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 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
74 * for saving names from getname(). If we get more names we will allocate
75 * a name dynamically and also add those to the list anchored by names_list. */
78 /* Indicates that audit should log the full pathname. */
79 #define AUDIT_NAME_FULL -1
81 /* no execve audit message should be longer than this (userspace limits) */
82 #define MAX_EXECVE_AUDIT_LEN 7500
84 /* number of audit rules */
87 /* determines whether we collect data for signals sent */
90 struct audit_cap_data {
91 kernel_cap_t permitted;
92 kernel_cap_t inheritable;
94 unsigned int fE; /* effective bit of a file capability */
95 kernel_cap_t effective; /* effective set of a process */
99 /* When fs/namei.c:getname() is called, we store the pointer in name and
100 * we don't let putname() free it (instead we free all of the saved
101 * pointers at syscall exit time).
103 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
105 struct list_head list; /* audit_context->names_list */
114 struct audit_cap_data fcap;
115 unsigned int fcap_ver;
116 int name_len; /* number of name's characters to log */
117 bool name_put; /* call __putname() for this name */
119 * This was an allocated audit_names and not from the array of
120 * names allocated in the task audit context. Thus this name
121 * should be freed on syscall exit
126 struct audit_aux_data {
127 struct audit_aux_data *next;
131 #define AUDIT_AUX_IPCPERM 0
133 /* Number of target pids per aux struct. */
134 #define AUDIT_AUX_PIDS 16
136 struct audit_aux_data_execve {
137 struct audit_aux_data d;
140 struct mm_struct *mm;
143 struct audit_aux_data_pids {
144 struct audit_aux_data d;
145 pid_t target_pid[AUDIT_AUX_PIDS];
146 uid_t target_auid[AUDIT_AUX_PIDS];
147 uid_t target_uid[AUDIT_AUX_PIDS];
148 unsigned int target_sessionid[AUDIT_AUX_PIDS];
149 u32 target_sid[AUDIT_AUX_PIDS];
150 char target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
154 struct audit_aux_data_bprm_fcaps {
155 struct audit_aux_data d;
156 struct audit_cap_data fcap;
157 unsigned int fcap_ver;
158 struct audit_cap_data old_pcap;
159 struct audit_cap_data new_pcap;
162 struct audit_aux_data_capset {
163 struct audit_aux_data d;
165 struct audit_cap_data cap;
168 struct audit_tree_refs {
169 struct audit_tree_refs *next;
170 struct audit_chunk *c[31];
173 /* The per-task audit context. */
174 struct audit_context {
175 int dummy; /* must be the first element */
176 int in_syscall; /* 1 if task is in a syscall */
177 enum audit_state state, current_state;
178 unsigned int serial; /* serial number for record */
179 int major; /* syscall number */
180 struct timespec ctime; /* time of syscall entry */
181 unsigned long argv[4]; /* syscall arguments */
182 long return_code;/* syscall return code */
184 int return_valid; /* return code is valid */
186 * The names_list is the list of all audit_names collected during this
187 * syscall. The first AUDIT_NAMES entries in the names_list will
188 * actually be from the preallocated_names array for performance
189 * reasons. Except during allocation they should never be referenced
190 * through the preallocated_names array and should only be found/used
191 * by running the names_list.
193 struct audit_names preallocated_names[AUDIT_NAMES];
194 int name_count; /* total records in names_list */
195 struct list_head names_list; /* anchor for struct audit_names->list */
196 char * filterkey; /* key for rule that triggered record */
198 struct audit_context *previous; /* For nested syscalls */
199 struct audit_aux_data *aux;
200 struct audit_aux_data *aux_pids;
201 struct sockaddr_storage *sockaddr;
203 /* Save things to print about task_struct */
205 uid_t uid, euid, suid, fsuid;
206 gid_t gid, egid, sgid, fsgid;
207 unsigned long personality;
213 unsigned int target_sessionid;
215 char target_comm[TASK_COMM_LEN];
217 struct audit_tree_refs *trees, *first_trees;
218 struct list_head killed_trees;
236 unsigned long qbytes;
240 struct mq_attr mqstat;
249 unsigned int msg_prio;
250 struct timespec abs_timeout;
259 struct audit_cap_data cap;
274 static inline int open_arg(int flags, int mask)
276 int n = ACC_MODE(flags);
277 if (flags & (O_TRUNC | O_CREAT))
278 n |= AUDIT_PERM_WRITE;
282 static int audit_match_perm(struct audit_context *ctx, int mask)
289 switch (audit_classify_syscall(ctx->arch, n)) {
291 if ((mask & AUDIT_PERM_WRITE) &&
292 audit_match_class(AUDIT_CLASS_WRITE, n))
294 if ((mask & AUDIT_PERM_READ) &&
295 audit_match_class(AUDIT_CLASS_READ, n))
297 if ((mask & AUDIT_PERM_ATTR) &&
298 audit_match_class(AUDIT_CLASS_CHATTR, n))
301 case 1: /* 32bit on biarch */
302 if ((mask & AUDIT_PERM_WRITE) &&
303 audit_match_class(AUDIT_CLASS_WRITE_32, n))
305 if ((mask & AUDIT_PERM_READ) &&
306 audit_match_class(AUDIT_CLASS_READ_32, n))
308 if ((mask & AUDIT_PERM_ATTR) &&
309 audit_match_class(AUDIT_CLASS_CHATTR_32, n))
313 return mask & ACC_MODE(ctx->argv[1]);
315 return mask & ACC_MODE(ctx->argv[2]);
316 case 4: /* socketcall */
317 return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
319 return mask & AUDIT_PERM_EXEC;
325 static int audit_match_filetype(struct audit_context *ctx, int val)
327 struct audit_names *n;
328 umode_t mode = (umode_t)val;
333 list_for_each_entry(n, &ctx->names_list, list) {
334 if ((n->ino != -1) &&
335 ((n->mode & S_IFMT) == mode))
343 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
344 * ->first_trees points to its beginning, ->trees - to the current end of data.
345 * ->tree_count is the number of free entries in array pointed to by ->trees.
346 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
347 * "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
348 * it's going to remain 1-element for almost any setup) until we free context itself.
349 * References in it _are_ dropped - at the same time we free/drop aux stuff.
352 #ifdef CONFIG_AUDIT_TREE
353 static void audit_set_auditable(struct audit_context *ctx)
357 ctx->current_state = AUDIT_RECORD_CONTEXT;
361 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
363 struct audit_tree_refs *p = ctx->trees;
364 int left = ctx->tree_count;
366 p->c[--left] = chunk;
367 ctx->tree_count = left;
376 ctx->tree_count = 30;
382 static int grow_tree_refs(struct audit_context *ctx)
384 struct audit_tree_refs *p = ctx->trees;
385 ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
391 p->next = ctx->trees;
393 ctx->first_trees = ctx->trees;
394 ctx->tree_count = 31;
399 static void unroll_tree_refs(struct audit_context *ctx,
400 struct audit_tree_refs *p, int count)
402 #ifdef CONFIG_AUDIT_TREE
403 struct audit_tree_refs *q;
406 /* we started with empty chain */
407 p = ctx->first_trees;
409 /* if the very first allocation has failed, nothing to do */
414 for (q = p; q != ctx->trees; q = q->next, n = 31) {
416 audit_put_chunk(q->c[n]);
420 while (n-- > ctx->tree_count) {
421 audit_put_chunk(q->c[n]);
425 ctx->tree_count = count;
429 static void free_tree_refs(struct audit_context *ctx)
431 struct audit_tree_refs *p, *q;
432 for (p = ctx->first_trees; p; p = q) {
438 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
440 #ifdef CONFIG_AUDIT_TREE
441 struct audit_tree_refs *p;
446 for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
447 for (n = 0; n < 31; n++)
448 if (audit_tree_match(p->c[n], tree))
453 for (n = ctx->tree_count; n < 31; n++)
454 if (audit_tree_match(p->c[n], tree))
461 /* Determine if any context name data matches a rule's watch data */
462 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
465 * If task_creation is true, this is an explicit indication that we are
466 * filtering a task rule at task creation time. This and tsk == current are
467 * the only situations where tsk->cred may be accessed without an rcu read lock.
469 static int audit_filter_rules(struct task_struct *tsk,
470 struct audit_krule *rule,
471 struct audit_context *ctx,
472 struct audit_names *name,
473 enum audit_state *state,
476 const struct cred *cred;
480 cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
482 for (i = 0; i < rule->field_count; i++) {
483 struct audit_field *f = &rule->fields[i];
484 struct audit_names *n;
489 result = audit_comparator(tsk->pid, f->op, f->val);
494 ctx->ppid = sys_getppid();
495 result = audit_comparator(ctx->ppid, f->op, f->val);
499 result = audit_comparator(cred->uid, f->op, f->val);
502 result = audit_comparator(cred->euid, f->op, f->val);
505 result = audit_comparator(cred->suid, f->op, f->val);
508 result = audit_comparator(cred->fsuid, f->op, f->val);
511 result = audit_comparator(cred->gid, f->op, f->val);
514 result = audit_comparator(cred->egid, f->op, f->val);
517 result = audit_comparator(cred->sgid, f->op, f->val);
520 result = audit_comparator(cred->fsgid, f->op, f->val);
523 result = audit_comparator(tsk->personality, f->op, f->val);
527 result = audit_comparator(ctx->arch, f->op, f->val);
531 if (ctx && ctx->return_valid)
532 result = audit_comparator(ctx->return_code, f->op, f->val);
535 if (ctx && ctx->return_valid) {
537 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
539 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
544 if (audit_comparator(MAJOR(name->dev), f->op, f->val) ||
545 audit_comparator(MAJOR(name->rdev), f->op, f->val))
548 list_for_each_entry(n, &ctx->names_list, list) {
549 if (audit_comparator(MAJOR(n->dev), f->op, f->val) ||
550 audit_comparator(MAJOR(n->rdev), f->op, f->val)) {
559 if (audit_comparator(MINOR(name->dev), f->op, f->val) ||
560 audit_comparator(MINOR(name->rdev), f->op, f->val))
563 list_for_each_entry(n, &ctx->names_list, list) {
564 if (audit_comparator(MINOR(n->dev), f->op, f->val) ||
565 audit_comparator(MINOR(n->rdev), f->op, f->val)) {
574 result = (name->ino == f->val);
576 list_for_each_entry(n, &ctx->names_list, list) {
577 if (audit_comparator(n->ino, f->op, f->val)) {
586 result = audit_watch_compare(rule->watch, name->ino, name->dev);
590 result = match_tree_refs(ctx, rule->tree);
595 result = audit_comparator(tsk->loginuid, f->op, f->val);
597 case AUDIT_SUBJ_USER:
598 case AUDIT_SUBJ_ROLE:
599 case AUDIT_SUBJ_TYPE:
602 /* NOTE: this may return negative values indicating
603 a temporary error. We simply treat this as a
604 match for now to avoid losing information that
605 may be wanted. An error message will also be
609 security_task_getsecid(tsk, &sid);
612 result = security_audit_rule_match(sid, f->type,
621 case AUDIT_OBJ_LEV_LOW:
622 case AUDIT_OBJ_LEV_HIGH:
623 /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
626 /* Find files that match */
628 result = security_audit_rule_match(
629 name->osid, f->type, f->op,
632 list_for_each_entry(n, &ctx->names_list, list) {
633 if (security_audit_rule_match(n->osid, f->type,
641 /* Find ipc objects that match */
642 if (!ctx || ctx->type != AUDIT_IPC)
644 if (security_audit_rule_match(ctx->ipc.osid,
655 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
657 case AUDIT_FILTERKEY:
658 /* ignore this field for filtering */
662 result = audit_match_perm(ctx, f->val);
665 result = audit_match_filetype(ctx, f->val);
674 if (rule->prio <= ctx->prio)
676 if (rule->filterkey) {
677 kfree(ctx->filterkey);
678 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
680 ctx->prio = rule->prio;
682 switch (rule->action) {
683 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
684 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
689 /* At process creation time, we can determine if system-call auditing is
690 * completely disabled for this task. Since we only have the task
691 * structure at this point, we can only check uid and gid.
693 static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
695 struct audit_entry *e;
696 enum audit_state state;
699 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
700 if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
702 if (state == AUDIT_RECORD_CONTEXT)
703 *key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
709 return AUDIT_BUILD_CONTEXT;
712 /* At syscall entry and exit time, this filter is called if the
713 * audit_state is not low enough that auditing cannot take place, but is
714 * also not high enough that we already know we have to write an audit
715 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
717 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
718 struct audit_context *ctx,
719 struct list_head *list)
721 struct audit_entry *e;
722 enum audit_state state;
724 if (audit_pid && tsk->tgid == audit_pid)
725 return AUDIT_DISABLED;
728 if (!list_empty(list)) {
729 int word = AUDIT_WORD(ctx->major);
730 int bit = AUDIT_BIT(ctx->major);
732 list_for_each_entry_rcu(e, list, list) {
733 if ((e->rule.mask[word] & bit) == bit &&
734 audit_filter_rules(tsk, &e->rule, ctx, NULL,
737 ctx->current_state = state;
743 return AUDIT_BUILD_CONTEXT;
747 * Given an audit_name check the inode hash table to see if they match.
748 * Called holding the rcu read lock to protect the use of audit_inode_hash
750 static int audit_filter_inode_name(struct task_struct *tsk,
751 struct audit_names *n,
752 struct audit_context *ctx) {
754 int h = audit_hash_ino((u32)n->ino);
755 struct list_head *list = &audit_inode_hash[h];
756 struct audit_entry *e;
757 enum audit_state state;
759 word = AUDIT_WORD(ctx->major);
760 bit = AUDIT_BIT(ctx->major);
762 if (list_empty(list))
765 list_for_each_entry_rcu(e, list, list) {
766 if ((e->rule.mask[word] & bit) == bit &&
767 audit_filter_rules(tsk, &e->rule, ctx, n, &state, false)) {
768 ctx->current_state = state;
776 /* At syscall exit time, this filter is called if any audit_names have been
777 * collected during syscall processing. We only check rules in sublists at hash
778 * buckets applicable to the inode numbers in audit_names.
779 * Regarding audit_state, same rules apply as for audit_filter_syscall().
781 void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
783 struct audit_names *n;
785 if (audit_pid && tsk->tgid == audit_pid)
790 list_for_each_entry(n, &ctx->names_list, list) {
791 if (audit_filter_inode_name(tsk, n, ctx))
797 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
801 struct audit_context *context = tsk->audit_context;
803 if (likely(!context))
805 context->return_valid = return_valid;
808 * we need to fix up the return code in the audit logs if the actual
809 * return codes are later going to be fixed up by the arch specific
812 * This is actually a test for:
813 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
814 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
816 * but is faster than a bunch of ||
818 if (unlikely(return_code <= -ERESTARTSYS) &&
819 (return_code >= -ERESTART_RESTARTBLOCK) &&
820 (return_code != -ENOIOCTLCMD))
821 context->return_code = -EINTR;
823 context->return_code = return_code;
825 if (context->in_syscall && !context->dummy) {
826 audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
827 audit_filter_inodes(tsk, context);
830 tsk->audit_context = NULL;
834 static inline void audit_free_names(struct audit_context *context)
836 struct audit_names *n, *next;
839 if (context->put_count + context->ino_count != context->name_count) {
840 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
841 " name_count=%d put_count=%d"
842 " ino_count=%d [NOT freeing]\n",
844 context->serial, context->major, context->in_syscall,
845 context->name_count, context->put_count,
847 list_for_each_entry(n, &context->names_list, list) {
848 printk(KERN_ERR "names[%d] = %p = %s\n", i,
849 n->name, n->name ?: "(null)");
856 context->put_count = 0;
857 context->ino_count = 0;
860 list_for_each_entry_safe(n, next, &context->names_list, list) {
862 if (n->name && n->name_put)
867 context->name_count = 0;
868 path_put(&context->pwd);
869 context->pwd.dentry = NULL;
870 context->pwd.mnt = NULL;
873 static inline void audit_free_aux(struct audit_context *context)
875 struct audit_aux_data *aux;
877 while ((aux = context->aux)) {
878 context->aux = aux->next;
881 while ((aux = context->aux_pids)) {
882 context->aux_pids = aux->next;
887 static inline void audit_zero_context(struct audit_context *context,
888 enum audit_state state)
890 memset(context, 0, sizeof(*context));
891 context->state = state;
892 context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
895 static inline struct audit_context *audit_alloc_context(enum audit_state state)
897 struct audit_context *context;
899 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
901 audit_zero_context(context, state);
902 INIT_LIST_HEAD(&context->killed_trees);
903 INIT_LIST_HEAD(&context->names_list);
908 * audit_alloc - allocate an audit context block for a task
911 * Filter on the task information and allocate a per-task audit context
912 * if necessary. Doing so turns on system call auditing for the
913 * specified task. This is called from copy_process, so no lock is
916 int audit_alloc(struct task_struct *tsk)
918 struct audit_context *context;
919 enum audit_state state;
922 if (likely(!audit_ever_enabled))
923 return 0; /* Return if not auditing. */
925 state = audit_filter_task(tsk, &key);
926 if (likely(state == AUDIT_DISABLED))
929 if (!(context = audit_alloc_context(state))) {
931 audit_log_lost("out of memory in audit_alloc");
934 context->filterkey = key;
936 tsk->audit_context = context;
937 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
941 static inline void audit_free_context(struct audit_context *context)
943 struct audit_context *previous;
947 previous = context->previous;
948 if (previous || (count && count < 10)) {
950 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
951 " freeing multiple contexts (%d)\n",
952 context->serial, context->major,
953 context->name_count, count);
955 audit_free_names(context);
956 unroll_tree_refs(context, NULL, 0);
957 free_tree_refs(context);
958 audit_free_aux(context);
959 kfree(context->filterkey);
960 kfree(context->sockaddr);
965 printk(KERN_ERR "audit: freed %d contexts\n", count);
968 void audit_log_task_context(struct audit_buffer *ab)
975 security_task_getsecid(current, &sid);
979 error = security_secid_to_secctx(sid, &ctx, &len);
981 if (error != -EINVAL)
986 audit_log_format(ab, " subj=%s", ctx);
987 security_release_secctx(ctx, len);
991 audit_panic("error in audit_log_task_context");
995 EXPORT_SYMBOL(audit_log_task_context);
997 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
999 char name[sizeof(tsk->comm)];
1000 struct mm_struct *mm = tsk->mm;
1001 struct vm_area_struct *vma;
1003 /* tsk == current */
1005 get_task_comm(name, tsk);
1006 audit_log_format(ab, " comm=");
1007 audit_log_untrustedstring(ab, name);
1010 down_read(&mm->mmap_sem);
1013 if ((vma->vm_flags & VM_EXECUTABLE) &&
1015 audit_log_d_path(ab, "exe=",
1016 &vma->vm_file->f_path);
1021 up_read(&mm->mmap_sem);
1023 audit_log_task_context(ab);
1026 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
1027 uid_t auid, uid_t uid, unsigned int sessionid,
1028 u32 sid, char *comm)
1030 struct audit_buffer *ab;
1035 ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
1039 audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
1041 if (security_secid_to_secctx(sid, &ctx, &len)) {
1042 audit_log_format(ab, " obj=(none)");
1045 audit_log_format(ab, " obj=%s", ctx);
1046 security_release_secctx(ctx, len);
1048 audit_log_format(ab, " ocomm=");
1049 audit_log_untrustedstring(ab, comm);
1056 * to_send and len_sent accounting are very loose estimates. We aren't
1057 * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1058 * within about 500 bytes (next page boundary)
1060 * why snprintf? an int is up to 12 digits long. if we just assumed when
1061 * logging that a[%d]= was going to be 16 characters long we would be wasting
1062 * space in every audit message. In one 7500 byte message we can log up to
1063 * about 1000 min size arguments. That comes down to about 50% waste of space
1064 * if we didn't do the snprintf to find out how long arg_num_len was.
1066 static int audit_log_single_execve_arg(struct audit_context *context,
1067 struct audit_buffer **ab,
1070 const char __user *p,
1073 char arg_num_len_buf[12];
1074 const char __user *tmp_p = p;
1075 /* how many digits are in arg_num? 5 is the length of ' a=""' */
1076 size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1077 size_t len, len_left, to_send;
1078 size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1079 unsigned int i, has_cntl = 0, too_long = 0;
1082 /* strnlen_user includes the null we don't want to send */
1083 len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1086 * We just created this mm, if we can't find the strings
1087 * we just copied into it something is _very_ wrong. Similar
1088 * for strings that are too long, we should not have created
1091 if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1093 send_sig(SIGKILL, current, 0);
1097 /* walk the whole argument looking for non-ascii chars */
1099 if (len_left > MAX_EXECVE_AUDIT_LEN)
1100 to_send = MAX_EXECVE_AUDIT_LEN;
1103 ret = copy_from_user(buf, tmp_p, to_send);
1105 * There is no reason for this copy to be short. We just
1106 * copied them here, and the mm hasn't been exposed to user-
1111 send_sig(SIGKILL, current, 0);
1114 buf[to_send] = '\0';
1115 has_cntl = audit_string_contains_control(buf, to_send);
1118 * hex messages get logged as 2 bytes, so we can only
1119 * send half as much in each message
1121 max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1124 len_left -= to_send;
1126 } while (len_left > 0);
1130 if (len > max_execve_audit_len)
1133 /* rewalk the argument actually logging the message */
1134 for (i = 0; len_left > 0; i++) {
1137 if (len_left > max_execve_audit_len)
1138 to_send = max_execve_audit_len;
1142 /* do we have space left to send this argument in this ab? */
1143 room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1145 room_left -= (to_send * 2);
1147 room_left -= to_send;
1148 if (room_left < 0) {
1151 *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1157 * first record needs to say how long the original string was
1158 * so we can be sure nothing was lost.
1160 if ((i == 0) && (too_long))
1161 audit_log_format(*ab, " a%d_len=%zu", arg_num,
1162 has_cntl ? 2*len : len);
1165 * normally arguments are small enough to fit and we already
1166 * filled buf above when we checked for control characters
1167 * so don't bother with another copy_from_user
1169 if (len >= max_execve_audit_len)
1170 ret = copy_from_user(buf, p, to_send);
1175 send_sig(SIGKILL, current, 0);
1178 buf[to_send] = '\0';
1180 /* actually log it */
1181 audit_log_format(*ab, " a%d", arg_num);
1183 audit_log_format(*ab, "[%d]", i);
1184 audit_log_format(*ab, "=");
1186 audit_log_n_hex(*ab, buf, to_send);
1188 audit_log_string(*ab, buf);
1191 len_left -= to_send;
1192 *len_sent += arg_num_len;
1194 *len_sent += to_send * 2;
1196 *len_sent += to_send;
1198 /* include the null we didn't log */
1202 static void audit_log_execve_info(struct audit_context *context,
1203 struct audit_buffer **ab,
1204 struct audit_aux_data_execve *axi)
1207 size_t len, len_sent = 0;
1208 const char __user *p;
1211 if (axi->mm != current->mm)
1212 return; /* execve failed, no additional info */
1214 p = (const char __user *)axi->mm->arg_start;
1216 audit_log_format(*ab, "argc=%d", axi->argc);
1219 * we need some kernel buffer to hold the userspace args. Just
1220 * allocate one big one rather than allocating one of the right size
1221 * for every single argument inside audit_log_single_execve_arg()
1222 * should be <8k allocation so should be pretty safe.
1224 buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1226 audit_panic("out of memory for argv string\n");
1230 for (i = 0; i < axi->argc; i++) {
1231 len = audit_log_single_execve_arg(context, ab, i,
1240 static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1244 audit_log_format(ab, " %s=", prefix);
1245 CAP_FOR_EACH_U32(i) {
1246 audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1250 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1252 kernel_cap_t *perm = &name->fcap.permitted;
1253 kernel_cap_t *inh = &name->fcap.inheritable;
1256 if (!cap_isclear(*perm)) {
1257 audit_log_cap(ab, "cap_fp", perm);
1260 if (!cap_isclear(*inh)) {
1261 audit_log_cap(ab, "cap_fi", inh);
1266 audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1269 static void show_special(struct audit_context *context, int *call_panic)
1271 struct audit_buffer *ab;
1274 ab = audit_log_start(context, GFP_KERNEL, context->type);
1278 switch (context->type) {
1279 case AUDIT_SOCKETCALL: {
1280 int nargs = context->socketcall.nargs;
1281 audit_log_format(ab, "nargs=%d", nargs);
1282 for (i = 0; i < nargs; i++)
1283 audit_log_format(ab, " a%d=%lx", i,
1284 context->socketcall.args[i]);
1287 u32 osid = context->ipc.osid;
1289 audit_log_format(ab, "ouid=%u ogid=%u mode=%#ho",
1290 context->ipc.uid, context->ipc.gid, context->ipc.mode);
1294 if (security_secid_to_secctx(osid, &ctx, &len)) {
1295 audit_log_format(ab, " osid=%u", osid);
1298 audit_log_format(ab, " obj=%s", ctx);
1299 security_release_secctx(ctx, len);
1302 if (context->ipc.has_perm) {
1304 ab = audit_log_start(context, GFP_KERNEL,
1305 AUDIT_IPC_SET_PERM);
1306 audit_log_format(ab,
1307 "qbytes=%lx ouid=%u ogid=%u mode=%#ho",
1308 context->ipc.qbytes,
1309 context->ipc.perm_uid,
1310 context->ipc.perm_gid,
1311 context->ipc.perm_mode);
1316 case AUDIT_MQ_OPEN: {
1317 audit_log_format(ab,
1318 "oflag=0x%x mode=%#ho mq_flags=0x%lx mq_maxmsg=%ld "
1319 "mq_msgsize=%ld mq_curmsgs=%ld",
1320 context->mq_open.oflag, context->mq_open.mode,
1321 context->mq_open.attr.mq_flags,
1322 context->mq_open.attr.mq_maxmsg,
1323 context->mq_open.attr.mq_msgsize,
1324 context->mq_open.attr.mq_curmsgs);
1326 case AUDIT_MQ_SENDRECV: {
1327 audit_log_format(ab,
1328 "mqdes=%d msg_len=%zd msg_prio=%u "
1329 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1330 context->mq_sendrecv.mqdes,
1331 context->mq_sendrecv.msg_len,
1332 context->mq_sendrecv.msg_prio,
1333 context->mq_sendrecv.abs_timeout.tv_sec,
1334 context->mq_sendrecv.abs_timeout.tv_nsec);
1336 case AUDIT_MQ_NOTIFY: {
1337 audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1338 context->mq_notify.mqdes,
1339 context->mq_notify.sigev_signo);
1341 case AUDIT_MQ_GETSETATTR: {
1342 struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1343 audit_log_format(ab,
1344 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1346 context->mq_getsetattr.mqdes,
1347 attr->mq_flags, attr->mq_maxmsg,
1348 attr->mq_msgsize, attr->mq_curmsgs);
1350 case AUDIT_CAPSET: {
1351 audit_log_format(ab, "pid=%d", context->capset.pid);
1352 audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1353 audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1354 audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1357 audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1358 context->mmap.flags);
1364 static void audit_log_name(struct audit_context *context, struct audit_names *n,
1365 int record_num, int *call_panic)
1367 struct audit_buffer *ab;
1368 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1370 return; /* audit_panic has been called */
1372 audit_log_format(ab, "item=%d", record_num);
1375 switch (n->name_len) {
1376 case AUDIT_NAME_FULL:
1377 /* log the full path */
1378 audit_log_format(ab, " name=");
1379 audit_log_untrustedstring(ab, n->name);
1382 /* name was specified as a relative path and the
1383 * directory component is the cwd */
1384 audit_log_d_path(ab, "name=", &context->pwd);
1387 /* log the name's directory component */
1388 audit_log_format(ab, " name=");
1389 audit_log_n_untrustedstring(ab, n->name,
1393 audit_log_format(ab, " name=(null)");
1395 if (n->ino != (unsigned long)-1) {
1396 audit_log_format(ab, " inode=%lu"
1397 " dev=%02x:%02x mode=%#ho"
1398 " ouid=%u ogid=%u rdev=%02x:%02x",
1411 if (security_secid_to_secctx(
1412 n->osid, &ctx, &len)) {
1413 audit_log_format(ab, " osid=%u", n->osid);
1416 audit_log_format(ab, " obj=%s", ctx);
1417 security_release_secctx(ctx, len);
1421 audit_log_fcaps(ab, n);
1426 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1428 const struct cred *cred;
1429 int i, call_panic = 0;
1430 struct audit_buffer *ab;
1431 struct audit_aux_data *aux;
1433 struct audit_names *n;
1435 /* tsk == current */
1436 context->pid = tsk->pid;
1438 context->ppid = sys_getppid();
1439 cred = current_cred();
1440 context->uid = cred->uid;
1441 context->gid = cred->gid;
1442 context->euid = cred->euid;
1443 context->suid = cred->suid;
1444 context->fsuid = cred->fsuid;
1445 context->egid = cred->egid;
1446 context->sgid = cred->sgid;
1447 context->fsgid = cred->fsgid;
1448 context->personality = tsk->personality;
1450 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1452 return; /* audit_panic has been called */
1453 audit_log_format(ab, "arch=%x syscall=%d",
1454 context->arch, context->major);
1455 if (context->personality != PER_LINUX)
1456 audit_log_format(ab, " per=%lx", context->personality);
1457 if (context->return_valid)
1458 audit_log_format(ab, " success=%s exit=%ld",
1459 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1460 context->return_code);
1462 spin_lock_irq(&tsk->sighand->siglock);
1463 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1464 tty = tsk->signal->tty->name;
1467 spin_unlock_irq(&tsk->sighand->siglock);
1469 audit_log_format(ab,
1470 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1471 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1472 " euid=%u suid=%u fsuid=%u"
1473 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1478 context->name_count,
1484 context->euid, context->suid, context->fsuid,
1485 context->egid, context->sgid, context->fsgid, tty,
1489 audit_log_task_info(ab, tsk);
1490 audit_log_key(ab, context->filterkey);
1493 for (aux = context->aux; aux; aux = aux->next) {
1495 ab = audit_log_start(context, GFP_KERNEL, aux->type);
1497 continue; /* audit_panic has been called */
1499 switch (aux->type) {
1501 case AUDIT_EXECVE: {
1502 struct audit_aux_data_execve *axi = (void *)aux;
1503 audit_log_execve_info(context, &ab, axi);
1506 case AUDIT_BPRM_FCAPS: {
1507 struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1508 audit_log_format(ab, "fver=%x", axs->fcap_ver);
1509 audit_log_cap(ab, "fp", &axs->fcap.permitted);
1510 audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1511 audit_log_format(ab, " fe=%d", axs->fcap.fE);
1512 audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1513 audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1514 audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1515 audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1516 audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1517 audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1525 show_special(context, &call_panic);
1527 if (context->fds[0] >= 0) {
1528 ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1530 audit_log_format(ab, "fd0=%d fd1=%d",
1531 context->fds[0], context->fds[1]);
1536 if (context->sockaddr_len) {
1537 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1539 audit_log_format(ab, "saddr=");
1540 audit_log_n_hex(ab, (void *)context->sockaddr,
1541 context->sockaddr_len);
1546 for (aux = context->aux_pids; aux; aux = aux->next) {
1547 struct audit_aux_data_pids *axs = (void *)aux;
1549 for (i = 0; i < axs->pid_count; i++)
1550 if (audit_log_pid_context(context, axs->target_pid[i],
1551 axs->target_auid[i],
1553 axs->target_sessionid[i],
1555 axs->target_comm[i]))
1559 if (context->target_pid &&
1560 audit_log_pid_context(context, context->target_pid,
1561 context->target_auid, context->target_uid,
1562 context->target_sessionid,
1563 context->target_sid, context->target_comm))
1566 if (context->pwd.dentry && context->pwd.mnt) {
1567 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1569 audit_log_d_path(ab, "cwd=", &context->pwd);
1575 list_for_each_entry(n, &context->names_list, list)
1576 audit_log_name(context, n, i++, &call_panic);
1578 /* Send end of event record to help user space know we are finished */
1579 ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1583 audit_panic("error converting sid to string");
1587 * audit_free - free a per-task audit context
1588 * @tsk: task whose audit context block to free
1590 * Called from copy_process and do_exit
1592 void audit_free(struct task_struct *tsk)
1594 struct audit_context *context;
1596 context = audit_get_context(tsk, 0, 0);
1597 if (likely(!context))
1600 /* Check for system calls that do not go through the exit
1601 * function (e.g., exit_group), then free context block.
1602 * We use GFP_ATOMIC here because we might be doing this
1603 * in the context of the idle thread */
1604 /* that can happen only if we are called from do_exit() */
1605 if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1606 audit_log_exit(context, tsk);
1607 if (!list_empty(&context->killed_trees))
1608 audit_kill_trees(&context->killed_trees);
1610 audit_free_context(context);
1614 * audit_syscall_entry - fill in an audit record at syscall entry
1615 * @arch: architecture type
1616 * @major: major syscall type (function)
1617 * @a1: additional syscall register 1
1618 * @a2: additional syscall register 2
1619 * @a3: additional syscall register 3
1620 * @a4: additional syscall register 4
1622 * Fill in audit context at syscall entry. This only happens if the
1623 * audit context was created when the task was created and the state or
1624 * filters demand the audit context be built. If the state from the
1625 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1626 * then the record will be written at syscall exit time (otherwise, it
1627 * will only be written if another part of the kernel requests that it
1630 void audit_syscall_entry(int arch, int major,
1631 unsigned long a1, unsigned long a2,
1632 unsigned long a3, unsigned long a4)
1634 struct task_struct *tsk = current;
1635 struct audit_context *context = tsk->audit_context;
1636 enum audit_state state;
1638 if (unlikely(!context))
1642 * This happens only on certain architectures that make system
1643 * calls in kernel_thread via the entry.S interface, instead of
1644 * with direct calls. (If you are porting to a new
1645 * architecture, hitting this condition can indicate that you
1646 * got the _exit/_leave calls backward in entry.S.)
1650 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1652 * This also happens with vm86 emulation in a non-nested manner
1653 * (entries without exits), so this case must be caught.
1655 if (context->in_syscall) {
1656 struct audit_context *newctx;
1660 "audit(:%d) pid=%d in syscall=%d;"
1661 " entering syscall=%d\n",
1662 context->serial, tsk->pid, context->major, major);
1664 newctx = audit_alloc_context(context->state);
1666 newctx->previous = context;
1668 tsk->audit_context = newctx;
1670 /* If we can't alloc a new context, the best we
1671 * can do is to leak memory (any pending putname
1672 * will be lost). The only other alternative is
1673 * to abandon auditing. */
1674 audit_zero_context(context, context->state);
1677 BUG_ON(context->in_syscall || context->name_count);
1682 context->arch = arch;
1683 context->major = major;
1684 context->argv[0] = a1;
1685 context->argv[1] = a2;
1686 context->argv[2] = a3;
1687 context->argv[3] = a4;
1689 state = context->state;
1690 context->dummy = !audit_n_rules;
1691 if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1693 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1695 if (likely(state == AUDIT_DISABLED))
1698 context->serial = 0;
1699 context->ctime = CURRENT_TIME;
1700 context->in_syscall = 1;
1701 context->current_state = state;
1705 void audit_finish_fork(struct task_struct *child)
1707 struct audit_context *ctx = current->audit_context;
1708 struct audit_context *p = child->audit_context;
1711 if (!ctx->in_syscall || ctx->current_state != AUDIT_RECORD_CONTEXT)
1713 p->arch = ctx->arch;
1714 p->major = ctx->major;
1715 memcpy(p->argv, ctx->argv, sizeof(ctx->argv));
1716 p->ctime = ctx->ctime;
1717 p->dummy = ctx->dummy;
1718 p->in_syscall = ctx->in_syscall;
1719 p->filterkey = kstrdup(ctx->filterkey, GFP_KERNEL);
1720 p->ppid = current->pid;
1721 p->prio = ctx->prio;
1722 p->current_state = ctx->current_state;
1726 * audit_syscall_exit - deallocate audit context after a system call
1727 * @valid: success/failure flag
1728 * @return_code: syscall return value
1730 * Tear down after system call. If the audit context has been marked as
1731 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1732 * filtering, or because some other part of the kernel write an audit
1733 * message), then write out the syscall information. In call cases,
1734 * free the names stored from getname().
1736 void audit_syscall_exit(int valid, long return_code)
1738 struct task_struct *tsk = current;
1739 struct audit_context *context;
1741 context = audit_get_context(tsk, valid, return_code);
1743 if (likely(!context))
1746 if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1747 audit_log_exit(context, tsk);
1749 context->in_syscall = 0;
1750 context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1752 if (!list_empty(&context->killed_trees))
1753 audit_kill_trees(&context->killed_trees);
1755 if (context->previous) {
1756 struct audit_context *new_context = context->previous;
1757 context->previous = NULL;
1758 audit_free_context(context);
1759 tsk->audit_context = new_context;
1761 audit_free_names(context);
1762 unroll_tree_refs(context, NULL, 0);
1763 audit_free_aux(context);
1764 context->aux = NULL;
1765 context->aux_pids = NULL;
1766 context->target_pid = 0;
1767 context->target_sid = 0;
1768 context->sockaddr_len = 0;
1770 context->fds[0] = -1;
1771 if (context->state != AUDIT_RECORD_CONTEXT) {
1772 kfree(context->filterkey);
1773 context->filterkey = NULL;
1775 tsk->audit_context = context;
1779 static inline void handle_one(const struct inode *inode)
1781 #ifdef CONFIG_AUDIT_TREE
1782 struct audit_context *context;
1783 struct audit_tree_refs *p;
1784 struct audit_chunk *chunk;
1786 if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1788 context = current->audit_context;
1790 count = context->tree_count;
1792 chunk = audit_tree_lookup(inode);
1796 if (likely(put_tree_ref(context, chunk)))
1798 if (unlikely(!grow_tree_refs(context))) {
1799 printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1800 audit_set_auditable(context);
1801 audit_put_chunk(chunk);
1802 unroll_tree_refs(context, p, count);
1805 put_tree_ref(context, chunk);
1809 static void handle_path(const struct dentry *dentry)
1811 #ifdef CONFIG_AUDIT_TREE
1812 struct audit_context *context;
1813 struct audit_tree_refs *p;
1814 const struct dentry *d, *parent;
1815 struct audit_chunk *drop;
1819 context = current->audit_context;
1821 count = context->tree_count;
1826 seq = read_seqbegin(&rename_lock);
1828 struct inode *inode = d->d_inode;
1829 if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1830 struct audit_chunk *chunk;
1831 chunk = audit_tree_lookup(inode);
1833 if (unlikely(!put_tree_ref(context, chunk))) {
1839 parent = d->d_parent;
1844 if (unlikely(read_seqretry(&rename_lock, seq) || drop)) { /* in this order */
1847 /* just a race with rename */
1848 unroll_tree_refs(context, p, count);
1851 audit_put_chunk(drop);
1852 if (grow_tree_refs(context)) {
1853 /* OK, got more space */
1854 unroll_tree_refs(context, p, count);
1859 "out of memory, audit has lost a tree reference\n");
1860 unroll_tree_refs(context, p, count);
1861 audit_set_auditable(context);
1868 static struct audit_names *audit_alloc_name(struct audit_context *context)
1870 struct audit_names *aname;
1872 if (context->name_count < AUDIT_NAMES) {
1873 aname = &context->preallocated_names[context->name_count];
1874 memset(aname, 0, sizeof(*aname));
1876 aname = kzalloc(sizeof(*aname), GFP_NOFS);
1879 aname->should_free = true;
1882 aname->ino = (unsigned long)-1;
1883 list_add_tail(&aname->list, &context->names_list);
1885 context->name_count++;
1887 context->ino_count++;
1893 * audit_getname - add a name to the list
1894 * @name: name to add
1896 * Add a name to the list of audit names for this context.
1897 * Called from fs/namei.c:getname().
1899 void __audit_getname(const char *name)
1901 struct audit_context *context = current->audit_context;
1902 struct audit_names *n;
1904 if (IS_ERR(name) || !name)
1907 if (!context->in_syscall) {
1908 #if AUDIT_DEBUG == 2
1909 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1910 __FILE__, __LINE__, context->serial, name);
1916 n = audit_alloc_name(context);
1921 n->name_len = AUDIT_NAME_FULL;
1924 if (!context->pwd.dentry)
1925 get_fs_pwd(current->fs, &context->pwd);
1928 /* audit_putname - intercept a putname request
1929 * @name: name to intercept and delay for putname
1931 * If we have stored the name from getname in the audit context,
1932 * then we delay the putname until syscall exit.
1933 * Called from include/linux/fs.h:putname().
1935 void audit_putname(const char *name)
1937 struct audit_context *context = current->audit_context;
1940 if (!context->in_syscall) {
1941 #if AUDIT_DEBUG == 2
1942 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1943 __FILE__, __LINE__, context->serial, name);
1944 if (context->name_count) {
1945 struct audit_names *n;
1948 list_for_each_entry(n, &context->names_list, list)
1949 printk(KERN_ERR "name[%d] = %p = %s\n", i,
1950 n->name, n->name ?: "(null)");
1957 ++context->put_count;
1958 if (context->put_count > context->name_count) {
1959 printk(KERN_ERR "%s:%d(:%d): major=%d"
1960 " in_syscall=%d putname(%p) name_count=%d"
1963 context->serial, context->major,
1964 context->in_syscall, name, context->name_count,
1965 context->put_count);
1972 static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
1974 struct cpu_vfs_cap_data caps;
1980 rc = get_vfs_caps_from_disk(dentry, &caps);
1984 name->fcap.permitted = caps.permitted;
1985 name->fcap.inheritable = caps.inheritable;
1986 name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
1987 name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
1993 /* Copy inode data into an audit_names. */
1994 static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
1995 const struct inode *inode)
1997 name->ino = inode->i_ino;
1998 name->dev = inode->i_sb->s_dev;
1999 name->mode = inode->i_mode;
2000 name->uid = inode->i_uid;
2001 name->gid = inode->i_gid;
2002 name->rdev = inode->i_rdev;
2003 security_inode_getsecid(inode, &name->osid);
2004 audit_copy_fcaps(name, dentry);
2008 * audit_inode - store the inode and device from a lookup
2009 * @name: name being audited
2010 * @dentry: dentry being audited
2012 * Called from fs/namei.c:path_lookup().
2014 void __audit_inode(const char *name, const struct dentry *dentry)
2016 struct audit_context *context = current->audit_context;
2017 const struct inode *inode = dentry->d_inode;
2018 struct audit_names *n;
2020 if (!context->in_syscall)
2023 list_for_each_entry_reverse(n, &context->names_list, list) {
2024 if (n->name && (n->name == name))
2028 /* unable to find the name from a previous getname() */
2029 n = audit_alloc_name(context);
2033 handle_path(dentry);
2034 audit_copy_inode(n, dentry, inode);
2038 * audit_inode_child - collect inode info for created/removed objects
2039 * @dentry: dentry being audited
2040 * @parent: inode of dentry parent
2042 * For syscalls that create or remove filesystem objects, audit_inode
2043 * can only collect information for the filesystem object's parent.
2044 * This call updates the audit context with the child's information.
2045 * Syscalls that create a new filesystem object must be hooked after
2046 * the object is created. Syscalls that remove a filesystem object
2047 * must be hooked prior, in order to capture the target inode during
2048 * unsuccessful attempts.
2050 void __audit_inode_child(const struct dentry *dentry,
2051 const struct inode *parent)
2053 struct audit_context *context = current->audit_context;
2054 const char *found_parent = NULL, *found_child = NULL;
2055 const struct inode *inode = dentry->d_inode;
2056 const char *dname = dentry->d_name.name;
2057 struct audit_names *n;
2060 if (!context->in_syscall)
2066 /* parent is more likely, look for it first */
2067 list_for_each_entry(n, &context->names_list, list) {
2071 if (n->ino == parent->i_ino &&
2072 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2073 n->name_len = dirlen; /* update parent data in place */
2074 found_parent = n->name;
2079 /* no matching parent, look for matching child */
2080 list_for_each_entry(n, &context->names_list, list) {
2084 /* strcmp() is the more likely scenario */
2085 if (!strcmp(dname, n->name) ||
2086 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2088 audit_copy_inode(n, NULL, inode);
2090 n->ino = (unsigned long)-1;
2091 found_child = n->name;
2097 if (!found_parent) {
2098 n = audit_alloc_name(context);
2101 audit_copy_inode(n, NULL, parent);
2105 n = audit_alloc_name(context);
2109 /* Re-use the name belonging to the slot for a matching parent
2110 * directory. All names for this context are relinquished in
2111 * audit_free_names() */
2113 n->name = found_parent;
2114 n->name_len = AUDIT_NAME_FULL;
2115 /* don't call __putname() */
2116 n->name_put = false;
2120 audit_copy_inode(n, NULL, inode);
2123 EXPORT_SYMBOL_GPL(__audit_inode_child);
2126 * auditsc_get_stamp - get local copies of audit_context values
2127 * @ctx: audit_context for the task
2128 * @t: timespec to store time recorded in the audit_context
2129 * @serial: serial value that is recorded in the audit_context
2131 * Also sets the context as auditable.
2133 int auditsc_get_stamp(struct audit_context *ctx,
2134 struct timespec *t, unsigned int *serial)
2136 if (!ctx->in_syscall)
2139 ctx->serial = audit_serial();
2140 t->tv_sec = ctx->ctime.tv_sec;
2141 t->tv_nsec = ctx->ctime.tv_nsec;
2142 *serial = ctx->serial;
2145 ctx->current_state = AUDIT_RECORD_CONTEXT;
2150 /* global counter which is incremented every time something logs in */
2151 static atomic_t session_id = ATOMIC_INIT(0);
2154 * audit_set_loginuid - set a task's audit_context loginuid
2155 * @task: task whose audit context is being modified
2156 * @loginuid: loginuid value
2160 * Called (set) from fs/proc/base.c::proc_loginuid_write().
2162 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
2164 unsigned int sessionid = atomic_inc_return(&session_id);
2165 struct audit_context *context = task->audit_context;
2167 if (context && context->in_syscall) {
2168 struct audit_buffer *ab;
2170 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2172 audit_log_format(ab, "login pid=%d uid=%u "
2173 "old auid=%u new auid=%u"
2174 " old ses=%u new ses=%u",
2175 task->pid, task_uid(task),
2176 task->loginuid, loginuid,
2177 task->sessionid, sessionid);
2181 task->sessionid = sessionid;
2182 task->loginuid = loginuid;
2187 * __audit_mq_open - record audit data for a POSIX MQ open
2190 * @attr: queue attributes
2193 void __audit_mq_open(int oflag, umode_t mode, struct mq_attr *attr)
2195 struct audit_context *context = current->audit_context;
2198 memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2200 memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2202 context->mq_open.oflag = oflag;
2203 context->mq_open.mode = mode;
2205 context->type = AUDIT_MQ_OPEN;
2209 * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2210 * @mqdes: MQ descriptor
2211 * @msg_len: Message length
2212 * @msg_prio: Message priority
2213 * @abs_timeout: Message timeout in absolute time
2216 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2217 const struct timespec *abs_timeout)
2219 struct audit_context *context = current->audit_context;
2220 struct timespec *p = &context->mq_sendrecv.abs_timeout;
2223 memcpy(p, abs_timeout, sizeof(struct timespec));
2225 memset(p, 0, sizeof(struct timespec));
2227 context->mq_sendrecv.mqdes = mqdes;
2228 context->mq_sendrecv.msg_len = msg_len;
2229 context->mq_sendrecv.msg_prio = msg_prio;
2231 context->type = AUDIT_MQ_SENDRECV;
2235 * __audit_mq_notify - record audit data for a POSIX MQ notify
2236 * @mqdes: MQ descriptor
2237 * @notification: Notification event
2241 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2243 struct audit_context *context = current->audit_context;
2246 context->mq_notify.sigev_signo = notification->sigev_signo;
2248 context->mq_notify.sigev_signo = 0;
2250 context->mq_notify.mqdes = mqdes;
2251 context->type = AUDIT_MQ_NOTIFY;
2255 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2256 * @mqdes: MQ descriptor
2260 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2262 struct audit_context *context = current->audit_context;
2263 context->mq_getsetattr.mqdes = mqdes;
2264 context->mq_getsetattr.mqstat = *mqstat;
2265 context->type = AUDIT_MQ_GETSETATTR;
2269 * audit_ipc_obj - record audit data for ipc object
2270 * @ipcp: ipc permissions
2273 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2275 struct audit_context *context = current->audit_context;
2276 context->ipc.uid = ipcp->uid;
2277 context->ipc.gid = ipcp->gid;
2278 context->ipc.mode = ipcp->mode;
2279 context->ipc.has_perm = 0;
2280 security_ipc_getsecid(ipcp, &context->ipc.osid);
2281 context->type = AUDIT_IPC;
2285 * audit_ipc_set_perm - record audit data for new ipc permissions
2286 * @qbytes: msgq bytes
2287 * @uid: msgq user id
2288 * @gid: msgq group id
2289 * @mode: msgq mode (permissions)
2291 * Called only after audit_ipc_obj().
2293 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, umode_t mode)
2295 struct audit_context *context = current->audit_context;
2297 context->ipc.qbytes = qbytes;
2298 context->ipc.perm_uid = uid;
2299 context->ipc.perm_gid = gid;
2300 context->ipc.perm_mode = mode;
2301 context->ipc.has_perm = 1;
2304 int audit_bprm(struct linux_binprm *bprm)
2306 struct audit_aux_data_execve *ax;
2307 struct audit_context *context = current->audit_context;
2309 if (likely(!audit_enabled || !context || context->dummy))
2312 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2316 ax->argc = bprm->argc;
2317 ax->envc = bprm->envc;
2319 ax->d.type = AUDIT_EXECVE;
2320 ax->d.next = context->aux;
2321 context->aux = (void *)ax;
2327 * audit_socketcall - record audit data for sys_socketcall
2328 * @nargs: number of args
2332 void audit_socketcall(int nargs, unsigned long *args)
2334 struct audit_context *context = current->audit_context;
2336 if (likely(!context || context->dummy))
2339 context->type = AUDIT_SOCKETCALL;
2340 context->socketcall.nargs = nargs;
2341 memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2345 * __audit_fd_pair - record audit data for pipe and socketpair
2346 * @fd1: the first file descriptor
2347 * @fd2: the second file descriptor
2350 void __audit_fd_pair(int fd1, int fd2)
2352 struct audit_context *context = current->audit_context;
2353 context->fds[0] = fd1;
2354 context->fds[1] = fd2;
2358 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2359 * @len: data length in user space
2360 * @a: data address in kernel space
2362 * Returns 0 for success or NULL context or < 0 on error.
2364 int audit_sockaddr(int len, void *a)
2366 struct audit_context *context = current->audit_context;
2368 if (likely(!context || context->dummy))
2371 if (!context->sockaddr) {
2372 void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2375 context->sockaddr = p;
2378 context->sockaddr_len = len;
2379 memcpy(context->sockaddr, a, len);
2383 void __audit_ptrace(struct task_struct *t)
2385 struct audit_context *context = current->audit_context;
2387 context->target_pid = t->pid;
2388 context->target_auid = audit_get_loginuid(t);
2389 context->target_uid = task_uid(t);
2390 context->target_sessionid = audit_get_sessionid(t);
2391 security_task_getsecid(t, &context->target_sid);
2392 memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2396 * audit_signal_info - record signal info for shutting down audit subsystem
2397 * @sig: signal value
2398 * @t: task being signaled
2400 * If the audit subsystem is being terminated, record the task (pid)
2401 * and uid that is doing that.
2403 int __audit_signal_info(int sig, struct task_struct *t)
2405 struct audit_aux_data_pids *axp;
2406 struct task_struct *tsk = current;
2407 struct audit_context *ctx = tsk->audit_context;
2408 uid_t uid = current_uid(), t_uid = task_uid(t);
2410 if (audit_pid && t->tgid == audit_pid) {
2411 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2412 audit_sig_pid = tsk->pid;
2413 if (tsk->loginuid != -1)
2414 audit_sig_uid = tsk->loginuid;
2416 audit_sig_uid = uid;
2417 security_task_getsecid(tsk, &audit_sig_sid);
2419 if (!audit_signals || audit_dummy_context())
2423 /* optimize the common case by putting first signal recipient directly
2424 * in audit_context */
2425 if (!ctx->target_pid) {
2426 ctx->target_pid = t->tgid;
2427 ctx->target_auid = audit_get_loginuid(t);
2428 ctx->target_uid = t_uid;
2429 ctx->target_sessionid = audit_get_sessionid(t);
2430 security_task_getsecid(t, &ctx->target_sid);
2431 memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2435 axp = (void *)ctx->aux_pids;
2436 if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2437 axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2441 axp->d.type = AUDIT_OBJ_PID;
2442 axp->d.next = ctx->aux_pids;
2443 ctx->aux_pids = (void *)axp;
2445 BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2447 axp->target_pid[axp->pid_count] = t->tgid;
2448 axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2449 axp->target_uid[axp->pid_count] = t_uid;
2450 axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2451 security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2452 memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2459 * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2460 * @bprm: pointer to the bprm being processed
2461 * @new: the proposed new credentials
2462 * @old: the old credentials
2464 * Simply check if the proc already has the caps given by the file and if not
2465 * store the priv escalation info for later auditing at the end of the syscall
2469 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2470 const struct cred *new, const struct cred *old)
2472 struct audit_aux_data_bprm_fcaps *ax;
2473 struct audit_context *context = current->audit_context;
2474 struct cpu_vfs_cap_data vcaps;
2475 struct dentry *dentry;
2477 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2481 ax->d.type = AUDIT_BPRM_FCAPS;
2482 ax->d.next = context->aux;
2483 context->aux = (void *)ax;
2485 dentry = dget(bprm->file->f_dentry);
2486 get_vfs_caps_from_disk(dentry, &vcaps);
2489 ax->fcap.permitted = vcaps.permitted;
2490 ax->fcap.inheritable = vcaps.inheritable;
2491 ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2492 ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2494 ax->old_pcap.permitted = old->cap_permitted;
2495 ax->old_pcap.inheritable = old->cap_inheritable;
2496 ax->old_pcap.effective = old->cap_effective;
2498 ax->new_pcap.permitted = new->cap_permitted;
2499 ax->new_pcap.inheritable = new->cap_inheritable;
2500 ax->new_pcap.effective = new->cap_effective;
2505 * __audit_log_capset - store information about the arguments to the capset syscall
2506 * @pid: target pid of the capset call
2507 * @new: the new credentials
2508 * @old: the old (current) credentials
2510 * Record the aguments userspace sent to sys_capset for later printing by the
2511 * audit system if applicable
2513 void __audit_log_capset(pid_t pid,
2514 const struct cred *new, const struct cred *old)
2516 struct audit_context *context = current->audit_context;
2517 context->capset.pid = pid;
2518 context->capset.cap.effective = new->cap_effective;
2519 context->capset.cap.inheritable = new->cap_effective;
2520 context->capset.cap.permitted = new->cap_permitted;
2521 context->type = AUDIT_CAPSET;
2524 void __audit_mmap_fd(int fd, int flags)
2526 struct audit_context *context = current->audit_context;
2527 context->mmap.fd = fd;
2528 context->mmap.flags = flags;
2529 context->type = AUDIT_MMAP;
2532 static void audit_log_abend(struct audit_buffer *ab, char *reason, long signr)
2536 unsigned int sessionid;
2538 auid = audit_get_loginuid(current);
2539 sessionid = audit_get_sessionid(current);
2540 current_uid_gid(&uid, &gid);
2542 audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2543 auid, uid, gid, sessionid);
2544 audit_log_task_context(ab);
2545 audit_log_format(ab, " pid=%d comm=", current->pid);
2546 audit_log_untrustedstring(ab, current->comm);
2547 audit_log_format(ab, " reason=");
2548 audit_log_string(ab, reason);
2549 audit_log_format(ab, " sig=%ld", signr);
2552 * audit_core_dumps - record information about processes that end abnormally
2553 * @signr: signal value
2555 * If a process ends with a core dump, something fishy is going on and we
2556 * should record the event for investigation.
2558 void audit_core_dumps(long signr)
2560 struct audit_buffer *ab;
2565 if (signr == SIGQUIT) /* don't care for those */
2568 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2569 audit_log_abend(ab, "memory violation", signr);
2573 void __audit_seccomp(unsigned long syscall)
2575 struct audit_buffer *ab;
2577 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2578 audit_log_abend(ab, "seccomp", SIGKILL);
2579 audit_log_format(ab, " syscall=%ld", syscall);
2583 struct list_head *audit_killed_trees(void)
2585 struct audit_context *ctx = current->audit_context;
2586 if (likely(!ctx || !ctx->in_syscall))
2588 return &ctx->killed_trees;