linux - kernel/module/main.c

3358 lines · 85.3 KiB

1	// SPDX-License-Identifier: GPL-2.0-or-later
2	/*
3	* Copyright (C) 2002 Richard Henderson
4	* Copyright (C) 2001 Rusty Russell, 2002, 2010 Rusty Russell IBM.
5	* Copyright (C) 2023 Luis Chamberlain <[email protected]>
6	*/
7
8	#define INCLUDE_VERMAGIC
9
10	#include <linux/export.h>
11	#include <linux/extable.h>
12	#include <linux/moduleloader.h>
13	#include <linux/module_signature.h>
14	#include <linux/trace_events.h>
15	#include <linux/init.h>
16	#include <linux/kallsyms.h>
17	#include <linux/buildid.h>
18	#include <linux/fs.h>
19	#include <linux/kernel.h>
20	#include <linux/kernel_read_file.h>
21	#include <linux/kstrtox.h>
22	#include <linux/slab.h>
23	#include <linux/vmalloc.h>
24	#include <linux/elf.h>
25	#include <linux/seq_file.h>
26	#include <linux/syscalls.h>
27	#include <linux/fcntl.h>
28	#include <linux/rcupdate.h>
29	#include <linux/capability.h>
30	#include <linux/cpu.h>
31	#include <linux/moduleparam.h>
32	#include <linux/errno.h>
33	#include <linux/err.h>
34	#include <linux/vermagic.h>
35	#include <linux/notifier.h>
36	#include <linux/sched.h>
37	#include <linux/device.h>
38	#include <linux/string.h>
39	#include <linux/mutex.h>
40	#include <linux/rculist.h>
41	#include <linux/uaccess.h>
42	#include <asm/cacheflush.h>
43	#include <linux/set_memory.h>
44	#include <asm/mmu_context.h>
45	#include <linux/license.h>
46	#include <asm/sections.h>
47	#include <linux/tracepoint.h>
48	#include <linux/ftrace.h>
49	#include <linux/livepatch.h>
50	#include <linux/async.h>
51	#include <linux/percpu.h>
52	#include <linux/kmemleak.h>
53	#include <linux/jump_label.h>
54	#include <linux/pfn.h>
55	#include <linux/bsearch.h>
56	#include <linux/dynamic_debug.h>
57	#include <linux/audit.h>
58	#include <linux/cfi.h>
59	#include <linux/debugfs.h>
60	#include <uapi/linux/module.h>
61	#include "internal.h"
62
63	#define CREATE_TRACE_POINTS
64	#include <trace/events/module.h>
65
66	/*
67	* Mutex protects:
68	* 1) List of modules (also safely readable with preempt_disable),
69	* 2) module_use links,
70	* 3) mod_tree.addr_min/mod_tree.addr_max.
71	* (delete and add uses RCU list operations).
72	*/
73	DEFINE_MUTEX(module_mutex);
74	LIST_HEAD(modules);
75
76	/* Work queue for freeing init sections in success case */
77	static void do_free_init(struct work_struct *w);
78	static DECLARE_WORK(init_free_wq, do_free_init);
79	static LLIST_HEAD(init_free_list);
80
81	struct mod_tree_root mod_tree __cacheline_aligned = {
82	.addr_min = -1UL,
83	};
84
85	struct symsearch {
86	const struct kernel_symbol start, stop;
87	const s32 *crcs;
88	enum mod_license license;
89	};
90
91	/*
92	* Bounds of module memory, for speeding up __module_address.
93	* Protected by module_mutex.
94	*/
95	static void __mod_update_bounds(enum mod_mem_type type __maybe_unused, void *base,
96	unsigned int size, struct mod_tree_root *tree)
97	{
98	unsigned long min = (unsigned long)base;
99	unsigned long max = min + size;
100
101	#ifdef CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC
102	if (mod_mem_type_is_core_data(type)) {
103	if (min < tree->data_addr_min)
104	tree->data_addr_min = min;
105	if (max > tree->data_addr_max)
106	tree->data_addr_max = max;
107	return;
108	}
109	#endif
110	if (min < tree->addr_min)
111	tree->addr_min = min;
112	if (max > tree->addr_max)
113	tree->addr_max = max;
114	}
115
116	static void mod_update_bounds(struct module *mod)
117	{
118	for_each_mod_mem_type(type) {
119	struct module_memory *mod_mem = &mod->mem[type];
120
121	if (mod_mem->size)
122	__mod_update_bounds(type, mod_mem->base, mod_mem->size, &mod_tree);
123	}
124	}
125
126	/* Block module loading/unloading? */
127	int modules_disabled;
128	core_param(nomodule, modules_disabled, bint, 0);
129
130	/* Waiting for a module to finish initializing? */
131	static DECLARE_WAIT_QUEUE_HEAD(module_wq);
132
133	static BLOCKING_NOTIFIER_HEAD(module_notify_list);
134
135	int register_module_notifier(struct notifier_block *nb)
136	{
137	return blocking_notifier_chain_register(&module_notify_list, nb);
138	}
139	EXPORT_SYMBOL(register_module_notifier);
140
141	int unregister_module_notifier(struct notifier_block *nb)
142	{
143	return blocking_notifier_chain_unregister(&module_notify_list, nb);
144	}
145	EXPORT_SYMBOL(unregister_module_notifier);
146
147	/*
148	* We require a truly strong try_module_get(): 0 means success.
149	* Otherwise an error is returned due to ongoing or failed
150	* initialization etc.
151	*/
152	static inline int strong_try_module_get(struct module *mod)
153	{
154	BUG_ON(mod && mod->state == MODULE_STATE_UNFORMED);
155	if (mod && mod->state == MODULE_STATE_COMING)
156	return -EBUSY;
157	if (try_module_get(mod))
158	return 0;
159	else
160	return -ENOENT;
161	}
162
163	static inline void add_taint_module(struct module *mod, unsigned flag,
164	enum lockdep_ok lockdep_ok)
165	{
166	add_taint(flag, lockdep_ok);
167	set_bit(flag, &mod->taints);
168	}
169
170	/*
171	* A thread that wants to hold a reference to a module only while it
172	* is running can call this to safely exit.
173	*/
174	void __noreturn __module_put_and_kthread_exit(struct module *mod, long code)
175	{
176	module_put(mod);
177	kthread_exit(code);
178	}
179	EXPORT_SYMBOL(__module_put_and_kthread_exit);
180
181	/* Find a module section: 0 means not found. */
182	static unsigned int find_sec(const struct load_info info, const char name)
183	{
184	unsigned int i;
185
186	for (i = 1; i < info->hdr->e_shnum; i++) {
187	Elf_Shdr *shdr = &info->sechdrs[i];
188	/* Alloc bit cleared means "ignore it." */
189	if ((shdr->sh_flags & SHF_ALLOC)
190	&& strcmp(info->secstrings + shdr->sh_name, name) == 0)
191	return i;
192	}
193	return 0;
194	}
195
196	/* Find a module section, or NULL. */
197	static void section_addr(const struct load_info info, const char *name)
198	{
199	/* Section 0 has sh_addr 0. */
200	return (void *)info->sechdrs[find_sec(info, name)].sh_addr;
201	}
202
203	/* Find a module section, or NULL. Fill in number of "objects" in section. */
204	static void section_objs(const struct load_info info,
205	const char *name,
206	size_t object_size,
207	unsigned int *num)
208	{
209	unsigned int sec = find_sec(info, name);
210
211	/* Section 0 has sh_addr 0 and sh_size 0. */
212	*num = info->sechdrs[sec].sh_size / object_size;
213	return (void *)info->sechdrs[sec].sh_addr;
214	}
215
216	/* Find a module section: 0 means not found. Ignores SHF_ALLOC flag. */
217	static unsigned int find_any_sec(const struct load_info info, const char name)
218	{
219	unsigned int i;
220
221	for (i = 1; i < info->hdr->e_shnum; i++) {
222	Elf_Shdr *shdr = &info->sechdrs[i];
223	if (strcmp(info->secstrings + shdr->sh_name, name) == 0)
224	return i;
225	}
226	return 0;
227	}
228
229	/*
230	* Find a module section, or NULL. Fill in number of "objects" in section.
231	* Ignores SHF_ALLOC flag.
232	*/
233	static __maybe_unused void any_section_objs(const struct load_info info,
234	const char *name,
235	size_t object_size,
236	unsigned int *num)
237	{
238	unsigned int sec = find_any_sec(info, name);
239
240	/* Section 0 has sh_addr 0 and sh_size 0. */
241	*num = info->sechdrs[sec].sh_size / object_size;
242	return (void *)info->sechdrs[sec].sh_addr;
243	}
244
245	#ifndef CONFIG_MODVERSIONS
246	#define symversion(base, idx) NULL
247	#else
248	#define symversion(base, idx) ((base != NULL) ? ((base) + (idx)) : NULL)
249	#endif
250
251	static const char kernel_symbol_name(const struct kernel_symbol sym)
252	{
253	#ifdef CONFIG_HAVE_ARCH_PREL32_RELOCATIONS
254	return offset_to_ptr(&sym->name_offset);
255	#else
256	return sym->name;
257	#endif
258	}
259
260	static const char kernel_symbol_namespace(const struct kernel_symbol sym)
261	{
262	#ifdef CONFIG_HAVE_ARCH_PREL32_RELOCATIONS
263	if (!sym->namespace_offset)
264	return NULL;
265	return offset_to_ptr(&sym->namespace_offset);
266	#else
267	return sym->namespace;
268	#endif
269	}
270
271	int cmp_name(const void name, const void sym)
272	{
273	return strcmp(name, kernel_symbol_name(sym));
274	}
275
276	static bool find_exported_symbol_in_section(const struct symsearch *syms,
277	struct module *owner,
278	struct find_symbol_arg *fsa)
279	{
280	struct kernel_symbol *sym;
281
282	if (!fsa->gplok && syms->license == GPL_ONLY)
283	return false;
284
285	sym = bsearch(fsa->name, syms->start, syms->stop - syms->start,
286	sizeof(struct kernel_symbol), cmp_name);
287	if (!sym)
288	return false;
289
290	fsa->owner = owner;
291	fsa->crc = symversion(syms->crcs, sym - syms->start);
292	fsa->sym = sym;
293	fsa->license = syms->license;
294
295	return true;
296	}
297
298	/*
299	* Find an exported symbol and return it, along with, (optional) crc and
300	* (optional) module which owns it. Needs preempt disabled or module_mutex.
301	*/
302	bool find_symbol(struct find_symbol_arg *fsa)
303	{
304	static const struct symsearch arr[] = {
305	{ __start___ksymtab, __stop___ksymtab, __start___kcrctab,
306	NOT_GPL_ONLY },
307	{ __start___ksymtab_gpl, __stop___ksymtab_gpl,
308	__start___kcrctab_gpl,
309	GPL_ONLY },
310	};
311	struct module *mod;
312	unsigned int i;
313
314	module_assert_mutex_or_preempt();
315
316	for (i = 0; i < ARRAY_SIZE(arr); i++)
317	if (find_exported_symbol_in_section(&arr[i], NULL, fsa))
318	return true;
319
320	list_for_each_entry_rcu(mod, &modules, list,
321	lockdep_is_held(&module_mutex)) {
322	struct symsearch arr[] = {
323	{ mod->syms, mod->syms + mod->num_syms, mod->crcs,
324	NOT_GPL_ONLY },
325	{ mod->gpl_syms, mod->gpl_syms + mod->num_gpl_syms,
326	mod->gpl_crcs,
327	GPL_ONLY },
328	};
329
330	if (mod->state == MODULE_STATE_UNFORMED)
331	continue;
332
333	for (i = 0; i < ARRAY_SIZE(arr); i++)
334	if (find_exported_symbol_in_section(&arr[i], mod, fsa))
335	return true;
336	}
337
338	pr_debug("Failed to find symbol %s\n", fsa->name);
339	return false;
340	}
341
342	/*
343	* Search for module by name: must hold module_mutex (or preempt disabled
344	* for read-only access).
345	*/
346	struct module find_module_all(const char name, size_t len,
347	bool even_unformed)
348	{
349	struct module *mod;
350
351	module_assert_mutex_or_preempt();
352
353	list_for_each_entry_rcu(mod, &modules, list,
354	lockdep_is_held(&module_mutex)) {
355	if (!even_unformed && mod->state == MODULE_STATE_UNFORMED)
356	continue;
357	if (strlen(mod->name) == len && !memcmp(mod->name, name, len))
358	return mod;
359	}
360	return NULL;
361	}
362
363	struct module find_module(const char name)
364	{
365	return find_module_all(name, strlen(name), false);
366	}
367
368	#ifdef CONFIG_SMP
369
370	static inline void __percpu mod_percpu(struct module mod)
371	{
372	return mod->percpu;
373	}
374
375	static int percpu_modalloc(struct module mod, struct load_info info)
376	{
377	Elf_Shdr *pcpusec = &info->sechdrs[info->index.pcpu];
378	unsigned long align = pcpusec->sh_addralign;
379
380	if (!pcpusec->sh_size)
381	return 0;
382
383	if (align > PAGE_SIZE) {
384	pr_warn("%s: per-cpu alignment %li > %li\n",
385	mod->name, align, PAGE_SIZE);
386	align = PAGE_SIZE;
387	}
388
389	mod->percpu = __alloc_reserved_percpu(pcpusec->sh_size, align);
390	if (!mod->percpu) {
391	pr_warn("%s: Could not allocate %lu bytes percpu data\n",
392	mod->name, (unsigned long)pcpusec->sh_size);
393	return -ENOMEM;
394	}
395	mod->percpu_size = pcpusec->sh_size;
396	return 0;
397	}
398
399	static void percpu_modfree(struct module *mod)
400	{
401	free_percpu(mod->percpu);
402	}
403
404	static unsigned int find_pcpusec(struct load_info *info)
405	{
406	return find_sec(info, ".data..percpu");
407	}
408
409	static void percpu_modcopy(struct module *mod,
410	const void *from, unsigned long size)
411	{
412	int cpu;
413
414	for_each_possible_cpu(cpu)
415	memcpy(per_cpu_ptr(mod->percpu, cpu), from, size);
416	}
417
418	bool __is_module_percpu_address(unsigned long addr, unsigned long *can_addr)
419	{
420	struct module *mod;
421	unsigned int cpu;
422
423	preempt_disable();
424
425	list_for_each_entry_rcu(mod, &modules, list) {
426	if (mod->state == MODULE_STATE_UNFORMED)
427	continue;
428	if (!mod->percpu_size)
429	continue;
430	for_each_possible_cpu(cpu) {
431	void *start = per_cpu_ptr(mod->percpu, cpu);
432	void va = (void )addr;
433
434	if (va >= start && va < start + mod->percpu_size) {
435	if (can_addr) {
436	*can_addr = (unsigned long) (va - start);
437	*can_addr += (unsigned long)
438	per_cpu_ptr(mod->percpu,
439	get_boot_cpu_id());
440	}
441	preempt_enable();
442	return true;
443	}
444	}
445	}
446
447	preempt_enable();
448	return false;
449	}
450
451	/**
452	* is_module_percpu_address() - test whether address is from module static percpu
453	* @addr: address to test
454	*
455	* Test whether @addr belongs to module static percpu area.
456	*
457	* Return: %true if @addr is from module static percpu area
458	*/
459	bool is_module_percpu_address(unsigned long addr)
460	{
461	return __is_module_percpu_address(addr, NULL);
462	}
463
464	#else /* ... !CONFIG_SMP */
465
466	static inline void __percpu mod_percpu(struct module mod)
467	{
468	return NULL;
469	}
470	static int percpu_modalloc(struct module mod, struct load_info info)
471	{
472	/* UP modules shouldn't have this section: ENOMEM isn't quite right */
473	if (info->sechdrs[info->index.pcpu].sh_size != 0)
474	return -ENOMEM;
475	return 0;
476	}
477	static inline void percpu_modfree(struct module *mod)
478	{
479	}
480	static unsigned int find_pcpusec(struct load_info *info)
481	{
482	return 0;
483	}
484	static inline void percpu_modcopy(struct module *mod,
485	const void *from, unsigned long size)
486	{
487	/* pcpusec should be 0, and size of that section should be 0. */
488	BUG_ON(size != 0);
489	}
490	bool is_module_percpu_address(unsigned long addr)
491	{
492	return false;
493	}
494
495	bool __is_module_percpu_address(unsigned long addr, unsigned long *can_addr)
496	{
497	return false;
498	}
499
500	#endif /* CONFIG_SMP */
501
502	#define MODINFO_ATTR(field) \
503	static void setup_modinfo_##field(struct module mod, const char s) \
504	{ \
505	mod->field = kstrdup(s, GFP_KERNEL); \
506	} \
507	static ssize_t show_modinfo_##field(struct module_attribute *mattr, \
508	struct module_kobject mk, char buffer) \
509	{ \
510	return scnprintf(buffer, PAGE_SIZE, "%s\n", mk->mod->field); \
511	} \
512	static int modinfo_##field##_exists(struct module *mod) \
513	{ \
514	return mod->field != NULL; \
515	} \
516	static void free_modinfo_##field(struct module *mod) \
517	{ \
518	kfree(mod->field); \
519	mod->field = NULL; \
520	} \
521	static struct module_attribute modinfo_##field = { \
522	.attr = { .name = __stringify(field), .mode = 0444 }, \
523	.show = show_modinfo_##field, \
524	.setup = setup_modinfo_##field, \
525	.test = modinfo_##field##_exists, \
526	.free = free_modinfo_##field, \
527	};
528
529	MODINFO_ATTR(version);
530	MODINFO_ATTR(srcversion);
531
532	static struct {
533	char name[MODULE_NAME_LEN + 1];
534	char taints[MODULE_FLAGS_BUF_SIZE];
535	} last_unloaded_module;
536
537	#ifdef CONFIG_MODULE_UNLOAD
538
539	EXPORT_TRACEPOINT_SYMBOL(module_get);
540
541	/* MODULE_REF_BASE is the base reference count by kmodule loader. */
542	#define MODULE_REF_BASE 1
543
544	/* Init the unload section of the module. */
545	static int module_unload_init(struct module *mod)
546	{
547	/*
548	* Initialize reference counter to MODULE_REF_BASE.
549	* refcnt == 0 means module is going.
550	*/
551	atomic_set(&mod->refcnt, MODULE_REF_BASE);
552
553	INIT_LIST_HEAD(&mod->source_list);
554	INIT_LIST_HEAD(&mod->target_list);
555
556	/* Hold reference count during initialization. */
557	atomic_inc(&mod->refcnt);
558
559	return 0;
560	}
561
562	/* Does a already use b? */
563	static int already_uses(struct module a, struct module b)
564	{
565	struct module_use *use;
566
567	list_for_each_entry(use, &b->source_list, source_list) {
568	if (use->source == a)
569	return 1;
570	}
571	pr_debug("%s does not use %s!\n", a->name, b->name);
572	return 0;
573	}
574
575	/*
576	* Module a uses b
577	* - we add 'a' as a "source", 'b' as a "target" of module use
578	* - the module_use is added to the list of 'b' sources (so
579	* 'b' can walk the list to see who sourced them), and of 'a'
580	* targets (so 'a' can see what modules it targets).
581	*/
582	static int add_module_usage(struct module a, struct module b)
583	{
584	struct module_use *use;
585
586	pr_debug("Allocating new usage for %s.\n", a->name);
587	use = kmalloc(sizeof(*use), GFP_ATOMIC);
588	if (!use)
589	return -ENOMEM;
590
591	use->source = a;
592	use->target = b;
593	list_add(&use->source_list, &b->source_list);
594	list_add(&use->target_list, &a->target_list);
595	return 0;
596	}
597
598	/* Module a uses b: caller needs module_mutex() */
599	static int ref_module(struct module a, struct module b)
600	{
601	int err;
602
603	if (b == NULL \|\| already_uses(a, b))
604	return 0;
605
606	/* If module isn't available, we fail. */
607	err = strong_try_module_get(b);
608	if (err)
609	return err;
610
611	err = add_module_usage(a, b);
612	if (err) {
613	module_put(b);
614	return err;
615	}
616	return 0;
617	}
618
619	/* Clear the unload stuff of the module. */
620	static void module_unload_free(struct module *mod)
621	{
622	struct module_use use, tmp;
623
624	mutex_lock(&module_mutex);
625	list_for_each_entry_safe(use, tmp, &mod->target_list, target_list) {
626	struct module *i = use->target;
627	pr_debug("%s unusing %s\n", mod->name, i->name);
628	module_put(i);
629	list_del(&use->source_list);
630	list_del(&use->target_list);
631	kfree(use);
632	}
633	mutex_unlock(&module_mutex);
634	}
635
636	#ifdef CONFIG_MODULE_FORCE_UNLOAD
637	static inline int try_force_unload(unsigned int flags)
638	{
639	int ret = (flags & O_TRUNC);
640	if (ret)
641	add_taint(TAINT_FORCED_RMMOD, LOCKDEP_NOW_UNRELIABLE);
642	return ret;
643	}
644	#else
645	static inline int try_force_unload(unsigned int flags)
646	{
647	return 0;
648	}
649	#endif /* CONFIG_MODULE_FORCE_UNLOAD */
650
651	/* Try to release refcount of module, 0 means success. */
652	static int try_release_module_ref(struct module *mod)
653	{
654	int ret;
655
656	/* Try to decrement refcnt which we set at loading */
657	ret = atomic_sub_return(MODULE_REF_BASE, &mod->refcnt);
658	BUG_ON(ret < 0);
659	if (ret)
660	/* Someone can put this right now, recover with checking */
661	ret = atomic_add_unless(&mod->refcnt, MODULE_REF_BASE, 0);
662
663	return ret;
664	}
665
666	static int try_stop_module(struct module mod, int flags, int forced)
667	{
668	/* If it's not unused, quit unless we're forcing. */
669	if (try_release_module_ref(mod) != 0) {
670	*forced = try_force_unload(flags);
671	if (!(*forced))
672	return -EWOULDBLOCK;
673	}
674
675	/* Mark it as dying. */
676	mod->state = MODULE_STATE_GOING;
677
678	return 0;
679	}
680
681	/**
682	* module_refcount() - return the refcount or -1 if unloading
683	* @mod: the module we're checking
684	*
685	* Return:
686	* -1 if the module is in the process of unloading
687	* otherwise the number of references in the kernel to the module
688	*/
689	int module_refcount(struct module *mod)
690	{
691	return atomic_read(&mod->refcnt) - MODULE_REF_BASE;
692	}
693	EXPORT_SYMBOL(module_refcount);
694
695	/* This exists whether we can unload or not */
696	static void free_module(struct module *mod);
697
698	SYSCALL_DEFINE2(delete_module, const char __user *, name_user,
699	unsigned int, flags)
700	{
701	struct module *mod;
702	char name[MODULE_NAME_LEN];
703	char buf[MODULE_FLAGS_BUF_SIZE];
704	int ret, forced = 0;
705
706	if (!capable(CAP_SYS_MODULE) \|\| modules_disabled)
707	return -EPERM;
708
709	if (strncpy_from_user(name, name_user, MODULE_NAME_LEN-1) < 0)
710	return -EFAULT;
711	name[MODULE_NAME_LEN-1] = '\0';
712
713	audit_log_kern_module(name);
714
715	if (mutex_lock_interruptible(&module_mutex) != 0)
716	return -EINTR;
717
718	mod = find_module(name);
719	if (!mod) {
720	ret = -ENOENT;
721	goto out;
722	}
723
724	if (!list_empty(&mod->source_list)) {
725	/* Other modules depend on us: get rid of them first. */
726	ret = -EWOULDBLOCK;
727	goto out;
728	}
729
730	/* Doing init or already dying? */
731	if (mod->state != MODULE_STATE_LIVE) {
732	/* FIXME: if (force), slam module count damn the torpedoes */
733	pr_debug("%s already dying\n", mod->name);
734	ret = -EBUSY;
735	goto out;
736	}
737
738	/* If it has an init func, it must have an exit func to unload */
739	if (mod->init && !mod->exit) {
740	forced = try_force_unload(flags);
741	if (!forced) {
742	/* This module can't be removed */
743	ret = -EBUSY;
744	goto out;
745	}
746	}
747
748	ret = try_stop_module(mod, flags, &forced);
749	if (ret != 0)
750	goto out;
751
752	mutex_unlock(&module_mutex);
753	/* Final destruction now no one is using it. */
754	if (mod->exit != NULL)
755	mod->exit();
756	blocking_notifier_call_chain(&module_notify_list,
757	MODULE_STATE_GOING, mod);
758	klp_module_going(mod);
759	ftrace_release_mod(mod);
760
761	async_synchronize_full();
762
763	/* Store the name and taints of the last unloaded module for diagnostic purposes */
764	strscpy(last_unloaded_module.name, mod->name, sizeof(last_unloaded_module.name));
765	strscpy(last_unloaded_module.taints, module_flags(mod, buf, false), sizeof(last_unloaded_module.taints));
766
767	free_module(mod);
768	/* someone could wait for the module in add_unformed_module() */
769	wake_up_all(&module_wq);
770	return 0;
771	out:
772	mutex_unlock(&module_mutex);
773	return ret;
774	}
775
776	void __symbol_put(const char *symbol)
777	{
778	struct find_symbol_arg fsa = {
779	.name = symbol,
780	.gplok = true,
781	};
782
783	preempt_disable();
784	BUG_ON(!find_symbol(&fsa));
785	module_put(fsa.owner);
786	preempt_enable();
787	}
788	EXPORT_SYMBOL(__symbol_put);
789
790	/* Note this assumes addr is a function, which it currently always is. */
791	void symbol_put_addr(void *addr)
792	{
793	struct module *modaddr;
794	unsigned long a = (unsigned long)dereference_function_descriptor(addr);
795
796	if (core_kernel_text(a))
797	return;
798
799	/*
800	* Even though we hold a reference on the module; we still need to
801	* disable preemption in order to safely traverse the data structure.
802	*/
803	preempt_disable();
804	modaddr = __module_text_address(a);
805	BUG_ON(!modaddr);
806	module_put(modaddr);
807	preempt_enable();
808	}
809	EXPORT_SYMBOL_GPL(symbol_put_addr);
810
811	static ssize_t show_refcnt(struct module_attribute *mattr,
812	struct module_kobject mk, char buffer)
813	{
814	return sprintf(buffer, "%i\n", module_refcount(mk->mod));
815	}
816
817	static struct module_attribute modinfo_refcnt =
818	__ATTR(refcnt, 0444, show_refcnt, NULL);
819
820	void __module_get(struct module *module)
821	{
822	if (module) {
823	atomic_inc(&module->refcnt);
824	trace_module_get(module, _RET_IP_);
825	}
826	}
827	EXPORT_SYMBOL(__module_get);
828
829	bool try_module_get(struct module *module)
830	{
831	bool ret = true;
832
833	if (module) {
834	/* Note: here, we can fail to get a reference */
835	if (likely(module_is_live(module) &&
836	atomic_inc_not_zero(&module->refcnt) != 0))
837	trace_module_get(module, _RET_IP_);
838	else
839	ret = false;
840	}
841	return ret;
842	}
843	EXPORT_SYMBOL(try_module_get);
844
845	void module_put(struct module *module)
846	{
847	int ret;
848
849	if (module) {
850	ret = atomic_dec_if_positive(&module->refcnt);
851	WARN_ON(ret < 0); /* Failed to put refcount */
852	trace_module_put(module, _RET_IP_);
853	}
854	}
855	EXPORT_SYMBOL(module_put);
856
857	#else /* !CONFIG_MODULE_UNLOAD */
858	static inline void module_unload_free(struct module *mod)
859	{
860	}
861
862	static int ref_module(struct module a, struct module b)
863	{
864	return strong_try_module_get(b);
865	}
866
867	static inline int module_unload_init(struct module *mod)
868	{
869	return 0;
870	}
871	#endif /* CONFIG_MODULE_UNLOAD */
872
873	size_t module_flags_taint(unsigned long taints, char *buf)
874	{
875	size_t l = 0;
876	int i;
877
878	for (i = 0; i < TAINT_FLAGS_COUNT; i++) {
879	if (taint_flags[i].module && test_bit(i, &taints))
880	buf[l++] = taint_flags[i].c_true;
881	}
882
883	return l;
884	}
885
886	static ssize_t show_initstate(struct module_attribute *mattr,
887	struct module_kobject mk, char buffer)
888	{
889	const char *state = "unknown";
890
891	switch (mk->mod->state) {
892	case MODULE_STATE_LIVE:
893	state = "live";
894	break;
895	case MODULE_STATE_COMING:
896	state = "coming";
897	break;
898	case MODULE_STATE_GOING:
899	state = "going";
900	break;
901	default:
902	BUG();
903	}
904	return sprintf(buffer, "%s\n", state);
905	}
906
907	static struct module_attribute modinfo_initstate =
908	__ATTR(initstate, 0444, show_initstate, NULL);
909
910	static ssize_t store_uevent(struct module_attribute *mattr,
911	struct module_kobject *mk,
912	const char *buffer, size_t count)
913	{
914	int rc;
915
916	rc = kobject_synth_uevent(&mk->kobj, buffer, count);
917	return rc ? rc : count;
918	}
919
920	struct module_attribute module_uevent =
921	__ATTR(uevent, 0200, NULL, store_uevent);
922
923	static ssize_t show_coresize(struct module_attribute *mattr,
924	struct module_kobject mk, char buffer)
925	{
926	unsigned int size = mk->mod->mem[MOD_TEXT].size;
927
928	if (!IS_ENABLED(CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC)) {
929	for_class_mod_mem_type(type, core_data)
930	size += mk->mod->mem[type].size;
931	}
932	return sprintf(buffer, "%u\n", size);
933	}
934
935	static struct module_attribute modinfo_coresize =
936	__ATTR(coresize, 0444, show_coresize, NULL);
937
938	#ifdef CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC
939	static ssize_t show_datasize(struct module_attribute *mattr,
940	struct module_kobject mk, char buffer)
941	{
942	unsigned int size = 0;
943
944	for_class_mod_mem_type(type, core_data)
945	size += mk->mod->mem[type].size;
946	return sprintf(buffer, "%u\n", size);
947	}
948
949	static struct module_attribute modinfo_datasize =
950	__ATTR(datasize, 0444, show_datasize, NULL);
951	#endif
952
953	static ssize_t show_initsize(struct module_attribute *mattr,
954	struct module_kobject mk, char buffer)
955	{
956	unsigned int size = 0;
957
958	for_class_mod_mem_type(type, init)
959	size += mk->mod->mem[type].size;
960	return sprintf(buffer, "%u\n", size);
961	}
962
963	static struct module_attribute modinfo_initsize =
964	__ATTR(initsize, 0444, show_initsize, NULL);
965
966	static ssize_t show_taint(struct module_attribute *mattr,
967	struct module_kobject mk, char buffer)
968	{
969	size_t l;
970
971	l = module_flags_taint(mk->mod->taints, buffer);
972	buffer[l++] = '\n';
973	return l;
974	}
975
976	static struct module_attribute modinfo_taint =
977	__ATTR(taint, 0444, show_taint, NULL);
978
979	struct module_attribute *modinfo_attrs[] = {
980	&module_uevent,
981	&modinfo_version,
982	&modinfo_srcversion,
983	&modinfo_initstate,
984	&modinfo_coresize,
985	#ifdef CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC
986	&modinfo_datasize,
987	#endif
988	&modinfo_initsize,
989	&modinfo_taint,
990	#ifdef CONFIG_MODULE_UNLOAD
991	&modinfo_refcnt,
992	#endif
993	NULL,
994	};
995
996	size_t modinfo_attrs_count = ARRAY_SIZE(modinfo_attrs);
997
998	static const char vermagic[] = VERMAGIC_STRING;
999
1000	int try_to_force_load(struct module mod, const char reason)
1001	{
1002	#ifdef CONFIG_MODULE_FORCE_LOAD
1003	if (!test_taint(TAINT_FORCED_MODULE))
1004	pr_warn("%s: %s: kernel tainted.\n", mod->name, reason);
1005	add_taint_module(mod, TAINT_FORCED_MODULE, LOCKDEP_NOW_UNRELIABLE);
1006	return 0;
1007	#else
1008	return -ENOEXEC;
1009	#endif
1010	}
1011
1012	/* Parse tag=value strings from .modinfo section */
1013	char module_next_tag_pair(char string, unsigned long *secsize)
1014	{
1015	/* Skip non-zero chars */
1016	while (string[0]) {
1017	string++;
1018	if ((*secsize)-- <= 1)
1019	return NULL;
1020	}
1021
1022	/* Skip any zero padding. */
1023	while (!string[0]) {
1024	string++;
1025	if ((*secsize)-- <= 1)
1026	return NULL;
1027	}
1028	return string;
1029	}
1030
1031	static char get_next_modinfo(const struct load_info info, const char *tag,
1032	char *prev)
1033	{
1034	char *p;
1035	unsigned int taglen = strlen(tag);
1036	Elf_Shdr *infosec = &info->sechdrs[info->index.info];
1037	unsigned long size = infosec->sh_size;
1038
1039	/*
1040	* get_modinfo() calls made before rewrite_section_headers()
1041	* must use sh_offset, as sh_addr isn't set!
1042	*/
1043	char modinfo = (char )info->hdr + infosec->sh_offset;
1044
1045	if (prev) {
1046	size -= prev - modinfo;
1047	modinfo = module_next_tag_pair(prev, &size);
1048	}
1049
1050	for (p = modinfo; p; p = module_next_tag_pair(p, &size)) {
1051	if (strncmp(p, tag, taglen) == 0 && p[taglen] == '=')
1052	return p + taglen + 1;
1053	}
1054	return NULL;
1055	}
1056
1057	static char get_modinfo(const struct load_info info, const char *tag)
1058	{
1059	return get_next_modinfo(info, tag, NULL);
1060	}
1061
1062	static int verify_namespace_is_imported(const struct load_info *info,
1063	const struct kernel_symbol *sym,
1064	struct module *mod)
1065	{
1066	const char *namespace;
1067	char *imported_namespace;
1068
1069	namespace = kernel_symbol_namespace(sym);
1070	if (namespace && namespace[0]) {
1071	for_each_modinfo_entry(imported_namespace, info, "import_ns") {
1072	if (strcmp(namespace, imported_namespace) == 0)
1073	return 0;
1074	}
1075	#ifdef CONFIG_MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS
1076	pr_warn(
1077	#else
1078	pr_err(
1079	#endif
1080	"%s: module uses symbol (%s) from namespace %s, but does not import it.\n",
1081	mod->name, kernel_symbol_name(sym), namespace);
1082	#ifndef CONFIG_MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS
1083	return -EINVAL;
1084	#endif
1085	}
1086	return 0;
1087	}
1088
1089	static bool inherit_taint(struct module mod, struct module owner, const char *name)
1090	{
1091	if (!owner \|\| !test_bit(TAINT_PROPRIETARY_MODULE, &owner->taints))
1092	return true;
1093
1094	if (mod->using_gplonly_symbols) {
1095	pr_err("%s: module using GPL-only symbols uses symbols %s from proprietary module %s.\n",
1096	mod->name, name, owner->name);
1097	return false;
1098	}
1099
1100	if (!test_bit(TAINT_PROPRIETARY_MODULE, &mod->taints)) {
1101	pr_warn("%s: module uses symbols %s from proprietary module %s, inheriting taint.\n",
1102	mod->name, name, owner->name);
1103	set_bit(TAINT_PROPRIETARY_MODULE, &mod->taints);
1104	}
1105	return true;
1106	}
1107
1108	/* Resolve a symbol for this module. I.e. if we find one, record usage. */
1109	static const struct kernel_symbol resolve_symbol(struct module mod,
1110	const struct load_info *info,
1111	const char *name,
1112	char ownername[])
1113	{
1114	struct find_symbol_arg fsa = {
1115	.name = name,
1116	.gplok = !(mod->taints & (1 << TAINT_PROPRIETARY_MODULE)),
1117	.warn = true,
1118	};
1119	int err;
1120
1121	/*
1122	* The module_mutex should not be a heavily contended lock;
1123	* if we get the occasional sleep here, we'll go an extra iteration
1124	* in the wait_event_interruptible(), which is harmless.
1125	*/
1126	sched_annotate_sleep();
1127	mutex_lock(&module_mutex);
1128	if (!find_symbol(&fsa))
1129	goto unlock;
1130
1131	if (fsa.license == GPL_ONLY)
1132	mod->using_gplonly_symbols = true;
1133
1134	if (!inherit_taint(mod, fsa.owner, name)) {
1135	fsa.sym = NULL;
1136	goto getname;
1137	}
1138
1139	if (!check_version(info, name, mod, fsa.crc)) {
1140	fsa.sym = ERR_PTR(-EINVAL);
1141	goto getname;
1142	}
1143
1144	err = verify_namespace_is_imported(info, fsa.sym, mod);
1145	if (err) {
1146	fsa.sym = ERR_PTR(err);
1147	goto getname;
1148	}
1149
1150	err = ref_module(mod, fsa.owner);
1151	if (err) {
1152	fsa.sym = ERR_PTR(err);
1153	goto getname;
1154	}
1155
1156	getname:
1157	/* We must make copy under the lock if we failed to get ref. */
1158	strncpy(ownername, module_name(fsa.owner), MODULE_NAME_LEN);
1159	unlock:
1160	mutex_unlock(&module_mutex);
1161	return fsa.sym;
1162	}
1163
1164	static const struct kernel_symbol *
1165	resolve_symbol_wait(struct module *mod,
1166	const struct load_info *info,
1167	const char *name)
1168	{
1169	const struct kernel_symbol *ksym;
1170	char owner[MODULE_NAME_LEN];
1171
1172	if (wait_event_interruptible_timeout(module_wq,
1173	!IS_ERR(ksym = resolve_symbol(mod, info, name, owner))
1174	\|\| PTR_ERR(ksym) != -EBUSY,
1175	30 * HZ) <= 0) {
1176	pr_warn("%s: gave up waiting for init of module %s.\n",
1177	mod->name, owner);
1178	}
1179	return ksym;
1180	}
1181
1182	void __weak module_memfree(void *module_region)
1183	{
1184	/*
1185	* This memory may be RO, and freeing RO memory in an interrupt is not
1186	* supported by vmalloc.
1187	*/
1188	WARN_ON(in_interrupt());
1189	vfree(module_region);
1190	}
1191
1192	void __weak module_arch_cleanup(struct module *mod)
1193	{
1194	}
1195
1196	void __weak module_arch_freeing_init(struct module *mod)
1197	{
1198	}
1199
1200	static bool mod_mem_use_vmalloc(enum mod_mem_type type)
1201	{
1202	return IS_ENABLED(CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC) &&
1203	mod_mem_type_is_core_data(type);
1204	}
1205
1206	static void *module_memory_alloc(unsigned int size, enum mod_mem_type type)
1207	{
1208	if (mod_mem_use_vmalloc(type))
1209	return vzalloc(size);
1210	return module_alloc(size);
1211	}
1212
1213	static void module_memory_free(void *ptr, enum mod_mem_type type)
1214	{
1215	if (mod_mem_use_vmalloc(type))
1216	vfree(ptr);
1217	else
1218	module_memfree(ptr);
1219	}
1220
1221	static void free_mod_mem(struct module *mod)
1222	{
1223	for_each_mod_mem_type(type) {
1224	struct module_memory *mod_mem = &mod->mem[type];
1225
1226	if (type == MOD_DATA)
1227	continue;
1228
1229	/* Free lock-classes; relies on the preceding sync_rcu(). */
1230	lockdep_free_key_range(mod_mem->base, mod_mem->size);
1231	if (mod_mem->size)
1232	module_memory_free(mod_mem->base, type);
1233	}
1234
1235	/* MOD_DATA hosts mod, so free it at last */
1236	lockdep_free_key_range(mod->mem[MOD_DATA].base, mod->mem[MOD_DATA].size);
1237	module_memory_free(mod->mem[MOD_DATA].base, MOD_DATA);
1238	}
1239
1240	/* Free a module, remove from lists, etc. */
1241	static void free_module(struct module *mod)
1242	{
1243	trace_module_free(mod);
1244
1245	mod_sysfs_teardown(mod);
1246
1247	/*
1248	* We leave it in list to prevent duplicate loads, but make sure
1249	* that noone uses it while it's being deconstructed.
1250	*/
1251	mutex_lock(&module_mutex);
1252	mod->state = MODULE_STATE_UNFORMED;
1253	mutex_unlock(&module_mutex);
1254
1255	/* Arch-specific cleanup. */
1256	module_arch_cleanup(mod);
1257
1258	/* Module unload stuff */
1259	module_unload_free(mod);
1260
1261	/* Free any allocated parameters. */
1262	destroy_params(mod->kp, mod->num_kp);
1263
1264	if (is_livepatch_module(mod))
1265	free_module_elf(mod);
1266
1267	/* Now we can delete it from the lists */
1268	mutex_lock(&module_mutex);
1269	/* Unlink carefully: kallsyms could be walking list. */
1270	list_del_rcu(&mod->list);
1271	mod_tree_remove(mod);
1272	/* Remove this module from bug list, this uses list_del_rcu */
1273	module_bug_cleanup(mod);
1274	/* Wait for RCU-sched synchronizing before releasing mod->list and buglist. */
1275	synchronize_rcu();
1276	if (try_add_tainted_module(mod))
1277	pr_err("%s: adding tainted module to the unloaded tainted modules list failed.\n",
1278	mod->name);
1279	mutex_unlock(&module_mutex);
1280
1281	/* This may be empty, but that's OK */
1282	module_arch_freeing_init(mod);
1283	kfree(mod->args);
1284	percpu_modfree(mod);
1285
1286	free_mod_mem(mod);
1287	}
1288
1289	void __symbol_get(const char symbol)
1290	{
1291	struct find_symbol_arg fsa = {
1292	.name = symbol,
1293	.gplok = true,
1294	.warn = true,
1295	};
1296
1297	preempt_disable();
1298	if (!find_symbol(&fsa) \|\| strong_try_module_get(fsa.owner)) {
1299	preempt_enable();
1300	return NULL;
1301	}
1302	preempt_enable();
1303	return (void *)kernel_symbol_value(fsa.sym);
1304	}
1305	EXPORT_SYMBOL_GPL(__symbol_get);
1306
1307	/*
1308	* Ensure that an exported symbol [global namespace] does not already exist
1309	* in the kernel or in some other module's exported symbol table.
1310	*
1311	* You must hold the module_mutex.
1312	*/
1313	static int verify_exported_symbols(struct module *mod)
1314	{
1315	unsigned int i;
1316	const struct kernel_symbol *s;
1317	struct {
1318	const struct kernel_symbol *sym;
1319	unsigned int num;
1320	} arr[] = {
1321	{ mod->syms, mod->num_syms },
1322	{ mod->gpl_syms, mod->num_gpl_syms },
1323	};
1324
1325	for (i = 0; i < ARRAY_SIZE(arr); i++) {
1326	for (s = arr[i].sym; s < arr[i].sym + arr[i].num; s++) {
1327	struct find_symbol_arg fsa = {
1328	.name = kernel_symbol_name(s),
1329	.gplok = true,
1330	};
1331	if (find_symbol(&fsa)) {
1332	pr_err("%s: exports duplicate symbol %s"
1333	" (owned by %s)\n",
1334	mod->name, kernel_symbol_name(s),
1335	module_name(fsa.owner));
1336	return -ENOEXEC;
1337	}
1338	}
1339	}
1340	return 0;
1341	}
1342
1343	static bool ignore_undef_symbol(Elf_Half emachine, const char *name)
1344	{
1345	/*
1346	* On x86, PIC code and Clang non-PIC code may have call foo@PLT. GNU as
1347	* before 2.37 produces an unreferenced _GLOBAL_OFFSET_TABLE_ on x86-64.
1348	* i386 has a similar problem but may not deserve a fix.
1349	*
1350	* If we ever have to ignore many symbols, consider refactoring the code to
1351	* only warn if referenced by a relocation.
1352	*/
1353	if (emachine == EM_386 \|\| emachine == EM_X86_64)
1354	return !strcmp(name, "_GLOBAL_OFFSET_TABLE_");
1355	return false;
1356	}
1357
1358	/* Change all symbols so that st_value encodes the pointer directly. */
1359	static int simplify_symbols(struct module mod, const struct load_info info)
1360	{
1361	Elf_Shdr *symsec = &info->sechdrs[info->index.sym];
1362	Elf_Sym sym = (void )symsec->sh_addr;
1363	unsigned long secbase;
1364	unsigned int i;
1365	int ret = 0;
1366	const struct kernel_symbol *ksym;
1367
1368	for (i = 1; i < symsec->sh_size / sizeof(Elf_Sym); i++) {
1369	const char *name = info->strtab + sym[i].st_name;
1370
1371	switch (sym[i].st_shndx) {
1372	case SHN_COMMON:
1373	/* Ignore common symbols */
1374	if (!strncmp(name, "__gnu_lto", 9))
1375	break;
1376
1377	/*
1378	* We compiled with -fno-common. These are not
1379	* supposed to happen.
1380	*/
1381	pr_debug("Common symbol: %s\n", name);
1382	pr_warn("%s: please compile with -fno-common\n",
1383	mod->name);
1384	ret = -ENOEXEC;
1385	break;
1386
1387	case SHN_ABS:
1388	/* Don't need to do anything */
1389	pr_debug("Absolute symbol: 0x%08lx %s\n",
1390	(long)sym[i].st_value, name);
1391	break;
1392
1393	case SHN_LIVEPATCH:
1394	/* Livepatch symbols are resolved by livepatch */
1395	break;
1396
1397	case SHN_UNDEF:
1398	ksym = resolve_symbol_wait(mod, info, name);
1399	/* Ok if resolved. */
1400	if (ksym && !IS_ERR(ksym)) {
1401	sym[i].st_value = kernel_symbol_value(ksym);
1402	break;
1403	}
1404
1405	/* Ok if weak or ignored. */
1406	if (!ksym &&
1407	(ELF_ST_BIND(sym[i].st_info) == STB_WEAK \|\|
1408	ignore_undef_symbol(info->hdr->e_machine, name)))
1409	break;
1410
1411	ret = PTR_ERR(ksym) ?: -ENOENT;
1412	pr_warn("%s: Unknown symbol %s (err %d)\n",
1413	mod->name, name, ret);
1414	break;
1415
1416	default:
1417	/* Divert to percpu allocation if a percpu var. */
1418	if (sym[i].st_shndx == info->index.pcpu)
1419	secbase = (unsigned long)mod_percpu(mod);
1420	else
1421	secbase = info->sechdrs[sym[i].st_shndx].sh_addr;
1422	sym[i].st_value += secbase;
1423	break;
1424	}
1425	}
1426
1427	return ret;
1428	}
1429
1430	static int apply_relocations(struct module mod, const struct load_info info)
1431	{
1432	unsigned int i;
1433	int err = 0;
1434
1435	/* Now do relocations. */
1436	for (i = 1; i < info->hdr->e_shnum; i++) {
1437	unsigned int infosec = info->sechdrs[i].sh_info;
1438
1439	/* Not a valid relocation section? */
1440	if (infosec >= info->hdr->e_shnum)
1441	continue;
1442
1443	/* Don't bother with non-allocated sections */
1444	if (!(info->sechdrs[infosec].sh_flags & SHF_ALLOC))
1445	continue;
1446
1447	if (info->sechdrs[i].sh_flags & SHF_RELA_LIVEPATCH)
1448	err = klp_apply_section_relocs(mod, info->sechdrs,
1449	info->secstrings,
1450	info->strtab,
1451	info->index.sym, i,
1452	NULL);
1453	else if (info->sechdrs[i].sh_type == SHT_REL)
1454	err = apply_relocate(info->sechdrs, info->strtab,
1455	info->index.sym, i, mod);
1456	else if (info->sechdrs[i].sh_type == SHT_RELA)
1457	err = apply_relocate_add(info->sechdrs, info->strtab,
1458	info->index.sym, i, mod);
1459	if (err < 0)
1460	break;
1461	}
1462	return err;
1463	}
1464
1465	/* Additional bytes needed by arch in front of individual sections */
1466	unsigned int __weak arch_mod_section_prepend(struct module *mod,
1467	unsigned int section)
1468	{
1469	/* default implementation just returns zero */
1470	return 0;
1471	}
1472
1473	long module_get_offset_and_type(struct module *mod, enum mod_mem_type type,
1474	Elf_Shdr *sechdr, unsigned int section)
1475	{
1476	long offset;
1477	long mask = ((unsigned long)(type) & SH_ENTSIZE_TYPE_MASK) << SH_ENTSIZE_TYPE_SHIFT;
1478
1479	mod->mem[type].size += arch_mod_section_prepend(mod, section);
1480	offset = ALIGN(mod->mem[type].size, sechdr->sh_addralign ?: 1);
1481	mod->mem[type].size = offset + sechdr->sh_size;
1482
1483	WARN_ON_ONCE(offset & mask);
1484	return offset \| mask;
1485	}
1486
1487	static bool module_init_layout_section(const char *sname)
1488	{
1489	#ifndef CONFIG_MODULE_UNLOAD
1490	if (module_exit_section(sname))
1491	return true;
1492	#endif
1493	return module_init_section(sname);
1494	}
1495
1496	static void __layout_sections(struct module mod, struct load_info info, bool is_init)
1497	{
1498	unsigned int m, i;
1499
1500	static const unsigned long masks[][2] = {
1501	/*
1502	* NOTE: all executable code must be the first section
1503	* in this array; otherwise modify the text_size
1504	* finder in the two loops below
1505	*/
1506	{ SHF_EXECINSTR \| SHF_ALLOC, ARCH_SHF_SMALL },
1507	{ SHF_ALLOC, SHF_WRITE \| ARCH_SHF_SMALL },
1508	{ SHF_RO_AFTER_INIT \| SHF_ALLOC, ARCH_SHF_SMALL },
1509	{ SHF_WRITE \| SHF_ALLOC, ARCH_SHF_SMALL },
1510	{ ARCH_SHF_SMALL \| SHF_ALLOC, 0 }
1511	};
1512	static const int core_m_to_mem_type[] = {
1513	MOD_TEXT,
1514	MOD_RODATA,
1515	MOD_RO_AFTER_INIT,
1516	MOD_DATA,
1517	MOD_DATA,
1518	};
1519	static const int init_m_to_mem_type[] = {
1520	MOD_INIT_TEXT,
1521	MOD_INIT_RODATA,
1522	MOD_INVALID,
1523	MOD_INIT_DATA,
1524	MOD_INIT_DATA,
1525	};
1526
1527	for (m = 0; m < ARRAY_SIZE(masks); ++m) {
1528	enum mod_mem_type type = is_init ? init_m_to_mem_type[m] : core_m_to_mem_type[m];
1529
1530	for (i = 0; i < info->hdr->e_shnum; ++i) {
1531	Elf_Shdr *s = &info->sechdrs[i];
1532	const char *sname = info->secstrings + s->sh_name;
1533
1534	if ((s->sh_flags & masks[m][0]) != masks[m][0]
1535	\|\| (s->sh_flags & masks[m][1])
1536	\|\| s->sh_entsize != ~0UL
1537	\|\| is_init != module_init_layout_section(sname))
1538	continue;
1539
1540	if (WARN_ON_ONCE(type == MOD_INVALID))
1541	continue;
1542
1543	s->sh_entsize = module_get_offset_and_type(mod, type, s, i);
1544	pr_debug("\t%s\n", sname);
1545	}
1546	}
1547	}
1548
1549	/*
1550	* Lay out the SHF_ALLOC sections in a way not dissimilar to how ld
1551	* might -- code, read-only data, read-write data, small data. Tally
1552	* sizes, and place the offsets into sh_entsize fields: high bit means it
1553	* belongs in init.
1554	*/
1555	static void layout_sections(struct module mod, struct load_info info)
1556	{
1557	unsigned int i;
1558
1559	for (i = 0; i < info->hdr->e_shnum; i++)
1560	info->sechdrs[i].sh_entsize = ~0UL;
1561
1562	pr_debug("Core section allocation order for %s:\n", mod->name);
1563	__layout_sections(mod, info, false);
1564
1565	pr_debug("Init section allocation order for %s:\n", mod->name);
1566	__layout_sections(mod, info, true);
1567	}
1568
1569	static void module_license_taint_check(struct module mod, const char license)
1570	{
1571	if (!license)
1572	license = "unspecified";
1573
1574	if (!license_is_gpl_compatible(license)) {
1575	if (!test_taint(TAINT_PROPRIETARY_MODULE))
1576	pr_warn("%s: module license '%s' taints kernel.\n",
1577	mod->name, license);
1578	add_taint_module(mod, TAINT_PROPRIETARY_MODULE,
1579	LOCKDEP_NOW_UNRELIABLE);
1580	}
1581	}
1582
1583	static void setup_modinfo(struct module mod, struct load_info info)
1584	{
1585	struct module_attribute *attr;
1586	int i;
1587
1588	for (i = 0; (attr = modinfo_attrs[i]); i++) {
1589	if (attr->setup)
1590	attr->setup(mod, get_modinfo(info, attr->attr.name));
1591	}
1592	}
1593
1594	static void free_modinfo(struct module *mod)
1595	{
1596	struct module_attribute *attr;
1597	int i;
1598
1599	for (i = 0; (attr = modinfo_attrs[i]); i++) {
1600	if (attr->free)
1601	attr->free(mod);
1602	}
1603	}
1604
1605	void * __weak module_alloc(unsigned long size)
1606	{
1607	return __vmalloc_node_range(size, 1, VMALLOC_START, VMALLOC_END,
1608	GFP_KERNEL, PAGE_KERNEL_EXEC, VM_FLUSH_RESET_PERMS,
1609	NUMA_NO_NODE, __builtin_return_address(0));
1610	}
1611
1612	bool __weak module_init_section(const char *name)
1613	{
1614	return strstarts(name, ".init");
1615	}
1616
1617	bool __weak module_exit_section(const char *name)
1618	{
1619	return strstarts(name, ".exit");
1620	}
1621
1622	static int validate_section_offset(struct load_info info, Elf_Shdr shdr)
1623	{
1624	#if defined(CONFIG_64BIT)
1625	unsigned long long secend;
1626	#else
1627	unsigned long secend;
1628	#endif
1629
1630	/*
1631	* Check for both overflow and offset/size being
1632	* too large.
1633	*/
1634	secend = shdr->sh_offset + shdr->sh_size;
1635	if (secend < shdr->sh_offset \|\| secend > info->len)
1636	return -ENOEXEC;
1637
1638	return 0;
1639	}
1640
1641	/*
1642	* Check userspace passed ELF module against our expectations, and cache
1643	* useful variables for further processing as we go.
1644	*
1645	* This does basic validity checks against section offsets and sizes, the
1646	* section name string table, and the indices used for it (sh_name).
1647	*
1648	* As a last step, since we're already checking the ELF sections we cache
1649	* useful variables which will be used later for our convenience:
1650	*
1651	* o pointers to section headers
1652	* o cache the modinfo symbol section
1653	* o cache the string symbol section
1654	* o cache the module section
1655	*
1656	* As a last step we set info->mod to the temporary copy of the module in
1657	* info->hdr. The final one will be allocated in move_module(). Any
1658	* modifications we make to our copy of the module will be carried over
1659	* to the final minted module.
1660	*/
1661	static int elf_validity_cache_copy(struct load_info *info, int flags)
1662	{
1663	unsigned int i;
1664	Elf_Shdr shdr, strhdr;
1665	int err;
1666	unsigned int num_mod_secs = 0, mod_idx;
1667	unsigned int num_info_secs = 0, info_idx;
1668	unsigned int num_sym_secs = 0, sym_idx;
1669
1670	if (info->len < sizeof(*(info->hdr))) {
1671	pr_err("Invalid ELF header len %lu\n", info->len);
1672	goto no_exec;
1673	}
1674
1675	if (memcmp(info->hdr->e_ident, ELFMAG, SELFMAG) != 0) {
1676	pr_err("Invalid ELF header magic: != %s\n", ELFMAG);
1677	goto no_exec;
1678	}
1679	if (info->hdr->e_type != ET_REL) {
1680	pr_err("Invalid ELF header type: %u != %u\n",
1681	info->hdr->e_type, ET_REL);
1682	goto no_exec;
1683	}
1684	if (!elf_check_arch(info->hdr)) {
1685	pr_err("Invalid architecture in ELF header: %u\n",
1686	info->hdr->e_machine);
1687	goto no_exec;
1688	}
1689	if (!module_elf_check_arch(info->hdr)) {
1690	pr_err("Invalid module architecture in ELF header: %u\n",
1691	info->hdr->e_machine);
1692	goto no_exec;
1693	}
1694	if (info->hdr->e_shentsize != sizeof(Elf_Shdr)) {
1695	pr_err("Invalid ELF section header size\n");
1696	goto no_exec;
1697	}
1698
1699	/*
1700	* e_shnum is 16 bits, and sizeof(Elf_Shdr) is
1701	* known and small. So e_shnum * sizeof(Elf_Shdr)
1702	* will not overflow unsigned long on any platform.
1703	*/
1704	if (info->hdr->e_shoff >= info->len
1705	\|\| (info->hdr->e_shnum * sizeof(Elf_Shdr) >
1706	info->len - info->hdr->e_shoff)) {
1707	pr_err("Invalid ELF section header overflow\n");
1708	goto no_exec;
1709	}
1710
1711	info->sechdrs = (void *)info->hdr + info->hdr->e_shoff;
1712
1713	/*
1714	* Verify if the section name table index is valid.
1715	*/
1716	if (info->hdr->e_shstrndx == SHN_UNDEF
1717	\|\| info->hdr->e_shstrndx >= info->hdr->e_shnum) {
1718	pr_err("Invalid ELF section name index: %d \|\| e_shstrndx (%d) >= e_shnum (%d)\n",
1719	info->hdr->e_shstrndx, info->hdr->e_shstrndx,
1720	info->hdr->e_shnum);
1721	goto no_exec;
1722	}
1723
1724	strhdr = &info->sechdrs[info->hdr->e_shstrndx];
1725	err = validate_section_offset(info, strhdr);
1726	if (err < 0) {
1727	pr_err("Invalid ELF section hdr(type %u)\n", strhdr->sh_type);
1728	return err;
1729	}
1730
1731	/*
1732	* The section name table must be NUL-terminated, as required
1733	* by the spec. This makes strcmp and pr_* calls that access
1734	* strings in the section safe.
1735	*/
1736	info->secstrings = (void *)info->hdr + strhdr->sh_offset;
1737	if (strhdr->sh_size == 0) {
1738	pr_err("empty section name table\n");
1739	goto no_exec;
1740	}
1741	if (info->secstrings[strhdr->sh_size - 1] != '\0') {
1742	pr_err("ELF Spec violation: section name table isn't null terminated\n");
1743	goto no_exec;
1744	}
1745
1746	/*
1747	* The code assumes that section 0 has a length of zero and
1748	* an addr of zero, so check for it.
1749	*/
1750	if (info->sechdrs[0].sh_type != SHT_NULL
1751	\|\| info->sechdrs[0].sh_size != 0
1752	\|\| info->sechdrs[0].sh_addr != 0) {
1753	pr_err("ELF Spec violation: section 0 type(%d)!=SH_NULL or non-zero len or addr\n",
1754	info->sechdrs[0].sh_type);
1755	goto no_exec;
1756	}
1757
1758	for (i = 1; i < info->hdr->e_shnum; i++) {
1759	shdr = &info->sechdrs[i];
1760	switch (shdr->sh_type) {
1761	case SHT_NULL:
1762	case SHT_NOBITS:
1763	continue;
1764	case SHT_SYMTAB:
1765	if (shdr->sh_link == SHN_UNDEF
1766	\|\| shdr->sh_link >= info->hdr->e_shnum) {
1767	pr_err("Invalid ELF sh_link!=SHN_UNDEF(%d) or (sh_link(%d) >= hdr->e_shnum(%d)\n",
1768	shdr->sh_link, shdr->sh_link,
1769	info->hdr->e_shnum);
1770	goto no_exec;
1771	}
1772	num_sym_secs++;
1773	sym_idx = i;
1774	fallthrough;
1775	default:
1776	err = validate_section_offset(info, shdr);
1777	if (err < 0) {
1778	pr_err("Invalid ELF section in module (section %u type %u)\n",
1779	i, shdr->sh_type);
1780	return err;
1781	}
1782	if (strcmp(info->secstrings + shdr->sh_name,
1783	".gnu.linkonce.this_module") == 0) {
1784	num_mod_secs++;
1785	mod_idx = i;
1786	} else if (strcmp(info->secstrings + shdr->sh_name,
1787	".modinfo") == 0) {
1788	num_info_secs++;
1789	info_idx = i;
1790	}
1791
1792	if (shdr->sh_flags & SHF_ALLOC) {
1793	if (shdr->sh_name >= strhdr->sh_size) {
1794	pr_err("Invalid ELF section name in module (section %u type %u)\n",
1795	i, shdr->sh_type);
1796	return -ENOEXEC;
1797	}
1798	}
1799	break;
1800	}
1801	}
1802
1803	if (num_info_secs > 1) {
1804	pr_err("Only one .modinfo section must exist.\n");
1805	goto no_exec;
1806	} else if (num_info_secs == 1) {
1807	/* Try to find a name early so we can log errors with a module name */
1808	info->index.info = info_idx;
1809	info->name = get_modinfo(info, "name");
1810	}
1811
1812	if (num_sym_secs != 1) {
1813	pr_warn("%s: module has no symbols (stripped?)\n",
1814	info->name ?: "(missing .modinfo section or name field)");
1815	goto no_exec;
1816	}
1817
1818	/* Sets internal symbols and strings. */
1819	info->index.sym = sym_idx;
1820	shdr = &info->sechdrs[sym_idx];
1821	info->index.str = shdr->sh_link;
1822	info->strtab = (char *)info->hdr + info->sechdrs[info->index.str].sh_offset;
1823
1824	/*
1825	* The ".gnu.linkonce.this_module" ELF section is special. It is
1826	* what modpost uses to refer to __this_module and let's use rely
1827	* on THIS_MODULE to point to &__this_module properly. The kernel's
1828	* modpost declares it on each modules's *.mod.c file. If the struct
1829	* module of the kernel changes a full kernel rebuild is required.
1830	*
1831	* We have a few expectaions for this special section, the following
1832	* code validates all this for us:
1833	*
1834	* o Only one section must exist
1835	* o We expect the kernel to always have to allocate it: SHF_ALLOC
1836	* o The section size must match the kernel's run time's struct module
1837	* size
1838	*/
1839	if (num_mod_secs != 1) {
1840	pr_err("module %s: Only one .gnu.linkonce.this_module section must exist.\n",
1841	info->name ?: "(missing .modinfo section or name field)");
1842	goto no_exec;
1843	}
1844
1845	shdr = &info->sechdrs[mod_idx];
1846
1847	/*
1848	* This is already implied on the switch above, however let's be
1849	* pedantic about it.
1850	*/
1851	if (shdr->sh_type == SHT_NOBITS) {
1852	pr_err("module %s: .gnu.linkonce.this_module section must have a size set\n",
1853	info->name ?: "(missing .modinfo section or name field)");
1854	goto no_exec;
1855	}
1856
1857	if (!(shdr->sh_flags & SHF_ALLOC)) {
1858	pr_err("module %s: .gnu.linkonce.this_module must occupy memory during process execution\n",
1859	info->name ?: "(missing .modinfo section or name field)");
1860	goto no_exec;
1861	}
1862
1863	if (shdr->sh_size != sizeof(struct module)) {
1864	pr_err("module %s: .gnu.linkonce.this_module section size must match the kernel's built struct module size at run time\n",
1865	info->name ?: "(missing .modinfo section or name field)");
1866	goto no_exec;
1867	}
1868
1869	info->index.mod = mod_idx;
1870
1871	/* This is temporary: point mod into copy of data. */
1872	info->mod = (void *)info->hdr + shdr->sh_offset;
1873
1874	/*
1875	* If we didn't load the .modinfo 'name' field earlier, fall back to
1876	* on-disk struct mod 'name' field.
1877	*/
1878	if (!info->name)
1879	info->name = info->mod->name;
1880
1881	if (flags & MODULE_INIT_IGNORE_MODVERSIONS)
1882	info->index.vers = 0; /* Pretend no __versions section! */
1883	else
1884	info->index.vers = find_sec(info, "__versions");
1885
1886	info->index.pcpu = find_pcpusec(info);
1887
1888	return 0;
1889
1890	no_exec:
1891	return -ENOEXEC;
1892	}
1893
1894	#define COPY_CHUNK_SIZE (16*PAGE_SIZE)
1895
1896	static int copy_chunked_from_user(void dst, const void __user usrc, unsigned long len)
1897	{
1898	do {
1899	unsigned long n = min(len, COPY_CHUNK_SIZE);
1900
1901	if (copy_from_user(dst, usrc, n) != 0)
1902	return -EFAULT;
1903	cond_resched();
1904	dst += n;
1905	usrc += n;
1906	len -= n;
1907	} while (len);
1908	return 0;
1909	}
1910
1911	static int check_modinfo_livepatch(struct module mod, struct load_info info)
1912	{
1913	if (!get_modinfo(info, "livepatch"))
1914	/* Nothing more to do */
1915	return 0;
1916
1917	if (set_livepatch_module(mod))
1918	return 0;
1919
1920	pr_err("%s: module is marked as livepatch module, but livepatch support is disabled",
1921	mod->name);
1922	return -ENOEXEC;
1923	}
1924
1925	static void check_modinfo_retpoline(struct module mod, struct load_info info)
1926	{
1927	if (retpoline_module_ok(get_modinfo(info, "retpoline")))
1928	return;
1929
1930	pr_warn("%s: loading module not compiled with retpoline compiler.\n",
1931	mod->name);
1932	}
1933
1934	/* Sets info->hdr and info->len. */
1935	static int copy_module_from_user(const void __user *umod, unsigned long len,
1936	struct load_info *info)
1937	{
1938	int err;
1939
1940	info->len = len;
1941	if (info->len < sizeof(*(info->hdr)))
1942	return -ENOEXEC;
1943
1944	err = security_kernel_load_data(LOADING_MODULE, true);
1945	if (err)
1946	return err;
1947
1948	/* Suck in entire file: we'll want most of it. */
1949	info->hdr = __vmalloc(info->len, GFP_KERNEL \| __GFP_NOWARN);
1950	if (!info->hdr)
1951	return -ENOMEM;
1952
1953	if (copy_chunked_from_user(info->hdr, umod, info->len) != 0) {
1954	err = -EFAULT;
1955	goto out;
1956	}
1957
1958	err = security_kernel_post_load_data((char *)info->hdr, info->len,
1959	LOADING_MODULE, "init_module");
1960	out:
1961	if (err)
1962	vfree(info->hdr);
1963
1964	return err;
1965	}
1966
1967	static void free_copy(struct load_info *info, int flags)
1968	{
1969	if (flags & MODULE_INIT_COMPRESSED_FILE)
1970	module_decompress_cleanup(info);
1971	else
1972	vfree(info->hdr);
1973	}
1974
1975	static int rewrite_section_headers(struct load_info *info, int flags)
1976	{
1977	unsigned int i;
1978
1979	/* This should always be true, but let's be sure. */
1980	info->sechdrs[0].sh_addr = 0;
1981
1982	for (i = 1; i < info->hdr->e_shnum; i++) {
1983	Elf_Shdr *shdr = &info->sechdrs[i];
1984
1985	/*
1986	* Mark all sections sh_addr with their address in the
1987	* temporary image.
1988	*/
1989	shdr->sh_addr = (size_t)info->hdr + shdr->sh_offset;
1990
1991	}
1992
1993	/* Track but don't keep modinfo and version sections. */
1994	info->sechdrs[info->index.vers].sh_flags &= ~(unsigned long)SHF_ALLOC;
1995	info->sechdrs[info->index.info].sh_flags &= ~(unsigned long)SHF_ALLOC;
1996
1997	return 0;
1998	}
1999
2000	/*
2001	* These calls taint the kernel depending certain module circumstances */
2002	static void module_augment_kernel_taints(struct module mod, struct load_info info)
2003	{
2004	int prev_taint = test_taint(TAINT_PROPRIETARY_MODULE);
2005
2006	if (!get_modinfo(info, "intree")) {
2007	if (!test_taint(TAINT_OOT_MODULE))
2008	pr_warn("%s: loading out-of-tree module taints kernel.\n",
2009	mod->name);
2010	add_taint_module(mod, TAINT_OOT_MODULE, LOCKDEP_STILL_OK);
2011	}
2012
2013	check_modinfo_retpoline(mod, info);
2014
2015	if (get_modinfo(info, "staging")) {
2016	add_taint_module(mod, TAINT_CRAP, LOCKDEP_STILL_OK);
2017	pr_warn("%s: module is from the staging directory, the quality "
2018	"is unknown, you have been warned.\n", mod->name);
2019	}
2020
2021	if (is_livepatch_module(mod)) {
2022	add_taint_module(mod, TAINT_LIVEPATCH, LOCKDEP_STILL_OK);
2023	pr_notice_once("%s: tainting kernel with TAINT_LIVEPATCH\n",
2024	mod->name);
2025	}
2026
2027	module_license_taint_check(mod, get_modinfo(info, "license"));
2028
2029	if (get_modinfo(info, "test")) {
2030	if (!test_taint(TAINT_TEST))
2031	pr_warn("%s: loading test module taints kernel.\n",
2032	mod->name);
2033	add_taint_module(mod, TAINT_TEST, LOCKDEP_STILL_OK);
2034	}
2035	#ifdef CONFIG_MODULE_SIG
2036	mod->sig_ok = info->sig_ok;
2037	if (!mod->sig_ok) {
2038	pr_notice_once("%s: module verification failed: signature "
2039	"and/or required key missing - tainting "
2040	"kernel\n", mod->name);
2041	add_taint_module(mod, TAINT_UNSIGNED_MODULE, LOCKDEP_STILL_OK);
2042	}
2043	#endif
2044
2045	/*
2046	* ndiswrapper is under GPL by itself, but loads proprietary modules.
2047	* Don't use add_taint_module(), as it would prevent ndiswrapper from
2048	* using GPL-only symbols it needs.
2049	*/
2050	if (strcmp(mod->name, "ndiswrapper") == 0)
2051	add_taint(TAINT_PROPRIETARY_MODULE, LOCKDEP_NOW_UNRELIABLE);
2052
2053	/* driverloader was caught wrongly pretending to be under GPL */
2054	if (strcmp(mod->name, "driverloader") == 0)
2055	add_taint_module(mod, TAINT_PROPRIETARY_MODULE,
2056	LOCKDEP_NOW_UNRELIABLE);
2057
2058	/* lve claims to be GPL but upstream won't provide source */
2059	if (strcmp(mod->name, "lve") == 0)
2060	add_taint_module(mod, TAINT_PROPRIETARY_MODULE,
2061	LOCKDEP_NOW_UNRELIABLE);
2062
2063	if (!prev_taint && test_taint(TAINT_PROPRIETARY_MODULE))
2064	pr_warn("%s: module license taints kernel.\n", mod->name);
2065
2066	}
2067
2068	static int check_modinfo(struct module mod, struct load_info info, int flags)
2069	{
2070	const char *modmagic = get_modinfo(info, "vermagic");
2071	int err;
2072
2073	if (flags & MODULE_INIT_IGNORE_VERMAGIC)
2074	modmagic = NULL;
2075
2076	/* This is allowed: modprobe --force will invalidate it. */
2077	if (!modmagic) {
2078	err = try_to_force_load(mod, "bad vermagic");
2079	if (err)
2080	return err;
2081	} else if (!same_magic(modmagic, vermagic, info->index.vers)) {
2082	pr_err("%s: version magic '%s' should be '%s'\n",
2083	info->name, modmagic, vermagic);
2084	return -ENOEXEC;
2085	}
2086
2087	err = check_modinfo_livepatch(mod, info);
2088	if (err)
2089	return err;
2090
2091	return 0;
2092	}
2093
2094	static int find_module_sections(struct module mod, struct load_info info)
2095	{
2096	mod->kp = section_objs(info, "__param",
2097	sizeof(*mod->kp), &mod->num_kp);
2098	mod->syms = section_objs(info, "__ksymtab",
2099	sizeof(*mod->syms), &mod->num_syms);
2100	mod->crcs = section_addr(info, "__kcrctab");
2101	mod->gpl_syms = section_objs(info, "__ksymtab_gpl",
2102	sizeof(*mod->gpl_syms),
2103	&mod->num_gpl_syms);
2104	mod->gpl_crcs = section_addr(info, "__kcrctab_gpl");
2105
2106	#ifdef CONFIG_CONSTRUCTORS
2107	mod->ctors = section_objs(info, ".ctors",
2108	sizeof(*mod->ctors), &mod->num_ctors);
2109	if (!mod->ctors)
2110	mod->ctors = section_objs(info, ".init_array",
2111	sizeof(*mod->ctors), &mod->num_ctors);
2112	else if (find_sec(info, ".init_array")) {
2113	/*
2114	* This shouldn't happen with same compiler and binutils
2115	* building all parts of the module.
2116	*/
2117	pr_warn("%s: has both .ctors and .init_array.\n",
2118	mod->name);
2119	return -EINVAL;
2120	}
2121	#endif
2122
2123	mod->noinstr_text_start = section_objs(info, ".noinstr.text", 1,
2124	&mod->noinstr_text_size);
2125
2126	#ifdef CONFIG_TRACEPOINTS
2127	mod->tracepoints_ptrs = section_objs(info, "__tracepoints_ptrs",
2128	sizeof(*mod->tracepoints_ptrs),
2129	&mod->num_tracepoints);
2130	#endif
2131	#ifdef CONFIG_TREE_SRCU
2132	mod->srcu_struct_ptrs = section_objs(info, "___srcu_struct_ptrs",
2133	sizeof(*mod->srcu_struct_ptrs),
2134	&mod->num_srcu_structs);
2135	#endif
2136	#ifdef CONFIG_BPF_EVENTS
2137	mod->bpf_raw_events = section_objs(info, "__bpf_raw_tp_map",
2138	sizeof(*mod->bpf_raw_events),
2139	&mod->num_bpf_raw_events);
2140	#endif
2141	#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
2142	mod->btf_data = any_section_objs(info, ".BTF", 1, &mod->btf_data_size);
2143	#endif
2144	#ifdef CONFIG_JUMP_LABEL
2145	mod->jump_entries = section_objs(info, "__jump_table",
2146	sizeof(*mod->jump_entries),
2147	&mod->num_jump_entries);
2148	#endif
2149	#ifdef CONFIG_EVENT_TRACING
2150	mod->trace_events = section_objs(info, "_ftrace_events",
2151	sizeof(*mod->trace_events),
2152	&mod->num_trace_events);
2153	mod->trace_evals = section_objs(info, "_ftrace_eval_map",
2154	sizeof(*mod->trace_evals),
2155	&mod->num_trace_evals);
2156	#endif
2157	#ifdef CONFIG_TRACING
2158	mod->trace_bprintk_fmt_start = section_objs(info, "__trace_printk_fmt",
2159	sizeof(*mod->trace_bprintk_fmt_start),
2160	&mod->num_trace_bprintk_fmt);
2161	#endif
2162	#ifdef CONFIG_FTRACE_MCOUNT_RECORD
2163	/* sechdrs[0].sh_size is always zero */
2164	mod->ftrace_callsites = section_objs(info, FTRACE_CALLSITE_SECTION,
2165	sizeof(*mod->ftrace_callsites),
2166	&mod->num_ftrace_callsites);
2167	#endif
2168	#ifdef CONFIG_FUNCTION_ERROR_INJECTION
2169	mod->ei_funcs = section_objs(info, "_error_injection_whitelist",
2170	sizeof(*mod->ei_funcs),
2171	&mod->num_ei_funcs);
2172	#endif
2173	#ifdef CONFIG_KPROBES
2174	mod->kprobes_text_start = section_objs(info, ".kprobes.text", 1,
2175	&mod->kprobes_text_size);
2176	mod->kprobe_blacklist = section_objs(info, "_kprobe_blacklist",
2177	sizeof(unsigned long),
2178	&mod->num_kprobe_blacklist);
2179	#endif
2180	#ifdef CONFIG_PRINTK_INDEX
2181	mod->printk_index_start = section_objs(info, ".printk_index",
2182	sizeof(*mod->printk_index_start),
2183	&mod->printk_index_size);
2184	#endif
2185	#ifdef CONFIG_HAVE_STATIC_CALL_INLINE
2186	mod->static_call_sites = section_objs(info, ".static_call_sites",
2187	sizeof(*mod->static_call_sites),
2188	&mod->num_static_call_sites);
2189	#endif
2190	#if IS_ENABLED(CONFIG_KUNIT)
2191	mod->kunit_suites = section_objs(info, ".kunit_test_suites",
2192	sizeof(*mod->kunit_suites),
2193	&mod->num_kunit_suites);
2194	#endif
2195
2196	mod->extable = section_objs(info, "__ex_table",
2197	sizeof(*mod->extable), &mod->num_exentries);
2198
2199	if (section_addr(info, "__obsparm"))
2200	pr_warn("%s: Ignoring obsolete parameters\n", mod->name);
2201
2202	#ifdef CONFIG_DYNAMIC_DEBUG_CORE
2203	mod->dyndbg_info.descs = section_objs(info, "__dyndbg",
2204	sizeof(*mod->dyndbg_info.descs),
2205	&mod->dyndbg_info.num_descs);
2206	mod->dyndbg_info.classes = section_objs(info, "__dyndbg_classes",
2207	sizeof(*mod->dyndbg_info.classes),
2208	&mod->dyndbg_info.num_classes);
2209	#endif
2210
2211	return 0;
2212	}
2213
2214	static int move_module(struct module mod, struct load_info info)
2215	{
2216	int i;
2217	void *ptr;
2218	enum mod_mem_type t = 0;
2219	int ret = -ENOMEM;
2220
2221	for_each_mod_mem_type(type) {
2222	if (!mod->mem[type].size) {
2223	mod->mem[type].base = NULL;
2224	continue;
2225	}
2226	mod->mem[type].size = PAGE_ALIGN(mod->mem[type].size);
2227	ptr = module_memory_alloc(mod->mem[type].size, type);
2228	/*
2229	* The pointer to these blocks of memory are stored on the module
2230	* structure and we keep that around so long as the module is
2231	* around. We only free that memory when we unload the module.
2232	* Just mark them as not being a leak then. The .init* ELF
2233	* sections do get freed after boot so we could treat them
2234	* slightly differently with kmemleak_ignore() and only grey
2235	* them out as they work as typical memory allocations which
2236	* do eventually get freed, but let's just keep things simple
2237	* and avoid any false positives.
2238	*/
2239	kmemleak_not_leak(ptr);
2240	if (!ptr) {
2241	t = type;
2242	goto out_enomem;
2243	}
2244	memset(ptr, 0, mod->mem[type].size);
2245	mod->mem[type].base = ptr;
2246	}
2247
2248	/* Transfer each section which specifies SHF_ALLOC */
2249	pr_debug("Final section addresses for %s:\n", mod->name);
2250	for (i = 0; i < info->hdr->e_shnum; i++) {
2251	void *dest;
2252	Elf_Shdr *shdr = &info->sechdrs[i];
2253	enum mod_mem_type type = shdr->sh_entsize >> SH_ENTSIZE_TYPE_SHIFT;
2254
2255	if (!(shdr->sh_flags & SHF_ALLOC))
2256	continue;
2257
2258	dest = mod->mem[type].base + (shdr->sh_entsize & SH_ENTSIZE_OFFSET_MASK);
2259
2260	if (shdr->sh_type != SHT_NOBITS) {
2261	/*
2262	* Our ELF checker already validated this, but let's
2263	* be pedantic and make the goal clearer. We actually
2264	* end up copying over all modifications made to the
2265	* userspace copy of the entire struct module.
2266	*/
2267	if (i == info->index.mod &&
2268	(WARN_ON_ONCE(shdr->sh_size != sizeof(struct module)))) {
2269	ret = -ENOEXEC;
2270	goto out_enomem;
2271	}
2272	memcpy(dest, (void *)shdr->sh_addr, shdr->sh_size);
2273	}
2274	/*
2275	* Update the userspace copy's ELF section address to point to
2276	* our newly allocated memory as a pure convenience so that
2277	* users of info can keep taking advantage and using the newly
2278	* minted official memory area.
2279	*/
2280	shdr->sh_addr = (unsigned long)dest;
2281	pr_debug("\t0x%lx 0x%.8lx %s\n", (long)shdr->sh_addr,
2282	(long)shdr->sh_size, info->secstrings + shdr->sh_name);
2283	}
2284
2285	return 0;
2286	out_enomem:
2287	for (t--; t >= 0; t--)
2288	module_memory_free(mod->mem[t].base, t);
2289	return ret;
2290	}
2291
2292	static int check_export_symbol_versions(struct module *mod)
2293	{
2294	#ifdef CONFIG_MODVERSIONS
2295	if ((mod->num_syms && !mod->crcs) \|\|
2296	(mod->num_gpl_syms && !mod->gpl_crcs)) {
2297	return try_to_force_load(mod,
2298	"no versions for exported symbols");
2299	}
2300	#endif
2301	return 0;
2302	}
2303
2304	static void flush_module_icache(const struct module *mod)
2305	{
2306	/*
2307	* Flush the instruction cache, since we've played with text.
2308	* Do it before processing of module parameters, so the module
2309	* can provide parameter accessor functions of its own.
2310	*/
2311	for_each_mod_mem_type(type) {
2312	const struct module_memory *mod_mem = &mod->mem[type];
2313
2314	if (mod_mem->size) {
2315	flush_icache_range((unsigned long)mod_mem->base,
2316	(unsigned long)mod_mem->base + mod_mem->size);
2317	}
2318	}
2319	}
2320
2321	bool __weak module_elf_check_arch(Elf_Ehdr *hdr)
2322	{
2323	return true;
2324	}
2325
2326	int __weak module_frob_arch_sections(Elf_Ehdr *hdr,
2327	Elf_Shdr *sechdrs,
2328	char *secstrings,
2329	struct module *mod)
2330	{
2331	return 0;
2332	}
2333
2334	/* module_blacklist is a comma-separated list of module names */
2335	static char *module_blacklist;
2336	static bool blacklisted(const char *module_name)
2337	{
2338	const char *p;
2339	size_t len;
2340
2341	if (!module_blacklist)
2342	return false;
2343
2344	for (p = module_blacklist; *p; p += len) {
2345	len = strcspn(p, ",");
2346	if (strlen(module_name) == len && !memcmp(module_name, p, len))
2347	return true;
2348	if (p[len] == ',')
2349	len++;
2350	}
2351	return false;
2352	}
2353	core_param(module_blacklist, module_blacklist, charp, 0400);
2354
2355	static struct module layout_and_allocate(struct load_info info, int flags)
2356	{
2357	struct module *mod;
2358	unsigned int ndx;
2359	int err;
2360
2361	/* Allow arches to frob section contents and sizes. */
2362	err = module_frob_arch_sections(info->hdr, info->sechdrs,
2363	info->secstrings, info->mod);
2364	if (err < 0)
2365	return ERR_PTR(err);
2366
2367	err = module_enforce_rwx_sections(info->hdr, info->sechdrs,
2368	info->secstrings, info->mod);
2369	if (err < 0)
2370	return ERR_PTR(err);
2371
2372	/* We will do a special allocation for per-cpu sections later. */
2373	info->sechdrs[info->index.pcpu].sh_flags &= ~(unsigned long)SHF_ALLOC;
2374
2375	/*
2376	* Mark ro_after_init section with SHF_RO_AFTER_INIT so that
2377	* layout_sections() can put it in the right place.
2378	* Note: ro_after_init sections also have SHF_{WRITE,ALLOC} set.
2379	*/
2380	ndx = find_sec(info, ".data..ro_after_init");
2381	if (ndx)
2382	info->sechdrs[ndx].sh_flags \|= SHF_RO_AFTER_INIT;
2383	/*
2384	* Mark the __jump_table section as ro_after_init as well: these data
2385	* structures are never modified, with the exception of entries that
2386	* refer to code in the __init section, which are annotated as such
2387	* at module load time.
2388	*/
2389	ndx = find_sec(info, "__jump_table");
2390	if (ndx)
2391	info->sechdrs[ndx].sh_flags \|= SHF_RO_AFTER_INIT;
2392
2393	/*
2394	* Determine total sizes, and put offsets in sh_entsize. For now
2395	* this is done generically; there doesn't appear to be any
2396	* special cases for the architectures.
2397	*/
2398	layout_sections(info->mod, info);
2399	layout_symtab(info->mod, info);
2400
2401	/* Allocate and move to the final place */
2402	err = move_module(info->mod, info);
2403	if (err)
2404	return ERR_PTR(err);
2405
2406	/* Module has been copied to its final place now: return it. */
2407	mod = (void *)info->sechdrs[info->index.mod].sh_addr;
2408	kmemleak_load_module(mod, info);
2409	return mod;
2410	}
2411
2412	/* mod is no longer valid after this! */
2413	static void module_deallocate(struct module mod, struct load_info info)
2414	{
2415	percpu_modfree(mod);
2416	module_arch_freeing_init(mod);
2417
2418	free_mod_mem(mod);
2419	}
2420
2421	int __weak module_finalize(const Elf_Ehdr *hdr,
2422	const Elf_Shdr *sechdrs,
2423	struct module *me)
2424	{
2425	return 0;
2426	}
2427
2428	static int post_relocation(struct module mod, const struct load_info info)
2429	{
2430	/* Sort exception table now relocations are done. */
2431	sort_extable(mod->extable, mod->extable + mod->num_exentries);
2432
2433	/* Copy relocated percpu area over. */
2434	percpu_modcopy(mod, (void *)info->sechdrs[info->index.pcpu].sh_addr,
2435	info->sechdrs[info->index.pcpu].sh_size);
2436
2437	/* Setup kallsyms-specific fields. */
2438	add_kallsyms(mod, info);
2439
2440	/* Arch-specific module finalizing. */
2441	return module_finalize(info->hdr, info->sechdrs, mod);
2442	}
2443
2444	/* Call module constructors. */
2445	static void do_mod_ctors(struct module *mod)
2446	{
2447	#ifdef CONFIG_CONSTRUCTORS
2448	unsigned long i;
2449
2450	for (i = 0; i < mod->num_ctors; i++)
2451	mod->ctors[i]();
2452	#endif
2453	}
2454
2455	/* For freeing module_init on success, in case kallsyms traversing */
2456	struct mod_initfree {
2457	struct llist_node node;
2458	void *init_text;
2459	void *init_data;
2460	void *init_rodata;
2461	};
2462
2463	static void do_free_init(struct work_struct *w)
2464	{
2465	struct llist_node pos, n, *list;
2466	struct mod_initfree *initfree;
2467
2468	list = llist_del_all(&init_free_list);
2469
2470	synchronize_rcu();
2471
2472	llist_for_each_safe(pos, n, list) {
2473	initfree = container_of(pos, struct mod_initfree, node);
2474	module_memfree(initfree->init_text);
2475	module_memfree(initfree->init_data);
2476	module_memfree(initfree->init_rodata);
2477	kfree(initfree);
2478	}
2479	}
2480
2481	#undef MODULE_PARAM_PREFIX
2482	#define MODULE_PARAM_PREFIX "module."
2483	/* Default value for module->async_probe_requested */
2484	static bool async_probe;
2485	module_param(async_probe, bool, 0644);
2486
2487	/*
2488	* This is where the real work happens.
2489	*
2490	* Keep it uninlined to provide a reliable breakpoint target, e.g. for the gdb
2491	* helper command 'lx-symbols'.
2492	*/
2493	static noinline int do_init_module(struct module *mod)
2494	{
2495	int ret = 0;
2496	struct mod_initfree *freeinit;
2497	#if defined(CONFIG_MODULE_STATS)
2498	unsigned int text_size = 0, total_size = 0;
2499
2500	for_each_mod_mem_type(type) {
2501	const struct module_memory *mod_mem = &mod->mem[type];
2502	if (mod_mem->size) {
2503	total_size += mod_mem->size;
2504	if (type == MOD_TEXT \|\| type == MOD_INIT_TEXT)
2505	text_size += mod_mem->size;
2506	}
2507	}
2508	#endif
2509
2510	freeinit = kmalloc(sizeof(*freeinit), GFP_KERNEL);
2511	if (!freeinit) {
2512	ret = -ENOMEM;
2513	goto fail;
2514	}
2515	freeinit->init_text = mod->mem[MOD_INIT_TEXT].base;
2516	freeinit->init_data = mod->mem[MOD_INIT_DATA].base;
2517	freeinit->init_rodata = mod->mem[MOD_INIT_RODATA].base;
2518
2519	do_mod_ctors(mod);
2520	/* Start the module */
2521	if (mod->init != NULL)
2522	ret = do_one_initcall(mod->init);
2523	if (ret < 0) {
2524	goto fail_free_freeinit;
2525	}
2526	if (ret > 0) {
2527	pr_warn("%s: '%s'->init suspiciously returned %d, it should "
2528	"follow 0/-E convention\n"
2529	"%s: loading module anyway...\n",
2530	__func__, mod->name, ret, __func__);
2531	dump_stack();
2532	}
2533
2534	/* Now it's a first class citizen! */
2535	mod->state = MODULE_STATE_LIVE;
2536	blocking_notifier_call_chain(&module_notify_list,
2537	MODULE_STATE_LIVE, mod);
2538
2539	/* Delay uevent until module has finished its init routine */
2540	kobject_uevent(&mod->mkobj.kobj, KOBJ_ADD);
2541
2542	/*
2543	* We need to finish all async code before the module init sequence
2544	* is done. This has potential to deadlock if synchronous module
2545	* loading is requested from async (which is not allowed!).
2546	*
2547	* See commit 0fdff3ec6d87 ("async, kmod: warn on synchronous
2548	* request_module() from async workers") for more details.
2549	*/
2550	if (!mod->async_probe_requested)
2551	async_synchronize_full();
2552
2553	ftrace_free_mem(mod, mod->mem[MOD_INIT_TEXT].base,
2554	mod->mem[MOD_INIT_TEXT].base + mod->mem[MOD_INIT_TEXT].size);
2555	mutex_lock(&module_mutex);
2556	/* Drop initial reference. */
2557	module_put(mod);
2558	trim_init_extable(mod);
2559	#ifdef CONFIG_KALLSYMS
2560	/* Switch to core kallsyms now init is done: kallsyms may be walking! */
2561	rcu_assign_pointer(mod->kallsyms, &mod->core_kallsyms);
2562	#endif
2563	module_enable_ro(mod, true);
2564	mod_tree_remove_init(mod);
2565	module_arch_freeing_init(mod);
2566	for_class_mod_mem_type(type, init) {
2567	mod->mem[type].base = NULL;
2568	mod->mem[type].size = 0;
2569	}
2570
2571	#ifdef CONFIG_DEBUG_INFO_BTF_MODULES
2572	/* .BTF is not SHF_ALLOC and will get removed, so sanitize pointer */
2573	mod->btf_data = NULL;
2574	#endif
2575	/*
2576	* We want to free module_init, but be aware that kallsyms may be
2577	* walking this with preempt disabled. In all the failure paths, we
2578	* call synchronize_rcu(), but we don't want to slow down the success
2579	* path. module_memfree() cannot be called in an interrupt, so do the
2580	* work and call synchronize_rcu() in a work queue.
2581	*
2582	* Note that module_alloc() on most architectures creates W+X page
2583	* mappings which won't be cleaned up until do_free_init() runs. Any
2584	* code such as mark_rodata_ro() which depends on those mappings to
2585	* be cleaned up needs to sync with the queued work - ie
2586	* rcu_barrier()
2587	*/
2588	if (llist_add(&freeinit->node, &init_free_list))
2589	schedule_work(&init_free_wq);
2590
2591	mutex_unlock(&module_mutex);
2592	wake_up_all(&module_wq);
2593
2594	mod_stat_add_long(text_size, &total_text_size);
2595	mod_stat_add_long(total_size, &total_mod_size);
2596
2597	mod_stat_inc(&modcount);
2598
2599	return 0;
2600
2601	fail_free_freeinit:
2602	kfree(freeinit);
2603	fail:
2604	/* Try to protect us from buggy refcounters. */
2605	mod->state = MODULE_STATE_GOING;
2606	synchronize_rcu();
2607	module_put(mod);
2608	blocking_notifier_call_chain(&module_notify_list,
2609	MODULE_STATE_GOING, mod);
2610	klp_module_going(mod);
2611	ftrace_release_mod(mod);
2612	free_module(mod);
2613	wake_up_all(&module_wq);
2614
2615	return ret;
2616	}
2617
2618	static int may_init_module(void)
2619	{
2620	if (!capable(CAP_SYS_MODULE) \|\| modules_disabled)
2621	return -EPERM;
2622
2623	return 0;
2624	}
2625
2626	/* Is this module of this name done loading? No locks held. */
2627	static bool finished_loading(const char *name)
2628	{
2629	struct module *mod;
2630	bool ret;
2631
2632	/*
2633	* The module_mutex should not be a heavily contended lock;
2634	* if we get the occasional sleep here, we'll go an extra iteration
2635	* in the wait_event_interruptible(), which is harmless.
2636	*/
2637	sched_annotate_sleep();
2638	mutex_lock(&module_mutex);
2639	mod = find_module_all(name, strlen(name), true);
2640	ret = !mod \|\| mod->state == MODULE_STATE_LIVE
2641	\|\| mod->state == MODULE_STATE_GOING;
2642	mutex_unlock(&module_mutex);
2643
2644	return ret;
2645	}
2646
2647	/* Must be called with module_mutex held */
2648	static int module_patient_check_exists(const char *name,
2649	enum fail_dup_mod_reason reason)
2650	{
2651	struct module *old;
2652	int err = 0;
2653
2654	old = find_module_all(name, strlen(name), true);
2655	if (old == NULL)
2656	return 0;
2657
2658	if (old->state == MODULE_STATE_COMING \|\|
2659	old->state == MODULE_STATE_UNFORMED) {
2660	/* Wait in case it fails to load. */
2661	mutex_unlock(&module_mutex);
2662	err = wait_event_interruptible(module_wq,
2663	finished_loading(name));
2664	mutex_lock(&module_mutex);
2665	if (err)
2666	return err;
2667
2668	/* The module might have gone in the meantime. */
2669	old = find_module_all(name, strlen(name), true);
2670	}
2671
2672	if (try_add_failed_module(name, reason))
2673	pr_warn("Could not add fail-tracking for module: %s\n", name);
2674
2675	/*
2676	* We are here only when the same module was being loaded. Do
2677	* not try to load it again right now. It prevents long delays
2678	* caused by serialized module load failures. It might happen
2679	* when more devices of the same type trigger load of
2680	* a particular module.
2681	*/
2682	if (old && old->state == MODULE_STATE_LIVE)
2683	return -EEXIST;
2684	return -EBUSY;
2685	}
2686
2687	/*
2688	* We try to place it in the list now to make sure it's unique before
2689	* we dedicate too many resources. In particular, temporary percpu
2690	* memory exhaustion.
2691	*/
2692	static int add_unformed_module(struct module *mod)
2693	{
2694	int err;
2695
2696	mod->state = MODULE_STATE_UNFORMED;
2697
2698	mutex_lock(&module_mutex);
2699	err = module_patient_check_exists(mod->name, FAIL_DUP_MOD_LOAD);
2700	if (err)
2701	goto out;
2702
2703	mod_update_bounds(mod);
2704	list_add_rcu(&mod->list, &modules);
2705	mod_tree_insert(mod);
2706	err = 0;
2707
2708	out:
2709	mutex_unlock(&module_mutex);
2710	return err;
2711	}
2712
2713	static int complete_formation(struct module mod, struct load_info info)
2714	{
2715	int err;
2716
2717	mutex_lock(&module_mutex);
2718
2719	/* Find duplicate symbols (must be called under lock). */
2720	err = verify_exported_symbols(mod);
2721	if (err < 0)
2722	goto out;
2723
2724	/* These rely on module_mutex for list integrity. */
2725	module_bug_finalize(info->hdr, info->sechdrs, mod);
2726	module_cfi_finalize(info->hdr, info->sechdrs, mod);
2727
2728	module_enable_ro(mod, false);
2729	module_enable_nx(mod);
2730	module_enable_x(mod);
2731
2732	/*
2733	* Mark state as coming so strong_try_module_get() ignores us,
2734	* but kallsyms etc. can see us.
2735	*/
2736	mod->state = MODULE_STATE_COMING;
2737	mutex_unlock(&module_mutex);
2738
2739	return 0;
2740
2741	out:
2742	mutex_unlock(&module_mutex);
2743	return err;
2744	}
2745
2746	static int prepare_coming_module(struct module *mod)
2747	{
2748	int err;
2749
2750	ftrace_module_enable(mod);
2751	err = klp_module_coming(mod);
2752	if (err)
2753	return err;
2754
2755	err = blocking_notifier_call_chain_robust(&module_notify_list,
2756	MODULE_STATE_COMING, MODULE_STATE_GOING, mod);
2757	err = notifier_to_errno(err);
2758	if (err)
2759	klp_module_going(mod);
2760
2761	return err;
2762	}
2763
2764	static int unknown_module_param_cb(char param, char val, const char *modname,
2765	void *arg)
2766	{
2767	struct module *mod = arg;
2768	int ret;
2769
2770	if (strcmp(param, "async_probe") == 0) {
2771	if (kstrtobool(val, &mod->async_probe_requested))
2772	mod->async_probe_requested = true;
2773	return 0;
2774	}
2775
2776	/* Check for magic 'dyndbg' arg */
2777	ret = ddebug_dyndbg_module_param_cb(param, val, modname);
2778	if (ret != 0)
2779	pr_warn("%s: unknown parameter '%s' ignored\n", modname, param);
2780	return 0;
2781	}
2782
2783	/* Module within temporary copy, this doesn't do any allocation */
2784	static int early_mod_check(struct load_info *info, int flags)
2785	{
2786	int err;
2787
2788	/*
2789	* Now that we know we have the correct module name, check
2790	* if it's blacklisted.
2791	*/
2792	if (blacklisted(info->name)) {
2793	pr_err("Module %s is blacklisted\n", info->name);
2794	return -EPERM;
2795	}
2796
2797	err = rewrite_section_headers(info, flags);
2798	if (err)
2799	return err;
2800
2801	/* Check module struct version now, before we try to use module. */
2802	if (!check_modstruct_version(info, info->mod))
2803	return -ENOEXEC;
2804
2805	err = check_modinfo(info->mod, info, flags);
2806	if (err)
2807	return err;
2808
2809	mutex_lock(&module_mutex);
2810	err = module_patient_check_exists(info->mod->name, FAIL_DUP_MOD_BECOMING);
2811	mutex_unlock(&module_mutex);
2812
2813	return err;
2814	}
2815
2816	/*
2817	* Allocate and load the module: note that size of section 0 is always
2818	* zero, and we rely on this for optional sections.
2819	*/
2820	static int load_module(struct load_info info, const char __user uargs,
2821	int flags)
2822	{
2823	struct module *mod;
2824	bool module_allocated = false;
2825	long err = 0;
2826	char *after_dashes;
2827
2828	/*
2829	* Do the signature check (if any) first. All that
2830	* the signature check needs is info->len, it does
2831	* not need any of the section info. That can be
2832	* set up later. This will minimize the chances
2833	* of a corrupt module causing problems before
2834	* we even get to the signature check.
2835	*
2836	* The check will also adjust info->len by stripping
2837	* off the sig length at the end of the module, making
2838	* checks against info->len more correct.
2839	*/
2840	err = module_sig_check(info, flags);
2841	if (err)
2842	goto free_copy;
2843
2844	/*
2845	* Do basic sanity checks against the ELF header and
2846	* sections. Cache useful sections and set the
2847	* info->mod to the userspace passed struct module.
2848	*/
2849	err = elf_validity_cache_copy(info, flags);
2850	if (err)
2851	goto free_copy;
2852
2853	err = early_mod_check(info, flags);
2854	if (err)
2855	goto free_copy;
2856
2857	/* Figure out module layout, and allocate all the memory. */
2858	mod = layout_and_allocate(info, flags);
2859	if (IS_ERR(mod)) {
2860	err = PTR_ERR(mod);
2861	goto free_copy;
2862	}
2863
2864	module_allocated = true;
2865
2866	audit_log_kern_module(mod->name);
2867
2868	/* Reserve our place in the list. */
2869	err = add_unformed_module(mod);
2870	if (err)
2871	goto free_module;
2872
2873	/*
2874	* We are tainting your kernel if your module gets into
2875	* the modules linked list somehow.
2876	*/
2877	module_augment_kernel_taints(mod, info);
2878
2879	/* To avoid stressing percpu allocator, do this once we're unique. */
2880	err = percpu_modalloc(mod, info);
2881	if (err)
2882	goto unlink_mod;
2883
2884	/* Now module is in final location, initialize linked lists, etc. */
2885	err = module_unload_init(mod);
2886	if (err)
2887	goto unlink_mod;
2888
2889	init_param_lock(mod);
2890
2891	/*
2892	* Now we've got everything in the final locations, we can
2893	* find optional sections.
2894	*/
2895	err = find_module_sections(mod, info);
2896	if (err)
2897	goto free_unload;
2898
2899	err = check_export_symbol_versions(mod);
2900	if (err)
2901	goto free_unload;
2902
2903	/* Set up MODINFO_ATTR fields */
2904	setup_modinfo(mod, info);
2905
2906	/* Fix up syms, so that st_value is a pointer to location. */
2907	err = simplify_symbols(mod, info);
2908	if (err < 0)
2909	goto free_modinfo;
2910
2911	err = apply_relocations(mod, info);
2912	if (err < 0)
2913	goto free_modinfo;
2914
2915	err = post_relocation(mod, info);
2916	if (err < 0)
2917	goto free_modinfo;
2918
2919	flush_module_icache(mod);
2920
2921	/* Now copy in args */
2922	mod->args = strndup_user(uargs, ~0UL >> 1);
2923	if (IS_ERR(mod->args)) {
2924	err = PTR_ERR(mod->args);
2925	goto free_arch_cleanup;
2926	}
2927
2928	init_build_id(mod, info);
2929
2930	/* Ftrace init must be called in the MODULE_STATE_UNFORMED state */
2931	ftrace_module_init(mod);
2932
2933	/* Finally it's fully formed, ready to start executing. */
2934	err = complete_formation(mod, info);
2935	if (err)
2936	goto ddebug_cleanup;
2937
2938	err = prepare_coming_module(mod);
2939	if (err)
2940	goto bug_cleanup;
2941
2942	mod->async_probe_requested = async_probe;
2943
2944	/* Module is ready to execute: parsing args may do that. */
2945	after_dashes = parse_args(mod->name, mod->args, mod->kp, mod->num_kp,
2946	-32768, 32767, mod,
2947	unknown_module_param_cb);
2948	if (IS_ERR(after_dashes)) {
2949	err = PTR_ERR(after_dashes);
2950	goto coming_cleanup;
2951	} else if (after_dashes) {
2952	pr_warn("%s: parameters '%s' after `--' ignored\n",
2953	mod->name, after_dashes);
2954	}
2955
2956	/* Link in to sysfs. */
2957	err = mod_sysfs_setup(mod, info, mod->kp, mod->num_kp);
2958	if (err < 0)
2959	goto coming_cleanup;
2960
2961	if (is_livepatch_module(mod)) {
2962	err = copy_module_elf(mod, info);
2963	if (err < 0)
2964	goto sysfs_cleanup;
2965	}
2966
2967	/* Get rid of temporary copy. */
2968	free_copy(info, flags);
2969
2970	/* Done! */
2971	trace_module_load(mod);
2972
2973	return do_init_module(mod);
2974
2975	sysfs_cleanup:
2976	mod_sysfs_teardown(mod);
2977	coming_cleanup:
2978	mod->state = MODULE_STATE_GOING;
2979	destroy_params(mod->kp, mod->num_kp);
2980	blocking_notifier_call_chain(&module_notify_list,
2981	MODULE_STATE_GOING, mod);
2982	klp_module_going(mod);
2983	bug_cleanup:
2984	mod->state = MODULE_STATE_GOING;
2985	/* module_bug_cleanup needs module_mutex protection */
2986	mutex_lock(&module_mutex);
2987	module_bug_cleanup(mod);
2988	mutex_unlock(&module_mutex);
2989
2990	ddebug_cleanup:
2991	ftrace_release_mod(mod);
2992	synchronize_rcu();
2993	kfree(mod->args);
2994	free_arch_cleanup:
2995	module_arch_cleanup(mod);
2996	free_modinfo:
2997	free_modinfo(mod);
2998	free_unload:
2999	module_unload_free(mod);
3000	unlink_mod:
3001	mutex_lock(&module_mutex);
3002	/* Unlink carefully: kallsyms could be walking list. */
3003	list_del_rcu(&mod->list);
3004	mod_tree_remove(mod);
3005	wake_up_all(&module_wq);
3006	/* Wait for RCU-sched synchronizing before releasing mod->list. */
3007	synchronize_rcu();
3008	mutex_unlock(&module_mutex);
3009	free_module:
3010	mod_stat_bump_invalid(info, flags);
3011	/* Free lock-classes; relies on the preceding sync_rcu() */
3012	for_class_mod_mem_type(type, core_data) {
3013	lockdep_free_key_range(mod->mem[type].base,
3014	mod->mem[type].size);
3015	}
3016
3017	module_deallocate(mod, info);
3018	free_copy:
3019	/*
3020	* The info->len is always set. We distinguish between
3021	* failures once the proper module was allocated and
3022	* before that.
3023	*/
3024	if (!module_allocated)
3025	mod_stat_bump_becoming(info, flags);
3026	free_copy(info, flags);
3027	return err;
3028	}
3029
3030	SYSCALL_DEFINE3(init_module, void __user *, umod,
3031	unsigned long, len, const char __user *, uargs)
3032	{
3033	int err;
3034	struct load_info info = { };
3035
3036	err = may_init_module();
3037	if (err)
3038	return err;
3039
3040	pr_debug("init_module: umod=%p, len=%lu, uargs=%p\n",
3041	umod, len, uargs);
3042
3043	err = copy_module_from_user(umod, len, &info);
3044	if (err) {
3045	mod_stat_inc(&failed_kreads);
3046	mod_stat_add_long(len, &invalid_kread_bytes);
3047	return err;
3048	}
3049
3050	return load_module(&info, uargs, 0);
3051	}
3052
3053	struct idempotent {
3054	const void *cookie;
3055	struct hlist_node entry;
3056	struct completion complete;
3057	int ret;
3058	};
3059
3060	#define IDEM_HASH_BITS 8
3061	static struct hlist_head idem_hash[1 << IDEM_HASH_BITS];
3062	static DEFINE_SPINLOCK(idem_lock);
3063
3064	static bool idempotent(struct idempotent u, const void cookie)
3065	{
3066	int hash = hash_ptr(cookie, IDEM_HASH_BITS);
3067	struct hlist_head *head = idem_hash + hash;
3068	struct idempotent *existing;
3069	bool first;
3070
3071	u->ret = 0;
3072	u->cookie = cookie;
3073	init_completion(&u->complete);
3074
3075	spin_lock(&idem_lock);
3076	first = true;
3077	hlist_for_each_entry(existing, head, entry) {
3078	if (existing->cookie != cookie)
3079	continue;
3080	first = false;
3081	break;
3082	}
3083	hlist_add_head(&u->entry, idem_hash + hash);
3084	spin_unlock(&idem_lock);
3085
3086	return !first;
3087	}
3088
3089	/*
3090	* We were the first one with 'cookie' on the list, and we ended
3091	* up completing the operation. We now need to walk the list,
3092	* remove everybody - which includes ourselves - fill in the return
3093	* value, and then complete the operation.
3094	*/
3095	static int idempotent_complete(struct idempotent *u, int ret)
3096	{
3097	const void *cookie = u->cookie;
3098	int hash = hash_ptr(cookie, IDEM_HASH_BITS);
3099	struct hlist_head *head = idem_hash + hash;
3100	struct hlist_node *next;
3101	struct idempotent *pos;
3102
3103	spin_lock(&idem_lock);
3104	hlist_for_each_entry_safe(pos, next, head, entry) {
3105	if (pos->cookie != cookie)
3106	continue;
3107	hlist_del(&pos->entry);
3108	pos->ret = ret;
3109	complete(&pos->complete);
3110	}
3111	spin_unlock(&idem_lock);
3112	return ret;
3113	}
3114
3115	static int init_module_from_file(struct file f, const char __user uargs, int flags)
3116	{
3117	struct load_info info = { };
3118	void *buf = NULL;
3119	int len;
3120
3121	len = kernel_read_file(f, 0, &buf, INT_MAX, NULL, READING_MODULE);
3122	if (len < 0) {
3123	mod_stat_inc(&failed_kreads);
3124	return len;
3125	}
3126
3127	if (flags & MODULE_INIT_COMPRESSED_FILE) {
3128	int err = module_decompress(&info, buf, len);
3129	vfree(buf); /* compressed data is no longer needed */
3130	if (err) {
3131	mod_stat_inc(&failed_decompress);
3132	mod_stat_add_long(len, &invalid_decompress_bytes);
3133	return err;
3134	}
3135	} else {
3136	info.hdr = buf;
3137	info.len = len;
3138	}
3139
3140	return load_module(&info, uargs, flags);
3141	}
3142
3143	static int idempotent_init_module(struct file f, const char __user uargs, int flags)
3144	{
3145	struct idempotent idem;
3146
3147	if (!f \|\| !(f->f_mode & FMODE_READ))
3148	return -EBADF;
3149
3150	/* See if somebody else is doing the operation? */
3151	if (idempotent(&idem, file_inode(f))) {
3152	wait_for_completion(&idem.complete);
3153	return idem.ret;
3154	}
3155
3156	/* Otherwise, we'll do it and complete others */
3157	return idempotent_complete(&idem,
3158	init_module_from_file(f, uargs, flags));
3159	}
3160
3161	SYSCALL_DEFINE3(finit_module, int, fd, const char __user *, uargs, int, flags)
3162	{
3163	int err;
3164	struct fd f;
3165
3166	err = may_init_module();
3167	if (err)
3168	return err;
3169
3170	pr_debug("finit_module: fd=%d, uargs=%p, flags=%i\n", fd, uargs, flags);
3171
3172	if (flags & ~(MODULE_INIT_IGNORE_MODVERSIONS
3173	\|MODULE_INIT_IGNORE_VERMAGIC
3174	\|MODULE_INIT_COMPRESSED_FILE))
3175	return -EINVAL;
3176
3177	f = fdget(fd);
3178	err = idempotent_init_module(f.file, uargs, flags);
3179	fdput(f);
3180	return err;
3181	}
3182
3183	/* Keep in sync with MODULE_FLAGS_BUF_SIZE !!! */
3184	char module_flags(struct module mod, char *buf, bool show_state)
3185	{
3186	int bx = 0;
3187
3188	BUG_ON(mod->state == MODULE_STATE_UNFORMED);
3189	if (!mod->taints && !show_state)
3190	goto out;
3191	if (mod->taints \|\|
3192	mod->state == MODULE_STATE_GOING \|\|
3193	mod->state == MODULE_STATE_COMING) {
3194	buf[bx++] = '(';
3195	bx += module_flags_taint(mod->taints, buf + bx);
3196	/* Show a - for module-is-being-unloaded */
3197	if (mod->state == MODULE_STATE_GOING && show_state)
3198	buf[bx++] = '-';
3199	/* Show a + for module-is-being-loaded */
3200	if (mod->state == MODULE_STATE_COMING && show_state)
3201	buf[bx++] = '+';
3202	buf[bx++] = ')';
3203	}
3204	out:
3205	buf[bx] = '\0';
3206
3207	return buf;
3208	}
3209
3210	/* Given an address, look for it in the module exception tables. */
3211	const struct exception_table_entry *search_module_extables(unsigned long addr)
3212	{
3213	const struct exception_table_entry *e = NULL;
3214	struct module *mod;
3215
3216	preempt_disable();
3217	mod = __module_address(addr);
3218	if (!mod)
3219	goto out;
3220
3221	if (!mod->num_exentries)
3222	goto out;
3223
3224	e = search_extable(mod->extable,
3225	mod->num_exentries,
3226	addr);
3227	out:
3228	preempt_enable();
3229
3230	/*
3231	* Now, if we found one, we are running inside it now, hence
3232	* we cannot unload the module, hence no refcnt needed.
3233	*/
3234	return e;
3235	}
3236
3237	/**
3238	* is_module_address() - is this address inside a module?
3239	* @addr: the address to check.
3240	*
3241	* See is_module_text_address() if you simply want to see if the address
3242	* is code (not data).
3243	*/
3244	bool is_module_address(unsigned long addr)
3245	{
3246	bool ret;
3247
3248	preempt_disable();
3249	ret = __module_address(addr) != NULL;
3250	preempt_enable();
3251
3252	return ret;
3253	}
3254
3255	/**
3256	* __module_address() - get the module which contains an address.
3257	* @addr: the address.
3258	*
3259	* Must be called with preempt disabled or module mutex held so that
3260	* module doesn't get freed during this.
3261	*/
3262	struct module *__module_address(unsigned long addr)
3263	{
3264	struct module *mod;
3265
3266	if (addr >= mod_tree.addr_min && addr <= mod_tree.addr_max)
3267	goto lookup;
3268
3269	#ifdef CONFIG_ARCH_WANTS_MODULES_DATA_IN_VMALLOC
3270	if (addr >= mod_tree.data_addr_min && addr <= mod_tree.data_addr_max)
3271	goto lookup;
3272	#endif
3273
3274	return NULL;
3275
3276	lookup:
3277	module_assert_mutex_or_preempt();
3278
3279	mod = mod_find(addr, &mod_tree);
3280	if (mod) {
3281	BUG_ON(!within_module(addr, mod));
3282	if (mod->state == MODULE_STATE_UNFORMED)
3283	mod = NULL;
3284	}
3285	return mod;
3286	}
3287
3288	/**
3289	* is_module_text_address() - is this address inside module code?
3290	* @addr: the address to check.
3291	*
3292	* See is_module_address() if you simply want to see if the address is
3293	* anywhere in a module. See kernel_text_address() for testing if an
3294	* address corresponds to kernel or module code.
3295	*/
3296	bool is_module_text_address(unsigned long addr)
3297	{
3298	bool ret;
3299
3300	preempt_disable();
3301	ret = __module_text_address(addr) != NULL;
3302	preempt_enable();
3303
3304	return ret;
3305	}
3306
3307	/**
3308	* __module_text_address() - get the module whose code contains an address.
3309	* @addr: the address.
3310	*
3311	* Must be called with preempt disabled or module mutex held so that
3312	* module doesn't get freed during this.
3313	*/
3314	struct module *__module_text_address(unsigned long addr)
3315	{
3316	struct module *mod = __module_address(addr);
3317	if (mod) {
3318	/* Make sure it's within the text section. */
3319	if (!within_module_mem_type(addr, mod, MOD_TEXT) &&
3320	!within_module_mem_type(addr, mod, MOD_INIT_TEXT))
3321	mod = NULL;
3322	}
3323	return mod;
3324	}
3325
3326	/* Don't grab lock, we're oopsing. */
3327	void print_modules(void)
3328	{
3329	struct module *mod;
3330	char buf[MODULE_FLAGS_BUF_SIZE];
3331
3332	printk(KERN_DEFAULT "Modules linked in:");
3333	/* Most callers should already have preempt disabled, but make sure */
3334	preempt_disable();
3335	list_for_each_entry_rcu(mod, &modules, list) {
3336	if (mod->state == MODULE_STATE_UNFORMED)
3337	continue;
3338	pr_cont(" %s%s", mod->name, module_flags(mod, buf, true));
3339	}
3340
3341	print_unloaded_tainted_modules();
3342	preempt_enable();
3343	if (last_unloaded_module.name[0])
3344	pr_cont(" [last unloaded: %s%s]", last_unloaded_module.name,
3345	last_unloaded_module.taints);
3346	pr_cont("\n");
3347	}
3348
3349	#ifdef CONFIG_MODULE_DEBUGFS
3350	struct dentry *mod_debugfs_root;
3351
3352	static int module_debugfs_init(void)
3353	{
3354	mod_debugfs_root = debugfs_create_dir("modules", NULL);
3355	return 0;
3356	}
3357	module_init(module_debugfs_init);
3358	#endif