1 /* auditsc.c -- System-call auditing support
2 * Handles all system-call specific auditing features.
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
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.
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.
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
23 * Written by Rickard E. (Rik) Faith <faith@redhat.com>
25 * Many of the ideas implemented here are from Stephen C. Tweedie,
26 * especially the idea of avoiding a copy by using getname.
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 * okir@suse.de and Copyright 2003 SuSE Linux AG.
32 * POSIX message queue support added by George Wilson <ltcgcw@us.ibm.com>,
35 * The support of additional filter rules compares (>, <, >=, <=) was
36 * added by Dustin Kirkland <dustin.kirkland@us.ibm.com>, 2005.
38 * Modified by Amy Griffis <amy.griffis@hp.com> to collect additional
39 * filesystem information.
41 * Subject and object context labeling support added by <danjones@us.ibm.com>
42 * and <dustin.kirkland@us.ibm.com> for LSPP certification compliance.
45 #include <linux/init.h>
46 #include <asm/types.h>
47 #include <linux/atomic.h>
49 #include <linux/namei.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>
73 /* flags stating the success for a syscall */
74 #define AUDITSC_INVALID 0
75 #define AUDITSC_SUCCESS 1
76 #define AUDITSC_FAILURE 2
78 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
79 * for saving names from getname(). If we get more names we will allocate
80 * a name dynamically and also add those to the list anchored by names_list. */
83 /* Indicates that audit should log the full pathname. */
84 #define AUDIT_NAME_FULL -1
86 /* no execve audit message should be longer than this (userspace limits) */
87 #define MAX_EXECVE_AUDIT_LEN 7500
89 /* number of audit rules */
92 /* determines whether we collect data for signals sent */
95 struct audit_cap_data {
96 kernel_cap_t permitted;
97 kernel_cap_t inheritable;
99 unsigned int fE; /* effective bit of a file capability */
100 kernel_cap_t effective; /* effective set of a process */
104 /* When fs/namei.c:getname() is called, we store the pointer in name and
105 * we don't let putname() free it (instead we free all of the saved
106 * pointers at syscall exit time).
108 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
110 struct list_head list; /* audit_context->names_list */
119 struct audit_cap_data fcap;
120 unsigned int fcap_ver;
121 int name_len; /* number of name's characters to log */
122 bool name_put; /* call __putname() for this name */
124 * This was an allocated audit_names and not from the array of
125 * names allocated in the task audit context. Thus this name
126 * should be freed on syscall exit
131 struct audit_aux_data {
132 struct audit_aux_data *next;
136 #define AUDIT_AUX_IPCPERM 0
138 /* Number of target pids per aux struct. */
139 #define AUDIT_AUX_PIDS 16
141 struct audit_aux_data_execve {
142 struct audit_aux_data d;
145 struct mm_struct *mm;
148 struct audit_aux_data_pids {
149 struct audit_aux_data d;
150 pid_t target_pid[AUDIT_AUX_PIDS];
151 uid_t target_auid[AUDIT_AUX_PIDS];
152 uid_t target_uid[AUDIT_AUX_PIDS];
153 unsigned int target_sessionid[AUDIT_AUX_PIDS];
154 u32 target_sid[AUDIT_AUX_PIDS];
155 char target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
159 struct audit_aux_data_bprm_fcaps {
160 struct audit_aux_data d;
161 struct audit_cap_data fcap;
162 unsigned int fcap_ver;
163 struct audit_cap_data old_pcap;
164 struct audit_cap_data new_pcap;
167 struct audit_aux_data_capset {
168 struct audit_aux_data d;
170 struct audit_cap_data cap;
173 struct audit_tree_refs {
174 struct audit_tree_refs *next;
175 struct audit_chunk *c[31];
178 /* The per-task audit context. */
179 struct audit_context {
180 int dummy; /* must be the first element */
181 int in_syscall; /* 1 if task is in a syscall */
182 enum audit_state state, current_state;
183 unsigned int serial; /* serial number for record */
184 int major; /* syscall number */
185 struct timespec ctime; /* time of syscall entry */
186 unsigned long argv[4]; /* syscall arguments */
187 long return_code;/* syscall return code */
189 int return_valid; /* return code is valid */
191 * The names_list is the list of all audit_names collected during this
192 * syscall. The first AUDIT_NAMES entries in the names_list will
193 * actually be from the preallocated_names array for performance
194 * reasons. Except during allocation they should never be referenced
195 * through the preallocated_names array and should only be found/used
196 * by running the names_list.
198 struct audit_names preallocated_names[AUDIT_NAMES];
199 int name_count; /* total records in names_list */
200 struct list_head names_list; /* anchor for struct audit_names->list */
201 char * filterkey; /* key for rule that triggered record */
203 struct audit_context *previous; /* For nested syscalls */
204 struct audit_aux_data *aux;
205 struct audit_aux_data *aux_pids;
206 struct sockaddr_storage *sockaddr;
208 /* Save things to print about task_struct */
210 uid_t uid, euid, suid, fsuid;
211 gid_t gid, egid, sgid, fsgid;
212 unsigned long personality;
218 unsigned int target_sessionid;
220 char target_comm[TASK_COMM_LEN];
222 struct audit_tree_refs *trees, *first_trees;
223 struct list_head killed_trees;
241 unsigned long qbytes;
245 struct mq_attr mqstat;
254 unsigned int msg_prio;
255 struct timespec abs_timeout;
264 struct audit_cap_data cap;
279 static inline int open_arg(int flags, int mask)
281 int n = ACC_MODE(flags);
282 if (flags & (O_TRUNC | O_CREAT))
283 n |= AUDIT_PERM_WRITE;
287 static int audit_match_perm(struct audit_context *ctx, int mask)
294 switch (audit_classify_syscall(ctx->arch, n)) {
296 if ((mask & AUDIT_PERM_WRITE) &&
297 audit_match_class(AUDIT_CLASS_WRITE, n))
299 if ((mask & AUDIT_PERM_READ) &&
300 audit_match_class(AUDIT_CLASS_READ, n))
302 if ((mask & AUDIT_PERM_ATTR) &&
303 audit_match_class(AUDIT_CLASS_CHATTR, n))
306 case 1: /* 32bit on biarch */
307 if ((mask & AUDIT_PERM_WRITE) &&
308 audit_match_class(AUDIT_CLASS_WRITE_32, n))
310 if ((mask & AUDIT_PERM_READ) &&
311 audit_match_class(AUDIT_CLASS_READ_32, n))
313 if ((mask & AUDIT_PERM_ATTR) &&
314 audit_match_class(AUDIT_CLASS_CHATTR_32, n))
318 return mask & ACC_MODE(ctx->argv[1]);
320 return mask & ACC_MODE(ctx->argv[2]);
321 case 4: /* socketcall */
322 return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
324 return mask & AUDIT_PERM_EXEC;
330 static int audit_match_filetype(struct audit_context *ctx, int val)
332 struct audit_names *n;
333 umode_t mode = (umode_t)val;
338 list_for_each_entry(n, &ctx->names_list, list) {
339 if ((n->ino != -1) &&
340 ((n->mode & S_IFMT) == mode))
348 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
349 * ->first_trees points to its beginning, ->trees - to the current end of data.
350 * ->tree_count is the number of free entries in array pointed to by ->trees.
351 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
352 * "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
353 * it's going to remain 1-element for almost any setup) until we free context itself.
354 * References in it _are_ dropped - at the same time we free/drop aux stuff.
357 #ifdef CONFIG_AUDIT_TREE
358 static void audit_set_auditable(struct audit_context *ctx)
362 ctx->current_state = AUDIT_RECORD_CONTEXT;
366 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
368 struct audit_tree_refs *p = ctx->trees;
369 int left = ctx->tree_count;
371 p->c[--left] = chunk;
372 ctx->tree_count = left;
381 ctx->tree_count = 30;
387 static int grow_tree_refs(struct audit_context *ctx)
389 struct audit_tree_refs *p = ctx->trees;
390 ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
396 p->next = ctx->trees;
398 ctx->first_trees = ctx->trees;
399 ctx->tree_count = 31;
404 static void unroll_tree_refs(struct audit_context *ctx,
405 struct audit_tree_refs *p, int count)
407 #ifdef CONFIG_AUDIT_TREE
408 struct audit_tree_refs *q;
411 /* we started with empty chain */
412 p = ctx->first_trees;
414 /* if the very first allocation has failed, nothing to do */
419 for (q = p; q != ctx->trees; q = q->next, n = 31) {
421 audit_put_chunk(q->c[n]);
425 while (n-- > ctx->tree_count) {
426 audit_put_chunk(q->c[n]);
430 ctx->tree_count = count;
434 static void free_tree_refs(struct audit_context *ctx)
436 struct audit_tree_refs *p, *q;
437 for (p = ctx->first_trees; p; p = q) {
443 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
445 #ifdef CONFIG_AUDIT_TREE
446 struct audit_tree_refs *p;
451 for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
452 for (n = 0; n < 31; n++)
453 if (audit_tree_match(p->c[n], tree))
458 for (n = ctx->tree_count; n < 31; n++)
459 if (audit_tree_match(p->c[n], tree))
466 static int audit_compare_id(uid_t uid1,
467 struct audit_names *name,
468 unsigned long name_offset,
469 struct audit_field *f,
470 struct audit_context *ctx)
472 struct audit_names *n;
477 BUILD_BUG_ON(sizeof(uid_t) != sizeof(gid_t));
480 addr = (unsigned long)name;
483 uid2 = *(uid_t *)addr;
484 rc = audit_comparator(uid1, f->op, uid2);
490 list_for_each_entry(n, &ctx->names_list, list) {
491 addr = (unsigned long)n;
494 uid2 = *(uid_t *)addr;
496 rc = audit_comparator(uid1, f->op, uid2);
504 static int audit_field_compare(struct task_struct *tsk,
505 const struct cred *cred,
506 struct audit_field *f,
507 struct audit_context *ctx,
508 struct audit_names *name)
511 /* process to file object comparisons */
512 case AUDIT_COMPARE_UID_TO_OBJ_UID:
513 return audit_compare_id(cred->uid,
514 name, offsetof(struct audit_names, uid),
516 case AUDIT_COMPARE_GID_TO_OBJ_GID:
517 return audit_compare_id(cred->gid,
518 name, offsetof(struct audit_names, gid),
520 case AUDIT_COMPARE_EUID_TO_OBJ_UID:
521 return audit_compare_id(cred->euid,
522 name, offsetof(struct audit_names, uid),
524 case AUDIT_COMPARE_EGID_TO_OBJ_GID:
525 return audit_compare_id(cred->egid,
526 name, offsetof(struct audit_names, gid),
528 case AUDIT_COMPARE_AUID_TO_OBJ_UID:
529 return audit_compare_id(tsk->loginuid,
530 name, offsetof(struct audit_names, uid),
532 case AUDIT_COMPARE_SUID_TO_OBJ_UID:
533 return audit_compare_id(cred->suid,
534 name, offsetof(struct audit_names, uid),
536 case AUDIT_COMPARE_SGID_TO_OBJ_GID:
537 return audit_compare_id(cred->sgid,
538 name, offsetof(struct audit_names, gid),
540 case AUDIT_COMPARE_FSUID_TO_OBJ_UID:
541 return audit_compare_id(cred->fsuid,
542 name, offsetof(struct audit_names, uid),
544 case AUDIT_COMPARE_FSGID_TO_OBJ_GID:
545 return audit_compare_id(cred->fsgid,
546 name, offsetof(struct audit_names, gid),
549 WARN(1, "Missing AUDIT_COMPARE define. Report as a bug\n");
555 /* Determine if any context name data matches a rule's watch data */
556 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
559 * If task_creation is true, this is an explicit indication that we are
560 * filtering a task rule at task creation time. This and tsk == current are
561 * the only situations where tsk->cred may be accessed without an rcu read lock.
563 static int audit_filter_rules(struct task_struct *tsk,
564 struct audit_krule *rule,
565 struct audit_context *ctx,
566 struct audit_names *name,
567 enum audit_state *state,
570 const struct cred *cred;
574 cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
576 for (i = 0; i < rule->field_count; i++) {
577 struct audit_field *f = &rule->fields[i];
578 struct audit_names *n;
583 result = audit_comparator(tsk->pid, f->op, f->val);
588 ctx->ppid = sys_getppid();
589 result = audit_comparator(ctx->ppid, f->op, f->val);
593 result = audit_comparator(cred->uid, f->op, f->val);
596 result = audit_comparator(cred->euid, f->op, f->val);
599 result = audit_comparator(cred->suid, f->op, f->val);
602 result = audit_comparator(cred->fsuid, f->op, f->val);
605 result = audit_comparator(cred->gid, f->op, f->val);
608 result = audit_comparator(cred->egid, f->op, f->val);
611 result = audit_comparator(cred->sgid, f->op, f->val);
614 result = audit_comparator(cred->fsgid, f->op, f->val);
617 result = audit_comparator(tsk->personality, f->op, f->val);
621 result = audit_comparator(ctx->arch, f->op, f->val);
625 if (ctx && ctx->return_valid)
626 result = audit_comparator(ctx->return_code, f->op, f->val);
629 if (ctx && ctx->return_valid) {
631 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
633 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
638 if (audit_comparator(MAJOR(name->dev), f->op, f->val) ||
639 audit_comparator(MAJOR(name->rdev), f->op, f->val))
642 list_for_each_entry(n, &ctx->names_list, list) {
643 if (audit_comparator(MAJOR(n->dev), f->op, f->val) ||
644 audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
653 if (audit_comparator(MINOR(name->dev), f->op, f->val) ||
654 audit_comparator(MINOR(name->rdev), f->op, f->val))
657 list_for_each_entry(n, &ctx->names_list, list) {
658 if (audit_comparator(MINOR(n->dev), f->op, f->val) ||
659 audit_comparator(MINOR(n->rdev), f->op, f->val)) {
668 result = (name->ino == f->val);
670 list_for_each_entry(n, &ctx->names_list, list) {
671 if (audit_comparator(n->ino, f->op, f->val)) {
680 result = audit_comparator(name->uid, f->op, f->val);
682 list_for_each_entry(n, &ctx->names_list, list) {
683 if (audit_comparator(n->uid, f->op, f->val)) {
692 result = audit_comparator(name->gid, f->op, f->val);
694 list_for_each_entry(n, &ctx->names_list, list) {
695 if (audit_comparator(n->gid, f->op, f->val)) {
704 result = audit_watch_compare(rule->watch, name->ino, name->dev);
708 result = match_tree_refs(ctx, rule->tree);
713 result = audit_comparator(tsk->loginuid, f->op, f->val);
715 case AUDIT_SUBJ_USER:
716 case AUDIT_SUBJ_ROLE:
717 case AUDIT_SUBJ_TYPE:
720 /* NOTE: this may return negative values indicating
721 a temporary error. We simply treat this as a
722 match for now to avoid losing information that
723 may be wanted. An error message will also be
727 security_task_getsecid(tsk, &sid);
730 result = security_audit_rule_match(sid, f->type,
739 case AUDIT_OBJ_LEV_LOW:
740 case AUDIT_OBJ_LEV_HIGH:
741 /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
744 /* Find files that match */
746 result = security_audit_rule_match(
747 name->osid, f->type, f->op,
750 list_for_each_entry(n, &ctx->names_list, list) {
751 if (security_audit_rule_match(n->osid, f->type,
759 /* Find ipc objects that match */
760 if (!ctx || ctx->type != AUDIT_IPC)
762 if (security_audit_rule_match(ctx->ipc.osid,
773 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
775 case AUDIT_FILTERKEY:
776 /* ignore this field for filtering */
780 result = audit_match_perm(ctx, f->val);
783 result = audit_match_filetype(ctx, f->val);
785 case AUDIT_FIELD_COMPARE:
786 result = audit_field_compare(tsk, cred, f, ctx, name);
794 if (rule->prio <= ctx->prio)
796 if (rule->filterkey) {
797 kfree(ctx->filterkey);
798 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
800 ctx->prio = rule->prio;
802 switch (rule->action) {
803 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
804 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
809 /* At process creation time, we can determine if system-call auditing is
810 * completely disabled for this task. Since we only have the task
811 * structure at this point, we can only check uid and gid.
813 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
815 struct audit_entry *e;
816 enum audit_state state;
819 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
820 if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
822 if (state == AUDIT_RECORD_CONTEXT)
823 *key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
829 return AUDIT_BUILD_CONTEXT;
832 /* At syscall entry and exit time, this filter is called if the
833 * audit_state is not low enough that auditing cannot take place, but is
834 * also not high enough that we already know we have to write an audit
835 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
837 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
838 struct audit_context *ctx,
839 struct list_head *list)
841 struct audit_entry *e;
842 enum audit_state state;
844 if (audit_pid && tsk->tgid == audit_pid)
845 return AUDIT_DISABLED;
848 if (!list_empty(list)) {
849 int word = AUDIT_WORD(ctx->major);
850 int bit = AUDIT_BIT(ctx->major);
852 list_for_each_entry_rcu(e, list, list) {
853 if ((e->rule.mask[word] & bit) == bit &&
854 audit_filter_rules(tsk, &e->rule, ctx, NULL,
857 ctx->current_state = state;
863 return AUDIT_BUILD_CONTEXT;
867 * Given an audit_name check the inode hash table to see if they match.
868 * Called holding the rcu read lock to protect the use of audit_inode_hash
870 static int audit_filter_inode_name(struct task_struct *tsk,
871 struct audit_names *n,
872 struct audit_context *ctx) {
874 int h = audit_hash_ino((u32)n->ino);
875 struct list_head *list = &audit_inode_hash[h];
876 struct audit_entry *e;
877 enum audit_state state;
879 word = AUDIT_WORD(ctx->major);
880 bit = AUDIT_BIT(ctx->major);
882 if (list_empty(list))
885 list_for_each_entry_rcu(e, list, list) {
886 if ((e->rule.mask[word] & bit) == bit &&
887 audit_filter_rules(tsk, &e->rule, ctx, n, &state, false)) {
888 ctx->current_state = state;
896 /* At syscall exit time, this filter is called if any audit_names have been
897 * collected during syscall processing. We only check rules in sublists at hash
898 * buckets applicable to the inode numbers in audit_names.
899 * Regarding audit_state, same rules apply as for audit_filter_syscall().
901 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
903 struct audit_names *n;
905 if (audit_pid && tsk->tgid == audit_pid)
910 list_for_each_entry(n, &ctx->names_list, list) {
911 if (audit_filter_inode_name(tsk, n, ctx))
917 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
921 struct audit_context *context = tsk->audit_context;
925 context->return_valid = return_valid;
928 * we need to fix up the return code in the audit logs if the actual
929 * return codes are later going to be fixed up by the arch specific
932 * This is actually a test for:
933 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
934 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
936 * but is faster than a bunch of ||
938 if (unlikely(return_code <= -ERESTARTSYS) &&
939 (return_code >= -ERESTART_RESTARTBLOCK) &&
940 (return_code != -ENOIOCTLCMD))
941 context->return_code = -EINTR;
943 context->return_code = return_code;
945 if (context->in_syscall && !context->dummy) {
946 audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
947 audit_filter_inodes(tsk, context);
950 tsk->audit_context = NULL;
954 static inline void audit_free_names(struct audit_context *context)
956 struct audit_names *n, *next;
959 if (context->put_count + context->ino_count != context->name_count) {
960 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
961 " name_count=%d put_count=%d"
962 " ino_count=%d [NOT freeing]\n",
964 context->serial, context->major, context->in_syscall,
965 context->name_count, context->put_count,
967 list_for_each_entry(n, &context->names_list, list) {
968 printk(KERN_ERR "names[%d] = %p = %s\n", i,
969 n->name, n->name ?: "(null)");
976 context->put_count = 0;
977 context->ino_count = 0;
980 list_for_each_entry_safe(n, next, &context->names_list, list) {
982 if (n->name && n->name_put)
987 context->name_count = 0;
988 path_put(&context->pwd);
989 context->pwd.dentry = NULL;
990 context->pwd.mnt = NULL;
993 static inline void audit_free_aux(struct audit_context *context)
995 struct audit_aux_data *aux;
997 while ((aux = context->aux)) {
998 context->aux = aux->next;
1001 while ((aux = context->aux_pids)) {
1002 context->aux_pids = aux->next;
1007 static inline void audit_zero_context(struct audit_context *context,
1008 enum audit_state state)
1010 memset(context, 0, sizeof(*context));
1011 context->state = state;
1012 context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1015 static inline struct audit_context *audit_alloc_context(enum audit_state state)
1017 struct audit_context *context;
1019 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
1021 audit_zero_context(context, state);
1022 INIT_LIST_HEAD(&context->killed_trees);
1023 INIT_LIST_HEAD(&context->names_list);
1028 * audit_alloc - allocate an audit context block for a task
1031 * Filter on the task information and allocate a per-task audit context
1032 * if necessary. Doing so turns on system call auditing for the
1033 * specified task. This is called from copy_process, so no lock is
1036 int audit_alloc(struct task_struct *tsk)
1038 struct audit_context *context;
1039 enum audit_state state;
1042 if (likely(!audit_ever_enabled))
1043 return 0; /* Return if not auditing. */
1045 state = audit_filter_task(tsk, &key);
1046 if (state == AUDIT_DISABLED)
1049 if (!(context = audit_alloc_context(state))) {
1051 audit_log_lost("out of memory in audit_alloc");
1054 context->filterkey = key;
1056 tsk->audit_context = context;
1057 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
1061 static inline void audit_free_context(struct audit_context *context)
1063 struct audit_context *previous;
1067 previous = context->previous;
1068 if (previous || (count && count < 10)) {
1070 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
1071 " freeing multiple contexts (%d)\n",
1072 context->serial, context->major,
1073 context->name_count, count);
1075 audit_free_names(context);
1076 unroll_tree_refs(context, NULL, 0);
1077 free_tree_refs(context);
1078 audit_free_aux(context);
1079 kfree(context->filterkey);
1080 kfree(context->sockaddr);
1085 printk(KERN_ERR "audit: freed %d contexts\n", count);
1088 void audit_log_task_context(struct audit_buffer *ab)
1095 security_task_getsecid(current, &sid);
1099 error = security_secid_to_secctx(sid, &ctx, &len);
1101 if (error != -EINVAL)
1106 audit_log_format(ab, " subj=%s", ctx);
1107 security_release_secctx(ctx, len);
1111 audit_panic("error in audit_log_task_context");
1115 EXPORT_SYMBOL(audit_log_task_context);
1117 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
1119 char name[sizeof(tsk->comm)];
1120 struct mm_struct *mm = tsk->mm;
1121 struct vm_area_struct *vma;
1123 /* tsk == current */
1125 get_task_comm(name, tsk);
1126 audit_log_format(ab, " comm=");
1127 audit_log_untrustedstring(ab, name);
1130 down_read(&mm->mmap_sem);
1133 if ((vma->vm_flags & VM_EXECUTABLE) &&
1135 audit_log_d_path(ab, "exe=",
1136 &vma->vm_file->f_path);
1141 up_read(&mm->mmap_sem);
1143 audit_log_task_context(ab);
1146 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
1147 uid_t auid, uid_t uid, unsigned int sessionid,
1148 u32 sid, char *comm)
1150 struct audit_buffer *ab;
1155 ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1159 audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
1161 if (security_secid_to_secctx(sid, &ctx, &len)) {
1162 audit_log_format(ab, " obj=(none)");
1165 audit_log_format(ab, " obj=%s", ctx);
1166 security_release_secctx(ctx, len);
1168 audit_log_format(ab, " ocomm=");
1169 audit_log_untrustedstring(ab, comm);
1176 * to_send and len_sent accounting are very loose estimates. We aren't
1177 * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1178 * within about 500 bytes (next page boundary)
1180 * why snprintf? an int is up to 12 digits long. if we just assumed when
1181 * logging that a[%d]= was going to be 16 characters long we would be wasting
1182 * space in every audit message. In one 7500 byte message we can log up to
1183 * about 1000 min size arguments. That comes down to about 50% waste of space
1184 * if we didn't do the snprintf to find out how long arg_num_len was.
1186 static int audit_log_single_execve_arg(struct audit_context *context,
1187 struct audit_buffer **ab,
1190 const char __user *p,
1193 char arg_num_len_buf[12];
1194 const char __user *tmp_p = p;
1195 /* how many digits are in arg_num? 5 is the length of ' a=""' */
1196 size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1197 size_t len, len_left, to_send;
1198 size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1199 unsigned int i, has_cntl = 0, too_long = 0;
1202 /* strnlen_user includes the null we don't want to send */
1203 len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1206 * We just created this mm, if we can't find the strings
1207 * we just copied into it something is _very_ wrong. Similar
1208 * for strings that are too long, we should not have created
1211 if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1213 send_sig(SIGKILL, current, 0);
1217 /* walk the whole argument looking for non-ascii chars */
1219 if (len_left > MAX_EXECVE_AUDIT_LEN)
1220 to_send = MAX_EXECVE_AUDIT_LEN;
1223 ret = copy_from_user(buf, tmp_p, to_send);
1225 * There is no reason for this copy to be short. We just
1226 * copied them here, and the mm hasn't been exposed to user-
1231 send_sig(SIGKILL, current, 0);
1234 buf[to_send] = '\0';
1235 has_cntl = audit_string_contains_control(buf, to_send);
1238 * hex messages get logged as 2 bytes, so we can only
1239 * send half as much in each message
1241 max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1244 len_left -= to_send;
1246 } while (len_left > 0);
1250 if (len > max_execve_audit_len)
1253 /* rewalk the argument actually logging the message */
1254 for (i = 0; len_left > 0; i++) {
1257 if (len_left > max_execve_audit_len)
1258 to_send = max_execve_audit_len;
1262 /* do we have space left to send this argument in this ab? */
1263 room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1265 room_left -= (to_send * 2);
1267 room_left -= to_send;
1268 if (room_left < 0) {
1271 *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1277 * first record needs to say how long the original string was
1278 * so we can be sure nothing was lost.
1280 if ((i == 0) && (too_long))
1281 audit_log_format(*ab, " a%d_len=%zu", arg_num,
1282 has_cntl ? 2*len : len);
1285 * normally arguments are small enough to fit and we already
1286 * filled buf above when we checked for control characters
1287 * so don't bother with another copy_from_user
1289 if (len >= max_execve_audit_len)
1290 ret = copy_from_user(buf, p, to_send);
1295 send_sig(SIGKILL, current, 0);
1298 buf[to_send] = '\0';
1300 /* actually log it */
1301 audit_log_format(*ab, " a%d", arg_num);
1303 audit_log_format(*ab, "[%d]", i);
1304 audit_log_format(*ab, "=");
1306 audit_log_n_hex(*ab, buf, to_send);
1308 audit_log_string(*ab, buf);
1311 len_left -= to_send;
1312 *len_sent += arg_num_len;
1314 *len_sent += to_send * 2;
1316 *len_sent += to_send;
1318 /* include the null we didn't log */
1322 static void audit_log_execve_info(struct audit_context *context,
1323 struct audit_buffer **ab,
1324 struct audit_aux_data_execve *axi)
1327 size_t len, len_sent = 0;
1328 const char __user *p;
1331 if (axi->mm != current->mm)
1332 return; /* execve failed, no additional info */
1334 p = (const char __user *)axi->mm->arg_start;
1336 audit_log_format(*ab, "argc=%d", axi->argc);
1339 * we need some kernel buffer to hold the userspace args. Just
1340 * allocate one big one rather than allocating one of the right size
1341 * for every single argument inside audit_log_single_execve_arg()
1342 * should be <8k allocation so should be pretty safe.
1344 buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1346 audit_panic("out of memory for argv string\n");
1350 for (i = 0; i < axi->argc; i++) {
1351 len = audit_log_single_execve_arg(context, ab, i,
1360 static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1364 audit_log_format(ab, " %s=", prefix);
1365 CAP_FOR_EACH_U32(i) {
1366 audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1370 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1372 kernel_cap_t *perm = &name->fcap.permitted;
1373 kernel_cap_t *inh = &name->fcap.inheritable;
1376 if (!cap_isclear(*perm)) {
1377 audit_log_cap(ab, "cap_fp", perm);
1380 if (!cap_isclear(*inh)) {
1381 audit_log_cap(ab, "cap_fi", inh);
1386 audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1389 static void show_special(struct audit_context *context, int *call_panic)
1391 struct audit_buffer *ab;
1394 ab = audit_log_start(context, GFP_KERNEL, context->type);
1398 switch (context->type) {
1399 case AUDIT_SOCKETCALL: {
1400 int nargs = context->socketcall.nargs;
1401 audit_log_format(ab, "nargs=%d", nargs);
1402 for (i = 0; i < nargs; i++)
1403 audit_log_format(ab, " a%d=%lx", i,
1404 context->socketcall.args[i]);
1407 u32 osid = context->ipc.osid;
1409 audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1410 context->ipc.uid, context->ipc.gid, context->ipc.mode);
1414 if (security_secid_to_secctx(osid, &ctx, &len)) {
1415 audit_log_format(ab, " osid=%u", osid);
1418 audit_log_format(ab, " obj=%s", ctx);
1419 security_release_secctx(ctx, len);
1422 if (context->ipc.has_perm) {
1424 ab = audit_log_start(context, GFP_KERNEL,
1425 AUDIT_IPC_SET_PERM);
1426 audit_log_format(ab,
1427 "qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1428 context->ipc.qbytes,
1429 context->ipc.perm_uid,
1430 context->ipc.perm_gid,
1431 context->ipc.perm_mode);
1436 case AUDIT_MQ_OPEN: {
1437 audit_log_format(ab,
1438 "oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1439 "mq_msgsize=%ld mq_curmsgs=%ld",
1440 context->mq_open.oflag, context->mq_open.mode,
1441 context->mq_open.attr.mq_flags,
1442 context->mq_open.attr.mq_maxmsg,
1443 context->mq_open.attr.mq_msgsize,
1444 context->mq_open.attr.mq_curmsgs);
1446 case AUDIT_MQ_SENDRECV: {
1447 audit_log_format(ab,
1448 "mqdes=%d msg_len=%zd msg_prio=%u "
1449 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1450 context->mq_sendrecv.mqdes,
1451 context->mq_sendrecv.msg_len,
1452 context->mq_sendrecv.msg_prio,
1453 context->mq_sendrecv.abs_timeout.tv_sec,
1454 context->mq_sendrecv.abs_timeout.tv_nsec);
1456 case AUDIT_MQ_NOTIFY: {
1457 audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1458 context->mq_notify.mqdes,
1459 context->mq_notify.sigev_signo);
1461 case AUDIT_MQ_GETSETATTR: {
1462 struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1463 audit_log_format(ab,
1464 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1466 context->mq_getsetattr.mqdes,
1467 attr->mq_flags, attr->mq_maxmsg,
1468 attr->mq_msgsize, attr->mq_curmsgs);
1470 case AUDIT_CAPSET: {
1471 audit_log_format(ab, "pid=%d", context->capset.pid);
1472 audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1473 audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1474 audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1477 audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1478 context->mmap.flags);
1484 static void audit_log_name(struct audit_context *context, struct audit_names *n,
1485 int record_num, int *call_panic)
1487 struct audit_buffer *ab;
1488 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1490 return; /* audit_panic has been called */
1492 audit_log_format(ab, "item=%d", record_num);
1495 switch (n->name_len) {
1496 case AUDIT_NAME_FULL:
1497 /* log the full path */
1498 audit_log_format(ab, " name=");
1499 audit_log_untrustedstring(ab, n->name);
1502 /* name was specified as a relative path and the
1503 * directory component is the cwd */
1504 audit_log_d_path(ab, "name=", &context->pwd);
1507 /* log the name's directory component */
1508 audit_log_format(ab, " name=");
1509 audit_log_n_untrustedstring(ab, n->name,
1513 audit_log_format(ab, " name=(null)");
1515 if (n->ino != (unsigned long)-1) {
1516 audit_log_format(ab, " inode=%lu"
1517 " dev=%02x:%02x mode=%#ho"
1518 " ouid=%u ogid=%u rdev=%02x:%02x",
1531 if (security_secid_to_secctx(
1532 n->osid, &ctx, &len)) {
1533 audit_log_format(ab, " osid=%u", n->osid);
1536 audit_log_format(ab, " obj=%s", ctx);
1537 security_release_secctx(ctx, len);
1541 audit_log_fcaps(ab, n);
1546 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1548 const struct cred *cred;
1549 int i, call_panic = 0;
1550 struct audit_buffer *ab;
1551 struct audit_aux_data *aux;
1553 struct audit_names *n;
1555 /* tsk == current */
1556 context->pid = tsk->pid;
1558 context->ppid = sys_getppid();
1559 cred = current_cred();
1560 context->uid = cred->uid;
1561 context->gid = cred->gid;
1562 context->euid = cred->euid;
1563 context->suid = cred->suid;
1564 context->fsuid = cred->fsuid;
1565 context->egid = cred->egid;
1566 context->sgid = cred->sgid;
1567 context->fsgid = cred->fsgid;
1568 context->personality = tsk->personality;
1570 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1572 return; /* audit_panic has been called */
1573 audit_log_format(ab, "arch=%x syscall=%d",
1574 context->arch, context->major);
1575 if (context->personality != PER_LINUX)
1576 audit_log_format(ab, " per=%lx", context->personality);
1577 if (context->return_valid)
1578 audit_log_format(ab, " success=%s exit=%ld",
1579 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1580 context->return_code);
1582 spin_lock_irq(&tsk->sighand->siglock);
1583 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1584 tty = tsk->signal->tty->name;
1587 spin_unlock_irq(&tsk->sighand->siglock);
1589 audit_log_format(ab,
1590 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1591 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1592 " euid=%u suid=%u fsuid=%u"
1593 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1598 context->name_count,
1604 context->euid, context->suid, context->fsuid,
1605 context->egid, context->sgid, context->fsgid, tty,
1609 audit_log_task_info(ab, tsk);
1610 audit_log_key(ab, context->filterkey);
1613 for (aux = context->aux; aux; aux = aux->next) {
1615 ab = audit_log_start(context, GFP_KERNEL, aux->type);
1617 continue; /* audit_panic has been called */
1619 switch (aux->type) {
1621 case AUDIT_EXECVE: {
1622 struct audit_aux_data_execve *axi = (void *)aux;
1623 audit_log_execve_info(context, &ab, axi);
1626 case AUDIT_BPRM_FCAPS: {
1627 struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1628 audit_log_format(ab, "fver=%x", axs->fcap_ver);
1629 audit_log_cap(ab, "fp", &axs->fcap.permitted);
1630 audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1631 audit_log_format(ab, " fe=%d", axs->fcap.fE);
1632 audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1633 audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1634 audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1635 audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1636 audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1637 audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1645 show_special(context, &call_panic);
1647 if (context->fds[0] >= 0) {
1648 ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1650 audit_log_format(ab, "fd0=%d fd1=%d",
1651 context->fds[0], context->fds[1]);
1656 if (context->sockaddr_len) {
1657 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1659 audit_log_format(ab, "saddr=");
1660 audit_log_n_hex(ab, (void *)context->sockaddr,
1661 context->sockaddr_len);
1666 for (aux = context->aux_pids; aux; aux = aux->next) {
1667 struct audit_aux_data_pids *axs = (void *)aux;
1669 for (i = 0; i < axs->pid_count; i++)
1670 if (audit_log_pid_context(context, axs->target_pid[i],
1671 axs->target_auid[i],
1673 axs->target_sessionid[i],
1675 axs->target_comm[i]))
1679 if (context->target_pid &&
1680 audit_log_pid_context(context, context->target_pid,
1681 context->target_auid, context->target_uid,
1682 context->target_sessionid,
1683 context->target_sid, context->target_comm))
1686 if (context->pwd.dentry && context->pwd.mnt) {
1687 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1689 audit_log_d_path(ab, "cwd=", &context->pwd);
1695 list_for_each_entry(n, &context->names_list, list)
1696 audit_log_name(context, n, i++, &call_panic);
1698 /* Send end of event record to help user space know we are finished */
1699 ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1703 audit_panic("error converting sid to string");
1707 * audit_free - free a per-task audit context
1708 * @tsk: task whose audit context block to free
1710 * Called from copy_process and do_exit
1712 void __audit_free(struct task_struct *tsk)
1714 struct audit_context *context;
1716 context = audit_get_context(tsk, 0, 0);
1720 /* Check for system calls that do not go through the exit
1721 * function (e.g., exit_group), then free context block.
1722 * We use GFP_ATOMIC here because we might be doing this
1723 * in the context of the idle thread */
1724 /* that can happen only if we are called from do_exit() */
1725 if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1726 audit_log_exit(context, tsk);
1727 if (!list_empty(&context->killed_trees))
1728 audit_kill_trees(&context->killed_trees);
1730 audit_free_context(context);
1734 * audit_syscall_entry - fill in an audit record at syscall entry
1735 * @arch: architecture type
1736 * @major: major syscall type (function)
1737 * @a1: additional syscall register 1
1738 * @a2: additional syscall register 2
1739 * @a3: additional syscall register 3
1740 * @a4: additional syscall register 4
1742 * Fill in audit context at syscall entry. This only happens if the
1743 * audit context was created when the task was created and the state or
1744 * filters demand the audit context be built. If the state from the
1745 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1746 * then the record will be written at syscall exit time (otherwise, it
1747 * will only be written if another part of the kernel requests that it
1750 void __audit_syscall_entry(int arch, int major,
1751 unsigned long a1, unsigned long a2,
1752 unsigned long a3, unsigned long a4)
1754 struct task_struct *tsk = current;
1755 struct audit_context *context = tsk->audit_context;
1756 enum audit_state state;
1762 * This happens only on certain architectures that make system
1763 * calls in kernel_thread via the entry.S interface, instead of
1764 * with direct calls. (If you are porting to a new
1765 * architecture, hitting this condition can indicate that you
1766 * got the _exit/_leave calls backward in entry.S.)
1770 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1772 * This also happens with vm86 emulation in a non-nested manner
1773 * (entries without exits), so this case must be caught.
1775 if (context->in_syscall) {
1776 struct audit_context *newctx;
1780 "audit(:%d) pid=%d in syscall=%d;"
1781 " entering syscall=%d\n",
1782 context->serial, tsk->pid, context->major, major);
1784 newctx = audit_alloc_context(context->state);
1786 newctx->previous = context;
1788 tsk->audit_context = newctx;
1790 /* If we can't alloc a new context, the best we
1791 * can do is to leak memory (any pending putname
1792 * will be lost). The only other alternative is
1793 * to abandon auditing. */
1794 audit_zero_context(context, context->state);
1797 BUG_ON(context->in_syscall || context->name_count);
1802 context->arch = arch;
1803 context->major = major;
1804 context->argv[0] = a1;
1805 context->argv[1] = a2;
1806 context->argv[2] = a3;
1807 context->argv[3] = a4;
1809 state = context->state;
1810 context->dummy = !audit_n_rules;
1811 if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1813 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1815 if (state == AUDIT_DISABLED)
1818 context->serial = 0;
1819 context->ctime = CURRENT_TIME;
1820 context->in_syscall = 1;
1821 context->current_state = state;
1826 * audit_syscall_exit - deallocate audit context after a system call
1827 * @pt_regs: syscall registers
1829 * Tear down after system call. If the audit context has been marked as
1830 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1831 * filtering, or because some other part of the kernel write an audit
1832 * message), then write out the syscall information. In call cases,
1833 * free the names stored from getname().
1835 void __audit_syscall_exit(int success, long return_code)
1837 struct task_struct *tsk = current;
1838 struct audit_context *context;
1841 success = AUDITSC_SUCCESS;
1843 success = AUDITSC_FAILURE;
1845 context = audit_get_context(tsk, success, return_code);
1849 if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1850 audit_log_exit(context, tsk);
1852 context->in_syscall = 0;
1853 context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1855 if (!list_empty(&context->killed_trees))
1856 audit_kill_trees(&context->killed_trees);
1858 if (context->previous) {
1859 struct audit_context *new_context = context->previous;
1860 context->previous = NULL;
1861 audit_free_context(context);
1862 tsk->audit_context = new_context;
1864 audit_free_names(context);
1865 unroll_tree_refs(context, NULL, 0);
1866 audit_free_aux(context);
1867 context->aux = NULL;
1868 context->aux_pids = NULL;
1869 context->target_pid = 0;
1870 context->target_sid = 0;
1871 context->sockaddr_len = 0;
1873 context->fds[0] = -1;
1874 if (context->state != AUDIT_RECORD_CONTEXT) {
1875 kfree(context->filterkey);
1876 context->filterkey = NULL;
1878 tsk->audit_context = context;
1882 static inline void handle_one(const struct inode *inode)
1884 #ifdef CONFIG_AUDIT_TREE
1885 struct audit_context *context;
1886 struct audit_tree_refs *p;
1887 struct audit_chunk *chunk;
1889 if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1891 context = current->audit_context;
1893 count = context->tree_count;
1895 chunk = audit_tree_lookup(inode);
1899 if (likely(put_tree_ref(context, chunk)))
1901 if (unlikely(!grow_tree_refs(context))) {
1902 printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1903 audit_set_auditable(context);
1904 audit_put_chunk(chunk);
1905 unroll_tree_refs(context, p, count);
1908 put_tree_ref(context, chunk);
1912 static void handle_path(const struct dentry *dentry)
1914 #ifdef CONFIG_AUDIT_TREE
1915 struct audit_context *context;
1916 struct audit_tree_refs *p;
1917 const struct dentry *d, *parent;
1918 struct audit_chunk *drop;
1922 context = current->audit_context;
1924 count = context->tree_count;
1929 seq = read_seqbegin(&rename_lock);
1931 struct inode *inode = d->d_inode;
1932 if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1933 struct audit_chunk *chunk;
1934 chunk = audit_tree_lookup(inode);
1936 if (unlikely(!put_tree_ref(context, chunk))) {
1942 parent = d->d_parent;
1947 if (unlikely(read_seqretry(&rename_lock, seq) || drop)) { /* in this order */
1950 /* just a race with rename */
1951 unroll_tree_refs(context, p, count);
1954 audit_put_chunk(drop);
1955 if (grow_tree_refs(context)) {
1956 /* OK, got more space */
1957 unroll_tree_refs(context, p, count);
1962 "out of memory, audit has lost a tree reference\n");
1963 unroll_tree_refs(context, p, count);
1964 audit_set_auditable(context);
1971 static struct audit_names *audit_alloc_name(struct audit_context *context)
1973 struct audit_names *aname;
1975 if (context->name_count < AUDIT_NAMES) {
1976 aname = &context->preallocated_names[context->name_count];
1977 memset(aname, 0, sizeof(*aname));
1979 aname = kzalloc(sizeof(*aname), GFP_NOFS);
1982 aname->should_free = true;
1985 aname->ino = (unsigned long)-1;
1986 list_add_tail(&aname->list, &context->names_list);
1988 context->name_count++;
1990 context->ino_count++;
1996 * audit_getname - add a name to the list
1997 * @name: name to add
1999 * Add a name to the list of audit names for this context.
2000 * Called from fs/namei.c:getname().
2002 void __audit_getname(const char *name)
2004 struct audit_context *context = current->audit_context;
2005 struct audit_names *n;
2007 if (!context->in_syscall) {
2008 #if AUDIT_DEBUG == 2
2009 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
2010 __FILE__, __LINE__, context->serial, name);
2016 n = audit_alloc_name(context);
2021 n->name_len = AUDIT_NAME_FULL;
2024 if (!context->pwd.dentry)
2025 get_fs_pwd(current->fs, &context->pwd);
2028 /* audit_putname - intercept a putname request
2029 * @name: name to intercept and delay for putname
2031 * If we have stored the name from getname in the audit context,
2032 * then we delay the putname until syscall exit.
2033 * Called from include/linux/fs.h:putname().
2035 void audit_putname(const char *name)
2037 struct audit_context *context = current->audit_context;
2040 if (!context->in_syscall) {
2041 #if AUDIT_DEBUG == 2
2042 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
2043 __FILE__, __LINE__, context->serial, name);
2044 if (context->name_count) {
2045 struct audit_names *n;
2048 list_for_each_entry(n, &context->names_list, list)
2049 printk(KERN_ERR "name[%d] = %p = %s\n", i,
2050 n->name, n->name ?: "(null)");
2057 ++context->put_count;
2058 if (context->put_count > context->name_count) {
2059 printk(KERN_ERR "%s:%d(:%d): major=%d"
2060 " in_syscall=%d putname(%p) name_count=%d"
2063 context->serial, context->major,
2064 context->in_syscall, name, context->name_count,
2065 context->put_count);
2072 static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
2074 struct cpu_vfs_cap_data caps;
2080 rc = get_vfs_caps_from_disk(dentry, &caps);
2084 name->fcap.permitted = caps.permitted;
2085 name->fcap.inheritable = caps.inheritable;
2086 name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2087 name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2093 /* Copy inode data into an audit_names. */
2094 static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
2095 const struct inode *inode)
2097 name->ino = inode->i_ino;
2098 name->dev = inode->i_sb->s_dev;
2099 name->mode = inode->i_mode;
2100 name->uid = inode->i_uid;
2101 name->gid = inode->i_gid;
2102 name->rdev = inode->i_rdev;
2103 security_inode_getsecid(inode, &name->osid);
2104 audit_copy_fcaps(name, dentry);
2108 * audit_inode - store the inode and device from a lookup
2109 * @name: name being audited
2110 * @dentry: dentry being audited
2112 * Called from fs/namei.c:path_lookup().
2114 void __audit_inode(const char *name, const struct dentry *dentry)
2116 struct audit_context *context = current->audit_context;
2117 const struct inode *inode = dentry->d_inode;
2118 struct audit_names *n;
2120 if (!context->in_syscall)
2123 list_for_each_entry_reverse(n, &context->names_list, list) {
2124 if (n->name && (n->name == name))
2128 /* unable to find the name from a previous getname() */
2129 n = audit_alloc_name(context);
2133 handle_path(dentry);
2134 audit_copy_inode(n, dentry, inode);
2138 * audit_inode_child - collect inode info for created/removed objects
2139 * @dentry: dentry being audited
2140 * @parent: inode of dentry parent
2142 * For syscalls that create or remove filesystem objects, audit_inode
2143 * can only collect information for the filesystem object's parent.
2144 * This call updates the audit context with the child's information.
2145 * Syscalls that create a new filesystem object must be hooked after
2146 * the object is created. Syscalls that remove a filesystem object
2147 * must be hooked prior, in order to capture the target inode during
2148 * unsuccessful attempts.
2150 void __audit_inode_child(const struct dentry *dentry,
2151 const struct inode *parent)
2153 struct audit_context *context = current->audit_context;
2154 const char *found_parent = NULL, *found_child = NULL;
2155 const struct inode *inode = dentry->d_inode;
2156 const char *dname = dentry->d_name.name;
2157 struct audit_names *n;
2160 if (!context->in_syscall)
2166 /* parent is more likely, look for it first */
2167 list_for_each_entry(n, &context->names_list, list) {
2171 if (n->ino == parent->i_ino &&
2172 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2173 n->name_len = dirlen; /* update parent data in place */
2174 found_parent = n->name;
2179 /* no matching parent, look for matching child */
2180 list_for_each_entry(n, &context->names_list, list) {
2184 /* strcmp() is the more likely scenario */
2185 if (!strcmp(dname, n->name) ||
2186 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2188 audit_copy_inode(n, NULL, inode);
2190 n->ino = (unsigned long)-1;
2191 found_child = n->name;
2197 if (!found_parent) {
2198 n = audit_alloc_name(context);
2201 audit_copy_inode(n, NULL, parent);
2205 n = audit_alloc_name(context);
2209 /* Re-use the name belonging to the slot for a matching parent
2210 * directory. All names for this context are relinquished in
2211 * audit_free_names() */
2213 n->name = found_parent;
2214 n->name_len = AUDIT_NAME_FULL;
2215 /* don't call __putname() */
2216 n->name_put = false;
2220 audit_copy_inode(n, NULL, inode);
2223 EXPORT_SYMBOL_GPL(__audit_inode_child);
2226 * auditsc_get_stamp - get local copies of audit_context values
2227 * @ctx: audit_context for the task
2228 * @t: timespec to store time recorded in the audit_context
2229 * @serial: serial value that is recorded in the audit_context
2231 * Also sets the context as auditable.
2233 int auditsc_get_stamp(struct audit_context *ctx,
2234 struct timespec *t, unsigned int *serial)
2236 if (!ctx->in_syscall)
2239 ctx->serial = audit_serial();
2240 t->tv_sec = ctx->ctime.tv_sec;
2241 t->tv_nsec = ctx->ctime.tv_nsec;
2242 *serial = ctx->serial;
2245 ctx->current_state = AUDIT_RECORD_CONTEXT;
2250 /* global counter which is incremented every time something logs in */
2251 static atomic_t session_id = ATOMIC_INIT(0);
2254 * audit_set_loginuid - set current task's audit_context loginuid
2255 * @loginuid: loginuid value
2259 * Called (set) from fs/proc/base.c::proc_loginuid_write().
2261 int audit_set_loginuid(uid_t loginuid)
2263 struct task_struct *task = current;
2264 struct audit_context *context = task->audit_context;
2265 unsigned int sessionid;
2267 #ifdef CONFIG_AUDIT_LOGINUID_IMMUTABLE
2268 if (task->loginuid != -1)
2270 #else /* CONFIG_AUDIT_LOGINUID_IMMUTABLE */
2271 if (!capable(CAP_AUDIT_CONTROL))
2273 #endif /* CONFIG_AUDIT_LOGINUID_IMMUTABLE */
2275 sessionid = atomic_inc_return(&session_id);
2276 if (context && context->in_syscall) {
2277 struct audit_buffer *ab;
2279 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2281 audit_log_format(ab, "login pid=%d uid=%u "
2282 "old auid=%u new auid=%u"
2283 " old ses=%u new ses=%u",
2284 task->pid, task_uid(task),
2285 task->loginuid, loginuid,
2286 task->sessionid, sessionid);
2290 task->sessionid = sessionid;
2291 task->loginuid = loginuid;
2296 * __audit_mq_open - record audit data for a POSIX MQ open
2299 * @attr: queue attributes
2302 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2304 struct audit_context *context = current->audit_context;
2307 memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2309 memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2311 context->mq_open.oflag = oflag;
2312 context->mq_open.mode = mode;
2314 context->type = AUDIT_MQ_OPEN;
2318 * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2319 * @mqdes: MQ descriptor
2320 * @msg_len: Message length
2321 * @msg_prio: Message priority
2322 * @abs_timeout: Message timeout in absolute time
2325 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2326 const struct timespec *abs_timeout)
2328 struct audit_context *context = current->audit_context;
2329 struct timespec *p = &context->mq_sendrecv.abs_timeout;
2332 memcpy(p, abs_timeout, sizeof(struct timespec));
2334 memset(p, 0, sizeof(struct timespec));
2336 context->mq_sendrecv.mqdes = mqdes;
2337 context->mq_sendrecv.msg_len = msg_len;
2338 context->mq_sendrecv.msg_prio = msg_prio;
2340 context->type = AUDIT_MQ_SENDRECV;
2344 * __audit_mq_notify - record audit data for a POSIX MQ notify
2345 * @mqdes: MQ descriptor
2346 * @notification: Notification event
2350 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2352 struct audit_context *context = current->audit_context;
2355 context->mq_notify.sigev_signo = notification->sigev_signo;
2357 context->mq_notify.sigev_signo = 0;
2359 context->mq_notify.mqdes = mqdes;
2360 context->type = AUDIT_MQ_NOTIFY;
2364 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2365 * @mqdes: MQ descriptor
2369 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2371 struct audit_context *context = current->audit_context;
2372 context->mq_getsetattr.mqdes = mqdes;
2373 context->mq_getsetattr.mqstat = *mqstat;
2374 context->type = AUDIT_MQ_GETSETATTR;
2378 * audit_ipc_obj - record audit data for ipc object
2379 * @ipcp: ipc permissions
2382 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2384 struct audit_context *context = current->audit_context;
2385 context->ipc.uid = ipcp->uid;
2386 context->ipc.gid = ipcp->gid;
2387 context->ipc.mode = ipcp->mode;
2388 context->ipc.has_perm = 0;
2389 security_ipc_getsecid(ipcp, &context->ipc.osid);
2390 context->type = AUDIT_IPC;
2394 * audit_ipc_set_perm - record audit data for new ipc permissions
2395 * @qbytes: msgq bytes
2396 * @uid: msgq user id
2397 * @gid: msgq group id
2398 * @mode: msgq mode (permissions)
2400 * Called only after audit_ipc_obj().
2402 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2404 struct audit_context *context = current->audit_context;
2406 context->ipc.qbytes = qbytes;
2407 context->ipc.perm_uid = uid;
2408 context->ipc.perm_gid = gid;
2409 context->ipc.perm_mode = mode;
2410 context->ipc.has_perm = 1;
2413 int __audit_bprm(struct linux_binprm *bprm)
2415 struct audit_aux_data_execve *ax;
2416 struct audit_context *context = current->audit_context;
2418 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2422 ax->argc = bprm->argc;
2423 ax->envc = bprm->envc;
2425 ax->d.type = AUDIT_EXECVE;
2426 ax->d.next = context->aux;
2427 context->aux = (void *)ax;
2433 * audit_socketcall - record audit data for sys_socketcall
2434 * @nargs: number of args
2438 void __audit_socketcall(int nargs, unsigned long *args)
2440 struct audit_context *context = current->audit_context;
2442 context->type = AUDIT_SOCKETCALL;
2443 context->socketcall.nargs = nargs;
2444 memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2448 * __audit_fd_pair - record audit data for pipe and socketpair
2449 * @fd1: the first file descriptor
2450 * @fd2: the second file descriptor
2453 void __audit_fd_pair(int fd1, int fd2)
2455 struct audit_context *context = current->audit_context;
2456 context->fds[0] = fd1;
2457 context->fds[1] = fd2;
2461 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2462 * @len: data length in user space
2463 * @a: data address in kernel space
2465 * Returns 0 for success or NULL context or < 0 on error.
2467 int __audit_sockaddr(int len, void *a)
2469 struct audit_context *context = current->audit_context;
2471 if (!context->sockaddr) {
2472 void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2475 context->sockaddr = p;
2478 context->sockaddr_len = len;
2479 memcpy(context->sockaddr, a, len);
2483 void __audit_ptrace(struct task_struct *t)
2485 struct audit_context *context = current->audit_context;
2487 context->target_pid = t->pid;
2488 context->target_auid = audit_get_loginuid(t);
2489 context->target_uid = task_uid(t);
2490 context->target_sessionid = audit_get_sessionid(t);
2491 security_task_getsecid(t, &context->target_sid);
2492 memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2496 * audit_signal_info - record signal info for shutting down audit subsystem
2497 * @sig: signal value
2498 * @t: task being signaled
2500 * If the audit subsystem is being terminated, record the task (pid)
2501 * and uid that is doing that.
2503 int __audit_signal_info(int sig, struct task_struct *t)
2505 struct audit_aux_data_pids *axp;
2506 struct task_struct *tsk = current;
2507 struct audit_context *ctx = tsk->audit_context;
2508 uid_t uid = current_uid(), t_uid = task_uid(t);
2510 if (audit_pid && t->tgid == audit_pid) {
2511 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2512 audit_sig_pid = tsk->pid;
2513 if (tsk->loginuid != -1)
2514 audit_sig_uid = tsk->loginuid;
2516 audit_sig_uid = uid;
2517 security_task_getsecid(tsk, &audit_sig_sid);
2519 if (!audit_signals || audit_dummy_context())
2523 /* optimize the common case by putting first signal recipient directly
2524 * in audit_context */
2525 if (!ctx->target_pid) {
2526 ctx->target_pid = t->tgid;
2527 ctx->target_auid = audit_get_loginuid(t);
2528 ctx->target_uid = t_uid;
2529 ctx->target_sessionid = audit_get_sessionid(t);
2530 security_task_getsecid(t, &ctx->target_sid);
2531 memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2535 axp = (void *)ctx->aux_pids;
2536 if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2537 axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2541 axp->d.type = AUDIT_OBJ_PID;
2542 axp->d.next = ctx->aux_pids;
2543 ctx->aux_pids = (void *)axp;
2545 BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2547 axp->target_pid[axp->pid_count] = t->tgid;
2548 axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2549 axp->target_uid[axp->pid_count] = t_uid;
2550 axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2551 security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2552 memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2559 * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2560 * @bprm: pointer to the bprm being processed
2561 * @new: the proposed new credentials
2562 * @old: the old credentials
2564 * Simply check if the proc already has the caps given by the file and if not
2565 * store the priv escalation info for later auditing at the end of the syscall
2569 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2570 const struct cred *new, const struct cred *old)
2572 struct audit_aux_data_bprm_fcaps *ax;
2573 struct audit_context *context = current->audit_context;
2574 struct cpu_vfs_cap_data vcaps;
2575 struct dentry *dentry;
2577 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2581 ax->d.type = AUDIT_BPRM_FCAPS;
2582 ax->d.next = context->aux;
2583 context->aux = (void *)ax;
2585 dentry = dget(bprm->file->f_dentry);
2586 get_vfs_caps_from_disk(dentry, &vcaps);
2589 ax->fcap.permitted = vcaps.permitted;
2590 ax->fcap.inheritable = vcaps.inheritable;
2591 ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2592 ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2594 ax->old_pcap.permitted = old->cap_permitted;
2595 ax->old_pcap.inheritable = old->cap_inheritable;
2596 ax->old_pcap.effective = old->cap_effective;
2598 ax->new_pcap.permitted = new->cap_permitted;
2599 ax->new_pcap.inheritable = new->cap_inheritable;
2600 ax->new_pcap.effective = new->cap_effective;
2605 * __audit_log_capset - store information about the arguments to the capset syscall
2606 * @pid: target pid of the capset call
2607 * @new: the new credentials
2608 * @old: the old (current) credentials
2610 * Record the aguments userspace sent to sys_capset for later printing by the
2611 * audit system if applicable
2613 void __audit_log_capset(pid_t pid,
2614 const struct cred *new, const struct cred *old)
2616 struct audit_context *context = current->audit_context;
2617 context->capset.pid = pid;
2618 context->capset.cap.effective = new->cap_effective;
2619 context->capset.cap.inheritable = new->cap_effective;
2620 context->capset.cap.permitted = new->cap_permitted;
2621 context->type = AUDIT_CAPSET;
2624 void __audit_mmap_fd(int fd, int flags)
2626 struct audit_context *context = current->audit_context;
2627 context->mmap.fd = fd;
2628 context->mmap.flags = flags;
2629 context->type = AUDIT_MMAP;
2632 static void audit_log_abend(struct audit_buffer *ab, char *reason, long signr)
2636 unsigned int sessionid;
2638 auid = audit_get_loginuid(current);
2639 sessionid = audit_get_sessionid(current);
2640 current_uid_gid(&uid, &gid);
2642 audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2643 auid, uid, gid, sessionid);
2644 audit_log_task_context(ab);
2645 audit_log_format(ab, " pid=%d comm=", current->pid);
2646 audit_log_untrustedstring(ab, current->comm);
2647 audit_log_format(ab, " reason=");
2648 audit_log_string(ab, reason);
2649 audit_log_format(ab, " sig=%ld", signr);
2652 * audit_core_dumps - record information about processes that end abnormally
2653 * @signr: signal value
2655 * If a process ends with a core dump, something fishy is going on and we
2656 * should record the event for investigation.
2658 void audit_core_dumps(long signr)
2660 struct audit_buffer *ab;
2665 if (signr == SIGQUIT) /* don't care for those */
2668 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2669 audit_log_abend(ab, "memory violation", signr);
2673 void __audit_seccomp(unsigned long syscall)
2675 struct audit_buffer *ab;
2677 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2678 audit_log_abend(ab, "seccomp", SIGKILL);
2679 audit_log_format(ab, " syscall=%ld", syscall);
2683 struct list_head *audit_killed_trees(void)
2685 struct audit_context *ctx = current->audit_context;
2686 if (likely(!ctx || !ctx->in_syscall))
2688 return &ctx->killed_trees;