xref: /openbmc/linux/kernel/auditsc.c (revision cd99b9eb)
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 #include <uapi/linux/fanotify.h>
68 
69 #include "audit.h"
70 
71 /* flags stating the success for a syscall */
72 #define AUDITSC_INVALID 0
73 #define AUDITSC_SUCCESS 1
74 #define AUDITSC_FAILURE 2
75 
76 /* no execve audit message should be longer than this (userspace limits),
77  * see the note near the top of audit_log_execve_info() about this value */
78 #define MAX_EXECVE_AUDIT_LEN 7500
79 
80 /* max length to print of cmdline/proctitle value during audit */
81 #define MAX_PROCTITLE_AUDIT_LEN 128
82 
83 /* number of audit rules */
84 int audit_n_rules;
85 
86 /* determines whether we collect data for signals sent */
87 int audit_signals;
88 
89 struct audit_aux_data {
90 	struct audit_aux_data	*next;
91 	int			type;
92 };
93 
94 /* Number of target pids per aux struct. */
95 #define AUDIT_AUX_PIDS	16
96 
97 struct audit_aux_data_pids {
98 	struct audit_aux_data	d;
99 	pid_t			target_pid[AUDIT_AUX_PIDS];
100 	kuid_t			target_auid[AUDIT_AUX_PIDS];
101 	kuid_t			target_uid[AUDIT_AUX_PIDS];
102 	unsigned int		target_sessionid[AUDIT_AUX_PIDS];
103 	u32			target_sid[AUDIT_AUX_PIDS];
104 	char 			target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
105 	int			pid_count;
106 };
107 
108 struct audit_aux_data_bprm_fcaps {
109 	struct audit_aux_data	d;
110 	struct audit_cap_data	fcap;
111 	unsigned int		fcap_ver;
112 	struct audit_cap_data	old_pcap;
113 	struct audit_cap_data	new_pcap;
114 };
115 
116 struct audit_tree_refs {
117 	struct audit_tree_refs *next;
118 	struct audit_chunk *c[31];
119 };
120 
121 struct audit_nfcfgop_tab {
122 	enum audit_nfcfgop	op;
123 	const char		*s;
124 };
125 
126 static const struct audit_nfcfgop_tab audit_nfcfgs[] = {
127 	{ AUDIT_XT_OP_REGISTER,			"xt_register"		   },
128 	{ AUDIT_XT_OP_REPLACE,			"xt_replace"		   },
129 	{ AUDIT_XT_OP_UNREGISTER,		"xt_unregister"		   },
130 	{ AUDIT_NFT_OP_TABLE_REGISTER,		"nft_register_table"	   },
131 	{ AUDIT_NFT_OP_TABLE_UNREGISTER,	"nft_unregister_table"	   },
132 	{ AUDIT_NFT_OP_CHAIN_REGISTER,		"nft_register_chain"	   },
133 	{ AUDIT_NFT_OP_CHAIN_UNREGISTER,	"nft_unregister_chain"	   },
134 	{ AUDIT_NFT_OP_RULE_REGISTER,		"nft_register_rule"	   },
135 	{ AUDIT_NFT_OP_RULE_UNREGISTER,		"nft_unregister_rule"	   },
136 	{ AUDIT_NFT_OP_SET_REGISTER,		"nft_register_set"	   },
137 	{ AUDIT_NFT_OP_SET_UNREGISTER,		"nft_unregister_set"	   },
138 	{ AUDIT_NFT_OP_SETELEM_REGISTER,	"nft_register_setelem"	   },
139 	{ AUDIT_NFT_OP_SETELEM_UNREGISTER,	"nft_unregister_setelem"   },
140 	{ AUDIT_NFT_OP_GEN_REGISTER,		"nft_register_gen"	   },
141 	{ AUDIT_NFT_OP_OBJ_REGISTER,		"nft_register_obj"	   },
142 	{ AUDIT_NFT_OP_OBJ_UNREGISTER,		"nft_unregister_obj"	   },
143 	{ AUDIT_NFT_OP_OBJ_RESET,		"nft_reset_obj"		   },
144 	{ AUDIT_NFT_OP_FLOWTABLE_REGISTER,	"nft_register_flowtable"   },
145 	{ AUDIT_NFT_OP_FLOWTABLE_UNREGISTER,	"nft_unregister_flowtable" },
146 	{ AUDIT_NFT_OP_INVALID,			"nft_invalid"		   },
147 };
148 
149 static int audit_match_perm(struct audit_context *ctx, int mask)
150 {
151 	unsigned n;
152 
153 	if (unlikely(!ctx))
154 		return 0;
155 	n = ctx->major;
156 
157 	switch (audit_classify_syscall(ctx->arch, n)) {
158 	case AUDITSC_NATIVE:
159 		if ((mask & AUDIT_PERM_WRITE) &&
160 		     audit_match_class(AUDIT_CLASS_WRITE, n))
161 			return 1;
162 		if ((mask & AUDIT_PERM_READ) &&
163 		     audit_match_class(AUDIT_CLASS_READ, n))
164 			return 1;
165 		if ((mask & AUDIT_PERM_ATTR) &&
166 		     audit_match_class(AUDIT_CLASS_CHATTR, n))
167 			return 1;
168 		return 0;
169 	case AUDITSC_COMPAT: /* 32bit on biarch */
170 		if ((mask & AUDIT_PERM_WRITE) &&
171 		     audit_match_class(AUDIT_CLASS_WRITE_32, n))
172 			return 1;
173 		if ((mask & AUDIT_PERM_READ) &&
174 		     audit_match_class(AUDIT_CLASS_READ_32, n))
175 			return 1;
176 		if ((mask & AUDIT_PERM_ATTR) &&
177 		     audit_match_class(AUDIT_CLASS_CHATTR_32, n))
178 			return 1;
179 		return 0;
180 	case AUDITSC_OPEN:
181 		return mask & ACC_MODE(ctx->argv[1]);
182 	case AUDITSC_OPENAT:
183 		return mask & ACC_MODE(ctx->argv[2]);
184 	case AUDITSC_SOCKETCALL:
185 		return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
186 	case AUDITSC_EXECVE:
187 		return mask & AUDIT_PERM_EXEC;
188 	case AUDITSC_OPENAT2:
189 		return mask & ACC_MODE((u32)ctx->openat2.flags);
190 	default:
191 		return 0;
192 	}
193 }
194 
195 static int audit_match_filetype(struct audit_context *ctx, int val)
196 {
197 	struct audit_names *n;
198 	umode_t mode = (umode_t)val;
199 
200 	if (unlikely(!ctx))
201 		return 0;
202 
203 	list_for_each_entry(n, &ctx->names_list, list) {
204 		if ((n->ino != AUDIT_INO_UNSET) &&
205 		    ((n->mode & S_IFMT) == mode))
206 			return 1;
207 	}
208 
209 	return 0;
210 }
211 
212 /*
213  * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
214  * ->first_trees points to its beginning, ->trees - to the current end of data.
215  * ->tree_count is the number of free entries in array pointed to by ->trees.
216  * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
217  * "empty" becomes (p, p, 31) afterwards.  We don't shrink the list (and seriously,
218  * it's going to remain 1-element for almost any setup) until we free context itself.
219  * References in it _are_ dropped - at the same time we free/drop aux stuff.
220  */
221 
222 static void audit_set_auditable(struct audit_context *ctx)
223 {
224 	if (!ctx->prio) {
225 		ctx->prio = 1;
226 		ctx->current_state = AUDIT_STATE_RECORD;
227 	}
228 }
229 
230 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
231 {
232 	struct audit_tree_refs *p = ctx->trees;
233 	int left = ctx->tree_count;
234 
235 	if (likely(left)) {
236 		p->c[--left] = chunk;
237 		ctx->tree_count = left;
238 		return 1;
239 	}
240 	if (!p)
241 		return 0;
242 	p = p->next;
243 	if (p) {
244 		p->c[30] = chunk;
245 		ctx->trees = p;
246 		ctx->tree_count = 30;
247 		return 1;
248 	}
249 	return 0;
250 }
251 
252 static int grow_tree_refs(struct audit_context *ctx)
253 {
254 	struct audit_tree_refs *p = ctx->trees;
255 
256 	ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
257 	if (!ctx->trees) {
258 		ctx->trees = p;
259 		return 0;
260 	}
261 	if (p)
262 		p->next = ctx->trees;
263 	else
264 		ctx->first_trees = ctx->trees;
265 	ctx->tree_count = 31;
266 	return 1;
267 }
268 
269 static void unroll_tree_refs(struct audit_context *ctx,
270 		      struct audit_tree_refs *p, int count)
271 {
272 	struct audit_tree_refs *q;
273 	int n;
274 
275 	if (!p) {
276 		/* we started with empty chain */
277 		p = ctx->first_trees;
278 		count = 31;
279 		/* if the very first allocation has failed, nothing to do */
280 		if (!p)
281 			return;
282 	}
283 	n = count;
284 	for (q = p; q != ctx->trees; q = q->next, n = 31) {
285 		while (n--) {
286 			audit_put_chunk(q->c[n]);
287 			q->c[n] = NULL;
288 		}
289 	}
290 	while (n-- > ctx->tree_count) {
291 		audit_put_chunk(q->c[n]);
292 		q->c[n] = NULL;
293 	}
294 	ctx->trees = p;
295 	ctx->tree_count = count;
296 }
297 
298 static void free_tree_refs(struct audit_context *ctx)
299 {
300 	struct audit_tree_refs *p, *q;
301 
302 	for (p = ctx->first_trees; p; p = q) {
303 		q = p->next;
304 		kfree(p);
305 	}
306 }
307 
308 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
309 {
310 	struct audit_tree_refs *p;
311 	int n;
312 
313 	if (!tree)
314 		return 0;
315 	/* full ones */
316 	for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
317 		for (n = 0; n < 31; n++)
318 			if (audit_tree_match(p->c[n], tree))
319 				return 1;
320 	}
321 	/* partial */
322 	if (p) {
323 		for (n = ctx->tree_count; n < 31; n++)
324 			if (audit_tree_match(p->c[n], tree))
325 				return 1;
326 	}
327 	return 0;
328 }
329 
330 static int audit_compare_uid(kuid_t uid,
331 			     struct audit_names *name,
332 			     struct audit_field *f,
333 			     struct audit_context *ctx)
334 {
335 	struct audit_names *n;
336 	int rc;
337 
338 	if (name) {
339 		rc = audit_uid_comparator(uid, f->op, name->uid);
340 		if (rc)
341 			return rc;
342 	}
343 
344 	if (ctx) {
345 		list_for_each_entry(n, &ctx->names_list, list) {
346 			rc = audit_uid_comparator(uid, f->op, n->uid);
347 			if (rc)
348 				return rc;
349 		}
350 	}
351 	return 0;
352 }
353 
354 static int audit_compare_gid(kgid_t gid,
355 			     struct audit_names *name,
356 			     struct audit_field *f,
357 			     struct audit_context *ctx)
358 {
359 	struct audit_names *n;
360 	int rc;
361 
362 	if (name) {
363 		rc = audit_gid_comparator(gid, f->op, name->gid);
364 		if (rc)
365 			return rc;
366 	}
367 
368 	if (ctx) {
369 		list_for_each_entry(n, &ctx->names_list, list) {
370 			rc = audit_gid_comparator(gid, f->op, n->gid);
371 			if (rc)
372 				return rc;
373 		}
374 	}
375 	return 0;
376 }
377 
378 static int audit_field_compare(struct task_struct *tsk,
379 			       const struct cred *cred,
380 			       struct audit_field *f,
381 			       struct audit_context *ctx,
382 			       struct audit_names *name)
383 {
384 	switch (f->val) {
385 	/* process to file object comparisons */
386 	case AUDIT_COMPARE_UID_TO_OBJ_UID:
387 		return audit_compare_uid(cred->uid, name, f, ctx);
388 	case AUDIT_COMPARE_GID_TO_OBJ_GID:
389 		return audit_compare_gid(cred->gid, name, f, ctx);
390 	case AUDIT_COMPARE_EUID_TO_OBJ_UID:
391 		return audit_compare_uid(cred->euid, name, f, ctx);
392 	case AUDIT_COMPARE_EGID_TO_OBJ_GID:
393 		return audit_compare_gid(cred->egid, name, f, ctx);
394 	case AUDIT_COMPARE_AUID_TO_OBJ_UID:
395 		return audit_compare_uid(audit_get_loginuid(tsk), name, f, ctx);
396 	case AUDIT_COMPARE_SUID_TO_OBJ_UID:
397 		return audit_compare_uid(cred->suid, name, f, ctx);
398 	case AUDIT_COMPARE_SGID_TO_OBJ_GID:
399 		return audit_compare_gid(cred->sgid, name, f, ctx);
400 	case AUDIT_COMPARE_FSUID_TO_OBJ_UID:
401 		return audit_compare_uid(cred->fsuid, name, f, ctx);
402 	case AUDIT_COMPARE_FSGID_TO_OBJ_GID:
403 		return audit_compare_gid(cred->fsgid, name, f, ctx);
404 	/* uid comparisons */
405 	case AUDIT_COMPARE_UID_TO_AUID:
406 		return audit_uid_comparator(cred->uid, f->op,
407 					    audit_get_loginuid(tsk));
408 	case AUDIT_COMPARE_UID_TO_EUID:
409 		return audit_uid_comparator(cred->uid, f->op, cred->euid);
410 	case AUDIT_COMPARE_UID_TO_SUID:
411 		return audit_uid_comparator(cred->uid, f->op, cred->suid);
412 	case AUDIT_COMPARE_UID_TO_FSUID:
413 		return audit_uid_comparator(cred->uid, f->op, cred->fsuid);
414 	/* auid comparisons */
415 	case AUDIT_COMPARE_AUID_TO_EUID:
416 		return audit_uid_comparator(audit_get_loginuid(tsk), f->op,
417 					    cred->euid);
418 	case AUDIT_COMPARE_AUID_TO_SUID:
419 		return audit_uid_comparator(audit_get_loginuid(tsk), f->op,
420 					    cred->suid);
421 	case AUDIT_COMPARE_AUID_TO_FSUID:
422 		return audit_uid_comparator(audit_get_loginuid(tsk), f->op,
423 					    cred->fsuid);
424 	/* euid comparisons */
425 	case AUDIT_COMPARE_EUID_TO_SUID:
426 		return audit_uid_comparator(cred->euid, f->op, cred->suid);
427 	case AUDIT_COMPARE_EUID_TO_FSUID:
428 		return audit_uid_comparator(cred->euid, f->op, cred->fsuid);
429 	/* suid comparisons */
430 	case AUDIT_COMPARE_SUID_TO_FSUID:
431 		return audit_uid_comparator(cred->suid, f->op, cred->fsuid);
432 	/* gid comparisons */
433 	case AUDIT_COMPARE_GID_TO_EGID:
434 		return audit_gid_comparator(cred->gid, f->op, cred->egid);
435 	case AUDIT_COMPARE_GID_TO_SGID:
436 		return audit_gid_comparator(cred->gid, f->op, cred->sgid);
437 	case AUDIT_COMPARE_GID_TO_FSGID:
438 		return audit_gid_comparator(cred->gid, f->op, cred->fsgid);
439 	/* egid comparisons */
440 	case AUDIT_COMPARE_EGID_TO_SGID:
441 		return audit_gid_comparator(cred->egid, f->op, cred->sgid);
442 	case AUDIT_COMPARE_EGID_TO_FSGID:
443 		return audit_gid_comparator(cred->egid, f->op, cred->fsgid);
444 	/* sgid comparison */
445 	case AUDIT_COMPARE_SGID_TO_FSGID:
446 		return audit_gid_comparator(cred->sgid, f->op, cred->fsgid);
447 	default:
448 		WARN(1, "Missing AUDIT_COMPARE define.  Report as a bug\n");
449 		return 0;
450 	}
451 	return 0;
452 }
453 
454 /* Determine if any context name data matches a rule's watch data */
455 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
456  * otherwise.
457  *
458  * If task_creation is true, this is an explicit indication that we are
459  * filtering a task rule at task creation time.  This and tsk == current are
460  * the only situations where tsk->cred may be accessed without an rcu read lock.
461  */
462 static int audit_filter_rules(struct task_struct *tsk,
463 			      struct audit_krule *rule,
464 			      struct audit_context *ctx,
465 			      struct audit_names *name,
466 			      enum audit_state *state,
467 			      bool task_creation)
468 {
469 	const struct cred *cred;
470 	int i, need_sid = 1;
471 	u32 sid;
472 	unsigned int sessionid;
473 
474 	if (ctx && rule->prio <= ctx->prio)
475 		return 0;
476 
477 	cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
478 
479 	for (i = 0; i < rule->field_count; i++) {
480 		struct audit_field *f = &rule->fields[i];
481 		struct audit_names *n;
482 		int result = 0;
483 		pid_t pid;
484 
485 		switch (f->type) {
486 		case AUDIT_PID:
487 			pid = task_tgid_nr(tsk);
488 			result = audit_comparator(pid, f->op, f->val);
489 			break;
490 		case AUDIT_PPID:
491 			if (ctx) {
492 				if (!ctx->ppid)
493 					ctx->ppid = task_ppid_nr(tsk);
494 				result = audit_comparator(ctx->ppid, f->op, f->val);
495 			}
496 			break;
497 		case AUDIT_EXE:
498 			result = audit_exe_compare(tsk, rule->exe);
499 			if (f->op == Audit_not_equal)
500 				result = !result;
501 			break;
502 		case AUDIT_UID:
503 			result = audit_uid_comparator(cred->uid, f->op, f->uid);
504 			break;
505 		case AUDIT_EUID:
506 			result = audit_uid_comparator(cred->euid, f->op, f->uid);
507 			break;
508 		case AUDIT_SUID:
509 			result = audit_uid_comparator(cred->suid, f->op, f->uid);
510 			break;
511 		case AUDIT_FSUID:
512 			result = audit_uid_comparator(cred->fsuid, f->op, f->uid);
513 			break;
514 		case AUDIT_GID:
515 			result = audit_gid_comparator(cred->gid, f->op, f->gid);
516 			if (f->op == Audit_equal) {
517 				if (!result)
518 					result = groups_search(cred->group_info, f->gid);
519 			} else if (f->op == Audit_not_equal) {
520 				if (result)
521 					result = !groups_search(cred->group_info, f->gid);
522 			}
523 			break;
524 		case AUDIT_EGID:
525 			result = audit_gid_comparator(cred->egid, f->op, f->gid);
526 			if (f->op == Audit_equal) {
527 				if (!result)
528 					result = groups_search(cred->group_info, f->gid);
529 			} else if (f->op == Audit_not_equal) {
530 				if (result)
531 					result = !groups_search(cred->group_info, f->gid);
532 			}
533 			break;
534 		case AUDIT_SGID:
535 			result = audit_gid_comparator(cred->sgid, f->op, f->gid);
536 			break;
537 		case AUDIT_FSGID:
538 			result = audit_gid_comparator(cred->fsgid, f->op, f->gid);
539 			break;
540 		case AUDIT_SESSIONID:
541 			sessionid = audit_get_sessionid(tsk);
542 			result = audit_comparator(sessionid, f->op, f->val);
543 			break;
544 		case AUDIT_PERS:
545 			result = audit_comparator(tsk->personality, f->op, f->val);
546 			break;
547 		case AUDIT_ARCH:
548 			if (ctx)
549 				result = audit_comparator(ctx->arch, f->op, f->val);
550 			break;
551 
552 		case AUDIT_EXIT:
553 			if (ctx && ctx->return_valid != AUDITSC_INVALID)
554 				result = audit_comparator(ctx->return_code, f->op, f->val);
555 			break;
556 		case AUDIT_SUCCESS:
557 			if (ctx && ctx->return_valid != AUDITSC_INVALID) {
558 				if (f->val)
559 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
560 				else
561 					result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
562 			}
563 			break;
564 		case AUDIT_DEVMAJOR:
565 			if (name) {
566 				if (audit_comparator(MAJOR(name->dev), f->op, f->val) ||
567 				    audit_comparator(MAJOR(name->rdev), f->op, f->val))
568 					++result;
569 			} else if (ctx) {
570 				list_for_each_entry(n, &ctx->names_list, list) {
571 					if (audit_comparator(MAJOR(n->dev), f->op, f->val) ||
572 					    audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
573 						++result;
574 						break;
575 					}
576 				}
577 			}
578 			break;
579 		case AUDIT_DEVMINOR:
580 			if (name) {
581 				if (audit_comparator(MINOR(name->dev), f->op, f->val) ||
582 				    audit_comparator(MINOR(name->rdev), f->op, f->val))
583 					++result;
584 			} else if (ctx) {
585 				list_for_each_entry(n, &ctx->names_list, list) {
586 					if (audit_comparator(MINOR(n->dev), f->op, f->val) ||
587 					    audit_comparator(MINOR(n->rdev), f->op, f->val)) {
588 						++result;
589 						break;
590 					}
591 				}
592 			}
593 			break;
594 		case AUDIT_INODE:
595 			if (name)
596 				result = audit_comparator(name->ino, f->op, f->val);
597 			else if (ctx) {
598 				list_for_each_entry(n, &ctx->names_list, list) {
599 					if (audit_comparator(n->ino, f->op, f->val)) {
600 						++result;
601 						break;
602 					}
603 				}
604 			}
605 			break;
606 		case AUDIT_OBJ_UID:
607 			if (name) {
608 				result = audit_uid_comparator(name->uid, f->op, f->uid);
609 			} else if (ctx) {
610 				list_for_each_entry(n, &ctx->names_list, list) {
611 					if (audit_uid_comparator(n->uid, f->op, f->uid)) {
612 						++result;
613 						break;
614 					}
615 				}
616 			}
617 			break;
618 		case AUDIT_OBJ_GID:
619 			if (name) {
620 				result = audit_gid_comparator(name->gid, f->op, f->gid);
621 			} else if (ctx) {
622 				list_for_each_entry(n, &ctx->names_list, list) {
623 					if (audit_gid_comparator(n->gid, f->op, f->gid)) {
624 						++result;
625 						break;
626 					}
627 				}
628 			}
629 			break;
630 		case AUDIT_WATCH:
631 			if (name) {
632 				result = audit_watch_compare(rule->watch,
633 							     name->ino,
634 							     name->dev);
635 				if (f->op == Audit_not_equal)
636 					result = !result;
637 			}
638 			break;
639 		case AUDIT_DIR:
640 			if (ctx) {
641 				result = match_tree_refs(ctx, rule->tree);
642 				if (f->op == Audit_not_equal)
643 					result = !result;
644 			}
645 			break;
646 		case AUDIT_LOGINUID:
647 			result = audit_uid_comparator(audit_get_loginuid(tsk),
648 						      f->op, f->uid);
649 			break;
650 		case AUDIT_LOGINUID_SET:
651 			result = audit_comparator(audit_loginuid_set(tsk), f->op, f->val);
652 			break;
653 		case AUDIT_SADDR_FAM:
654 			if (ctx && ctx->sockaddr)
655 				result = audit_comparator(ctx->sockaddr->ss_family,
656 							  f->op, f->val);
657 			break;
658 		case AUDIT_SUBJ_USER:
659 		case AUDIT_SUBJ_ROLE:
660 		case AUDIT_SUBJ_TYPE:
661 		case AUDIT_SUBJ_SEN:
662 		case AUDIT_SUBJ_CLR:
663 			/* NOTE: this may return negative values indicating
664 			   a temporary error.  We simply treat this as a
665 			   match for now to avoid losing information that
666 			   may be wanted.   An error message will also be
667 			   logged upon error */
668 			if (f->lsm_rule) {
669 				if (need_sid) {
670 					/* @tsk should always be equal to
671 					 * @current with the exception of
672 					 * fork()/copy_process() in which case
673 					 * the new @tsk creds are still a dup
674 					 * of @current's creds so we can still
675 					 * use security_current_getsecid_subj()
676 					 * here even though it always refs
677 					 * @current's creds
678 					 */
679 					security_current_getsecid_subj(&sid);
680 					need_sid = 0;
681 				}
682 				result = security_audit_rule_match(sid, f->type,
683 								   f->op,
684 								   f->lsm_rule);
685 			}
686 			break;
687 		case AUDIT_OBJ_USER:
688 		case AUDIT_OBJ_ROLE:
689 		case AUDIT_OBJ_TYPE:
690 		case AUDIT_OBJ_LEV_LOW:
691 		case AUDIT_OBJ_LEV_HIGH:
692 			/* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
693 			   also applies here */
694 			if (f->lsm_rule) {
695 				/* Find files that match */
696 				if (name) {
697 					result = security_audit_rule_match(
698 								name->osid,
699 								f->type,
700 								f->op,
701 								f->lsm_rule);
702 				} else if (ctx) {
703 					list_for_each_entry(n, &ctx->names_list, list) {
704 						if (security_audit_rule_match(
705 								n->osid,
706 								f->type,
707 								f->op,
708 								f->lsm_rule)) {
709 							++result;
710 							break;
711 						}
712 					}
713 				}
714 				/* Find ipc objects that match */
715 				if (!ctx || ctx->type != AUDIT_IPC)
716 					break;
717 				if (security_audit_rule_match(ctx->ipc.osid,
718 							      f->type, f->op,
719 							      f->lsm_rule))
720 					++result;
721 			}
722 			break;
723 		case AUDIT_ARG0:
724 		case AUDIT_ARG1:
725 		case AUDIT_ARG2:
726 		case AUDIT_ARG3:
727 			if (ctx)
728 				result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
729 			break;
730 		case AUDIT_FILTERKEY:
731 			/* ignore this field for filtering */
732 			result = 1;
733 			break;
734 		case AUDIT_PERM:
735 			result = audit_match_perm(ctx, f->val);
736 			if (f->op == Audit_not_equal)
737 				result = !result;
738 			break;
739 		case AUDIT_FILETYPE:
740 			result = audit_match_filetype(ctx, f->val);
741 			if (f->op == Audit_not_equal)
742 				result = !result;
743 			break;
744 		case AUDIT_FIELD_COMPARE:
745 			result = audit_field_compare(tsk, cred, f, ctx, name);
746 			break;
747 		}
748 		if (!result)
749 			return 0;
750 	}
751 
752 	if (ctx) {
753 		if (rule->filterkey) {
754 			kfree(ctx->filterkey);
755 			ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
756 		}
757 		ctx->prio = rule->prio;
758 	}
759 	switch (rule->action) {
760 	case AUDIT_NEVER:
761 		*state = AUDIT_STATE_DISABLED;
762 		break;
763 	case AUDIT_ALWAYS:
764 		*state = AUDIT_STATE_RECORD;
765 		break;
766 	}
767 	return 1;
768 }
769 
770 /* At process creation time, we can determine if system-call auditing is
771  * completely disabled for this task.  Since we only have the task
772  * structure at this point, we can only check uid and gid.
773  */
774 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
775 {
776 	struct audit_entry *e;
777 	enum audit_state   state;
778 
779 	rcu_read_lock();
780 	list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
781 		if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
782 				       &state, true)) {
783 			if (state == AUDIT_STATE_RECORD)
784 				*key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
785 			rcu_read_unlock();
786 			return state;
787 		}
788 	}
789 	rcu_read_unlock();
790 	return AUDIT_STATE_BUILD;
791 }
792 
793 static int audit_in_mask(const struct audit_krule *rule, unsigned long val)
794 {
795 	int word, bit;
796 
797 	if (val > 0xffffffff)
798 		return false;
799 
800 	word = AUDIT_WORD(val);
801 	if (word >= AUDIT_BITMASK_SIZE)
802 		return false;
803 
804 	bit = AUDIT_BIT(val);
805 
806 	return rule->mask[word] & bit;
807 }
808 
809 /**
810  * __audit_filter_op - common filter helper for operations (syscall/uring/etc)
811  * @tsk: associated task
812  * @ctx: audit context
813  * @list: audit filter list
814  * @name: audit_name (can be NULL)
815  * @op: current syscall/uring_op
816  *
817  * Run the udit filters specified in @list against @tsk using @ctx,
818  * @name, and @op, as necessary; the caller is responsible for ensuring
819  * that the call is made while the RCU read lock is held. The @name
820  * parameter can be NULL, but all others must be specified.
821  * Returns 1/true if the filter finds a match, 0/false if none are found.
822  */
823 static int __audit_filter_op(struct task_struct *tsk,
824 			   struct audit_context *ctx,
825 			   struct list_head *list,
826 			   struct audit_names *name,
827 			   unsigned long op)
828 {
829 	struct audit_entry *e;
830 	enum audit_state state;
831 
832 	list_for_each_entry_rcu(e, list, list) {
833 		if (audit_in_mask(&e->rule, op) &&
834 		    audit_filter_rules(tsk, &e->rule, ctx, name,
835 				       &state, false)) {
836 			ctx->current_state = state;
837 			return 1;
838 		}
839 	}
840 	return 0;
841 }
842 
843 /**
844  * audit_filter_uring - apply filters to an io_uring operation
845  * @tsk: associated task
846  * @ctx: audit context
847  */
848 static void audit_filter_uring(struct task_struct *tsk,
849 			       struct audit_context *ctx)
850 {
851 	if (auditd_test_task(tsk))
852 		return;
853 
854 	rcu_read_lock();
855 	__audit_filter_op(tsk, ctx, &audit_filter_list[AUDIT_FILTER_URING_EXIT],
856 			NULL, ctx->uring_op);
857 	rcu_read_unlock();
858 }
859 
860 /* At syscall exit time, this filter is called if the audit_state is
861  * not low enough that auditing cannot take place, but is also not
862  * high enough that we already know we have to write an audit record
863  * (i.e., the state is AUDIT_STATE_BUILD).
864  */
865 static void audit_filter_syscall(struct task_struct *tsk,
866 				 struct audit_context *ctx)
867 {
868 	if (auditd_test_task(tsk))
869 		return;
870 
871 	rcu_read_lock();
872 	__audit_filter_op(tsk, ctx, &audit_filter_list[AUDIT_FILTER_EXIT],
873 			NULL, ctx->major);
874 	rcu_read_unlock();
875 }
876 
877 /*
878  * Given an audit_name check the inode hash table to see if they match.
879  * Called holding the rcu read lock to protect the use of audit_inode_hash
880  */
881 static int audit_filter_inode_name(struct task_struct *tsk,
882 				   struct audit_names *n,
883 				   struct audit_context *ctx)
884 {
885 	int h = audit_hash_ino((u32)n->ino);
886 	struct list_head *list = &audit_inode_hash[h];
887 
888 	return __audit_filter_op(tsk, ctx, list, n, ctx->major);
889 }
890 
891 /* At syscall exit time, this filter is called if any audit_names have been
892  * collected during syscall processing.  We only check rules in sublists at hash
893  * buckets applicable to the inode numbers in audit_names.
894  * Regarding audit_state, same rules apply as for audit_filter_syscall().
895  */
896 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
897 {
898 	struct audit_names *n;
899 
900 	if (auditd_test_task(tsk))
901 		return;
902 
903 	rcu_read_lock();
904 
905 	list_for_each_entry(n, &ctx->names_list, list) {
906 		if (audit_filter_inode_name(tsk, n, ctx))
907 			break;
908 	}
909 	rcu_read_unlock();
910 }
911 
912 static inline void audit_proctitle_free(struct audit_context *context)
913 {
914 	kfree(context->proctitle.value);
915 	context->proctitle.value = NULL;
916 	context->proctitle.len = 0;
917 }
918 
919 static inline void audit_free_module(struct audit_context *context)
920 {
921 	if (context->type == AUDIT_KERN_MODULE) {
922 		kfree(context->module.name);
923 		context->module.name = NULL;
924 	}
925 }
926 static inline void audit_free_names(struct audit_context *context)
927 {
928 	struct audit_names *n, *next;
929 
930 	list_for_each_entry_safe(n, next, &context->names_list, list) {
931 		list_del(&n->list);
932 		if (n->name)
933 			putname(n->name);
934 		if (n->should_free)
935 			kfree(n);
936 	}
937 	context->name_count = 0;
938 	path_put(&context->pwd);
939 	context->pwd.dentry = NULL;
940 	context->pwd.mnt = NULL;
941 }
942 
943 static inline void audit_free_aux(struct audit_context *context)
944 {
945 	struct audit_aux_data *aux;
946 
947 	while ((aux = context->aux)) {
948 		context->aux = aux->next;
949 		kfree(aux);
950 	}
951 	context->aux = NULL;
952 	while ((aux = context->aux_pids)) {
953 		context->aux_pids = aux->next;
954 		kfree(aux);
955 	}
956 	context->aux_pids = NULL;
957 }
958 
959 /**
960  * audit_reset_context - reset a audit_context structure
961  * @ctx: the audit_context to reset
962  *
963  * All fields in the audit_context will be reset to an initial state, all
964  * references held by fields will be dropped, and private memory will be
965  * released.  When this function returns the audit_context will be suitable
966  * for reuse, so long as the passed context is not NULL or a dummy context.
967  */
968 static void audit_reset_context(struct audit_context *ctx)
969 {
970 	if (!ctx)
971 		return;
972 
973 	/* if ctx is non-null, reset the "ctx->context" regardless */
974 	ctx->context = AUDIT_CTX_UNUSED;
975 	if (ctx->dummy)
976 		return;
977 
978 	/*
979 	 * NOTE: It shouldn't matter in what order we release the fields, so
980 	 *       release them in the order in which they appear in the struct;
981 	 *       this gives us some hope of quickly making sure we are
982 	 *       resetting the audit_context properly.
983 	 *
984 	 *       Other things worth mentioning:
985 	 *       - we don't reset "dummy"
986 	 *       - we don't reset "state", we do reset "current_state"
987 	 *       - we preserve "filterkey" if "state" is AUDIT_STATE_RECORD
988 	 *       - much of this is likely overkill, but play it safe for now
989 	 *       - we really need to work on improving the audit_context struct
990 	 */
991 
992 	ctx->current_state = ctx->state;
993 	ctx->serial = 0;
994 	ctx->major = 0;
995 	ctx->uring_op = 0;
996 	ctx->ctime = (struct timespec64){ .tv_sec = 0, .tv_nsec = 0 };
997 	memset(ctx->argv, 0, sizeof(ctx->argv));
998 	ctx->return_code = 0;
999 	ctx->prio = (ctx->state == AUDIT_STATE_RECORD ? ~0ULL : 0);
1000 	ctx->return_valid = AUDITSC_INVALID;
1001 	audit_free_names(ctx);
1002 	if (ctx->state != AUDIT_STATE_RECORD) {
1003 		kfree(ctx->filterkey);
1004 		ctx->filterkey = NULL;
1005 	}
1006 	audit_free_aux(ctx);
1007 	kfree(ctx->sockaddr);
1008 	ctx->sockaddr = NULL;
1009 	ctx->sockaddr_len = 0;
1010 	ctx->ppid = 0;
1011 	ctx->uid = ctx->euid = ctx->suid = ctx->fsuid = KUIDT_INIT(0);
1012 	ctx->gid = ctx->egid = ctx->sgid = ctx->fsgid = KGIDT_INIT(0);
1013 	ctx->personality = 0;
1014 	ctx->arch = 0;
1015 	ctx->target_pid = 0;
1016 	ctx->target_auid = ctx->target_uid = KUIDT_INIT(0);
1017 	ctx->target_sessionid = 0;
1018 	ctx->target_sid = 0;
1019 	ctx->target_comm[0] = '\0';
1020 	unroll_tree_refs(ctx, NULL, 0);
1021 	WARN_ON(!list_empty(&ctx->killed_trees));
1022 	audit_free_module(ctx);
1023 	ctx->fds[0] = -1;
1024 	ctx->type = 0; /* reset last for audit_free_*() */
1025 }
1026 
1027 static inline struct audit_context *audit_alloc_context(enum audit_state state)
1028 {
1029 	struct audit_context *context;
1030 
1031 	context = kzalloc(sizeof(*context), GFP_KERNEL);
1032 	if (!context)
1033 		return NULL;
1034 	context->context = AUDIT_CTX_UNUSED;
1035 	context->state = state;
1036 	context->prio = state == AUDIT_STATE_RECORD ? ~0ULL : 0;
1037 	INIT_LIST_HEAD(&context->killed_trees);
1038 	INIT_LIST_HEAD(&context->names_list);
1039 	context->fds[0] = -1;
1040 	context->return_valid = AUDITSC_INVALID;
1041 	return context;
1042 }
1043 
1044 /**
1045  * audit_alloc - allocate an audit context block for a task
1046  * @tsk: task
1047  *
1048  * Filter on the task information and allocate a per-task audit context
1049  * if necessary.  Doing so turns on system call auditing for the
1050  * specified task.  This is called from copy_process, so no lock is
1051  * needed.
1052  */
1053 int audit_alloc(struct task_struct *tsk)
1054 {
1055 	struct audit_context *context;
1056 	enum audit_state     state;
1057 	char *key = NULL;
1058 
1059 	if (likely(!audit_ever_enabled))
1060 		return 0;
1061 
1062 	state = audit_filter_task(tsk, &key);
1063 	if (state == AUDIT_STATE_DISABLED) {
1064 		clear_task_syscall_work(tsk, SYSCALL_AUDIT);
1065 		return 0;
1066 	}
1067 
1068 	context = audit_alloc_context(state);
1069 	if (!context) {
1070 		kfree(key);
1071 		audit_log_lost("out of memory in audit_alloc");
1072 		return -ENOMEM;
1073 	}
1074 	context->filterkey = key;
1075 
1076 	audit_set_context(tsk, context);
1077 	set_task_syscall_work(tsk, SYSCALL_AUDIT);
1078 	return 0;
1079 }
1080 
1081 static inline void audit_free_context(struct audit_context *context)
1082 {
1083 	/* resetting is extra work, but it is likely just noise */
1084 	audit_reset_context(context);
1085 	audit_proctitle_free(context);
1086 	free_tree_refs(context);
1087 	kfree(context->filterkey);
1088 	kfree(context);
1089 }
1090 
1091 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
1092 				 kuid_t auid, kuid_t uid, unsigned int sessionid,
1093 				 u32 sid, char *comm)
1094 {
1095 	struct audit_buffer *ab;
1096 	char *ctx = NULL;
1097 	u32 len;
1098 	int rc = 0;
1099 
1100 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1101 	if (!ab)
1102 		return rc;
1103 
1104 	audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid,
1105 			 from_kuid(&init_user_ns, auid),
1106 			 from_kuid(&init_user_ns, uid), sessionid);
1107 	if (sid) {
1108 		if (security_secid_to_secctx(sid, &ctx, &len)) {
1109 			audit_log_format(ab, " obj=(none)");
1110 			rc = 1;
1111 		} else {
1112 			audit_log_format(ab, " obj=%s", ctx);
1113 			security_release_secctx(ctx, len);
1114 		}
1115 	}
1116 	audit_log_format(ab, " ocomm=");
1117 	audit_log_untrustedstring(ab, comm);
1118 	audit_log_end(ab);
1119 
1120 	return rc;
1121 }
1122 
1123 static void audit_log_execve_info(struct audit_context *context,
1124 				  struct audit_buffer **ab)
1125 {
1126 	long len_max;
1127 	long len_rem;
1128 	long len_full;
1129 	long len_buf;
1130 	long len_abuf = 0;
1131 	long len_tmp;
1132 	bool require_data;
1133 	bool encode;
1134 	unsigned int iter;
1135 	unsigned int arg;
1136 	char *buf_head;
1137 	char *buf;
1138 	const char __user *p = (const char __user *)current->mm->arg_start;
1139 
1140 	/* NOTE: this buffer needs to be large enough to hold all the non-arg
1141 	 *       data we put in the audit record for this argument (see the
1142 	 *       code below) ... at this point in time 96 is plenty */
1143 	char abuf[96];
1144 
1145 	/* NOTE: we set MAX_EXECVE_AUDIT_LEN to a rather arbitrary limit, the
1146 	 *       current value of 7500 is not as important as the fact that it
1147 	 *       is less than 8k, a setting of 7500 gives us plenty of wiggle
1148 	 *       room if we go over a little bit in the logging below */
1149 	WARN_ON_ONCE(MAX_EXECVE_AUDIT_LEN > 7500);
1150 	len_max = MAX_EXECVE_AUDIT_LEN;
1151 
1152 	/* scratch buffer to hold the userspace args */
1153 	buf_head = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1154 	if (!buf_head) {
1155 		audit_panic("out of memory for argv string");
1156 		return;
1157 	}
1158 	buf = buf_head;
1159 
1160 	audit_log_format(*ab, "argc=%d", context->execve.argc);
1161 
1162 	len_rem = len_max;
1163 	len_buf = 0;
1164 	len_full = 0;
1165 	require_data = true;
1166 	encode = false;
1167 	iter = 0;
1168 	arg = 0;
1169 	do {
1170 		/* NOTE: we don't ever want to trust this value for anything
1171 		 *       serious, but the audit record format insists we
1172 		 *       provide an argument length for really long arguments,
1173 		 *       e.g. > MAX_EXECVE_AUDIT_LEN, so we have no choice but
1174 		 *       to use strncpy_from_user() to obtain this value for
1175 		 *       recording in the log, although we don't use it
1176 		 *       anywhere here to avoid a double-fetch problem */
1177 		if (len_full == 0)
1178 			len_full = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1179 
1180 		/* read more data from userspace */
1181 		if (require_data) {
1182 			/* can we make more room in the buffer? */
1183 			if (buf != buf_head) {
1184 				memmove(buf_head, buf, len_buf);
1185 				buf = buf_head;
1186 			}
1187 
1188 			/* fetch as much as we can of the argument */
1189 			len_tmp = strncpy_from_user(&buf_head[len_buf], p,
1190 						    len_max - len_buf);
1191 			if (len_tmp == -EFAULT) {
1192 				/* unable to copy from userspace */
1193 				send_sig(SIGKILL, current, 0);
1194 				goto out;
1195 			} else if (len_tmp == (len_max - len_buf)) {
1196 				/* buffer is not large enough */
1197 				require_data = true;
1198 				/* NOTE: if we are going to span multiple
1199 				 *       buffers force the encoding so we stand
1200 				 *       a chance at a sane len_full value and
1201 				 *       consistent record encoding */
1202 				encode = true;
1203 				len_full = len_full * 2;
1204 				p += len_tmp;
1205 			} else {
1206 				require_data = false;
1207 				if (!encode)
1208 					encode = audit_string_contains_control(
1209 								buf, len_tmp);
1210 				/* try to use a trusted value for len_full */
1211 				if (len_full < len_max)
1212 					len_full = (encode ?
1213 						    len_tmp * 2 : len_tmp);
1214 				p += len_tmp + 1;
1215 			}
1216 			len_buf += len_tmp;
1217 			buf_head[len_buf] = '\0';
1218 
1219 			/* length of the buffer in the audit record? */
1220 			len_abuf = (encode ? len_buf * 2 : len_buf + 2);
1221 		}
1222 
1223 		/* write as much as we can to the audit log */
1224 		if (len_buf >= 0) {
1225 			/* NOTE: some magic numbers here - basically if we
1226 			 *       can't fit a reasonable amount of data into the
1227 			 *       existing audit buffer, flush it and start with
1228 			 *       a new buffer */
1229 			if ((sizeof(abuf) + 8) > len_rem) {
1230 				len_rem = len_max;
1231 				audit_log_end(*ab);
1232 				*ab = audit_log_start(context,
1233 						      GFP_KERNEL, AUDIT_EXECVE);
1234 				if (!*ab)
1235 					goto out;
1236 			}
1237 
1238 			/* create the non-arg portion of the arg record */
1239 			len_tmp = 0;
1240 			if (require_data || (iter > 0) ||
1241 			    ((len_abuf + sizeof(abuf)) > len_rem)) {
1242 				if (iter == 0) {
1243 					len_tmp += snprintf(&abuf[len_tmp],
1244 							sizeof(abuf) - len_tmp,
1245 							" a%d_len=%lu",
1246 							arg, len_full);
1247 				}
1248 				len_tmp += snprintf(&abuf[len_tmp],
1249 						    sizeof(abuf) - len_tmp,
1250 						    " a%d[%d]=", arg, iter++);
1251 			} else
1252 				len_tmp += snprintf(&abuf[len_tmp],
1253 						    sizeof(abuf) - len_tmp,
1254 						    " a%d=", arg);
1255 			WARN_ON(len_tmp >= sizeof(abuf));
1256 			abuf[sizeof(abuf) - 1] = '\0';
1257 
1258 			/* log the arg in the audit record */
1259 			audit_log_format(*ab, "%s", abuf);
1260 			len_rem -= len_tmp;
1261 			len_tmp = len_buf;
1262 			if (encode) {
1263 				if (len_abuf > len_rem)
1264 					len_tmp = len_rem / 2; /* encoding */
1265 				audit_log_n_hex(*ab, buf, len_tmp);
1266 				len_rem -= len_tmp * 2;
1267 				len_abuf -= len_tmp * 2;
1268 			} else {
1269 				if (len_abuf > len_rem)
1270 					len_tmp = len_rem - 2; /* quotes */
1271 				audit_log_n_string(*ab, buf, len_tmp);
1272 				len_rem -= len_tmp + 2;
1273 				/* don't subtract the "2" because we still need
1274 				 * to add quotes to the remaining string */
1275 				len_abuf -= len_tmp;
1276 			}
1277 			len_buf -= len_tmp;
1278 			buf += len_tmp;
1279 		}
1280 
1281 		/* ready to move to the next argument? */
1282 		if ((len_buf == 0) && !require_data) {
1283 			arg++;
1284 			iter = 0;
1285 			len_full = 0;
1286 			require_data = true;
1287 			encode = false;
1288 		}
1289 	} while (arg < context->execve.argc);
1290 
1291 	/* NOTE: the caller handles the final audit_log_end() call */
1292 
1293 out:
1294 	kfree(buf_head);
1295 }
1296 
1297 static void audit_log_cap(struct audit_buffer *ab, char *prefix,
1298 			  kernel_cap_t *cap)
1299 {
1300 	if (cap_isclear(*cap)) {
1301 		audit_log_format(ab, " %s=0", prefix);
1302 		return;
1303 	}
1304 	audit_log_format(ab, " %s=%016llx", prefix, cap->val);
1305 }
1306 
1307 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1308 {
1309 	if (name->fcap_ver == -1) {
1310 		audit_log_format(ab, " cap_fe=? cap_fver=? cap_fp=? cap_fi=?");
1311 		return;
1312 	}
1313 	audit_log_cap(ab, "cap_fp", &name->fcap.permitted);
1314 	audit_log_cap(ab, "cap_fi", &name->fcap.inheritable);
1315 	audit_log_format(ab, " cap_fe=%d cap_fver=%x cap_frootid=%d",
1316 			 name->fcap.fE, name->fcap_ver,
1317 			 from_kuid(&init_user_ns, name->fcap.rootid));
1318 }
1319 
1320 static void audit_log_time(struct audit_context *context, struct audit_buffer **ab)
1321 {
1322 	const struct audit_ntp_data *ntp = &context->time.ntp_data;
1323 	const struct timespec64 *tk = &context->time.tk_injoffset;
1324 	static const char * const ntp_name[] = {
1325 		"offset",
1326 		"freq",
1327 		"status",
1328 		"tai",
1329 		"tick",
1330 		"adjust",
1331 	};
1332 	int type;
1333 
1334 	if (context->type == AUDIT_TIME_ADJNTPVAL) {
1335 		for (type = 0; type < AUDIT_NTP_NVALS; type++) {
1336 			if (ntp->vals[type].newval != ntp->vals[type].oldval) {
1337 				if (!*ab) {
1338 					*ab = audit_log_start(context,
1339 							GFP_KERNEL,
1340 							AUDIT_TIME_ADJNTPVAL);
1341 					if (!*ab)
1342 						return;
1343 				}
1344 				audit_log_format(*ab, "op=%s old=%lli new=%lli",
1345 						 ntp_name[type],
1346 						 ntp->vals[type].oldval,
1347 						 ntp->vals[type].newval);
1348 				audit_log_end(*ab);
1349 				*ab = NULL;
1350 			}
1351 		}
1352 	}
1353 	if (tk->tv_sec != 0 || tk->tv_nsec != 0) {
1354 		if (!*ab) {
1355 			*ab = audit_log_start(context, GFP_KERNEL,
1356 					      AUDIT_TIME_INJOFFSET);
1357 			if (!*ab)
1358 				return;
1359 		}
1360 		audit_log_format(*ab, "sec=%lli nsec=%li",
1361 				 (long long)tk->tv_sec, tk->tv_nsec);
1362 		audit_log_end(*ab);
1363 		*ab = NULL;
1364 	}
1365 }
1366 
1367 static void show_special(struct audit_context *context, int *call_panic)
1368 {
1369 	struct audit_buffer *ab;
1370 	int i;
1371 
1372 	ab = audit_log_start(context, GFP_KERNEL, context->type);
1373 	if (!ab)
1374 		return;
1375 
1376 	switch (context->type) {
1377 	case AUDIT_SOCKETCALL: {
1378 		int nargs = context->socketcall.nargs;
1379 
1380 		audit_log_format(ab, "nargs=%d", nargs);
1381 		for (i = 0; i < nargs; i++)
1382 			audit_log_format(ab, " a%d=%lx", i,
1383 				context->socketcall.args[i]);
1384 		break; }
1385 	case AUDIT_IPC: {
1386 		u32 osid = context->ipc.osid;
1387 
1388 		audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1389 				 from_kuid(&init_user_ns, context->ipc.uid),
1390 				 from_kgid(&init_user_ns, context->ipc.gid),
1391 				 context->ipc.mode);
1392 		if (osid) {
1393 			char *ctx = NULL;
1394 			u32 len;
1395 
1396 			if (security_secid_to_secctx(osid, &ctx, &len)) {
1397 				audit_log_format(ab, " osid=%u", osid);
1398 				*call_panic = 1;
1399 			} else {
1400 				audit_log_format(ab, " obj=%s", ctx);
1401 				security_release_secctx(ctx, len);
1402 			}
1403 		}
1404 		if (context->ipc.has_perm) {
1405 			audit_log_end(ab);
1406 			ab = audit_log_start(context, GFP_KERNEL,
1407 					     AUDIT_IPC_SET_PERM);
1408 			if (unlikely(!ab))
1409 				return;
1410 			audit_log_format(ab,
1411 				"qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1412 				context->ipc.qbytes,
1413 				context->ipc.perm_uid,
1414 				context->ipc.perm_gid,
1415 				context->ipc.perm_mode);
1416 		}
1417 		break; }
1418 	case AUDIT_MQ_OPEN:
1419 		audit_log_format(ab,
1420 			"oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1421 			"mq_msgsize=%ld mq_curmsgs=%ld",
1422 			context->mq_open.oflag, context->mq_open.mode,
1423 			context->mq_open.attr.mq_flags,
1424 			context->mq_open.attr.mq_maxmsg,
1425 			context->mq_open.attr.mq_msgsize,
1426 			context->mq_open.attr.mq_curmsgs);
1427 		break;
1428 	case AUDIT_MQ_SENDRECV:
1429 		audit_log_format(ab,
1430 			"mqdes=%d msg_len=%zd msg_prio=%u "
1431 			"abs_timeout_sec=%lld abs_timeout_nsec=%ld",
1432 			context->mq_sendrecv.mqdes,
1433 			context->mq_sendrecv.msg_len,
1434 			context->mq_sendrecv.msg_prio,
1435 			(long long) context->mq_sendrecv.abs_timeout.tv_sec,
1436 			context->mq_sendrecv.abs_timeout.tv_nsec);
1437 		break;
1438 	case AUDIT_MQ_NOTIFY:
1439 		audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1440 				context->mq_notify.mqdes,
1441 				context->mq_notify.sigev_signo);
1442 		break;
1443 	case AUDIT_MQ_GETSETATTR: {
1444 		struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1445 
1446 		audit_log_format(ab,
1447 			"mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1448 			"mq_curmsgs=%ld ",
1449 			context->mq_getsetattr.mqdes,
1450 			attr->mq_flags, attr->mq_maxmsg,
1451 			attr->mq_msgsize, attr->mq_curmsgs);
1452 		break; }
1453 	case AUDIT_CAPSET:
1454 		audit_log_format(ab, "pid=%d", context->capset.pid);
1455 		audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1456 		audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1457 		audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1458 		audit_log_cap(ab, "cap_pa", &context->capset.cap.ambient);
1459 		break;
1460 	case AUDIT_MMAP:
1461 		audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1462 				 context->mmap.flags);
1463 		break;
1464 	case AUDIT_OPENAT2:
1465 		audit_log_format(ab, "oflag=0%llo mode=0%llo resolve=0x%llx",
1466 				 context->openat2.flags,
1467 				 context->openat2.mode,
1468 				 context->openat2.resolve);
1469 		break;
1470 	case AUDIT_EXECVE:
1471 		audit_log_execve_info(context, &ab);
1472 		break;
1473 	case AUDIT_KERN_MODULE:
1474 		audit_log_format(ab, "name=");
1475 		if (context->module.name) {
1476 			audit_log_untrustedstring(ab, context->module.name);
1477 		} else
1478 			audit_log_format(ab, "(null)");
1479 
1480 		break;
1481 	case AUDIT_TIME_ADJNTPVAL:
1482 	case AUDIT_TIME_INJOFFSET:
1483 		/* this call deviates from the rest, eating the buffer */
1484 		audit_log_time(context, &ab);
1485 		break;
1486 	}
1487 	audit_log_end(ab);
1488 }
1489 
1490 static inline int audit_proctitle_rtrim(char *proctitle, int len)
1491 {
1492 	char *end = proctitle + len - 1;
1493 
1494 	while (end > proctitle && !isprint(*end))
1495 		end--;
1496 
1497 	/* catch the case where proctitle is only 1 non-print character */
1498 	len = end - proctitle + 1;
1499 	len -= isprint(proctitle[len-1]) == 0;
1500 	return len;
1501 }
1502 
1503 /*
1504  * audit_log_name - produce AUDIT_PATH record from struct audit_names
1505  * @context: audit_context for the task
1506  * @n: audit_names structure with reportable details
1507  * @path: optional path to report instead of audit_names->name
1508  * @record_num: record number to report when handling a list of names
1509  * @call_panic: optional pointer to int that will be updated if secid fails
1510  */
1511 static void audit_log_name(struct audit_context *context, struct audit_names *n,
1512 		    const struct path *path, int record_num, int *call_panic)
1513 {
1514 	struct audit_buffer *ab;
1515 
1516 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1517 	if (!ab)
1518 		return;
1519 
1520 	audit_log_format(ab, "item=%d", record_num);
1521 
1522 	if (path)
1523 		audit_log_d_path(ab, " name=", path);
1524 	else if (n->name) {
1525 		switch (n->name_len) {
1526 		case AUDIT_NAME_FULL:
1527 			/* log the full path */
1528 			audit_log_format(ab, " name=");
1529 			audit_log_untrustedstring(ab, n->name->name);
1530 			break;
1531 		case 0:
1532 			/* name was specified as a relative path and the
1533 			 * directory component is the cwd
1534 			 */
1535 			if (context->pwd.dentry && context->pwd.mnt)
1536 				audit_log_d_path(ab, " name=", &context->pwd);
1537 			else
1538 				audit_log_format(ab, " name=(null)");
1539 			break;
1540 		default:
1541 			/* log the name's directory component */
1542 			audit_log_format(ab, " name=");
1543 			audit_log_n_untrustedstring(ab, n->name->name,
1544 						    n->name_len);
1545 		}
1546 	} else
1547 		audit_log_format(ab, " name=(null)");
1548 
1549 	if (n->ino != AUDIT_INO_UNSET)
1550 		audit_log_format(ab, " inode=%lu dev=%02x:%02x mode=%#ho ouid=%u ogid=%u rdev=%02x:%02x",
1551 				 n->ino,
1552 				 MAJOR(n->dev),
1553 				 MINOR(n->dev),
1554 				 n->mode,
1555 				 from_kuid(&init_user_ns, n->uid),
1556 				 from_kgid(&init_user_ns, n->gid),
1557 				 MAJOR(n->rdev),
1558 				 MINOR(n->rdev));
1559 	if (n->osid != 0) {
1560 		char *ctx = NULL;
1561 		u32 len;
1562 
1563 		if (security_secid_to_secctx(
1564 			n->osid, &ctx, &len)) {
1565 			audit_log_format(ab, " osid=%u", n->osid);
1566 			if (call_panic)
1567 				*call_panic = 2;
1568 		} else {
1569 			audit_log_format(ab, " obj=%s", ctx);
1570 			security_release_secctx(ctx, len);
1571 		}
1572 	}
1573 
1574 	/* log the audit_names record type */
1575 	switch (n->type) {
1576 	case AUDIT_TYPE_NORMAL:
1577 		audit_log_format(ab, " nametype=NORMAL");
1578 		break;
1579 	case AUDIT_TYPE_PARENT:
1580 		audit_log_format(ab, " nametype=PARENT");
1581 		break;
1582 	case AUDIT_TYPE_CHILD_DELETE:
1583 		audit_log_format(ab, " nametype=DELETE");
1584 		break;
1585 	case AUDIT_TYPE_CHILD_CREATE:
1586 		audit_log_format(ab, " nametype=CREATE");
1587 		break;
1588 	default:
1589 		audit_log_format(ab, " nametype=UNKNOWN");
1590 		break;
1591 	}
1592 
1593 	audit_log_fcaps(ab, n);
1594 	audit_log_end(ab);
1595 }
1596 
1597 static void audit_log_proctitle(void)
1598 {
1599 	int res;
1600 	char *buf;
1601 	char *msg = "(null)";
1602 	int len = strlen(msg);
1603 	struct audit_context *context = audit_context();
1604 	struct audit_buffer *ab;
1605 
1606 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PROCTITLE);
1607 	if (!ab)
1608 		return;	/* audit_panic or being filtered */
1609 
1610 	audit_log_format(ab, "proctitle=");
1611 
1612 	/* Not  cached */
1613 	if (!context->proctitle.value) {
1614 		buf = kmalloc(MAX_PROCTITLE_AUDIT_LEN, GFP_KERNEL);
1615 		if (!buf)
1616 			goto out;
1617 		/* Historically called this from procfs naming */
1618 		res = get_cmdline(current, buf, MAX_PROCTITLE_AUDIT_LEN);
1619 		if (res == 0) {
1620 			kfree(buf);
1621 			goto out;
1622 		}
1623 		res = audit_proctitle_rtrim(buf, res);
1624 		if (res == 0) {
1625 			kfree(buf);
1626 			goto out;
1627 		}
1628 		context->proctitle.value = buf;
1629 		context->proctitle.len = res;
1630 	}
1631 	msg = context->proctitle.value;
1632 	len = context->proctitle.len;
1633 out:
1634 	audit_log_n_untrustedstring(ab, msg, len);
1635 	audit_log_end(ab);
1636 }
1637 
1638 /**
1639  * audit_log_uring - generate a AUDIT_URINGOP record
1640  * @ctx: the audit context
1641  */
1642 static void audit_log_uring(struct audit_context *ctx)
1643 {
1644 	struct audit_buffer *ab;
1645 	const struct cred *cred;
1646 
1647 	ab = audit_log_start(ctx, GFP_ATOMIC, AUDIT_URINGOP);
1648 	if (!ab)
1649 		return;
1650 	cred = current_cred();
1651 	audit_log_format(ab, "uring_op=%d", ctx->uring_op);
1652 	if (ctx->return_valid != AUDITSC_INVALID)
1653 		audit_log_format(ab, " success=%s exit=%ld",
1654 				 (ctx->return_valid == AUDITSC_SUCCESS ?
1655 				  "yes" : "no"),
1656 				 ctx->return_code);
1657 	audit_log_format(ab,
1658 			 " items=%d"
1659 			 " ppid=%d pid=%d uid=%u gid=%u euid=%u suid=%u"
1660 			 " fsuid=%u egid=%u sgid=%u fsgid=%u",
1661 			 ctx->name_count,
1662 			 task_ppid_nr(current), task_tgid_nr(current),
1663 			 from_kuid(&init_user_ns, cred->uid),
1664 			 from_kgid(&init_user_ns, cred->gid),
1665 			 from_kuid(&init_user_ns, cred->euid),
1666 			 from_kuid(&init_user_ns, cred->suid),
1667 			 from_kuid(&init_user_ns, cred->fsuid),
1668 			 from_kgid(&init_user_ns, cred->egid),
1669 			 from_kgid(&init_user_ns, cred->sgid),
1670 			 from_kgid(&init_user_ns, cred->fsgid));
1671 	audit_log_task_context(ab);
1672 	audit_log_key(ab, ctx->filterkey);
1673 	audit_log_end(ab);
1674 }
1675 
1676 static void audit_log_exit(void)
1677 {
1678 	int i, call_panic = 0;
1679 	struct audit_context *context = audit_context();
1680 	struct audit_buffer *ab;
1681 	struct audit_aux_data *aux;
1682 	struct audit_names *n;
1683 
1684 	context->personality = current->personality;
1685 
1686 	switch (context->context) {
1687 	case AUDIT_CTX_SYSCALL:
1688 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1689 		if (!ab)
1690 			return;
1691 		audit_log_format(ab, "arch=%x syscall=%d",
1692 				 context->arch, context->major);
1693 		if (context->personality != PER_LINUX)
1694 			audit_log_format(ab, " per=%lx", context->personality);
1695 		if (context->return_valid != AUDITSC_INVALID)
1696 			audit_log_format(ab, " success=%s exit=%ld",
1697 					 (context->return_valid == AUDITSC_SUCCESS ?
1698 					  "yes" : "no"),
1699 					 context->return_code);
1700 		audit_log_format(ab,
1701 				 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d",
1702 				 context->argv[0],
1703 				 context->argv[1],
1704 				 context->argv[2],
1705 				 context->argv[3],
1706 				 context->name_count);
1707 		audit_log_task_info(ab);
1708 		audit_log_key(ab, context->filterkey);
1709 		audit_log_end(ab);
1710 		break;
1711 	case AUDIT_CTX_URING:
1712 		audit_log_uring(context);
1713 		break;
1714 	default:
1715 		BUG();
1716 		break;
1717 	}
1718 
1719 	for (aux = context->aux; aux; aux = aux->next) {
1720 
1721 		ab = audit_log_start(context, GFP_KERNEL, aux->type);
1722 		if (!ab)
1723 			continue; /* audit_panic has been called */
1724 
1725 		switch (aux->type) {
1726 
1727 		case AUDIT_BPRM_FCAPS: {
1728 			struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1729 
1730 			audit_log_format(ab, "fver=%x", axs->fcap_ver);
1731 			audit_log_cap(ab, "fp", &axs->fcap.permitted);
1732 			audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1733 			audit_log_format(ab, " fe=%d", axs->fcap.fE);
1734 			audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1735 			audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1736 			audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1737 			audit_log_cap(ab, "old_pa", &axs->old_pcap.ambient);
1738 			audit_log_cap(ab, "pp", &axs->new_pcap.permitted);
1739 			audit_log_cap(ab, "pi", &axs->new_pcap.inheritable);
1740 			audit_log_cap(ab, "pe", &axs->new_pcap.effective);
1741 			audit_log_cap(ab, "pa", &axs->new_pcap.ambient);
1742 			audit_log_format(ab, " frootid=%d",
1743 					 from_kuid(&init_user_ns,
1744 						   axs->fcap.rootid));
1745 			break; }
1746 
1747 		}
1748 		audit_log_end(ab);
1749 	}
1750 
1751 	if (context->type)
1752 		show_special(context, &call_panic);
1753 
1754 	if (context->fds[0] >= 0) {
1755 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1756 		if (ab) {
1757 			audit_log_format(ab, "fd0=%d fd1=%d",
1758 					context->fds[0], context->fds[1]);
1759 			audit_log_end(ab);
1760 		}
1761 	}
1762 
1763 	if (context->sockaddr_len) {
1764 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1765 		if (ab) {
1766 			audit_log_format(ab, "saddr=");
1767 			audit_log_n_hex(ab, (void *)context->sockaddr,
1768 					context->sockaddr_len);
1769 			audit_log_end(ab);
1770 		}
1771 	}
1772 
1773 	for (aux = context->aux_pids; aux; aux = aux->next) {
1774 		struct audit_aux_data_pids *axs = (void *)aux;
1775 
1776 		for (i = 0; i < axs->pid_count; i++)
1777 			if (audit_log_pid_context(context, axs->target_pid[i],
1778 						  axs->target_auid[i],
1779 						  axs->target_uid[i],
1780 						  axs->target_sessionid[i],
1781 						  axs->target_sid[i],
1782 						  axs->target_comm[i]))
1783 				call_panic = 1;
1784 	}
1785 
1786 	if (context->target_pid &&
1787 	    audit_log_pid_context(context, context->target_pid,
1788 				  context->target_auid, context->target_uid,
1789 				  context->target_sessionid,
1790 				  context->target_sid, context->target_comm))
1791 			call_panic = 1;
1792 
1793 	if (context->pwd.dentry && context->pwd.mnt) {
1794 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1795 		if (ab) {
1796 			audit_log_d_path(ab, "cwd=", &context->pwd);
1797 			audit_log_end(ab);
1798 		}
1799 	}
1800 
1801 	i = 0;
1802 	list_for_each_entry(n, &context->names_list, list) {
1803 		if (n->hidden)
1804 			continue;
1805 		audit_log_name(context, n, NULL, i++, &call_panic);
1806 	}
1807 
1808 	if (context->context == AUDIT_CTX_SYSCALL)
1809 		audit_log_proctitle();
1810 
1811 	/* Send end of event record to help user space know we are finished */
1812 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1813 	if (ab)
1814 		audit_log_end(ab);
1815 	if (call_panic)
1816 		audit_panic("error in audit_log_exit()");
1817 }
1818 
1819 /**
1820  * __audit_free - free a per-task audit context
1821  * @tsk: task whose audit context block to free
1822  *
1823  * Called from copy_process, do_exit, and the io_uring code
1824  */
1825 void __audit_free(struct task_struct *tsk)
1826 {
1827 	struct audit_context *context = tsk->audit_context;
1828 
1829 	if (!context)
1830 		return;
1831 
1832 	/* this may generate CONFIG_CHANGE records */
1833 	if (!list_empty(&context->killed_trees))
1834 		audit_kill_trees(context);
1835 
1836 	/* We are called either by do_exit() or the fork() error handling code;
1837 	 * in the former case tsk == current and in the latter tsk is a
1838 	 * random task_struct that doesn't have any meaningful data we
1839 	 * need to log via audit_log_exit().
1840 	 */
1841 	if (tsk == current && !context->dummy) {
1842 		context->return_valid = AUDITSC_INVALID;
1843 		context->return_code = 0;
1844 		if (context->context == AUDIT_CTX_SYSCALL) {
1845 			audit_filter_syscall(tsk, context);
1846 			audit_filter_inodes(tsk, context);
1847 			if (context->current_state == AUDIT_STATE_RECORD)
1848 				audit_log_exit();
1849 		} else if (context->context == AUDIT_CTX_URING) {
1850 			/* TODO: verify this case is real and valid */
1851 			audit_filter_uring(tsk, context);
1852 			audit_filter_inodes(tsk, context);
1853 			if (context->current_state == AUDIT_STATE_RECORD)
1854 				audit_log_uring(context);
1855 		}
1856 	}
1857 
1858 	audit_set_context(tsk, NULL);
1859 	audit_free_context(context);
1860 }
1861 
1862 /**
1863  * audit_return_fixup - fixup the return codes in the audit_context
1864  * @ctx: the audit_context
1865  * @success: true/false value to indicate if the operation succeeded or not
1866  * @code: operation return code
1867  *
1868  * We need to fixup the return code in the audit logs if the actual return
1869  * codes are later going to be fixed by the arch specific signal handlers.
1870  */
1871 static void audit_return_fixup(struct audit_context *ctx,
1872 			       int success, long code)
1873 {
1874 	/*
1875 	 * This is actually a test for:
1876 	 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
1877 	 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
1878 	 *
1879 	 * but is faster than a bunch of ||
1880 	 */
1881 	if (unlikely(code <= -ERESTARTSYS) &&
1882 	    (code >= -ERESTART_RESTARTBLOCK) &&
1883 	    (code != -ENOIOCTLCMD))
1884 		ctx->return_code = -EINTR;
1885 	else
1886 		ctx->return_code  = code;
1887 	ctx->return_valid = (success ? AUDITSC_SUCCESS : AUDITSC_FAILURE);
1888 }
1889 
1890 /**
1891  * __audit_uring_entry - prepare the kernel task's audit context for io_uring
1892  * @op: the io_uring opcode
1893  *
1894  * This is similar to audit_syscall_entry() but is intended for use by io_uring
1895  * operations.  This function should only ever be called from
1896  * audit_uring_entry() as we rely on the audit context checking present in that
1897  * function.
1898  */
1899 void __audit_uring_entry(u8 op)
1900 {
1901 	struct audit_context *ctx = audit_context();
1902 
1903 	if (ctx->state == AUDIT_STATE_DISABLED)
1904 		return;
1905 
1906 	/*
1907 	 * NOTE: It's possible that we can be called from the process' context
1908 	 *       before it returns to userspace, and before audit_syscall_exit()
1909 	 *       is called.  In this case there is not much to do, just record
1910 	 *       the io_uring details and return.
1911 	 */
1912 	ctx->uring_op = op;
1913 	if (ctx->context == AUDIT_CTX_SYSCALL)
1914 		return;
1915 
1916 	ctx->dummy = !audit_n_rules;
1917 	if (!ctx->dummy && ctx->state == AUDIT_STATE_BUILD)
1918 		ctx->prio = 0;
1919 
1920 	ctx->context = AUDIT_CTX_URING;
1921 	ctx->current_state = ctx->state;
1922 	ktime_get_coarse_real_ts64(&ctx->ctime);
1923 }
1924 
1925 /**
1926  * __audit_uring_exit - wrap up the kernel task's audit context after io_uring
1927  * @success: true/false value to indicate if the operation succeeded or not
1928  * @code: operation return code
1929  *
1930  * This is similar to audit_syscall_exit() but is intended for use by io_uring
1931  * operations.  This function should only ever be called from
1932  * audit_uring_exit() as we rely on the audit context checking present in that
1933  * function.
1934  */
1935 void __audit_uring_exit(int success, long code)
1936 {
1937 	struct audit_context *ctx = audit_context();
1938 
1939 	if (ctx->dummy) {
1940 		if (ctx->context != AUDIT_CTX_URING)
1941 			return;
1942 		goto out;
1943 	}
1944 
1945 	audit_return_fixup(ctx, success, code);
1946 	if (ctx->context == AUDIT_CTX_SYSCALL) {
1947 		/*
1948 		 * NOTE: See the note in __audit_uring_entry() about the case
1949 		 *       where we may be called from process context before we
1950 		 *       return to userspace via audit_syscall_exit().  In this
1951 		 *       case we simply emit a URINGOP record and bail, the
1952 		 *       normal syscall exit handling will take care of
1953 		 *       everything else.
1954 		 *       It is also worth mentioning that when we are called,
1955 		 *       the current process creds may differ from the creds
1956 		 *       used during the normal syscall processing; keep that
1957 		 *       in mind if/when we move the record generation code.
1958 		 */
1959 
1960 		/*
1961 		 * We need to filter on the syscall info here to decide if we
1962 		 * should emit a URINGOP record.  I know it seems odd but this
1963 		 * solves the problem where users have a filter to block *all*
1964 		 * syscall records in the "exit" filter; we want to preserve
1965 		 * the behavior here.
1966 		 */
1967 		audit_filter_syscall(current, ctx);
1968 		if (ctx->current_state != AUDIT_STATE_RECORD)
1969 			audit_filter_uring(current, ctx);
1970 		audit_filter_inodes(current, ctx);
1971 		if (ctx->current_state != AUDIT_STATE_RECORD)
1972 			return;
1973 
1974 		audit_log_uring(ctx);
1975 		return;
1976 	}
1977 
1978 	/* this may generate CONFIG_CHANGE records */
1979 	if (!list_empty(&ctx->killed_trees))
1980 		audit_kill_trees(ctx);
1981 
1982 	/* run through both filters to ensure we set the filterkey properly */
1983 	audit_filter_uring(current, ctx);
1984 	audit_filter_inodes(current, ctx);
1985 	if (ctx->current_state != AUDIT_STATE_RECORD)
1986 		goto out;
1987 	audit_log_exit();
1988 
1989 out:
1990 	audit_reset_context(ctx);
1991 }
1992 
1993 /**
1994  * __audit_syscall_entry - fill in an audit record at syscall entry
1995  * @major: major syscall type (function)
1996  * @a1: additional syscall register 1
1997  * @a2: additional syscall register 2
1998  * @a3: additional syscall register 3
1999  * @a4: additional syscall register 4
2000  *
2001  * Fill in audit context at syscall entry.  This only happens if the
2002  * audit context was created when the task was created and the state or
2003  * filters demand the audit context be built.  If the state from the
2004  * per-task filter or from the per-syscall filter is AUDIT_STATE_RECORD,
2005  * then the record will be written at syscall exit time (otherwise, it
2006  * will only be written if another part of the kernel requests that it
2007  * be written).
2008  */
2009 void __audit_syscall_entry(int major, unsigned long a1, unsigned long a2,
2010 			   unsigned long a3, unsigned long a4)
2011 {
2012 	struct audit_context *context = audit_context();
2013 	enum audit_state     state;
2014 
2015 	if (!audit_enabled || !context)
2016 		return;
2017 
2018 	WARN_ON(context->context != AUDIT_CTX_UNUSED);
2019 	WARN_ON(context->name_count);
2020 	if (context->context != AUDIT_CTX_UNUSED || context->name_count) {
2021 		audit_panic("unrecoverable error in audit_syscall_entry()");
2022 		return;
2023 	}
2024 
2025 	state = context->state;
2026 	if (state == AUDIT_STATE_DISABLED)
2027 		return;
2028 
2029 	context->dummy = !audit_n_rules;
2030 	if (!context->dummy && state == AUDIT_STATE_BUILD) {
2031 		context->prio = 0;
2032 		if (auditd_test_task(current))
2033 			return;
2034 	}
2035 
2036 	context->arch	    = syscall_get_arch(current);
2037 	context->major      = major;
2038 	context->argv[0]    = a1;
2039 	context->argv[1]    = a2;
2040 	context->argv[2]    = a3;
2041 	context->argv[3]    = a4;
2042 	context->context = AUDIT_CTX_SYSCALL;
2043 	context->current_state  = state;
2044 	ktime_get_coarse_real_ts64(&context->ctime);
2045 }
2046 
2047 /**
2048  * __audit_syscall_exit - deallocate audit context after a system call
2049  * @success: success value of the syscall
2050  * @return_code: return value of the syscall
2051  *
2052  * Tear down after system call.  If the audit context has been marked as
2053  * auditable (either because of the AUDIT_STATE_RECORD state from
2054  * filtering, or because some other part of the kernel wrote an audit
2055  * message), then write out the syscall information.  In call cases,
2056  * free the names stored from getname().
2057  */
2058 void __audit_syscall_exit(int success, long return_code)
2059 {
2060 	struct audit_context *context = audit_context();
2061 
2062 	if (!context || context->dummy ||
2063 	    context->context != AUDIT_CTX_SYSCALL)
2064 		goto out;
2065 
2066 	/* this may generate CONFIG_CHANGE records */
2067 	if (!list_empty(&context->killed_trees))
2068 		audit_kill_trees(context);
2069 
2070 	audit_return_fixup(context, success, return_code);
2071 	/* run through both filters to ensure we set the filterkey properly */
2072 	audit_filter_syscall(current, context);
2073 	audit_filter_inodes(current, context);
2074 	if (context->current_state != AUDIT_STATE_RECORD)
2075 		goto out;
2076 
2077 	audit_log_exit();
2078 
2079 out:
2080 	audit_reset_context(context);
2081 }
2082 
2083 static inline void handle_one(const struct inode *inode)
2084 {
2085 	struct audit_context *context;
2086 	struct audit_tree_refs *p;
2087 	struct audit_chunk *chunk;
2088 	int count;
2089 
2090 	if (likely(!inode->i_fsnotify_marks))
2091 		return;
2092 	context = audit_context();
2093 	p = context->trees;
2094 	count = context->tree_count;
2095 	rcu_read_lock();
2096 	chunk = audit_tree_lookup(inode);
2097 	rcu_read_unlock();
2098 	if (!chunk)
2099 		return;
2100 	if (likely(put_tree_ref(context, chunk)))
2101 		return;
2102 	if (unlikely(!grow_tree_refs(context))) {
2103 		pr_warn("out of memory, audit has lost a tree reference\n");
2104 		audit_set_auditable(context);
2105 		audit_put_chunk(chunk);
2106 		unroll_tree_refs(context, p, count);
2107 		return;
2108 	}
2109 	put_tree_ref(context, chunk);
2110 }
2111 
2112 static void handle_path(const struct dentry *dentry)
2113 {
2114 	struct audit_context *context;
2115 	struct audit_tree_refs *p;
2116 	const struct dentry *d, *parent;
2117 	struct audit_chunk *drop;
2118 	unsigned long seq;
2119 	int count;
2120 
2121 	context = audit_context();
2122 	p = context->trees;
2123 	count = context->tree_count;
2124 retry:
2125 	drop = NULL;
2126 	d = dentry;
2127 	rcu_read_lock();
2128 	seq = read_seqbegin(&rename_lock);
2129 	for (;;) {
2130 		struct inode *inode = d_backing_inode(d);
2131 
2132 		if (inode && unlikely(inode->i_fsnotify_marks)) {
2133 			struct audit_chunk *chunk;
2134 
2135 			chunk = audit_tree_lookup(inode);
2136 			if (chunk) {
2137 				if (unlikely(!put_tree_ref(context, chunk))) {
2138 					drop = chunk;
2139 					break;
2140 				}
2141 			}
2142 		}
2143 		parent = d->d_parent;
2144 		if (parent == d)
2145 			break;
2146 		d = parent;
2147 	}
2148 	if (unlikely(read_seqretry(&rename_lock, seq) || drop)) {  /* in this order */
2149 		rcu_read_unlock();
2150 		if (!drop) {
2151 			/* just a race with rename */
2152 			unroll_tree_refs(context, p, count);
2153 			goto retry;
2154 		}
2155 		audit_put_chunk(drop);
2156 		if (grow_tree_refs(context)) {
2157 			/* OK, got more space */
2158 			unroll_tree_refs(context, p, count);
2159 			goto retry;
2160 		}
2161 		/* too bad */
2162 		pr_warn("out of memory, audit has lost a tree reference\n");
2163 		unroll_tree_refs(context, p, count);
2164 		audit_set_auditable(context);
2165 		return;
2166 	}
2167 	rcu_read_unlock();
2168 }
2169 
2170 static struct audit_names *audit_alloc_name(struct audit_context *context,
2171 						unsigned char type)
2172 {
2173 	struct audit_names *aname;
2174 
2175 	if (context->name_count < AUDIT_NAMES) {
2176 		aname = &context->preallocated_names[context->name_count];
2177 		memset(aname, 0, sizeof(*aname));
2178 	} else {
2179 		aname = kzalloc(sizeof(*aname), GFP_NOFS);
2180 		if (!aname)
2181 			return NULL;
2182 		aname->should_free = true;
2183 	}
2184 
2185 	aname->ino = AUDIT_INO_UNSET;
2186 	aname->type = type;
2187 	list_add_tail(&aname->list, &context->names_list);
2188 
2189 	context->name_count++;
2190 	if (!context->pwd.dentry)
2191 		get_fs_pwd(current->fs, &context->pwd);
2192 	return aname;
2193 }
2194 
2195 /**
2196  * __audit_reusename - fill out filename with info from existing entry
2197  * @uptr: userland ptr to pathname
2198  *
2199  * Search the audit_names list for the current audit context. If there is an
2200  * existing entry with a matching "uptr" then return the filename
2201  * associated with that audit_name. If not, return NULL.
2202  */
2203 struct filename *
2204 __audit_reusename(const __user char *uptr)
2205 {
2206 	struct audit_context *context = audit_context();
2207 	struct audit_names *n;
2208 
2209 	list_for_each_entry(n, &context->names_list, list) {
2210 		if (!n->name)
2211 			continue;
2212 		if (n->name->uptr == uptr) {
2213 			n->name->refcnt++;
2214 			return n->name;
2215 		}
2216 	}
2217 	return NULL;
2218 }
2219 
2220 /**
2221  * __audit_getname - add a name to the list
2222  * @name: name to add
2223  *
2224  * Add a name to the list of audit names for this context.
2225  * Called from fs/namei.c:getname().
2226  */
2227 void __audit_getname(struct filename *name)
2228 {
2229 	struct audit_context *context = audit_context();
2230 	struct audit_names *n;
2231 
2232 	if (context->context == AUDIT_CTX_UNUSED)
2233 		return;
2234 
2235 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
2236 	if (!n)
2237 		return;
2238 
2239 	n->name = name;
2240 	n->name_len = AUDIT_NAME_FULL;
2241 	name->aname = n;
2242 	name->refcnt++;
2243 }
2244 
2245 static inline int audit_copy_fcaps(struct audit_names *name,
2246 				   const struct dentry *dentry)
2247 {
2248 	struct cpu_vfs_cap_data caps;
2249 	int rc;
2250 
2251 	if (!dentry)
2252 		return 0;
2253 
2254 	rc = get_vfs_caps_from_disk(&nop_mnt_idmap, dentry, &caps);
2255 	if (rc)
2256 		return rc;
2257 
2258 	name->fcap.permitted = caps.permitted;
2259 	name->fcap.inheritable = caps.inheritable;
2260 	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2261 	name->fcap.rootid = caps.rootid;
2262 	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >>
2263 				VFS_CAP_REVISION_SHIFT;
2264 
2265 	return 0;
2266 }
2267 
2268 /* Copy inode data into an audit_names. */
2269 static void audit_copy_inode(struct audit_names *name,
2270 			     const struct dentry *dentry,
2271 			     struct inode *inode, unsigned int flags)
2272 {
2273 	name->ino   = inode->i_ino;
2274 	name->dev   = inode->i_sb->s_dev;
2275 	name->mode  = inode->i_mode;
2276 	name->uid   = inode->i_uid;
2277 	name->gid   = inode->i_gid;
2278 	name->rdev  = inode->i_rdev;
2279 	security_inode_getsecid(inode, &name->osid);
2280 	if (flags & AUDIT_INODE_NOEVAL) {
2281 		name->fcap_ver = -1;
2282 		return;
2283 	}
2284 	audit_copy_fcaps(name, dentry);
2285 }
2286 
2287 /**
2288  * __audit_inode - store the inode and device from a lookup
2289  * @name: name being audited
2290  * @dentry: dentry being audited
2291  * @flags: attributes for this particular entry
2292  */
2293 void __audit_inode(struct filename *name, const struct dentry *dentry,
2294 		   unsigned int flags)
2295 {
2296 	struct audit_context *context = audit_context();
2297 	struct inode *inode = d_backing_inode(dentry);
2298 	struct audit_names *n;
2299 	bool parent = flags & AUDIT_INODE_PARENT;
2300 	struct audit_entry *e;
2301 	struct list_head *list = &audit_filter_list[AUDIT_FILTER_FS];
2302 	int i;
2303 
2304 	if (context->context == AUDIT_CTX_UNUSED)
2305 		return;
2306 
2307 	rcu_read_lock();
2308 	list_for_each_entry_rcu(e, list, list) {
2309 		for (i = 0; i < e->rule.field_count; i++) {
2310 			struct audit_field *f = &e->rule.fields[i];
2311 
2312 			if (f->type == AUDIT_FSTYPE
2313 			    && audit_comparator(inode->i_sb->s_magic,
2314 						f->op, f->val)
2315 			    && e->rule.action == AUDIT_NEVER) {
2316 				rcu_read_unlock();
2317 				return;
2318 			}
2319 		}
2320 	}
2321 	rcu_read_unlock();
2322 
2323 	if (!name)
2324 		goto out_alloc;
2325 
2326 	/*
2327 	 * If we have a pointer to an audit_names entry already, then we can
2328 	 * just use it directly if the type is correct.
2329 	 */
2330 	n = name->aname;
2331 	if (n) {
2332 		if (parent) {
2333 			if (n->type == AUDIT_TYPE_PARENT ||
2334 			    n->type == AUDIT_TYPE_UNKNOWN)
2335 				goto out;
2336 		} else {
2337 			if (n->type != AUDIT_TYPE_PARENT)
2338 				goto out;
2339 		}
2340 	}
2341 
2342 	list_for_each_entry_reverse(n, &context->names_list, list) {
2343 		if (n->ino) {
2344 			/* valid inode number, use that for the comparison */
2345 			if (n->ino != inode->i_ino ||
2346 			    n->dev != inode->i_sb->s_dev)
2347 				continue;
2348 		} else if (n->name) {
2349 			/* inode number has not been set, check the name */
2350 			if (strcmp(n->name->name, name->name))
2351 				continue;
2352 		} else
2353 			/* no inode and no name (?!) ... this is odd ... */
2354 			continue;
2355 
2356 		/* match the correct record type */
2357 		if (parent) {
2358 			if (n->type == AUDIT_TYPE_PARENT ||
2359 			    n->type == AUDIT_TYPE_UNKNOWN)
2360 				goto out;
2361 		} else {
2362 			if (n->type != AUDIT_TYPE_PARENT)
2363 				goto out;
2364 		}
2365 	}
2366 
2367 out_alloc:
2368 	/* unable to find an entry with both a matching name and type */
2369 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
2370 	if (!n)
2371 		return;
2372 	if (name) {
2373 		n->name = name;
2374 		name->refcnt++;
2375 	}
2376 
2377 out:
2378 	if (parent) {
2379 		n->name_len = n->name ? parent_len(n->name->name) : AUDIT_NAME_FULL;
2380 		n->type = AUDIT_TYPE_PARENT;
2381 		if (flags & AUDIT_INODE_HIDDEN)
2382 			n->hidden = true;
2383 	} else {
2384 		n->name_len = AUDIT_NAME_FULL;
2385 		n->type = AUDIT_TYPE_NORMAL;
2386 	}
2387 	handle_path(dentry);
2388 	audit_copy_inode(n, dentry, inode, flags & AUDIT_INODE_NOEVAL);
2389 }
2390 
2391 void __audit_file(const struct file *file)
2392 {
2393 	__audit_inode(NULL, file->f_path.dentry, 0);
2394 }
2395 
2396 /**
2397  * __audit_inode_child - collect inode info for created/removed objects
2398  * @parent: inode of dentry parent
2399  * @dentry: dentry being audited
2400  * @type:   AUDIT_TYPE_* value that we're looking for
2401  *
2402  * For syscalls that create or remove filesystem objects, audit_inode
2403  * can only collect information for the filesystem object's parent.
2404  * This call updates the audit context with the child's information.
2405  * Syscalls that create a new filesystem object must be hooked after
2406  * the object is created.  Syscalls that remove a filesystem object
2407  * must be hooked prior, in order to capture the target inode during
2408  * unsuccessful attempts.
2409  */
2410 void __audit_inode_child(struct inode *parent,
2411 			 const struct dentry *dentry,
2412 			 const unsigned char type)
2413 {
2414 	struct audit_context *context = audit_context();
2415 	struct inode *inode = d_backing_inode(dentry);
2416 	const struct qstr *dname = &dentry->d_name;
2417 	struct audit_names *n, *found_parent = NULL, *found_child = NULL;
2418 	struct audit_entry *e;
2419 	struct list_head *list = &audit_filter_list[AUDIT_FILTER_FS];
2420 	int i;
2421 
2422 	if (context->context == AUDIT_CTX_UNUSED)
2423 		return;
2424 
2425 	rcu_read_lock();
2426 	list_for_each_entry_rcu(e, list, list) {
2427 		for (i = 0; i < e->rule.field_count; i++) {
2428 			struct audit_field *f = &e->rule.fields[i];
2429 
2430 			if (f->type == AUDIT_FSTYPE
2431 			    && audit_comparator(parent->i_sb->s_magic,
2432 						f->op, f->val)
2433 			    && e->rule.action == AUDIT_NEVER) {
2434 				rcu_read_unlock();
2435 				return;
2436 			}
2437 		}
2438 	}
2439 	rcu_read_unlock();
2440 
2441 	if (inode)
2442 		handle_one(inode);
2443 
2444 	/* look for a parent entry first */
2445 	list_for_each_entry(n, &context->names_list, list) {
2446 		if (!n->name ||
2447 		    (n->type != AUDIT_TYPE_PARENT &&
2448 		     n->type != AUDIT_TYPE_UNKNOWN))
2449 			continue;
2450 
2451 		if (n->ino == parent->i_ino && n->dev == parent->i_sb->s_dev &&
2452 		    !audit_compare_dname_path(dname,
2453 					      n->name->name, n->name_len)) {
2454 			if (n->type == AUDIT_TYPE_UNKNOWN)
2455 				n->type = AUDIT_TYPE_PARENT;
2456 			found_parent = n;
2457 			break;
2458 		}
2459 	}
2460 
2461 	cond_resched();
2462 
2463 	/* is there a matching child entry? */
2464 	list_for_each_entry(n, &context->names_list, list) {
2465 		/* can only match entries that have a name */
2466 		if (!n->name ||
2467 		    (n->type != type && n->type != AUDIT_TYPE_UNKNOWN))
2468 			continue;
2469 
2470 		if (!strcmp(dname->name, n->name->name) ||
2471 		    !audit_compare_dname_path(dname, n->name->name,
2472 						found_parent ?
2473 						found_parent->name_len :
2474 						AUDIT_NAME_FULL)) {
2475 			if (n->type == AUDIT_TYPE_UNKNOWN)
2476 				n->type = type;
2477 			found_child = n;
2478 			break;
2479 		}
2480 	}
2481 
2482 	if (!found_parent) {
2483 		/* create a new, "anonymous" parent record */
2484 		n = audit_alloc_name(context, AUDIT_TYPE_PARENT);
2485 		if (!n)
2486 			return;
2487 		audit_copy_inode(n, NULL, parent, 0);
2488 	}
2489 
2490 	if (!found_child) {
2491 		found_child = audit_alloc_name(context, type);
2492 		if (!found_child)
2493 			return;
2494 
2495 		/* Re-use the name belonging to the slot for a matching parent
2496 		 * directory. All names for this context are relinquished in
2497 		 * audit_free_names() */
2498 		if (found_parent) {
2499 			found_child->name = found_parent->name;
2500 			found_child->name_len = AUDIT_NAME_FULL;
2501 			found_child->name->refcnt++;
2502 		}
2503 	}
2504 
2505 	if (inode)
2506 		audit_copy_inode(found_child, dentry, inode, 0);
2507 	else
2508 		found_child->ino = AUDIT_INO_UNSET;
2509 }
2510 EXPORT_SYMBOL_GPL(__audit_inode_child);
2511 
2512 /**
2513  * auditsc_get_stamp - get local copies of audit_context values
2514  * @ctx: audit_context for the task
2515  * @t: timespec64 to store time recorded in the audit_context
2516  * @serial: serial value that is recorded in the audit_context
2517  *
2518  * Also sets the context as auditable.
2519  */
2520 int auditsc_get_stamp(struct audit_context *ctx,
2521 		       struct timespec64 *t, unsigned int *serial)
2522 {
2523 	if (ctx->context == AUDIT_CTX_UNUSED)
2524 		return 0;
2525 	if (!ctx->serial)
2526 		ctx->serial = audit_serial();
2527 	t->tv_sec  = ctx->ctime.tv_sec;
2528 	t->tv_nsec = ctx->ctime.tv_nsec;
2529 	*serial    = ctx->serial;
2530 	if (!ctx->prio) {
2531 		ctx->prio = 1;
2532 		ctx->current_state = AUDIT_STATE_RECORD;
2533 	}
2534 	return 1;
2535 }
2536 
2537 /**
2538  * __audit_mq_open - record audit data for a POSIX MQ open
2539  * @oflag: open flag
2540  * @mode: mode bits
2541  * @attr: queue attributes
2542  *
2543  */
2544 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2545 {
2546 	struct audit_context *context = audit_context();
2547 
2548 	if (attr)
2549 		memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2550 	else
2551 		memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2552 
2553 	context->mq_open.oflag = oflag;
2554 	context->mq_open.mode = mode;
2555 
2556 	context->type = AUDIT_MQ_OPEN;
2557 }
2558 
2559 /**
2560  * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2561  * @mqdes: MQ descriptor
2562  * @msg_len: Message length
2563  * @msg_prio: Message priority
2564  * @abs_timeout: Message timeout in absolute time
2565  *
2566  */
2567 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2568 			const struct timespec64 *abs_timeout)
2569 {
2570 	struct audit_context *context = audit_context();
2571 	struct timespec64 *p = &context->mq_sendrecv.abs_timeout;
2572 
2573 	if (abs_timeout)
2574 		memcpy(p, abs_timeout, sizeof(*p));
2575 	else
2576 		memset(p, 0, sizeof(*p));
2577 
2578 	context->mq_sendrecv.mqdes = mqdes;
2579 	context->mq_sendrecv.msg_len = msg_len;
2580 	context->mq_sendrecv.msg_prio = msg_prio;
2581 
2582 	context->type = AUDIT_MQ_SENDRECV;
2583 }
2584 
2585 /**
2586  * __audit_mq_notify - record audit data for a POSIX MQ notify
2587  * @mqdes: MQ descriptor
2588  * @notification: Notification event
2589  *
2590  */
2591 
2592 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2593 {
2594 	struct audit_context *context = audit_context();
2595 
2596 	if (notification)
2597 		context->mq_notify.sigev_signo = notification->sigev_signo;
2598 	else
2599 		context->mq_notify.sigev_signo = 0;
2600 
2601 	context->mq_notify.mqdes = mqdes;
2602 	context->type = AUDIT_MQ_NOTIFY;
2603 }
2604 
2605 /**
2606  * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2607  * @mqdes: MQ descriptor
2608  * @mqstat: MQ flags
2609  *
2610  */
2611 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2612 {
2613 	struct audit_context *context = audit_context();
2614 
2615 	context->mq_getsetattr.mqdes = mqdes;
2616 	context->mq_getsetattr.mqstat = *mqstat;
2617 	context->type = AUDIT_MQ_GETSETATTR;
2618 }
2619 
2620 /**
2621  * __audit_ipc_obj - record audit data for ipc object
2622  * @ipcp: ipc permissions
2623  *
2624  */
2625 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2626 {
2627 	struct audit_context *context = audit_context();
2628 
2629 	context->ipc.uid = ipcp->uid;
2630 	context->ipc.gid = ipcp->gid;
2631 	context->ipc.mode = ipcp->mode;
2632 	context->ipc.has_perm = 0;
2633 	security_ipc_getsecid(ipcp, &context->ipc.osid);
2634 	context->type = AUDIT_IPC;
2635 }
2636 
2637 /**
2638  * __audit_ipc_set_perm - record audit data for new ipc permissions
2639  * @qbytes: msgq bytes
2640  * @uid: msgq user id
2641  * @gid: msgq group id
2642  * @mode: msgq mode (permissions)
2643  *
2644  * Called only after audit_ipc_obj().
2645  */
2646 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2647 {
2648 	struct audit_context *context = audit_context();
2649 
2650 	context->ipc.qbytes = qbytes;
2651 	context->ipc.perm_uid = uid;
2652 	context->ipc.perm_gid = gid;
2653 	context->ipc.perm_mode = mode;
2654 	context->ipc.has_perm = 1;
2655 }
2656 
2657 void __audit_bprm(struct linux_binprm *bprm)
2658 {
2659 	struct audit_context *context = audit_context();
2660 
2661 	context->type = AUDIT_EXECVE;
2662 	context->execve.argc = bprm->argc;
2663 }
2664 
2665 
2666 /**
2667  * __audit_socketcall - record audit data for sys_socketcall
2668  * @nargs: number of args, which should not be more than AUDITSC_ARGS.
2669  * @args: args array
2670  *
2671  */
2672 int __audit_socketcall(int nargs, unsigned long *args)
2673 {
2674 	struct audit_context *context = audit_context();
2675 
2676 	if (nargs <= 0 || nargs > AUDITSC_ARGS || !args)
2677 		return -EINVAL;
2678 	context->type = AUDIT_SOCKETCALL;
2679 	context->socketcall.nargs = nargs;
2680 	memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2681 	return 0;
2682 }
2683 
2684 /**
2685  * __audit_fd_pair - record audit data for pipe and socketpair
2686  * @fd1: the first file descriptor
2687  * @fd2: the second file descriptor
2688  *
2689  */
2690 void __audit_fd_pair(int fd1, int fd2)
2691 {
2692 	struct audit_context *context = audit_context();
2693 
2694 	context->fds[0] = fd1;
2695 	context->fds[1] = fd2;
2696 }
2697 
2698 /**
2699  * __audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2700  * @len: data length in user space
2701  * @a: data address in kernel space
2702  *
2703  * Returns 0 for success or NULL context or < 0 on error.
2704  */
2705 int __audit_sockaddr(int len, void *a)
2706 {
2707 	struct audit_context *context = audit_context();
2708 
2709 	if (!context->sockaddr) {
2710 		void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2711 
2712 		if (!p)
2713 			return -ENOMEM;
2714 		context->sockaddr = p;
2715 	}
2716 
2717 	context->sockaddr_len = len;
2718 	memcpy(context->sockaddr, a, len);
2719 	return 0;
2720 }
2721 
2722 void __audit_ptrace(struct task_struct *t)
2723 {
2724 	struct audit_context *context = audit_context();
2725 
2726 	context->target_pid = task_tgid_nr(t);
2727 	context->target_auid = audit_get_loginuid(t);
2728 	context->target_uid = task_uid(t);
2729 	context->target_sessionid = audit_get_sessionid(t);
2730 	security_task_getsecid_obj(t, &context->target_sid);
2731 	memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2732 }
2733 
2734 /**
2735  * audit_signal_info_syscall - record signal info for syscalls
2736  * @t: task being signaled
2737  *
2738  * If the audit subsystem is being terminated, record the task (pid)
2739  * and uid that is doing that.
2740  */
2741 int audit_signal_info_syscall(struct task_struct *t)
2742 {
2743 	struct audit_aux_data_pids *axp;
2744 	struct audit_context *ctx = audit_context();
2745 	kuid_t t_uid = task_uid(t);
2746 
2747 	if (!audit_signals || audit_dummy_context())
2748 		return 0;
2749 
2750 	/* optimize the common case by putting first signal recipient directly
2751 	 * in audit_context */
2752 	if (!ctx->target_pid) {
2753 		ctx->target_pid = task_tgid_nr(t);
2754 		ctx->target_auid = audit_get_loginuid(t);
2755 		ctx->target_uid = t_uid;
2756 		ctx->target_sessionid = audit_get_sessionid(t);
2757 		security_task_getsecid_obj(t, &ctx->target_sid);
2758 		memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2759 		return 0;
2760 	}
2761 
2762 	axp = (void *)ctx->aux_pids;
2763 	if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2764 		axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2765 		if (!axp)
2766 			return -ENOMEM;
2767 
2768 		axp->d.type = AUDIT_OBJ_PID;
2769 		axp->d.next = ctx->aux_pids;
2770 		ctx->aux_pids = (void *)axp;
2771 	}
2772 	BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2773 
2774 	axp->target_pid[axp->pid_count] = task_tgid_nr(t);
2775 	axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2776 	axp->target_uid[axp->pid_count] = t_uid;
2777 	axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2778 	security_task_getsecid_obj(t, &axp->target_sid[axp->pid_count]);
2779 	memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2780 	axp->pid_count++;
2781 
2782 	return 0;
2783 }
2784 
2785 /**
2786  * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2787  * @bprm: pointer to the bprm being processed
2788  * @new: the proposed new credentials
2789  * @old: the old credentials
2790  *
2791  * Simply check if the proc already has the caps given by the file and if not
2792  * store the priv escalation info for later auditing at the end of the syscall
2793  *
2794  * -Eric
2795  */
2796 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2797 			   const struct cred *new, const struct cred *old)
2798 {
2799 	struct audit_aux_data_bprm_fcaps *ax;
2800 	struct audit_context *context = audit_context();
2801 	struct cpu_vfs_cap_data vcaps;
2802 
2803 	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2804 	if (!ax)
2805 		return -ENOMEM;
2806 
2807 	ax->d.type = AUDIT_BPRM_FCAPS;
2808 	ax->d.next = context->aux;
2809 	context->aux = (void *)ax;
2810 
2811 	get_vfs_caps_from_disk(&nop_mnt_idmap,
2812 			       bprm->file->f_path.dentry, &vcaps);
2813 
2814 	ax->fcap.permitted = vcaps.permitted;
2815 	ax->fcap.inheritable = vcaps.inheritable;
2816 	ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2817 	ax->fcap.rootid = vcaps.rootid;
2818 	ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2819 
2820 	ax->old_pcap.permitted   = old->cap_permitted;
2821 	ax->old_pcap.inheritable = old->cap_inheritable;
2822 	ax->old_pcap.effective   = old->cap_effective;
2823 	ax->old_pcap.ambient     = old->cap_ambient;
2824 
2825 	ax->new_pcap.permitted   = new->cap_permitted;
2826 	ax->new_pcap.inheritable = new->cap_inheritable;
2827 	ax->new_pcap.effective   = new->cap_effective;
2828 	ax->new_pcap.ambient     = new->cap_ambient;
2829 	return 0;
2830 }
2831 
2832 /**
2833  * __audit_log_capset - store information about the arguments to the capset syscall
2834  * @new: the new credentials
2835  * @old: the old (current) credentials
2836  *
2837  * Record the arguments userspace sent to sys_capset for later printing by the
2838  * audit system if applicable
2839  */
2840 void __audit_log_capset(const struct cred *new, const struct cred *old)
2841 {
2842 	struct audit_context *context = audit_context();
2843 
2844 	context->capset.pid = task_tgid_nr(current);
2845 	context->capset.cap.effective   = new->cap_effective;
2846 	context->capset.cap.inheritable = new->cap_effective;
2847 	context->capset.cap.permitted   = new->cap_permitted;
2848 	context->capset.cap.ambient     = new->cap_ambient;
2849 	context->type = AUDIT_CAPSET;
2850 }
2851 
2852 void __audit_mmap_fd(int fd, int flags)
2853 {
2854 	struct audit_context *context = audit_context();
2855 
2856 	context->mmap.fd = fd;
2857 	context->mmap.flags = flags;
2858 	context->type = AUDIT_MMAP;
2859 }
2860 
2861 void __audit_openat2_how(struct open_how *how)
2862 {
2863 	struct audit_context *context = audit_context();
2864 
2865 	context->openat2.flags = how->flags;
2866 	context->openat2.mode = how->mode;
2867 	context->openat2.resolve = how->resolve;
2868 	context->type = AUDIT_OPENAT2;
2869 }
2870 
2871 void __audit_log_kern_module(char *name)
2872 {
2873 	struct audit_context *context = audit_context();
2874 
2875 	context->module.name = kstrdup(name, GFP_KERNEL);
2876 	if (!context->module.name)
2877 		audit_log_lost("out of memory in __audit_log_kern_module");
2878 	context->type = AUDIT_KERN_MODULE;
2879 }
2880 
2881 void __audit_fanotify(u32 response, struct fanotify_response_info_audit_rule *friar)
2882 {
2883 	/* {subj,obj}_trust values are {0,1,2}: no,yes,unknown */
2884 	switch (friar->hdr.type) {
2885 	case FAN_RESPONSE_INFO_NONE:
2886 		audit_log(audit_context(), GFP_KERNEL, AUDIT_FANOTIFY,
2887 			  "resp=%u fan_type=%u fan_info=0 subj_trust=2 obj_trust=2",
2888 			  response, FAN_RESPONSE_INFO_NONE);
2889 		break;
2890 	case FAN_RESPONSE_INFO_AUDIT_RULE:
2891 		audit_log(audit_context(), GFP_KERNEL, AUDIT_FANOTIFY,
2892 			  "resp=%u fan_type=%u fan_info=%X subj_trust=%u obj_trust=%u",
2893 			  response, friar->hdr.type, friar->rule_number,
2894 			  friar->subj_trust, friar->obj_trust);
2895 	}
2896 }
2897 
2898 void __audit_tk_injoffset(struct timespec64 offset)
2899 {
2900 	struct audit_context *context = audit_context();
2901 
2902 	/* only set type if not already set by NTP */
2903 	if (!context->type)
2904 		context->type = AUDIT_TIME_INJOFFSET;
2905 	memcpy(&context->time.tk_injoffset, &offset, sizeof(offset));
2906 }
2907 
2908 void __audit_ntp_log(const struct audit_ntp_data *ad)
2909 {
2910 	struct audit_context *context = audit_context();
2911 	int type;
2912 
2913 	for (type = 0; type < AUDIT_NTP_NVALS; type++)
2914 		if (ad->vals[type].newval != ad->vals[type].oldval) {
2915 			/* unconditionally set type, overwriting TK */
2916 			context->type = AUDIT_TIME_ADJNTPVAL;
2917 			memcpy(&context->time.ntp_data, ad, sizeof(*ad));
2918 			break;
2919 		}
2920 }
2921 
2922 void __audit_log_nfcfg(const char *name, u8 af, unsigned int nentries,
2923 		       enum audit_nfcfgop op, gfp_t gfp)
2924 {
2925 	struct audit_buffer *ab;
2926 	char comm[sizeof(current->comm)];
2927 
2928 	ab = audit_log_start(audit_context(), gfp, AUDIT_NETFILTER_CFG);
2929 	if (!ab)
2930 		return;
2931 	audit_log_format(ab, "table=%s family=%u entries=%u op=%s",
2932 			 name, af, nentries, audit_nfcfgs[op].s);
2933 
2934 	audit_log_format(ab, " pid=%u", task_pid_nr(current));
2935 	audit_log_task_context(ab); /* subj= */
2936 	audit_log_format(ab, " comm=");
2937 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2938 	audit_log_end(ab);
2939 }
2940 EXPORT_SYMBOL_GPL(__audit_log_nfcfg);
2941 
2942 static void audit_log_task(struct audit_buffer *ab)
2943 {
2944 	kuid_t auid, uid;
2945 	kgid_t gid;
2946 	unsigned int sessionid;
2947 	char comm[sizeof(current->comm)];
2948 
2949 	auid = audit_get_loginuid(current);
2950 	sessionid = audit_get_sessionid(current);
2951 	current_uid_gid(&uid, &gid);
2952 
2953 	audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2954 			 from_kuid(&init_user_ns, auid),
2955 			 from_kuid(&init_user_ns, uid),
2956 			 from_kgid(&init_user_ns, gid),
2957 			 sessionid);
2958 	audit_log_task_context(ab);
2959 	audit_log_format(ab, " pid=%d comm=", task_tgid_nr(current));
2960 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2961 	audit_log_d_path_exe(ab, current->mm);
2962 }
2963 
2964 /**
2965  * audit_core_dumps - record information about processes that end abnormally
2966  * @signr: signal value
2967  *
2968  * If a process ends with a core dump, something fishy is going on and we
2969  * should record the event for investigation.
2970  */
2971 void audit_core_dumps(long signr)
2972 {
2973 	struct audit_buffer *ab;
2974 
2975 	if (!audit_enabled)
2976 		return;
2977 
2978 	if (signr == SIGQUIT)	/* don't care for those */
2979 		return;
2980 
2981 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_ANOM_ABEND);
2982 	if (unlikely(!ab))
2983 		return;
2984 	audit_log_task(ab);
2985 	audit_log_format(ab, " sig=%ld res=1", signr);
2986 	audit_log_end(ab);
2987 }
2988 
2989 /**
2990  * audit_seccomp - record information about a seccomp action
2991  * @syscall: syscall number
2992  * @signr: signal value
2993  * @code: the seccomp action
2994  *
2995  * Record the information associated with a seccomp action. Event filtering for
2996  * seccomp actions that are not to be logged is done in seccomp_log().
2997  * Therefore, this function forces auditing independent of the audit_enabled
2998  * and dummy context state because seccomp actions should be logged even when
2999  * audit is not in use.
3000  */
3001 void audit_seccomp(unsigned long syscall, long signr, int code)
3002 {
3003 	struct audit_buffer *ab;
3004 
3005 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_SECCOMP);
3006 	if (unlikely(!ab))
3007 		return;
3008 	audit_log_task(ab);
3009 	audit_log_format(ab, " sig=%ld arch=%x syscall=%ld compat=%d ip=0x%lx code=0x%x",
3010 			 signr, syscall_get_arch(current), syscall,
3011 			 in_compat_syscall(), KSTK_EIP(current), code);
3012 	audit_log_end(ab);
3013 }
3014 
3015 void audit_seccomp_actions_logged(const char *names, const char *old_names,
3016 				  int res)
3017 {
3018 	struct audit_buffer *ab;
3019 
3020 	if (!audit_enabled)
3021 		return;
3022 
3023 	ab = audit_log_start(audit_context(), GFP_KERNEL,
3024 			     AUDIT_CONFIG_CHANGE);
3025 	if (unlikely(!ab))
3026 		return;
3027 
3028 	audit_log_format(ab,
3029 			 "op=seccomp-logging actions=%s old-actions=%s res=%d",
3030 			 names, old_names, res);
3031 	audit_log_end(ab);
3032 }
3033 
3034 struct list_head *audit_killed_trees(void)
3035 {
3036 	struct audit_context *ctx = audit_context();
3037 	if (likely(!ctx || ctx->context == AUDIT_CTX_UNUSED))
3038 		return NULL;
3039 	return &ctx->killed_trees;
3040 }
3041