linux - kernel/auditsc.c

2733 lines · 71.5 KiB

1	/* auditsc.c -- System-call auditing support
2	* Handles all system-call specific auditing features.
3	*
4	* Copyright 2003-2004 Red Hat Inc., Durham, North Carolina.
5	* Copyright 2005 Hewlett-Packard Development Company, L.P.
6	* Copyright (C) 2005, 2006 IBM Corporation
7	* All Rights Reserved.
8	*
9	* This program is free software; you can redistribute it and/or modify
10	* it under the terms of the GNU General Public License as published by
11	* the Free Software Foundation; either version 2 of the License, or
12	* (at your option) any later version.
13	*
14	* This program is distributed in the hope that it will be useful,
15	* but WITHOUT ANY WARRANTY; without even the implied warranty of
16	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17	* GNU General Public License for more details.
18	*
19	* You should have received a copy of the GNU General Public License
20	* along with this program; if not, write to the Free Software
21	* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22	*
23	* Written by Rickard E. (Rik) Faith <[email protected]>
24	*
25	* Many of the ideas implemented here are from Stephen C. Tweedie,
26	* especially the idea of avoiding a copy by using getname.
27	*
28	* The method for actual interception of syscall entry and exit (not in
29	* this file -- see entry.S) is based on a GPL'd patch written by
30	* [email protected] and Copyright 2003 SuSE Linux AG.
31	*
32	* POSIX message queue support added by George Wilson <[email protected]>,
33	* 2006.
34	*
35	* The support of additional filter rules compares (>, <, >=, <=) was
36	* added by Dustin Kirkland <[email protected]>, 2005.
37	*
38	* Modified by Amy Griffis <[email protected]> to collect additional
39	* filesystem information.
40	*
41	* Subject and object context labeling support added by <[email protected]>
42	* and <[email protected]> for LSPP certification compliance.
43	*/
44
45	#include <linux/init.h>
46	#include <asm/types.h>
47	#include <linux/atomic.h>
48	#include <linux/fs.h>
49	#include <linux/namei.h>
50	#include <linux/mm.h>
51	#include <linux/export.h>
52	#include <linux/slab.h>
53	#include <linux/mount.h>
54	#include <linux/socket.h>
55	#include <linux/mqueue.h>
56	#include <linux/audit.h>
57	#include <linux/personality.h>
58	#include <linux/time.h>
59	#include <linux/netlink.h>
60	#include <linux/compiler.h>
61	#include <asm/unistd.h>
62	#include <linux/security.h>
63	#include <linux/list.h>
64	#include <linux/tty.h>
65	#include <linux/binfmts.h>
66	#include <linux/highmem.h>
67	#include <linux/syscalls.h>
68	#include <linux/capability.h>
69	#include <linux/fs_struct.h>
70	#include <linux/compat.h>
71
72	#include "audit.h"
73
74	/* flags stating the success for a syscall */
75	#define AUDITSC_INVALID 0
76	#define AUDITSC_SUCCESS 1
77	#define AUDITSC_FAILURE 2
78
79	/* AUDIT_NAMES is the number of slots we reserve in the audit_context
80	* for saving names from getname(). If we get more names we will allocate
81	* a name dynamically and also add those to the list anchored by names_list. */
82	#define AUDIT_NAMES 5
83
84	/* Indicates that audit should log the full pathname. */
85	#define AUDIT_NAME_FULL -1
86
87	/* no execve audit message should be longer than this (userspace limits) */
88	#define MAX_EXECVE_AUDIT_LEN 7500
89
90	/* number of audit rules */
91	int audit_n_rules;
92
93	/* determines whether we collect data for signals sent */
94	int audit_signals;
95
96	struct audit_cap_data {
97	kernel_cap_t permitted;
98	kernel_cap_t inheritable;
99	union {
100	unsigned int fE; /* effective bit of a file capability */
101	kernel_cap_t effective; /* effective set of a process */
102	};
103	};
104
105	/* When fs/namei.c:getname() is called, we store the pointer in name and
106	* we don't let putname() free it (instead we free all of the saved
107	* pointers at syscall exit time).
108	*
109	* Further, in fs/namei.c:path_lookup() we store the inode and device. */
110	struct audit_names {
111	struct list_head list; /* audit_context->names_list */
112	const char *name;
113	unsigned long ino;
114	dev_t dev;
115	umode_t mode;
116	uid_t uid;
117	gid_t gid;
118	dev_t rdev;
119	u32 osid;
120	struct audit_cap_data fcap;
121	unsigned int fcap_ver;
122	int name_len; /* number of name's characters to log */
123	bool name_put; /* call __putname() for this name */
124	/*
125	* This was an allocated audit_names and not from the array of
126	* names allocated in the task audit context. Thus this name
127	* should be freed on syscall exit
128	*/
129	bool should_free;
130	};
131
132	struct audit_aux_data {
133	struct audit_aux_data *next;
134	int type;
135	};
136
137	#define AUDIT_AUX_IPCPERM 0
138
139	/* Number of target pids per aux struct. */
140	#define AUDIT_AUX_PIDS 16
141
142	struct audit_aux_data_execve {
143	struct audit_aux_data d;
144	int argc;
145	int envc;
146	struct mm_struct *mm;
147	};
148
149	struct audit_aux_data_pids {
150	struct audit_aux_data d;
151	pid_t target_pid[AUDIT_AUX_PIDS];
152	uid_t target_auid[AUDIT_AUX_PIDS];
153	uid_t target_uid[AUDIT_AUX_PIDS];
154	unsigned int target_sessionid[AUDIT_AUX_PIDS];
155	u32 target_sid[AUDIT_AUX_PIDS];
156	char target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
157	int pid_count;
158	};
159
160	struct audit_aux_data_bprm_fcaps {
161	struct audit_aux_data d;
162	struct audit_cap_data fcap;
163	unsigned int fcap_ver;
164	struct audit_cap_data old_pcap;
165	struct audit_cap_data new_pcap;
166	};
167
168	struct audit_aux_data_capset {
169	struct audit_aux_data d;
170	pid_t pid;
171	struct audit_cap_data cap;
172	};
173
174	struct audit_tree_refs {
175	struct audit_tree_refs *next;
176	struct audit_chunk *c[31];
177	};
178
179	/* The per-task audit context. */
180	struct audit_context {
181	int dummy; /* must be the first element */
182	int in_syscall; /* 1 if task is in a syscall */
183	enum audit_state state, current_state;
184	unsigned int serial; /* serial number for record */
185	int major; /* syscall number */
186	struct timespec ctime; /* time of syscall entry */
187	unsigned long argv[4]; /* syscall arguments */
188	long return_code;/* syscall return code */
189	u64 prio;
190	int return_valid; /* return code is valid */
191	/*
192	* The names_list is the list of all audit_names collected during this
193	* syscall. The first AUDIT_NAMES entries in the names_list will
194	* actually be from the preallocated_names array for performance
195	* reasons. Except during allocation they should never be referenced
196	* through the preallocated_names array and should only be found/used
197	* by running the names_list.
198	*/
199	struct audit_names preallocated_names[AUDIT_NAMES];
200	int name_count; /* total records in names_list */
201	struct list_head names_list; /* anchor for struct audit_names->list */
202	char * filterkey; /* key for rule that triggered record */
203	struct path pwd;
204	struct audit_context previous; / For nested syscalls */
205	struct audit_aux_data *aux;
206	struct audit_aux_data *aux_pids;
207	struct sockaddr_storage *sockaddr;
208	size_t sockaddr_len;
209	/* Save things to print about task_struct */
210	pid_t pid, ppid;
211	uid_t uid, euid, suid, fsuid;
212	gid_t gid, egid, sgid, fsgid;
213	unsigned long personality;
214	int arch;
215
216	pid_t target_pid;
217	uid_t target_auid;
218	uid_t target_uid;
219	unsigned int target_sessionid;
220	u32 target_sid;
221	char target_comm[TASK_COMM_LEN];
222
223	struct audit_tree_refs trees, first_trees;
224	struct list_head killed_trees;
225	int tree_count;
226
227	int type;
228	union {
229	struct {
230	int nargs;
231	long args[6];
232	} socketcall;
233	struct {
234	uid_t uid;
235	gid_t gid;
236	umode_t mode;
237	u32 osid;
238	int has_perm;
239	uid_t perm_uid;
240	gid_t perm_gid;
241	umode_t perm_mode;
242	unsigned long qbytes;
243	} ipc;
244	struct {
245	mqd_t mqdes;
246	struct mq_attr mqstat;
247	} mq_getsetattr;
248	struct {
249	mqd_t mqdes;
250	int sigev_signo;
251	} mq_notify;
252	struct {
253	mqd_t mqdes;
254	size_t msg_len;
255	unsigned int msg_prio;
256	struct timespec abs_timeout;
257	} mq_sendrecv;
258	struct {
259	int oflag;
260	umode_t mode;
261	struct mq_attr attr;
262	} mq_open;
263	struct {
264	pid_t pid;
265	struct audit_cap_data cap;
266	} capset;
267	struct {
268	int fd;
269	int flags;
270	} mmap;
271	};
272	int fds[2];
273
274	#if AUDIT_DEBUG
275	int put_count;
276	int ino_count;
277	#endif
278	};
279
280	static inline int open_arg(int flags, int mask)
281	{
282	int n = ACC_MODE(flags);
283	if (flags & (O_TRUNC \| O_CREAT))
284	n \|= AUDIT_PERM_WRITE;
285	return n & mask;
286	}
287
288	static int audit_match_perm(struct audit_context *ctx, int mask)
289	{
290	unsigned n;
291	if (unlikely(!ctx))
292	return 0;
293	n = ctx->major;
294
295	switch (audit_classify_syscall(ctx->arch, n)) {
296	case 0: /* native */
297	if ((mask & AUDIT_PERM_WRITE) &&
298	audit_match_class(AUDIT_CLASS_WRITE, n))
299	return 1;
300	if ((mask & AUDIT_PERM_READ) &&
301	audit_match_class(AUDIT_CLASS_READ, n))
302	return 1;
303	if ((mask & AUDIT_PERM_ATTR) &&
304	audit_match_class(AUDIT_CLASS_CHATTR, n))
305	return 1;
306	return 0;
307	case 1: /* 32bit on biarch */
308	if ((mask & AUDIT_PERM_WRITE) &&
309	audit_match_class(AUDIT_CLASS_WRITE_32, n))
310	return 1;
311	if ((mask & AUDIT_PERM_READ) &&
312	audit_match_class(AUDIT_CLASS_READ_32, n))
313	return 1;
314	if ((mask & AUDIT_PERM_ATTR) &&
315	audit_match_class(AUDIT_CLASS_CHATTR_32, n))
316	return 1;
317	return 0;
318	case 2: /* open */
319	return mask & ACC_MODE(ctx->argv[1]);
320	case 3: /* openat */
321	return mask & ACC_MODE(ctx->argv[2]);
322	case 4: /* socketcall */
323	return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
324	case 5: /* execve */
325	return mask & AUDIT_PERM_EXEC;
326	default:
327	return 0;
328	}
329	}
330
331	static int audit_match_filetype(struct audit_context *ctx, int val)
332	{
333	struct audit_names *n;
334	umode_t mode = (umode_t)val;
335
336	if (unlikely(!ctx))
337	return 0;
338
339	list_for_each_entry(n, &ctx->names_list, list) {
340	if ((n->ino != -1) &&
341	((n->mode & S_IFMT) == mode))
342	return 1;
343	}
344
345	return 0;
346	}
347
348	/*
349	* We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
350	* ->first_trees points to its beginning, ->trees - to the current end of data.
351	* ->tree_count is the number of free entries in array pointed to by ->trees.
352	* Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
353	* "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
354	* it's going to remain 1-element for almost any setup) until we free context itself.
355	* References in it _are_ dropped - at the same time we free/drop aux stuff.
356	*/
357
358	#ifdef CONFIG_AUDIT_TREE
359	static void audit_set_auditable(struct audit_context *ctx)
360	{
361	if (!ctx->prio) {
362	ctx->prio = 1;
363	ctx->current_state = AUDIT_RECORD_CONTEXT;
364	}
365	}
366
367	static int put_tree_ref(struct audit_context ctx, struct audit_chunk chunk)
368	{
369	struct audit_tree_refs *p = ctx->trees;
370	int left = ctx->tree_count;
371	if (likely(left)) {
372	p->c[--left] = chunk;
373	ctx->tree_count = left;
374	return 1;
375	}
376	if (!p)
377	return 0;
378	p = p->next;
379	if (p) {
380	p->c[30] = chunk;
381	ctx->trees = p;
382	ctx->tree_count = 30;
383	return 1;
384	}
385	return 0;
386	}
387
388	static int grow_tree_refs(struct audit_context *ctx)
389	{
390	struct audit_tree_refs *p = ctx->trees;
391	ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
392	if (!ctx->trees) {
393	ctx->trees = p;
394	return 0;
395	}
396	if (p)
397	p->next = ctx->trees;
398	else
399	ctx->first_trees = ctx->trees;
400	ctx->tree_count = 31;
401	return 1;
402	}
403	#endif
404
405	static void unroll_tree_refs(struct audit_context *ctx,
406	struct audit_tree_refs *p, int count)
407	{
408	#ifdef CONFIG_AUDIT_TREE
409	struct audit_tree_refs *q;
410	int n;
411	if (!p) {
412	/* we started with empty chain */
413	p = ctx->first_trees;
414	count = 31;
415	/* if the very first allocation has failed, nothing to do */
416	if (!p)
417	return;
418	}
419	n = count;
420	for (q = p; q != ctx->trees; q = q->next, n = 31) {
421	while (n--) {
422	audit_put_chunk(q->c[n]);
423	q->c[n] = NULL;
424	}
425	}
426	while (n-- > ctx->tree_count) {
427	audit_put_chunk(q->c[n]);
428	q->c[n] = NULL;
429	}
430	ctx->trees = p;
431	ctx->tree_count = count;
432	#endif
433	}
434
435	static void free_tree_refs(struct audit_context *ctx)
436	{
437	struct audit_tree_refs p, q;
438	for (p = ctx->first_trees; p; p = q) {
439	q = p->next;
440	kfree(p);
441	}
442	}
443
444	static int match_tree_refs(struct audit_context ctx, struct audit_tree tree)
445	{
446	#ifdef CONFIG_AUDIT_TREE
447	struct audit_tree_refs *p;
448	int n;
449	if (!tree)
450	return 0;
451	/* full ones */
452	for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
453	for (n = 0; n < 31; n++)
454	if (audit_tree_match(p->c[n], tree))
455	return 1;
456	}
457	/* partial */
458	if (p) {
459	for (n = ctx->tree_count; n < 31; n++)
460	if (audit_tree_match(p->c[n], tree))
461	return 1;
462	}
463	#endif
464	return 0;
465	}
466
467	static int audit_compare_id(uid_t uid1,
468	struct audit_names *name,
469	unsigned long name_offset,
470	struct audit_field *f,
471	struct audit_context *ctx)
472	{
473	struct audit_names *n;
474	unsigned long addr;
475	uid_t uid2;
476	int rc;
477
478	BUILD_BUG_ON(sizeof(uid_t) != sizeof(gid_t));
479
480	if (name) {
481	addr = (unsigned long)name;
482	addr += name_offset;
483
484	uid2 = (uid_t )addr;
485	rc = audit_comparator(uid1, f->op, uid2);
486	if (rc)
487	return rc;
488	}
489
490	if (ctx) {
491	list_for_each_entry(n, &ctx->names_list, list) {
492	addr = (unsigned long)n;
493	addr += name_offset;
494
495	uid2 = (uid_t )addr;
496
497	rc = audit_comparator(uid1, f->op, uid2);
498	if (rc)
499	return rc;
500	}
501	}
502	return 0;
503	}
504
505	static int audit_field_compare(struct task_struct *tsk,
506	const struct cred *cred,
507	struct audit_field *f,
508	struct audit_context *ctx,
509	struct audit_names *name)
510	{
511	switch (f->val) {
512	/* process to file object comparisons */
513	case AUDIT_COMPARE_UID_TO_OBJ_UID:
514	return audit_compare_id(cred->uid,
515	name, offsetof(struct audit_names, uid),
516	f, ctx);
517	case AUDIT_COMPARE_GID_TO_OBJ_GID:
518	return audit_compare_id(cred->gid,
519	name, offsetof(struct audit_names, gid),
520	f, ctx);
521	case AUDIT_COMPARE_EUID_TO_OBJ_UID:
522	return audit_compare_id(cred->euid,
523	name, offsetof(struct audit_names, uid),
524	f, ctx);
525	case AUDIT_COMPARE_EGID_TO_OBJ_GID:
526	return audit_compare_id(cred->egid,
527	name, offsetof(struct audit_names, gid),
528	f, ctx);
529	case AUDIT_COMPARE_AUID_TO_OBJ_UID:
530	return audit_compare_id(tsk->loginuid,
531	name, offsetof(struct audit_names, uid),
532	f, ctx);
533	case AUDIT_COMPARE_SUID_TO_OBJ_UID:
534	return audit_compare_id(cred->suid,
535	name, offsetof(struct audit_names, uid),
536	f, ctx);
537	case AUDIT_COMPARE_SGID_TO_OBJ_GID:
538	return audit_compare_id(cred->sgid,
539	name, offsetof(struct audit_names, gid),
540	f, ctx);
541	case AUDIT_COMPARE_FSUID_TO_OBJ_UID:
542	return audit_compare_id(cred->fsuid,
543	name, offsetof(struct audit_names, uid),
544	f, ctx);
545	case AUDIT_COMPARE_FSGID_TO_OBJ_GID:
546	return audit_compare_id(cred->fsgid,
547	name, offsetof(struct audit_names, gid),
548	f, ctx);
549	/* uid comparisons */
550	case AUDIT_COMPARE_UID_TO_AUID:
551	return audit_comparator(cred->uid, f->op, tsk->loginuid);
552	case AUDIT_COMPARE_UID_TO_EUID:
553	return audit_comparator(cred->uid, f->op, cred->euid);
554	case AUDIT_COMPARE_UID_TO_SUID:
555	return audit_comparator(cred->uid, f->op, cred->suid);
556	case AUDIT_COMPARE_UID_TO_FSUID:
557	return audit_comparator(cred->uid, f->op, cred->fsuid);
558	/* auid comparisons */
559	case AUDIT_COMPARE_AUID_TO_EUID:
560	return audit_comparator(tsk->loginuid, f->op, cred->euid);
561	case AUDIT_COMPARE_AUID_TO_SUID:
562	return audit_comparator(tsk->loginuid, f->op, cred->suid);
563	case AUDIT_COMPARE_AUID_TO_FSUID:
564	return audit_comparator(tsk->loginuid, f->op, cred->fsuid);
565	/* euid comparisons */
566	case AUDIT_COMPARE_EUID_TO_SUID:
567	return audit_comparator(cred->euid, f->op, cred->suid);
568	case AUDIT_COMPARE_EUID_TO_FSUID:
569	return audit_comparator(cred->euid, f->op, cred->fsuid);
570	/* suid comparisons */
571	case AUDIT_COMPARE_SUID_TO_FSUID:
572	return audit_comparator(cred->suid, f->op, cred->fsuid);
573	/* gid comparisons */
574	case AUDIT_COMPARE_GID_TO_EGID:
575	return audit_comparator(cred->gid, f->op, cred->egid);
576	case AUDIT_COMPARE_GID_TO_SGID:
577	return audit_comparator(cred->gid, f->op, cred->sgid);
578	case AUDIT_COMPARE_GID_TO_FSGID:
579	return audit_comparator(cred->gid, f->op, cred->fsgid);
580	/* egid comparisons */
581	case AUDIT_COMPARE_EGID_TO_SGID:
582	return audit_comparator(cred->egid, f->op, cred->sgid);
583	case AUDIT_COMPARE_EGID_TO_FSGID:
584	return audit_comparator(cred->egid, f->op, cred->fsgid);
585	/* sgid comparison */
586	case AUDIT_COMPARE_SGID_TO_FSGID:
587	return audit_comparator(cred->sgid, f->op, cred->fsgid);
588	default:
589	WARN(1, "Missing AUDIT_COMPARE define. Report as a bug\n");
590	return 0;
591	}
592	return 0;
593	}
594
595	/* Determine if any context name data matches a rule's watch data */
596	/* Compare a task_struct with an audit_rule. Return 1 on match, 0
597	* otherwise.
598	*
599	* If task_creation is true, this is an explicit indication that we are
600	* filtering a task rule at task creation time. This and tsk == current are
601	* the only situations where tsk->cred may be accessed without an rcu read lock.
602	*/
603	static int audit_filter_rules(struct task_struct *tsk,
604	struct audit_krule *rule,
605	struct audit_context *ctx,
606	struct audit_names *name,
607	enum audit_state *state,
608	bool task_creation)
609	{
610	const struct cred *cred;
611	int i, need_sid = 1;
612	u32 sid;
613
614	cred = rcu_dereference_check(tsk->cred, tsk == current \|\| task_creation);
615
616	for (i = 0; i < rule->field_count; i++) {
617	struct audit_field *f = &rule->fields[i];
618	struct audit_names *n;
619	int result = 0;
620
621	switch (f->type) {
622	case AUDIT_PID:
623	result = audit_comparator(tsk->pid, f->op, f->val);
624	break;
625	case AUDIT_PPID:
626	if (ctx) {
627	if (!ctx->ppid)
628	ctx->ppid = sys_getppid();
629	result = audit_comparator(ctx->ppid, f->op, f->val);
630	}
631	break;
632	case AUDIT_UID:
633	result = audit_comparator(cred->uid, f->op, f->val);
634	break;
635	case AUDIT_EUID:
636	result = audit_comparator(cred->euid, f->op, f->val);
637	break;
638	case AUDIT_SUID:
639	result = audit_comparator(cred->suid, f->op, f->val);
640	break;
641	case AUDIT_FSUID:
642	result = audit_comparator(cred->fsuid, f->op, f->val);
643	break;
644	case AUDIT_GID:
645	result = audit_comparator(cred->gid, f->op, f->val);
646	break;
647	case AUDIT_EGID:
648	result = audit_comparator(cred->egid, f->op, f->val);
649	break;
650	case AUDIT_SGID:
651	result = audit_comparator(cred->sgid, f->op, f->val);
652	break;
653	case AUDIT_FSGID:
654	result = audit_comparator(cred->fsgid, f->op, f->val);
655	break;
656	case AUDIT_PERS:
657	result = audit_comparator(tsk->personality, f->op, f->val);
658	break;
659	case AUDIT_ARCH:
660	if (ctx)
661	result = audit_comparator(ctx->arch, f->op, f->val);
662	break;
663
664	case AUDIT_EXIT:
665	if (ctx && ctx->return_valid)
666	result = audit_comparator(ctx->return_code, f->op, f->val);
667	break;
668	case AUDIT_SUCCESS:
669	if (ctx && ctx->return_valid) {
670	if (f->val)
671	result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
672	else
673	result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
674	}
675	break;
676	case AUDIT_DEVMAJOR:
677	if (name) {
678	if (audit_comparator(MAJOR(name->dev), f->op, f->val) \|\|
679	audit_comparator(MAJOR(name->rdev), f->op, f->val))
680	++result;
681	} else if (ctx) {
682	list_for_each_entry(n, &ctx->names_list, list) {
683	if (audit_comparator(MAJOR(n->dev), f->op, f->val) \|\|
684	audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
685	++result;
686	break;
687	}
688	}
689	}
690	break;
691	case AUDIT_DEVMINOR:
692	if (name) {
693	if (audit_comparator(MINOR(name->dev), f->op, f->val) \|\|
694	audit_comparator(MINOR(name->rdev), f->op, f->val))
695	++result;
696	} else if (ctx) {
697	list_for_each_entry(n, &ctx->names_list, list) {
698	if (audit_comparator(MINOR(n->dev), f->op, f->val) \|\|
699	audit_comparator(MINOR(n->rdev), f->op, f->val)) {
700	++result;
701	break;
702	}
703	}
704	}
705	break;
706	case AUDIT_INODE:
707	if (name)
708	result = (name->ino == f->val);
709	else if (ctx) {
710	list_for_each_entry(n, &ctx->names_list, list) {
711	if (audit_comparator(n->ino, f->op, f->val)) {
712	++result;
713	break;
714	}
715	}
716	}
717	break;
718	case AUDIT_OBJ_UID:
719	if (name) {
720	result = audit_comparator(name->uid, f->op, f->val);
721	} else if (ctx) {
722	list_for_each_entry(n, &ctx->names_list, list) {
723	if (audit_comparator(n->uid, f->op, f->val)) {
724	++result;
725	break;
726	}
727	}
728	}
729	break;
730	case AUDIT_OBJ_GID:
731	if (name) {
732	result = audit_comparator(name->gid, f->op, f->val);
733	} else if (ctx) {
734	list_for_each_entry(n, &ctx->names_list, list) {
735	if (audit_comparator(n->gid, f->op, f->val)) {
736	++result;
737	break;
738	}
739	}
740	}
741	break;
742	case AUDIT_WATCH:
743	if (name)
744	result = audit_watch_compare(rule->watch, name->ino, name->dev);
745	break;
746	case AUDIT_DIR:
747	if (ctx)
748	result = match_tree_refs(ctx, rule->tree);
749	break;
750	case AUDIT_LOGINUID:
751	result = 0;
752	if (ctx)
753	result = audit_comparator(tsk->loginuid, f->op, f->val);
754	break;
755	case AUDIT_SUBJ_USER:
756	case AUDIT_SUBJ_ROLE:
757	case AUDIT_SUBJ_TYPE:
758	case AUDIT_SUBJ_SEN:
759	case AUDIT_SUBJ_CLR:
760	/* NOTE: this may return negative values indicating
761	a temporary error. We simply treat this as a
762	match for now to avoid losing information that
763	may be wanted. An error message will also be
764	logged upon error */
765	if (f->lsm_rule) {
766	if (need_sid) {
767	security_task_getsecid(tsk, &sid);
768	need_sid = 0;
769	}
770	result = security_audit_rule_match(sid, f->type,
771	f->op,
772	f->lsm_rule,
773	ctx);
774	}
775	break;
776	case AUDIT_OBJ_USER:
777	case AUDIT_OBJ_ROLE:
778	case AUDIT_OBJ_TYPE:
779	case AUDIT_OBJ_LEV_LOW:
780	case AUDIT_OBJ_LEV_HIGH:
781	/* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
782	also applies here */
783	if (f->lsm_rule) {
784	/* Find files that match */
785	if (name) {
786	result = security_audit_rule_match(
787	name->osid, f->type, f->op,
788	f->lsm_rule, ctx);
789	} else if (ctx) {
790	list_for_each_entry(n, &ctx->names_list, list) {
791	if (security_audit_rule_match(n->osid, f->type,
792	f->op, f->lsm_rule,
793	ctx)) {
794	++result;
795	break;
796	}
797	}
798	}
799	/* Find ipc objects that match */
800	if (!ctx \|\| ctx->type != AUDIT_IPC)
801	break;
802	if (security_audit_rule_match(ctx->ipc.osid,
803	f->type, f->op,
804	f->lsm_rule, ctx))
805	++result;
806	}
807	break;
808	case AUDIT_ARG0:
809	case AUDIT_ARG1:
810	case AUDIT_ARG2:
811	case AUDIT_ARG3:
812	if (ctx)
813	result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
814	break;
815	case AUDIT_FILTERKEY:
816	/* ignore this field for filtering */
817	result = 1;
818	break;
819	case AUDIT_PERM:
820	result = audit_match_perm(ctx, f->val);
821	break;
822	case AUDIT_FILETYPE:
823	result = audit_match_filetype(ctx, f->val);
824	break;
825	case AUDIT_FIELD_COMPARE:
826	result = audit_field_compare(tsk, cred, f, ctx, name);
827	break;
828	}
829	if (!result)
830	return 0;
831	}
832
833	if (ctx) {
834	if (rule->prio <= ctx->prio)
835	return 0;
836	if (rule->filterkey) {
837	kfree(ctx->filterkey);
838	ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
839	}
840	ctx->prio = rule->prio;
841	}
842	switch (rule->action) {
843	case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
844	case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
845	}
846	return 1;
847	}
848
849	/* At process creation time, we can determine if system-call auditing is
850	* completely disabled for this task. Since we only have the task
851	* structure at this point, we can only check uid and gid.
852	*/
853	static enum audit_state audit_filter_task(struct task_struct tsk, char *key)
854	{
855	struct audit_entry *e;
856	enum audit_state state;
857
858	rcu_read_lock();
859	list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
860	if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
861	&state, true)) {
862	if (state == AUDIT_RECORD_CONTEXT)
863	*key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
864	rcu_read_unlock();
865	return state;
866	}
867	}
868	rcu_read_unlock();
869	return AUDIT_BUILD_CONTEXT;
870	}
871
872	/* At syscall entry and exit time, this filter is called if the
873	* audit_state is not low enough that auditing cannot take place, but is
874	* also not high enough that we already know we have to write an audit
875	* record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
876	*/
877	static enum audit_state audit_filter_syscall(struct task_struct *tsk,
878	struct audit_context *ctx,
879	struct list_head *list)
880	{
881	struct audit_entry *e;
882	enum audit_state state;
883
884	if (audit_pid && tsk->tgid == audit_pid)
885	return AUDIT_DISABLED;
886
887	rcu_read_lock();
888	if (!list_empty(list)) {
889	int word = AUDIT_WORD(ctx->major);
890	int bit = AUDIT_BIT(ctx->major);
891
892	list_for_each_entry_rcu(e, list, list) {
893	if ((e->rule.mask[word] & bit) == bit &&
894	audit_filter_rules(tsk, &e->rule, ctx, NULL,
895	&state, false)) {
896	rcu_read_unlock();
897	ctx->current_state = state;
898	return state;
899	}
900	}
901	}
902	rcu_read_unlock();
903	return AUDIT_BUILD_CONTEXT;
904	}
905
906	/*
907	* Given an audit_name check the inode hash table to see if they match.
908	* Called holding the rcu read lock to protect the use of audit_inode_hash
909	*/
910	static int audit_filter_inode_name(struct task_struct *tsk,
911	struct audit_names *n,
912	struct audit_context *ctx) {
913	int word, bit;
914	int h = audit_hash_ino((u32)n->ino);
915	struct list_head *list = &audit_inode_hash[h];
916	struct audit_entry *e;
917	enum audit_state state;
918
919	word = AUDIT_WORD(ctx->major);
920	bit = AUDIT_BIT(ctx->major);
921
922	if (list_empty(list))
923	return 0;
924
925	list_for_each_entry_rcu(e, list, list) {
926	if ((e->rule.mask[word] & bit) == bit &&
927	audit_filter_rules(tsk, &e->rule, ctx, n, &state, false)) {
928	ctx->current_state = state;
929	return 1;
930	}
931	}
932
933	return 0;
934	}
935
936	/* At syscall exit time, this filter is called if any audit_names have been
937	* collected during syscall processing. We only check rules in sublists at hash
938	* buckets applicable to the inode numbers in audit_names.
939	* Regarding audit_state, same rules apply as for audit_filter_syscall().
940	*/
941	void audit_filter_inodes(struct task_struct tsk, struct audit_context ctx)
942	{
943	struct audit_names *n;
944
945	if (audit_pid && tsk->tgid == audit_pid)
946	return;
947
948	rcu_read_lock();
949
950	list_for_each_entry(n, &ctx->names_list, list) {
951	if (audit_filter_inode_name(tsk, n, ctx))
952	break;
953	}
954	rcu_read_unlock();
955	}
956
957	static inline struct audit_context audit_get_context(struct task_struct tsk,
958	int return_valid,
959	long return_code)
960	{
961	struct audit_context *context = tsk->audit_context;
962
963	if (!context)
964	return NULL;
965	context->return_valid = return_valid;
966
967	/*
968	* we need to fix up the return code in the audit logs if the actual
969	* return codes are later going to be fixed up by the arch specific
970	* signal handlers
971	*
972	* This is actually a test for:
973	* (rc == ERESTARTSYS ) \|\| (rc == ERESTARTNOINTR) \|\|
974	* (rc == ERESTARTNOHAND) \|\| (rc == ERESTART_RESTARTBLOCK)
975	*
976	* but is faster than a bunch of \|\|
977	*/
978	if (unlikely(return_code <= -ERESTARTSYS) &&
979	(return_code >= -ERESTART_RESTARTBLOCK) &&
980	(return_code != -ENOIOCTLCMD))
981	context->return_code = -EINTR;
982	else
983	context->return_code = return_code;
984
985	if (context->in_syscall && !context->dummy) {
986	audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
987	audit_filter_inodes(tsk, context);
988	}
989
990	tsk->audit_context = NULL;
991	return context;
992	}
993
994	static inline void audit_free_names(struct audit_context *context)
995	{
996	struct audit_names n, next;
997
998	#if AUDIT_DEBUG == 2
999	if (context->put_count + context->ino_count != context->name_count) {
1000	printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
1001	" name_count=%d put_count=%d"
1002	" ino_count=%d [NOT freeing]\n",
1003	__FILE__, __LINE__,
1004	context->serial, context->major, context->in_syscall,
1005	context->name_count, context->put_count,
1006	context->ino_count);
1007	list_for_each_entry(n, &context->names_list, list) {
1008	printk(KERN_ERR "names[%d] = %p = %s\n", i,
1009	n->name, n->name ?: "(null)");
1010	}
1011	dump_stack();
1012	return;
1013	}
1014	#endif
1015	#if AUDIT_DEBUG
1016	context->put_count = 0;
1017	context->ino_count = 0;
1018	#endif
1019
1020	list_for_each_entry_safe(n, next, &context->names_list, list) {
1021	list_del(&n->list);
1022	if (n->name && n->name_put)
1023	__putname(n->name);
1024	if (n->should_free)
1025	kfree(n);
1026	}
1027	context->name_count = 0;
1028	path_put(&context->pwd);
1029	context->pwd.dentry = NULL;
1030	context->pwd.mnt = NULL;
1031	}
1032
1033	static inline void audit_free_aux(struct audit_context *context)
1034	{
1035	struct audit_aux_data *aux;
1036
1037	while ((aux = context->aux)) {
1038	context->aux = aux->next;
1039	kfree(aux);
1040	}
1041	while ((aux = context->aux_pids)) {
1042	context->aux_pids = aux->next;
1043	kfree(aux);
1044	}
1045	}
1046
1047	static inline void audit_zero_context(struct audit_context *context,
1048	enum audit_state state)
1049	{
1050	memset(context, 0, sizeof(*context));
1051	context->state = state;
1052	context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1053	}
1054
1055	static inline struct audit_context *audit_alloc_context(enum audit_state state)
1056	{
1057	struct audit_context *context;
1058
1059	if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
1060	return NULL;
1061	audit_zero_context(context, state);
1062	INIT_LIST_HEAD(&context->killed_trees);
1063	INIT_LIST_HEAD(&context->names_list);
1064	return context;
1065	}
1066
1067	/**
1068	* audit_alloc - allocate an audit context block for a task
1069	* @tsk: task
1070	*
1071	* Filter on the task information and allocate a per-task audit context
1072	* if necessary. Doing so turns on system call auditing for the
1073	* specified task. This is called from copy_process, so no lock is
1074	* needed.
1075	*/
1076	int audit_alloc(struct task_struct *tsk)
1077	{
1078	struct audit_context *context;
1079	enum audit_state state;
1080	char *key = NULL;
1081
1082	if (likely(!audit_ever_enabled))
1083	return 0; /* Return if not auditing. */
1084
1085	state = audit_filter_task(tsk, &key);
1086	if (state == AUDIT_DISABLED)
1087	return 0;
1088
1089	if (!(context = audit_alloc_context(state))) {
1090	kfree(key);
1091	audit_log_lost("out of memory in audit_alloc");
1092	return -ENOMEM;
1093	}
1094	context->filterkey = key;
1095
1096	tsk->audit_context = context;
1097	set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
1098	return 0;
1099	}
1100
1101	static inline void audit_free_context(struct audit_context *context)
1102	{
1103	struct audit_context *previous;
1104	int count = 0;
1105
1106	do {
1107	previous = context->previous;
1108	if (previous \|\| (count && count < 10)) {
1109	++count;
1110	printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
1111	" freeing multiple contexts (%d)\n",
1112	context->serial, context->major,
1113	context->name_count, count);
1114	}
1115	audit_free_names(context);
1116	unroll_tree_refs(context, NULL, 0);
1117	free_tree_refs(context);
1118	audit_free_aux(context);
1119	kfree(context->filterkey);
1120	kfree(context->sockaddr);
1121	kfree(context);
1122	context = previous;
1123	} while (context);
1124	if (count >= 10)
1125	printk(KERN_ERR "audit: freed %d contexts\n", count);
1126	}
1127
1128	void audit_log_task_context(struct audit_buffer *ab)
1129	{
1130	char *ctx = NULL;
1131	unsigned len;
1132	int error;
1133	u32 sid;
1134
1135	security_task_getsecid(current, &sid);
1136	if (!sid)
1137	return;
1138
1139	error = security_secid_to_secctx(sid, &ctx, &len);
1140	if (error) {
1141	if (error != -EINVAL)
1142	goto error_path;
1143	return;
1144	}
1145
1146	audit_log_format(ab, " subj=%s", ctx);
1147	security_release_secctx(ctx, len);
1148	return;
1149
1150	error_path:
1151	audit_panic("error in audit_log_task_context");
1152	return;
1153	}
1154
1155	EXPORT_SYMBOL(audit_log_task_context);
1156
1157	static void audit_log_task_info(struct audit_buffer ab, struct task_struct tsk)
1158	{
1159	char name[sizeof(tsk->comm)];
1160	struct mm_struct *mm = tsk->mm;
1161	struct vm_area_struct *vma;
1162
1163	/* tsk == current */
1164
1165	get_task_comm(name, tsk);
1166	audit_log_format(ab, " comm=");
1167	audit_log_untrustedstring(ab, name);
1168
1169	if (mm) {
1170	down_read(&mm->mmap_sem);
1171	vma = mm->mmap;
1172	while (vma) {
1173	if ((vma->vm_flags & VM_EXECUTABLE) &&
1174	vma->vm_file) {
1175	audit_log_d_path(ab, " exe=",
1176	&vma->vm_file->f_path);
1177	break;
1178	}
1179	vma = vma->vm_next;
1180	}
1181	up_read(&mm->mmap_sem);
1182	}
1183	audit_log_task_context(ab);
1184	}
1185
1186	static int audit_log_pid_context(struct audit_context *context, pid_t pid,
1187	uid_t auid, uid_t uid, unsigned int sessionid,
1188	u32 sid, char *comm)
1189	{
1190	struct audit_buffer *ab;
1191	char *ctx = NULL;
1192	u32 len;
1193	int rc = 0;
1194
1195	ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1196	if (!ab)
1197	return rc;
1198
1199	audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
1200	uid, sessionid);
1201	if (security_secid_to_secctx(sid, &ctx, &len)) {
1202	audit_log_format(ab, " obj=(none)");
1203	rc = 1;
1204	} else {
1205	audit_log_format(ab, " obj=%s", ctx);
1206	security_release_secctx(ctx, len);
1207	}
1208	audit_log_format(ab, " ocomm=");
1209	audit_log_untrustedstring(ab, comm);
1210	audit_log_end(ab);
1211
1212	return rc;
1213	}
1214
1215	/*
1216	* to_send and len_sent accounting are very loose estimates. We aren't
1217	* really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1218	* within about 500 bytes (next page boundary)
1219	*
1220	* why snprintf? an int is up to 12 digits long. if we just assumed when
1221	* logging that a[%d]= was going to be 16 characters long we would be wasting
1222	* space in every audit message. In one 7500 byte message we can log up to
1223	* about 1000 min size arguments. That comes down to about 50% waste of space
1224	* if we didn't do the snprintf to find out how long arg_num_len was.
1225	*/
1226	static int audit_log_single_execve_arg(struct audit_context *context,
1227	struct audit_buffer **ab,
1228	int arg_num,
1229	size_t *len_sent,
1230	const char __user *p,
1231	char *buf)
1232	{
1233	char arg_num_len_buf[12];
1234	const char __user *tmp_p = p;
1235	/* how many digits are in arg_num? 5 is the length of ' a=""' */
1236	size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1237	size_t len, len_left, to_send;
1238	size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1239	unsigned int i, has_cntl = 0, too_long = 0;
1240	int ret;
1241
1242	/* strnlen_user includes the null we don't want to send */
1243	len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1244
1245	/*
1246	* We just created this mm, if we can't find the strings
1247	* we just copied into it something is _very_ wrong. Similar
1248	* for strings that are too long, we should not have created
1249	* any.
1250	*/
1251	if (unlikely((len == -1) \|\| len > MAX_ARG_STRLEN - 1)) {
1252	WARN_ON(1);
1253	send_sig(SIGKILL, current, 0);
1254	return -1;
1255	}
1256
1257	/* walk the whole argument looking for non-ascii chars */
1258	do {
1259	if (len_left > MAX_EXECVE_AUDIT_LEN)
1260	to_send = MAX_EXECVE_AUDIT_LEN;
1261	else
1262	to_send = len_left;
1263	ret = copy_from_user(buf, tmp_p, to_send);
1264	/*
1265	* There is no reason for this copy to be short. We just
1266	* copied them here, and the mm hasn't been exposed to user-
1267	* space yet.
1268	*/
1269	if (ret) {
1270	WARN_ON(1);
1271	send_sig(SIGKILL, current, 0);
1272	return -1;
1273	}
1274	buf[to_send] = '\0';
1275	has_cntl = audit_string_contains_control(buf, to_send);
1276	if (has_cntl) {
1277	/*
1278	* hex messages get logged as 2 bytes, so we can only
1279	* send half as much in each message
1280	*/
1281	max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1282	break;
1283	}
1284	len_left -= to_send;
1285	tmp_p += to_send;
1286	} while (len_left > 0);
1287
1288	len_left = len;
1289
1290	if (len > max_execve_audit_len)
1291	too_long = 1;
1292
1293	/* rewalk the argument actually logging the message */
1294	for (i = 0; len_left > 0; i++) {
1295	int room_left;
1296
1297	if (len_left > max_execve_audit_len)
1298	to_send = max_execve_audit_len;
1299	else
1300	to_send = len_left;
1301
1302	/* do we have space left to send this argument in this ab? */
1303	room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1304	if (has_cntl)
1305	room_left -= (to_send * 2);
1306	else
1307	room_left -= to_send;
1308	if (room_left < 0) {
1309	*len_sent = 0;
1310	audit_log_end(*ab);
1311	*ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1312	if (!*ab)
1313	return 0;
1314	}
1315
1316	/*
1317	* first record needs to say how long the original string was
1318	* so we can be sure nothing was lost.
1319	*/
1320	if ((i == 0) && (too_long))
1321	audit_log_format(*ab, " a%d_len=%zu", arg_num,
1322	has_cntl ? 2*len : len);
1323
1324	/*
1325	* normally arguments are small enough to fit and we already
1326	* filled buf above when we checked for control characters
1327	* so don't bother with another copy_from_user
1328	*/
1329	if (len >= max_execve_audit_len)
1330	ret = copy_from_user(buf, p, to_send);
1331	else
1332	ret = 0;
1333	if (ret) {
1334	WARN_ON(1);
1335	send_sig(SIGKILL, current, 0);
1336	return -1;
1337	}
1338	buf[to_send] = '\0';
1339
1340	/* actually log it */
1341	audit_log_format(*ab, " a%d", arg_num);
1342	if (too_long)
1343	audit_log_format(*ab, "[%d]", i);
1344	audit_log_format(*ab, "=");
1345	if (has_cntl)
1346	audit_log_n_hex(*ab, buf, to_send);
1347	else
1348	audit_log_string(*ab, buf);
1349
1350	p += to_send;
1351	len_left -= to_send;
1352	*len_sent += arg_num_len;
1353	if (has_cntl)
1354	len_sent += to_send 2;
1355	else
1356	*len_sent += to_send;
1357	}
1358	/* include the null we didn't log */
1359	return len + 1;
1360	}
1361
1362	static void audit_log_execve_info(struct audit_context *context,
1363	struct audit_buffer **ab,
1364	struct audit_aux_data_execve *axi)
1365	{
1366	int i, len;
1367	size_t len_sent = 0;
1368	const char __user *p;
1369	char *buf;
1370
1371	if (axi->mm != current->mm)
1372	return; /* execve failed, no additional info */
1373
1374	p = (const char __user *)axi->mm->arg_start;
1375
1376	audit_log_format(*ab, "argc=%d", axi->argc);
1377
1378	/*
1379	* we need some kernel buffer to hold the userspace args. Just
1380	* allocate one big one rather than allocating one of the right size
1381	* for every single argument inside audit_log_single_execve_arg()
1382	* should be <8k allocation so should be pretty safe.
1383	*/
1384	buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1385	if (!buf) {
1386	audit_panic("out of memory for argv string\n");
1387	return;
1388	}
1389
1390	for (i = 0; i < axi->argc; i++) {
1391	len = audit_log_single_execve_arg(context, ab, i,
1392	&len_sent, p, buf);
1393	if (len <= 0)
1394	break;
1395	p += len;
1396	}
1397	kfree(buf);
1398	}
1399
1400	static void audit_log_cap(struct audit_buffer ab, char prefix, kernel_cap_t *cap)
1401	{
1402	int i;
1403
1404	audit_log_format(ab, " %s=", prefix);
1405	CAP_FOR_EACH_U32(i) {
1406	audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1407	}
1408	}
1409
1410	static void audit_log_fcaps(struct audit_buffer ab, struct audit_names name)
1411	{
1412	kernel_cap_t *perm = &name->fcap.permitted;
1413	kernel_cap_t *inh = &name->fcap.inheritable;
1414	int log = 0;
1415
1416	if (!cap_isclear(*perm)) {
1417	audit_log_cap(ab, "cap_fp", perm);
1418	log = 1;
1419	}
1420	if (!cap_isclear(*inh)) {
1421	audit_log_cap(ab, "cap_fi", inh);
1422	log = 1;
1423	}
1424
1425	if (log)
1426	audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1427	}
1428
1429	static void show_special(struct audit_context context, int call_panic)
1430	{
1431	struct audit_buffer *ab;
1432	int i;
1433
1434	ab = audit_log_start(context, GFP_KERNEL, context->type);
1435	if (!ab)
1436	return;
1437
1438	switch (context->type) {
1439	case AUDIT_SOCKETCALL: {
1440	int nargs = context->socketcall.nargs;
1441	audit_log_format(ab, "nargs=%d", nargs);
1442	for (i = 0; i < nargs; i++)
1443	audit_log_format(ab, " a%d=%lx", i,
1444	context->socketcall.args[i]);
1445	break; }
1446	case AUDIT_IPC: {
1447	u32 osid = context->ipc.osid;
1448
1449	audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1450	context->ipc.uid, context->ipc.gid, context->ipc.mode);
1451	if (osid) {
1452	char *ctx = NULL;
1453	u32 len;
1454	if (security_secid_to_secctx(osid, &ctx, &len)) {
1455	audit_log_format(ab, " osid=%u", osid);
1456	*call_panic = 1;
1457	} else {
1458	audit_log_format(ab, " obj=%s", ctx);
1459	security_release_secctx(ctx, len);
1460	}
1461	}
1462	if (context->ipc.has_perm) {
1463	audit_log_end(ab);
1464	ab = audit_log_start(context, GFP_KERNEL,
1465	AUDIT_IPC_SET_PERM);
1466	audit_log_format(ab,
1467	"qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1468	context->ipc.qbytes,
1469	context->ipc.perm_uid,
1470	context->ipc.perm_gid,
1471	context->ipc.perm_mode);
1472	if (!ab)
1473	return;
1474	}
1475	break; }
1476	case AUDIT_MQ_OPEN: {
1477	audit_log_format(ab,
1478	"oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1479	"mq_msgsize=%ld mq_curmsgs=%ld",
1480	context->mq_open.oflag, context->mq_open.mode,
1481	context->mq_open.attr.mq_flags,
1482	context->mq_open.attr.mq_maxmsg,
1483	context->mq_open.attr.mq_msgsize,
1484	context->mq_open.attr.mq_curmsgs);
1485	break; }
1486	case AUDIT_MQ_SENDRECV: {
1487	audit_log_format(ab,
1488	"mqdes=%d msg_len=%zd msg_prio=%u "
1489	"abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1490	context->mq_sendrecv.mqdes,
1491	context->mq_sendrecv.msg_len,
1492	context->mq_sendrecv.msg_prio,
1493	context->mq_sendrecv.abs_timeout.tv_sec,
1494	context->mq_sendrecv.abs_timeout.tv_nsec);
1495	break; }
1496	case AUDIT_MQ_NOTIFY: {
1497	audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1498	context->mq_notify.mqdes,
1499	context->mq_notify.sigev_signo);
1500	break; }
1501	case AUDIT_MQ_GETSETATTR: {
1502	struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1503	audit_log_format(ab,
1504	"mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1505	"mq_curmsgs=%ld ",
1506	context->mq_getsetattr.mqdes,
1507	attr->mq_flags, attr->mq_maxmsg,
1508	attr->mq_msgsize, attr->mq_curmsgs);
1509	break; }
1510	case AUDIT_CAPSET: {
1511	audit_log_format(ab, "pid=%d", context->capset.pid);
1512	audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1513	audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1514	audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1515	break; }
1516	case AUDIT_MMAP: {
1517	audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1518	context->mmap.flags);
1519	break; }
1520	}
1521	audit_log_end(ab);
1522	}
1523
1524	static void audit_log_name(struct audit_context context, struct audit_names n,
1525	int record_num, int *call_panic)
1526	{
1527	struct audit_buffer *ab;
1528	ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1529	if (!ab)
1530	return; /* audit_panic has been called */
1531
1532	audit_log_format(ab, "item=%d", record_num);
1533
1534	if (n->name) {
1535	switch (n->name_len) {
1536	case AUDIT_NAME_FULL:
1537	/* log the full path */
1538	audit_log_format(ab, " name=");
1539	audit_log_untrustedstring(ab, n->name);
1540	break;
1541	case 0:
1542	/* name was specified as a relative path and the
1543	* directory component is the cwd */
1544	audit_log_d_path(ab, " name=", &context->pwd);
1545	break;
1546	default:
1547	/* log the name's directory component */
1548	audit_log_format(ab, " name=");
1549	audit_log_n_untrustedstring(ab, n->name,
1550	n->name_len);
1551	}
1552	} else
1553	audit_log_format(ab, " name=(null)");
1554
1555	if (n->ino != (unsigned long)-1) {
1556	audit_log_format(ab, " inode=%lu"
1557	" dev=%02x:%02x mode=%#ho"
1558	" ouid=%u ogid=%u rdev=%02x:%02x",
1559	n->ino,
1560	MAJOR(n->dev),
1561	MINOR(n->dev),
1562	n->mode,
1563	n->uid,
1564	n->gid,
1565	MAJOR(n->rdev),
1566	MINOR(n->rdev));
1567	}
1568	if (n->osid != 0) {
1569	char *ctx = NULL;
1570	u32 len;
1571	if (security_secid_to_secctx(
1572	n->osid, &ctx, &len)) {
1573	audit_log_format(ab, " osid=%u", n->osid);
1574	*call_panic = 2;
1575	} else {
1576	audit_log_format(ab, " obj=%s", ctx);
1577	security_release_secctx(ctx, len);
1578	}
1579	}
1580
1581	audit_log_fcaps(ab, n);
1582
1583	audit_log_end(ab);
1584	}
1585
1586	static void audit_log_exit(struct audit_context context, struct task_struct tsk)
1587	{
1588	const struct cred *cred;
1589	int i, call_panic = 0;
1590	struct audit_buffer *ab;
1591	struct audit_aux_data *aux;
1592	const char *tty;
1593	struct audit_names *n;
1594
1595	/* tsk == current */
1596	context->pid = tsk->pid;
1597	if (!context->ppid)
1598	context->ppid = sys_getppid();
1599	cred = current_cred();
1600	context->uid = cred->uid;
1601	context->gid = cred->gid;
1602	context->euid = cred->euid;
1603	context->suid = cred->suid;
1604	context->fsuid = cred->fsuid;
1605	context->egid = cred->egid;
1606	context->sgid = cred->sgid;
1607	context->fsgid = cred->fsgid;
1608	context->personality = tsk->personality;
1609
1610	ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1611	if (!ab)
1612	return; /* audit_panic has been called */
1613	audit_log_format(ab, "arch=%x syscall=%d",
1614	context->arch, context->major);
1615	if (context->personality != PER_LINUX)
1616	audit_log_format(ab, " per=%lx", context->personality);
1617	if (context->return_valid)
1618	audit_log_format(ab, " success=%s exit=%ld",
1619	(context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1620	context->return_code);
1621
1622	spin_lock_irq(&tsk->sighand->siglock);
1623	if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1624	tty = tsk->signal->tty->name;
1625	else
1626	tty = "(none)";
1627	spin_unlock_irq(&tsk->sighand->siglock);
1628
1629	audit_log_format(ab,
1630	" a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1631	" ppid=%d pid=%d auid=%u uid=%u gid=%u"
1632	" euid=%u suid=%u fsuid=%u"
1633	" egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1634	context->argv[0],
1635	context->argv[1],
1636	context->argv[2],
1637	context->argv[3],
1638	context->name_count,
1639	context->ppid,
1640	context->pid,
1641	tsk->loginuid,
1642	context->uid,
1643	context->gid,
1644	context->euid, context->suid, context->fsuid,
1645	context->egid, context->sgid, context->fsgid, tty,
1646	tsk->sessionid);
1647
1648
1649	audit_log_task_info(ab, tsk);
1650	audit_log_key(ab, context->filterkey);
1651	audit_log_end(ab);
1652
1653	for (aux = context->aux; aux; aux = aux->next) {
1654
1655	ab = audit_log_start(context, GFP_KERNEL, aux->type);
1656	if (!ab)
1657	continue; /* audit_panic has been called */
1658
1659	switch (aux->type) {
1660
1661	case AUDIT_EXECVE: {
1662	struct audit_aux_data_execve axi = (void )aux;
1663	audit_log_execve_info(context, &ab, axi);
1664	break; }
1665
1666	case AUDIT_BPRM_FCAPS: {
1667	struct audit_aux_data_bprm_fcaps axs = (void )aux;
1668	audit_log_format(ab, "fver=%x", axs->fcap_ver);
1669	audit_log_cap(ab, "fp", &axs->fcap.permitted);
1670	audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1671	audit_log_format(ab, " fe=%d", axs->fcap.fE);
1672	audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1673	audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1674	audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1675	audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1676	audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1677	audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1678	break; }
1679
1680	}
1681	audit_log_end(ab);
1682	}
1683
1684	if (context->type)
1685	show_special(context, &call_panic);
1686
1687	if (context->fds[0] >= 0) {
1688	ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1689	if (ab) {
1690	audit_log_format(ab, "fd0=%d fd1=%d",
1691	context->fds[0], context->fds[1]);
1692	audit_log_end(ab);
1693	}
1694	}
1695
1696	if (context->sockaddr_len) {
1697	ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1698	if (ab) {
1699	audit_log_format(ab, "saddr=");
1700	audit_log_n_hex(ab, (void *)context->sockaddr,
1701	context->sockaddr_len);
1702	audit_log_end(ab);
1703	}
1704	}
1705
1706	for (aux = context->aux_pids; aux; aux = aux->next) {
1707	struct audit_aux_data_pids axs = (void )aux;
1708
1709	for (i = 0; i < axs->pid_count; i++)
1710	if (audit_log_pid_context(context, axs->target_pid[i],
1711	axs->target_auid[i],
1712	axs->target_uid[i],
1713	axs->target_sessionid[i],
1714	axs->target_sid[i],
1715	axs->target_comm[i]))
1716	call_panic = 1;
1717	}
1718
1719	if (context->target_pid &&
1720	audit_log_pid_context(context, context->target_pid,
1721	context->target_auid, context->target_uid,
1722	context->target_sessionid,
1723	context->target_sid, context->target_comm))
1724	call_panic = 1;
1725
1726	if (context->pwd.dentry && context->pwd.mnt) {
1727	ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1728	if (ab) {
1729	audit_log_d_path(ab, " cwd=", &context->pwd);
1730	audit_log_end(ab);
1731	}
1732	}
1733
1734	i = 0;
1735	list_for_each_entry(n, &context->names_list, list)
1736	audit_log_name(context, n, i++, &call_panic);
1737
1738	/* Send end of event record to help user space know we are finished */
1739	ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1740	if (ab)
1741	audit_log_end(ab);
1742	if (call_panic)
1743	audit_panic("error converting sid to string");
1744	}
1745
1746	/**
1747	* audit_free - free a per-task audit context
1748	* @tsk: task whose audit context block to free
1749	*
1750	* Called from copy_process and do_exit
1751	*/
1752	void __audit_free(struct task_struct *tsk)
1753	{
1754	struct audit_context *context;
1755
1756	context = audit_get_context(tsk, 0, 0);
1757	if (!context)
1758	return;
1759
1760	/* Check for system calls that do not go through the exit
1761	* function (e.g., exit_group), then free context block.
1762	* We use GFP_ATOMIC here because we might be doing this
1763	* in the context of the idle thread */
1764	/* that can happen only if we are called from do_exit() */
1765	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1766	audit_log_exit(context, tsk);
1767	if (!list_empty(&context->killed_trees))
1768	audit_kill_trees(&context->killed_trees);
1769
1770	audit_free_context(context);
1771	}
1772
1773	/**
1774	* audit_syscall_entry - fill in an audit record at syscall entry
1775	* @arch: architecture type
1776	* @major: major syscall type (function)
1777	* @a1: additional syscall register 1
1778	* @a2: additional syscall register 2
1779	* @a3: additional syscall register 3
1780	* @a4: additional syscall register 4
1781	*
1782	* Fill in audit context at syscall entry. This only happens if the
1783	* audit context was created when the task was created and the state or
1784	* filters demand the audit context be built. If the state from the
1785	* per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1786	* then the record will be written at syscall exit time (otherwise, it
1787	* will only be written if another part of the kernel requests that it
1788	* be written).
1789	*/
1790	void __audit_syscall_entry(int arch, int major,
1791	unsigned long a1, unsigned long a2,
1792	unsigned long a3, unsigned long a4)
1793	{
1794	struct task_struct *tsk = current;
1795	struct audit_context *context = tsk->audit_context;
1796	enum audit_state state;
1797
1798	if (!context)
1799	return;
1800
1801	/*
1802	* This happens only on certain architectures that make system
1803	* calls in kernel_thread via the entry.S interface, instead of
1804	* with direct calls. (If you are porting to a new
1805	* architecture, hitting this condition can indicate that you
1806	* got the _exit/_leave calls backward in entry.S.)
1807	*
1808	* i386 no
1809	* x86_64 no
1810	* ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1811	*
1812	* This also happens with vm86 emulation in a non-nested manner
1813	* (entries without exits), so this case must be caught.
1814	*/
1815	if (context->in_syscall) {
1816	struct audit_context *newctx;
1817
1818	#if AUDIT_DEBUG
1819	printk(KERN_ERR
1820	"audit(:%d) pid=%d in syscall=%d;"
1821	" entering syscall=%d\n",
1822	context->serial, tsk->pid, context->major, major);
1823	#endif
1824	newctx = audit_alloc_context(context->state);
1825	if (newctx) {
1826	newctx->previous = context;
1827	context = newctx;
1828	tsk->audit_context = newctx;
1829	} else {
1830	/* If we can't alloc a new context, the best we
1831	* can do is to leak memory (any pending putname
1832	* will be lost). The only other alternative is
1833	* to abandon auditing. */
1834	audit_zero_context(context, context->state);
1835	}
1836	}
1837	BUG_ON(context->in_syscall \|\| context->name_count);
1838
1839	if (!audit_enabled)
1840	return;
1841
1842	context->arch = arch;
1843	context->major = major;
1844	context->argv[0] = a1;
1845	context->argv[1] = a2;
1846	context->argv[2] = a3;
1847	context->argv[3] = a4;
1848
1849	state = context->state;
1850	context->dummy = !audit_n_rules;
1851	if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1852	context->prio = 0;
1853	state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1854	}
1855	if (state == AUDIT_DISABLED)
1856	return;
1857
1858	context->serial = 0;
1859	context->ctime = CURRENT_TIME;
1860	context->in_syscall = 1;
1861	context->current_state = state;
1862	context->ppid = 0;
1863	}
1864
1865	/**
1866	* audit_syscall_exit - deallocate audit context after a system call
1867	* @success: success value of the syscall
1868	* @return_code: return value of the syscall
1869	*
1870	* Tear down after system call. If the audit context has been marked as
1871	* auditable (either because of the AUDIT_RECORD_CONTEXT state from
1872	* filtering, or because some other part of the kernel wrote an audit
1873	* message), then write out the syscall information. In call cases,
1874	* free the names stored from getname().
1875	*/
1876	void __audit_syscall_exit(int success, long return_code)
1877	{
1878	struct task_struct *tsk = current;
1879	struct audit_context *context;
1880
1881	if (success)
1882	success = AUDITSC_SUCCESS;
1883	else
1884	success = AUDITSC_FAILURE;
1885
1886	context = audit_get_context(tsk, success, return_code);
1887	if (!context)
1888	return;
1889
1890	if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1891	audit_log_exit(context, tsk);
1892
1893	context->in_syscall = 0;
1894	context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1895
1896	if (!list_empty(&context->killed_trees))
1897	audit_kill_trees(&context->killed_trees);
1898
1899	if (context->previous) {
1900	struct audit_context *new_context = context->previous;
1901	context->previous = NULL;
1902	audit_free_context(context);
1903	tsk->audit_context = new_context;
1904	} else {
1905	audit_free_names(context);
1906	unroll_tree_refs(context, NULL, 0);
1907	audit_free_aux(context);
1908	context->aux = NULL;
1909	context->aux_pids = NULL;
1910	context->target_pid = 0;
1911	context->target_sid = 0;
1912	context->sockaddr_len = 0;
1913	context->type = 0;
1914	context->fds[0] = -1;
1915	if (context->state != AUDIT_RECORD_CONTEXT) {
1916	kfree(context->filterkey);
1917	context->filterkey = NULL;
1918	}
1919	tsk->audit_context = context;
1920	}
1921	}
1922
1923	static inline void handle_one(const struct inode *inode)
1924	{
1925	#ifdef CONFIG_AUDIT_TREE
1926	struct audit_context *context;
1927	struct audit_tree_refs *p;
1928	struct audit_chunk *chunk;
1929	int count;
1930	if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1931	return;
1932	context = current->audit_context;
1933	p = context->trees;
1934	count = context->tree_count;
1935	rcu_read_lock();
1936	chunk = audit_tree_lookup(inode);
1937	rcu_read_unlock();
1938	if (!chunk)
1939	return;
1940	if (likely(put_tree_ref(context, chunk)))
1941	return;
1942	if (unlikely(!grow_tree_refs(context))) {
1943	printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1944	audit_set_auditable(context);
1945	audit_put_chunk(chunk);
1946	unroll_tree_refs(context, p, count);
1947	return;
1948	}
1949	put_tree_ref(context, chunk);
1950	#endif
1951	}
1952
1953	static void handle_path(const struct dentry *dentry)
1954	{
1955	#ifdef CONFIG_AUDIT_TREE
1956	struct audit_context *context;
1957	struct audit_tree_refs *p;
1958	const struct dentry d, parent;
1959	struct audit_chunk *drop;
1960	unsigned long seq;
1961	int count;
1962
1963	context = current->audit_context;
1964	p = context->trees;
1965	count = context->tree_count;
1966	retry:
1967	drop = NULL;
1968	d = dentry;
1969	rcu_read_lock();
1970	seq = read_seqbegin(&rename_lock);
1971	for(;;) {
1972	struct inode *inode = d->d_inode;
1973	if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1974	struct audit_chunk *chunk;
1975	chunk = audit_tree_lookup(inode);
1976	if (chunk) {
1977	if (unlikely(!put_tree_ref(context, chunk))) {
1978	drop = chunk;
1979	break;
1980	}
1981	}
1982	}
1983	parent = d->d_parent;
1984	if (parent == d)
1985	break;
1986	d = parent;
1987	}
1988	if (unlikely(read_seqretry(&rename_lock, seq) \|\| drop)) { /* in this order */
1989	rcu_read_unlock();
1990	if (!drop) {
1991	/* just a race with rename */
1992	unroll_tree_refs(context, p, count);
1993	goto retry;
1994	}
1995	audit_put_chunk(drop);
1996	if (grow_tree_refs(context)) {
1997	/* OK, got more space */
1998	unroll_tree_refs(context, p, count);
1999	goto retry;
2000	}
2001	/* too bad */
2002	printk(KERN_WARNING
2003	"out of memory, audit has lost a tree reference\n");
2004	unroll_tree_refs(context, p, count);
2005	audit_set_auditable(context);
2006	return;
2007	}
2008	rcu_read_unlock();
2009	#endif
2010	}
2011
2012	static struct audit_names audit_alloc_name(struct audit_context context)
2013	{
2014	struct audit_names *aname;
2015
2016	if (context->name_count < AUDIT_NAMES) {
2017	aname = &context->preallocated_names[context->name_count];
2018	memset(aname, 0, sizeof(*aname));
2019	} else {
2020	aname = kzalloc(sizeof(*aname), GFP_NOFS);
2021	if (!aname)
2022	return NULL;
2023	aname->should_free = true;
2024	}
2025
2026	aname->ino = (unsigned long)-1;
2027	list_add_tail(&aname->list, &context->names_list);
2028
2029	context->name_count++;
2030	#if AUDIT_DEBUG
2031	context->ino_count++;
2032	#endif
2033	return aname;
2034	}
2035
2036	/**
2037	* audit_getname - add a name to the list
2038	* @name: name to add
2039	*
2040	* Add a name to the list of audit names for this context.
2041	* Called from fs/namei.c:getname().
2042	*/
2043	void __audit_getname(const char *name)
2044	{
2045	struct audit_context *context = current->audit_context;
2046	struct audit_names *n;
2047
2048	if (!context->in_syscall) {
2049	#if AUDIT_DEBUG == 2
2050	printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
2051	__FILE__, __LINE__, context->serial, name);
2052	dump_stack();
2053	#endif
2054	return;
2055	}
2056
2057	n = audit_alloc_name(context);
2058	if (!n)
2059	return;
2060
2061	n->name = name;
2062	n->name_len = AUDIT_NAME_FULL;
2063	n->name_put = true;
2064
2065	if (!context->pwd.dentry)
2066	get_fs_pwd(current->fs, &context->pwd);
2067	}
2068
2069	/* audit_putname - intercept a putname request
2070	* @name: name to intercept and delay for putname
2071	*
2072	* If we have stored the name from getname in the audit context,
2073	* then we delay the putname until syscall exit.
2074	* Called from include/linux/fs.h:putname().
2075	*/
2076	void audit_putname(const char *name)
2077	{
2078	struct audit_context *context = current->audit_context;
2079
2080	BUG_ON(!context);
2081	if (!context->in_syscall) {
2082	#if AUDIT_DEBUG == 2
2083	printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
2084	__FILE__, __LINE__, context->serial, name);
2085	if (context->name_count) {
2086	struct audit_names *n;
2087	int i;
2088
2089	list_for_each_entry(n, &context->names_list, list)
2090	printk(KERN_ERR "name[%d] = %p = %s\n", i,
2091	n->name, n->name ?: "(null)");
2092	}
2093	#endif
2094	__putname(name);
2095	}
2096	#if AUDIT_DEBUG
2097	else {
2098	++context->put_count;
2099	if (context->put_count > context->name_count) {
2100	printk(KERN_ERR "%s:%d(:%d): major=%d"
2101	" in_syscall=%d putname(%p) name_count=%d"
2102	" put_count=%d\n",
2103	__FILE__, __LINE__,
2104	context->serial, context->major,
2105	context->in_syscall, name, context->name_count,
2106	context->put_count);
2107	dump_stack();
2108	}
2109	}
2110	#endif
2111	}
2112
2113	static inline int audit_copy_fcaps(struct audit_names name, const struct dentry dentry)
2114	{
2115	struct cpu_vfs_cap_data caps;
2116	int rc;
2117
2118	if (!dentry)
2119	return 0;
2120
2121	rc = get_vfs_caps_from_disk(dentry, &caps);
2122	if (rc)
2123	return rc;
2124
2125	name->fcap.permitted = caps.permitted;
2126	name->fcap.inheritable = caps.inheritable;
2127	name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2128	name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2129
2130	return 0;
2131	}
2132
2133
2134	/* Copy inode data into an audit_names. */
2135	static void audit_copy_inode(struct audit_names name, const struct dentry dentry,
2136	const struct inode *inode)
2137	{
2138	name->ino = inode->i_ino;
2139	name->dev = inode->i_sb->s_dev;
2140	name->mode = inode->i_mode;
2141	name->uid = inode->i_uid;
2142	name->gid = inode->i_gid;
2143	name->rdev = inode->i_rdev;
2144	security_inode_getsecid(inode, &name->osid);
2145	audit_copy_fcaps(name, dentry);
2146	}
2147
2148	/**
2149	* audit_inode - store the inode and device from a lookup
2150	* @name: name being audited
2151	* @dentry: dentry being audited
2152	*
2153	* Called from fs/namei.c:path_lookup().
2154	*/
2155	void __audit_inode(const char name, const struct dentry dentry)
2156	{
2157	struct audit_context *context = current->audit_context;
2158	const struct inode *inode = dentry->d_inode;
2159	struct audit_names *n;
2160
2161	if (!context->in_syscall)
2162	return;
2163
2164	list_for_each_entry_reverse(n, &context->names_list, list) {
2165	if (n->name && (n->name == name))
2166	goto out;
2167	}
2168
2169	/* unable to find the name from a previous getname() */
2170	n = audit_alloc_name(context);
2171	if (!n)
2172	return;
2173	out:
2174	handle_path(dentry);
2175	audit_copy_inode(n, dentry, inode);
2176	}
2177
2178	/**
2179	* audit_inode_child - collect inode info for created/removed objects
2180	* @dentry: dentry being audited
2181	* @parent: inode of dentry parent
2182	*
2183	* For syscalls that create or remove filesystem objects, audit_inode
2184	* can only collect information for the filesystem object's parent.
2185	* This call updates the audit context with the child's information.
2186	* Syscalls that create a new filesystem object must be hooked after
2187	* the object is created. Syscalls that remove a filesystem object
2188	* must be hooked prior, in order to capture the target inode during
2189	* unsuccessful attempts.
2190	*/
2191	void __audit_inode_child(const struct dentry *dentry,
2192	const struct inode *parent)
2193	{
2194	struct audit_context *context = current->audit_context;
2195	const char found_parent = NULL, found_child = NULL;
2196	const struct inode *inode = dentry->d_inode;
2197	const char *dname = dentry->d_name.name;
2198	struct audit_names *n;
2199	int dirlen = 0;
2200
2201	if (!context->in_syscall)
2202	return;
2203
2204	if (inode)
2205	handle_one(inode);
2206
2207	/* parent is more likely, look for it first */
2208	list_for_each_entry(n, &context->names_list, list) {
2209	if (!n->name)
2210	continue;
2211
2212	if (n->ino == parent->i_ino &&
2213	!audit_compare_dname_path(dname, n->name, &dirlen)) {
2214	n->name_len = dirlen; /* update parent data in place */
2215	found_parent = n->name;
2216	goto add_names;
2217	}
2218	}
2219
2220	/* no matching parent, look for matching child */
2221	list_for_each_entry(n, &context->names_list, list) {
2222	if (!n->name)
2223	continue;
2224
2225	/* strcmp() is the more likely scenario */
2226	if (!strcmp(dname, n->name) \|\|
2227	!audit_compare_dname_path(dname, n->name, &dirlen)) {
2228	if (inode)
2229	audit_copy_inode(n, NULL, inode);
2230	else
2231	n->ino = (unsigned long)-1;
2232	found_child = n->name;
2233	goto add_names;
2234	}
2235	}
2236
2237	add_names:
2238	if (!found_parent) {
2239	n = audit_alloc_name(context);
2240	if (!n)
2241	return;
2242	audit_copy_inode(n, NULL, parent);
2243	}
2244
2245	if (!found_child) {
2246	n = audit_alloc_name(context);
2247	if (!n)
2248	return;
2249
2250	/* Re-use the name belonging to the slot for a matching parent
2251	* directory. All names for this context are relinquished in
2252	* audit_free_names() */
2253	if (found_parent) {
2254	n->name = found_parent;
2255	n->name_len = AUDIT_NAME_FULL;
2256	/* don't call __putname() */
2257	n->name_put = false;
2258	}
2259
2260	if (inode)
2261	audit_copy_inode(n, NULL, inode);
2262	}
2263	}
2264	EXPORT_SYMBOL_GPL(__audit_inode_child);
2265
2266	/**
2267	* auditsc_get_stamp - get local copies of audit_context values
2268	* @ctx: audit_context for the task
2269	* @t: timespec to store time recorded in the audit_context
2270	* @serial: serial value that is recorded in the audit_context
2271	*
2272	* Also sets the context as auditable.
2273	*/
2274	int auditsc_get_stamp(struct audit_context *ctx,
2275	struct timespec t, unsigned int serial)
2276	{
2277	if (!ctx->in_syscall)
2278	return 0;
2279	if (!ctx->serial)
2280	ctx->serial = audit_serial();
2281	t->tv_sec = ctx->ctime.tv_sec;
2282	t->tv_nsec = ctx->ctime.tv_nsec;
2283	*serial = ctx->serial;
2284	if (!ctx->prio) {
2285	ctx->prio = 1;
2286	ctx->current_state = AUDIT_RECORD_CONTEXT;
2287	}
2288	return 1;
2289	}
2290
2291	/* global counter which is incremented every time something logs in */
2292	static atomic_t session_id = ATOMIC_INIT(0);
2293
2294	/**
2295	* audit_set_loginuid - set current task's audit_context loginuid
2296	* @loginuid: loginuid value
2297	*
2298	* Returns 0.
2299	*
2300	* Called (set) from fs/proc/base.c::proc_loginuid_write().
2301	*/
2302	int audit_set_loginuid(uid_t loginuid)
2303	{
2304	struct task_struct *task = current;
2305	struct audit_context *context = task->audit_context;
2306	unsigned int sessionid;
2307
2308	#ifdef CONFIG_AUDIT_LOGINUID_IMMUTABLE
2309	if (task->loginuid != -1)
2310	return -EPERM;
2311	#else /* CONFIG_AUDIT_LOGINUID_IMMUTABLE */
2312	if (!capable(CAP_AUDIT_CONTROL))
2313	return -EPERM;
2314	#endif /* CONFIG_AUDIT_LOGINUID_IMMUTABLE */
2315
2316	sessionid = atomic_inc_return(&session_id);
2317	if (context && context->in_syscall) {
2318	struct audit_buffer *ab;
2319
2320	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2321	if (ab) {
2322	audit_log_format(ab, "login pid=%d uid=%u "
2323	"old auid=%u new auid=%u"
2324	" old ses=%u new ses=%u",
2325	task->pid, task_uid(task),
2326	task->loginuid, loginuid,
2327	task->sessionid, sessionid);
2328	audit_log_end(ab);
2329	}
2330	}
2331	task->sessionid = sessionid;
2332	task->loginuid = loginuid;
2333	return 0;
2334	}
2335
2336	/**
2337	* __audit_mq_open - record audit data for a POSIX MQ open
2338	* @oflag: open flag
2339	* @mode: mode bits
2340	* @attr: queue attributes
2341	*
2342	*/
2343	void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2344	{
2345	struct audit_context *context = current->audit_context;
2346
2347	if (attr)
2348	memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2349	else
2350	memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2351
2352	context->mq_open.oflag = oflag;
2353	context->mq_open.mode = mode;
2354
2355	context->type = AUDIT_MQ_OPEN;
2356	}
2357
2358	/**
2359	* __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2360	* @mqdes: MQ descriptor
2361	* @msg_len: Message length
2362	* @msg_prio: Message priority
2363	* @abs_timeout: Message timeout in absolute time
2364	*
2365	*/
2366	void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2367	const struct timespec *abs_timeout)
2368	{
2369	struct audit_context *context = current->audit_context;
2370	struct timespec *p = &context->mq_sendrecv.abs_timeout;
2371
2372	if (abs_timeout)
2373	memcpy(p, abs_timeout, sizeof(struct timespec));
2374	else
2375	memset(p, 0, sizeof(struct timespec));
2376
2377	context->mq_sendrecv.mqdes = mqdes;
2378	context->mq_sendrecv.msg_len = msg_len;
2379	context->mq_sendrecv.msg_prio = msg_prio;
2380
2381	context->type = AUDIT_MQ_SENDRECV;
2382	}
2383
2384	/**
2385	* __audit_mq_notify - record audit data for a POSIX MQ notify
2386	* @mqdes: MQ descriptor
2387	* @notification: Notification event
2388	*
2389	*/
2390
2391	void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2392	{
2393	struct audit_context *context = current->audit_context;
2394
2395	if (notification)
2396	context->mq_notify.sigev_signo = notification->sigev_signo;
2397	else
2398	context->mq_notify.sigev_signo = 0;
2399
2400	context->mq_notify.mqdes = mqdes;
2401	context->type = AUDIT_MQ_NOTIFY;
2402	}
2403
2404	/**
2405	* __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2406	* @mqdes: MQ descriptor
2407	* @mqstat: MQ flags
2408	*
2409	*/
2410	void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2411	{
2412	struct audit_context *context = current->audit_context;
2413	context->mq_getsetattr.mqdes = mqdes;
2414	context->mq_getsetattr.mqstat = *mqstat;
2415	context->type = AUDIT_MQ_GETSETATTR;
2416	}
2417
2418	/**
2419	* audit_ipc_obj - record audit data for ipc object
2420	* @ipcp: ipc permissions
2421	*
2422	*/
2423	void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2424	{
2425	struct audit_context *context = current->audit_context;
2426	context->ipc.uid = ipcp->uid;
2427	context->ipc.gid = ipcp->gid;
2428	context->ipc.mode = ipcp->mode;
2429	context->ipc.has_perm = 0;
2430	security_ipc_getsecid(ipcp, &context->ipc.osid);
2431	context->type = AUDIT_IPC;
2432	}
2433
2434	/**
2435	* audit_ipc_set_perm - record audit data for new ipc permissions
2436	* @qbytes: msgq bytes
2437	* @uid: msgq user id
2438	* @gid: msgq group id
2439	* @mode: msgq mode (permissions)
2440	*
2441	* Called only after audit_ipc_obj().
2442	*/
2443	void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2444	{
2445	struct audit_context *context = current->audit_context;
2446
2447	context->ipc.qbytes = qbytes;
2448	context->ipc.perm_uid = uid;
2449	context->ipc.perm_gid = gid;
2450	context->ipc.perm_mode = mode;
2451	context->ipc.has_perm = 1;
2452	}
2453
2454	int __audit_bprm(struct linux_binprm *bprm)
2455	{
2456	struct audit_aux_data_execve *ax;
2457	struct audit_context *context = current->audit_context;
2458
2459	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2460	if (!ax)
2461	return -ENOMEM;
2462
2463	ax->argc = bprm->argc;
2464	ax->envc = bprm->envc;
2465	ax->mm = bprm->mm;
2466	ax->d.type = AUDIT_EXECVE;
2467	ax->d.next = context->aux;
2468	context->aux = (void *)ax;
2469	return 0;
2470	}
2471
2472
2473	/**
2474	* audit_socketcall - record audit data for sys_socketcall
2475	* @nargs: number of args
2476	* @args: args array
2477	*
2478	*/
2479	void __audit_socketcall(int nargs, unsigned long *args)
2480	{
2481	struct audit_context *context = current->audit_context;
2482
2483	context->type = AUDIT_SOCKETCALL;
2484	context->socketcall.nargs = nargs;
2485	memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2486	}
2487
2488	/**
2489	* __audit_fd_pair - record audit data for pipe and socketpair
2490	* @fd1: the first file descriptor
2491	* @fd2: the second file descriptor
2492	*
2493	*/
2494	void __audit_fd_pair(int fd1, int fd2)
2495	{
2496	struct audit_context *context = current->audit_context;
2497	context->fds[0] = fd1;
2498	context->fds[1] = fd2;
2499	}
2500
2501	/**
2502	* audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2503	* @len: data length in user space
2504	* @a: data address in kernel space
2505	*
2506	* Returns 0 for success or NULL context or < 0 on error.
2507	*/
2508	int __audit_sockaddr(int len, void *a)
2509	{
2510	struct audit_context *context = current->audit_context;
2511
2512	if (!context->sockaddr) {
2513	void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2514	if (!p)
2515	return -ENOMEM;
2516	context->sockaddr = p;
2517	}
2518
2519	context->sockaddr_len = len;
2520	memcpy(context->sockaddr, a, len);
2521	return 0;
2522	}
2523
2524	void __audit_ptrace(struct task_struct *t)
2525	{
2526	struct audit_context *context = current->audit_context;
2527
2528	context->target_pid = t->pid;
2529	context->target_auid = audit_get_loginuid(t);
2530	context->target_uid = task_uid(t);
2531	context->target_sessionid = audit_get_sessionid(t);
2532	security_task_getsecid(t, &context->target_sid);
2533	memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2534	}
2535
2536	/**
2537	* audit_signal_info - record signal info for shutting down audit subsystem
2538	* @sig: signal value
2539	* @t: task being signaled
2540	*
2541	* If the audit subsystem is being terminated, record the task (pid)
2542	* and uid that is doing that.
2543	*/
2544	int __audit_signal_info(int sig, struct task_struct *t)
2545	{
2546	struct audit_aux_data_pids *axp;
2547	struct task_struct *tsk = current;
2548	struct audit_context *ctx = tsk->audit_context;
2549	uid_t uid = current_uid(), t_uid = task_uid(t);
2550
2551	if (audit_pid && t->tgid == audit_pid) {
2552	if (sig == SIGTERM \|\| sig == SIGHUP \|\| sig == SIGUSR1 \|\| sig == SIGUSR2) {
2553	audit_sig_pid = tsk->pid;
2554	if (tsk->loginuid != -1)
2555	audit_sig_uid = tsk->loginuid;
2556	else
2557	audit_sig_uid = uid;
2558	security_task_getsecid(tsk, &audit_sig_sid);
2559	}
2560	if (!audit_signals \|\| audit_dummy_context())
2561	return 0;
2562	}
2563
2564	/* optimize the common case by putting first signal recipient directly
2565	* in audit_context */
2566	if (!ctx->target_pid) {
2567	ctx->target_pid = t->tgid;
2568	ctx->target_auid = audit_get_loginuid(t);
2569	ctx->target_uid = t_uid;
2570	ctx->target_sessionid = audit_get_sessionid(t);
2571	security_task_getsecid(t, &ctx->target_sid);
2572	memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2573	return 0;
2574	}
2575
2576	axp = (void *)ctx->aux_pids;
2577	if (!axp \|\| axp->pid_count == AUDIT_AUX_PIDS) {
2578	axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2579	if (!axp)
2580	return -ENOMEM;
2581
2582	axp->d.type = AUDIT_OBJ_PID;
2583	axp->d.next = ctx->aux_pids;
2584	ctx->aux_pids = (void *)axp;
2585	}
2586	BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2587
2588	axp->target_pid[axp->pid_count] = t->tgid;
2589	axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2590	axp->target_uid[axp->pid_count] = t_uid;
2591	axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2592	security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2593	memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2594	axp->pid_count++;
2595
2596	return 0;
2597	}
2598
2599	/**
2600	* __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2601	* @bprm: pointer to the bprm being processed
2602	* @new: the proposed new credentials
2603	* @old: the old credentials
2604	*
2605	* Simply check if the proc already has the caps given by the file and if not
2606	* store the priv escalation info for later auditing at the end of the syscall
2607	*
2608	* -Eric
2609	*/
2610	int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2611	const struct cred new, const struct cred old)
2612	{
2613	struct audit_aux_data_bprm_fcaps *ax;
2614	struct audit_context *context = current->audit_context;
2615	struct cpu_vfs_cap_data vcaps;
2616	struct dentry *dentry;
2617
2618	ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2619	if (!ax)
2620	return -ENOMEM;
2621
2622	ax->d.type = AUDIT_BPRM_FCAPS;
2623	ax->d.next = context->aux;
2624	context->aux = (void *)ax;
2625
2626	dentry = dget(bprm->file->f_dentry);
2627	get_vfs_caps_from_disk(dentry, &vcaps);
2628	dput(dentry);
2629
2630	ax->fcap.permitted = vcaps.permitted;
2631	ax->fcap.inheritable = vcaps.inheritable;
2632	ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2633	ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2634
2635	ax->old_pcap.permitted = old->cap_permitted;
2636	ax->old_pcap.inheritable = old->cap_inheritable;
2637	ax->old_pcap.effective = old->cap_effective;
2638
2639	ax->new_pcap.permitted = new->cap_permitted;
2640	ax->new_pcap.inheritable = new->cap_inheritable;
2641	ax->new_pcap.effective = new->cap_effective;
2642	return 0;
2643	}
2644
2645	/**
2646	* __audit_log_capset - store information about the arguments to the capset syscall
2647	* @pid: target pid of the capset call
2648	* @new: the new credentials
2649	* @old: the old (current) credentials
2650	*
2651	* Record the aguments userspace sent to sys_capset for later printing by the
2652	* audit system if applicable
2653	*/
2654	void __audit_log_capset(pid_t pid,
2655	const struct cred new, const struct cred old)
2656	{
2657	struct audit_context *context = current->audit_context;
2658	context->capset.pid = pid;
2659	context->capset.cap.effective = new->cap_effective;
2660	context->capset.cap.inheritable = new->cap_effective;
2661	context->capset.cap.permitted = new->cap_permitted;
2662	context->type = AUDIT_CAPSET;
2663	}
2664
2665	void __audit_mmap_fd(int fd, int flags)
2666	{
2667	struct audit_context *context = current->audit_context;
2668	context->mmap.fd = fd;
2669	context->mmap.flags = flags;
2670	context->type = AUDIT_MMAP;
2671	}
2672
2673	static void audit_log_abend(struct audit_buffer ab, char reason, long signr)
2674	{
2675	uid_t auid, uid;
2676	gid_t gid;
2677	unsigned int sessionid;
2678
2679	auid = audit_get_loginuid(current);
2680	sessionid = audit_get_sessionid(current);
2681	current_uid_gid(&uid, &gid);
2682
2683	audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2684	auid, uid, gid, sessionid);
2685	audit_log_task_context(ab);
2686	audit_log_format(ab, " pid=%d comm=", current->pid);
2687	audit_log_untrustedstring(ab, current->comm);
2688	audit_log_format(ab, " reason=");
2689	audit_log_string(ab, reason);
2690	audit_log_format(ab, " sig=%ld", signr);
2691	}
2692	/**
2693	* audit_core_dumps - record information about processes that end abnormally
2694	* @signr: signal value
2695	*
2696	* If a process ends with a core dump, something fishy is going on and we
2697	* should record the event for investigation.
2698	*/
2699	void audit_core_dumps(long signr)
2700	{
2701	struct audit_buffer *ab;
2702
2703	if (!audit_enabled)
2704	return;
2705
2706	if (signr == SIGQUIT) /* don't care for those */
2707	return;
2708
2709	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2710	audit_log_abend(ab, "memory violation", signr);
2711	audit_log_end(ab);
2712	}
2713
2714	void __audit_seccomp(unsigned long syscall, long signr, int code)
2715	{
2716	struct audit_buffer *ab;
2717
2718	ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2719	audit_log_abend(ab, "seccomp", signr);
2720	audit_log_format(ab, " syscall=%ld", syscall);
2721	audit_log_format(ab, " compat=%d", is_compat_task());
2722	audit_log_format(ab, " ip=0x%lx", KSTK_EIP(current));
2723	audit_log_format(ab, " code=0x%x", code);
2724	audit_log_end(ab);
2725	}
2726
2727	struct list_head *audit_killed_trees(void)
2728	{
2729	struct audit_context *ctx = current->audit_context;
2730	if (likely(!ctx \|\| !ctx->in_syscall))
2731	return NULL;
2732	return &ctx->killed_trees;
2733	}