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/mount.h>
53 #include <linux/socket.h>
54 #include <linux/mqueue.h>
55 #include <linux/audit.h>
56 #include <linux/personality.h>
57 #include <linux/time.h>
58 #include <linux/netlink.h>
59 #include <linux/compiler.h>
60 #include <asm/unistd.h>
61 #include <linux/security.h>
62 #include <linux/list.h>
63 #include <linux/tty.h>
64 #include <linux/binfmts.h>
65 #include <linux/highmem.h>
66 #include <linux/syscalls.h>
67 #include <linux/inotify.h>
68 #include <linux/capability.h>
72 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
73 * for saving names from getname(). */
74 #define AUDIT_NAMES 20
76 /* Indicates that audit should log the full pathname. */
77 #define AUDIT_NAME_FULL -1
79 /* no execve audit message should be longer than this (userspace limits) */
80 #define MAX_EXECVE_AUDIT_LEN 7500
82 /* number of audit rules */
85 /* determines whether we collect data for signals sent */
88 struct audit_cap_data {
89 kernel_cap_t permitted;
90 kernel_cap_t inheritable;
92 unsigned int fE; /* effective bit of a file capability */
93 kernel_cap_t effective; /* effective set of a process */
97 /* When fs/namei.c:getname() is called, we store the pointer in name and
98 * we don't let putname() free it (instead we free all of the saved
99 * pointers at syscall exit time).
101 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
104 int name_len; /* number of name's characters to log */
105 unsigned name_put; /* call __putname() for this name */
113 struct audit_cap_data fcap;
114 unsigned int fcap_ver;
117 struct audit_aux_data {
118 struct audit_aux_data *next;
122 #define AUDIT_AUX_IPCPERM 0
124 /* Number of target pids per aux struct. */
125 #define AUDIT_AUX_PIDS 16
127 struct audit_aux_data_execve {
128 struct audit_aux_data d;
131 struct mm_struct *mm;
134 struct audit_aux_data_pids {
135 struct audit_aux_data d;
136 pid_t target_pid[AUDIT_AUX_PIDS];
137 uid_t target_auid[AUDIT_AUX_PIDS];
138 uid_t target_uid[AUDIT_AUX_PIDS];
139 unsigned int target_sessionid[AUDIT_AUX_PIDS];
140 u32 target_sid[AUDIT_AUX_PIDS];
141 char target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
145 struct audit_aux_data_bprm_fcaps {
146 struct audit_aux_data d;
147 struct audit_cap_data fcap;
148 unsigned int fcap_ver;
149 struct audit_cap_data old_pcap;
150 struct audit_cap_data new_pcap;
153 struct audit_aux_data_capset {
154 struct audit_aux_data d;
156 struct audit_cap_data cap;
159 struct audit_tree_refs {
160 struct audit_tree_refs *next;
161 struct audit_chunk *c[31];
164 /* The per-task audit context. */
165 struct audit_context {
166 int dummy; /* must be the first element */
167 int in_syscall; /* 1 if task is in a syscall */
168 enum audit_state state;
169 unsigned int serial; /* serial number for record */
170 struct timespec ctime; /* time of syscall entry */
171 int major; /* syscall number */
172 unsigned long argv[4]; /* syscall arguments */
173 int return_valid; /* return code is valid */
174 long return_code;/* syscall return code */
175 int auditable; /* 1 if record should be written */
177 struct audit_names names[AUDIT_NAMES];
178 char * filterkey; /* key for rule that triggered record */
180 struct audit_context *previous; /* For nested syscalls */
181 struct audit_aux_data *aux;
182 struct audit_aux_data *aux_pids;
183 struct sockaddr_storage *sockaddr;
185 /* Save things to print about task_struct */
187 uid_t uid, euid, suid, fsuid;
188 gid_t gid, egid, sgid, fsgid;
189 unsigned long personality;
195 unsigned int target_sessionid;
197 char target_comm[TASK_COMM_LEN];
199 struct audit_tree_refs *trees, *first_trees;
217 unsigned long qbytes;
221 struct mq_attr mqstat;
230 unsigned int msg_prio;
231 struct timespec abs_timeout;
240 struct audit_cap_data cap;
251 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
252 static inline int open_arg(int flags, int mask)
254 int n = ACC_MODE(flags);
255 if (flags & (O_TRUNC | O_CREAT))
256 n |= AUDIT_PERM_WRITE;
260 static int audit_match_perm(struct audit_context *ctx, int mask)
267 switch (audit_classify_syscall(ctx->arch, n)) {
269 if ((mask & AUDIT_PERM_WRITE) &&
270 audit_match_class(AUDIT_CLASS_WRITE, n))
272 if ((mask & AUDIT_PERM_READ) &&
273 audit_match_class(AUDIT_CLASS_READ, n))
275 if ((mask & AUDIT_PERM_ATTR) &&
276 audit_match_class(AUDIT_CLASS_CHATTR, n))
279 case 1: /* 32bit on biarch */
280 if ((mask & AUDIT_PERM_WRITE) &&
281 audit_match_class(AUDIT_CLASS_WRITE_32, n))
283 if ((mask & AUDIT_PERM_READ) &&
284 audit_match_class(AUDIT_CLASS_READ_32, n))
286 if ((mask & AUDIT_PERM_ATTR) &&
287 audit_match_class(AUDIT_CLASS_CHATTR_32, n))
291 return mask & ACC_MODE(ctx->argv[1]);
293 return mask & ACC_MODE(ctx->argv[2]);
294 case 4: /* socketcall */
295 return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
297 return mask & AUDIT_PERM_EXEC;
303 static int audit_match_filetype(struct audit_context *ctx, int which)
305 unsigned index = which & ~S_IFMT;
306 mode_t mode = which & S_IFMT;
311 if (index >= ctx->name_count)
313 if (ctx->names[index].ino == -1)
315 if ((ctx->names[index].mode ^ mode) & S_IFMT)
321 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
322 * ->first_trees points to its beginning, ->trees - to the current end of data.
323 * ->tree_count is the number of free entries in array pointed to by ->trees.
324 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
325 * "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
326 * it's going to remain 1-element for almost any setup) until we free context itself.
327 * References in it _are_ dropped - at the same time we free/drop aux stuff.
330 #ifdef CONFIG_AUDIT_TREE
331 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
333 struct audit_tree_refs *p = ctx->trees;
334 int left = ctx->tree_count;
336 p->c[--left] = chunk;
337 ctx->tree_count = left;
346 ctx->tree_count = 30;
352 static int grow_tree_refs(struct audit_context *ctx)
354 struct audit_tree_refs *p = ctx->trees;
355 ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
361 p->next = ctx->trees;
363 ctx->first_trees = ctx->trees;
364 ctx->tree_count = 31;
369 static void unroll_tree_refs(struct audit_context *ctx,
370 struct audit_tree_refs *p, int count)
372 #ifdef CONFIG_AUDIT_TREE
373 struct audit_tree_refs *q;
376 /* we started with empty chain */
377 p = ctx->first_trees;
379 /* if the very first allocation has failed, nothing to do */
384 for (q = p; q != ctx->trees; q = q->next, n = 31) {
386 audit_put_chunk(q->c[n]);
390 while (n-- > ctx->tree_count) {
391 audit_put_chunk(q->c[n]);
395 ctx->tree_count = count;
399 static void free_tree_refs(struct audit_context *ctx)
401 struct audit_tree_refs *p, *q;
402 for (p = ctx->first_trees; p; p = q) {
408 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
410 #ifdef CONFIG_AUDIT_TREE
411 struct audit_tree_refs *p;
416 for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
417 for (n = 0; n < 31; n++)
418 if (audit_tree_match(p->c[n], tree))
423 for (n = ctx->tree_count; n < 31; n++)
424 if (audit_tree_match(p->c[n], tree))
431 /* Determine if any context name data matches a rule's watch data */
432 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
434 static int audit_filter_rules(struct task_struct *tsk,
435 struct audit_krule *rule,
436 struct audit_context *ctx,
437 struct audit_names *name,
438 enum audit_state *state)
440 const struct cred *cred = get_task_cred(tsk);
441 int i, j, need_sid = 1;
444 for (i = 0; i < rule->field_count; i++) {
445 struct audit_field *f = &rule->fields[i];
450 result = audit_comparator(tsk->pid, f->op, f->val);
455 ctx->ppid = sys_getppid();
456 result = audit_comparator(ctx->ppid, f->op, f->val);
460 result = audit_comparator(cred->uid, f->op, f->val);
463 result = audit_comparator(cred->euid, f->op, f->val);
466 result = audit_comparator(cred->suid, f->op, f->val);
469 result = audit_comparator(cred->fsuid, f->op, f->val);
472 result = audit_comparator(cred->gid, f->op, f->val);
475 result = audit_comparator(cred->egid, f->op, f->val);
478 result = audit_comparator(cred->sgid, f->op, f->val);
481 result = audit_comparator(cred->fsgid, f->op, f->val);
484 result = audit_comparator(tsk->personality, f->op, f->val);
488 result = audit_comparator(ctx->arch, f->op, f->val);
492 if (ctx && ctx->return_valid)
493 result = audit_comparator(ctx->return_code, f->op, f->val);
496 if (ctx && ctx->return_valid) {
498 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
500 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
505 result = audit_comparator(MAJOR(name->dev),
508 for (j = 0; j < ctx->name_count; j++) {
509 if (audit_comparator(MAJOR(ctx->names[j].dev), f->op, f->val)) {
518 result = audit_comparator(MINOR(name->dev),
521 for (j = 0; j < ctx->name_count; j++) {
522 if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
531 result = (name->ino == f->val);
533 for (j = 0; j < ctx->name_count; j++) {
534 if (audit_comparator(ctx->names[j].ino, f->op, f->val)) {
542 if (name && rule->watch->ino != (unsigned long)-1)
543 result = (name->dev == rule->watch->dev &&
544 name->ino == rule->watch->ino);
548 result = match_tree_refs(ctx, rule->tree);
553 result = audit_comparator(tsk->loginuid, f->op, f->val);
555 case AUDIT_SUBJ_USER:
556 case AUDIT_SUBJ_ROLE:
557 case AUDIT_SUBJ_TYPE:
560 /* NOTE: this may return negative values indicating
561 a temporary error. We simply treat this as a
562 match for now to avoid losing information that
563 may be wanted. An error message will also be
567 security_task_getsecid(tsk, &sid);
570 result = security_audit_rule_match(sid, f->type,
579 case AUDIT_OBJ_LEV_LOW:
580 case AUDIT_OBJ_LEV_HIGH:
581 /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
584 /* Find files that match */
586 result = security_audit_rule_match(
587 name->osid, f->type, f->op,
590 for (j = 0; j < ctx->name_count; j++) {
591 if (security_audit_rule_match(
600 /* Find ipc objects that match */
601 if (!ctx || ctx->type != AUDIT_IPC)
603 if (security_audit_rule_match(ctx->ipc.osid,
614 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
616 case AUDIT_FILTERKEY:
617 /* ignore this field for filtering */
621 result = audit_match_perm(ctx, f->val);
624 result = audit_match_filetype(ctx, f->val);
633 if (rule->filterkey && ctx)
634 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
635 switch (rule->action) {
636 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
637 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
643 /* At process creation time, we can determine if system-call auditing is
644 * completely disabled for this task. Since we only have the task
645 * structure at this point, we can only check uid and gid.
647 static enum audit_state audit_filter_task(struct task_struct *tsk)
649 struct audit_entry *e;
650 enum audit_state state;
653 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
654 if (audit_filter_rules(tsk, &e->rule, NULL, NULL, &state)) {
660 return AUDIT_BUILD_CONTEXT;
663 /* At syscall entry and exit time, this filter is called if the
664 * audit_state is not low enough that auditing cannot take place, but is
665 * also not high enough that we already know we have to write an audit
666 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
668 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
669 struct audit_context *ctx,
670 struct list_head *list)
672 struct audit_entry *e;
673 enum audit_state state;
675 if (audit_pid && tsk->tgid == audit_pid)
676 return AUDIT_DISABLED;
679 if (!list_empty(list)) {
680 int word = AUDIT_WORD(ctx->major);
681 int bit = AUDIT_BIT(ctx->major);
683 list_for_each_entry_rcu(e, list, list) {
684 if ((e->rule.mask[word] & bit) == bit &&
685 audit_filter_rules(tsk, &e->rule, ctx, NULL,
693 return AUDIT_BUILD_CONTEXT;
696 /* At syscall exit time, this filter is called if any audit_names[] have been
697 * collected during syscall processing. We only check rules in sublists at hash
698 * buckets applicable to the inode numbers in audit_names[].
699 * Regarding audit_state, same rules apply as for audit_filter_syscall().
701 enum audit_state audit_filter_inodes(struct task_struct *tsk,
702 struct audit_context *ctx)
705 struct audit_entry *e;
706 enum audit_state state;
708 if (audit_pid && tsk->tgid == audit_pid)
709 return AUDIT_DISABLED;
712 for (i = 0; i < ctx->name_count; i++) {
713 int word = AUDIT_WORD(ctx->major);
714 int bit = AUDIT_BIT(ctx->major);
715 struct audit_names *n = &ctx->names[i];
716 int h = audit_hash_ino((u32)n->ino);
717 struct list_head *list = &audit_inode_hash[h];
719 if (list_empty(list))
722 list_for_each_entry_rcu(e, list, list) {
723 if ((e->rule.mask[word] & bit) == bit &&
724 audit_filter_rules(tsk, &e->rule, ctx, n, &state)) {
731 return AUDIT_BUILD_CONTEXT;
734 void audit_set_auditable(struct audit_context *ctx)
739 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
743 struct audit_context *context = tsk->audit_context;
745 if (likely(!context))
747 context->return_valid = return_valid;
750 * we need to fix up the return code in the audit logs if the actual
751 * return codes are later going to be fixed up by the arch specific
754 * This is actually a test for:
755 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
756 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
758 * but is faster than a bunch of ||
760 if (unlikely(return_code <= -ERESTARTSYS) &&
761 (return_code >= -ERESTART_RESTARTBLOCK) &&
762 (return_code != -ENOIOCTLCMD))
763 context->return_code = -EINTR;
765 context->return_code = return_code;
767 if (context->in_syscall && !context->dummy && !context->auditable) {
768 enum audit_state state;
770 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
771 if (state == AUDIT_RECORD_CONTEXT) {
772 context->auditable = 1;
776 state = audit_filter_inodes(tsk, context);
777 if (state == AUDIT_RECORD_CONTEXT)
778 context->auditable = 1;
784 tsk->audit_context = NULL;
788 static inline void audit_free_names(struct audit_context *context)
793 if (context->auditable
794 ||context->put_count + context->ino_count != context->name_count) {
795 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
796 " name_count=%d put_count=%d"
797 " ino_count=%d [NOT freeing]\n",
799 context->serial, context->major, context->in_syscall,
800 context->name_count, context->put_count,
802 for (i = 0; i < context->name_count; i++) {
803 printk(KERN_ERR "names[%d] = %p = %s\n", i,
804 context->names[i].name,
805 context->names[i].name ?: "(null)");
812 context->put_count = 0;
813 context->ino_count = 0;
816 for (i = 0; i < context->name_count; i++) {
817 if (context->names[i].name && context->names[i].name_put)
818 __putname(context->names[i].name);
820 context->name_count = 0;
821 path_put(&context->pwd);
822 context->pwd.dentry = NULL;
823 context->pwd.mnt = NULL;
826 static inline void audit_free_aux(struct audit_context *context)
828 struct audit_aux_data *aux;
830 while ((aux = context->aux)) {
831 context->aux = aux->next;
834 while ((aux = context->aux_pids)) {
835 context->aux_pids = aux->next;
840 static inline void audit_zero_context(struct audit_context *context,
841 enum audit_state state)
843 memset(context, 0, sizeof(*context));
844 context->state = state;
847 static inline struct audit_context *audit_alloc_context(enum audit_state state)
849 struct audit_context *context;
851 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
853 audit_zero_context(context, state);
858 * audit_alloc - allocate an audit context block for a task
861 * Filter on the task information and allocate a per-task audit context
862 * if necessary. Doing so turns on system call auditing for the
863 * specified task. This is called from copy_process, so no lock is
866 int audit_alloc(struct task_struct *tsk)
868 struct audit_context *context;
869 enum audit_state state;
871 if (likely(!audit_ever_enabled))
872 return 0; /* Return if not auditing. */
874 state = audit_filter_task(tsk);
875 if (likely(state == AUDIT_DISABLED))
878 if (!(context = audit_alloc_context(state))) {
879 audit_log_lost("out of memory in audit_alloc");
883 tsk->audit_context = context;
884 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
888 static inline void audit_free_context(struct audit_context *context)
890 struct audit_context *previous;
894 previous = context->previous;
895 if (previous || (count && count < 10)) {
897 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
898 " freeing multiple contexts (%d)\n",
899 context->serial, context->major,
900 context->name_count, count);
902 audit_free_names(context);
903 unroll_tree_refs(context, NULL, 0);
904 free_tree_refs(context);
905 audit_free_aux(context);
906 kfree(context->filterkey);
907 kfree(context->sockaddr);
912 printk(KERN_ERR "audit: freed %d contexts\n", count);
915 void audit_log_task_context(struct audit_buffer *ab)
922 security_task_getsecid(current, &sid);
926 error = security_secid_to_secctx(sid, &ctx, &len);
928 if (error != -EINVAL)
933 audit_log_format(ab, " subj=%s", ctx);
934 security_release_secctx(ctx, len);
938 audit_panic("error in audit_log_task_context");
942 EXPORT_SYMBOL(audit_log_task_context);
944 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
946 char name[sizeof(tsk->comm)];
947 struct mm_struct *mm = tsk->mm;
948 struct vm_area_struct *vma;
952 get_task_comm(name, tsk);
953 audit_log_format(ab, " comm=");
954 audit_log_untrustedstring(ab, name);
957 down_read(&mm->mmap_sem);
960 if ((vma->vm_flags & VM_EXECUTABLE) &&
962 audit_log_d_path(ab, "exe=",
963 &vma->vm_file->f_path);
968 up_read(&mm->mmap_sem);
970 audit_log_task_context(ab);
973 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
974 uid_t auid, uid_t uid, unsigned int sessionid,
977 struct audit_buffer *ab;
982 ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
986 audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
988 if (security_secid_to_secctx(sid, &ctx, &len)) {
989 audit_log_format(ab, " obj=(none)");
992 audit_log_format(ab, " obj=%s", ctx);
993 security_release_secctx(ctx, len);
995 audit_log_format(ab, " ocomm=");
996 audit_log_untrustedstring(ab, comm);
1003 * to_send and len_sent accounting are very loose estimates. We aren't
1004 * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1005 * within about 500 bytes (next page boundry)
1007 * why snprintf? an int is up to 12 digits long. if we just assumed when
1008 * logging that a[%d]= was going to be 16 characters long we would be wasting
1009 * space in every audit message. In one 7500 byte message we can log up to
1010 * about 1000 min size arguments. That comes down to about 50% waste of space
1011 * if we didn't do the snprintf to find out how long arg_num_len was.
1013 static int audit_log_single_execve_arg(struct audit_context *context,
1014 struct audit_buffer **ab,
1017 const char __user *p,
1020 char arg_num_len_buf[12];
1021 const char __user *tmp_p = p;
1022 /* how many digits are in arg_num? 3 is the length of a=\n */
1023 size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 3;
1024 size_t len, len_left, to_send;
1025 size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1026 unsigned int i, has_cntl = 0, too_long = 0;
1029 /* strnlen_user includes the null we don't want to send */
1030 len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1033 * We just created this mm, if we can't find the strings
1034 * we just copied into it something is _very_ wrong. Similar
1035 * for strings that are too long, we should not have created
1038 if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1040 send_sig(SIGKILL, current, 0);
1044 /* walk the whole argument looking for non-ascii chars */
1046 if (len_left > MAX_EXECVE_AUDIT_LEN)
1047 to_send = MAX_EXECVE_AUDIT_LEN;
1050 ret = copy_from_user(buf, tmp_p, to_send);
1052 * There is no reason for this copy to be short. We just
1053 * copied them here, and the mm hasn't been exposed to user-
1058 send_sig(SIGKILL, current, 0);
1061 buf[to_send] = '\0';
1062 has_cntl = audit_string_contains_control(buf, to_send);
1065 * hex messages get logged as 2 bytes, so we can only
1066 * send half as much in each message
1068 max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1071 len_left -= to_send;
1073 } while (len_left > 0);
1077 if (len > max_execve_audit_len)
1080 /* rewalk the argument actually logging the message */
1081 for (i = 0; len_left > 0; i++) {
1084 if (len_left > max_execve_audit_len)
1085 to_send = max_execve_audit_len;
1089 /* do we have space left to send this argument in this ab? */
1090 room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1092 room_left -= (to_send * 2);
1094 room_left -= to_send;
1095 if (room_left < 0) {
1098 *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1104 * first record needs to say how long the original string was
1105 * so we can be sure nothing was lost.
1107 if ((i == 0) && (too_long))
1108 audit_log_format(*ab, "a%d_len=%zu ", arg_num,
1109 has_cntl ? 2*len : len);
1112 * normally arguments are small enough to fit and we already
1113 * filled buf above when we checked for control characters
1114 * so don't bother with another copy_from_user
1116 if (len >= max_execve_audit_len)
1117 ret = copy_from_user(buf, p, to_send);
1122 send_sig(SIGKILL, current, 0);
1125 buf[to_send] = '\0';
1127 /* actually log it */
1128 audit_log_format(*ab, "a%d", arg_num);
1130 audit_log_format(*ab, "[%d]", i);
1131 audit_log_format(*ab, "=");
1133 audit_log_n_hex(*ab, buf, to_send);
1135 audit_log_format(*ab, "\"%s\"", buf);
1136 audit_log_format(*ab, "\n");
1139 len_left -= to_send;
1140 *len_sent += arg_num_len;
1142 *len_sent += to_send * 2;
1144 *len_sent += to_send;
1146 /* include the null we didn't log */
1150 static void audit_log_execve_info(struct audit_context *context,
1151 struct audit_buffer **ab,
1152 struct audit_aux_data_execve *axi)
1155 size_t len, len_sent = 0;
1156 const char __user *p;
1159 if (axi->mm != current->mm)
1160 return; /* execve failed, no additional info */
1162 p = (const char __user *)axi->mm->arg_start;
1164 audit_log_format(*ab, "argc=%d ", axi->argc);
1167 * we need some kernel buffer to hold the userspace args. Just
1168 * allocate one big one rather than allocating one of the right size
1169 * for every single argument inside audit_log_single_execve_arg()
1170 * should be <8k allocation so should be pretty safe.
1172 buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1174 audit_panic("out of memory for argv string\n");
1178 for (i = 0; i < axi->argc; i++) {
1179 len = audit_log_single_execve_arg(context, ab, i,
1188 static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1192 audit_log_format(ab, " %s=", prefix);
1193 CAP_FOR_EACH_U32(i) {
1194 audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1198 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1200 kernel_cap_t *perm = &name->fcap.permitted;
1201 kernel_cap_t *inh = &name->fcap.inheritable;
1204 if (!cap_isclear(*perm)) {
1205 audit_log_cap(ab, "cap_fp", perm);
1208 if (!cap_isclear(*inh)) {
1209 audit_log_cap(ab, "cap_fi", inh);
1214 audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1217 static void show_special(struct audit_context *context, int *call_panic)
1219 struct audit_buffer *ab;
1222 ab = audit_log_start(context, GFP_KERNEL, context->type);
1226 switch (context->type) {
1227 case AUDIT_SOCKETCALL: {
1228 int nargs = context->socketcall.nargs;
1229 audit_log_format(ab, "nargs=%d", nargs);
1230 for (i = 0; i < nargs; i++)
1231 audit_log_format(ab, " a%d=%lx", i,
1232 context->socketcall.args[i]);
1235 u32 osid = context->ipc.osid;
1237 audit_log_format(ab, "ouid=%u ogid=%u mode=%#o",
1238 context->ipc.uid, context->ipc.gid, context->ipc.mode);
1242 if (security_secid_to_secctx(osid, &ctx, &len)) {
1243 audit_log_format(ab, " osid=%u", osid);
1246 audit_log_format(ab, " obj=%s", ctx);
1247 security_release_secctx(ctx, len);
1250 if (context->ipc.has_perm) {
1252 ab = audit_log_start(context, GFP_KERNEL,
1253 AUDIT_IPC_SET_PERM);
1254 audit_log_format(ab,
1255 "qbytes=%lx ouid=%u ogid=%u mode=%#o",
1256 context->ipc.qbytes,
1257 context->ipc.perm_uid,
1258 context->ipc.perm_gid,
1259 context->ipc.perm_mode);
1264 case AUDIT_MQ_OPEN: {
1265 audit_log_format(ab,
1266 "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
1267 "mq_msgsize=%ld mq_curmsgs=%ld",
1268 context->mq_open.oflag, context->mq_open.mode,
1269 context->mq_open.attr.mq_flags,
1270 context->mq_open.attr.mq_maxmsg,
1271 context->mq_open.attr.mq_msgsize,
1272 context->mq_open.attr.mq_curmsgs);
1274 case AUDIT_MQ_SENDRECV: {
1275 audit_log_format(ab,
1276 "mqdes=%d msg_len=%zd msg_prio=%u "
1277 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1278 context->mq_sendrecv.mqdes,
1279 context->mq_sendrecv.msg_len,
1280 context->mq_sendrecv.msg_prio,
1281 context->mq_sendrecv.abs_timeout.tv_sec,
1282 context->mq_sendrecv.abs_timeout.tv_nsec);
1284 case AUDIT_MQ_NOTIFY: {
1285 audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1286 context->mq_notify.mqdes,
1287 context->mq_notify.sigev_signo);
1289 case AUDIT_MQ_GETSETATTR: {
1290 struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1291 audit_log_format(ab,
1292 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1294 context->mq_getsetattr.mqdes,
1295 attr->mq_flags, attr->mq_maxmsg,
1296 attr->mq_msgsize, attr->mq_curmsgs);
1298 case AUDIT_CAPSET: {
1299 audit_log_format(ab, "pid=%d", context->capset.pid);
1300 audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1301 audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1302 audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1308 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1310 const struct cred *cred;
1311 int i, call_panic = 0;
1312 struct audit_buffer *ab;
1313 struct audit_aux_data *aux;
1316 /* tsk == current */
1317 context->pid = tsk->pid;
1319 context->ppid = sys_getppid();
1320 cred = current_cred();
1321 context->uid = cred->uid;
1322 context->gid = cred->gid;
1323 context->euid = cred->euid;
1324 context->suid = cred->suid;
1325 context->fsuid = cred->fsuid;
1326 context->egid = cred->egid;
1327 context->sgid = cred->sgid;
1328 context->fsgid = cred->fsgid;
1329 context->personality = tsk->personality;
1331 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1333 return; /* audit_panic has been called */
1334 audit_log_format(ab, "arch=%x syscall=%d",
1335 context->arch, context->major);
1336 if (context->personality != PER_LINUX)
1337 audit_log_format(ab, " per=%lx", context->personality);
1338 if (context->return_valid)
1339 audit_log_format(ab, " success=%s exit=%ld",
1340 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1341 context->return_code);
1343 spin_lock_irq(&tsk->sighand->siglock);
1344 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1345 tty = tsk->signal->tty->name;
1348 spin_unlock_irq(&tsk->sighand->siglock);
1350 audit_log_format(ab,
1351 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1352 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1353 " euid=%u suid=%u fsuid=%u"
1354 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1359 context->name_count,
1365 context->euid, context->suid, context->fsuid,
1366 context->egid, context->sgid, context->fsgid, tty,
1370 audit_log_task_info(ab, tsk);
1371 if (context->filterkey) {
1372 audit_log_format(ab, " key=");
1373 audit_log_untrustedstring(ab, context->filterkey);
1375 audit_log_format(ab, " key=(null)");
1378 for (aux = context->aux; aux; aux = aux->next) {
1380 ab = audit_log_start(context, GFP_KERNEL, aux->type);
1382 continue; /* audit_panic has been called */
1384 switch (aux->type) {
1386 case AUDIT_EXECVE: {
1387 struct audit_aux_data_execve *axi = (void *)aux;
1388 audit_log_execve_info(context, &ab, axi);
1391 case AUDIT_BPRM_FCAPS: {
1392 struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1393 audit_log_format(ab, "fver=%x", axs->fcap_ver);
1394 audit_log_cap(ab, "fp", &axs->fcap.permitted);
1395 audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1396 audit_log_format(ab, " fe=%d", axs->fcap.fE);
1397 audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1398 audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1399 audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1400 audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1401 audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1402 audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1410 show_special(context, &call_panic);
1412 if (context->fds[0] >= 0) {
1413 ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1415 audit_log_format(ab, "fd0=%d fd1=%d",
1416 context->fds[0], context->fds[1]);
1421 if (context->sockaddr_len) {
1422 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1424 audit_log_format(ab, "saddr=");
1425 audit_log_n_hex(ab, (void *)context->sockaddr,
1426 context->sockaddr_len);
1431 for (aux = context->aux_pids; aux; aux = aux->next) {
1432 struct audit_aux_data_pids *axs = (void *)aux;
1434 for (i = 0; i < axs->pid_count; i++)
1435 if (audit_log_pid_context(context, axs->target_pid[i],
1436 axs->target_auid[i],
1438 axs->target_sessionid[i],
1440 axs->target_comm[i]))
1444 if (context->target_pid &&
1445 audit_log_pid_context(context, context->target_pid,
1446 context->target_auid, context->target_uid,
1447 context->target_sessionid,
1448 context->target_sid, context->target_comm))
1451 if (context->pwd.dentry && context->pwd.mnt) {
1452 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1454 audit_log_d_path(ab, "cwd=", &context->pwd);
1458 for (i = 0; i < context->name_count; i++) {
1459 struct audit_names *n = &context->names[i];
1461 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1463 continue; /* audit_panic has been called */
1465 audit_log_format(ab, "item=%d", i);
1468 switch(n->name_len) {
1469 case AUDIT_NAME_FULL:
1470 /* log the full path */
1471 audit_log_format(ab, " name=");
1472 audit_log_untrustedstring(ab, n->name);
1475 /* name was specified as a relative path and the
1476 * directory component is the cwd */
1477 audit_log_d_path(ab, " name=", &context->pwd);
1480 /* log the name's directory component */
1481 audit_log_format(ab, " name=");
1482 audit_log_n_untrustedstring(ab, n->name,
1486 audit_log_format(ab, " name=(null)");
1488 if (n->ino != (unsigned long)-1) {
1489 audit_log_format(ab, " inode=%lu"
1490 " dev=%02x:%02x mode=%#o"
1491 " ouid=%u ogid=%u rdev=%02x:%02x",
1504 if (security_secid_to_secctx(
1505 n->osid, &ctx, &len)) {
1506 audit_log_format(ab, " osid=%u", n->osid);
1509 audit_log_format(ab, " obj=%s", ctx);
1510 security_release_secctx(ctx, len);
1514 audit_log_fcaps(ab, n);
1519 /* Send end of event record to help user space know we are finished */
1520 ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1524 audit_panic("error converting sid to string");
1528 * audit_free - free a per-task audit context
1529 * @tsk: task whose audit context block to free
1531 * Called from copy_process and do_exit
1533 void audit_free(struct task_struct *tsk)
1535 struct audit_context *context;
1537 context = audit_get_context(tsk, 0, 0);
1538 if (likely(!context))
1541 /* Check for system calls that do not go through the exit
1542 * function (e.g., exit_group), then free context block.
1543 * We use GFP_ATOMIC here because we might be doing this
1544 * in the context of the idle thread */
1545 /* that can happen only if we are called from do_exit() */
1546 if (context->in_syscall && context->auditable)
1547 audit_log_exit(context, tsk);
1549 audit_free_context(context);
1553 * audit_syscall_entry - fill in an audit record at syscall entry
1554 * @arch: architecture type
1555 * @major: major syscall type (function)
1556 * @a1: additional syscall register 1
1557 * @a2: additional syscall register 2
1558 * @a3: additional syscall register 3
1559 * @a4: additional syscall register 4
1561 * Fill in audit context at syscall entry. This only happens if the
1562 * audit context was created when the task was created and the state or
1563 * filters demand the audit context be built. If the state from the
1564 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1565 * then the record will be written at syscall exit time (otherwise, it
1566 * will only be written if another part of the kernel requests that it
1569 void audit_syscall_entry(int arch, int major,
1570 unsigned long a1, unsigned long a2,
1571 unsigned long a3, unsigned long a4)
1573 struct task_struct *tsk = current;
1574 struct audit_context *context = tsk->audit_context;
1575 enum audit_state state;
1577 if (unlikely(!context))
1581 * This happens only on certain architectures that make system
1582 * calls in kernel_thread via the entry.S interface, instead of
1583 * with direct calls. (If you are porting to a new
1584 * architecture, hitting this condition can indicate that you
1585 * got the _exit/_leave calls backward in entry.S.)
1589 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1591 * This also happens with vm86 emulation in a non-nested manner
1592 * (entries without exits), so this case must be caught.
1594 if (context->in_syscall) {
1595 struct audit_context *newctx;
1599 "audit(:%d) pid=%d in syscall=%d;"
1600 " entering syscall=%d\n",
1601 context->serial, tsk->pid, context->major, major);
1603 newctx = audit_alloc_context(context->state);
1605 newctx->previous = context;
1607 tsk->audit_context = newctx;
1609 /* If we can't alloc a new context, the best we
1610 * can do is to leak memory (any pending putname
1611 * will be lost). The only other alternative is
1612 * to abandon auditing. */
1613 audit_zero_context(context, context->state);
1616 BUG_ON(context->in_syscall || context->name_count);
1621 context->arch = arch;
1622 context->major = major;
1623 context->argv[0] = a1;
1624 context->argv[1] = a2;
1625 context->argv[2] = a3;
1626 context->argv[3] = a4;
1628 state = context->state;
1629 context->dummy = !audit_n_rules;
1630 if (!context->dummy && (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT))
1631 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1632 if (likely(state == AUDIT_DISABLED))
1635 context->serial = 0;
1636 context->ctime = CURRENT_TIME;
1637 context->in_syscall = 1;
1638 context->auditable = !!(state == AUDIT_RECORD_CONTEXT);
1642 void audit_finish_fork(struct task_struct *child)
1644 struct audit_context *ctx = current->audit_context;
1645 struct audit_context *p = child->audit_context;
1646 if (!p || !ctx || !ctx->auditable)
1648 p->arch = ctx->arch;
1649 p->major = ctx->major;
1650 memcpy(p->argv, ctx->argv, sizeof(ctx->argv));
1651 p->ctime = ctx->ctime;
1652 p->dummy = ctx->dummy;
1653 p->auditable = ctx->auditable;
1654 p->in_syscall = ctx->in_syscall;
1655 p->filterkey = kstrdup(ctx->filterkey, GFP_KERNEL);
1656 p->ppid = current->pid;
1660 * audit_syscall_exit - deallocate audit context after a system call
1661 * @valid: success/failure flag
1662 * @return_code: syscall return value
1664 * Tear down after system call. If the audit context has been marked as
1665 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1666 * filtering, or because some other part of the kernel write an audit
1667 * message), then write out the syscall information. In call cases,
1668 * free the names stored from getname().
1670 void audit_syscall_exit(int valid, long return_code)
1672 struct task_struct *tsk = current;
1673 struct audit_context *context;
1675 context = audit_get_context(tsk, valid, return_code);
1677 if (likely(!context))
1680 if (context->in_syscall && context->auditable)
1681 audit_log_exit(context, tsk);
1683 context->in_syscall = 0;
1684 context->auditable = 0;
1686 if (context->previous) {
1687 struct audit_context *new_context = context->previous;
1688 context->previous = NULL;
1689 audit_free_context(context);
1690 tsk->audit_context = new_context;
1692 audit_free_names(context);
1693 unroll_tree_refs(context, NULL, 0);
1694 audit_free_aux(context);
1695 context->aux = NULL;
1696 context->aux_pids = NULL;
1697 context->target_pid = 0;
1698 context->target_sid = 0;
1699 context->sockaddr_len = 0;
1701 context->fds[0] = -1;
1702 kfree(context->filterkey);
1703 context->filterkey = NULL;
1704 tsk->audit_context = context;
1708 static inline void handle_one(const struct inode *inode)
1710 #ifdef CONFIG_AUDIT_TREE
1711 struct audit_context *context;
1712 struct audit_tree_refs *p;
1713 struct audit_chunk *chunk;
1715 if (likely(list_empty(&inode->inotify_watches)))
1717 context = current->audit_context;
1719 count = context->tree_count;
1721 chunk = audit_tree_lookup(inode);
1725 if (likely(put_tree_ref(context, chunk)))
1727 if (unlikely(!grow_tree_refs(context))) {
1728 printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1729 audit_set_auditable(context);
1730 audit_put_chunk(chunk);
1731 unroll_tree_refs(context, p, count);
1734 put_tree_ref(context, chunk);
1738 static void handle_path(const struct dentry *dentry)
1740 #ifdef CONFIG_AUDIT_TREE
1741 struct audit_context *context;
1742 struct audit_tree_refs *p;
1743 const struct dentry *d, *parent;
1744 struct audit_chunk *drop;
1748 context = current->audit_context;
1750 count = context->tree_count;
1755 seq = read_seqbegin(&rename_lock);
1757 struct inode *inode = d->d_inode;
1758 if (inode && unlikely(!list_empty(&inode->inotify_watches))) {
1759 struct audit_chunk *chunk;
1760 chunk = audit_tree_lookup(inode);
1762 if (unlikely(!put_tree_ref(context, chunk))) {
1768 parent = d->d_parent;
1773 if (unlikely(read_seqretry(&rename_lock, seq) || drop)) { /* in this order */
1776 /* just a race with rename */
1777 unroll_tree_refs(context, p, count);
1780 audit_put_chunk(drop);
1781 if (grow_tree_refs(context)) {
1782 /* OK, got more space */
1783 unroll_tree_refs(context, p, count);
1788 "out of memory, audit has lost a tree reference\n");
1789 unroll_tree_refs(context, p, count);
1790 audit_set_auditable(context);
1798 * audit_getname - add a name to the list
1799 * @name: name to add
1801 * Add a name to the list of audit names for this context.
1802 * Called from fs/namei.c:getname().
1804 void __audit_getname(const char *name)
1806 struct audit_context *context = current->audit_context;
1808 if (IS_ERR(name) || !name)
1811 if (!context->in_syscall) {
1812 #if AUDIT_DEBUG == 2
1813 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1814 __FILE__, __LINE__, context->serial, name);
1819 BUG_ON(context->name_count >= AUDIT_NAMES);
1820 context->names[context->name_count].name = name;
1821 context->names[context->name_count].name_len = AUDIT_NAME_FULL;
1822 context->names[context->name_count].name_put = 1;
1823 context->names[context->name_count].ino = (unsigned long)-1;
1824 context->names[context->name_count].osid = 0;
1825 ++context->name_count;
1826 if (!context->pwd.dentry) {
1827 read_lock(¤t->fs->lock);
1828 context->pwd = current->fs->pwd;
1829 path_get(¤t->fs->pwd);
1830 read_unlock(¤t->fs->lock);
1835 /* audit_putname - intercept a putname request
1836 * @name: name to intercept and delay for putname
1838 * If we have stored the name from getname in the audit context,
1839 * then we delay the putname until syscall exit.
1840 * Called from include/linux/fs.h:putname().
1842 void audit_putname(const char *name)
1844 struct audit_context *context = current->audit_context;
1847 if (!context->in_syscall) {
1848 #if AUDIT_DEBUG == 2
1849 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1850 __FILE__, __LINE__, context->serial, name);
1851 if (context->name_count) {
1853 for (i = 0; i < context->name_count; i++)
1854 printk(KERN_ERR "name[%d] = %p = %s\n", i,
1855 context->names[i].name,
1856 context->names[i].name ?: "(null)");
1863 ++context->put_count;
1864 if (context->put_count > context->name_count) {
1865 printk(KERN_ERR "%s:%d(:%d): major=%d"
1866 " in_syscall=%d putname(%p) name_count=%d"
1869 context->serial, context->major,
1870 context->in_syscall, name, context->name_count,
1871 context->put_count);
1878 static int audit_inc_name_count(struct audit_context *context,
1879 const struct inode *inode)
1881 if (context->name_count >= AUDIT_NAMES) {
1883 printk(KERN_DEBUG "name_count maxed, losing inode data: "
1884 "dev=%02x:%02x, inode=%lu\n",
1885 MAJOR(inode->i_sb->s_dev),
1886 MINOR(inode->i_sb->s_dev),
1890 printk(KERN_DEBUG "name_count maxed, losing inode data\n");
1893 context->name_count++;
1895 context->ino_count++;
1901 static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
1903 struct cpu_vfs_cap_data caps;
1906 memset(&name->fcap.permitted, 0, sizeof(kernel_cap_t));
1907 memset(&name->fcap.inheritable, 0, sizeof(kernel_cap_t));
1914 rc = get_vfs_caps_from_disk(dentry, &caps);
1918 name->fcap.permitted = caps.permitted;
1919 name->fcap.inheritable = caps.inheritable;
1920 name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
1921 name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
1927 /* Copy inode data into an audit_names. */
1928 static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
1929 const struct inode *inode)
1931 name->ino = inode->i_ino;
1932 name->dev = inode->i_sb->s_dev;
1933 name->mode = inode->i_mode;
1934 name->uid = inode->i_uid;
1935 name->gid = inode->i_gid;
1936 name->rdev = inode->i_rdev;
1937 security_inode_getsecid(inode, &name->osid);
1938 audit_copy_fcaps(name, dentry);
1942 * audit_inode - store the inode and device from a lookup
1943 * @name: name being audited
1944 * @dentry: dentry being audited
1946 * Called from fs/namei.c:path_lookup().
1948 void __audit_inode(const char *name, const struct dentry *dentry)
1951 struct audit_context *context = current->audit_context;
1952 const struct inode *inode = dentry->d_inode;
1954 if (!context->in_syscall)
1956 if (context->name_count
1957 && context->names[context->name_count-1].name
1958 && context->names[context->name_count-1].name == name)
1959 idx = context->name_count - 1;
1960 else if (context->name_count > 1
1961 && context->names[context->name_count-2].name
1962 && context->names[context->name_count-2].name == name)
1963 idx = context->name_count - 2;
1965 /* FIXME: how much do we care about inodes that have no
1966 * associated name? */
1967 if (audit_inc_name_count(context, inode))
1969 idx = context->name_count - 1;
1970 context->names[idx].name = NULL;
1972 handle_path(dentry);
1973 audit_copy_inode(&context->names[idx], dentry, inode);
1977 * audit_inode_child - collect inode info for created/removed objects
1978 * @dname: inode's dentry name
1979 * @dentry: dentry being audited
1980 * @parent: inode of dentry parent
1982 * For syscalls that create or remove filesystem objects, audit_inode
1983 * can only collect information for the filesystem object's parent.
1984 * This call updates the audit context with the child's information.
1985 * Syscalls that create a new filesystem object must be hooked after
1986 * the object is created. Syscalls that remove a filesystem object
1987 * must be hooked prior, in order to capture the target inode during
1988 * unsuccessful attempts.
1990 void __audit_inode_child(const char *dname, const struct dentry *dentry,
1991 const struct inode *parent)
1994 struct audit_context *context = current->audit_context;
1995 const char *found_parent = NULL, *found_child = NULL;
1996 const struct inode *inode = dentry->d_inode;
1999 if (!context->in_syscall)
2004 /* determine matching parent */
2008 /* parent is more likely, look for it first */
2009 for (idx = 0; idx < context->name_count; idx++) {
2010 struct audit_names *n = &context->names[idx];
2015 if (n->ino == parent->i_ino &&
2016 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2017 n->name_len = dirlen; /* update parent data in place */
2018 found_parent = n->name;
2023 /* no matching parent, look for matching child */
2024 for (idx = 0; idx < context->name_count; idx++) {
2025 struct audit_names *n = &context->names[idx];
2030 /* strcmp() is the more likely scenario */
2031 if (!strcmp(dname, n->name) ||
2032 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2034 audit_copy_inode(n, NULL, inode);
2036 n->ino = (unsigned long)-1;
2037 found_child = n->name;
2043 if (!found_parent) {
2044 if (audit_inc_name_count(context, parent))
2046 idx = context->name_count - 1;
2047 context->names[idx].name = NULL;
2048 audit_copy_inode(&context->names[idx], NULL, parent);
2052 if (audit_inc_name_count(context, inode))
2054 idx = context->name_count - 1;
2056 /* Re-use the name belonging to the slot for a matching parent
2057 * directory. All names for this context are relinquished in
2058 * audit_free_names() */
2060 context->names[idx].name = found_parent;
2061 context->names[idx].name_len = AUDIT_NAME_FULL;
2062 /* don't call __putname() */
2063 context->names[idx].name_put = 0;
2065 context->names[idx].name = NULL;
2069 audit_copy_inode(&context->names[idx], NULL, inode);
2071 context->names[idx].ino = (unsigned long)-1;
2074 EXPORT_SYMBOL_GPL(__audit_inode_child);
2077 * auditsc_get_stamp - get local copies of audit_context values
2078 * @ctx: audit_context for the task
2079 * @t: timespec to store time recorded in the audit_context
2080 * @serial: serial value that is recorded in the audit_context
2082 * Also sets the context as auditable.
2084 int auditsc_get_stamp(struct audit_context *ctx,
2085 struct timespec *t, unsigned int *serial)
2087 if (!ctx->in_syscall)
2090 ctx->serial = audit_serial();
2091 t->tv_sec = ctx->ctime.tv_sec;
2092 t->tv_nsec = ctx->ctime.tv_nsec;
2093 *serial = ctx->serial;
2098 /* global counter which is incremented every time something logs in */
2099 static atomic_t session_id = ATOMIC_INIT(0);
2102 * audit_set_loginuid - set a task's audit_context loginuid
2103 * @task: task whose audit context is being modified
2104 * @loginuid: loginuid value
2108 * Called (set) from fs/proc/base.c::proc_loginuid_write().
2110 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
2112 unsigned int sessionid = atomic_inc_return(&session_id);
2113 struct audit_context *context = task->audit_context;
2115 if (context && context->in_syscall) {
2116 struct audit_buffer *ab;
2118 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2120 audit_log_format(ab, "login pid=%d uid=%u "
2121 "old auid=%u new auid=%u"
2122 " old ses=%u new ses=%u",
2123 task->pid, task_uid(task),
2124 task->loginuid, loginuid,
2125 task->sessionid, sessionid);
2129 task->sessionid = sessionid;
2130 task->loginuid = loginuid;
2135 * __audit_mq_open - record audit data for a POSIX MQ open
2138 * @u_attr: queue attributes
2141 void __audit_mq_open(int oflag, mode_t mode, struct mq_attr *attr)
2143 struct audit_context *context = current->audit_context;
2146 memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2148 memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2150 context->mq_open.oflag = oflag;
2151 context->mq_open.mode = mode;
2153 context->type = AUDIT_MQ_OPEN;
2157 * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2158 * @mqdes: MQ descriptor
2159 * @msg_len: Message length
2160 * @msg_prio: Message priority
2161 * @abs_timeout: Message timeout in absolute time
2164 void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2165 const struct timespec *abs_timeout)
2167 struct audit_context *context = current->audit_context;
2168 struct timespec *p = &context->mq_sendrecv.abs_timeout;
2171 memcpy(p, abs_timeout, sizeof(struct timespec));
2173 memset(p, 0, sizeof(struct timespec));
2175 context->mq_sendrecv.mqdes = mqdes;
2176 context->mq_sendrecv.msg_len = msg_len;
2177 context->mq_sendrecv.msg_prio = msg_prio;
2179 context->type = AUDIT_MQ_SENDRECV;
2183 * __audit_mq_notify - record audit data for a POSIX MQ notify
2184 * @mqdes: MQ descriptor
2185 * @u_notification: Notification event
2189 void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2191 struct audit_context *context = current->audit_context;
2194 context->mq_notify.sigev_signo = notification->sigev_signo;
2196 context->mq_notify.sigev_signo = 0;
2198 context->mq_notify.mqdes = mqdes;
2199 context->type = AUDIT_MQ_NOTIFY;
2203 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2204 * @mqdes: MQ descriptor
2208 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2210 struct audit_context *context = current->audit_context;
2211 context->mq_getsetattr.mqdes = mqdes;
2212 context->mq_getsetattr.mqstat = *mqstat;
2213 context->type = AUDIT_MQ_GETSETATTR;
2217 * audit_ipc_obj - record audit data for ipc object
2218 * @ipcp: ipc permissions
2221 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2223 struct audit_context *context = current->audit_context;
2224 context->ipc.uid = ipcp->uid;
2225 context->ipc.gid = ipcp->gid;
2226 context->ipc.mode = ipcp->mode;
2227 context->ipc.has_perm = 0;
2228 security_ipc_getsecid(ipcp, &context->ipc.osid);
2229 context->type = AUDIT_IPC;
2233 * audit_ipc_set_perm - record audit data for new ipc permissions
2234 * @qbytes: msgq bytes
2235 * @uid: msgq user id
2236 * @gid: msgq group id
2237 * @mode: msgq mode (permissions)
2239 * Called only after audit_ipc_obj().
2241 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
2243 struct audit_context *context = current->audit_context;
2245 context->ipc.qbytes = qbytes;
2246 context->ipc.perm_uid = uid;
2247 context->ipc.perm_gid = gid;
2248 context->ipc.perm_mode = mode;
2249 context->ipc.has_perm = 1;
2252 int audit_bprm(struct linux_binprm *bprm)
2254 struct audit_aux_data_execve *ax;
2255 struct audit_context *context = current->audit_context;
2257 if (likely(!audit_enabled || !context || context->dummy))
2260 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2264 ax->argc = bprm->argc;
2265 ax->envc = bprm->envc;
2267 ax->d.type = AUDIT_EXECVE;
2268 ax->d.next = context->aux;
2269 context->aux = (void *)ax;
2275 * audit_socketcall - record audit data for sys_socketcall
2276 * @nargs: number of args
2280 void audit_socketcall(int nargs, unsigned long *args)
2282 struct audit_context *context = current->audit_context;
2284 if (likely(!context || context->dummy))
2287 context->type = AUDIT_SOCKETCALL;
2288 context->socketcall.nargs = nargs;
2289 memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2293 * __audit_fd_pair - record audit data for pipe and socketpair
2294 * @fd1: the first file descriptor
2295 * @fd2: the second file descriptor
2298 void __audit_fd_pair(int fd1, int fd2)
2300 struct audit_context *context = current->audit_context;
2301 context->fds[0] = fd1;
2302 context->fds[1] = fd2;
2306 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2307 * @len: data length in user space
2308 * @a: data address in kernel space
2310 * Returns 0 for success or NULL context or < 0 on error.
2312 int audit_sockaddr(int len, void *a)
2314 struct audit_context *context = current->audit_context;
2316 if (likely(!context || context->dummy))
2319 if (!context->sockaddr) {
2320 void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2323 context->sockaddr = p;
2326 context->sockaddr_len = len;
2327 memcpy(context->sockaddr, a, len);
2331 void __audit_ptrace(struct task_struct *t)
2333 struct audit_context *context = current->audit_context;
2335 context->target_pid = t->pid;
2336 context->target_auid = audit_get_loginuid(t);
2337 context->target_uid = task_uid(t);
2338 context->target_sessionid = audit_get_sessionid(t);
2339 security_task_getsecid(t, &context->target_sid);
2340 memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2344 * audit_signal_info - record signal info for shutting down audit subsystem
2345 * @sig: signal value
2346 * @t: task being signaled
2348 * If the audit subsystem is being terminated, record the task (pid)
2349 * and uid that is doing that.
2351 int __audit_signal_info(int sig, struct task_struct *t)
2353 struct audit_aux_data_pids *axp;
2354 struct task_struct *tsk = current;
2355 struct audit_context *ctx = tsk->audit_context;
2356 uid_t uid = current_uid(), t_uid = task_uid(t);
2358 if (audit_pid && t->tgid == audit_pid) {
2359 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2360 audit_sig_pid = tsk->pid;
2361 if (tsk->loginuid != -1)
2362 audit_sig_uid = tsk->loginuid;
2364 audit_sig_uid = uid;
2365 security_task_getsecid(tsk, &audit_sig_sid);
2367 if (!audit_signals || audit_dummy_context())
2371 /* optimize the common case by putting first signal recipient directly
2372 * in audit_context */
2373 if (!ctx->target_pid) {
2374 ctx->target_pid = t->tgid;
2375 ctx->target_auid = audit_get_loginuid(t);
2376 ctx->target_uid = t_uid;
2377 ctx->target_sessionid = audit_get_sessionid(t);
2378 security_task_getsecid(t, &ctx->target_sid);
2379 memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2383 axp = (void *)ctx->aux_pids;
2384 if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2385 axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2389 axp->d.type = AUDIT_OBJ_PID;
2390 axp->d.next = ctx->aux_pids;
2391 ctx->aux_pids = (void *)axp;
2393 BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2395 axp->target_pid[axp->pid_count] = t->tgid;
2396 axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2397 axp->target_uid[axp->pid_count] = t_uid;
2398 axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2399 security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2400 memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2407 * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2408 * @bprm: pointer to the bprm being processed
2409 * @new: the proposed new credentials
2410 * @old: the old credentials
2412 * Simply check if the proc already has the caps given by the file and if not
2413 * store the priv escalation info for later auditing at the end of the syscall
2417 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2418 const struct cred *new, const struct cred *old)
2420 struct audit_aux_data_bprm_fcaps *ax;
2421 struct audit_context *context = current->audit_context;
2422 struct cpu_vfs_cap_data vcaps;
2423 struct dentry *dentry;
2425 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2429 ax->d.type = AUDIT_BPRM_FCAPS;
2430 ax->d.next = context->aux;
2431 context->aux = (void *)ax;
2433 dentry = dget(bprm->file->f_dentry);
2434 get_vfs_caps_from_disk(dentry, &vcaps);
2437 ax->fcap.permitted = vcaps.permitted;
2438 ax->fcap.inheritable = vcaps.inheritable;
2439 ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2440 ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2442 ax->old_pcap.permitted = old->cap_permitted;
2443 ax->old_pcap.inheritable = old->cap_inheritable;
2444 ax->old_pcap.effective = old->cap_effective;
2446 ax->new_pcap.permitted = new->cap_permitted;
2447 ax->new_pcap.inheritable = new->cap_inheritable;
2448 ax->new_pcap.effective = new->cap_effective;
2453 * __audit_log_capset - store information about the arguments to the capset syscall
2454 * @pid: target pid of the capset call
2455 * @new: the new credentials
2456 * @old: the old (current) credentials
2458 * Record the aguments userspace sent to sys_capset for later printing by the
2459 * audit system if applicable
2461 void __audit_log_capset(pid_t pid,
2462 const struct cred *new, const struct cred *old)
2464 struct audit_context *context = current->audit_context;
2465 context->capset.pid = pid;
2466 context->capset.cap.effective = new->cap_effective;
2467 context->capset.cap.inheritable = new->cap_effective;
2468 context->capset.cap.permitted = new->cap_permitted;
2469 context->type = AUDIT_CAPSET;
2473 * audit_core_dumps - record information about processes that end abnormally
2474 * @signr: signal value
2476 * If a process ends with a core dump, something fishy is going on and we
2477 * should record the event for investigation.
2479 void audit_core_dumps(long signr)
2481 struct audit_buffer *ab;
2483 uid_t auid = audit_get_loginuid(current), uid;
2485 unsigned int sessionid = audit_get_sessionid(current);
2490 if (signr == SIGQUIT) /* don't care for those */
2493 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2494 current_uid_gid(&uid, &gid);
2495 audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2496 auid, uid, gid, sessionid);
2497 security_task_getsecid(current, &sid);
2502 if (security_secid_to_secctx(sid, &ctx, &len))
2503 audit_log_format(ab, " ssid=%u", sid);
2505 audit_log_format(ab, " subj=%s", ctx);
2506 security_release_secctx(ctx, len);
2509 audit_log_format(ab, " pid=%d comm=", current->pid);
2510 audit_log_untrustedstring(ab, current->comm);
2511 audit_log_format(ab, " sig=%ld", signr);