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 <asm/atomic.h>
49 #include <linux/namei.h>
51 #include <linux/module.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 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
74 * for saving names from getname(). */
75 #define AUDIT_NAMES 20
77 /* Indicates that audit should log the full pathname. */
78 #define AUDIT_NAME_FULL -1
80 /* no execve audit message should be longer than this (userspace limits) */
81 #define MAX_EXECVE_AUDIT_LEN 7500
83 /* number of audit rules */
86 /* determines whether we collect data for signals sent */
89 struct audit_cap_data {
90 kernel_cap_t permitted;
91 kernel_cap_t inheritable;
93 unsigned int fE; /* effective bit of a file capability */
94 kernel_cap_t effective; /* effective set of a process */
98 /* When fs/namei.c:getname() is called, we store the pointer in name and
99 * we don't let putname() free it (instead we free all of the saved
100 * pointers at syscall exit time).
102 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
105 int name_len; /* number of name's characters to log */
106 unsigned name_put; /* call __putname() for this name */
114 struct audit_cap_data fcap;
115 unsigned int fcap_ver;
118 struct audit_aux_data {
119 struct audit_aux_data *next;
123 #define AUDIT_AUX_IPCPERM 0
125 /* Number of target pids per aux struct. */
126 #define AUDIT_AUX_PIDS 16
128 struct audit_aux_data_execve {
129 struct audit_aux_data d;
132 struct mm_struct *mm;
135 struct audit_aux_data_pids {
136 struct audit_aux_data d;
137 pid_t target_pid[AUDIT_AUX_PIDS];
138 uid_t target_auid[AUDIT_AUX_PIDS];
139 uid_t target_uid[AUDIT_AUX_PIDS];
140 unsigned int target_sessionid[AUDIT_AUX_PIDS];
141 u32 target_sid[AUDIT_AUX_PIDS];
142 char target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
146 struct audit_aux_data_bprm_fcaps {
147 struct audit_aux_data d;
148 struct audit_cap_data fcap;
149 unsigned int fcap_ver;
150 struct audit_cap_data old_pcap;
151 struct audit_cap_data new_pcap;
154 struct audit_aux_data_capset {
155 struct audit_aux_data d;
157 struct audit_cap_data cap;
160 struct audit_tree_refs {
161 struct audit_tree_refs *next;
162 struct audit_chunk *c[31];
165 /* The per-task audit context. */
166 struct audit_context {
167 int dummy; /* must be the first element */
168 int in_syscall; /* 1 if task is in a syscall */
169 enum audit_state state, current_state;
170 unsigned int serial; /* serial number for record */
171 int major; /* syscall number */
172 struct timespec ctime; /* time of syscall entry */
173 unsigned long argv[4]; /* syscall arguments */
174 long return_code;/* syscall return code */
176 int return_valid; /* return code is valid */
178 struct audit_names names[AUDIT_NAMES];
179 char * filterkey; /* key for rule that triggered record */
181 struct audit_context *previous; /* For nested syscalls */
182 struct audit_aux_data *aux;
183 struct audit_aux_data *aux_pids;
184 struct sockaddr_storage *sockaddr;
186 /* Save things to print about task_struct */
188 uid_t uid, euid, suid, fsuid;
189 gid_t gid, egid, sgid, fsgid;
190 unsigned long personality;
196 unsigned int target_sessionid;
198 char target_comm[TASK_COMM_LEN];
200 struct audit_tree_refs *trees, *first_trees;
201 struct list_head killed_trees;
219 unsigned long qbytes;
223 struct mq_attr mqstat;
232 unsigned int msg_prio;
233 struct timespec abs_timeout;
242 struct audit_cap_data cap;
257 static inline int open_arg(int flags, int mask)
259 int n = ACC_MODE(flags);
260 if (flags & (O_TRUNC | O_CREAT))
261 n |= AUDIT_PERM_WRITE;
265 static int audit_match_perm(struct audit_context *ctx, int mask)
272 switch (audit_classify_syscall(ctx->arch, n)) {
274 if ((mask & AUDIT_PERM_WRITE) &&
275 audit_match_class(AUDIT_CLASS_WRITE, n))
277 if ((mask & AUDIT_PERM_READ) &&
278 audit_match_class(AUDIT_CLASS_READ, n))
280 if ((mask & AUDIT_PERM_ATTR) &&
281 audit_match_class(AUDIT_CLASS_CHATTR, n))
284 case 1: /* 32bit on biarch */
285 if ((mask & AUDIT_PERM_WRITE) &&
286 audit_match_class(AUDIT_CLASS_WRITE_32, n))
288 if ((mask & AUDIT_PERM_READ) &&
289 audit_match_class(AUDIT_CLASS_READ_32, n))
291 if ((mask & AUDIT_PERM_ATTR) &&
292 audit_match_class(AUDIT_CLASS_CHATTR_32, n))
296 return mask & ACC_MODE(ctx->argv[1]);
298 return mask & ACC_MODE(ctx->argv[2]);
299 case 4: /* socketcall */
300 return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
302 return mask & AUDIT_PERM_EXEC;
308 static int audit_match_filetype(struct audit_context *ctx, int which)
310 unsigned index = which & ~S_IFMT;
311 mode_t mode = which & S_IFMT;
316 if (index >= ctx->name_count)
318 if (ctx->names[index].ino == -1)
320 if ((ctx->names[index].mode ^ mode) & S_IFMT)
326 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
327 * ->first_trees points to its beginning, ->trees - to the current end of data.
328 * ->tree_count is the number of free entries in array pointed to by ->trees.
329 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
330 * "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
331 * it's going to remain 1-element for almost any setup) until we free context itself.
332 * References in it _are_ dropped - at the same time we free/drop aux stuff.
335 #ifdef CONFIG_AUDIT_TREE
336 static void audit_set_auditable(struct audit_context *ctx)
340 ctx->current_state = AUDIT_RECORD_CONTEXT;
344 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
346 struct audit_tree_refs *p = ctx->trees;
347 int left = ctx->tree_count;
349 p->c[--left] = chunk;
350 ctx->tree_count = left;
359 ctx->tree_count = 30;
365 static int grow_tree_refs(struct audit_context *ctx)
367 struct audit_tree_refs *p = ctx->trees;
368 ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
374 p->next = ctx->trees;
376 ctx->first_trees = ctx->trees;
377 ctx->tree_count = 31;
382 static void unroll_tree_refs(struct audit_context *ctx,
383 struct audit_tree_refs *p, int count)
385 #ifdef CONFIG_AUDIT_TREE
386 struct audit_tree_refs *q;
389 /* we started with empty chain */
390 p = ctx->first_trees;
392 /* if the very first allocation has failed, nothing to do */
397 for (q = p; q != ctx->trees; q = q->next, n = 31) {
399 audit_put_chunk(q->c[n]);
403 while (n-- > ctx->tree_count) {
404 audit_put_chunk(q->c[n]);
408 ctx->tree_count = count;
412 static void free_tree_refs(struct audit_context *ctx)
414 struct audit_tree_refs *p, *q;
415 for (p = ctx->first_trees; p; p = q) {
421 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
423 #ifdef CONFIG_AUDIT_TREE
424 struct audit_tree_refs *p;
429 for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
430 for (n = 0; n < 31; n++)
431 if (audit_tree_match(p->c[n], tree))
436 for (n = ctx->tree_count; n < 31; n++)
437 if (audit_tree_match(p->c[n], tree))
444 /* Determine if any context name data matches a rule's watch data */
445 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
448 * If task_creation is true, this is an explicit indication that we are
449 * filtering a task rule at task creation time. This and tsk == current are
450 * the only situations where tsk->cred may be accessed without an rcu read lock.
452 static int audit_filter_rules(struct task_struct *tsk,
453 struct audit_krule *rule,
454 struct audit_context *ctx,
455 struct audit_names *name,
456 enum audit_state *state,
459 const struct cred *cred;
460 int i, j, need_sid = 1;
463 cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
465 for (i = 0; i < rule->field_count; i++) {
466 struct audit_field *f = &rule->fields[i];
471 result = audit_comparator(tsk->pid, f->op, f->val);
476 ctx->ppid = sys_getppid();
477 result = audit_comparator(ctx->ppid, f->op, f->val);
481 result = audit_comparator(cred->uid, f->op, f->val);
484 result = audit_comparator(cred->euid, f->op, f->val);
487 result = audit_comparator(cred->suid, f->op, f->val);
490 result = audit_comparator(cred->fsuid, f->op, f->val);
493 result = audit_comparator(cred->gid, f->op, f->val);
496 result = audit_comparator(cred->egid, f->op, f->val);
499 result = audit_comparator(cred->sgid, f->op, f->val);
502 result = audit_comparator(cred->fsgid, f->op, f->val);
505 result = audit_comparator(tsk->personality, f->op, f->val);
509 result = audit_comparator(ctx->arch, f->op, f->val);
513 if (ctx && ctx->return_valid)
514 result = audit_comparator(ctx->return_code, f->op, f->val);
517 if (ctx && ctx->return_valid) {
519 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
521 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
526 result = audit_comparator(MAJOR(name->dev),
529 for (j = 0; j < ctx->name_count; j++) {
530 if (audit_comparator(MAJOR(ctx->names[j].dev), f->op, f->val)) {
539 result = audit_comparator(MINOR(name->dev),
542 for (j = 0; j < ctx->name_count; j++) {
543 if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
552 result = (name->ino == f->val);
554 for (j = 0; j < ctx->name_count; j++) {
555 if (audit_comparator(ctx->names[j].ino, f->op, f->val)) {
564 result = audit_watch_compare(rule->watch, name->ino, name->dev);
568 result = match_tree_refs(ctx, rule->tree);
573 result = audit_comparator(tsk->loginuid, f->op, f->val);
575 case AUDIT_SUBJ_USER:
576 case AUDIT_SUBJ_ROLE:
577 case AUDIT_SUBJ_TYPE:
580 /* NOTE: this may return negative values indicating
581 a temporary error. We simply treat this as a
582 match for now to avoid losing information that
583 may be wanted. An error message will also be
587 security_task_getsecid(tsk, &sid);
590 result = security_audit_rule_match(sid, f->type,
599 case AUDIT_OBJ_LEV_LOW:
600 case AUDIT_OBJ_LEV_HIGH:
601 /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
604 /* Find files that match */
606 result = security_audit_rule_match(
607 name->osid, f->type, f->op,
610 for (j = 0; j < ctx->name_count; j++) {
611 if (security_audit_rule_match(
620 /* Find ipc objects that match */
621 if (!ctx || ctx->type != AUDIT_IPC)
623 if (security_audit_rule_match(ctx->ipc.osid,
634 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
636 case AUDIT_FILTERKEY:
637 /* ignore this field for filtering */
641 result = audit_match_perm(ctx, f->val);
644 result = audit_match_filetype(ctx, f->val);
653 if (rule->prio <= ctx->prio)
655 if (rule->filterkey) {
656 kfree(ctx->filterkey);
657 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
659 ctx->prio = rule->prio;
661 switch (rule->action) {
662 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
663 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
668 /* At process creation time, we can determine if system-call auditing is
669 * completely disabled for this task. Since we only have the task
670 * structure at this point, we can only check uid and gid.
672 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
674 struct audit_entry *e;
675 enum audit_state state;
678 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
679 if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
681 if (state == AUDIT_RECORD_CONTEXT)
682 *key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
688 return AUDIT_BUILD_CONTEXT;
691 /* At syscall entry and exit time, this filter is called if the
692 * audit_state is not low enough that auditing cannot take place, but is
693 * also not high enough that we already know we have to write an audit
694 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
696 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
697 struct audit_context *ctx,
698 struct list_head *list)
700 struct audit_entry *e;
701 enum audit_state state;
703 if (audit_pid && tsk->tgid == audit_pid)
704 return AUDIT_DISABLED;
707 if (!list_empty(list)) {
708 int word = AUDIT_WORD(ctx->major);
709 int bit = AUDIT_BIT(ctx->major);
711 list_for_each_entry_rcu(e, list, list) {
712 if ((e->rule.mask[word] & bit) == bit &&
713 audit_filter_rules(tsk, &e->rule, ctx, NULL,
716 ctx->current_state = state;
722 return AUDIT_BUILD_CONTEXT;
725 /* At syscall exit time, this filter is called if any audit_names[] have been
726 * collected during syscall processing. We only check rules in sublists at hash
727 * buckets applicable to the inode numbers in audit_names[].
728 * Regarding audit_state, same rules apply as for audit_filter_syscall().
730 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
733 struct audit_entry *e;
734 enum audit_state state;
736 if (audit_pid && tsk->tgid == audit_pid)
740 for (i = 0; i < ctx->name_count; i++) {
741 int word = AUDIT_WORD(ctx->major);
742 int bit = AUDIT_BIT(ctx->major);
743 struct audit_names *n = &ctx->names[i];
744 int h = audit_hash_ino((u32)n->ino);
745 struct list_head *list = &audit_inode_hash[h];
747 if (list_empty(list))
750 list_for_each_entry_rcu(e, list, list) {
751 if ((e->rule.mask[word] & bit) == bit &&
752 audit_filter_rules(tsk, &e->rule, ctx, n,
755 ctx->current_state = state;
763 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
767 struct audit_context *context = tsk->audit_context;
769 if (likely(!context))
771 context->return_valid = return_valid;
774 * we need to fix up the return code in the audit logs if the actual
775 * return codes are later going to be fixed up by the arch specific
778 * This is actually a test for:
779 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
780 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
782 * but is faster than a bunch of ||
784 if (unlikely(return_code <= -ERESTARTSYS) &&
785 (return_code >= -ERESTART_RESTARTBLOCK) &&
786 (return_code != -ENOIOCTLCMD))
787 context->return_code = -EINTR;
789 context->return_code = return_code;
791 if (context->in_syscall && !context->dummy) {
792 audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
793 audit_filter_inodes(tsk, context);
796 tsk->audit_context = NULL;
800 static inline void audit_free_names(struct audit_context *context)
805 if (context->put_count + context->ino_count != context->name_count) {
806 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
807 " name_count=%d put_count=%d"
808 " ino_count=%d [NOT freeing]\n",
810 context->serial, context->major, context->in_syscall,
811 context->name_count, context->put_count,
813 for (i = 0; i < context->name_count; i++) {
814 printk(KERN_ERR "names[%d] = %p = %s\n", i,
815 context->names[i].name,
816 context->names[i].name ?: "(null)");
823 context->put_count = 0;
824 context->ino_count = 0;
827 for (i = 0; i < context->name_count; i++) {
828 if (context->names[i].name && context->names[i].name_put)
829 __putname(context->names[i].name);
831 context->name_count = 0;
832 path_put(&context->pwd);
833 context->pwd.dentry = NULL;
834 context->pwd.mnt = NULL;
837 static inline void audit_free_aux(struct audit_context *context)
839 struct audit_aux_data *aux;
841 while ((aux = context->aux)) {
842 context->aux = aux->next;
845 while ((aux = context->aux_pids)) {
846 context->aux_pids = aux->next;
851 static inline void audit_zero_context(struct audit_context *context,
852 enum audit_state state)
854 memset(context, 0, sizeof(*context));
855 context->state = state;
856 context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
859 static inline struct audit_context *audit_alloc_context(enum audit_state state)
861 struct audit_context *context;
863 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
865 audit_zero_context(context, state);
866 INIT_LIST_HEAD(&context->killed_trees);
871 * audit_alloc - allocate an audit context block for a task
874 * Filter on the task information and allocate a per-task audit context
875 * if necessary. Doing so turns on system call auditing for the
876 * specified task. This is called from copy_process, so no lock is
879 int audit_alloc(struct task_struct *tsk)
881 struct audit_context *context;
882 enum audit_state state;
885 if (likely(!audit_ever_enabled))
886 return 0; /* Return if not auditing. */
888 state = audit_filter_task(tsk, &key);
889 if (likely(state == AUDIT_DISABLED))
892 if (!(context = audit_alloc_context(state))) {
894 audit_log_lost("out of memory in audit_alloc");
897 context->filterkey = key;
899 tsk->audit_context = context;
900 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
904 static inline void audit_free_context(struct audit_context *context)
906 struct audit_context *previous;
910 previous = context->previous;
911 if (previous || (count && count < 10)) {
913 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
914 " freeing multiple contexts (%d)\n",
915 context->serial, context->major,
916 context->name_count, count);
918 audit_free_names(context);
919 unroll_tree_refs(context, NULL, 0);
920 free_tree_refs(context);
921 audit_free_aux(context);
922 kfree(context->filterkey);
923 kfree(context->sockaddr);
928 printk(KERN_ERR "audit: freed %d contexts\n", count);
931 void audit_log_task_context(struct audit_buffer *ab)
938 security_task_getsecid(current, &sid);
942 error = security_secid_to_secctx(sid, &ctx, &len);
944 if (error != -EINVAL)
949 audit_log_format(ab, " subj=%s", ctx);
950 security_release_secctx(ctx, len);
954 audit_panic("error in audit_log_task_context");
958 EXPORT_SYMBOL(audit_log_task_context);
960 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
962 char name[sizeof(tsk->comm)];
963 struct mm_struct *mm = tsk->mm;
964 struct vm_area_struct *vma;
968 get_task_comm(name, tsk);
969 audit_log_format(ab, " comm=");
970 audit_log_untrustedstring(ab, name);
973 down_read(&mm->mmap_sem);
976 if ((vma->vm_flags & VM_EXECUTABLE) &&
978 audit_log_d_path(ab, "exe=",
979 &vma->vm_file->f_path);
984 up_read(&mm->mmap_sem);
986 audit_log_task_context(ab);
989 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
990 uid_t auid, uid_t uid, unsigned int sessionid,
993 struct audit_buffer *ab;
998 ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1002 audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
1004 if (security_secid_to_secctx(sid, &ctx, &len)) {
1005 audit_log_format(ab, " obj=(none)");
1008 audit_log_format(ab, " obj=%s", ctx);
1009 security_release_secctx(ctx, len);
1011 audit_log_format(ab, " ocomm=");
1012 audit_log_untrustedstring(ab, comm);
1019 * to_send and len_sent accounting are very loose estimates. We aren't
1020 * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1021 * within about 500 bytes (next page boundary)
1023 * why snprintf? an int is up to 12 digits long. if we just assumed when
1024 * logging that a[%d]= was going to be 16 characters long we would be wasting
1025 * space in every audit message. In one 7500 byte message we can log up to
1026 * about 1000 min size arguments. That comes down to about 50% waste of space
1027 * if we didn't do the snprintf to find out how long arg_num_len was.
1029 static int audit_log_single_execve_arg(struct audit_context *context,
1030 struct audit_buffer **ab,
1033 const char __user *p,
1036 char arg_num_len_buf[12];
1037 const char __user *tmp_p = p;
1038 /* how many digits are in arg_num? 5 is the length of ' a=""' */
1039 size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1040 size_t len, len_left, to_send;
1041 size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1042 unsigned int i, has_cntl = 0, too_long = 0;
1045 /* strnlen_user includes the null we don't want to send */
1046 len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1049 * We just created this mm, if we can't find the strings
1050 * we just copied into it something is _very_ wrong. Similar
1051 * for strings that are too long, we should not have created
1054 if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1056 send_sig(SIGKILL, current, 0);
1060 /* walk the whole argument looking for non-ascii chars */
1062 if (len_left > MAX_EXECVE_AUDIT_LEN)
1063 to_send = MAX_EXECVE_AUDIT_LEN;
1066 ret = copy_from_user(buf, tmp_p, to_send);
1068 * There is no reason for this copy to be short. We just
1069 * copied them here, and the mm hasn't been exposed to user-
1074 send_sig(SIGKILL, current, 0);
1077 buf[to_send] = '\0';
1078 has_cntl = audit_string_contains_control(buf, to_send);
1081 * hex messages get logged as 2 bytes, so we can only
1082 * send half as much in each message
1084 max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1087 len_left -= to_send;
1089 } while (len_left > 0);
1093 if (len > max_execve_audit_len)
1096 /* rewalk the argument actually logging the message */
1097 for (i = 0; len_left > 0; i++) {
1100 if (len_left > max_execve_audit_len)
1101 to_send = max_execve_audit_len;
1105 /* do we have space left to send this argument in this ab? */
1106 room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1108 room_left -= (to_send * 2);
1110 room_left -= to_send;
1111 if (room_left < 0) {
1114 *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1120 * first record needs to say how long the original string was
1121 * so we can be sure nothing was lost.
1123 if ((i == 0) && (too_long))
1124 audit_log_format(*ab, " a%d_len=%zu", arg_num,
1125 has_cntl ? 2*len : len);
1128 * normally arguments are small enough to fit and we already
1129 * filled buf above when we checked for control characters
1130 * so don't bother with another copy_from_user
1132 if (len >= max_execve_audit_len)
1133 ret = copy_from_user(buf, p, to_send);
1138 send_sig(SIGKILL, current, 0);
1141 buf[to_send] = '\0';
1143 /* actually log it */
1144 audit_log_format(*ab, " a%d", arg_num);
1146 audit_log_format(*ab, "[%d]", i);
1147 audit_log_format(*ab, "=");
1149 audit_log_n_hex(*ab, buf, to_send);
1151 audit_log_string(*ab, buf);
1154 len_left -= to_send;
1155 *len_sent += arg_num_len;
1157 *len_sent += to_send * 2;
1159 *len_sent += to_send;
1161 /* include the null we didn't log */
1165 static void audit_log_execve_info(struct audit_context *context,
1166 struct audit_buffer **ab,
1167 struct audit_aux_data_execve *axi)
1170 size_t len, len_sent = 0;
1171 const char __user *p;
1174 if (axi->mm != current->mm)
1175 return; /* execve failed, no additional info */
1177 p = (const char __user *)axi->mm->arg_start;
1179 audit_log_format(*ab, "argc=%d", axi->argc);
1182 * we need some kernel buffer to hold the userspace args. Just
1183 * allocate one big one rather than allocating one of the right size
1184 * for every single argument inside audit_log_single_execve_arg()
1185 * should be <8k allocation so should be pretty safe.
1187 buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1189 audit_panic("out of memory for argv string\n");
1193 for (i = 0; i < axi->argc; i++) {
1194 len = audit_log_single_execve_arg(context, ab, i,
1203 static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1207 audit_log_format(ab, " %s=", prefix);
1208 CAP_FOR_EACH_U32(i) {
1209 audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1213 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1215 kernel_cap_t *perm = &name->fcap.permitted;
1216 kernel_cap_t *inh = &name->fcap.inheritable;
1219 if (!cap_isclear(*perm)) {
1220 audit_log_cap(ab, "cap_fp", perm);
1223 if (!cap_isclear(*inh)) {
1224 audit_log_cap(ab, "cap_fi", inh);
1229 audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1232 static void show_special(struct audit_context *context, int *call_panic)
1234 struct audit_buffer *ab;
1237 ab = audit_log_start(context, GFP_KERNEL, context->type);
1241 switch (context->type) {
1242 case AUDIT_SOCKETCALL: {
1243 int nargs = context->socketcall.nargs;
1244 audit_log_format(ab, "nargs=%d", nargs);
1245 for (i = 0; i < nargs; i++)
1246 audit_log_format(ab, " a%d=%lx", i,
1247 context->socketcall.args[i]);
1250 u32 osid = context->ipc.osid;
1252 audit_log_format(ab, "ouid=%u ogid=%u mode=%#o",
1253 context->ipc.uid, context->ipc.gid, context->ipc.mode);
1257 if (security_secid_to_secctx(osid, &ctx, &len)) {
1258 audit_log_format(ab, " osid=%u", osid);
1261 audit_log_format(ab, " obj=%s", ctx);
1262 security_release_secctx(ctx, len);
1265 if (context->ipc.has_perm) {
1267 ab = audit_log_start(context, GFP_KERNEL,
1268 AUDIT_IPC_SET_PERM);
1269 audit_log_format(ab,
1270 "qbytes=%lx ouid=%u ogid=%u mode=%#o",
1271 context->ipc.qbytes,
1272 context->ipc.perm_uid,
1273 context->ipc.perm_gid,
1274 context->ipc.perm_mode);
1279 case AUDIT_MQ_OPEN: {
1280 audit_log_format(ab,
1281 "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
1282 "mq_msgsize=%ld mq_curmsgs=%ld",
1283 context->mq_open.oflag, context->mq_open.mode,
1284 context->mq_open.attr.mq_flags,
1285 context->mq_open.attr.mq_maxmsg,
1286 context->mq_open.attr.mq_msgsize,
1287 context->mq_open.attr.mq_curmsgs);
1289 case AUDIT_MQ_SENDRECV: {
1290 audit_log_format(ab,
1291 "mqdes=%d msg_len=%zd msg_prio=%u "
1292 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1293 context->mq_sendrecv.mqdes,
1294 context->mq_sendrecv.msg_len,
1295 context->mq_sendrecv.msg_prio,
1296 context->mq_sendrecv.abs_timeout.tv_sec,
1297 context->mq_sendrecv.abs_timeout.tv_nsec);
1299 case AUDIT_MQ_NOTIFY: {
1300 audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1301 context->mq_notify.mqdes,
1302 context->mq_notify.sigev_signo);
1304 case AUDIT_MQ_GETSETATTR: {
1305 struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1306 audit_log_format(ab,
1307 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1309 context->mq_getsetattr.mqdes,
1310 attr->mq_flags, attr->mq_maxmsg,
1311 attr->mq_msgsize, attr->mq_curmsgs);
1313 case AUDIT_CAPSET: {
1314 audit_log_format(ab, "pid=%d", context->capset.pid);
1315 audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1316 audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1317 audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1320 audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1321 context->mmap.flags);
1327 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1329 const struct cred *cred;
1330 int i, call_panic = 0;
1331 struct audit_buffer *ab;
1332 struct audit_aux_data *aux;
1335 /* tsk == current */
1336 context->pid = tsk->pid;
1338 context->ppid = sys_getppid();
1339 cred = current_cred();
1340 context->uid = cred->uid;
1341 context->gid = cred->gid;
1342 context->euid = cred->euid;
1343 context->suid = cred->suid;
1344 context->fsuid = cred->fsuid;
1345 context->egid = cred->egid;
1346 context->sgid = cred->sgid;
1347 context->fsgid = cred->fsgid;
1348 context->personality = tsk->personality;
1350 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1352 return; /* audit_panic has been called */
1353 audit_log_format(ab, "arch=%x syscall=%d",
1354 context->arch, context->major);
1355 if (context->personality != PER_LINUX)
1356 audit_log_format(ab, " per=%lx", context->personality);
1357 if (context->return_valid)
1358 audit_log_format(ab, " success=%s exit=%ld",
1359 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1360 context->return_code);
1362 spin_lock_irq(&tsk->sighand->siglock);
1363 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1364 tty = tsk->signal->tty->name;
1367 spin_unlock_irq(&tsk->sighand->siglock);
1369 audit_log_format(ab,
1370 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1371 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1372 " euid=%u suid=%u fsuid=%u"
1373 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1378 context->name_count,
1384 context->euid, context->suid, context->fsuid,
1385 context->egid, context->sgid, context->fsgid, tty,
1389 audit_log_task_info(ab, tsk);
1390 audit_log_key(ab, context->filterkey);
1393 for (aux = context->aux; aux; aux = aux->next) {
1395 ab = audit_log_start(context, GFP_KERNEL, aux->type);
1397 continue; /* audit_panic has been called */
1399 switch (aux->type) {
1401 case AUDIT_EXECVE: {
1402 struct audit_aux_data_execve *axi = (void *)aux;
1403 audit_log_execve_info(context, &ab, axi);
1406 case AUDIT_BPRM_FCAPS: {
1407 struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1408 audit_log_format(ab, "fver=%x", axs->fcap_ver);
1409 audit_log_cap(ab, "fp", &axs->fcap.permitted);
1410 audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1411 audit_log_format(ab, " fe=%d", axs->fcap.fE);
1412 audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1413 audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1414 audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1415 audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1416 audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1417 audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1425 show_special(context, &call_panic);
1427 if (context->fds[0] >= 0) {
1428 ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1430 audit_log_format(ab, "fd0=%d fd1=%d",
1431 context->fds[0], context->fds[1]);
1436 if (context->sockaddr_len) {
1437 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1439 audit_log_format(ab, "saddr=");
1440 audit_log_n_hex(ab, (void *)context->sockaddr,
1441 context->sockaddr_len);
1446 for (aux = context->aux_pids; aux; aux = aux->next) {
1447 struct audit_aux_data_pids *axs = (void *)aux;
1449 for (i = 0; i < axs->pid_count; i++)
1450 if (audit_log_pid_context(context, axs->target_pid[i],
1451 axs->target_auid[i],
1453 axs->target_sessionid[i],
1455 axs->target_comm[i]))
1459 if (context->target_pid &&
1460 audit_log_pid_context(context, context->target_pid,
1461 context->target_auid, context->target_uid,
1462 context->target_sessionid,
1463 context->target_sid, context->target_comm))
1466 if (context->pwd.dentry && context->pwd.mnt) {
1467 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1469 audit_log_d_path(ab, "cwd=", &context->pwd);
1473 for (i = 0; i < context->name_count; i++) {
1474 struct audit_names *n = &context->names[i];
1476 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1478 continue; /* audit_panic has been called */
1480 audit_log_format(ab, "item=%d", i);
1483 switch(n->name_len) {
1484 case AUDIT_NAME_FULL:
1485 /* log the full path */
1486 audit_log_format(ab, " name=");
1487 audit_log_untrustedstring(ab, n->name);
1490 /* name was specified as a relative path and the
1491 * directory component is the cwd */
1492 audit_log_d_path(ab, "name=", &context->pwd);
1495 /* log the name's directory component */
1496 audit_log_format(ab, " name=");
1497 audit_log_n_untrustedstring(ab, n->name,
1501 audit_log_format(ab, " name=(null)");
1503 if (n->ino != (unsigned long)-1) {
1504 audit_log_format(ab, " inode=%lu"
1505 " dev=%02x:%02x mode=%#o"
1506 " ouid=%u ogid=%u rdev=%02x:%02x",
1519 if (security_secid_to_secctx(
1520 n->osid, &ctx, &len)) {
1521 audit_log_format(ab, " osid=%u", n->osid);
1524 audit_log_format(ab, " obj=%s", ctx);
1525 security_release_secctx(ctx, len);
1529 audit_log_fcaps(ab, n);
1534 /* Send end of event record to help user space know we are finished */
1535 ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1539 audit_panic("error converting sid to string");
1543 * audit_free - free a per-task audit context
1544 * @tsk: task whose audit context block to free
1546 * Called from copy_process and do_exit
1548 void audit_free(struct task_struct *tsk)
1550 struct audit_context *context;
1552 context = audit_get_context(tsk, 0, 0);
1553 if (likely(!context))
1556 /* Check for system calls that do not go through the exit
1557 * function (e.g., exit_group), then free context block.
1558 * We use GFP_ATOMIC here because we might be doing this
1559 * in the context of the idle thread */
1560 /* that can happen only if we are called from do_exit() */
1561 if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1562 audit_log_exit(context, tsk);
1563 if (!list_empty(&context->killed_trees))
1564 audit_kill_trees(&context->killed_trees);
1566 audit_free_context(context);
1570 * audit_syscall_entry - fill in an audit record at syscall entry
1571 * @arch: architecture type
1572 * @major: major syscall type (function)
1573 * @a1: additional syscall register 1
1574 * @a2: additional syscall register 2
1575 * @a3: additional syscall register 3
1576 * @a4: additional syscall register 4
1578 * Fill in audit context at syscall entry. This only happens if the
1579 * audit context was created when the task was created and the state or
1580 * filters demand the audit context be built. If the state from the
1581 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1582 * then the record will be written at syscall exit time (otherwise, it
1583 * will only be written if another part of the kernel requests that it
1586 void audit_syscall_entry(int arch, int major,
1587 unsigned long a1, unsigned long a2,
1588 unsigned long a3, unsigned long a4)
1590 struct task_struct *tsk = current;
1591 struct audit_context *context = tsk->audit_context;
1592 enum audit_state state;
1594 if (unlikely(!context))
1598 * This happens only on certain architectures that make system
1599 * calls in kernel_thread via the entry.S interface, instead of
1600 * with direct calls. (If you are porting to a new
1601 * architecture, hitting this condition can indicate that you
1602 * got the _exit/_leave calls backward in entry.S.)
1606 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1608 * This also happens with vm86 emulation in a non-nested manner
1609 * (entries without exits), so this case must be caught.
1611 if (context->in_syscall) {
1612 struct audit_context *newctx;
1616 "audit(:%d) pid=%d in syscall=%d;"
1617 " entering syscall=%d\n",
1618 context->serial, tsk->pid, context->major, major);
1620 newctx = audit_alloc_context(context->state);
1622 newctx->previous = context;
1624 tsk->audit_context = newctx;
1626 /* If we can't alloc a new context, the best we
1627 * can do is to leak memory (any pending putname
1628 * will be lost). The only other alternative is
1629 * to abandon auditing. */
1630 audit_zero_context(context, context->state);
1633 BUG_ON(context->in_syscall || context->name_count);
1638 context->arch = arch;
1639 context->major = major;
1640 context->argv[0] = a1;
1641 context->argv[1] = a2;
1642 context->argv[2] = a3;
1643 context->argv[3] = a4;
1645 state = context->state;
1646 context->dummy = !audit_n_rules;
1647 if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1649 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1651 if (likely(state == AUDIT_DISABLED))
1654 context->serial = 0;
1655 context->ctime = CURRENT_TIME;
1656 context->in_syscall = 1;
1657 context->current_state = state;
1661 void audit_finish_fork(struct task_struct *child)
1663 struct audit_context *ctx = current->audit_context;
1664 struct audit_context *p = child->audit_context;
1667 if (!ctx->in_syscall || ctx->current_state != AUDIT_RECORD_CONTEXT)
1669 p->arch = ctx->arch;
1670 p->major = ctx->major;
1671 memcpy(p->argv, ctx->argv, sizeof(ctx->argv));
1672 p->ctime = ctx->ctime;
1673 p->dummy = ctx->dummy;
1674 p->in_syscall = ctx->in_syscall;
1675 p->filterkey = kstrdup(ctx->filterkey, GFP_KERNEL);
1676 p->ppid = current->pid;
1677 p->prio = ctx->prio;
1678 p->current_state = ctx->current_state;
1682 * audit_syscall_exit - deallocate audit context after a system call
1683 * @valid: success/failure flag
1684 * @return_code: syscall return value
1686 * Tear down after system call. If the audit context has been marked as
1687 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1688 * filtering, or because some other part of the kernel write an audit
1689 * message), then write out the syscall information. In call cases,
1690 * free the names stored from getname().
1692 void audit_syscall_exit(int valid, long return_code)
1694 struct task_struct *tsk = current;
1695 struct audit_context *context;
1697 context = audit_get_context(tsk, valid, return_code);
1699 if (likely(!context))
1702 if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1703 audit_log_exit(context, tsk);
1705 context->in_syscall = 0;
1706 context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1708 if (!list_empty(&context->killed_trees))
1709 audit_kill_trees(&context->killed_trees);
1711 if (context->previous) {
1712 struct audit_context *new_context = context->previous;
1713 context->previous = NULL;
1714 audit_free_context(context);
1715 tsk->audit_context = new_context;
1717 audit_free_names(context);
1718 unroll_tree_refs(context, NULL, 0);
1719 audit_free_aux(context);
1720 context->aux = NULL;
1721 context->aux_pids = NULL;
1722 context->target_pid = 0;
1723 context->target_sid = 0;
1724 context->sockaddr_len = 0;
1726 context->fds[0] = -1;
1727 if (context->state != AUDIT_RECORD_CONTEXT) {
1728 kfree(context->filterkey);
1729 context->filterkey = NULL;
1731 tsk->audit_context = context;
1735 static inline void handle_one(const struct inode *inode)
1737 #ifdef CONFIG_AUDIT_TREE
1738 struct audit_context *context;
1739 struct audit_tree_refs *p;
1740 struct audit_chunk *chunk;
1742 if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1744 context = current->audit_context;
1746 count = context->tree_count;
1748 chunk = audit_tree_lookup(inode);
1752 if (likely(put_tree_ref(context, chunk)))
1754 if (unlikely(!grow_tree_refs(context))) {
1755 printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1756 audit_set_auditable(context);
1757 audit_put_chunk(chunk);
1758 unroll_tree_refs(context, p, count);
1761 put_tree_ref(context, chunk);
1765 static void handle_path(const struct dentry *dentry)
1767 #ifdef CONFIG_AUDIT_TREE
1768 struct audit_context *context;
1769 struct audit_tree_refs *p;
1770 const struct dentry *d, *parent;
1771 struct audit_chunk *drop;
1775 context = current->audit_context;
1777 count = context->tree_count;
1782 seq = read_seqbegin(&rename_lock);
1784 struct inode *inode = d->d_inode;
1785 if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1786 struct audit_chunk *chunk;
1787 chunk = audit_tree_lookup(inode);
1789 if (unlikely(!put_tree_ref(context, chunk))) {
1795 parent = d->d_parent;
1800 if (unlikely(read_seqretry(&rename_lock, seq) || drop)) { /* in this order */
1803 /* just a race with rename */
1804 unroll_tree_refs(context, p, count);
1807 audit_put_chunk(drop);
1808 if (grow_tree_refs(context)) {
1809 /* OK, got more space */
1810 unroll_tree_refs(context, p, count);
1815 "out of memory, audit has lost a tree reference\n");
1816 unroll_tree_refs(context, p, count);
1817 audit_set_auditable(context);
1825 * audit_getname - add a name to the list
1826 * @name: name to add
1828 * Add a name to the list of audit names for this context.
1829 * Called from fs/namei.c:getname().
1831 void __audit_getname(const char *name)
1833 struct audit_context *context = current->audit_context;
1835 if (IS_ERR(name) || !name)
1838 if (!context->in_syscall) {
1839 #if AUDIT_DEBUG == 2
1840 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1841 __FILE__, __LINE__, context->serial, name);
1846 BUG_ON(context->name_count >= AUDIT_NAMES);
1847 context->names[context->name_count].name = name;
1848 context->names[context->name_count].name_len = AUDIT_NAME_FULL;
1849 context->names[context->name_count].name_put = 1;
1850 context->names[context->name_count].ino = (unsigned long)-1;
1851 context->names[context->name_count].osid = 0;
1852 ++context->name_count;
1853 if (!context->pwd.dentry)
1854 get_fs_pwd(current->fs, &context->pwd);
1857 /* audit_putname - intercept a putname request
1858 * @name: name to intercept and delay for putname
1860 * If we have stored the name from getname in the audit context,
1861 * then we delay the putname until syscall exit.
1862 * Called from include/linux/fs.h:putname().
1864 void audit_putname(const char *name)
1866 struct audit_context *context = current->audit_context;
1869 if (!context->in_syscall) {
1870 #if AUDIT_DEBUG == 2
1871 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1872 __FILE__, __LINE__, context->serial, name);
1873 if (context->name_count) {
1875 for (i = 0; i < context->name_count; i++)
1876 printk(KERN_ERR "name[%d] = %p = %s\n", i,
1877 context->names[i].name,
1878 context->names[i].name ?: "(null)");
1885 ++context->put_count;
1886 if (context->put_count > context->name_count) {
1887 printk(KERN_ERR "%s:%d(:%d): major=%d"
1888 " in_syscall=%d putname(%p) name_count=%d"
1891 context->serial, context->major,
1892 context->in_syscall, name, context->name_count,
1893 context->put_count);
1900 static int audit_inc_name_count(struct audit_context *context,
1901 const struct inode *inode)
1903 if (context->name_count >= AUDIT_NAMES) {
1905 printk(KERN_DEBUG "audit: name_count maxed, losing inode data: "
1906 "dev=%02x:%02x, inode=%lu\n",
1907 MAJOR(inode->i_sb->s_dev),
1908 MINOR(inode->i_sb->s_dev),
1912 printk(KERN_DEBUG "name_count maxed, losing inode data\n");
1915 context->name_count++;
1917 context->ino_count++;
1923 static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
1925 struct cpu_vfs_cap_data caps;
1928 memset(&name->fcap.permitted, 0, sizeof(kernel_cap_t));
1929 memset(&name->fcap.inheritable, 0, sizeof(kernel_cap_t));
1936 rc = get_vfs_caps_from_disk(dentry, &caps);
1940 name->fcap.permitted = caps.permitted;
1941 name->fcap.inheritable = caps.inheritable;
1942 name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
1943 name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
1949 /* Copy inode data into an audit_names. */
1950 static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
1951 const struct inode *inode)
1953 name->ino = inode->i_ino;
1954 name->dev = inode->i_sb->s_dev;
1955 name->mode = inode->i_mode;
1956 name->uid = inode->i_uid;
1957 name->gid = inode->i_gid;
1958 name->rdev = inode->i_rdev;
1959 security_inode_getsecid(inode, &name->osid);
1960 audit_copy_fcaps(name, dentry);
1964 * audit_inode - store the inode and device from a lookup
1965 * @name: name being audited
1966 * @dentry: dentry being audited
1968 * Called from fs/namei.c:path_lookup().
1970 void __audit_inode(const char *name, const struct dentry *dentry)
1973 struct audit_context *context = current->audit_context;
1974 const struct inode *inode = dentry->d_inode;
1976 if (!context->in_syscall)
1978 if (context->name_count
1979 && context->names[context->name_count-1].name
1980 && context->names[context->name_count-1].name == name)
1981 idx = context->name_count - 1;
1982 else if (context->name_count > 1
1983 && context->names[context->name_count-2].name
1984 && context->names[context->name_count-2].name == name)
1985 idx = context->name_count - 2;
1987 /* FIXME: how much do we care about inodes that have no
1988 * associated name? */
1989 if (audit_inc_name_count(context, inode))
1991 idx = context->name_count - 1;
1992 context->names[idx].name = NULL;
1994 handle_path(dentry);
1995 audit_copy_inode(&context->names[idx], dentry, inode);
1999 * audit_inode_child - collect inode info for created/removed objects
2000 * @dentry: dentry being audited
2001 * @parent: inode of dentry parent
2003 * For syscalls that create or remove filesystem objects, audit_inode
2004 * can only collect information for the filesystem object's parent.
2005 * This call updates the audit context with the child's information.
2006 * Syscalls that create a new filesystem object must be hooked after
2007 * the object is created. Syscalls that remove a filesystem object
2008 * must be hooked prior, in order to capture the target inode during
2009 * unsuccessful attempts.
2011 void __audit_inode_child(const struct dentry *dentry,
2012 const struct inode *parent)
2015 struct audit_context *context = current->audit_context;
2016 const char *found_parent = NULL, *found_child = NULL;
2017 const struct inode *inode = dentry->d_inode;
2018 const char *dname = dentry->d_name.name;
2021 if (!context->in_syscall)
2027 /* parent is more likely, look for it first */
2028 for (idx = 0; idx < context->name_count; idx++) {
2029 struct audit_names *n = &context->names[idx];
2034 if (n->ino == parent->i_ino &&
2035 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2036 n->name_len = dirlen; /* update parent data in place */
2037 found_parent = n->name;
2042 /* no matching parent, look for matching child */
2043 for (idx = 0; idx < context->name_count; idx++) {
2044 struct audit_names *n = &context->names[idx];
2049 /* strcmp() is the more likely scenario */
2050 if (!strcmp(dname, n->name) ||
2051 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2053 audit_copy_inode(n, NULL, inode);
2055 n->ino = (unsigned long)-1;
2056 found_child = n->name;
2062 if (!found_parent) {
2063 if (audit_inc_name_count(context, parent))
2065 idx = context->name_count - 1;
2066 context->names[idx].name = NULL;
2067 audit_copy_inode(&context->names[idx], NULL, parent);
2071 if (audit_inc_name_count(context, inode))
2073 idx = context->name_count - 1;
2075 /* Re-use the name belonging to the slot for a matching parent
2076 * directory. All names for this context are relinquished in
2077 * audit_free_names() */
2079 context->names[idx].name = found_parent;
2080 context->names[idx].name_len = AUDIT_NAME_FULL;
2081 /* don't call __putname() */
2082 context->names[idx].name_put = 0;
2084 context->names[idx].name = NULL;
2088 audit_copy_inode(&context->names[idx], NULL, inode);
2090 context->names[idx].ino = (unsigned long)-1;
2093 EXPORT_SYMBOL_GPL(__audit_inode_child);
2096 * auditsc_get_stamp - get local copies of audit_context values
2097 * @ctx: audit_context for the task
2098 * @t: timespec to store time recorded in the audit_context
2099 * @serial: serial value that is recorded in the audit_context
2101 * Also sets the context as auditable.
2103 int auditsc_get_stamp(struct audit_context *ctx,
2104 struct timespec *t, unsigned int *serial)
2106 if (!ctx->in_syscall)
2109 ctx->serial = audit_serial();
2110 t->tv_sec = ctx->ctime.tv_sec;
2111 t->tv_nsec = ctx->ctime.tv_nsec;
2112 *serial = ctx->serial;
2115 ctx->current_state = AUDIT_RECORD_CONTEXT;
2120 /* global counter which is incremented every time something logs in */
2121 static atomic_t session_id = ATOMIC_INIT(0);
2124 * audit_set_loginuid - set a task's audit_context loginuid
2125 * @task: task whose audit context is being modified
2126 * @loginuid: loginuid value
2130 * Called (set) from fs/proc/base.c::proc_loginuid_write().
2132 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
2134 unsigned int sessionid = atomic_inc_return(&session_id);
2135 struct audit_context *context = task->audit_context;
2137 if (context && context->in_syscall) {
2138 struct audit_buffer *ab;
2140 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2142 audit_log_format(ab, "login pid=%d uid=%u "
2143 "old auid=%u new auid=%u"
2144 " old ses=%u new ses=%u",
2145 task->pid, task_uid(task),
2146 task->loginuid, loginuid,
2147 task->sessionid, sessionid);
2151 task->sessionid = sessionid;
2152 task->loginuid = loginuid;
2157 * __audit_mq_open - record audit data for a POSIX MQ open
2160 * @attr: queue attributes
2163 void __audit_mq_open(int oflag, mode_t mode, struct mq_attr *attr)
2165 struct audit_context *context = current->audit_context;
2168 memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2170 memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2172 context->mq_open.oflag = oflag;
2173 context->mq_open.mode = mode;
2175 context->type = AUDIT_MQ_OPEN;
2179 * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2180 * @mqdes: MQ descriptor
2181 * @msg_len: Message length
2182 * @msg_prio: Message priority
2183 * @abs_timeout: Message timeout in absolute time
2186 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2187 const struct timespec *abs_timeout)
2189 struct audit_context *context = current->audit_context;
2190 struct timespec *p = &context->mq_sendrecv.abs_timeout;
2193 memcpy(p, abs_timeout, sizeof(struct timespec));
2195 memset(p, 0, sizeof(struct timespec));
2197 context->mq_sendrecv.mqdes = mqdes;
2198 context->mq_sendrecv.msg_len = msg_len;
2199 context->mq_sendrecv.msg_prio = msg_prio;
2201 context->type = AUDIT_MQ_SENDRECV;
2205 * __audit_mq_notify - record audit data for a POSIX MQ notify
2206 * @mqdes: MQ descriptor
2207 * @notification: Notification event
2211 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2213 struct audit_context *context = current->audit_context;
2216 context->mq_notify.sigev_signo = notification->sigev_signo;
2218 context->mq_notify.sigev_signo = 0;
2220 context->mq_notify.mqdes = mqdes;
2221 context->type = AUDIT_MQ_NOTIFY;
2225 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2226 * @mqdes: MQ descriptor
2230 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2232 struct audit_context *context = current->audit_context;
2233 context->mq_getsetattr.mqdes = mqdes;
2234 context->mq_getsetattr.mqstat = *mqstat;
2235 context->type = AUDIT_MQ_GETSETATTR;
2239 * audit_ipc_obj - record audit data for ipc object
2240 * @ipcp: ipc permissions
2243 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2245 struct audit_context *context = current->audit_context;
2246 context->ipc.uid = ipcp->uid;
2247 context->ipc.gid = ipcp->gid;
2248 context->ipc.mode = ipcp->mode;
2249 context->ipc.has_perm = 0;
2250 security_ipc_getsecid(ipcp, &context->ipc.osid);
2251 context->type = AUDIT_IPC;
2255 * audit_ipc_set_perm - record audit data for new ipc permissions
2256 * @qbytes: msgq bytes
2257 * @uid: msgq user id
2258 * @gid: msgq group id
2259 * @mode: msgq mode (permissions)
2261 * Called only after audit_ipc_obj().
2263 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
2265 struct audit_context *context = current->audit_context;
2267 context->ipc.qbytes = qbytes;
2268 context->ipc.perm_uid = uid;
2269 context->ipc.perm_gid = gid;
2270 context->ipc.perm_mode = mode;
2271 context->ipc.has_perm = 1;
2274 int audit_bprm(struct linux_binprm *bprm)
2276 struct audit_aux_data_execve *ax;
2277 struct audit_context *context = current->audit_context;
2279 if (likely(!audit_enabled || !context || context->dummy))
2282 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2286 ax->argc = bprm->argc;
2287 ax->envc = bprm->envc;
2289 ax->d.type = AUDIT_EXECVE;
2290 ax->d.next = context->aux;
2291 context->aux = (void *)ax;
2297 * audit_socketcall - record audit data for sys_socketcall
2298 * @nargs: number of args
2302 void audit_socketcall(int nargs, unsigned long *args)
2304 struct audit_context *context = current->audit_context;
2306 if (likely(!context || context->dummy))
2309 context->type = AUDIT_SOCKETCALL;
2310 context->socketcall.nargs = nargs;
2311 memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2315 * __audit_fd_pair - record audit data for pipe and socketpair
2316 * @fd1: the first file descriptor
2317 * @fd2: the second file descriptor
2320 void __audit_fd_pair(int fd1, int fd2)
2322 struct audit_context *context = current->audit_context;
2323 context->fds[0] = fd1;
2324 context->fds[1] = fd2;
2328 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2329 * @len: data length in user space
2330 * @a: data address in kernel space
2332 * Returns 0 for success or NULL context or < 0 on error.
2334 int audit_sockaddr(int len, void *a)
2336 struct audit_context *context = current->audit_context;
2338 if (likely(!context || context->dummy))
2341 if (!context->sockaddr) {
2342 void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2345 context->sockaddr = p;
2348 context->sockaddr_len = len;
2349 memcpy(context->sockaddr, a, len);
2353 void __audit_ptrace(struct task_struct *t)
2355 struct audit_context *context = current->audit_context;
2357 context->target_pid = t->pid;
2358 context->target_auid = audit_get_loginuid(t);
2359 context->target_uid = task_uid(t);
2360 context->target_sessionid = audit_get_sessionid(t);
2361 security_task_getsecid(t, &context->target_sid);
2362 memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2366 * audit_signal_info - record signal info for shutting down audit subsystem
2367 * @sig: signal value
2368 * @t: task being signaled
2370 * If the audit subsystem is being terminated, record the task (pid)
2371 * and uid that is doing that.
2373 int __audit_signal_info(int sig, struct task_struct *t)
2375 struct audit_aux_data_pids *axp;
2376 struct task_struct *tsk = current;
2377 struct audit_context *ctx = tsk->audit_context;
2378 uid_t uid = current_uid(), t_uid = task_uid(t);
2380 if (audit_pid && t->tgid == audit_pid) {
2381 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2382 audit_sig_pid = tsk->pid;
2383 if (tsk->loginuid != -1)
2384 audit_sig_uid = tsk->loginuid;
2386 audit_sig_uid = uid;
2387 security_task_getsecid(tsk, &audit_sig_sid);
2389 if (!audit_signals || audit_dummy_context())
2393 /* optimize the common case by putting first signal recipient directly
2394 * in audit_context */
2395 if (!ctx->target_pid) {
2396 ctx->target_pid = t->tgid;
2397 ctx->target_auid = audit_get_loginuid(t);
2398 ctx->target_uid = t_uid;
2399 ctx->target_sessionid = audit_get_sessionid(t);
2400 security_task_getsecid(t, &ctx->target_sid);
2401 memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2405 axp = (void *)ctx->aux_pids;
2406 if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2407 axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2411 axp->d.type = AUDIT_OBJ_PID;
2412 axp->d.next = ctx->aux_pids;
2413 ctx->aux_pids = (void *)axp;
2415 BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2417 axp->target_pid[axp->pid_count] = t->tgid;
2418 axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2419 axp->target_uid[axp->pid_count] = t_uid;
2420 axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2421 security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2422 memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2429 * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2430 * @bprm: pointer to the bprm being processed
2431 * @new: the proposed new credentials
2432 * @old: the old credentials
2434 * Simply check if the proc already has the caps given by the file and if not
2435 * store the priv escalation info for later auditing at the end of the syscall
2439 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2440 const struct cred *new, const struct cred *old)
2442 struct audit_aux_data_bprm_fcaps *ax;
2443 struct audit_context *context = current->audit_context;
2444 struct cpu_vfs_cap_data vcaps;
2445 struct dentry *dentry;
2447 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2451 ax->d.type = AUDIT_BPRM_FCAPS;
2452 ax->d.next = context->aux;
2453 context->aux = (void *)ax;
2455 dentry = dget(bprm->file->f_dentry);
2456 get_vfs_caps_from_disk(dentry, &vcaps);
2459 ax->fcap.permitted = vcaps.permitted;
2460 ax->fcap.inheritable = vcaps.inheritable;
2461 ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2462 ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2464 ax->old_pcap.permitted = old->cap_permitted;
2465 ax->old_pcap.inheritable = old->cap_inheritable;
2466 ax->old_pcap.effective = old->cap_effective;
2468 ax->new_pcap.permitted = new->cap_permitted;
2469 ax->new_pcap.inheritable = new->cap_inheritable;
2470 ax->new_pcap.effective = new->cap_effective;
2475 * __audit_log_capset - store information about the arguments to the capset syscall
2476 * @pid: target pid of the capset call
2477 * @new: the new credentials
2478 * @old: the old (current) credentials
2480 * Record the aguments userspace sent to sys_capset for later printing by the
2481 * audit system if applicable
2483 void __audit_log_capset(pid_t pid,
2484 const struct cred *new, const struct cred *old)
2486 struct audit_context *context = current->audit_context;
2487 context->capset.pid = pid;
2488 context->capset.cap.effective = new->cap_effective;
2489 context->capset.cap.inheritable = new->cap_effective;
2490 context->capset.cap.permitted = new->cap_permitted;
2491 context->type = AUDIT_CAPSET;
2494 void __audit_mmap_fd(int fd, int flags)
2496 struct audit_context *context = current->audit_context;
2497 context->mmap.fd = fd;
2498 context->mmap.flags = flags;
2499 context->type = AUDIT_MMAP;
2503 * audit_core_dumps - record information about processes that end abnormally
2504 * @signr: signal value
2506 * If a process ends with a core dump, something fishy is going on and we
2507 * should record the event for investigation.
2509 void audit_core_dumps(long signr)
2511 struct audit_buffer *ab;
2513 uid_t auid = audit_get_loginuid(current), uid;
2515 unsigned int sessionid = audit_get_sessionid(current);
2520 if (signr == SIGQUIT) /* don't care for those */
2523 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2524 current_uid_gid(&uid, &gid);
2525 audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2526 auid, uid, gid, sessionid);
2527 security_task_getsecid(current, &sid);
2532 if (security_secid_to_secctx(sid, &ctx, &len))
2533 audit_log_format(ab, " ssid=%u", sid);
2535 audit_log_format(ab, " subj=%s", ctx);
2536 security_release_secctx(ctx, len);
2539 audit_log_format(ab, " pid=%d comm=", current->pid);
2540 audit_log_untrustedstring(ab, current->comm);
2541 audit_log_format(ab, " sig=%ld", signr);
2545 struct list_head *audit_killed_trees(void)
2547 struct audit_context *ctx = current->audit_context;
2548 if (likely(!ctx || !ctx->in_syscall))
2550 return &ctx->killed_trees;