Thomas Gleixner | 1a59d1b8 | 2019-05-27 08:55:05 +0200 | [diff] [blame] | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2 | /* Rewritten by Rusty Russell, on the backs of many others... |
| 3 | Copyright (C) 2001 Rusty Russell, 2002 Rusty Russell IBM. |
| 4 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 5 | */ |
Frederic Weisbecker | 8b96f01 | 2008-12-06 03:40:00 +0100 | [diff] [blame] | 6 | #include <linux/ftrace.h> |
Dmitri Vorobiev | f80d2d7 | 2009-03-22 19:11:10 +0200 | [diff] [blame] | 7 | #include <linux/memory.h> |
Paul Gortmaker | 8a293be | 2016-07-23 14:01:45 -0400 | [diff] [blame] | 8 | #include <linux/extable.h> |
Ingo Molnar | 505f2b9 | 2009-03-20 11:05:04 +0100 | [diff] [blame] | 9 | #include <linux/module.h> |
| 10 | #include <linux/mutex.h> |
| 11 | #include <linux/init.h> |
Masami Hiramatsu | 5b48562 | 2017-01-08 23:58:09 +0900 | [diff] [blame] | 12 | #include <linux/kprobes.h> |
Daniel Borkmann | 74451e66 | 2017-02-16 22:24:50 +0100 | [diff] [blame] | 13 | #include <linux/filter.h> |
Ingo Molnar | 505f2b9 | 2009-03-20 11:05:04 +0100 | [diff] [blame] | 14 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 15 | #include <asm/sections.h> |
Linus Torvalds | 7c0f6ba | 2016-12-24 11:46:01 -0800 | [diff] [blame] | 16 | #include <linux/uaccess.h> |
Ingo Molnar | 505f2b9 | 2009-03-20 11:05:04 +0100 | [diff] [blame] | 17 | |
| 18 | /* |
| 19 | * mutex protecting text section modification (dynamic code patching). |
| 20 | * some users need to sleep (allocating memory...) while they hold this lock. |
| 21 | * |
Zhou Chengming | e846d13 | 2017-11-02 09:18:21 +0800 | [diff] [blame] | 22 | * Note: Also protects SMP-alternatives modification on x86. |
| 23 | * |
Ingo Molnar | 505f2b9 | 2009-03-20 11:05:04 +0100 | [diff] [blame] | 24 | * NOT exported to modules - patching kernel text is a really delicate matter. |
| 25 | */ |
| 26 | DEFINE_MUTEX(text_mutex); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 27 | |
| 28 | extern struct exception_table_entry __start___ex_table[]; |
| 29 | extern struct exception_table_entry __stop___ex_table[]; |
| 30 | |
David Daney | d219e2e | 2012-04-19 14:59:56 -0700 | [diff] [blame] | 31 | /* Cleared by build time tools if the table is already sorted. */ |
Andi Kleen | 00b7103 | 2014-02-08 08:52:04 +0100 | [diff] [blame] | 32 | u32 __initdata __visible main_extable_sort_needed = 1; |
David Daney | d219e2e | 2012-04-19 14:59:56 -0700 | [diff] [blame] | 33 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 34 | /* Sort the kernel's built-in exception table */ |
| 35 | void __init sort_main_extable(void) |
| 36 | { |
Nathan Chancellor | 63174f6 | 2020-04-06 20:09:27 -0700 | [diff] [blame] | 37 | if (main_extable_sort_needed && |
| 38 | &__stop___ex_table > &__start___ex_table) { |
Borislav Petkov | bec1b9e | 2013-04-15 12:51:49 +0200 | [diff] [blame] | 39 | pr_notice("Sorting __ex_table...\n"); |
David Daney | d219e2e | 2012-04-19 14:59:56 -0700 | [diff] [blame] | 40 | sort_extable(__start___ex_table, __stop___ex_table); |
Borislav Petkov | bec1b9e | 2013-04-15 12:51:49 +0200 | [diff] [blame] | 41 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 42 | } |
| 43 | |
Santosh Sivaraj | 49ec917 | 2019-08-20 13:43:49 +0530 | [diff] [blame] | 44 | /* Given an address, look for it in the kernel exception table */ |
| 45 | const |
| 46 | struct exception_table_entry *search_kernel_exception_table(unsigned long addr) |
| 47 | { |
| 48 | return search_extable(__start___ex_table, |
| 49 | __stop___ex_table - __start___ex_table, addr); |
| 50 | } |
| 51 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 52 | /* Given an address, look for it in the exception tables. */ |
| 53 | const struct exception_table_entry *search_exception_tables(unsigned long addr) |
| 54 | { |
| 55 | const struct exception_table_entry *e; |
| 56 | |
Santosh Sivaraj | 49ec917 | 2019-08-20 13:43:49 +0530 | [diff] [blame] | 57 | e = search_kernel_exception_table(addr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 58 | if (!e) |
| 59 | e = search_module_extables(addr); |
Alexei Starovoitov | 3dec541 | 2019-10-15 20:25:03 -0700 | [diff] [blame] | 60 | if (!e) |
| 61 | e = search_bpf_extables(addr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 62 | return e; |
| 63 | } |
| 64 | |
Marcin Nowakowski | c0d80dd | 2017-07-06 15:35:31 -0700 | [diff] [blame] | 65 | int notrace core_kernel_text(unsigned long addr) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 66 | { |
Kefeng Wang | 808b645 | 2021-11-08 18:34:09 -0800 | [diff] [blame] | 67 | if (is_kernel_text(addr)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 68 | return 1; |
| 69 | |
Christophe Leroy | d2635f2 | 2021-11-05 13:40:40 -0700 | [diff] [blame] | 70 | if (system_state < SYSTEM_FREEING_INITMEM && |
Kefeng Wang | b9ad8fe | 2021-11-08 18:33:54 -0800 | [diff] [blame] | 71 | is_kernel_inittext(addr)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 72 | return 1; |
| 73 | return 0; |
| 74 | } |
| 75 | |
Frederic Weisbecker | 3861a17 | 2009-02-08 00:04:02 +0100 | [diff] [blame] | 76 | int __kernel_text_address(unsigned long addr) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 77 | { |
Steven Rostedt (VMware) | 9aadde9 | 2017-09-22 17:22:19 -0400 | [diff] [blame] | 78 | if (kernel_text_address(addr)) |
Daniel Borkmann | 74451e66 | 2017-02-16 22:24:50 +0100 | [diff] [blame] | 79 | return 1; |
Ingo Molnar | 4a44bac | 2009-03-19 13:21:44 +0100 | [diff] [blame] | 80 | /* |
| 81 | * There might be init symbols in saved stacktraces. |
| 82 | * Give those symbols a chance to be printed in |
| 83 | * backtraces (such as lockdep traces). |
| 84 | * |
| 85 | * Since we are after the module-symbols check, there's |
| 86 | * no danger of address overlap: |
| 87 | */ |
Kefeng Wang | b9ad8fe | 2021-11-08 18:33:54 -0800 | [diff] [blame] | 88 | if (is_kernel_inittext(addr)) |
Ingo Molnar | 4a44bac | 2009-03-19 13:21:44 +0100 | [diff] [blame] | 89 | return 1; |
| 90 | return 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 91 | } |
| 92 | |
| 93 | int kernel_text_address(unsigned long addr) |
| 94 | { |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 95 | bool no_rcu; |
| 96 | int ret = 1; |
| 97 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 98 | if (core_kernel_text(addr)) |
| 99 | return 1; |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 100 | |
| 101 | /* |
| 102 | * If a stack dump happens while RCU is not watching, then |
| 103 | * RCU needs to be notified that it requires to start |
| 104 | * watching again. This can happen either by tracing that |
| 105 | * triggers a stack trace, or a WARN() that happens during |
| 106 | * coming back from idle, or cpu on or offlining. |
| 107 | * |
Jiri Olsa | e9b4e60 | 2020-01-23 17:15:07 +0100 | [diff] [blame] | 108 | * is_module_text_address() as well as the kprobe slots, |
| 109 | * is_bpf_text_address() and is_bpf_image_address require |
| 110 | * RCU to be watching. |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 111 | */ |
| 112 | no_rcu = !rcu_is_watching(); |
| 113 | |
| 114 | /* Treat this like an NMI as it can happen anywhere */ |
| 115 | if (no_rcu) |
| 116 | rcu_nmi_enter(); |
| 117 | |
Steven Rostedt (Red Hat) | aec0be2 | 2014-11-18 21:14:11 -0500 | [diff] [blame] | 118 | if (is_module_text_address(addr)) |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 119 | goto out; |
Masami Hiramatsu | 5b48562 | 2017-01-08 23:58:09 +0900 | [diff] [blame] | 120 | if (is_ftrace_trampoline(addr)) |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 121 | goto out; |
Masami Hiramatsu | 5b48562 | 2017-01-08 23:58:09 +0900 | [diff] [blame] | 122 | if (is_kprobe_optinsn_slot(addr) || is_kprobe_insn_slot(addr)) |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 123 | goto out; |
Daniel Borkmann | 74451e66 | 2017-02-16 22:24:50 +0100 | [diff] [blame] | 124 | if (is_bpf_text_address(addr)) |
Steven Rostedt (VMware) | e8cac8b | 2017-09-22 17:36:32 -0400 | [diff] [blame] | 125 | goto out; |
| 126 | ret = 0; |
| 127 | out: |
| 128 | if (no_rcu) |
| 129 | rcu_nmi_exit(); |
| 130 | |
| 131 | return ret; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 132 | } |
Arjan van de Ven | ab7476c | 2008-08-15 15:29:38 -0700 | [diff] [blame] | 133 | |
| 134 | /* |
| 135 | * On some architectures (PPC64, IA64) function pointers |
| 136 | * are actually only tokens to some data that then holds the |
| 137 | * real function address. As a result, to find if a function |
| 138 | * pointer is part of the kernel text, we need to do some |
| 139 | * special dereferencing first. |
| 140 | */ |
| 141 | int func_ptr_is_kernel_text(void *ptr) |
| 142 | { |
| 143 | unsigned long addr; |
| 144 | addr = (unsigned long) dereference_function_descriptor(ptr); |
| 145 | if (core_kernel_text(addr)) |
| 146 | return 1; |
Rusty Russell | a6e6abd | 2009-03-31 13:05:31 -0600 | [diff] [blame] | 147 | return is_module_text_address(addr); |
Arjan van de Ven | ab7476c | 2008-08-15 15:29:38 -0700 | [diff] [blame] | 148 | } |