xref: /linux/kernel/auditsc.c (revision 9f2c9170934eace462499ba0bfe042cc72900173)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* auditsc.c -- System-call auditing support
3  * Handles all system-call specific auditing features.
4  *
5  * Copyright 2003-2004 Red Hat Inc., Durham, North Carolina.
6  * Copyright 2005 Hewlett-Packard Development Company, L.P.
7  * Copyright (C) 2005, 2006 IBM Corporation
8  * All Rights Reserved.
9  *
10  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
11  *
12  * Many of the ideas implemented here are from Stephen C. Tweedie,
13  * especially the idea of avoiding a copy by using getname.
14  *
15  * The method for actual interception of syscall entry and exit (not in
16  * this file -- see entry.S) is based on a GPL'd patch written by
17  * okir@suse.de and Copyright 2003 SuSE Linux AG.
18  *
19  * POSIX message queue support added by George Wilson <ltcgcw@us.ibm.com>,
20  * 2006.
21  *
22  * The support of additional filter rules compares (>, <, >=, <=) was
23  * added by Dustin Kirkland <dustin.kirkland@us.ibm.com>, 2005.
24  *
25  * Modified by Amy Griffis <amy.griffis@hp.com> to collect additional
26  * filesystem information.
27  *
28  * Subject and object context labeling support added by <danjones@us.ibm.com>
29  * and <dustin.kirkland@us.ibm.com> for LSPP certification compliance.
30  */
31 
32 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
33 
34 #include <linux/init.h>
35 #include <asm/types.h>
36 #include <linux/atomic.h>
37 #include <linux/fs.h>
38 #include <linux/namei.h>
39 #include <linux/mm.h>
40 #include <linux/export.h>
41 #include <linux/slab.h>
42 #include <linux/mount.h>
43 #include <linux/socket.h>
44 #include <linux/mqueue.h>
45 #include <linux/audit.h>
46 #include <linux/personality.h>
47 #include <linux/time.h>
48 #include <linux/netlink.h>
49 #include <linux/compiler.h>
50 #include <asm/unistd.h>
51 #include <linux/security.h>
52 #include <linux/list.h>
53 #include <linux/binfmts.h>
54 #include <linux/highmem.h>
55 #include <linux/syscalls.h>
56 #include <asm/syscall.h>
57 #include <linux/capability.h>
58 #include <linux/fs_struct.h>
59 #include <linux/compat.h>
60 #include <linux/ctype.h>
61 #include <linux/string.h>
62 #include <linux/uaccess.h>
63 #include <linux/fsnotify_backend.h>
64 #include <uapi/linux/limits.h>
65 #include <uapi/linux/netfilter/nf_tables.h>
66 #include <uapi/linux/openat2.h> // struct open_how
67 
68 #include "audit.h"
69 
70 /* flags stating the success for a syscall */
71 #define AUDITSC_INVALID 0
72 #define AUDITSC_SUCCESS 1
73 #define AUDITSC_FAILURE 2
74 
75 /* no execve audit message should be longer than this (userspace limits),
76  * see the note near the top of audit_log_execve_info() about this value */
77 #define MAX_EXECVE_AUDIT_LEN 7500
78 
79 /* max length to print of cmdline/proctitle value during audit */
80 #define MAX_PROCTITLE_AUDIT_LEN 128
81 
82 /* number of audit rules */
83 int audit_n_rules;
84 
85 /* determines whether we collect data for signals sent */
86 int audit_signals;
87 
88 struct audit_aux_data {
89 	struct audit_aux_data	*next;
90 	int			type;
91 };
92 
93 /* Number of target pids per aux struct. */
94 #define AUDIT_AUX_PIDS	16
95 
96 struct audit_aux_data_pids {
97 	struct audit_aux_data	d;
98 	pid_t			target_pid[AUDIT_AUX_PIDS];
99 	kuid_t			target_auid[AUDIT_AUX_PIDS];
100 	kuid_t			target_uid[AUDIT_AUX_PIDS];
101 	unsigned int		target_sessionid[AUDIT_AUX_PIDS];
102 	u32			target_sid[AUDIT_AUX_PIDS];
103 	char 			target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
104 	int			pid_count;
105 };
106 
107 struct audit_aux_data_bprm_fcaps {
108 	struct audit_aux_data	d;
109 	struct audit_cap_data	fcap;
110 	unsigned int		fcap_ver;
111 	struct audit_cap_data	old_pcap;
112 	struct audit_cap_data	new_pcap;
113 };
114 
115 struct audit_tree_refs {
116 	struct audit_tree_refs *next;
117 	struct audit_chunk *c[31];
118 };
119 
120 struct audit_nfcfgop_tab {
121 	enum audit_nfcfgop	op;
122 	const char		*s;
123 };
124 
125 static const struct audit_nfcfgop_tab audit_nfcfgs[] = {
126 	{ AUDIT_XT_OP_REGISTER,			"xt_register"		   },
127 	{ AUDIT_XT_OP_REPLACE,			"xt_replace"		   },
128 	{ AUDIT_XT_OP_UNREGISTER,		"xt_unregister"		   },
129 	{ AUDIT_NFT_OP_TABLE_REGISTER,		"nft_register_table"	   },
130 	{ AUDIT_NFT_OP_TABLE_UNREGISTER,	"nft_unregister_table"	   },
131 	{ AUDIT_NFT_OP_CHAIN_REGISTER,		"nft_register_chain"	   },
132 	{ AUDIT_NFT_OP_CHAIN_UNREGISTER,	"nft_unregister_chain"	   },
133 	{ AUDIT_NFT_OP_RULE_REGISTER,		"nft_register_rule"	   },
134 	{ AUDIT_NFT_OP_RULE_UNREGISTER,		"nft_unregister_rule"	   },
135 	{ AUDIT_NFT_OP_SET_REGISTER,		"nft_register_set"	   },
136 	{ AUDIT_NFT_OP_SET_UNREGISTER,		"nft_unregister_set"	   },
137 	{ AUDIT_NFT_OP_SETELEM_REGISTER,	"nft_register_setelem"	   },
138 	{ AUDIT_NFT_OP_SETELEM_UNREGISTER,	"nft_unregister_setelem"   },
139 	{ AUDIT_NFT_OP_GEN_REGISTER,		"nft_register_gen"	   },
140 	{ AUDIT_NFT_OP_OBJ_REGISTER,		"nft_register_obj"	   },
141 	{ AUDIT_NFT_OP_OBJ_UNREGISTER,		"nft_unregister_obj"	   },
142 	{ AUDIT_NFT_OP_OBJ_RESET,		"nft_reset_obj"		   },
143 	{ AUDIT_NFT_OP_FLOWTABLE_REGISTER,	"nft_register_flowtable"   },
144 	{ AUDIT_NFT_OP_FLOWTABLE_UNREGISTER,	"nft_unregister_flowtable" },
145 	{ AUDIT_NFT_OP_INVALID,			"nft_invalid"		   },
146 };
147 
148 static int audit_match_perm(struct audit_context *ctx, int mask)
149 {
150 	unsigned n;
151 
152 	if (unlikely(!ctx))
153 		return 0;
154 	n = ctx->major;
155 
156 	switch (audit_classify_syscall(ctx->arch, n)) {
157 	case AUDITSC_NATIVE:
158 		if ((mask & AUDIT_PERM_WRITE) &&
159 		     audit_match_class(AUDIT_CLASS_WRITE, n))
160 			return 1;
161 		if ((mask & AUDIT_PERM_READ) &&
162 		     audit_match_class(AUDIT_CLASS_READ, n))
163 			return 1;
164 		if ((mask & AUDIT_PERM_ATTR) &&
165 		     audit_match_class(AUDIT_CLASS_CHATTR, n))
166 			return 1;
167 		return 0;
168 	case AUDITSC_COMPAT: /* 32bit on biarch */
169 		if ((mask & AUDIT_PERM_WRITE) &&
170 		     audit_match_class(AUDIT_CLASS_WRITE_32, n))
171 			return 1;
172 		if ((mask & AUDIT_PERM_READ) &&
173 		     audit_match_class(AUDIT_CLASS_READ_32, n))
174 			return 1;
175 		if ((mask & AUDIT_PERM_ATTR) &&
176 		     audit_match_class(AUDIT_CLASS_CHATTR_32, n))
177 			return 1;
178 		return 0;
179 	case AUDITSC_OPEN:
180 		return mask & ACC_MODE(ctx->argv[1]);
181 	case AUDITSC_OPENAT:
182 		return mask & ACC_MODE(ctx->argv[2]);
183 	case AUDITSC_SOCKETCALL:
184 		return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
185 	case AUDITSC_EXECVE:
186 		return mask & AUDIT_PERM_EXEC;
187 	case AUDITSC_OPENAT2:
188 		return mask & ACC_MODE((u32)ctx->openat2.flags);
189 	default:
190 		return 0;
191 	}
192 }
193 
194 static int audit_match_filetype(struct audit_context *ctx, int val)
195 {
196 	struct audit_names *n;
197 	umode_t mode = (umode_t)val;
198 
199 	if (unlikely(!ctx))
200 		return 0;
201 
202 	list_for_each_entry(n, &ctx->names_list, list) {
203 		if ((n->ino != AUDIT_INO_UNSET) &&
204 		    ((n->mode & S_IFMT) == mode))
205 			return 1;
206 	}
207 
208 	return 0;
209 }
210 
211 /*
212  * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
213  * ->first_trees points to its beginning, ->trees - to the current end of data.
214  * ->tree_count is the number of free entries in array pointed to by ->trees.
215  * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
216  * "empty" becomes (p, p, 31) afterwards.  We don't shrink the list (and seriously,
217  * it's going to remain 1-element for almost any setup) until we free context itself.
218  * References in it _are_ dropped - at the same time we free/drop aux stuff.
219  */
220 
221 static void audit_set_auditable(struct audit_context *ctx)
222 {
223 	if (!ctx->prio) {
224 		ctx->prio = 1;
225 		ctx->current_state = AUDIT_STATE_RECORD;
226 	}
227 }
228 
229 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
230 {
231 	struct audit_tree_refs *p = ctx->trees;
232 	int left = ctx->tree_count;
233 
234 	if (likely(left)) {
235 		p->c[--left] = chunk;
236 		ctx->tree_count = left;
237 		return 1;
238 	}
239 	if (!p)
240 		return 0;
241 	p = p->next;
242 	if (p) {
243 		p->c[30] = chunk;
244 		ctx->trees = p;
245 		ctx->tree_count = 30;
246 		return 1;
247 	}
248 	return 0;
249 }
250 
251 static int grow_tree_refs(struct audit_context *ctx)
252 {
253 	struct audit_tree_refs *p = ctx->trees;
254 
255 	ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
256 	if (!ctx->trees) {
257 		ctx->trees = p;
258 		return 0;
259 	}
260 	if (p)
261 		p->next = ctx->trees;
262 	else
263 		ctx->first_trees = ctx->trees;
264 	ctx->tree_count = 31;
265 	return 1;
266 }
267 
268 static void unroll_tree_refs(struct audit_context *ctx,
269 		      struct audit_tree_refs *p, int count)
270 {
271 	struct audit_tree_refs *q;
272 	int n;
273 
274 	if (!p) {
275 		/* we started with empty chain */
276 		p = ctx->first_trees;
277 		count = 31;
278 		/* if the very first allocation has failed, nothing to do */
279 		if (!p)
280 			return;
281 	}
282 	n = count;
283 	for (q = p; q != ctx->trees; q = q->next, n = 31) {
284 		while (n--) {
285 			audit_put_chunk(q->c[n]);
286 			q->c[n] = NULL;
287 		}
288 	}
289 	while (n-- > ctx->tree_count) {
290 		audit_put_chunk(q->c[n]);
291 		q->c[n] = NULL;
292 	}
293 	ctx->trees = p;
294 	ctx->tree_count = count;
295 }
296 
297 static void free_tree_refs(struct audit_context *ctx)
298 {
299 	struct audit_tree_refs *p, *q;
300 
301 	for (p = ctx->first_trees; p; p = q) {
302 		q = p->next;
303 		kfree(p);
304 	}
305 }
306 
307 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
308 {
309 	struct audit_tree_refs *p;
310 	int n;
311 
312 	if (!tree)
313 		return 0;
314 	/* full ones */
315 	for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
316 		for (n = 0; n < 31; n++)
317 			if (audit_tree_match(p->c[n], tree))
318 				return 1;
319 	}
320 	/* partial */
321 	if (p) {
322 		for (n = ctx->tree_count; n < 31; n++)
323 			if (audit_tree_match(p->c[n], tree))
324 				return 1;
325 	}
326 	return 0;
327 }
328 
329 static int audit_compare_uid(kuid_t uid,
330 			     struct audit_names *name,
331 			     struct audit_field *f,
332 			     struct audit_context *ctx)
333 {
334 	struct audit_names *n;
335 	int rc;
336 
337 	if (name) {
338 		rc = audit_uid_comparator(uid, f->op, name->uid);
339 		if (rc)
340 			return rc;
341 	}
342 
343 	if (ctx) {
344 		list_for_each_entry(n, &ctx->names_list, list) {
345 			rc = audit_uid_comparator(uid, f->op, n->uid);
346 			if (rc)
347 				return rc;
348 		}
349 	}
350 	return 0;
351 }
352 
353 static int audit_compare_gid(kgid_t gid,
354 			     struct audit_names *name,
355 			     struct audit_field *f,
356 			     struct audit_context *ctx)
357 {
358 	struct audit_names *n;
359 	int rc;
360 
361 	if (name) {
362 		rc = audit_gid_comparator(gid, f->op, name->gid);
363 		if (rc)
364 			return rc;
365 	}
366 
367 	if (ctx) {
368 		list_for_each_entry(n, &ctx->names_list, list) {
369 			rc = audit_gid_comparator(gid, f->op, n->gid);
370 			if (rc)
371 				return rc;
372 		}
373 	}
374 	return 0;
375 }
376 
377 static int audit_field_compare(struct task_struct *tsk,
378 			       const struct cred *cred,
379 			       struct audit_field *f,
380 			       struct audit_context *ctx,
381 			       struct audit_names *name)
382 {
383 	switch (f->val) {
384 	/* process to file object comparisons */
385 	case AUDIT_COMPARE_UID_TO_OBJ_UID:
386 		return audit_compare_uid(cred->uid, name, f, ctx);
387 	case AUDIT_COMPARE_GID_TO_OBJ_GID:
388 		return audit_compare_gid(cred->gid, name, f, ctx);
389 	case AUDIT_COMPARE_EUID_TO_OBJ_UID:
390 		return audit_compare_uid(cred->euid, name, f, ctx);
391 	case AUDIT_COMPARE_EGID_TO_OBJ_GID:
392 		return audit_compare_gid(cred->egid, name, f, ctx);
393 	case AUDIT_COMPARE_AUID_TO_OBJ_UID:
394 		return audit_compare_uid(audit_get_loginuid(tsk), name, f, ctx);
395 	case AUDIT_COMPARE_SUID_TO_OBJ_UID:
396 		return audit_compare_uid(cred->suid, name, f, ctx);
397 	case AUDIT_COMPARE_SGID_TO_OBJ_GID:
398 		return audit_compare_gid(cred->sgid, name, f, ctx);
399 	case AUDIT_COMPARE_FSUID_TO_OBJ_UID:
400 		return audit_compare_uid(cred->fsuid, name, f, ctx);
401 	case AUDIT_COMPARE_FSGID_TO_OBJ_GID:
402 		return audit_compare_gid(cred->fsgid, name, f, ctx);
403 	/* uid comparisons */
404 	case AUDIT_COMPARE_UID_TO_AUID:
405 		return audit_uid_comparator(cred->uid, f->op,
406 					    audit_get_loginuid(tsk));
407 	case AUDIT_COMPARE_UID_TO_EUID:
408 		return audit_uid_comparator(cred->uid, f->op, cred->euid);
409 	case AUDIT_COMPARE_UID_TO_SUID:
410 		return audit_uid_comparator(cred->uid, f->op, cred->suid);
411 	case AUDIT_COMPARE_UID_TO_FSUID:
412 		return audit_uid_comparator(cred->uid, f->op, cred->fsuid);
413 	/* auid comparisons */
414 	case AUDIT_COMPARE_AUID_TO_EUID:
415 		return audit_uid_comparator(audit_get_loginuid(tsk), f->op,
416 					    cred->euid);
417 	case AUDIT_COMPARE_AUID_TO_SUID:
418 		return audit_uid_comparator(audit_get_loginuid(tsk), f->op,
419 					    cred->suid);
420 	case AUDIT_COMPARE_AUID_TO_FSUID:
421 		return audit_uid_comparator(audit_get_loginuid(tsk), f->op,
422 					    cred->fsuid);
423 	/* euid comparisons */
424 	case AUDIT_COMPARE_EUID_TO_SUID:
425 		return audit_uid_comparator(cred->euid, f->op, cred->suid);
426 	case AUDIT_COMPARE_EUID_TO_FSUID:
427 		return audit_uid_comparator(cred->euid, f->op, cred->fsuid);
428 	/* suid comparisons */
429 	case AUDIT_COMPARE_SUID_TO_FSUID:
430 		return audit_uid_comparator(cred->suid, f->op, cred->fsuid);
431 	/* gid comparisons */
432 	case AUDIT_COMPARE_GID_TO_EGID:
433 		return audit_gid_comparator(cred->gid, f->op, cred->egid);
434 	case AUDIT_COMPARE_GID_TO_SGID:
435 		return audit_gid_comparator(cred->gid, f->op, cred->sgid);
436 	case AUDIT_COMPARE_GID_TO_FSGID:
437 		return audit_gid_comparator(cred->gid, f->op, cred->fsgid);
438 	/* egid comparisons */
439 	case AUDIT_COMPARE_EGID_TO_SGID:
440 		return audit_gid_comparator(cred->egid, f->op, cred->sgid);
441 	case AUDIT_COMPARE_EGID_TO_FSGID:
442 		return audit_gid_comparator(cred->egid, f->op, cred->fsgid);
443 	/* sgid comparison */
444 	case AUDIT_COMPARE_SGID_TO_FSGID:
445 		return audit_gid_comparator(cred->sgid, f->op, cred->fsgid);
446 	default:
447 		WARN(1, "Missing AUDIT_COMPARE define.  Report as a bug\n");
448 		return 0;
449 	}
450 	return 0;
451 }
452 
453 /* Determine if any context name data matches a rule's watch data */
454 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
455  * otherwise.
456  *
457  * If task_creation is true, this is an explicit indication that we are
458  * filtering a task rule at task creation time.  This and tsk == current are
459  * the only situations where tsk->cred may be accessed without an rcu read lock.
460  */
461 static int audit_filter_rules(struct task_struct *tsk,
462 			      struct audit_krule *rule,
463 			      struct audit_context *ctx,
464 			      struct audit_names *name,
465 			      enum audit_state *state,
466 			      bool task_creation)
467 {
468 	const struct cred *cred;
469 	int i, need_sid = 1;
470 	u32 sid;
471 	unsigned int sessionid;
472 
473 	if (ctx && rule->prio <= ctx->prio)
474 		return 0;
475 
476 	cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
477 
478 	for (i = 0; i < rule->field_count; i++) {
479 		struct audit_field *f = &rule->fields[i];
480 		struct audit_names *n;
481 		int result = 0;
482 		pid_t pid;
483 
484 		switch (f->type) {
485 		case AUDIT_PID:
486 			pid = task_tgid_nr(tsk);
487 			result = audit_comparator(pid, f->op, f->val);
488 			break;
489 		case AUDIT_PPID:
490 			if (ctx) {
491 				if (!ctx->ppid)
492 					ctx->ppid = task_ppid_nr(tsk);
493 				result = audit_comparator(ctx->ppid, f->op, f->val);
494 			}
495 			break;
496 		case AUDIT_EXE:
497 			result = audit_exe_compare(tsk, rule->exe);
498 			if (f->op == Audit_not_equal)
499 				result = !result;
500 			break;
501 		case AUDIT_UID:
502 			result = audit_uid_comparator(cred->uid, f->op, f->uid);
503 			break;
504 		case AUDIT_EUID:
505 			result = audit_uid_comparator(cred->euid, f->op, f->uid);
506 			break;
507 		case AUDIT_SUID:
508 			result = audit_uid_comparator(cred->suid, f->op, f->uid);
509 			break;
510 		case AUDIT_FSUID:
511 			result = audit_uid_comparator(cred->fsuid, f->op, f->uid);
512 			break;
513 		case AUDIT_GID:
514 			result = audit_gid_comparator(cred->gid, f->op, f->gid);
515 			if (f->op == Audit_equal) {
516 				if (!result)
517 					result = groups_search(cred->group_info, f->gid);
518 			} else if (f->op == Audit_not_equal) {
519 				if (result)
520 					result = !groups_search(cred->group_info, f->gid);
521 			}
522 			break;
523 		case AUDIT_EGID:
524 			result = audit_gid_comparator(cred->egid, f->op, f->gid);
525 			if (f->op == Audit_equal) {
526 				if (!result)
527 					result = groups_search(cred->group_info, f->gid);
528 			} else if (f->op == Audit_not_equal) {
529 				if (result)
530 					result = !groups_search(cred->group_info, f->gid);
531 			}
532 			break;
533 		case AUDIT_SGID:
534 			result = audit_gid_comparator(cred->sgid, f->op, f->gid);
535 			break;
536 		case AUDIT_FSGID:
537 			result = audit_gid_comparator(cred->fsgid, f->op, f->gid);
538 			break;
539 		case AUDIT_SESSIONID:
540 			sessionid = audit_get_sessionid(tsk);
541 			result = audit_comparator(sessionid, f->op, f->val);
542 			break;
543 		case AUDIT_PERS:
544 			result = audit_comparator(tsk->personality, f->op, f->val);
545 			break;
546 		case AUDIT_ARCH:
547 			if (ctx)
548 				result = audit_comparator(ctx->arch, f->op, f->val);
549 			break;
550 
551 		case AUDIT_EXIT:
552 			if (ctx && ctx->return_valid != AUDITSC_INVALID)
553 				result = audit_comparator(ctx->return_code, f->op, f->val);
554 			break;
555 		case AUDIT_SUCCESS:
556 			if (ctx && ctx->return_valid != AUDITSC_INVALID) {
557 				if (f->val)
558 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
559 				else
560 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
561 			}
562 			break;
563 		case AUDIT_DEVMAJOR:
564 			if (name) {
565 				if (audit_comparator(MAJOR(name->dev), f->op, f->val) ||
566 				    audit_comparator(MAJOR(name->rdev), f->op, f->val))
567 					++result;
568 			} else if (ctx) {
569 				list_for_each_entry(n, &ctx->names_list, list) {
570 					if (audit_comparator(MAJOR(n->dev), f->op, f->val) ||
571 					    audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
572 						++result;
573 						break;
574 					}
575 				}
576 			}
577 			break;
578 		case AUDIT_DEVMINOR:
579 			if (name) {
580 				if (audit_comparator(MINOR(name->dev), f->op, f->val) ||
581 				    audit_comparator(MINOR(name->rdev), f->op, f->val))
582 					++result;
583 			} else if (ctx) {
584 				list_for_each_entry(n, &ctx->names_list, list) {
585 					if (audit_comparator(MINOR(n->dev), f->op, f->val) ||
586 					    audit_comparator(MINOR(n->rdev), f->op, f->val)) {
587 						++result;
588 						break;
589 					}
590 				}
591 			}
592 			break;
593 		case AUDIT_INODE:
594 			if (name)
595 				result = audit_comparator(name->ino, f->op, f->val);
596 			else if (ctx) {
597 				list_for_each_entry(n, &ctx->names_list, list) {
598 					if (audit_comparator(n->ino, f->op, f->val)) {
599 						++result;
600 						break;
601 					}
602 				}
603 			}
604 			break;
605 		case AUDIT_OBJ_UID:
606 			if (name) {
607 				result = audit_uid_comparator(name->uid, f->op, f->uid);
608 			} else if (ctx) {
609 				list_for_each_entry(n, &ctx->names_list, list) {
610 					if (audit_uid_comparator(n->uid, f->op, f->uid)) {
611 						++result;
612 						break;
613 					}
614 				}
615 			}
616 			break;
617 		case AUDIT_OBJ_GID:
618 			if (name) {
619 				result = audit_gid_comparator(name->gid, f->op, f->gid);
620 			} else if (ctx) {
621 				list_for_each_entry(n, &ctx->names_list, list) {
622 					if (audit_gid_comparator(n->gid, f->op, f->gid)) {
623 						++result;
624 						break;
625 					}
626 				}
627 			}
628 			break;
629 		case AUDIT_WATCH:
630 			if (name) {
631 				result = audit_watch_compare(rule->watch,
632 							     name->ino,
633 							     name->dev);
634 				if (f->op == Audit_not_equal)
635 					result = !result;
636 			}
637 			break;
638 		case AUDIT_DIR:
639 			if (ctx) {
640 				result = match_tree_refs(ctx, rule->tree);
641 				if (f->op == Audit_not_equal)
642 					result = !result;
643 			}
644 			break;
645 		case AUDIT_LOGINUID:
646 			result = audit_uid_comparator(audit_get_loginuid(tsk),
647 						      f->op, f->uid);
648 			break;
649 		case AUDIT_LOGINUID_SET:
650 			result = audit_comparator(audit_loginuid_set(tsk), f->op, f->val);
651 			break;
652 		case AUDIT_SADDR_FAM:
653 			if (ctx && ctx->sockaddr)
654 				result = audit_comparator(ctx->sockaddr->ss_family,
655 							  f->op, f->val);
656 			break;
657 		case AUDIT_SUBJ_USER:
658 		case AUDIT_SUBJ_ROLE:
659 		case AUDIT_SUBJ_TYPE:
660 		case AUDIT_SUBJ_SEN:
661 		case AUDIT_SUBJ_CLR:
662 			/* NOTE: this may return negative values indicating
663 			   a temporary error.  We simply treat this as a
664 			   match for now to avoid losing information that
665 			   may be wanted.   An error message will also be
666 			   logged upon error */
667 			if (f->lsm_rule) {
668 				if (need_sid) {
669 					/* @tsk should always be equal to
670 					 * @current with the exception of
671 					 * fork()/copy_process() in which case
672 					 * the new @tsk creds are still a dup
673 					 * of @current's creds so we can still
674 					 * use security_current_getsecid_subj()
675 					 * here even though it always refs
676 					 * @current's creds
677 					 */
678 					security_current_getsecid_subj(&sid);
679 					need_sid = 0;
680 				}
681 				result = security_audit_rule_match(sid, f->type,
682 								   f->op,
683 								   f->lsm_rule);
684 			}
685 			break;
686 		case AUDIT_OBJ_USER:
687 		case AUDIT_OBJ_ROLE:
688 		case AUDIT_OBJ_TYPE:
689 		case AUDIT_OBJ_LEV_LOW:
690 		case AUDIT_OBJ_LEV_HIGH:
691 			/* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
692 			   also applies here */
693 			if (f->lsm_rule) {
694 				/* Find files that match */
695 				if (name) {
696 					result = security_audit_rule_match(
697 								name->osid,
698 								f->type,
699 								f->op,
700 								f->lsm_rule);
701 				} else if (ctx) {
702 					list_for_each_entry(n, &ctx->names_list, list) {
703 						if (security_audit_rule_match(
704 								n->osid,
705 								f->type,
706 								f->op,
707 								f->lsm_rule)) {
708 							++result;
709 							break;
710 						}
711 					}
712 				}
713 				/* Find ipc objects that match */
714 				if (!ctx || ctx->type != AUDIT_IPC)
715 					break;
716 				if (security_audit_rule_match(ctx->ipc.osid,
717 							      f->type, f->op,
718 							      f->lsm_rule))
719 					++result;
720 			}
721 			break;
722 		case AUDIT_ARG0:
723 		case AUDIT_ARG1:
724 		case AUDIT_ARG2:
725 		case AUDIT_ARG3:
726 			if (ctx)
727 				result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
728 			break;
729 		case AUDIT_FILTERKEY:
730 			/* ignore this field for filtering */
731 			result = 1;
732 			break;
733 		case AUDIT_PERM:
734 			result = audit_match_perm(ctx, f->val);
735 			if (f->op == Audit_not_equal)
736 				result = !result;
737 			break;
738 		case AUDIT_FILETYPE:
739 			result = audit_match_filetype(ctx, f->val);
740 			if (f->op == Audit_not_equal)
741 				result = !result;
742 			break;
743 		case AUDIT_FIELD_COMPARE:
744 			result = audit_field_compare(tsk, cred, f, ctx, name);
745 			break;
746 		}
747 		if (!result)
748 			return 0;
749 	}
750 
751 	if (ctx) {
752 		if (rule->filterkey) {
753 			kfree(ctx->filterkey);
754 			ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
755 		}
756 		ctx->prio = rule->prio;
757 	}
758 	switch (rule->action) {
759 	case AUDIT_NEVER:
760 		*state = AUDIT_STATE_DISABLED;
761 		break;
762 	case AUDIT_ALWAYS:
763 		*state = AUDIT_STATE_RECORD;
764 		break;
765 	}
766 	return 1;
767 }
768 
769 /* At process creation time, we can determine if system-call auditing is
770  * completely disabled for this task.  Since we only have the task
771  * structure at this point, we can only check uid and gid.
772  */
773 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
774 {
775 	struct audit_entry *e;
776 	enum audit_state   state;
777 
778 	rcu_read_lock();
779 	list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
780 		if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
781 				       &state, true)) {
782 			if (state == AUDIT_STATE_RECORD)
783 				*key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
784 			rcu_read_unlock();
785 			return state;
786 		}
787 	}
788 	rcu_read_unlock();
789 	return AUDIT_STATE_BUILD;
790 }
791 
792 static int audit_in_mask(const struct audit_krule *rule, unsigned long val)
793 {
794 	int word, bit;
795 
796 	if (val > 0xffffffff)
797 		return false;
798 
799 	word = AUDIT_WORD(val);
800 	if (word >= AUDIT_BITMASK_SIZE)
801 		return false;
802 
803 	bit = AUDIT_BIT(val);
804 
805 	return rule->mask[word] & bit;
806 }
807 
808 /**
809  * __audit_filter_op - common filter helper for operations (syscall/uring/etc)
810  * @tsk: associated task
811  * @ctx: audit context
812  * @list: audit filter list
813  * @name: audit_name (can be NULL)
814  * @op: current syscall/uring_op
815  *
816  * Run the udit filters specified in @list against @tsk using @ctx,
817  * @name, and @op, as necessary; the caller is responsible for ensuring
818  * that the call is made while the RCU read lock is held. The @name
819  * parameter can be NULL, but all others must be specified.
820  * Returns 1/true if the filter finds a match, 0/false if none are found.
821  */
822 static int __audit_filter_op(struct task_struct *tsk,
823 			   struct audit_context *ctx,
824 			   struct list_head *list,
825 			   struct audit_names *name,
826 			   unsigned long op)
827 {
828 	struct audit_entry *e;
829 	enum audit_state state;
830 
831 	list_for_each_entry_rcu(e, list, list) {
832 		if (audit_in_mask(&e->rule, op) &&
833 		    audit_filter_rules(tsk, &e->rule, ctx, name,
834 				       &state, false)) {
835 			ctx->current_state = state;
836 			return 1;
837 		}
838 	}
839 	return 0;
840 }
841 
842 /**
843  * audit_filter_uring - apply filters to an io_uring operation
844  * @tsk: associated task
845  * @ctx: audit context
846  */
847 static void audit_filter_uring(struct task_struct *tsk,
848 			       struct audit_context *ctx)
849 {
850 	if (auditd_test_task(tsk))
851 		return;
852 
853 	rcu_read_lock();
854 	__audit_filter_op(tsk, ctx, &audit_filter_list[AUDIT_FILTER_URING_EXIT],
855 			NULL, ctx->uring_op);
856 	rcu_read_unlock();
857 }
858 
859 /* At syscall exit time, this filter is called if the audit_state is
860  * not low enough that auditing cannot take place, but is also not
861  * high enough that we already know we have to write an audit record
862  * (i.e., the state is AUDIT_STATE_BUILD).
863  */
864 static void audit_filter_syscall(struct task_struct *tsk,
865 				 struct audit_context *ctx)
866 {
867 	if (auditd_test_task(tsk))
868 		return;
869 
870 	rcu_read_lock();
871 	__audit_filter_op(tsk, ctx, &audit_filter_list[AUDIT_FILTER_EXIT],
872 			NULL, ctx->major);
873 	rcu_read_unlock();
874 }
875 
876 /*
877  * Given an audit_name check the inode hash table to see if they match.
878  * Called holding the rcu read lock to protect the use of audit_inode_hash
879  */
880 static int audit_filter_inode_name(struct task_struct *tsk,
881 				   struct audit_names *n,
882 				   struct audit_context *ctx) {
883 	int h = audit_hash_ino((u32)n->ino);
884 	struct list_head *list = &audit_inode_hash[h];
885 
886 	return __audit_filter_op(tsk, ctx, list, n, ctx->major);
887 }
888 
889 /* At syscall exit time, this filter is called if any audit_names have been
890  * collected during syscall processing.  We only check rules in sublists at hash
891  * buckets applicable to the inode numbers in audit_names.
892  * Regarding audit_state, same rules apply as for audit_filter_syscall().
893  */
894 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
895 {
896 	struct audit_names *n;
897 
898 	if (auditd_test_task(tsk))
899 		return;
900 
901 	rcu_read_lock();
902 
903 	list_for_each_entry(n, &ctx->names_list, list) {
904 		if (audit_filter_inode_name(tsk, n, ctx))
905 			break;
906 	}
907 	rcu_read_unlock();
908 }
909 
910 static inline void audit_proctitle_free(struct audit_context *context)
911 {
912 	kfree(context->proctitle.value);
913 	context->proctitle.value = NULL;
914 	context->proctitle.len = 0;
915 }
916 
917 static inline void audit_free_module(struct audit_context *context)
918 {
919 	if (context->type == AUDIT_KERN_MODULE) {
920 		kfree(context->module.name);
921 		context->module.name = NULL;
922 	}
923 }
924 static inline void audit_free_names(struct audit_context *context)
925 {
926 	struct audit_names *n, *next;
927 
928 	list_for_each_entry_safe(n, next, &context->names_list, list) {
929 		list_del(&n->list);
930 		if (n->name)
931 			putname(n->name);
932 		if (n->should_free)
933 			kfree(n);
934 	}
935 	context->name_count = 0;
936 	path_put(&context->pwd);
937 	context->pwd.dentry = NULL;
938 	context->pwd.mnt = NULL;
939 }
940 
941 static inline void audit_free_aux(struct audit_context *context)
942 {
943 	struct audit_aux_data *aux;
944 
945 	while ((aux = context->aux)) {
946 		context->aux = aux->next;
947 		kfree(aux);
948 	}
949 	context->aux = NULL;
950 	while ((aux = context->aux_pids)) {
951 		context->aux_pids = aux->next;
952 		kfree(aux);
953 	}
954 	context->aux_pids = NULL;
955 }
956 
957 /**
958  * audit_reset_context - reset a audit_context structure
959  * @ctx: the audit_context to reset
960  *
961  * All fields in the audit_context will be reset to an initial state, all
962  * references held by fields will be dropped, and private memory will be
963  * released.  When this function returns the audit_context will be suitable
964  * for reuse, so long as the passed context is not NULL or a dummy context.
965  */
966 static void audit_reset_context(struct audit_context *ctx)
967 {
968 	if (!ctx)
969 		return;
970 
971 	/* if ctx is non-null, reset the "ctx->context" regardless */
972 	ctx->context = AUDIT_CTX_UNUSED;
973 	if (ctx->dummy)
974 		return;
975 
976 	/*
977 	 * NOTE: It shouldn't matter in what order we release the fields, so
978 	 *       release them in the order in which they appear in the struct;
979 	 *       this gives us some hope of quickly making sure we are
980 	 *       resetting the audit_context properly.
981 	 *
982 	 *       Other things worth mentioning:
983 	 *       - we don't reset "dummy"
984 	 *       - we don't reset "state", we do reset "current_state"
985 	 *       - we preserve "filterkey" if "state" is AUDIT_STATE_RECORD
986 	 *       - much of this is likely overkill, but play it safe for now
987 	 *       - we really need to work on improving the audit_context struct
988 	 */
989 
990 	ctx->current_state = ctx->state;
991 	ctx->serial = 0;
992 	ctx->major = 0;
993 	ctx->uring_op = 0;
994 	ctx->ctime = (struct timespec64){ .tv_sec = 0, .tv_nsec = 0 };
995 	memset(ctx->argv, 0, sizeof(ctx->argv));
996 	ctx->return_code = 0;
997 	ctx->prio = (ctx->state == AUDIT_STATE_RECORD ? ~0ULL : 0);
998 	ctx->return_valid = AUDITSC_INVALID;
999 	audit_free_names(ctx);
1000 	if (ctx->state != AUDIT_STATE_RECORD) {
1001 		kfree(ctx->filterkey);
1002 		ctx->filterkey = NULL;
1003 	}
1004 	audit_free_aux(ctx);
1005 	kfree(ctx->sockaddr);
1006 	ctx->sockaddr = NULL;
1007 	ctx->sockaddr_len = 0;
1008 	ctx->ppid = 0;
1009 	ctx->uid = ctx->euid = ctx->suid = ctx->fsuid = KUIDT_INIT(0);
1010 	ctx->gid = ctx->egid = ctx->sgid = ctx->fsgid = KGIDT_INIT(0);
1011 	ctx->personality = 0;
1012 	ctx->arch = 0;
1013 	ctx->target_pid = 0;
1014 	ctx->target_auid = ctx->target_uid = KUIDT_INIT(0);
1015 	ctx->target_sessionid = 0;
1016 	ctx->target_sid = 0;
1017 	ctx->target_comm[0] = '\0';
1018 	unroll_tree_refs(ctx, NULL, 0);
1019 	WARN_ON(!list_empty(&ctx->killed_trees));
1020 	audit_free_module(ctx);
1021 	ctx->fds[0] = -1;
1022 	ctx->type = 0; /* reset last for audit_free_*() */
1023 }
1024 
1025 static inline struct audit_context *audit_alloc_context(enum audit_state state)
1026 {
1027 	struct audit_context *context;
1028 
1029 	context = kzalloc(sizeof(*context), GFP_KERNEL);
1030 	if (!context)
1031 		return NULL;
1032 	context->context = AUDIT_CTX_UNUSED;
1033 	context->state = state;
1034 	context->prio = state == AUDIT_STATE_RECORD ? ~0ULL : 0;
1035 	INIT_LIST_HEAD(&context->killed_trees);
1036 	INIT_LIST_HEAD(&context->names_list);
1037 	context->fds[0] = -1;
1038 	context->return_valid = AUDITSC_INVALID;
1039 	return context;
1040 }
1041 
1042 /**
1043  * audit_alloc - allocate an audit context block for a task
1044  * @tsk: task
1045  *
1046  * Filter on the task information and allocate a per-task audit context
1047  * if necessary.  Doing so turns on system call auditing for the
1048  * specified task.  This is called from copy_process, so no lock is
1049  * needed.
1050  */
1051 int audit_alloc(struct task_struct *tsk)
1052 {
1053 	struct audit_context *context;
1054 	enum audit_state     state;
1055 	char *key = NULL;
1056 
1057 	if (likely(!audit_ever_enabled))
1058 		return 0;
1059 
1060 	state = audit_filter_task(tsk, &key);
1061 	if (state == AUDIT_STATE_DISABLED) {
1062 		clear_task_syscall_work(tsk, SYSCALL_AUDIT);
1063 		return 0;
1064 	}
1065 
1066 	if (!(context = audit_alloc_context(state))) {
1067 		kfree(key);
1068 		audit_log_lost("out of memory in audit_alloc");
1069 		return -ENOMEM;
1070 	}
1071 	context->filterkey = key;
1072 
1073 	audit_set_context(tsk, context);
1074 	set_task_syscall_work(tsk, SYSCALL_AUDIT);
1075 	return 0;
1076 }
1077 
1078 static inline void audit_free_context(struct audit_context *context)
1079 {
1080 	/* resetting is extra work, but it is likely just noise */
1081 	audit_reset_context(context);
1082 	audit_proctitle_free(context);
1083 	free_tree_refs(context);
1084 	kfree(context->filterkey);
1085 	kfree(context);
1086 }
1087 
1088 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
1089 				 kuid_t auid, kuid_t uid, unsigned int sessionid,
1090 				 u32 sid, char *comm)
1091 {
1092 	struct audit_buffer *ab;
1093 	char *ctx = NULL;
1094 	u32 len;
1095 	int rc = 0;
1096 
1097 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1098 	if (!ab)
1099 		return rc;
1100 
1101 	audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid,
1102 			 from_kuid(&init_user_ns, auid),
1103 			 from_kuid(&init_user_ns, uid), sessionid);
1104 	if (sid) {
1105 		if (security_secid_to_secctx(sid, &ctx, &len)) {
1106 			audit_log_format(ab, " obj=(none)");
1107 			rc = 1;
1108 		} else {
1109 			audit_log_format(ab, " obj=%s", ctx);
1110 			security_release_secctx(ctx, len);
1111 		}
1112 	}
1113 	audit_log_format(ab, " ocomm=");
1114 	audit_log_untrustedstring(ab, comm);
1115 	audit_log_end(ab);
1116 
1117 	return rc;
1118 }
1119 
1120 static void audit_log_execve_info(struct audit_context *context,
1121 				  struct audit_buffer **ab)
1122 {
1123 	long len_max;
1124 	long len_rem;
1125 	long len_full;
1126 	long len_buf;
1127 	long len_abuf = 0;
1128 	long len_tmp;
1129 	bool require_data;
1130 	bool encode;
1131 	unsigned int iter;
1132 	unsigned int arg;
1133 	char *buf_head;
1134 	char *buf;
1135 	const char __user *p = (const char __user *)current->mm->arg_start;
1136 
1137 	/* NOTE: this buffer needs to be large enough to hold all the non-arg
1138 	 *       data we put in the audit record for this argument (see the
1139 	 *       code below) ... at this point in time 96 is plenty */
1140 	char abuf[96];
1141 
1142 	/* NOTE: we set MAX_EXECVE_AUDIT_LEN to a rather arbitrary limit, the
1143 	 *       current value of 7500 is not as important as the fact that it
1144 	 *       is less than 8k, a setting of 7500 gives us plenty of wiggle
1145 	 *       room if we go over a little bit in the logging below */
1146 	WARN_ON_ONCE(MAX_EXECVE_AUDIT_LEN > 7500);
1147 	len_max = MAX_EXECVE_AUDIT_LEN;
1148 
1149 	/* scratch buffer to hold the userspace args */
1150 	buf_head = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1151 	if (!buf_head) {
1152 		audit_panic("out of memory for argv string");
1153 		return;
1154 	}
1155 	buf = buf_head;
1156 
1157 	audit_log_format(*ab, "argc=%d", context->execve.argc);
1158 
1159 	len_rem = len_max;
1160 	len_buf = 0;
1161 	len_full = 0;
1162 	require_data = true;
1163 	encode = false;
1164 	iter = 0;
1165 	arg = 0;
1166 	do {
1167 		/* NOTE: we don't ever want to trust this value for anything
1168 		 *       serious, but the audit record format insists we
1169 		 *       provide an argument length for really long arguments,
1170 		 *       e.g. > MAX_EXECVE_AUDIT_LEN, so we have no choice but
1171 		 *       to use strncpy_from_user() to obtain this value for
1172 		 *       recording in the log, although we don't use it
1173 		 *       anywhere here to avoid a double-fetch problem */
1174 		if (len_full == 0)
1175 			len_full = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1176 
1177 		/* read more data from userspace */
1178 		if (require_data) {
1179 			/* can we make more room in the buffer? */
1180 			if (buf != buf_head) {
1181 				memmove(buf_head, buf, len_buf);
1182 				buf = buf_head;
1183 			}
1184 
1185 			/* fetch as much as we can of the argument */
1186 			len_tmp = strncpy_from_user(&buf_head[len_buf], p,
1187 						    len_max - len_buf);
1188 			if (len_tmp == -EFAULT) {
1189 				/* unable to copy from userspace */
1190 				send_sig(SIGKILL, current, 0);
1191 				goto out;
1192 			} else if (len_tmp == (len_max - len_buf)) {
1193 				/* buffer is not large enough */
1194 				require_data = true;
1195 				/* NOTE: if we are going to span multiple
1196 				 *       buffers force the encoding so we stand
1197 				 *       a chance at a sane len_full value and
1198 				 *       consistent record encoding */
1199 				encode = true;
1200 				len_full = len_full * 2;
1201 				p += len_tmp;
1202 			} else {
1203 				require_data = false;
1204 				if (!encode)
1205 					encode = audit_string_contains_control(
1206 								buf, len_tmp);
1207 				/* try to use a trusted value for len_full */
1208 				if (len_full < len_max)
1209 					len_full = (encode ?
1210 						    len_tmp * 2 : len_tmp);
1211 				p += len_tmp + 1;
1212 			}
1213 			len_buf += len_tmp;
1214 			buf_head[len_buf] = '\0';
1215 
1216 			/* length of the buffer in the audit record? */
1217 			len_abuf = (encode ? len_buf * 2 : len_buf + 2);
1218 		}
1219 
1220 		/* write as much as we can to the audit log */
1221 		if (len_buf >= 0) {
1222 			/* NOTE: some magic numbers here - basically if we
1223 			 *       can't fit a reasonable amount of data into the
1224 			 *       existing audit buffer, flush it and start with
1225 			 *       a new buffer */
1226 			if ((sizeof(abuf) + 8) > len_rem) {
1227 				len_rem = len_max;
1228 				audit_log_end(*ab);
1229 				*ab = audit_log_start(context,
1230 						      GFP_KERNEL, AUDIT_EXECVE);
1231 				if (!*ab)
1232 					goto out;
1233 			}
1234 
1235 			/* create the non-arg portion of the arg record */
1236 			len_tmp = 0;
1237 			if (require_data || (iter > 0) ||
1238 			    ((len_abuf + sizeof(abuf)) > len_rem)) {
1239 				if (iter == 0) {
1240 					len_tmp += snprintf(&abuf[len_tmp],
1241 							sizeof(abuf) - len_tmp,
1242 							" a%d_len=%lu",
1243 							arg, len_full);
1244 				}
1245 				len_tmp += snprintf(&abuf[len_tmp],
1246 						    sizeof(abuf) - len_tmp,
1247 						    " a%d[%d]=", arg, iter++);
1248 			} else
1249 				len_tmp += snprintf(&abuf[len_tmp],
1250 						    sizeof(abuf) - len_tmp,
1251 						    " a%d=", arg);
1252 			WARN_ON(len_tmp >= sizeof(abuf));
1253 			abuf[sizeof(abuf) - 1] = '\0';
1254 
1255 			/* log the arg in the audit record */
1256 			audit_log_format(*ab, "%s", abuf);
1257 			len_rem -= len_tmp;
1258 			len_tmp = len_buf;
1259 			if (encode) {
1260 				if (len_abuf > len_rem)
1261 					len_tmp = len_rem / 2; /* encoding */
1262 				audit_log_n_hex(*ab, buf, len_tmp);
1263 				len_rem -= len_tmp * 2;
1264 				len_abuf -= len_tmp * 2;
1265 			} else {
1266 				if (len_abuf > len_rem)
1267 					len_tmp = len_rem - 2; /* quotes */
1268 				audit_log_n_string(*ab, buf, len_tmp);
1269 				len_rem -= len_tmp + 2;
1270 				/* don't subtract the "2" because we still need
1271 				 * to add quotes to the remaining string */
1272 				len_abuf -= len_tmp;
1273 			}
1274 			len_buf -= len_tmp;
1275 			buf += len_tmp;
1276 		}
1277 
1278 		/* ready to move to the next argument? */
1279 		if ((len_buf == 0) && !require_data) {
1280 			arg++;
1281 			iter = 0;
1282 			len_full = 0;
1283 			require_data = true;
1284 			encode = false;
1285 		}
1286 	} while (arg < context->execve.argc);
1287 
1288 	/* NOTE: the caller handles the final audit_log_end() call */
1289 
1290 out:
1291 	kfree(buf_head);
1292 }
1293 
1294 static void audit_log_cap(struct audit_buffer *ab, char *prefix,
1295 			  kernel_cap_t *cap)
1296 {
1297 	int i;
1298 
1299 	if (cap_isclear(*cap)) {
1300 		audit_log_format(ab, " %s=0", prefix);
1301 		return;
1302 	}
1303 	audit_log_format(ab, " %s=", prefix);
1304 	CAP_FOR_EACH_U32(i)
1305 		audit_log_format(ab, "%08x", cap->cap[CAP_LAST_U32 - i]);
1306 }
1307 
1308 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1309 {
1310 	if (name->fcap_ver == -1) {
1311 		audit_log_format(ab, " cap_fe=? cap_fver=? cap_fp=? cap_fi=?");
1312 		return;
1313 	}
1314 	audit_log_cap(ab, "cap_fp", &name->fcap.permitted);
1315 	audit_log_cap(ab, "cap_fi", &name->fcap.inheritable);
1316 	audit_log_format(ab, " cap_fe=%d cap_fver=%x cap_frootid=%d",
1317 			 name->fcap.fE, name->fcap_ver,
1318 			 from_kuid(&init_user_ns, name->fcap.rootid));
1319 }
1320 
1321 static void audit_log_time(struct audit_context *context, struct audit_buffer **ab)
1322 {
1323 	const struct audit_ntp_data *ntp = &context->time.ntp_data;
1324 	const struct timespec64 *tk = &context->time.tk_injoffset;
1325 	static const char * const ntp_name[] = {
1326 		"offset",
1327 		"freq",
1328 		"status",
1329 		"tai",
1330 		"tick",
1331 		"adjust",
1332 	};
1333 	int type;
1334 
1335 	if (context->type == AUDIT_TIME_ADJNTPVAL) {
1336 		for (type = 0; type < AUDIT_NTP_NVALS; type++) {
1337 			if (ntp->vals[type].newval != ntp->vals[type].oldval) {
1338 				if (!*ab) {
1339 					*ab = audit_log_start(context,
1340 							GFP_KERNEL,
1341 							AUDIT_TIME_ADJNTPVAL);
1342 					if (!*ab)
1343 						return;
1344 				}
1345 				audit_log_format(*ab, "op=%s old=%lli new=%lli",
1346 						 ntp_name[type],
1347 						 ntp->vals[type].oldval,
1348 						 ntp->vals[type].newval);
1349 				audit_log_end(*ab);
1350 				*ab = NULL;
1351 			}
1352 		}
1353 	}
1354 	if (tk->tv_sec != 0 || tk->tv_nsec != 0) {
1355 		if (!*ab) {
1356 			*ab = audit_log_start(context, GFP_KERNEL,
1357 					      AUDIT_TIME_INJOFFSET);
1358 			if (!*ab)
1359 				return;
1360 		}
1361 		audit_log_format(*ab, "sec=%lli nsec=%li",
1362 				 (long long)tk->tv_sec, tk->tv_nsec);
1363 		audit_log_end(*ab);
1364 		*ab = NULL;
1365 	}
1366 }
1367 
1368 static void show_special(struct audit_context *context, int *call_panic)
1369 {
1370 	struct audit_buffer *ab;
1371 	int i;
1372 
1373 	ab = audit_log_start(context, GFP_KERNEL, context->type);
1374 	if (!ab)
1375 		return;
1376 
1377 	switch (context->type) {
1378 	case AUDIT_SOCKETCALL: {
1379 		int nargs = context->socketcall.nargs;
1380 
1381 		audit_log_format(ab, "nargs=%d", nargs);
1382 		for (i = 0; i < nargs; i++)
1383 			audit_log_format(ab, " a%d=%lx", i,
1384 				context->socketcall.args[i]);
1385 		break; }
1386 	case AUDIT_IPC: {
1387 		u32 osid = context->ipc.osid;
1388 
1389 		audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1390 				 from_kuid(&init_user_ns, context->ipc.uid),
1391 				 from_kgid(&init_user_ns, context->ipc.gid),
1392 				 context->ipc.mode);
1393 		if (osid) {
1394 			char *ctx = NULL;
1395 			u32 len;
1396 
1397 			if (security_secid_to_secctx(osid, &ctx, &len)) {
1398 				audit_log_format(ab, " osid=%u", osid);
1399 				*call_panic = 1;
1400 			} else {
1401 				audit_log_format(ab, " obj=%s", ctx);
1402 				security_release_secctx(ctx, len);
1403 			}
1404 		}
1405 		if (context->ipc.has_perm) {
1406 			audit_log_end(ab);
1407 			ab = audit_log_start(context, GFP_KERNEL,
1408 					     AUDIT_IPC_SET_PERM);
1409 			if (unlikely(!ab))
1410 				return;
1411 			audit_log_format(ab,
1412 				"qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1413 				context->ipc.qbytes,
1414 				context->ipc.perm_uid,
1415 				context->ipc.perm_gid,
1416 				context->ipc.perm_mode);
1417 		}
1418 		break; }
1419 	case AUDIT_MQ_OPEN:
1420 		audit_log_format(ab,
1421 			"oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1422 			"mq_msgsize=%ld mq_curmsgs=%ld",
1423 			context->mq_open.oflag, context->mq_open.mode,
1424 			context->mq_open.attr.mq_flags,
1425 			context->mq_open.attr.mq_maxmsg,
1426 			context->mq_open.attr.mq_msgsize,
1427 			context->mq_open.attr.mq_curmsgs);
1428 		break;
1429 	case AUDIT_MQ_SENDRECV:
1430 		audit_log_format(ab,
1431 			"mqdes=%d msg_len=%zd msg_prio=%u "
1432 			"abs_timeout_sec=%lld abs_timeout_nsec=%ld",
1433 			context->mq_sendrecv.mqdes,
1434 			context->mq_sendrecv.msg_len,
1435 			context->mq_sendrecv.msg_prio,
1436 			(long long) context->mq_sendrecv.abs_timeout.tv_sec,
1437 			context->mq_sendrecv.abs_timeout.tv_nsec);
1438 		break;
1439 	case AUDIT_MQ_NOTIFY:
1440 		audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1441 				context->mq_notify.mqdes,
1442 				context->mq_notify.sigev_signo);
1443 		break;
1444 	case AUDIT_MQ_GETSETATTR: {
1445 		struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1446 
1447 		audit_log_format(ab,
1448 			"mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1449 			"mq_curmsgs=%ld ",
1450 			context->mq_getsetattr.mqdes,
1451 			attr->mq_flags, attr->mq_maxmsg,
1452 			attr->mq_msgsize, attr->mq_curmsgs);
1453 		break; }
1454 	case AUDIT_CAPSET:
1455 		audit_log_format(ab, "pid=%d", context->capset.pid);
1456 		audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1457 		audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1458 		audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1459 		audit_log_cap(ab, "cap_pa", &context->capset.cap.ambient);
1460 		break;
1461 	case AUDIT_MMAP:
1462 		audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1463 				 context->mmap.flags);
1464 		break;
1465 	case AUDIT_OPENAT2:
1466 		audit_log_format(ab, "oflag=0%llo mode=0%llo resolve=0x%llx",
1467 				 context->openat2.flags,
1468 				 context->openat2.mode,
1469 				 context->openat2.resolve);
1470 		break;
1471 	case AUDIT_EXECVE:
1472 		audit_log_execve_info(context, &ab);
1473 		break;
1474 	case AUDIT_KERN_MODULE:
1475 		audit_log_format(ab, "name=");
1476 		if (context->module.name) {
1477 			audit_log_untrustedstring(ab, context->module.name);
1478 		} else
1479 			audit_log_format(ab, "(null)");
1480 
1481 		break;
1482 	case AUDIT_TIME_ADJNTPVAL:
1483 	case AUDIT_TIME_INJOFFSET:
1484 		/* this call deviates from the rest, eating the buffer */
1485 		audit_log_time(context, &ab);
1486 		break;
1487 	}
1488 	audit_log_end(ab);
1489 }
1490 
1491 static inline int audit_proctitle_rtrim(char *proctitle, int len)
1492 {
1493 	char *end = proctitle + len - 1;
1494 
1495 	while (end > proctitle && !isprint(*end))
1496 		end--;
1497 
1498 	/* catch the case where proctitle is only 1 non-print character */
1499 	len = end - proctitle + 1;
1500 	len -= isprint(proctitle[len-1]) == 0;
1501 	return len;
1502 }
1503 
1504 /*
1505  * audit_log_name - produce AUDIT_PATH record from struct audit_names
1506  * @context: audit_context for the task
1507  * @n: audit_names structure with reportable details
1508  * @path: optional path to report instead of audit_names->name
1509  * @record_num: record number to report when handling a list of names
1510  * @call_panic: optional pointer to int that will be updated if secid fails
1511  */
1512 static void audit_log_name(struct audit_context *context, struct audit_names *n,
1513 		    const struct path *path, int record_num, int *call_panic)
1514 {
1515 	struct audit_buffer *ab;
1516 
1517 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1518 	if (!ab)
1519 		return;
1520 
1521 	audit_log_format(ab, "item=%d", record_num);
1522 
1523 	if (path)
1524 		audit_log_d_path(ab, " name=", path);
1525 	else if (n->name) {
1526 		switch (n->name_len) {
1527 		case AUDIT_NAME_FULL:
1528 			/* log the full path */
1529 			audit_log_format(ab, " name=");
1530 			audit_log_untrustedstring(ab, n->name->name);
1531 			break;
1532 		case 0:
1533 			/* name was specified as a relative path and the
1534 			 * directory component is the cwd
1535 			 */
1536 			if (context->pwd.dentry && context->pwd.mnt)
1537 				audit_log_d_path(ab, " name=", &context->pwd);
1538 			else
1539 				audit_log_format(ab, " name=(null)");
1540 			break;
1541 		default:
1542 			/* log the name's directory component */
1543 			audit_log_format(ab, " name=");
1544 			audit_log_n_untrustedstring(ab, n->name->name,
1545 						    n->name_len);
1546 		}
1547 	} else
1548 		audit_log_format(ab, " name=(null)");
1549 
1550 	if (n->ino != AUDIT_INO_UNSET)
1551 		audit_log_format(ab, " inode=%lu dev=%02x:%02x mode=%#ho ouid=%u ogid=%u rdev=%02x:%02x",
1552 				 n->ino,
1553 				 MAJOR(n->dev),
1554 				 MINOR(n->dev),
1555 				 n->mode,
1556 				 from_kuid(&init_user_ns, n->uid),
1557 				 from_kgid(&init_user_ns, n->gid),
1558 				 MAJOR(n->rdev),
1559 				 MINOR(n->rdev));
1560 	if (n->osid != 0) {
1561 		char *ctx = NULL;
1562 		u32 len;
1563 
1564 		if (security_secid_to_secctx(
1565 			n->osid, &ctx, &len)) {
1566 			audit_log_format(ab, " osid=%u", n->osid);
1567 			if (call_panic)
1568 				*call_panic = 2;
1569 		} else {
1570 			audit_log_format(ab, " obj=%s", ctx);
1571 			security_release_secctx(ctx, len);
1572 		}
1573 	}
1574 
1575 	/* log the audit_names record type */
1576 	switch (n->type) {
1577 	case AUDIT_TYPE_NORMAL:
1578 		audit_log_format(ab, " nametype=NORMAL");
1579 		break;
1580 	case AUDIT_TYPE_PARENT:
1581 		audit_log_format(ab, " nametype=PARENT");
1582 		break;
1583 	case AUDIT_TYPE_CHILD_DELETE:
1584 		audit_log_format(ab, " nametype=DELETE");
1585 		break;
1586 	case AUDIT_TYPE_CHILD_CREATE:
1587 		audit_log_format(ab, " nametype=CREATE");
1588 		break;
1589 	default:
1590 		audit_log_format(ab, " nametype=UNKNOWN");
1591 		break;
1592 	}
1593 
1594 	audit_log_fcaps(ab, n);
1595 	audit_log_end(ab);
1596 }
1597 
1598 static void audit_log_proctitle(void)
1599 {
1600 	int res;
1601 	char *buf;
1602 	char *msg = "(null)";
1603 	int len = strlen(msg);
1604 	struct audit_context *context = audit_context();
1605 	struct audit_buffer *ab;
1606 
1607 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PROCTITLE);
1608 	if (!ab)
1609 		return;	/* audit_panic or being filtered */
1610 
1611 	audit_log_format(ab, "proctitle=");
1612 
1613 	/* Not  cached */
1614 	if (!context->proctitle.value) {
1615 		buf = kmalloc(MAX_PROCTITLE_AUDIT_LEN, GFP_KERNEL);
1616 		if (!buf)
1617 			goto out;
1618 		/* Historically called this from procfs naming */
1619 		res = get_cmdline(current, buf, MAX_PROCTITLE_AUDIT_LEN);
1620 		if (res == 0) {
1621 			kfree(buf);
1622 			goto out;
1623 		}
1624 		res = audit_proctitle_rtrim(buf, res);
1625 		if (res == 0) {
1626 			kfree(buf);
1627 			goto out;
1628 		}
1629 		context->proctitle.value = buf;
1630 		context->proctitle.len = res;
1631 	}
1632 	msg = context->proctitle.value;
1633 	len = context->proctitle.len;
1634 out:
1635 	audit_log_n_untrustedstring(ab, msg, len);
1636 	audit_log_end(ab);
1637 }
1638 
1639 /**
1640  * audit_log_uring - generate a AUDIT_URINGOP record
1641  * @ctx: the audit context
1642  */
1643 static void audit_log_uring(struct audit_context *ctx)
1644 {
1645 	struct audit_buffer *ab;
1646 	const struct cred *cred;
1647 
1648 	ab = audit_log_start(ctx, GFP_ATOMIC, AUDIT_URINGOP);
1649 	if (!ab)
1650 		return;
1651 	cred = current_cred();
1652 	audit_log_format(ab, "uring_op=%d", ctx->uring_op);
1653 	if (ctx->return_valid != AUDITSC_INVALID)
1654 		audit_log_format(ab, " success=%s exit=%ld",
1655 				 (ctx->return_valid == AUDITSC_SUCCESS ?
1656 				  "yes" : "no"),
1657 				 ctx->return_code);
1658 	audit_log_format(ab,
1659 			 " items=%d"
1660 			 " ppid=%d pid=%d uid=%u gid=%u euid=%u suid=%u"
1661 			 " fsuid=%u egid=%u sgid=%u fsgid=%u",
1662 			 ctx->name_count,
1663 			 task_ppid_nr(current), task_tgid_nr(current),
1664 			 from_kuid(&init_user_ns, cred->uid),
1665 			 from_kgid(&init_user_ns, cred->gid),
1666 			 from_kuid(&init_user_ns, cred->euid),
1667 			 from_kuid(&init_user_ns, cred->suid),
1668 			 from_kuid(&init_user_ns, cred->fsuid),
1669 			 from_kgid(&init_user_ns, cred->egid),
1670 			 from_kgid(&init_user_ns, cred->sgid),
1671 			 from_kgid(&init_user_ns, cred->fsgid));
1672 	audit_log_task_context(ab);
1673 	audit_log_key(ab, ctx->filterkey);
1674 	audit_log_end(ab);
1675 }
1676 
1677 static void audit_log_exit(void)
1678 {
1679 	int i, call_panic = 0;
1680 	struct audit_context *context = audit_context();
1681 	struct audit_buffer *ab;
1682 	struct audit_aux_data *aux;
1683 	struct audit_names *n;
1684 
1685 	context->personality = current->personality;
1686 
1687 	switch (context->context) {
1688 	case AUDIT_CTX_SYSCALL:
1689 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1690 		if (!ab)
1691 			return;
1692 		audit_log_format(ab, "arch=%x syscall=%d",
1693 				 context->arch, context->major);
1694 		if (context->personality != PER_LINUX)
1695 			audit_log_format(ab, " per=%lx", context->personality);
1696 		if (context->return_valid != AUDITSC_INVALID)
1697 			audit_log_format(ab, " success=%s exit=%ld",
1698 					 (context->return_valid == AUDITSC_SUCCESS ?
1699 					  "yes" : "no"),
1700 					 context->return_code);
1701 		audit_log_format(ab,
1702 				 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d",
1703 				 context->argv[0],
1704 				 context->argv[1],
1705 				 context->argv[2],
1706 				 context->argv[3],
1707 				 context->name_count);
1708 		audit_log_task_info(ab);
1709 		audit_log_key(ab, context->filterkey);
1710 		audit_log_end(ab);
1711 		break;
1712 	case AUDIT_CTX_URING:
1713 		audit_log_uring(context);
1714 		break;
1715 	default:
1716 		BUG();
1717 		break;
1718 	}
1719 
1720 	for (aux = context->aux; aux; aux = aux->next) {
1721 
1722 		ab = audit_log_start(context, GFP_KERNEL, aux->type);
1723 		if (!ab)
1724 			continue; /* audit_panic has been called */
1725 
1726 		switch (aux->type) {
1727 
1728 		case AUDIT_BPRM_FCAPS: {
1729 			struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1730 
1731 			audit_log_format(ab, "fver=%x", axs->fcap_ver);
1732 			audit_log_cap(ab, "fp", &axs->fcap.permitted);
1733 			audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1734 			audit_log_format(ab, " fe=%d", axs->fcap.fE);
1735 			audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1736 			audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1737 			audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1738 			audit_log_cap(ab, "old_pa", &axs->old_pcap.ambient);
1739 			audit_log_cap(ab, "pp", &axs->new_pcap.permitted);
1740 			audit_log_cap(ab, "pi", &axs->new_pcap.inheritable);
1741 			audit_log_cap(ab, "pe", &axs->new_pcap.effective);
1742 			audit_log_cap(ab, "pa", &axs->new_pcap.ambient);
1743 			audit_log_format(ab, " frootid=%d",
1744 					 from_kuid(&init_user_ns,
1745 						   axs->fcap.rootid));
1746 			break; }
1747 
1748 		}
1749 		audit_log_end(ab);
1750 	}
1751 
1752 	if (context->type)
1753 		show_special(context, &call_panic);
1754 
1755 	if (context->fds[0] >= 0) {
1756 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1757 		if (ab) {
1758 			audit_log_format(ab, "fd0=%d fd1=%d",
1759 					context->fds[0], context->fds[1]);
1760 			audit_log_end(ab);
1761 		}
1762 	}
1763 
1764 	if (context->sockaddr_len) {
1765 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1766 		if (ab) {
1767 			audit_log_format(ab, "saddr=");
1768 			audit_log_n_hex(ab, (void *)context->sockaddr,
1769 					context->sockaddr_len);
1770 			audit_log_end(ab);
1771 		}
1772 	}
1773 
1774 	for (aux = context->aux_pids; aux; aux = aux->next) {
1775 		struct audit_aux_data_pids *axs = (void *)aux;
1776 
1777 		for (i = 0; i < axs->pid_count; i++)
1778 			if (audit_log_pid_context(context, axs->target_pid[i],
1779 						  axs->target_auid[i],
1780 						  axs->target_uid[i],
1781 						  axs->target_sessionid[i],
1782 						  axs->target_sid[i],
1783 						  axs->target_comm[i]))
1784 				call_panic = 1;
1785 	}
1786 
1787 	if (context->target_pid &&
1788 	    audit_log_pid_context(context, context->target_pid,
1789 				  context->target_auid, context->target_uid,
1790 				  context->target_sessionid,
1791 				  context->target_sid, context->target_comm))
1792 			call_panic = 1;
1793 
1794 	if (context->pwd.dentry && context->pwd.mnt) {
1795 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1796 		if (ab) {
1797 			audit_log_d_path(ab, "cwd=", &context->pwd);
1798 			audit_log_end(ab);
1799 		}
1800 	}
1801 
1802 	i = 0;
1803 	list_for_each_entry(n, &context->names_list, list) {
1804 		if (n->hidden)
1805 			continue;
1806 		audit_log_name(context, n, NULL, i++, &call_panic);
1807 	}
1808 
1809 	if (context->context == AUDIT_CTX_SYSCALL)
1810 		audit_log_proctitle();
1811 
1812 	/* Send end of event record to help user space know we are finished */
1813 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1814 	if (ab)
1815 		audit_log_end(ab);
1816 	if (call_panic)
1817 		audit_panic("error in audit_log_exit()");
1818 }
1819 
1820 /**
1821  * __audit_free - free a per-task audit context
1822  * @tsk: task whose audit context block to free
1823  *
1824  * Called from copy_process, do_exit, and the io_uring code
1825  */
1826 void __audit_free(struct task_struct *tsk)
1827 {
1828 	struct audit_context *context = tsk->audit_context;
1829 
1830 	if (!context)
1831 		return;
1832 
1833 	/* this may generate CONFIG_CHANGE records */
1834 	if (!list_empty(&context->killed_trees))
1835 		audit_kill_trees(context);
1836 
1837 	/* We are called either by do_exit() or the fork() error handling code;
1838 	 * in the former case tsk == current and in the latter tsk is a
1839 	 * random task_struct that doesn't have any meaningful data we
1840 	 * need to log via audit_log_exit().
1841 	 */
1842 	if (tsk == current && !context->dummy) {
1843 		context->return_valid = AUDITSC_INVALID;
1844 		context->return_code = 0;
1845 		if (context->context == AUDIT_CTX_SYSCALL) {
1846 			audit_filter_syscall(tsk, context);
1847 			audit_filter_inodes(tsk, context);
1848 			if (context->current_state == AUDIT_STATE_RECORD)
1849 				audit_log_exit();
1850 		} else if (context->context == AUDIT_CTX_URING) {
1851 			/* TODO: verify this case is real and valid */
1852 			audit_filter_uring(tsk, context);
1853 			audit_filter_inodes(tsk, context);
1854 			if (context->current_state == AUDIT_STATE_RECORD)
1855 				audit_log_uring(context);
1856 		}
1857 	}
1858 
1859 	audit_set_context(tsk, NULL);
1860 	audit_free_context(context);
1861 }
1862 
1863 /**
1864  * audit_return_fixup - fixup the return codes in the audit_context
1865  * @ctx: the audit_context
1866  * @success: true/false value to indicate if the operation succeeded or not
1867  * @code: operation return code
1868  *
1869  * We need to fixup the return code in the audit logs if the actual return
1870  * codes are later going to be fixed by the arch specific signal handlers.
1871  */
1872 static void audit_return_fixup(struct audit_context *ctx,
1873 			       int success, long code)
1874 {
1875 	/*
1876 	 * This is actually a test for:
1877 	 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
1878 	 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
1879 	 *
1880 	 * but is faster than a bunch of ||
1881 	 */
1882 	if (unlikely(code <= -ERESTARTSYS) &&
1883 	    (code >= -ERESTART_RESTARTBLOCK) &&
1884 	    (code != -ENOIOCTLCMD))
1885 		ctx->return_code = -EINTR;
1886 	else
1887 		ctx->return_code  = code;
1888 	ctx->return_valid = (success ? AUDITSC_SUCCESS : AUDITSC_FAILURE);
1889 }
1890 
1891 /**
1892  * __audit_uring_entry - prepare the kernel task's audit context for io_uring
1893  * @op: the io_uring opcode
1894  *
1895  * This is similar to audit_syscall_entry() but is intended for use by io_uring
1896  * operations.  This function should only ever be called from
1897  * audit_uring_entry() as we rely on the audit context checking present in that
1898  * function.
1899  */
1900 void __audit_uring_entry(u8 op)
1901 {
1902 	struct audit_context *ctx = audit_context();
1903 
1904 	if (ctx->state == AUDIT_STATE_DISABLED)
1905 		return;
1906 
1907 	/*
1908 	 * NOTE: It's possible that we can be called from the process' context
1909 	 *       before it returns to userspace, and before audit_syscall_exit()
1910 	 *       is called.  In this case there is not much to do, just record
1911 	 *       the io_uring details and return.
1912 	 */
1913 	ctx->uring_op = op;
1914 	if (ctx->context == AUDIT_CTX_SYSCALL)
1915 		return;
1916 
1917 	ctx->dummy = !audit_n_rules;
1918 	if (!ctx->dummy && ctx->state == AUDIT_STATE_BUILD)
1919 		ctx->prio = 0;
1920 
1921 	ctx->context = AUDIT_CTX_URING;
1922 	ctx->current_state = ctx->state;
1923 	ktime_get_coarse_real_ts64(&ctx->ctime);
1924 }
1925 
1926 /**
1927  * __audit_uring_exit - wrap up the kernel task's audit context after io_uring
1928  * @success: true/false value to indicate if the operation succeeded or not
1929  * @code: operation return code
1930  *
1931  * This is similar to audit_syscall_exit() but is intended for use by io_uring
1932  * operations.  This function should only ever be called from
1933  * audit_uring_exit() as we rely on the audit context checking present in that
1934  * function.
1935  */
1936 void __audit_uring_exit(int success, long code)
1937 {
1938 	struct audit_context *ctx = audit_context();
1939 
1940 	if (ctx->dummy) {
1941 		if (ctx->context != AUDIT_CTX_URING)
1942 			return;
1943 		goto out;
1944 	}
1945 
1946 	audit_return_fixup(ctx, success, code);
1947 	if (ctx->context == AUDIT_CTX_SYSCALL) {
1948 		/*
1949 		 * NOTE: See the note in __audit_uring_entry() about the case
1950 		 *       where we may be called from process context before we
1951 		 *       return to userspace via audit_syscall_exit().  In this
1952 		 *       case we simply emit a URINGOP record and bail, the
1953 		 *       normal syscall exit handling will take care of
1954 		 *       everything else.
1955 		 *       It is also worth mentioning that when we are called,
1956 		 *       the current process creds may differ from the creds
1957 		 *       used during the normal syscall processing; keep that
1958 		 *       in mind if/when we move the record generation code.
1959 		 */
1960 
1961 		/*
1962 		 * We need to filter on the syscall info here to decide if we
1963 		 * should emit a URINGOP record.  I know it seems odd but this
1964 		 * solves the problem where users have a filter to block *all*
1965 		 * syscall records in the "exit" filter; we want to preserve
1966 		 * the behavior here.
1967 		 */
1968 		audit_filter_syscall(current, ctx);
1969 		if (ctx->current_state != AUDIT_STATE_RECORD)
1970 			audit_filter_uring(current, ctx);
1971 		audit_filter_inodes(current, ctx);
1972 		if (ctx->current_state != AUDIT_STATE_RECORD)
1973 			return;
1974 
1975 		audit_log_uring(ctx);
1976 		return;
1977 	}
1978 
1979 	/* this may generate CONFIG_CHANGE records */
1980 	if (!list_empty(&ctx->killed_trees))
1981 		audit_kill_trees(ctx);
1982 
1983 	/* run through both filters to ensure we set the filterkey properly */
1984 	audit_filter_uring(current, ctx);
1985 	audit_filter_inodes(current, ctx);
1986 	if (ctx->current_state != AUDIT_STATE_RECORD)
1987 		goto out;
1988 	audit_log_exit();
1989 
1990 out:
1991 	audit_reset_context(ctx);
1992 }
1993 
1994 /**
1995  * __audit_syscall_entry - fill in an audit record at syscall entry
1996  * @major: major syscall type (function)
1997  * @a1: additional syscall register 1
1998  * @a2: additional syscall register 2
1999  * @a3: additional syscall register 3
2000  * @a4: additional syscall register 4
2001  *
2002  * Fill in audit context at syscall entry.  This only happens if the
2003  * audit context was created when the task was created and the state or
2004  * filters demand the audit context be built.  If the state from the
2005  * per-task filter or from the per-syscall filter is AUDIT_STATE_RECORD,
2006  * then the record will be written at syscall exit time (otherwise, it
2007  * will only be written if another part of the kernel requests that it
2008  * be written).
2009  */
2010 void __audit_syscall_entry(int major, unsigned long a1, unsigned long a2,
2011 			   unsigned long a3, unsigned long a4)
2012 {
2013 	struct audit_context *context = audit_context();
2014 	enum audit_state     state;
2015 
2016 	if (!audit_enabled || !context)
2017 		return;
2018 
2019 	WARN_ON(context->context != AUDIT_CTX_UNUSED);
2020 	WARN_ON(context->name_count);
2021 	if (context->context != AUDIT_CTX_UNUSED || context->name_count) {
2022 		audit_panic("unrecoverable error in audit_syscall_entry()");
2023 		return;
2024 	}
2025 
2026 	state = context->state;
2027 	if (state == AUDIT_STATE_DISABLED)
2028 		return;
2029 
2030 	context->dummy = !audit_n_rules;
2031 	if (!context->dummy && state == AUDIT_STATE_BUILD) {
2032 		context->prio = 0;
2033 		if (auditd_test_task(current))
2034 			return;
2035 	}
2036 
2037 	context->arch	    = syscall_get_arch(current);
2038 	context->major      = major;
2039 	context->argv[0]    = a1;
2040 	context->argv[1]    = a2;
2041 	context->argv[2]    = a3;
2042 	context->argv[3]    = a4;
2043 	context->context = AUDIT_CTX_SYSCALL;
2044 	context->current_state  = state;
2045 	ktime_get_coarse_real_ts64(&context->ctime);
2046 }
2047 
2048 /**
2049  * __audit_syscall_exit - deallocate audit context after a system call
2050  * @success: success value of the syscall
2051  * @return_code: return value of the syscall
2052  *
2053  * Tear down after system call.  If the audit context has been marked as
2054  * auditable (either because of the AUDIT_STATE_RECORD state from
2055  * filtering, or because some other part of the kernel wrote an audit
2056  * message), then write out the syscall information.  In call cases,
2057  * free the names stored from getname().
2058  */
2059 void __audit_syscall_exit(int success, long return_code)
2060 {
2061 	struct audit_context *context = audit_context();
2062 
2063 	if (!context || context->dummy ||
2064 	    context->context != AUDIT_CTX_SYSCALL)
2065 		goto out;
2066 
2067 	/* this may generate CONFIG_CHANGE records */
2068 	if (!list_empty(&context->killed_trees))
2069 		audit_kill_trees(context);
2070 
2071 	audit_return_fixup(context, success, return_code);
2072 	/* run through both filters to ensure we set the filterkey properly */
2073 	audit_filter_syscall(current, context);
2074 	audit_filter_inodes(current, context);
2075 	if (context->current_state != AUDIT_STATE_RECORD)
2076 		goto out;
2077 
2078 	audit_log_exit();
2079 
2080 out:
2081 	audit_reset_context(context);
2082 }
2083 
2084 static inline void handle_one(const struct inode *inode)
2085 {
2086 	struct audit_context *context;
2087 	struct audit_tree_refs *p;
2088 	struct audit_chunk *chunk;
2089 	int count;
2090 
2091 	if (likely(!inode->i_fsnotify_marks))
2092 		return;
2093 	context = audit_context();
2094 	p = context->trees;
2095 	count = context->tree_count;
2096 	rcu_read_lock();
2097 	chunk = audit_tree_lookup(inode);
2098 	rcu_read_unlock();
2099 	if (!chunk)
2100 		return;
2101 	if (likely(put_tree_ref(context, chunk)))
2102 		return;
2103 	if (unlikely(!grow_tree_refs(context))) {
2104 		pr_warn("out of memory, audit has lost a tree reference\n");
2105 		audit_set_auditable(context);
2106 		audit_put_chunk(chunk);
2107 		unroll_tree_refs(context, p, count);
2108 		return;
2109 	}
2110 	put_tree_ref(context, chunk);
2111 }
2112 
2113 static void handle_path(const struct dentry *dentry)
2114 {
2115 	struct audit_context *context;
2116 	struct audit_tree_refs *p;
2117 	const struct dentry *d, *parent;
2118 	struct audit_chunk *drop;
2119 	unsigned long seq;
2120 	int count;
2121 
2122 	context = audit_context();
2123 	p = context->trees;
2124 	count = context->tree_count;
2125 retry:
2126 	drop = NULL;
2127 	d = dentry;
2128 	rcu_read_lock();
2129 	seq = read_seqbegin(&rename_lock);
2130 	for(;;) {
2131 		struct inode *inode = d_backing_inode(d);
2132 
2133 		if (inode && unlikely(inode->i_fsnotify_marks)) {
2134 			struct audit_chunk *chunk;
2135 
2136 			chunk = audit_tree_lookup(inode);
2137 			if (chunk) {
2138 				if (unlikely(!put_tree_ref(context, chunk))) {
2139 					drop = chunk;
2140 					break;
2141 				}
2142 			}
2143 		}
2144 		parent = d->d_parent;
2145 		if (parent == d)
2146 			break;
2147 		d = parent;
2148 	}
2149 	if (unlikely(read_seqretry(&rename_lock, seq) || drop)) {  /* in this order */
2150 		rcu_read_unlock();
2151 		if (!drop) {
2152 			/* just a race with rename */
2153 			unroll_tree_refs(context, p, count);
2154 			goto retry;
2155 		}
2156 		audit_put_chunk(drop);
2157 		if (grow_tree_refs(context)) {
2158 			/* OK, got more space */
2159 			unroll_tree_refs(context, p, count);
2160 			goto retry;
2161 		}
2162 		/* too bad */
2163 		pr_warn("out of memory, audit has lost a tree reference\n");
2164 		unroll_tree_refs(context, p, count);
2165 		audit_set_auditable(context);
2166 		return;
2167 	}
2168 	rcu_read_unlock();
2169 }
2170 
2171 static struct audit_names *audit_alloc_name(struct audit_context *context,
2172 						unsigned char type)
2173 {
2174 	struct audit_names *aname;
2175 
2176 	if (context->name_count < AUDIT_NAMES) {
2177 		aname = &context->preallocated_names[context->name_count];
2178 		memset(aname, 0, sizeof(*aname));
2179 	} else {
2180 		aname = kzalloc(sizeof(*aname), GFP_NOFS);
2181 		if (!aname)
2182 			return NULL;
2183 		aname->should_free = true;
2184 	}
2185 
2186 	aname->ino = AUDIT_INO_UNSET;
2187 	aname->type = type;
2188 	list_add_tail(&aname->list, &context->names_list);
2189 
2190 	context->name_count++;
2191 	if (!context->pwd.dentry)
2192 		get_fs_pwd(current->fs, &context->pwd);
2193 	return aname;
2194 }
2195 
2196 /**
2197  * __audit_reusename - fill out filename with info from existing entry
2198  * @uptr: userland ptr to pathname
2199  *
2200  * Search the audit_names list for the current audit context. If there is an
2201  * existing entry with a matching "uptr" then return the filename
2202  * associated with that audit_name. If not, return NULL.
2203  */
2204 struct filename *
2205 __audit_reusename(const __user char *uptr)
2206 {
2207 	struct audit_context *context = audit_context();
2208 	struct audit_names *n;
2209 
2210 	list_for_each_entry(n, &context->names_list, list) {
2211 		if (!n->name)
2212 			continue;
2213 		if (n->name->uptr == uptr) {
2214 			n->name->refcnt++;
2215 			return n->name;
2216 		}
2217 	}
2218 	return NULL;
2219 }
2220 
2221 /**
2222  * __audit_getname - add a name to the list
2223  * @name: name to add
2224  *
2225  * Add a name to the list of audit names for this context.
2226  * Called from fs/namei.c:getname().
2227  */
2228 void __audit_getname(struct filename *name)
2229 {
2230 	struct audit_context *context = audit_context();
2231 	struct audit_names *n;
2232 
2233 	if (context->context == AUDIT_CTX_UNUSED)
2234 		return;
2235 
2236 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
2237 	if (!n)
2238 		return;
2239 
2240 	n->name = name;
2241 	n->name_len = AUDIT_NAME_FULL;
2242 	name->aname = n;
2243 	name->refcnt++;
2244 }
2245 
2246 static inline int audit_copy_fcaps(struct audit_names *name,
2247 				   const struct dentry *dentry)
2248 {
2249 	struct cpu_vfs_cap_data caps;
2250 	int rc;
2251 
2252 	if (!dentry)
2253 		return 0;
2254 
2255 	rc = get_vfs_caps_from_disk(&init_user_ns, dentry, &caps);
2256 	if (rc)
2257 		return rc;
2258 
2259 	name->fcap.permitted = caps.permitted;
2260 	name->fcap.inheritable = caps.inheritable;
2261 	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2262 	name->fcap.rootid = caps.rootid;
2263 	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >>
2264 				VFS_CAP_REVISION_SHIFT;
2265 
2266 	return 0;
2267 }
2268 
2269 /* Copy inode data into an audit_names. */
2270 static void audit_copy_inode(struct audit_names *name,
2271 			     const struct dentry *dentry,
2272 			     struct inode *inode, unsigned int flags)
2273 {
2274 	name->ino   = inode->i_ino;
2275 	name->dev   = inode->i_sb->s_dev;
2276 	name->mode  = inode->i_mode;
2277 	name->uid   = inode->i_uid;
2278 	name->gid   = inode->i_gid;
2279 	name->rdev  = inode->i_rdev;
2280 	security_inode_getsecid(inode, &name->osid);
2281 	if (flags & AUDIT_INODE_NOEVAL) {
2282 		name->fcap_ver = -1;
2283 		return;
2284 	}
2285 	audit_copy_fcaps(name, dentry);
2286 }
2287 
2288 /**
2289  * __audit_inode - store the inode and device from a lookup
2290  * @name: name being audited
2291  * @dentry: dentry being audited
2292  * @flags: attributes for this particular entry
2293  */
2294 void __audit_inode(struct filename *name, const struct dentry *dentry,
2295 		   unsigned int flags)
2296 {
2297 	struct audit_context *context = audit_context();
2298 	struct inode *inode = d_backing_inode(dentry);
2299 	struct audit_names *n;
2300 	bool parent = flags & AUDIT_INODE_PARENT;
2301 	struct audit_entry *e;
2302 	struct list_head *list = &audit_filter_list[AUDIT_FILTER_FS];
2303 	int i;
2304 
2305 	if (context->context == AUDIT_CTX_UNUSED)
2306 		return;
2307 
2308 	rcu_read_lock();
2309 	list_for_each_entry_rcu(e, list, list) {
2310 		for (i = 0; i < e->rule.field_count; i++) {
2311 			struct audit_field *f = &e->rule.fields[i];
2312 
2313 			if (f->type == AUDIT_FSTYPE
2314 			    && audit_comparator(inode->i_sb->s_magic,
2315 						f->op, f->val)
2316 			    && e->rule.action == AUDIT_NEVER) {
2317 				rcu_read_unlock();
2318 				return;
2319 			}
2320 		}
2321 	}
2322 	rcu_read_unlock();
2323 
2324 	if (!name)
2325 		goto out_alloc;
2326 
2327 	/*
2328 	 * If we have a pointer to an audit_names entry already, then we can
2329 	 * just use it directly if the type is correct.
2330 	 */
2331 	n = name->aname;
2332 	if (n) {
2333 		if (parent) {
2334 			if (n->type == AUDIT_TYPE_PARENT ||
2335 			    n->type == AUDIT_TYPE_UNKNOWN)
2336 				goto out;
2337 		} else {
2338 			if (n->type != AUDIT_TYPE_PARENT)
2339 				goto out;
2340 		}
2341 	}
2342 
2343 	list_for_each_entry_reverse(n, &context->names_list, list) {
2344 		if (n->ino) {
2345 			/* valid inode number, use that for the comparison */
2346 			if (n->ino != inode->i_ino ||
2347 			    n->dev != inode->i_sb->s_dev)
2348 				continue;
2349 		} else if (n->name) {
2350 			/* inode number has not been set, check the name */
2351 			if (strcmp(n->name->name, name->name))
2352 				continue;
2353 		} else
2354 			/* no inode and no name (?!) ... this is odd ... */
2355 			continue;
2356 
2357 		/* match the correct record type */
2358 		if (parent) {
2359 			if (n->type == AUDIT_TYPE_PARENT ||
2360 			    n->type == AUDIT_TYPE_UNKNOWN)
2361 				goto out;
2362 		} else {
2363 			if (n->type != AUDIT_TYPE_PARENT)
2364 				goto out;
2365 		}
2366 	}
2367 
2368 out_alloc:
2369 	/* unable to find an entry with both a matching name and type */
2370 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
2371 	if (!n)
2372 		return;
2373 	if (name) {
2374 		n->name = name;
2375 		name->refcnt++;
2376 	}
2377 
2378 out:
2379 	if (parent) {
2380 		n->name_len = n->name ? parent_len(n->name->name) : AUDIT_NAME_FULL;
2381 		n->type = AUDIT_TYPE_PARENT;
2382 		if (flags & AUDIT_INODE_HIDDEN)
2383 			n->hidden = true;
2384 	} else {
2385 		n->name_len = AUDIT_NAME_FULL;
2386 		n->type = AUDIT_TYPE_NORMAL;
2387 	}
2388 	handle_path(dentry);
2389 	audit_copy_inode(n, dentry, inode, flags & AUDIT_INODE_NOEVAL);
2390 }
2391 
2392 void __audit_file(const struct file *file)
2393 {
2394 	__audit_inode(NULL, file->f_path.dentry, 0);
2395 }
2396 
2397 /**
2398  * __audit_inode_child - collect inode info for created/removed objects
2399  * @parent: inode of dentry parent
2400  * @dentry: dentry being audited
2401  * @type:   AUDIT_TYPE_* value that we're looking for
2402  *
2403  * For syscalls that create or remove filesystem objects, audit_inode
2404  * can only collect information for the filesystem object's parent.
2405  * This call updates the audit context with the child's information.
2406  * Syscalls that create a new filesystem object must be hooked after
2407  * the object is created.  Syscalls that remove a filesystem object
2408  * must be hooked prior, in order to capture the target inode during
2409  * unsuccessful attempts.
2410  */
2411 void __audit_inode_child(struct inode *parent,
2412 			 const struct dentry *dentry,
2413 			 const unsigned char type)
2414 {
2415 	struct audit_context *context = audit_context();
2416 	struct inode *inode = d_backing_inode(dentry);
2417 	const struct qstr *dname = &dentry->d_name;
2418 	struct audit_names *n, *found_parent = NULL, *found_child = NULL;
2419 	struct audit_entry *e;
2420 	struct list_head *list = &audit_filter_list[AUDIT_FILTER_FS];
2421 	int i;
2422 
2423 	if (context->context == AUDIT_CTX_UNUSED)
2424 		return;
2425 
2426 	rcu_read_lock();
2427 	list_for_each_entry_rcu(e, list, list) {
2428 		for (i = 0; i < e->rule.field_count; i++) {
2429 			struct audit_field *f = &e->rule.fields[i];
2430 
2431 			if (f->type == AUDIT_FSTYPE
2432 			    && audit_comparator(parent->i_sb->s_magic,
2433 						f->op, f->val)
2434 			    && e->rule.action == AUDIT_NEVER) {
2435 				rcu_read_unlock();
2436 				return;
2437 			}
2438 		}
2439 	}
2440 	rcu_read_unlock();
2441 
2442 	if (inode)
2443 		handle_one(inode);
2444 
2445 	/* look for a parent entry first */
2446 	list_for_each_entry(n, &context->names_list, list) {
2447 		if (!n->name ||
2448 		    (n->type != AUDIT_TYPE_PARENT &&
2449 		     n->type != AUDIT_TYPE_UNKNOWN))
2450 			continue;
2451 
2452 		if (n->ino == parent->i_ino && n->dev == parent->i_sb->s_dev &&
2453 		    !audit_compare_dname_path(dname,
2454 					      n->name->name, n->name_len)) {
2455 			if (n->type == AUDIT_TYPE_UNKNOWN)
2456 				n->type = AUDIT_TYPE_PARENT;
2457 			found_parent = n;
2458 			break;
2459 		}
2460 	}
2461 
2462 	/* is there a matching child entry? */
2463 	list_for_each_entry(n, &context->names_list, list) {
2464 		/* can only match entries that have a name */
2465 		if (!n->name ||
2466 		    (n->type != type && n->type != AUDIT_TYPE_UNKNOWN))
2467 			continue;
2468 
2469 		if (!strcmp(dname->name, n->name->name) ||
2470 		    !audit_compare_dname_path(dname, n->name->name,
2471 						found_parent ?
2472 						found_parent->name_len :
2473 						AUDIT_NAME_FULL)) {
2474 			if (n->type == AUDIT_TYPE_UNKNOWN)
2475 				n->type = type;
2476 			found_child = n;
2477 			break;
2478 		}
2479 	}
2480 
2481 	if (!found_parent) {
2482 		/* create a new, "anonymous" parent record */
2483 		n = audit_alloc_name(context, AUDIT_TYPE_PARENT);
2484 		if (!n)
2485 			return;
2486 		audit_copy_inode(n, NULL, parent, 0);
2487 	}
2488 
2489 	if (!found_child) {
2490 		found_child = audit_alloc_name(context, type);
2491 		if (!found_child)
2492 			return;
2493 
2494 		/* Re-use the name belonging to the slot for a matching parent
2495 		 * directory. All names for this context are relinquished in
2496 		 * audit_free_names() */
2497 		if (found_parent) {
2498 			found_child->name = found_parent->name;
2499 			found_child->name_len = AUDIT_NAME_FULL;
2500 			found_child->name->refcnt++;
2501 		}
2502 	}
2503 
2504 	if (inode)
2505 		audit_copy_inode(found_child, dentry, inode, 0);
2506 	else
2507 		found_child->ino = AUDIT_INO_UNSET;
2508 }
2509 EXPORT_SYMBOL_GPL(__audit_inode_child);
2510 
2511 /**
2512  * auditsc_get_stamp - get local copies of audit_context values
2513  * @ctx: audit_context for the task
2514  * @t: timespec64 to store time recorded in the audit_context
2515  * @serial: serial value that is recorded in the audit_context
2516  *
2517  * Also sets the context as auditable.
2518  */
2519 int auditsc_get_stamp(struct audit_context *ctx,
2520 		       struct timespec64 *t, unsigned int *serial)
2521 {
2522 	if (ctx->context == AUDIT_CTX_UNUSED)
2523 		return 0;
2524 	if (!ctx->serial)
2525 		ctx->serial = audit_serial();
2526 	t->tv_sec  = ctx->ctime.tv_sec;
2527 	t->tv_nsec = ctx->ctime.tv_nsec;
2528 	*serial    = ctx->serial;
2529 	if (!ctx->prio) {
2530 		ctx->prio = 1;
2531 		ctx->current_state = AUDIT_STATE_RECORD;
2532 	}
2533 	return 1;
2534 }
2535 
2536 /**
2537  * __audit_mq_open - record audit data for a POSIX MQ open
2538  * @oflag: open flag
2539  * @mode: mode bits
2540  * @attr: queue attributes
2541  *
2542  */
2543 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2544 {
2545 	struct audit_context *context = audit_context();
2546 
2547 	if (attr)
2548 		memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2549 	else
2550 		memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2551 
2552 	context->mq_open.oflag = oflag;
2553 	context->mq_open.mode = mode;
2554 
2555 	context->type = AUDIT_MQ_OPEN;
2556 }
2557 
2558 /**
2559  * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2560  * @mqdes: MQ descriptor
2561  * @msg_len: Message length
2562  * @msg_prio: Message priority
2563  * @abs_timeout: Message timeout in absolute time
2564  *
2565  */
2566 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2567 			const struct timespec64 *abs_timeout)
2568 {
2569 	struct audit_context *context = audit_context();
2570 	struct timespec64 *p = &context->mq_sendrecv.abs_timeout;
2571 
2572 	if (abs_timeout)
2573 		memcpy(p, abs_timeout, sizeof(*p));
2574 	else
2575 		memset(p, 0, sizeof(*p));
2576 
2577 	context->mq_sendrecv.mqdes = mqdes;
2578 	context->mq_sendrecv.msg_len = msg_len;
2579 	context->mq_sendrecv.msg_prio = msg_prio;
2580 
2581 	context->type = AUDIT_MQ_SENDRECV;
2582 }
2583 
2584 /**
2585  * __audit_mq_notify - record audit data for a POSIX MQ notify
2586  * @mqdes: MQ descriptor
2587  * @notification: Notification event
2588  *
2589  */
2590 
2591 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2592 {
2593 	struct audit_context *context = audit_context();
2594 
2595 	if (notification)
2596 		context->mq_notify.sigev_signo = notification->sigev_signo;
2597 	else
2598 		context->mq_notify.sigev_signo = 0;
2599 
2600 	context->mq_notify.mqdes = mqdes;
2601 	context->type = AUDIT_MQ_NOTIFY;
2602 }
2603 
2604 /**
2605  * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2606  * @mqdes: MQ descriptor
2607  * @mqstat: MQ flags
2608  *
2609  */
2610 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2611 {
2612 	struct audit_context *context = audit_context();
2613 
2614 	context->mq_getsetattr.mqdes = mqdes;
2615 	context->mq_getsetattr.mqstat = *mqstat;
2616 	context->type = AUDIT_MQ_GETSETATTR;
2617 }
2618 
2619 /**
2620  * __audit_ipc_obj - record audit data for ipc object
2621  * @ipcp: ipc permissions
2622  *
2623  */
2624 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2625 {
2626 	struct audit_context *context = audit_context();
2627 
2628 	context->ipc.uid = ipcp->uid;
2629 	context->ipc.gid = ipcp->gid;
2630 	context->ipc.mode = ipcp->mode;
2631 	context->ipc.has_perm = 0;
2632 	security_ipc_getsecid(ipcp, &context->ipc.osid);
2633 	context->type = AUDIT_IPC;
2634 }
2635 
2636 /**
2637  * __audit_ipc_set_perm - record audit data for new ipc permissions
2638  * @qbytes: msgq bytes
2639  * @uid: msgq user id
2640  * @gid: msgq group id
2641  * @mode: msgq mode (permissions)
2642  *
2643  * Called only after audit_ipc_obj().
2644  */
2645 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2646 {
2647 	struct audit_context *context = audit_context();
2648 
2649 	context->ipc.qbytes = qbytes;
2650 	context->ipc.perm_uid = uid;
2651 	context->ipc.perm_gid = gid;
2652 	context->ipc.perm_mode = mode;
2653 	context->ipc.has_perm = 1;
2654 }
2655 
2656 void __audit_bprm(struct linux_binprm *bprm)
2657 {
2658 	struct audit_context *context = audit_context();
2659 
2660 	context->type = AUDIT_EXECVE;
2661 	context->execve.argc = bprm->argc;
2662 }
2663 
2664 
2665 /**
2666  * __audit_socketcall - record audit data for sys_socketcall
2667  * @nargs: number of args, which should not be more than AUDITSC_ARGS.
2668  * @args: args array
2669  *
2670  */
2671 int __audit_socketcall(int nargs, unsigned long *args)
2672 {
2673 	struct audit_context *context = audit_context();
2674 
2675 	if (nargs <= 0 || nargs > AUDITSC_ARGS || !args)
2676 		return -EINVAL;
2677 	context->type = AUDIT_SOCKETCALL;
2678 	context->socketcall.nargs = nargs;
2679 	memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2680 	return 0;
2681 }
2682 
2683 /**
2684  * __audit_fd_pair - record audit data for pipe and socketpair
2685  * @fd1: the first file descriptor
2686  * @fd2: the second file descriptor
2687  *
2688  */
2689 void __audit_fd_pair(int fd1, int fd2)
2690 {
2691 	struct audit_context *context = audit_context();
2692 
2693 	context->fds[0] = fd1;
2694 	context->fds[1] = fd2;
2695 }
2696 
2697 /**
2698  * __audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2699  * @len: data length in user space
2700  * @a: data address in kernel space
2701  *
2702  * Returns 0 for success or NULL context or < 0 on error.
2703  */
2704 int __audit_sockaddr(int len, void *a)
2705 {
2706 	struct audit_context *context = audit_context();
2707 
2708 	if (!context->sockaddr) {
2709 		void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2710 
2711 		if (!p)
2712 			return -ENOMEM;
2713 		context->sockaddr = p;
2714 	}
2715 
2716 	context->sockaddr_len = len;
2717 	memcpy(context->sockaddr, a, len);
2718 	return 0;
2719 }
2720 
2721 void __audit_ptrace(struct task_struct *t)
2722 {
2723 	struct audit_context *context = audit_context();
2724 
2725 	context->target_pid = task_tgid_nr(t);
2726 	context->target_auid = audit_get_loginuid(t);
2727 	context->target_uid = task_uid(t);
2728 	context->target_sessionid = audit_get_sessionid(t);
2729 	security_task_getsecid_obj(t, &context->target_sid);
2730 	memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2731 }
2732 
2733 /**
2734  * audit_signal_info_syscall - record signal info for syscalls
2735  * @t: task being signaled
2736  *
2737  * If the audit subsystem is being terminated, record the task (pid)
2738  * and uid that is doing that.
2739  */
2740 int audit_signal_info_syscall(struct task_struct *t)
2741 {
2742 	struct audit_aux_data_pids *axp;
2743 	struct audit_context *ctx = audit_context();
2744 	kuid_t t_uid = task_uid(t);
2745 
2746 	if (!audit_signals || audit_dummy_context())
2747 		return 0;
2748 
2749 	/* optimize the common case by putting first signal recipient directly
2750 	 * in audit_context */
2751 	if (!ctx->target_pid) {
2752 		ctx->target_pid = task_tgid_nr(t);
2753 		ctx->target_auid = audit_get_loginuid(t);
2754 		ctx->target_uid = t_uid;
2755 		ctx->target_sessionid = audit_get_sessionid(t);
2756 		security_task_getsecid_obj(t, &ctx->target_sid);
2757 		memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2758 		return 0;
2759 	}
2760 
2761 	axp = (void *)ctx->aux_pids;
2762 	if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2763 		axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2764 		if (!axp)
2765 			return -ENOMEM;
2766 
2767 		axp->d.type = AUDIT_OBJ_PID;
2768 		axp->d.next = ctx->aux_pids;
2769 		ctx->aux_pids = (void *)axp;
2770 	}
2771 	BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2772 
2773 	axp->target_pid[axp->pid_count] = task_tgid_nr(t);
2774 	axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2775 	axp->target_uid[axp->pid_count] = t_uid;
2776 	axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2777 	security_task_getsecid_obj(t, &axp->target_sid[axp->pid_count]);
2778 	memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2779 	axp->pid_count++;
2780 
2781 	return 0;
2782 }
2783 
2784 /**
2785  * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2786  * @bprm: pointer to the bprm being processed
2787  * @new: the proposed new credentials
2788  * @old: the old credentials
2789  *
2790  * Simply check if the proc already has the caps given by the file and if not
2791  * store the priv escalation info for later auditing at the end of the syscall
2792  *
2793  * -Eric
2794  */
2795 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2796 			   const struct cred *new, const struct cred *old)
2797 {
2798 	struct audit_aux_data_bprm_fcaps *ax;
2799 	struct audit_context *context = audit_context();
2800 	struct cpu_vfs_cap_data vcaps;
2801 
2802 	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2803 	if (!ax)
2804 		return -ENOMEM;
2805 
2806 	ax->d.type = AUDIT_BPRM_FCAPS;
2807 	ax->d.next = context->aux;
2808 	context->aux = (void *)ax;
2809 
2810 	get_vfs_caps_from_disk(&init_user_ns,
2811 			       bprm->file->f_path.dentry, &vcaps);
2812 
2813 	ax->fcap.permitted = vcaps.permitted;
2814 	ax->fcap.inheritable = vcaps.inheritable;
2815 	ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2816 	ax->fcap.rootid = vcaps.rootid;
2817 	ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2818 
2819 	ax->old_pcap.permitted   = old->cap_permitted;
2820 	ax->old_pcap.inheritable = old->cap_inheritable;
2821 	ax->old_pcap.effective   = old->cap_effective;
2822 	ax->old_pcap.ambient     = old->cap_ambient;
2823 
2824 	ax->new_pcap.permitted   = new->cap_permitted;
2825 	ax->new_pcap.inheritable = new->cap_inheritable;
2826 	ax->new_pcap.effective   = new->cap_effective;
2827 	ax->new_pcap.ambient     = new->cap_ambient;
2828 	return 0;
2829 }
2830 
2831 /**
2832  * __audit_log_capset - store information about the arguments to the capset syscall
2833  * @new: the new credentials
2834  * @old: the old (current) credentials
2835  *
2836  * Record the arguments userspace sent to sys_capset for later printing by the
2837  * audit system if applicable
2838  */
2839 void __audit_log_capset(const struct cred *new, const struct cred *old)
2840 {
2841 	struct audit_context *context = audit_context();
2842 
2843 	context->capset.pid = task_tgid_nr(current);
2844 	context->capset.cap.effective   = new->cap_effective;
2845 	context->capset.cap.inheritable = new->cap_effective;
2846 	context->capset.cap.permitted   = new->cap_permitted;
2847 	context->capset.cap.ambient     = new->cap_ambient;
2848 	context->type = AUDIT_CAPSET;
2849 }
2850 
2851 void __audit_mmap_fd(int fd, int flags)
2852 {
2853 	struct audit_context *context = audit_context();
2854 
2855 	context->mmap.fd = fd;
2856 	context->mmap.flags = flags;
2857 	context->type = AUDIT_MMAP;
2858 }
2859 
2860 void __audit_openat2_how(struct open_how *how)
2861 {
2862 	struct audit_context *context = audit_context();
2863 
2864 	context->openat2.flags = how->flags;
2865 	context->openat2.mode = how->mode;
2866 	context->openat2.resolve = how->resolve;
2867 	context->type = AUDIT_OPENAT2;
2868 }
2869 
2870 void __audit_log_kern_module(char *name)
2871 {
2872 	struct audit_context *context = audit_context();
2873 
2874 	context->module.name = kstrdup(name, GFP_KERNEL);
2875 	if (!context->module.name)
2876 		audit_log_lost("out of memory in __audit_log_kern_module");
2877 	context->type = AUDIT_KERN_MODULE;
2878 }
2879 
2880 void __audit_fanotify(unsigned int response)
2881 {
2882 	audit_log(audit_context(), GFP_KERNEL,
2883 		AUDIT_FANOTIFY,	"resp=%u", response);
2884 }
2885 
2886 void __audit_tk_injoffset(struct timespec64 offset)
2887 {
2888 	struct audit_context *context = audit_context();
2889 
2890 	/* only set type if not already set by NTP */
2891 	if (!context->type)
2892 		context->type = AUDIT_TIME_INJOFFSET;
2893 	memcpy(&context->time.tk_injoffset, &offset, sizeof(offset));
2894 }
2895 
2896 void __audit_ntp_log(const struct audit_ntp_data *ad)
2897 {
2898 	struct audit_context *context = audit_context();
2899 	int type;
2900 
2901 	for (type = 0; type < AUDIT_NTP_NVALS; type++)
2902 		if (ad->vals[type].newval != ad->vals[type].oldval) {
2903 			/* unconditionally set type, overwriting TK */
2904 			context->type = AUDIT_TIME_ADJNTPVAL;
2905 			memcpy(&context->time.ntp_data, ad, sizeof(*ad));
2906 			break;
2907 		}
2908 }
2909 
2910 void __audit_log_nfcfg(const char *name, u8 af, unsigned int nentries,
2911 		       enum audit_nfcfgop op, gfp_t gfp)
2912 {
2913 	struct audit_buffer *ab;
2914 	char comm[sizeof(current->comm)];
2915 
2916 	ab = audit_log_start(audit_context(), gfp, AUDIT_NETFILTER_CFG);
2917 	if (!ab)
2918 		return;
2919 	audit_log_format(ab, "table=%s family=%u entries=%u op=%s",
2920 			 name, af, nentries, audit_nfcfgs[op].s);
2921 
2922 	audit_log_format(ab, " pid=%u", task_pid_nr(current));
2923 	audit_log_task_context(ab); /* subj= */
2924 	audit_log_format(ab, " comm=");
2925 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2926 	audit_log_end(ab);
2927 }
2928 EXPORT_SYMBOL_GPL(__audit_log_nfcfg);
2929 
2930 static void audit_log_task(struct audit_buffer *ab)
2931 {
2932 	kuid_t auid, uid;
2933 	kgid_t gid;
2934 	unsigned int sessionid;
2935 	char comm[sizeof(current->comm)];
2936 
2937 	auid = audit_get_loginuid(current);
2938 	sessionid = audit_get_sessionid(current);
2939 	current_uid_gid(&uid, &gid);
2940 
2941 	audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2942 			 from_kuid(&init_user_ns, auid),
2943 			 from_kuid(&init_user_ns, uid),
2944 			 from_kgid(&init_user_ns, gid),
2945 			 sessionid);
2946 	audit_log_task_context(ab);
2947 	audit_log_format(ab, " pid=%d comm=", task_tgid_nr(current));
2948 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2949 	audit_log_d_path_exe(ab, current->mm);
2950 }
2951 
2952 /**
2953  * audit_core_dumps - record information about processes that end abnormally
2954  * @signr: signal value
2955  *
2956  * If a process ends with a core dump, something fishy is going on and we
2957  * should record the event for investigation.
2958  */
2959 void audit_core_dumps(long signr)
2960 {
2961 	struct audit_buffer *ab;
2962 
2963 	if (!audit_enabled)
2964 		return;
2965 
2966 	if (signr == SIGQUIT)	/* don't care for those */
2967 		return;
2968 
2969 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_ANOM_ABEND);
2970 	if (unlikely(!ab))
2971 		return;
2972 	audit_log_task(ab);
2973 	audit_log_format(ab, " sig=%ld res=1", signr);
2974 	audit_log_end(ab);
2975 }
2976 
2977 /**
2978  * audit_seccomp - record information about a seccomp action
2979  * @syscall: syscall number
2980  * @signr: signal value
2981  * @code: the seccomp action
2982  *
2983  * Record the information associated with a seccomp action. Event filtering for
2984  * seccomp actions that are not to be logged is done in seccomp_log().
2985  * Therefore, this function forces auditing independent of the audit_enabled
2986  * and dummy context state because seccomp actions should be logged even when
2987  * audit is not in use.
2988  */
2989 void audit_seccomp(unsigned long syscall, long signr, int code)
2990 {
2991 	struct audit_buffer *ab;
2992 
2993 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_SECCOMP);
2994 	if (unlikely(!ab))
2995 		return;
2996 	audit_log_task(ab);
2997 	audit_log_format(ab, " sig=%ld arch=%x syscall=%ld compat=%d ip=0x%lx code=0x%x",
2998 			 signr, syscall_get_arch(current), syscall,
2999 			 in_compat_syscall(), KSTK_EIP(current), code);
3000 	audit_log_end(ab);
3001 }
3002 
3003 void audit_seccomp_actions_logged(const char *names, const char *old_names,
3004 				  int res)
3005 {
3006 	struct audit_buffer *ab;
3007 
3008 	if (!audit_enabled)
3009 		return;
3010 
3011 	ab = audit_log_start(audit_context(), GFP_KERNEL,
3012 			     AUDIT_CONFIG_CHANGE);
3013 	if (unlikely(!ab))
3014 		return;
3015 
3016 	audit_log_format(ab,
3017 			 "op=seccomp-logging actions=%s old-actions=%s res=%d",
3018 			 names, old_names, res);
3019 	audit_log_end(ab);
3020 }
3021 
3022 struct list_head *audit_killed_trees(void)
3023 {
3024 	struct audit_context *ctx = audit_context();
3025 	if (likely(!ctx || ctx->context == AUDIT_CTX_UNUSED))
3026 		return NULL;
3027 	return &ctx->killed_trees;
3028 }
3029