ret = false; + + mutex_lock(&ei_mutex); + list_for_each_entry(ent, &error_injection_list, list) { + if (addr >= ent->start_addr && addr < ent->end_addr) { + ret = true; + break; + } + } + mutex_unlock(&ei_mutex); + return ret; +} + +int get_injectable_error_type(unsigned long addr) +{ + struct ei_entry *ent; + int ei_type = -EINVAL; + + mutex_lock(&ei_mutex); + list_for_each_entry(ent, &error_injection_list, list) { + if (addr >= ent->start_addr && addr < ent->end_addr) { + ei_type = ent->etype; + break; + } + } + mutex_unlock(&ei_mutex); + + return ei_type; +} + +/* + * Lookup and populate the error_injection_list. + * + * For safety reasons we only allow certain functions to be overridden with + * bpf_error_injection, so we need to populate the list of the symbols that have + * been marked as safe for overriding. + */ +static void populate_error_injection_list(struct error_injection_entry *start, + struct error_injection_entry *end, + void *priv) +{ + struct error_injection_entry *iter; + struct ei_entry *ent; + unsigned long entry, offset = 0, size = 0; + + mutex_lock(&ei_mutex); + for (iter = start; iter < end; iter++) { + entry = (unsigned long)dereference_symbol_descriptor((void *)iter->addr); + + if (!kernel_text_address(entry) || + !kallsyms_lookup_size_offset(entry, &size, &offset)) { + pr_err("Failed to find error inject entry at %p\n", + (void *)entry); + continue; + } + + ent = kmalloc_obj(*ent); + if (!ent) + break; + ent->start_addr = entry; + ent->end_addr = entry + size; + ent->etype = iter->etype; + ent->priv = priv; + INIT_LIST_HEAD(&ent->list); + list_add_tail(&ent->list, &error_injection_list); + } + mutex_unlock(&ei_mutex); +} + +/* Markers of the _error_inject_whitelist section */ +extern struct error_injection_entry __start_error_injection_whitelist[]; +extern struct error_injection_entry __stop_error_injection_whitelist[]; + +static void __init populate_kernel_ei_list(void) +{ + populate_error_injection_list(__start_error_injection_whitelist, + __stop_error_injection_whitelist, + NULL); +} + +#ifdef CONFIG_MODULES +static void module_load_ei_list(struct module *mod) +{ + if (!mod->num_ei_funcs) + return; + + populate_error_injection_list(mod->ei_funcs, + mod->ei_funcs + mod->num_ei_funcs, mod); +} + +static void module_unload_ei_list(struct module *mod) +{ + struct ei_entry *ent, *n; + + if (!mod->num_ei_funcs) + return; + + mutex_lock(&ei_mutex); + list_for_each_entry_safe(ent, n, &error_injection_list, list) { + if (ent->priv == mod) { + list_del_init(&ent->list); + kfree(ent); + } + } + mutex_unlock(&ei_mutex); +} + +/* Module notifier call back, checking error injection table on the module */ +static int ei_module_callback(struct notifier_block *nb, + unsigned long val, void *data) +{ + struct module *mod = data; + + if (val == MODULE_STATE_COMING) + module_load_ei_list(mod); + else if (val == MODULE_STATE_GOING) + module_unload_ei_list(mod); + + return NOTIFY_DONE; +} + +static struct notifier_block ei_module_nb = { + .notifier_call = ei_module_callback, + .priority = 0 +}; + +static __init int module_ei_init(void) +{ + return register_module_notifier(&ei_module_nb); +} +#else /* !CONFIG_MODULES */ +#define module_ei_init() (0) +#endif + +/* + * error_injection/whitelist -- shows which functions can be overridden for + * error injection. + */ +static void *ei_seq_start(struct seq_file *m, loff_t *pos) +{ + mutex_lock(&ei_mutex); + return seq_list_start(&error_injection_list, *pos); +} + +static void ei_seq_stop(struct seq_file *m, void *v) +{ + mutex_unlock(&ei_mutex); +} + +static void *ei_seq_next(struct seq_file *m, void *v, loff_t *pos) +{ + return seq_list_next(v, &error_injection_list, pos); +} + +static const char *error_type_string(int etype) +{ + switch (etype) { + case EI_ETYPE_NULL: + return "NULL"; + case EI_ETYPE_ERRNO: + return "ERRNO"; + case EI_ETYPE_ERRNO_NULL: + return "ERRNO_NULL"; + case EI_ETYPE_TRUE: + return "TRUE"; + default: + return "(unknown)"; + } +} + +static int ei_seq_show(struct seq_file *m, void *v) +{ + struct ei_entry *ent = list_entry(v, struct ei_entry, list); + + seq_printf(m, "%ps\t%s\n", (void *)ent->start_addr, + error_type_string(ent->etype)); + return 0; +} + +static const struct seq_operations ei_sops = { + .start = ei_seq_start, + .next = ei_seq_next, + .stop = ei_seq_stop, + .show = ei_seq_show, +}; + +DEFINE_SEQ_ATTRIBUTE(ei); + +static int __init ei_debugfs_init(void) +{ + struct dentry *dir, *file; + + dir = debugfs_create_dir("error_injection", NULL); + + file = debugfs_create_file("list", 0444, dir, NULL, &ei_fops); + if (IS_ERR(file)) { + debugfs_remove(dir); + return PTR_ERR(file); + } + + return 0; +} + +static int __init init_error_injection(void) +{ + populate_kernel_ei_list(); + + if (!module_ei_init()) + ei_debugfs_init(); + + return 0; +} +late_initcall(init_error_injection); -- cgit v1.3.1