xref: /openbmc/linux/kernel/auditsc.c (revision 61f4d204)
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 	int h = audit_hash_ino((u32)n->ino);
885 	struct list_head *list = &audit_inode_hash[h];
886 
887 	return __audit_filter_op(tsk, ctx, list, n, ctx->major);
888 }
889 
890 /* At syscall exit time, this filter is called if any audit_names have been
891  * collected during syscall processing.  We only check rules in sublists at hash
892  * buckets applicable to the inode numbers in audit_names.
893  * Regarding audit_state, same rules apply as for audit_filter_syscall().
894  */
895 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
896 {
897 	struct audit_names *n;
898 
899 	if (auditd_test_task(tsk))
900 		return;
901 
902 	rcu_read_lock();
903 
904 	list_for_each_entry(n, &ctx->names_list, list) {
905 		if (audit_filter_inode_name(tsk, n, ctx))
906 			break;
907 	}
908 	rcu_read_unlock();
909 }
910 
911 static inline void audit_proctitle_free(struct audit_context *context)
912 {
913 	kfree(context->proctitle.value);
914 	context->proctitle.value = NULL;
915 	context->proctitle.len = 0;
916 }
917 
918 static inline void audit_free_module(struct audit_context *context)
919 {
920 	if (context->type == AUDIT_KERN_MODULE) {
921 		kfree(context->module.name);
922 		context->module.name = NULL;
923 	}
924 }
925 static inline void audit_free_names(struct audit_context *context)
926 {
927 	struct audit_names *n, *next;
928 
929 	list_for_each_entry_safe(n, next, &context->names_list, list) {
930 		list_del(&n->list);
931 		if (n->name)
932 			putname(n->name);
933 		if (n->should_free)
934 			kfree(n);
935 	}
936 	context->name_count = 0;
937 	path_put(&context->pwd);
938 	context->pwd.dentry = NULL;
939 	context->pwd.mnt = NULL;
940 }
941 
942 static inline void audit_free_aux(struct audit_context *context)
943 {
944 	struct audit_aux_data *aux;
945 
946 	while ((aux = context->aux)) {
947 		context->aux = aux->next;
948 		kfree(aux);
949 	}
950 	context->aux = NULL;
951 	while ((aux = context->aux_pids)) {
952 		context->aux_pids = aux->next;
953 		kfree(aux);
954 	}
955 	context->aux_pids = NULL;
956 }
957 
958 /**
959  * audit_reset_context - reset a audit_context structure
960  * @ctx: the audit_context to reset
961  *
962  * All fields in the audit_context will be reset to an initial state, all
963  * references held by fields will be dropped, and private memory will be
964  * released.  When this function returns the audit_context will be suitable
965  * for reuse, so long as the passed context is not NULL or a dummy context.
966  */
967 static void audit_reset_context(struct audit_context *ctx)
968 {
969 	if (!ctx)
970 		return;
971 
972 	/* if ctx is non-null, reset the "ctx->context" regardless */
973 	ctx->context = AUDIT_CTX_UNUSED;
974 	if (ctx->dummy)
975 		return;
976 
977 	/*
978 	 * NOTE: It shouldn't matter in what order we release the fields, so
979 	 *       release them in the order in which they appear in the struct;
980 	 *       this gives us some hope of quickly making sure we are
981 	 *       resetting the audit_context properly.
982 	 *
983 	 *       Other things worth mentioning:
984 	 *       - we don't reset "dummy"
985 	 *       - we don't reset "state", we do reset "current_state"
986 	 *       - we preserve "filterkey" if "state" is AUDIT_STATE_RECORD
987 	 *       - much of this is likely overkill, but play it safe for now
988 	 *       - we really need to work on improving the audit_context struct
989 	 */
990 
991 	ctx->current_state = ctx->state;
992 	ctx->serial = 0;
993 	ctx->major = 0;
994 	ctx->uring_op = 0;
995 	ctx->ctime = (struct timespec64){ .tv_sec = 0, .tv_nsec = 0 };
996 	memset(ctx->argv, 0, sizeof(ctx->argv));
997 	ctx->return_code = 0;
998 	ctx->prio = (ctx->state == AUDIT_STATE_RECORD ? ~0ULL : 0);
999 	ctx->return_valid = AUDITSC_INVALID;
1000 	audit_free_names(ctx);
1001 	if (ctx->state != AUDIT_STATE_RECORD) {
1002 		kfree(ctx->filterkey);
1003 		ctx->filterkey = NULL;
1004 	}
1005 	audit_free_aux(ctx);
1006 	kfree(ctx->sockaddr);
1007 	ctx->sockaddr = NULL;
1008 	ctx->sockaddr_len = 0;
1009 	ctx->ppid = 0;
1010 	ctx->uid = ctx->euid = ctx->suid = ctx->fsuid = KUIDT_INIT(0);
1011 	ctx->gid = ctx->egid = ctx->sgid = ctx->fsgid = KGIDT_INIT(0);
1012 	ctx->personality = 0;
1013 	ctx->arch = 0;
1014 	ctx->target_pid = 0;
1015 	ctx->target_auid = ctx->target_uid = KUIDT_INIT(0);
1016 	ctx->target_sessionid = 0;
1017 	ctx->target_sid = 0;
1018 	ctx->target_comm[0] = '\0';
1019 	unroll_tree_refs(ctx, NULL, 0);
1020 	WARN_ON(!list_empty(&ctx->killed_trees));
1021 	audit_free_module(ctx);
1022 	ctx->fds[0] = -1;
1023 	ctx->type = 0; /* reset last for audit_free_*() */
1024 }
1025 
1026 static inline struct audit_context *audit_alloc_context(enum audit_state state)
1027 {
1028 	struct audit_context *context;
1029 
1030 	context = kzalloc(sizeof(*context), GFP_KERNEL);
1031 	if (!context)
1032 		return NULL;
1033 	context->context = AUDIT_CTX_UNUSED;
1034 	context->state = state;
1035 	context->prio = state == AUDIT_STATE_RECORD ? ~0ULL : 0;
1036 	INIT_LIST_HEAD(&context->killed_trees);
1037 	INIT_LIST_HEAD(&context->names_list);
1038 	context->fds[0] = -1;
1039 	context->return_valid = AUDITSC_INVALID;
1040 	return context;
1041 }
1042 
1043 /**
1044  * audit_alloc - allocate an audit context block for a task
1045  * @tsk: task
1046  *
1047  * Filter on the task information and allocate a per-task audit context
1048  * if necessary.  Doing so turns on system call auditing for the
1049  * specified task.  This is called from copy_process, so no lock is
1050  * needed.
1051  */
1052 int audit_alloc(struct task_struct *tsk)
1053 {
1054 	struct audit_context *context;
1055 	enum audit_state     state;
1056 	char *key = NULL;
1057 
1058 	if (likely(!audit_ever_enabled))
1059 		return 0;
1060 
1061 	state = audit_filter_task(tsk, &key);
1062 	if (state == AUDIT_STATE_DISABLED) {
1063 		clear_task_syscall_work(tsk, SYSCALL_AUDIT);
1064 		return 0;
1065 	}
1066 
1067 	if (!(context = audit_alloc_context(state))) {
1068 		kfree(key);
1069 		audit_log_lost("out of memory in audit_alloc");
1070 		return -ENOMEM;
1071 	}
1072 	context->filterkey = key;
1073 
1074 	audit_set_context(tsk, context);
1075 	set_task_syscall_work(tsk, SYSCALL_AUDIT);
1076 	return 0;
1077 }
1078 
1079 static inline void audit_free_context(struct audit_context *context)
1080 {
1081 	/* resetting is extra work, but it is likely just noise */
1082 	audit_reset_context(context);
1083 	audit_proctitle_free(context);
1084 	free_tree_refs(context);
1085 	kfree(context->filterkey);
1086 	kfree(context);
1087 }
1088 
1089 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
1090 				 kuid_t auid, kuid_t uid, unsigned int sessionid,
1091 				 u32 sid, char *comm)
1092 {
1093 	struct audit_buffer *ab;
1094 	char *ctx = NULL;
1095 	u32 len;
1096 	int rc = 0;
1097 
1098 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1099 	if (!ab)
1100 		return rc;
1101 
1102 	audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid,
1103 			 from_kuid(&init_user_ns, auid),
1104 			 from_kuid(&init_user_ns, uid), sessionid);
1105 	if (sid) {
1106 		if (security_secid_to_secctx(sid, &ctx, &len)) {
1107 			audit_log_format(ab, " obj=(none)");
1108 			rc = 1;
1109 		} else {
1110 			audit_log_format(ab, " obj=%s", ctx);
1111 			security_release_secctx(ctx, len);
1112 		}
1113 	}
1114 	audit_log_format(ab, " ocomm=");
1115 	audit_log_untrustedstring(ab, comm);
1116 	audit_log_end(ab);
1117 
1118 	return rc;
1119 }
1120 
1121 static void audit_log_execve_info(struct audit_context *context,
1122 				  struct audit_buffer **ab)
1123 {
1124 	long len_max;
1125 	long len_rem;
1126 	long len_full;
1127 	long len_buf;
1128 	long len_abuf = 0;
1129 	long len_tmp;
1130 	bool require_data;
1131 	bool encode;
1132 	unsigned int iter;
1133 	unsigned int arg;
1134 	char *buf_head;
1135 	char *buf;
1136 	const char __user *p = (const char __user *)current->mm->arg_start;
1137 
1138 	/* NOTE: this buffer needs to be large enough to hold all the non-arg
1139 	 *       data we put in the audit record for this argument (see the
1140 	 *       code below) ... at this point in time 96 is plenty */
1141 	char abuf[96];
1142 
1143 	/* NOTE: we set MAX_EXECVE_AUDIT_LEN to a rather arbitrary limit, the
1144 	 *       current value of 7500 is not as important as the fact that it
1145 	 *       is less than 8k, a setting of 7500 gives us plenty of wiggle
1146 	 *       room if we go over a little bit in the logging below */
1147 	WARN_ON_ONCE(MAX_EXECVE_AUDIT_LEN > 7500);
1148 	len_max = MAX_EXECVE_AUDIT_LEN;
1149 
1150 	/* scratch buffer to hold the userspace args */
1151 	buf_head = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1152 	if (!buf_head) {
1153 		audit_panic("out of memory for argv string");
1154 		return;
1155 	}
1156 	buf = buf_head;
1157 
1158 	audit_log_format(*ab, "argc=%d", context->execve.argc);
1159 
1160 	len_rem = len_max;
1161 	len_buf = 0;
1162 	len_full = 0;
1163 	require_data = true;
1164 	encode = false;
1165 	iter = 0;
1166 	arg = 0;
1167 	do {
1168 		/* NOTE: we don't ever want to trust this value for anything
1169 		 *       serious, but the audit record format insists we
1170 		 *       provide an argument length for really long arguments,
1171 		 *       e.g. > MAX_EXECVE_AUDIT_LEN, so we have no choice but
1172 		 *       to use strncpy_from_user() to obtain this value for
1173 		 *       recording in the log, although we don't use it
1174 		 *       anywhere here to avoid a double-fetch problem */
1175 		if (len_full == 0)
1176 			len_full = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1177 
1178 		/* read more data from userspace */
1179 		if (require_data) {
1180 			/* can we make more room in the buffer? */
1181 			if (buf != buf_head) {
1182 				memmove(buf_head, buf, len_buf);
1183 				buf = buf_head;
1184 			}
1185 
1186 			/* fetch as much as we can of the argument */
1187 			len_tmp = strncpy_from_user(&buf_head[len_buf], p,
1188 						    len_max - len_buf);
1189 			if (len_tmp == -EFAULT) {
1190 				/* unable to copy from userspace */
1191 				send_sig(SIGKILL, current, 0);
1192 				goto out;
1193 			} else if (len_tmp == (len_max - len_buf)) {
1194 				/* buffer is not large enough */
1195 				require_data = true;
1196 				/* NOTE: if we are going to span multiple
1197 				 *       buffers force the encoding so we stand
1198 				 *       a chance at a sane len_full value and
1199 				 *       consistent record encoding */
1200 				encode = true;
1201 				len_full = len_full * 2;
1202 				p += len_tmp;
1203 			} else {
1204 				require_data = false;
1205 				if (!encode)
1206 					encode = audit_string_contains_control(
1207 								buf, len_tmp);
1208 				/* try to use a trusted value for len_full */
1209 				if (len_full < len_max)
1210 					len_full = (encode ?
1211 						    len_tmp * 2 : len_tmp);
1212 				p += len_tmp + 1;
1213 			}
1214 			len_buf += len_tmp;
1215 			buf_head[len_buf] = '\0';
1216 
1217 			/* length of the buffer in the audit record? */
1218 			len_abuf = (encode ? len_buf * 2 : len_buf + 2);
1219 		}
1220 
1221 		/* write as much as we can to the audit log */
1222 		if (len_buf >= 0) {
1223 			/* NOTE: some magic numbers here - basically if we
1224 			 *       can't fit a reasonable amount of data into the
1225 			 *       existing audit buffer, flush it and start with
1226 			 *       a new buffer */
1227 			if ((sizeof(abuf) + 8) > len_rem) {
1228 				len_rem = len_max;
1229 				audit_log_end(*ab);
1230 				*ab = audit_log_start(context,
1231 						      GFP_KERNEL, AUDIT_EXECVE);
1232 				if (!*ab)
1233 					goto out;
1234 			}
1235 
1236 			/* create the non-arg portion of the arg record */
1237 			len_tmp = 0;
1238 			if (require_data || (iter > 0) ||
1239 			    ((len_abuf + sizeof(abuf)) > len_rem)) {
1240 				if (iter == 0) {
1241 					len_tmp += snprintf(&abuf[len_tmp],
1242 							sizeof(abuf) - len_tmp,
1243 							" a%d_len=%lu",
1244 							arg, len_full);
1245 				}
1246 				len_tmp += snprintf(&abuf[len_tmp],
1247 						    sizeof(abuf) - len_tmp,
1248 						    " a%d[%d]=", arg, iter++);
1249 			} else
1250 				len_tmp += snprintf(&abuf[len_tmp],
1251 						    sizeof(abuf) - len_tmp,
1252 						    " a%d=", arg);
1253 			WARN_ON(len_tmp >= sizeof(abuf));
1254 			abuf[sizeof(abuf) - 1] = '\0';
1255 
1256 			/* log the arg in the audit record */
1257 			audit_log_format(*ab, "%s", abuf);
1258 			len_rem -= len_tmp;
1259 			len_tmp = len_buf;
1260 			if (encode) {
1261 				if (len_abuf > len_rem)
1262 					len_tmp = len_rem / 2; /* encoding */
1263 				audit_log_n_hex(*ab, buf, len_tmp);
1264 				len_rem -= len_tmp * 2;
1265 				len_abuf -= len_tmp * 2;
1266 			} else {
1267 				if (len_abuf > len_rem)
1268 					len_tmp = len_rem - 2; /* quotes */
1269 				audit_log_n_string(*ab, buf, len_tmp);
1270 				len_rem -= len_tmp + 2;
1271 				/* don't subtract the "2" because we still need
1272 				 * to add quotes to the remaining string */
1273 				len_abuf -= len_tmp;
1274 			}
1275 			len_buf -= len_tmp;
1276 			buf += len_tmp;
1277 		}
1278 
1279 		/* ready to move to the next argument? */
1280 		if ((len_buf == 0) && !require_data) {
1281 			arg++;
1282 			iter = 0;
1283 			len_full = 0;
1284 			require_data = true;
1285 			encode = false;
1286 		}
1287 	} while (arg < context->execve.argc);
1288 
1289 	/* NOTE: the caller handles the final audit_log_end() call */
1290 
1291 out:
1292 	kfree(buf_head);
1293 }
1294 
1295 static void audit_log_cap(struct audit_buffer *ab, char *prefix,
1296 			  kernel_cap_t *cap)
1297 {
1298 	if (cap_isclear(*cap)) {
1299 		audit_log_format(ab, " %s=0", prefix);
1300 		return;
1301 	}
1302 	audit_log_format(ab, " %s=%016llx", prefix, cap->val);
1303 }
1304 
1305 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1306 {
1307 	if (name->fcap_ver == -1) {
1308 		audit_log_format(ab, " cap_fe=? cap_fver=? cap_fp=? cap_fi=?");
1309 		return;
1310 	}
1311 	audit_log_cap(ab, "cap_fp", &name->fcap.permitted);
1312 	audit_log_cap(ab, "cap_fi", &name->fcap.inheritable);
1313 	audit_log_format(ab, " cap_fe=%d cap_fver=%x cap_frootid=%d",
1314 			 name->fcap.fE, name->fcap_ver,
1315 			 from_kuid(&init_user_ns, name->fcap.rootid));
1316 }
1317 
1318 static void audit_log_time(struct audit_context *context, struct audit_buffer **ab)
1319 {
1320 	const struct audit_ntp_data *ntp = &context->time.ntp_data;
1321 	const struct timespec64 *tk = &context->time.tk_injoffset;
1322 	static const char * const ntp_name[] = {
1323 		"offset",
1324 		"freq",
1325 		"status",
1326 		"tai",
1327 		"tick",
1328 		"adjust",
1329 	};
1330 	int type;
1331 
1332 	if (context->type == AUDIT_TIME_ADJNTPVAL) {
1333 		for (type = 0; type < AUDIT_NTP_NVALS; type++) {
1334 			if (ntp->vals[type].newval != ntp->vals[type].oldval) {
1335 				if (!*ab) {
1336 					*ab = audit_log_start(context,
1337 							GFP_KERNEL,
1338 							AUDIT_TIME_ADJNTPVAL);
1339 					if (!*ab)
1340 						return;
1341 				}
1342 				audit_log_format(*ab, "op=%s old=%lli new=%lli",
1343 						 ntp_name[type],
1344 						 ntp->vals[type].oldval,
1345 						 ntp->vals[type].newval);
1346 				audit_log_end(*ab);
1347 				*ab = NULL;
1348 			}
1349 		}
1350 	}
1351 	if (tk->tv_sec != 0 || tk->tv_nsec != 0) {
1352 		if (!*ab) {
1353 			*ab = audit_log_start(context, GFP_KERNEL,
1354 					      AUDIT_TIME_INJOFFSET);
1355 			if (!*ab)
1356 				return;
1357 		}
1358 		audit_log_format(*ab, "sec=%lli nsec=%li",
1359 				 (long long)tk->tv_sec, tk->tv_nsec);
1360 		audit_log_end(*ab);
1361 		*ab = NULL;
1362 	}
1363 }
1364 
1365 static void show_special(struct audit_context *context, int *call_panic)
1366 {
1367 	struct audit_buffer *ab;
1368 	int i;
1369 
1370 	ab = audit_log_start(context, GFP_KERNEL, context->type);
1371 	if (!ab)
1372 		return;
1373 
1374 	switch (context->type) {
1375 	case AUDIT_SOCKETCALL: {
1376 		int nargs = context->socketcall.nargs;
1377 
1378 		audit_log_format(ab, "nargs=%d", nargs);
1379 		for (i = 0; i < nargs; i++)
1380 			audit_log_format(ab, " a%d=%lx", i,
1381 				context->socketcall.args[i]);
1382 		break; }
1383 	case AUDIT_IPC: {
1384 		u32 osid = context->ipc.osid;
1385 
1386 		audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1387 				 from_kuid(&init_user_ns, context->ipc.uid),
1388 				 from_kgid(&init_user_ns, context->ipc.gid),
1389 				 context->ipc.mode);
1390 		if (osid) {
1391 			char *ctx = NULL;
1392 			u32 len;
1393 
1394 			if (security_secid_to_secctx(osid, &ctx, &len)) {
1395 				audit_log_format(ab, " osid=%u", osid);
1396 				*call_panic = 1;
1397 			} else {
1398 				audit_log_format(ab, " obj=%s", ctx);
1399 				security_release_secctx(ctx, len);
1400 			}
1401 		}
1402 		if (context->ipc.has_perm) {
1403 			audit_log_end(ab);
1404 			ab = audit_log_start(context, GFP_KERNEL,
1405 					     AUDIT_IPC_SET_PERM);
1406 			if (unlikely(!ab))
1407 				return;
1408 			audit_log_format(ab,
1409 				"qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1410 				context->ipc.qbytes,
1411 				context->ipc.perm_uid,
1412 				context->ipc.perm_gid,
1413 				context->ipc.perm_mode);
1414 		}
1415 		break; }
1416 	case AUDIT_MQ_OPEN:
1417 		audit_log_format(ab,
1418 			"oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1419 			"mq_msgsize=%ld mq_curmsgs=%ld",
1420 			context->mq_open.oflag, context->mq_open.mode,
1421 			context->mq_open.attr.mq_flags,
1422 			context->mq_open.attr.mq_maxmsg,
1423 			context->mq_open.attr.mq_msgsize,
1424 			context->mq_open.attr.mq_curmsgs);
1425 		break;
1426 	case AUDIT_MQ_SENDRECV:
1427 		audit_log_format(ab,
1428 			"mqdes=%d msg_len=%zd msg_prio=%u "
1429 			"abs_timeout_sec=%lld abs_timeout_nsec=%ld",
1430 			context->mq_sendrecv.mqdes,
1431 			context->mq_sendrecv.msg_len,
1432 			context->mq_sendrecv.msg_prio,
1433 			(long long) context->mq_sendrecv.abs_timeout.tv_sec,
1434 			context->mq_sendrecv.abs_timeout.tv_nsec);
1435 		break;
1436 	case AUDIT_MQ_NOTIFY:
1437 		audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1438 				context->mq_notify.mqdes,
1439 				context->mq_notify.sigev_signo);
1440 		break;
1441 	case AUDIT_MQ_GETSETATTR: {
1442 		struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1443 
1444 		audit_log_format(ab,
1445 			"mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1446 			"mq_curmsgs=%ld ",
1447 			context->mq_getsetattr.mqdes,
1448 			attr->mq_flags, attr->mq_maxmsg,
1449 			attr->mq_msgsize, attr->mq_curmsgs);
1450 		break; }
1451 	case AUDIT_CAPSET:
1452 		audit_log_format(ab, "pid=%d", context->capset.pid);
1453 		audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1454 		audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1455 		audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1456 		audit_log_cap(ab, "cap_pa", &context->capset.cap.ambient);
1457 		break;
1458 	case AUDIT_MMAP:
1459 		audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1460 				 context->mmap.flags);
1461 		break;
1462 	case AUDIT_OPENAT2:
1463 		audit_log_format(ab, "oflag=0%llo mode=0%llo resolve=0x%llx",
1464 				 context->openat2.flags,
1465 				 context->openat2.mode,
1466 				 context->openat2.resolve);
1467 		break;
1468 	case AUDIT_EXECVE:
1469 		audit_log_execve_info(context, &ab);
1470 		break;
1471 	case AUDIT_KERN_MODULE:
1472 		audit_log_format(ab, "name=");
1473 		if (context->module.name) {
1474 			audit_log_untrustedstring(ab, context->module.name);
1475 		} else
1476 			audit_log_format(ab, "(null)");
1477 
1478 		break;
1479 	case AUDIT_TIME_ADJNTPVAL:
1480 	case AUDIT_TIME_INJOFFSET:
1481 		/* this call deviates from the rest, eating the buffer */
1482 		audit_log_time(context, &ab);
1483 		break;
1484 	}
1485 	audit_log_end(ab);
1486 }
1487 
1488 static inline int audit_proctitle_rtrim(char *proctitle, int len)
1489 {
1490 	char *end = proctitle + len - 1;
1491 
1492 	while (end > proctitle && !isprint(*end))
1493 		end--;
1494 
1495 	/* catch the case where proctitle is only 1 non-print character */
1496 	len = end - proctitle + 1;
1497 	len -= isprint(proctitle[len-1]) == 0;
1498 	return len;
1499 }
1500 
1501 /*
1502  * audit_log_name - produce AUDIT_PATH record from struct audit_names
1503  * @context: audit_context for the task
1504  * @n: audit_names structure with reportable details
1505  * @path: optional path to report instead of audit_names->name
1506  * @record_num: record number to report when handling a list of names
1507  * @call_panic: optional pointer to int that will be updated if secid fails
1508  */
1509 static void audit_log_name(struct audit_context *context, struct audit_names *n,
1510 		    const struct path *path, int record_num, int *call_panic)
1511 {
1512 	struct audit_buffer *ab;
1513 
1514 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1515 	if (!ab)
1516 		return;
1517 
1518 	audit_log_format(ab, "item=%d", record_num);
1519 
1520 	if (path)
1521 		audit_log_d_path(ab, " name=", path);
1522 	else if (n->name) {
1523 		switch (n->name_len) {
1524 		case AUDIT_NAME_FULL:
1525 			/* log the full path */
1526 			audit_log_format(ab, " name=");
1527 			audit_log_untrustedstring(ab, n->name->name);
1528 			break;
1529 		case 0:
1530 			/* name was specified as a relative path and the
1531 			 * directory component is the cwd
1532 			 */
1533 			if (context->pwd.dentry && context->pwd.mnt)
1534 				audit_log_d_path(ab, " name=", &context->pwd);
1535 			else
1536 				audit_log_format(ab, " name=(null)");
1537 			break;
1538 		default:
1539 			/* log the name's directory component */
1540 			audit_log_format(ab, " name=");
1541 			audit_log_n_untrustedstring(ab, n->name->name,
1542 						    n->name_len);
1543 		}
1544 	} else
1545 		audit_log_format(ab, " name=(null)");
1546 
1547 	if (n->ino != AUDIT_INO_UNSET)
1548 		audit_log_format(ab, " inode=%lu dev=%02x:%02x mode=%#ho ouid=%u ogid=%u rdev=%02x:%02x",
1549 				 n->ino,
1550 				 MAJOR(n->dev),
1551 				 MINOR(n->dev),
1552 				 n->mode,
1553 				 from_kuid(&init_user_ns, n->uid),
1554 				 from_kgid(&init_user_ns, n->gid),
1555 				 MAJOR(n->rdev),
1556 				 MINOR(n->rdev));
1557 	if (n->osid != 0) {
1558 		char *ctx = NULL;
1559 		u32 len;
1560 
1561 		if (security_secid_to_secctx(
1562 			n->osid, &ctx, &len)) {
1563 			audit_log_format(ab, " osid=%u", n->osid);
1564 			if (call_panic)
1565 				*call_panic = 2;
1566 		} else {
1567 			audit_log_format(ab, " obj=%s", ctx);
1568 			security_release_secctx(ctx, len);
1569 		}
1570 	}
1571 
1572 	/* log the audit_names record type */
1573 	switch (n->type) {
1574 	case AUDIT_TYPE_NORMAL:
1575 		audit_log_format(ab, " nametype=NORMAL");
1576 		break;
1577 	case AUDIT_TYPE_PARENT:
1578 		audit_log_format(ab, " nametype=PARENT");
1579 		break;
1580 	case AUDIT_TYPE_CHILD_DELETE:
1581 		audit_log_format(ab, " nametype=DELETE");
1582 		break;
1583 	case AUDIT_TYPE_CHILD_CREATE:
1584 		audit_log_format(ab, " nametype=CREATE");
1585 		break;
1586 	default:
1587 		audit_log_format(ab, " nametype=UNKNOWN");
1588 		break;
1589 	}
1590 
1591 	audit_log_fcaps(ab, n);
1592 	audit_log_end(ab);
1593 }
1594 
1595 static void audit_log_proctitle(void)
1596 {
1597 	int res;
1598 	char *buf;
1599 	char *msg = "(null)";
1600 	int len = strlen(msg);
1601 	struct audit_context *context = audit_context();
1602 	struct audit_buffer *ab;
1603 
1604 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PROCTITLE);
1605 	if (!ab)
1606 		return;	/* audit_panic or being filtered */
1607 
1608 	audit_log_format(ab, "proctitle=");
1609 
1610 	/* Not  cached */
1611 	if (!context->proctitle.value) {
1612 		buf = kmalloc(MAX_PROCTITLE_AUDIT_LEN, GFP_KERNEL);
1613 		if (!buf)
1614 			goto out;
1615 		/* Historically called this from procfs naming */
1616 		res = get_cmdline(current, buf, MAX_PROCTITLE_AUDIT_LEN);
1617 		if (res == 0) {
1618 			kfree(buf);
1619 			goto out;
1620 		}
1621 		res = audit_proctitle_rtrim(buf, res);
1622 		if (res == 0) {
1623 			kfree(buf);
1624 			goto out;
1625 		}
1626 		context->proctitle.value = buf;
1627 		context->proctitle.len = res;
1628 	}
1629 	msg = context->proctitle.value;
1630 	len = context->proctitle.len;
1631 out:
1632 	audit_log_n_untrustedstring(ab, msg, len);
1633 	audit_log_end(ab);
1634 }
1635 
1636 /**
1637  * audit_log_uring - generate a AUDIT_URINGOP record
1638  * @ctx: the audit context
1639  */
1640 static void audit_log_uring(struct audit_context *ctx)
1641 {
1642 	struct audit_buffer *ab;
1643 	const struct cred *cred;
1644 
1645 	ab = audit_log_start(ctx, GFP_ATOMIC, AUDIT_URINGOP);
1646 	if (!ab)
1647 		return;
1648 	cred = current_cred();
1649 	audit_log_format(ab, "uring_op=%d", ctx->uring_op);
1650 	if (ctx->return_valid != AUDITSC_INVALID)
1651 		audit_log_format(ab, " success=%s exit=%ld",
1652 				 (ctx->return_valid == AUDITSC_SUCCESS ?
1653 				  "yes" : "no"),
1654 				 ctx->return_code);
1655 	audit_log_format(ab,
1656 			 " items=%d"
1657 			 " ppid=%d pid=%d uid=%u gid=%u euid=%u suid=%u"
1658 			 " fsuid=%u egid=%u sgid=%u fsgid=%u",
1659 			 ctx->name_count,
1660 			 task_ppid_nr(current), task_tgid_nr(current),
1661 			 from_kuid(&init_user_ns, cred->uid),
1662 			 from_kgid(&init_user_ns, cred->gid),
1663 			 from_kuid(&init_user_ns, cred->euid),
1664 			 from_kuid(&init_user_ns, cred->suid),
1665 			 from_kuid(&init_user_ns, cred->fsuid),
1666 			 from_kgid(&init_user_ns, cred->egid),
1667 			 from_kgid(&init_user_ns, cred->sgid),
1668 			 from_kgid(&init_user_ns, cred->fsgid));
1669 	audit_log_task_context(ab);
1670 	audit_log_key(ab, ctx->filterkey);
1671 	audit_log_end(ab);
1672 }
1673 
1674 static void audit_log_exit(void)
1675 {
1676 	int i, call_panic = 0;
1677 	struct audit_context *context = audit_context();
1678 	struct audit_buffer *ab;
1679 	struct audit_aux_data *aux;
1680 	struct audit_names *n;
1681 
1682 	context->personality = current->personality;
1683 
1684 	switch (context->context) {
1685 	case AUDIT_CTX_SYSCALL:
1686 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1687 		if (!ab)
1688 			return;
1689 		audit_log_format(ab, "arch=%x syscall=%d",
1690 				 context->arch, context->major);
1691 		if (context->personality != PER_LINUX)
1692 			audit_log_format(ab, " per=%lx", context->personality);
1693 		if (context->return_valid != AUDITSC_INVALID)
1694 			audit_log_format(ab, " success=%s exit=%ld",
1695 					 (context->return_valid == AUDITSC_SUCCESS ?
1696 					  "yes" : "no"),
1697 					 context->return_code);
1698 		audit_log_format(ab,
1699 				 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d",
1700 				 context->argv[0],
1701 				 context->argv[1],
1702 				 context->argv[2],
1703 				 context->argv[3],
1704 				 context->name_count);
1705 		audit_log_task_info(ab);
1706 		audit_log_key(ab, context->filterkey);
1707 		audit_log_end(ab);
1708 		break;
1709 	case AUDIT_CTX_URING:
1710 		audit_log_uring(context);
1711 		break;
1712 	default:
1713 		BUG();
1714 		break;
1715 	}
1716 
1717 	for (aux = context->aux; aux; aux = aux->next) {
1718 
1719 		ab = audit_log_start(context, GFP_KERNEL, aux->type);
1720 		if (!ab)
1721 			continue; /* audit_panic has been called */
1722 
1723 		switch (aux->type) {
1724 
1725 		case AUDIT_BPRM_FCAPS: {
1726 			struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1727 
1728 			audit_log_format(ab, "fver=%x", axs->fcap_ver);
1729 			audit_log_cap(ab, "fp", &axs->fcap.permitted);
1730 			audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1731 			audit_log_format(ab, " fe=%d", axs->fcap.fE);
1732 			audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1733 			audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1734 			audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1735 			audit_log_cap(ab, "old_pa", &axs->old_pcap.ambient);
1736 			audit_log_cap(ab, "pp", &axs->new_pcap.permitted);
1737 			audit_log_cap(ab, "pi", &axs->new_pcap.inheritable);
1738 			audit_log_cap(ab, "pe", &axs->new_pcap.effective);
1739 			audit_log_cap(ab, "pa", &axs->new_pcap.ambient);
1740 			audit_log_format(ab, " frootid=%d",
1741 					 from_kuid(&init_user_ns,
1742 						   axs->fcap.rootid));
1743 			break; }
1744 
1745 		}
1746 		audit_log_end(ab);
1747 	}
1748 
1749 	if (context->type)
1750 		show_special(context, &call_panic);
1751 
1752 	if (context->fds[0] >= 0) {
1753 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1754 		if (ab) {
1755 			audit_log_format(ab, "fd0=%d fd1=%d",
1756 					context->fds[0], context->fds[1]);
1757 			audit_log_end(ab);
1758 		}
1759 	}
1760 
1761 	if (context->sockaddr_len) {
1762 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1763 		if (ab) {
1764 			audit_log_format(ab, "saddr=");
1765 			audit_log_n_hex(ab, (void *)context->sockaddr,
1766 					context->sockaddr_len);
1767 			audit_log_end(ab);
1768 		}
1769 	}
1770 
1771 	for (aux = context->aux_pids; aux; aux = aux->next) {
1772 		struct audit_aux_data_pids *axs = (void *)aux;
1773 
1774 		for (i = 0; i < axs->pid_count; i++)
1775 			if (audit_log_pid_context(context, axs->target_pid[i],
1776 						  axs->target_auid[i],
1777 						  axs->target_uid[i],
1778 						  axs->target_sessionid[i],
1779 						  axs->target_sid[i],
1780 						  axs->target_comm[i]))
1781 				call_panic = 1;
1782 	}
1783 
1784 	if (context->target_pid &&
1785 	    audit_log_pid_context(context, context->target_pid,
1786 				  context->target_auid, context->target_uid,
1787 				  context->target_sessionid,
1788 				  context->target_sid, context->target_comm))
1789 			call_panic = 1;
1790 
1791 	if (context->pwd.dentry && context->pwd.mnt) {
1792 		ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1793 		if (ab) {
1794 			audit_log_d_path(ab, "cwd=", &context->pwd);
1795 			audit_log_end(ab);
1796 		}
1797 	}
1798 
1799 	i = 0;
1800 	list_for_each_entry(n, &context->names_list, list) {
1801 		if (n->hidden)
1802 			continue;
1803 		audit_log_name(context, n, NULL, i++, &call_panic);
1804 	}
1805 
1806 	if (context->context == AUDIT_CTX_SYSCALL)
1807 		audit_log_proctitle();
1808 
1809 	/* Send end of event record to help user space know we are finished */
1810 	ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1811 	if (ab)
1812 		audit_log_end(ab);
1813 	if (call_panic)
1814 		audit_panic("error in audit_log_exit()");
1815 }
1816 
1817 /**
1818  * __audit_free - free a per-task audit context
1819  * @tsk: task whose audit context block to free
1820  *
1821  * Called from copy_process, do_exit, and the io_uring code
1822  */
1823 void __audit_free(struct task_struct *tsk)
1824 {
1825 	struct audit_context *context = tsk->audit_context;
1826 
1827 	if (!context)
1828 		return;
1829 
1830 	/* this may generate CONFIG_CHANGE records */
1831 	if (!list_empty(&context->killed_trees))
1832 		audit_kill_trees(context);
1833 
1834 	/* We are called either by do_exit() or the fork() error handling code;
1835 	 * in the former case tsk == current and in the latter tsk is a
1836 	 * random task_struct that doesn't have any meaningful data we
1837 	 * need to log via audit_log_exit().
1838 	 */
1839 	if (tsk == current && !context->dummy) {
1840 		context->return_valid = AUDITSC_INVALID;
1841 		context->return_code = 0;
1842 		if (context->context == AUDIT_CTX_SYSCALL) {
1843 			audit_filter_syscall(tsk, context);
1844 			audit_filter_inodes(tsk, context);
1845 			if (context->current_state == AUDIT_STATE_RECORD)
1846 				audit_log_exit();
1847 		} else if (context->context == AUDIT_CTX_URING) {
1848 			/* TODO: verify this case is real and valid */
1849 			audit_filter_uring(tsk, context);
1850 			audit_filter_inodes(tsk, context);
1851 			if (context->current_state == AUDIT_STATE_RECORD)
1852 				audit_log_uring(context);
1853 		}
1854 	}
1855 
1856 	audit_set_context(tsk, NULL);
1857 	audit_free_context(context);
1858 }
1859 
1860 /**
1861  * audit_return_fixup - fixup the return codes in the audit_context
1862  * @ctx: the audit_context
1863  * @success: true/false value to indicate if the operation succeeded or not
1864  * @code: operation return code
1865  *
1866  * We need to fixup the return code in the audit logs if the actual return
1867  * codes are later going to be fixed by the arch specific signal handlers.
1868  */
1869 static void audit_return_fixup(struct audit_context *ctx,
1870 			       int success, long code)
1871 {
1872 	/*
1873 	 * This is actually a test for:
1874 	 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
1875 	 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
1876 	 *
1877 	 * but is faster than a bunch of ||
1878 	 */
1879 	if (unlikely(code <= -ERESTARTSYS) &&
1880 	    (code >= -ERESTART_RESTARTBLOCK) &&
1881 	    (code != -ENOIOCTLCMD))
1882 		ctx->return_code = -EINTR;
1883 	else
1884 		ctx->return_code  = code;
1885 	ctx->return_valid = (success ? AUDITSC_SUCCESS : AUDITSC_FAILURE);
1886 }
1887 
1888 /**
1889  * __audit_uring_entry - prepare the kernel task's audit context for io_uring
1890  * @op: the io_uring opcode
1891  *
1892  * This is similar to audit_syscall_entry() but is intended for use by io_uring
1893  * operations.  This function should only ever be called from
1894  * audit_uring_entry() as we rely on the audit context checking present in that
1895  * function.
1896  */
1897 void __audit_uring_entry(u8 op)
1898 {
1899 	struct audit_context *ctx = audit_context();
1900 
1901 	if (ctx->state == AUDIT_STATE_DISABLED)
1902 		return;
1903 
1904 	/*
1905 	 * NOTE: It's possible that we can be called from the process' context
1906 	 *       before it returns to userspace, and before audit_syscall_exit()
1907 	 *       is called.  In this case there is not much to do, just record
1908 	 *       the io_uring details and return.
1909 	 */
1910 	ctx->uring_op = op;
1911 	if (ctx->context == AUDIT_CTX_SYSCALL)
1912 		return;
1913 
1914 	ctx->dummy = !audit_n_rules;
1915 	if (!ctx->dummy && ctx->state == AUDIT_STATE_BUILD)
1916 		ctx->prio = 0;
1917 
1918 	ctx->context = AUDIT_CTX_URING;
1919 	ctx->current_state = ctx->state;
1920 	ktime_get_coarse_real_ts64(&ctx->ctime);
1921 }
1922 
1923 /**
1924  * __audit_uring_exit - wrap up the kernel task's audit context after io_uring
1925  * @success: true/false value to indicate if the operation succeeded or not
1926  * @code: operation return code
1927  *
1928  * This is similar to audit_syscall_exit() but is intended for use by io_uring
1929  * operations.  This function should only ever be called from
1930  * audit_uring_exit() as we rely on the audit context checking present in that
1931  * function.
1932  */
1933 void __audit_uring_exit(int success, long code)
1934 {
1935 	struct audit_context *ctx = audit_context();
1936 
1937 	if (ctx->dummy) {
1938 		if (ctx->context != AUDIT_CTX_URING)
1939 			return;
1940 		goto out;
1941 	}
1942 
1943 	audit_return_fixup(ctx, success, code);
1944 	if (ctx->context == AUDIT_CTX_SYSCALL) {
1945 		/*
1946 		 * NOTE: See the note in __audit_uring_entry() about the case
1947 		 *       where we may be called from process context before we
1948 		 *       return to userspace via audit_syscall_exit().  In this
1949 		 *       case we simply emit a URINGOP record and bail, the
1950 		 *       normal syscall exit handling will take care of
1951 		 *       everything else.
1952 		 *       It is also worth mentioning that when we are called,
1953 		 *       the current process creds may differ from the creds
1954 		 *       used during the normal syscall processing; keep that
1955 		 *       in mind if/when we move the record generation code.
1956 		 */
1957 
1958 		/*
1959 		 * We need to filter on the syscall info here to decide if we
1960 		 * should emit a URINGOP record.  I know it seems odd but this
1961 		 * solves the problem where users have a filter to block *all*
1962 		 * syscall records in the "exit" filter; we want to preserve
1963 		 * the behavior here.
1964 		 */
1965 		audit_filter_syscall(current, ctx);
1966 		if (ctx->current_state != AUDIT_STATE_RECORD)
1967 			audit_filter_uring(current, ctx);
1968 		audit_filter_inodes(current, ctx);
1969 		if (ctx->current_state != AUDIT_STATE_RECORD)
1970 			return;
1971 
1972 		audit_log_uring(ctx);
1973 		return;
1974 	}
1975 
1976 	/* this may generate CONFIG_CHANGE records */
1977 	if (!list_empty(&ctx->killed_trees))
1978 		audit_kill_trees(ctx);
1979 
1980 	/* run through both filters to ensure we set the filterkey properly */
1981 	audit_filter_uring(current, ctx);
1982 	audit_filter_inodes(current, ctx);
1983 	if (ctx->current_state != AUDIT_STATE_RECORD)
1984 		goto out;
1985 	audit_log_exit();
1986 
1987 out:
1988 	audit_reset_context(ctx);
1989 }
1990 
1991 /**
1992  * __audit_syscall_entry - fill in an audit record at syscall entry
1993  * @major: major syscall type (function)
1994  * @a1: additional syscall register 1
1995  * @a2: additional syscall register 2
1996  * @a3: additional syscall register 3
1997  * @a4: additional syscall register 4
1998  *
1999  * Fill in audit context at syscall entry.  This only happens if the
2000  * audit context was created when the task was created and the state or
2001  * filters demand the audit context be built.  If the state from the
2002  * per-task filter or from the per-syscall filter is AUDIT_STATE_RECORD,
2003  * then the record will be written at syscall exit time (otherwise, it
2004  * will only be written if another part of the kernel requests that it
2005  * be written).
2006  */
2007 void __audit_syscall_entry(int major, unsigned long a1, unsigned long a2,
2008 			   unsigned long a3, unsigned long a4)
2009 {
2010 	struct audit_context *context = audit_context();
2011 	enum audit_state     state;
2012 
2013 	if (!audit_enabled || !context)
2014 		return;
2015 
2016 	WARN_ON(context->context != AUDIT_CTX_UNUSED);
2017 	WARN_ON(context->name_count);
2018 	if (context->context != AUDIT_CTX_UNUSED || context->name_count) {
2019 		audit_panic("unrecoverable error in audit_syscall_entry()");
2020 		return;
2021 	}
2022 
2023 	state = context->state;
2024 	if (state == AUDIT_STATE_DISABLED)
2025 		return;
2026 
2027 	context->dummy = !audit_n_rules;
2028 	if (!context->dummy && state == AUDIT_STATE_BUILD) {
2029 		context->prio = 0;
2030 		if (auditd_test_task(current))
2031 			return;
2032 	}
2033 
2034 	context->arch	    = syscall_get_arch(current);
2035 	context->major      = major;
2036 	context->argv[0]    = a1;
2037 	context->argv[1]    = a2;
2038 	context->argv[2]    = a3;
2039 	context->argv[3]    = a4;
2040 	context->context = AUDIT_CTX_SYSCALL;
2041 	context->current_state  = state;
2042 	ktime_get_coarse_real_ts64(&context->ctime);
2043 }
2044 
2045 /**
2046  * __audit_syscall_exit - deallocate audit context after a system call
2047  * @success: success value of the syscall
2048  * @return_code: return value of the syscall
2049  *
2050  * Tear down after system call.  If the audit context has been marked as
2051  * auditable (either because of the AUDIT_STATE_RECORD state from
2052  * filtering, or because some other part of the kernel wrote an audit
2053  * message), then write out the syscall information.  In call cases,
2054  * free the names stored from getname().
2055  */
2056 void __audit_syscall_exit(int success, long return_code)
2057 {
2058 	struct audit_context *context = audit_context();
2059 
2060 	if (!context || context->dummy ||
2061 	    context->context != AUDIT_CTX_SYSCALL)
2062 		goto out;
2063 
2064 	/* this may generate CONFIG_CHANGE records */
2065 	if (!list_empty(&context->killed_trees))
2066 		audit_kill_trees(context);
2067 
2068 	audit_return_fixup(context, success, return_code);
2069 	/* run through both filters to ensure we set the filterkey properly */
2070 	audit_filter_syscall(current, context);
2071 	audit_filter_inodes(current, context);
2072 	if (context->current_state != AUDIT_STATE_RECORD)
2073 		goto out;
2074 
2075 	audit_log_exit();
2076 
2077 out:
2078 	audit_reset_context(context);
2079 }
2080 
2081 static inline void handle_one(const struct inode *inode)
2082 {
2083 	struct audit_context *context;
2084 	struct audit_tree_refs *p;
2085 	struct audit_chunk *chunk;
2086 	int count;
2087 
2088 	if (likely(!inode->i_fsnotify_marks))
2089 		return;
2090 	context = audit_context();
2091 	p = context->trees;
2092 	count = context->tree_count;
2093 	rcu_read_lock();
2094 	chunk = audit_tree_lookup(inode);
2095 	rcu_read_unlock();
2096 	if (!chunk)
2097 		return;
2098 	if (likely(put_tree_ref(context, chunk)))
2099 		return;
2100 	if (unlikely(!grow_tree_refs(context))) {
2101 		pr_warn("out of memory, audit has lost a tree reference\n");
2102 		audit_set_auditable(context);
2103 		audit_put_chunk(chunk);
2104 		unroll_tree_refs(context, p, count);
2105 		return;
2106 	}
2107 	put_tree_ref(context, chunk);
2108 }
2109 
2110 static void handle_path(const struct dentry *dentry)
2111 {
2112 	struct audit_context *context;
2113 	struct audit_tree_refs *p;
2114 	const struct dentry *d, *parent;
2115 	struct audit_chunk *drop;
2116 	unsigned long seq;
2117 	int count;
2118 
2119 	context = audit_context();
2120 	p = context->trees;
2121 	count = context->tree_count;
2122 retry:
2123 	drop = NULL;
2124 	d = dentry;
2125 	rcu_read_lock();
2126 	seq = read_seqbegin(&rename_lock);
2127 	for(;;) {
2128 		struct inode *inode = d_backing_inode(d);
2129 
2130 		if (inode && unlikely(inode->i_fsnotify_marks)) {
2131 			struct audit_chunk *chunk;
2132 
2133 			chunk = audit_tree_lookup(inode);
2134 			if (chunk) {
2135 				if (unlikely(!put_tree_ref(context, chunk))) {
2136 					drop = chunk;
2137 					break;
2138 				}
2139 			}
2140 		}
2141 		parent = d->d_parent;
2142 		if (parent == d)
2143 			break;
2144 		d = parent;
2145 	}
2146 	if (unlikely(read_seqretry(&rename_lock, seq) || drop)) {  /* in this order */
2147 		rcu_read_unlock();
2148 		if (!drop) {
2149 			/* just a race with rename */
2150 			unroll_tree_refs(context, p, count);
2151 			goto retry;
2152 		}
2153 		audit_put_chunk(drop);
2154 		if (grow_tree_refs(context)) {
2155 			/* OK, got more space */
2156 			unroll_tree_refs(context, p, count);
2157 			goto retry;
2158 		}
2159 		/* too bad */
2160 		pr_warn("out of memory, audit has lost a tree reference\n");
2161 		unroll_tree_refs(context, p, count);
2162 		audit_set_auditable(context);
2163 		return;
2164 	}
2165 	rcu_read_unlock();
2166 }
2167 
2168 static struct audit_names *audit_alloc_name(struct audit_context *context,
2169 						unsigned char type)
2170 {
2171 	struct audit_names *aname;
2172 
2173 	if (context->name_count < AUDIT_NAMES) {
2174 		aname = &context->preallocated_names[context->name_count];
2175 		memset(aname, 0, sizeof(*aname));
2176 	} else {
2177 		aname = kzalloc(sizeof(*aname), GFP_NOFS);
2178 		if (!aname)
2179 			return NULL;
2180 		aname->should_free = true;
2181 	}
2182 
2183 	aname->ino = AUDIT_INO_UNSET;
2184 	aname->type = type;
2185 	list_add_tail(&aname->list, &context->names_list);
2186 
2187 	context->name_count++;
2188 	if (!context->pwd.dentry)
2189 		get_fs_pwd(current->fs, &context->pwd);
2190 	return aname;
2191 }
2192 
2193 /**
2194  * __audit_reusename - fill out filename with info from existing entry
2195  * @uptr: userland ptr to pathname
2196  *
2197  * Search the audit_names list for the current audit context. If there is an
2198  * existing entry with a matching "uptr" then return the filename
2199  * associated with that audit_name. If not, return NULL.
2200  */
2201 struct filename *
2202 __audit_reusename(const __user char *uptr)
2203 {
2204 	struct audit_context *context = audit_context();
2205 	struct audit_names *n;
2206 
2207 	list_for_each_entry(n, &context->names_list, list) {
2208 		if (!n->name)
2209 			continue;
2210 		if (n->name->uptr == uptr) {
2211 			n->name->refcnt++;
2212 			return n->name;
2213 		}
2214 	}
2215 	return NULL;
2216 }
2217 
2218 /**
2219  * __audit_getname - add a name to the list
2220  * @name: name to add
2221  *
2222  * Add a name to the list of audit names for this context.
2223  * Called from fs/namei.c:getname().
2224  */
2225 void __audit_getname(struct filename *name)
2226 {
2227 	struct audit_context *context = audit_context();
2228 	struct audit_names *n;
2229 
2230 	if (context->context == AUDIT_CTX_UNUSED)
2231 		return;
2232 
2233 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
2234 	if (!n)
2235 		return;
2236 
2237 	n->name = name;
2238 	n->name_len = AUDIT_NAME_FULL;
2239 	name->aname = n;
2240 	name->refcnt++;
2241 }
2242 
2243 static inline int audit_copy_fcaps(struct audit_names *name,
2244 				   const struct dentry *dentry)
2245 {
2246 	struct cpu_vfs_cap_data caps;
2247 	int rc;
2248 
2249 	if (!dentry)
2250 		return 0;
2251 
2252 	rc = get_vfs_caps_from_disk(&nop_mnt_idmap, dentry, &caps);
2253 	if (rc)
2254 		return rc;
2255 
2256 	name->fcap.permitted = caps.permitted;
2257 	name->fcap.inheritable = caps.inheritable;
2258 	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2259 	name->fcap.rootid = caps.rootid;
2260 	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >>
2261 				VFS_CAP_REVISION_SHIFT;
2262 
2263 	return 0;
2264 }
2265 
2266 /* Copy inode data into an audit_names. */
2267 static void audit_copy_inode(struct audit_names *name,
2268 			     const struct dentry *dentry,
2269 			     struct inode *inode, unsigned int flags)
2270 {
2271 	name->ino   = inode->i_ino;
2272 	name->dev   = inode->i_sb->s_dev;
2273 	name->mode  = inode->i_mode;
2274 	name->uid   = inode->i_uid;
2275 	name->gid   = inode->i_gid;
2276 	name->rdev  = inode->i_rdev;
2277 	security_inode_getsecid(inode, &name->osid);
2278 	if (flags & AUDIT_INODE_NOEVAL) {
2279 		name->fcap_ver = -1;
2280 		return;
2281 	}
2282 	audit_copy_fcaps(name, dentry);
2283 }
2284 
2285 /**
2286  * __audit_inode - store the inode and device from a lookup
2287  * @name: name being audited
2288  * @dentry: dentry being audited
2289  * @flags: attributes for this particular entry
2290  */
2291 void __audit_inode(struct filename *name, const struct dentry *dentry,
2292 		   unsigned int flags)
2293 {
2294 	struct audit_context *context = audit_context();
2295 	struct inode *inode = d_backing_inode(dentry);
2296 	struct audit_names *n;
2297 	bool parent = flags & AUDIT_INODE_PARENT;
2298 	struct audit_entry *e;
2299 	struct list_head *list = &audit_filter_list[AUDIT_FILTER_FS];
2300 	int i;
2301 
2302 	if (context->context == AUDIT_CTX_UNUSED)
2303 		return;
2304 
2305 	rcu_read_lock();
2306 	list_for_each_entry_rcu(e, list, list) {
2307 		for (i = 0; i < e->rule.field_count; i++) {
2308 			struct audit_field *f = &e->rule.fields[i];
2309 
2310 			if (f->type == AUDIT_FSTYPE
2311 			    && audit_comparator(inode->i_sb->s_magic,
2312 						f->op, f->val)
2313 			    && e->rule.action == AUDIT_NEVER) {
2314 				rcu_read_unlock();
2315 				return;
2316 			}
2317 		}
2318 	}
2319 	rcu_read_unlock();
2320 
2321 	if (!name)
2322 		goto out_alloc;
2323 
2324 	/*
2325 	 * If we have a pointer to an audit_names entry already, then we can
2326 	 * just use it directly if the type is correct.
2327 	 */
2328 	n = name->aname;
2329 	if (n) {
2330 		if (parent) {
2331 			if (n->type == AUDIT_TYPE_PARENT ||
2332 			    n->type == AUDIT_TYPE_UNKNOWN)
2333 				goto out;
2334 		} else {
2335 			if (n->type != AUDIT_TYPE_PARENT)
2336 				goto out;
2337 		}
2338 	}
2339 
2340 	list_for_each_entry_reverse(n, &context->names_list, list) {
2341 		if (n->ino) {
2342 			/* valid inode number, use that for the comparison */
2343 			if (n->ino != inode->i_ino ||
2344 			    n->dev != inode->i_sb->s_dev)
2345 				continue;
2346 		} else if (n->name) {
2347 			/* inode number has not been set, check the name */
2348 			if (strcmp(n->name->name, name->name))
2349 				continue;
2350 		} else
2351 			/* no inode and no name (?!) ... this is odd ... */
2352 			continue;
2353 
2354 		/* match the correct record type */
2355 		if (parent) {
2356 			if (n->type == AUDIT_TYPE_PARENT ||
2357 			    n->type == AUDIT_TYPE_UNKNOWN)
2358 				goto out;
2359 		} else {
2360 			if (n->type != AUDIT_TYPE_PARENT)
2361 				goto out;
2362 		}
2363 	}
2364 
2365 out_alloc:
2366 	/* unable to find an entry with both a matching name and type */
2367 	n = audit_alloc_name(context, AUDIT_TYPE_UNKNOWN);
2368 	if (!n)
2369 		return;
2370 	if (name) {
2371 		n->name = name;
2372 		name->refcnt++;
2373 	}
2374 
2375 out:
2376 	if (parent) {
2377 		n->name_len = n->name ? parent_len(n->name->name) : AUDIT_NAME_FULL;
2378 		n->type = AUDIT_TYPE_PARENT;
2379 		if (flags & AUDIT_INODE_HIDDEN)
2380 			n->hidden = true;
2381 	} else {
2382 		n->name_len = AUDIT_NAME_FULL;
2383 		n->type = AUDIT_TYPE_NORMAL;
2384 	}
2385 	handle_path(dentry);
2386 	audit_copy_inode(n, dentry, inode, flags & AUDIT_INODE_NOEVAL);
2387 }
2388 
2389 void __audit_file(const struct file *file)
2390 {
2391 	__audit_inode(NULL, file->f_path.dentry, 0);
2392 }
2393 
2394 /**
2395  * __audit_inode_child - collect inode info for created/removed objects
2396  * @parent: inode of dentry parent
2397  * @dentry: dentry being audited
2398  * @type:   AUDIT_TYPE_* value that we're looking for
2399  *
2400  * For syscalls that create or remove filesystem objects, audit_inode
2401  * can only collect information for the filesystem object's parent.
2402  * This call updates the audit context with the child's information.
2403  * Syscalls that create a new filesystem object must be hooked after
2404  * the object is created.  Syscalls that remove a filesystem object
2405  * must be hooked prior, in order to capture the target inode during
2406  * unsuccessful attempts.
2407  */
2408 void __audit_inode_child(struct inode *parent,
2409 			 const struct dentry *dentry,
2410 			 const unsigned char type)
2411 {
2412 	struct audit_context *context = audit_context();
2413 	struct inode *inode = d_backing_inode(dentry);
2414 	const struct qstr *dname = &dentry->d_name;
2415 	struct audit_names *n, *found_parent = NULL, *found_child = NULL;
2416 	struct audit_entry *e;
2417 	struct list_head *list = &audit_filter_list[AUDIT_FILTER_FS];
2418 	int i;
2419 
2420 	if (context->context == AUDIT_CTX_UNUSED)
2421 		return;
2422 
2423 	rcu_read_lock();
2424 	list_for_each_entry_rcu(e, list, list) {
2425 		for (i = 0; i < e->rule.field_count; i++) {
2426 			struct audit_field *f = &e->rule.fields[i];
2427 
2428 			if (f->type == AUDIT_FSTYPE
2429 			    && audit_comparator(parent->i_sb->s_magic,
2430 						f->op, f->val)
2431 			    && e->rule.action == AUDIT_NEVER) {
2432 				rcu_read_unlock();
2433 				return;
2434 			}
2435 		}
2436 	}
2437 	rcu_read_unlock();
2438 
2439 	if (inode)
2440 		handle_one(inode);
2441 
2442 	/* look for a parent entry first */
2443 	list_for_each_entry(n, &context->names_list, list) {
2444 		if (!n->name ||
2445 		    (n->type != AUDIT_TYPE_PARENT &&
2446 		     n->type != AUDIT_TYPE_UNKNOWN))
2447 			continue;
2448 
2449 		if (n->ino == parent->i_ino && n->dev == parent->i_sb->s_dev &&
2450 		    !audit_compare_dname_path(dname,
2451 					      n->name->name, n->name_len)) {
2452 			if (n->type == AUDIT_TYPE_UNKNOWN)
2453 				n->type = AUDIT_TYPE_PARENT;
2454 			found_parent = n;
2455 			break;
2456 		}
2457 	}
2458 
2459 	/* is there a matching child entry? */
2460 	list_for_each_entry(n, &context->names_list, list) {
2461 		/* can only match entries that have a name */
2462 		if (!n->name ||
2463 		    (n->type != type && n->type != AUDIT_TYPE_UNKNOWN))
2464 			continue;
2465 
2466 		if (!strcmp(dname->name, n->name->name) ||
2467 		    !audit_compare_dname_path(dname, n->name->name,
2468 						found_parent ?
2469 						found_parent->name_len :
2470 						AUDIT_NAME_FULL)) {
2471 			if (n->type == AUDIT_TYPE_UNKNOWN)
2472 				n->type = type;
2473 			found_child = n;
2474 			break;
2475 		}
2476 	}
2477 
2478 	if (!found_parent) {
2479 		/* create a new, "anonymous" parent record */
2480 		n = audit_alloc_name(context, AUDIT_TYPE_PARENT);
2481 		if (!n)
2482 			return;
2483 		audit_copy_inode(n, NULL, parent, 0);
2484 	}
2485 
2486 	if (!found_child) {
2487 		found_child = audit_alloc_name(context, type);
2488 		if (!found_child)
2489 			return;
2490 
2491 		/* Re-use the name belonging to the slot for a matching parent
2492 		 * directory. All names for this context are relinquished in
2493 		 * audit_free_names() */
2494 		if (found_parent) {
2495 			found_child->name = found_parent->name;
2496 			found_child->name_len = AUDIT_NAME_FULL;
2497 			found_child->name->refcnt++;
2498 		}
2499 	}
2500 
2501 	if (inode)
2502 		audit_copy_inode(found_child, dentry, inode, 0);
2503 	else
2504 		found_child->ino = AUDIT_INO_UNSET;
2505 }
2506 EXPORT_SYMBOL_GPL(__audit_inode_child);
2507 
2508 /**
2509  * auditsc_get_stamp - get local copies of audit_context values
2510  * @ctx: audit_context for the task
2511  * @t: timespec64 to store time recorded in the audit_context
2512  * @serial: serial value that is recorded in the audit_context
2513  *
2514  * Also sets the context as auditable.
2515  */
2516 int auditsc_get_stamp(struct audit_context *ctx,
2517 		       struct timespec64 *t, unsigned int *serial)
2518 {
2519 	if (ctx->context == AUDIT_CTX_UNUSED)
2520 		return 0;
2521 	if (!ctx->serial)
2522 		ctx->serial = audit_serial();
2523 	t->tv_sec  = ctx->ctime.tv_sec;
2524 	t->tv_nsec = ctx->ctime.tv_nsec;
2525 	*serial    = ctx->serial;
2526 	if (!ctx->prio) {
2527 		ctx->prio = 1;
2528 		ctx->current_state = AUDIT_STATE_RECORD;
2529 	}
2530 	return 1;
2531 }
2532 
2533 /**
2534  * __audit_mq_open - record audit data for a POSIX MQ open
2535  * @oflag: open flag
2536  * @mode: mode bits
2537  * @attr: queue attributes
2538  *
2539  */
2540 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2541 {
2542 	struct audit_context *context = audit_context();
2543 
2544 	if (attr)
2545 		memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2546 	else
2547 		memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2548 
2549 	context->mq_open.oflag = oflag;
2550 	context->mq_open.mode = mode;
2551 
2552 	context->type = AUDIT_MQ_OPEN;
2553 }
2554 
2555 /**
2556  * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2557  * @mqdes: MQ descriptor
2558  * @msg_len: Message length
2559  * @msg_prio: Message priority
2560  * @abs_timeout: Message timeout in absolute time
2561  *
2562  */
2563 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2564 			const struct timespec64 *abs_timeout)
2565 {
2566 	struct audit_context *context = audit_context();
2567 	struct timespec64 *p = &context->mq_sendrecv.abs_timeout;
2568 
2569 	if (abs_timeout)
2570 		memcpy(p, abs_timeout, sizeof(*p));
2571 	else
2572 		memset(p, 0, sizeof(*p));
2573 
2574 	context->mq_sendrecv.mqdes = mqdes;
2575 	context->mq_sendrecv.msg_len = msg_len;
2576 	context->mq_sendrecv.msg_prio = msg_prio;
2577 
2578 	context->type = AUDIT_MQ_SENDRECV;
2579 }
2580 
2581 /**
2582  * __audit_mq_notify - record audit data for a POSIX MQ notify
2583  * @mqdes: MQ descriptor
2584  * @notification: Notification event
2585  *
2586  */
2587 
2588 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2589 {
2590 	struct audit_context *context = audit_context();
2591 
2592 	if (notification)
2593 		context->mq_notify.sigev_signo = notification->sigev_signo;
2594 	else
2595 		context->mq_notify.sigev_signo = 0;
2596 
2597 	context->mq_notify.mqdes = mqdes;
2598 	context->type = AUDIT_MQ_NOTIFY;
2599 }
2600 
2601 /**
2602  * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2603  * @mqdes: MQ descriptor
2604  * @mqstat: MQ flags
2605  *
2606  */
2607 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2608 {
2609 	struct audit_context *context = audit_context();
2610 
2611 	context->mq_getsetattr.mqdes = mqdes;
2612 	context->mq_getsetattr.mqstat = *mqstat;
2613 	context->type = AUDIT_MQ_GETSETATTR;
2614 }
2615 
2616 /**
2617  * __audit_ipc_obj - record audit data for ipc object
2618  * @ipcp: ipc permissions
2619  *
2620  */
2621 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2622 {
2623 	struct audit_context *context = audit_context();
2624 
2625 	context->ipc.uid = ipcp->uid;
2626 	context->ipc.gid = ipcp->gid;
2627 	context->ipc.mode = ipcp->mode;
2628 	context->ipc.has_perm = 0;
2629 	security_ipc_getsecid(ipcp, &context->ipc.osid);
2630 	context->type = AUDIT_IPC;
2631 }
2632 
2633 /**
2634  * __audit_ipc_set_perm - record audit data for new ipc permissions
2635  * @qbytes: msgq bytes
2636  * @uid: msgq user id
2637  * @gid: msgq group id
2638  * @mode: msgq mode (permissions)
2639  *
2640  * Called only after audit_ipc_obj().
2641  */
2642 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2643 {
2644 	struct audit_context *context = audit_context();
2645 
2646 	context->ipc.qbytes = qbytes;
2647 	context->ipc.perm_uid = uid;
2648 	context->ipc.perm_gid = gid;
2649 	context->ipc.perm_mode = mode;
2650 	context->ipc.has_perm = 1;
2651 }
2652 
2653 void __audit_bprm(struct linux_binprm *bprm)
2654 {
2655 	struct audit_context *context = audit_context();
2656 
2657 	context->type = AUDIT_EXECVE;
2658 	context->execve.argc = bprm->argc;
2659 }
2660 
2661 
2662 /**
2663  * __audit_socketcall - record audit data for sys_socketcall
2664  * @nargs: number of args, which should not be more than AUDITSC_ARGS.
2665  * @args: args array
2666  *
2667  */
2668 int __audit_socketcall(int nargs, unsigned long *args)
2669 {
2670 	struct audit_context *context = audit_context();
2671 
2672 	if (nargs <= 0 || nargs > AUDITSC_ARGS || !args)
2673 		return -EINVAL;
2674 	context->type = AUDIT_SOCKETCALL;
2675 	context->socketcall.nargs = nargs;
2676 	memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2677 	return 0;
2678 }
2679 
2680 /**
2681  * __audit_fd_pair - record audit data for pipe and socketpair
2682  * @fd1: the first file descriptor
2683  * @fd2: the second file descriptor
2684  *
2685  */
2686 void __audit_fd_pair(int fd1, int fd2)
2687 {
2688 	struct audit_context *context = audit_context();
2689 
2690 	context->fds[0] = fd1;
2691 	context->fds[1] = fd2;
2692 }
2693 
2694 /**
2695  * __audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2696  * @len: data length in user space
2697  * @a: data address in kernel space
2698  *
2699  * Returns 0 for success or NULL context or < 0 on error.
2700  */
2701 int __audit_sockaddr(int len, void *a)
2702 {
2703 	struct audit_context *context = audit_context();
2704 
2705 	if (!context->sockaddr) {
2706 		void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2707 
2708 		if (!p)
2709 			return -ENOMEM;
2710 		context->sockaddr = p;
2711 	}
2712 
2713 	context->sockaddr_len = len;
2714 	memcpy(context->sockaddr, a, len);
2715 	return 0;
2716 }
2717 
2718 void __audit_ptrace(struct task_struct *t)
2719 {
2720 	struct audit_context *context = audit_context();
2721 
2722 	context->target_pid = task_tgid_nr(t);
2723 	context->target_auid = audit_get_loginuid(t);
2724 	context->target_uid = task_uid(t);
2725 	context->target_sessionid = audit_get_sessionid(t);
2726 	security_task_getsecid_obj(t, &context->target_sid);
2727 	memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2728 }
2729 
2730 /**
2731  * audit_signal_info_syscall - record signal info for syscalls
2732  * @t: task being signaled
2733  *
2734  * If the audit subsystem is being terminated, record the task (pid)
2735  * and uid that is doing that.
2736  */
2737 int audit_signal_info_syscall(struct task_struct *t)
2738 {
2739 	struct audit_aux_data_pids *axp;
2740 	struct audit_context *ctx = audit_context();
2741 	kuid_t t_uid = task_uid(t);
2742 
2743 	if (!audit_signals || audit_dummy_context())
2744 		return 0;
2745 
2746 	/* optimize the common case by putting first signal recipient directly
2747 	 * in audit_context */
2748 	if (!ctx->target_pid) {
2749 		ctx->target_pid = task_tgid_nr(t);
2750 		ctx->target_auid = audit_get_loginuid(t);
2751 		ctx->target_uid = t_uid;
2752 		ctx->target_sessionid = audit_get_sessionid(t);
2753 		security_task_getsecid_obj(t, &ctx->target_sid);
2754 		memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2755 		return 0;
2756 	}
2757 
2758 	axp = (void *)ctx->aux_pids;
2759 	if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2760 		axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2761 		if (!axp)
2762 			return -ENOMEM;
2763 
2764 		axp->d.type = AUDIT_OBJ_PID;
2765 		axp->d.next = ctx->aux_pids;
2766 		ctx->aux_pids = (void *)axp;
2767 	}
2768 	BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2769 
2770 	axp->target_pid[axp->pid_count] = task_tgid_nr(t);
2771 	axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2772 	axp->target_uid[axp->pid_count] = t_uid;
2773 	axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2774 	security_task_getsecid_obj(t, &axp->target_sid[axp->pid_count]);
2775 	memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2776 	axp->pid_count++;
2777 
2778 	return 0;
2779 }
2780 
2781 /**
2782  * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2783  * @bprm: pointer to the bprm being processed
2784  * @new: the proposed new credentials
2785  * @old: the old credentials
2786  *
2787  * Simply check if the proc already has the caps given by the file and if not
2788  * store the priv escalation info for later auditing at the end of the syscall
2789  *
2790  * -Eric
2791  */
2792 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2793 			   const struct cred *new, const struct cred *old)
2794 {
2795 	struct audit_aux_data_bprm_fcaps *ax;
2796 	struct audit_context *context = audit_context();
2797 	struct cpu_vfs_cap_data vcaps;
2798 
2799 	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2800 	if (!ax)
2801 		return -ENOMEM;
2802 
2803 	ax->d.type = AUDIT_BPRM_FCAPS;
2804 	ax->d.next = context->aux;
2805 	context->aux = (void *)ax;
2806 
2807 	get_vfs_caps_from_disk(&nop_mnt_idmap,
2808 			       bprm->file->f_path.dentry, &vcaps);
2809 
2810 	ax->fcap.permitted = vcaps.permitted;
2811 	ax->fcap.inheritable = vcaps.inheritable;
2812 	ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2813 	ax->fcap.rootid = vcaps.rootid;
2814 	ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2815 
2816 	ax->old_pcap.permitted   = old->cap_permitted;
2817 	ax->old_pcap.inheritable = old->cap_inheritable;
2818 	ax->old_pcap.effective   = old->cap_effective;
2819 	ax->old_pcap.ambient     = old->cap_ambient;
2820 
2821 	ax->new_pcap.permitted   = new->cap_permitted;
2822 	ax->new_pcap.inheritable = new->cap_inheritable;
2823 	ax->new_pcap.effective   = new->cap_effective;
2824 	ax->new_pcap.ambient     = new->cap_ambient;
2825 	return 0;
2826 }
2827 
2828 /**
2829  * __audit_log_capset - store information about the arguments to the capset syscall
2830  * @new: the new credentials
2831  * @old: the old (current) credentials
2832  *
2833  * Record the arguments userspace sent to sys_capset for later printing by the
2834  * audit system if applicable
2835  */
2836 void __audit_log_capset(const struct cred *new, const struct cred *old)
2837 {
2838 	struct audit_context *context = audit_context();
2839 
2840 	context->capset.pid = task_tgid_nr(current);
2841 	context->capset.cap.effective   = new->cap_effective;
2842 	context->capset.cap.inheritable = new->cap_effective;
2843 	context->capset.cap.permitted   = new->cap_permitted;
2844 	context->capset.cap.ambient     = new->cap_ambient;
2845 	context->type = AUDIT_CAPSET;
2846 }
2847 
2848 void __audit_mmap_fd(int fd, int flags)
2849 {
2850 	struct audit_context *context = audit_context();
2851 
2852 	context->mmap.fd = fd;
2853 	context->mmap.flags = flags;
2854 	context->type = AUDIT_MMAP;
2855 }
2856 
2857 void __audit_openat2_how(struct open_how *how)
2858 {
2859 	struct audit_context *context = audit_context();
2860 
2861 	context->openat2.flags = how->flags;
2862 	context->openat2.mode = how->mode;
2863 	context->openat2.resolve = how->resolve;
2864 	context->type = AUDIT_OPENAT2;
2865 }
2866 
2867 void __audit_log_kern_module(char *name)
2868 {
2869 	struct audit_context *context = audit_context();
2870 
2871 	context->module.name = kstrdup(name, GFP_KERNEL);
2872 	if (!context->module.name)
2873 		audit_log_lost("out of memory in __audit_log_kern_module");
2874 	context->type = AUDIT_KERN_MODULE;
2875 }
2876 
2877 void __audit_fanotify(u32 response, struct fanotify_response_info_audit_rule *friar)
2878 {
2879 	/* {subj,obj}_trust values are {0,1,2}: no,yes,unknown */
2880 	switch (friar->hdr.type) {
2881 	case FAN_RESPONSE_INFO_NONE:
2882 		audit_log(audit_context(), GFP_KERNEL, AUDIT_FANOTIFY,
2883 			  "resp=%u fan_type=%u fan_info=0 subj_trust=2 obj_trust=2",
2884 			  response, FAN_RESPONSE_INFO_NONE);
2885 		break;
2886 	case FAN_RESPONSE_INFO_AUDIT_RULE:
2887 		audit_log(audit_context(), GFP_KERNEL, AUDIT_FANOTIFY,
2888 			  "resp=%u fan_type=%u fan_info=%X subj_trust=%u obj_trust=%u",
2889 			  response, friar->hdr.type, friar->rule_number,
2890 			  friar->subj_trust, friar->obj_trust);
2891 	}
2892 }
2893 
2894 void __audit_tk_injoffset(struct timespec64 offset)
2895 {
2896 	struct audit_context *context = audit_context();
2897 
2898 	/* only set type if not already set by NTP */
2899 	if (!context->type)
2900 		context->type = AUDIT_TIME_INJOFFSET;
2901 	memcpy(&context->time.tk_injoffset, &offset, sizeof(offset));
2902 }
2903 
2904 void __audit_ntp_log(const struct audit_ntp_data *ad)
2905 {
2906 	struct audit_context *context = audit_context();
2907 	int type;
2908 
2909 	for (type = 0; type < AUDIT_NTP_NVALS; type++)
2910 		if (ad->vals[type].newval != ad->vals[type].oldval) {
2911 			/* unconditionally set type, overwriting TK */
2912 			context->type = AUDIT_TIME_ADJNTPVAL;
2913 			memcpy(&context->time.ntp_data, ad, sizeof(*ad));
2914 			break;
2915 		}
2916 }
2917 
2918 void __audit_log_nfcfg(const char *name, u8 af, unsigned int nentries,
2919 		       enum audit_nfcfgop op, gfp_t gfp)
2920 {
2921 	struct audit_buffer *ab;
2922 	char comm[sizeof(current->comm)];
2923 
2924 	ab = audit_log_start(audit_context(), gfp, AUDIT_NETFILTER_CFG);
2925 	if (!ab)
2926 		return;
2927 	audit_log_format(ab, "table=%s family=%u entries=%u op=%s",
2928 			 name, af, nentries, audit_nfcfgs[op].s);
2929 
2930 	audit_log_format(ab, " pid=%u", task_pid_nr(current));
2931 	audit_log_task_context(ab); /* subj= */
2932 	audit_log_format(ab, " comm=");
2933 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2934 	audit_log_end(ab);
2935 }
2936 EXPORT_SYMBOL_GPL(__audit_log_nfcfg);
2937 
2938 static void audit_log_task(struct audit_buffer *ab)
2939 {
2940 	kuid_t auid, uid;
2941 	kgid_t gid;
2942 	unsigned int sessionid;
2943 	char comm[sizeof(current->comm)];
2944 
2945 	auid = audit_get_loginuid(current);
2946 	sessionid = audit_get_sessionid(current);
2947 	current_uid_gid(&uid, &gid);
2948 
2949 	audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2950 			 from_kuid(&init_user_ns, auid),
2951 			 from_kuid(&init_user_ns, uid),
2952 			 from_kgid(&init_user_ns, gid),
2953 			 sessionid);
2954 	audit_log_task_context(ab);
2955 	audit_log_format(ab, " pid=%d comm=", task_tgid_nr(current));
2956 	audit_log_untrustedstring(ab, get_task_comm(comm, current));
2957 	audit_log_d_path_exe(ab, current->mm);
2958 }
2959 
2960 /**
2961  * audit_core_dumps - record information about processes that end abnormally
2962  * @signr: signal value
2963  *
2964  * If a process ends with a core dump, something fishy is going on and we
2965  * should record the event for investigation.
2966  */
2967 void audit_core_dumps(long signr)
2968 {
2969 	struct audit_buffer *ab;
2970 
2971 	if (!audit_enabled)
2972 		return;
2973 
2974 	if (signr == SIGQUIT)	/* don't care for those */
2975 		return;
2976 
2977 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_ANOM_ABEND);
2978 	if (unlikely(!ab))
2979 		return;
2980 	audit_log_task(ab);
2981 	audit_log_format(ab, " sig=%ld res=1", signr);
2982 	audit_log_end(ab);
2983 }
2984 
2985 /**
2986  * audit_seccomp - record information about a seccomp action
2987  * @syscall: syscall number
2988  * @signr: signal value
2989  * @code: the seccomp action
2990  *
2991  * Record the information associated with a seccomp action. Event filtering for
2992  * seccomp actions that are not to be logged is done in seccomp_log().
2993  * Therefore, this function forces auditing independent of the audit_enabled
2994  * and dummy context state because seccomp actions should be logged even when
2995  * audit is not in use.
2996  */
2997 void audit_seccomp(unsigned long syscall, long signr, int code)
2998 {
2999 	struct audit_buffer *ab;
3000 
3001 	ab = audit_log_start(audit_context(), GFP_KERNEL, AUDIT_SECCOMP);
3002 	if (unlikely(!ab))
3003 		return;
3004 	audit_log_task(ab);
3005 	audit_log_format(ab, " sig=%ld arch=%x syscall=%ld compat=%d ip=0x%lx code=0x%x",
3006 			 signr, syscall_get_arch(current), syscall,
3007 			 in_compat_syscall(), KSTK_EIP(current), code);
3008 	audit_log_end(ab);
3009 }
3010 
3011 void audit_seccomp_actions_logged(const char *names, const char *old_names,
3012 				  int res)
3013 {
3014 	struct audit_buffer *ab;
3015 
3016 	if (!audit_enabled)
3017 		return;
3018 
3019 	ab = audit_log_start(audit_context(), GFP_KERNEL,
3020 			     AUDIT_CONFIG_CHANGE);
3021 	if (unlikely(!ab))
3022 		return;
3023 
3024 	audit_log_format(ab,
3025 			 "op=seccomp-logging actions=%s old-actions=%s res=%d",
3026 			 names, old_names, res);
3027 	audit_log_end(ab);
3028 }
3029 
3030 struct list_head *audit_killed_trees(void)
3031 {
3032 	struct audit_context *ctx = audit_context();
3033 	if (likely(!ctx || ctx->context == AUDIT_CTX_UNUSED))
3034 		return NULL;
3035 	return &ctx->killed_trees;
3036 }
3037