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>
48 #include <asm/types.h>
50 #include <linux/namei.h>
52 #include <linux/module.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/selinux.h>
66 #include <linux/binfmts.h>
67 #include <linux/syscalls.h>
71 extern struct list_head audit_filter_list[];
73 /* No syscall auditing will take place unless audit_enabled != 0. */
74 extern int audit_enabled;
76 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
77 * for saving names from getname(). */
78 #define AUDIT_NAMES 20
80 /* AUDIT_NAMES_RESERVED is the number of slots we reserve in the
81 * audit_context from being used for nameless inodes from
83 #define AUDIT_NAMES_RESERVED 7
85 /* Indicates that audit should log the full pathname. */
86 #define AUDIT_NAME_FULL -1
88 /* When fs/namei.c:getname() is called, we store the pointer in name and
89 * we don't let putname() free it (instead we free all of the saved
90 * pointers at syscall exit time).
92 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
95 int name_len; /* number of name's characters to log */
96 unsigned name_put; /* call __putname() for this name */
106 struct audit_aux_data {
107 struct audit_aux_data *next;
111 #define AUDIT_AUX_IPCPERM 0
113 struct audit_aux_data_mq_open {
114 struct audit_aux_data d;
120 struct audit_aux_data_mq_sendrecv {
121 struct audit_aux_data d;
124 unsigned int msg_prio;
125 struct timespec abs_timeout;
128 struct audit_aux_data_mq_notify {
129 struct audit_aux_data d;
131 struct sigevent notification;
134 struct audit_aux_data_mq_getsetattr {
135 struct audit_aux_data d;
137 struct mq_attr mqstat;
140 struct audit_aux_data_ipcctl {
141 struct audit_aux_data d;
143 unsigned long qbytes;
150 struct audit_aux_data_execve {
151 struct audit_aux_data d;
157 struct audit_aux_data_socketcall {
158 struct audit_aux_data d;
160 unsigned long args[0];
163 struct audit_aux_data_sockaddr {
164 struct audit_aux_data d;
169 struct audit_aux_data_path {
170 struct audit_aux_data d;
171 struct dentry *dentry;
172 struct vfsmount *mnt;
175 /* The per-task audit context. */
176 struct audit_context {
177 int in_syscall; /* 1 if task is in a syscall */
178 enum audit_state state;
179 unsigned int serial; /* serial number for record */
180 struct timespec ctime; /* time of syscall entry */
181 uid_t loginuid; /* login uid (identity) */
182 int major; /* syscall number */
183 unsigned long argv[4]; /* syscall arguments */
184 int return_valid; /* return code is valid */
185 long return_code;/* syscall return code */
186 int auditable; /* 1 if record should be written */
188 struct audit_names names[AUDIT_NAMES];
189 char * filterkey; /* key for rule that triggered record */
191 struct vfsmount * pwdmnt;
192 struct audit_context *previous; /* For nested syscalls */
193 struct audit_aux_data *aux;
195 /* Save things to print about task_struct */
197 uid_t uid, euid, suid, fsuid;
198 gid_t gid, egid, sgid, fsgid;
199 unsigned long personality;
208 /* Determine if any context name data matches a rule's watch data */
209 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
211 static int audit_filter_rules(struct task_struct *tsk,
212 struct audit_krule *rule,
213 struct audit_context *ctx,
214 struct audit_names *name,
215 enum audit_state *state)
217 int i, j, need_sid = 1;
220 for (i = 0; i < rule->field_count; i++) {
221 struct audit_field *f = &rule->fields[i];
226 result = audit_comparator(tsk->pid, f->op, f->val);
230 result = audit_comparator(ctx->ppid, f->op, f->val);
233 result = audit_comparator(tsk->uid, f->op, f->val);
236 result = audit_comparator(tsk->euid, f->op, f->val);
239 result = audit_comparator(tsk->suid, f->op, f->val);
242 result = audit_comparator(tsk->fsuid, f->op, f->val);
245 result = audit_comparator(tsk->gid, f->op, f->val);
248 result = audit_comparator(tsk->egid, f->op, f->val);
251 result = audit_comparator(tsk->sgid, f->op, f->val);
254 result = audit_comparator(tsk->fsgid, f->op, f->val);
257 result = audit_comparator(tsk->personality, f->op, f->val);
261 result = audit_comparator(ctx->arch, f->op, f->val);
265 if (ctx && ctx->return_valid)
266 result = audit_comparator(ctx->return_code, f->op, f->val);
269 if (ctx && ctx->return_valid) {
271 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
273 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
278 result = audit_comparator(MAJOR(name->dev),
281 for (j = 0; j < ctx->name_count; j++) {
282 if (audit_comparator(MAJOR(ctx->names[j].dev), f->op, f->val)) {
291 result = audit_comparator(MINOR(name->dev),
294 for (j = 0; j < ctx->name_count; j++) {
295 if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
304 result = (name->ino == f->val);
306 for (j = 0; j < ctx->name_count; j++) {
307 if (audit_comparator(ctx->names[j].ino, f->op, f->val)) {
315 if (name && rule->watch->ino != (unsigned long)-1)
316 result = (name->dev == rule->watch->dev &&
317 name->ino == rule->watch->ino);
322 result = audit_comparator(ctx->loginuid, f->op, f->val);
324 case AUDIT_SUBJ_USER:
325 case AUDIT_SUBJ_ROLE:
326 case AUDIT_SUBJ_TYPE:
329 /* NOTE: this may return negative values indicating
330 a temporary error. We simply treat this as a
331 match for now to avoid losing information that
332 may be wanted. An error message will also be
336 selinux_task_ctxid(tsk, &sid);
339 result = selinux_audit_rule_match(sid, f->type,
350 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
352 case AUDIT_FILTERKEY:
353 /* ignore this field for filtering */
362 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
363 switch (rule->action) {
364 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
365 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
370 /* At process creation time, we can determine if system-call auditing is
371 * completely disabled for this task. Since we only have the task
372 * structure at this point, we can only check uid and gid.
374 static enum audit_state audit_filter_task(struct task_struct *tsk)
376 struct audit_entry *e;
377 enum audit_state state;
380 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
381 if (audit_filter_rules(tsk, &e->rule, NULL, NULL, &state)) {
387 return AUDIT_BUILD_CONTEXT;
390 /* At syscall entry and exit time, this filter is called if the
391 * audit_state is not low enough that auditing cannot take place, but is
392 * also not high enough that we already know we have to write an audit
393 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
395 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
396 struct audit_context *ctx,
397 struct list_head *list)
399 struct audit_entry *e;
400 enum audit_state state;
402 if (audit_pid && tsk->tgid == audit_pid)
403 return AUDIT_DISABLED;
406 if (!list_empty(list)) {
407 int word = AUDIT_WORD(ctx->major);
408 int bit = AUDIT_BIT(ctx->major);
410 list_for_each_entry_rcu(e, list, list) {
411 if ((e->rule.mask[word] & bit) == bit &&
412 audit_filter_rules(tsk, &e->rule, ctx, NULL,
420 return AUDIT_BUILD_CONTEXT;
423 /* At syscall exit time, this filter is called if any audit_names[] have been
424 * collected during syscall processing. We only check rules in sublists at hash
425 * buckets applicable to the inode numbers in audit_names[].
426 * Regarding audit_state, same rules apply as for audit_filter_syscall().
428 enum audit_state audit_filter_inodes(struct task_struct *tsk,
429 struct audit_context *ctx)
432 struct audit_entry *e;
433 enum audit_state state;
435 if (audit_pid && tsk->tgid == audit_pid)
436 return AUDIT_DISABLED;
439 for (i = 0; i < ctx->name_count; i++) {
440 int word = AUDIT_WORD(ctx->major);
441 int bit = AUDIT_BIT(ctx->major);
442 struct audit_names *n = &ctx->names[i];
443 int h = audit_hash_ino((u32)n->ino);
444 struct list_head *list = &audit_inode_hash[h];
446 if (list_empty(list))
449 list_for_each_entry_rcu(e, list, list) {
450 if ((e->rule.mask[word] & bit) == bit &&
451 audit_filter_rules(tsk, &e->rule, ctx, n, &state)) {
458 return AUDIT_BUILD_CONTEXT;
461 void audit_set_auditable(struct audit_context *ctx)
466 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
470 struct audit_context *context = tsk->audit_context;
472 if (likely(!context))
474 context->return_valid = return_valid;
475 context->return_code = return_code;
477 if (context->in_syscall && !context->auditable) {
478 enum audit_state state;
480 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
481 if (state == AUDIT_RECORD_CONTEXT) {
482 context->auditable = 1;
486 state = audit_filter_inodes(tsk, context);
487 if (state == AUDIT_RECORD_CONTEXT)
488 context->auditable = 1;
493 context->pid = tsk->pid;
494 context->ppid = sys_getppid(); /* sic. tsk == current in all cases */
495 context->uid = tsk->uid;
496 context->gid = tsk->gid;
497 context->euid = tsk->euid;
498 context->suid = tsk->suid;
499 context->fsuid = tsk->fsuid;
500 context->egid = tsk->egid;
501 context->sgid = tsk->sgid;
502 context->fsgid = tsk->fsgid;
503 context->personality = tsk->personality;
504 tsk->audit_context = NULL;
508 static inline void audit_free_names(struct audit_context *context)
513 if (context->auditable
514 ||context->put_count + context->ino_count != context->name_count) {
515 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
516 " name_count=%d put_count=%d"
517 " ino_count=%d [NOT freeing]\n",
519 context->serial, context->major, context->in_syscall,
520 context->name_count, context->put_count,
522 for (i = 0; i < context->name_count; i++) {
523 printk(KERN_ERR "names[%d] = %p = %s\n", i,
524 context->names[i].name,
525 context->names[i].name ?: "(null)");
532 context->put_count = 0;
533 context->ino_count = 0;
536 for (i = 0; i < context->name_count; i++) {
537 if (context->names[i].name && context->names[i].name_put)
538 __putname(context->names[i].name);
540 context->name_count = 0;
544 mntput(context->pwdmnt);
546 context->pwdmnt = NULL;
549 static inline void audit_free_aux(struct audit_context *context)
551 struct audit_aux_data *aux;
553 while ((aux = context->aux)) {
554 if (aux->type == AUDIT_AVC_PATH) {
555 struct audit_aux_data_path *axi = (void *)aux;
560 context->aux = aux->next;
565 static inline void audit_zero_context(struct audit_context *context,
566 enum audit_state state)
568 uid_t loginuid = context->loginuid;
570 memset(context, 0, sizeof(*context));
571 context->state = state;
572 context->loginuid = loginuid;
575 static inline struct audit_context *audit_alloc_context(enum audit_state state)
577 struct audit_context *context;
579 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
581 audit_zero_context(context, state);
586 * audit_alloc - allocate an audit context block for a task
589 * Filter on the task information and allocate a per-task audit context
590 * if necessary. Doing so turns on system call auditing for the
591 * specified task. This is called from copy_process, so no lock is
594 int audit_alloc(struct task_struct *tsk)
596 struct audit_context *context;
597 enum audit_state state;
599 if (likely(!audit_enabled))
600 return 0; /* Return if not auditing. */
602 state = audit_filter_task(tsk);
603 if (likely(state == AUDIT_DISABLED))
606 if (!(context = audit_alloc_context(state))) {
607 audit_log_lost("out of memory in audit_alloc");
611 /* Preserve login uid */
612 context->loginuid = -1;
613 if (current->audit_context)
614 context->loginuid = current->audit_context->loginuid;
616 tsk->audit_context = context;
617 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
621 static inline void audit_free_context(struct audit_context *context)
623 struct audit_context *previous;
627 previous = context->previous;
628 if (previous || (count && count < 10)) {
630 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
631 " freeing multiple contexts (%d)\n",
632 context->serial, context->major,
633 context->name_count, count);
635 audit_free_names(context);
636 audit_free_aux(context);
637 kfree(context->filterkey);
642 printk(KERN_ERR "audit: freed %d contexts\n", count);
645 static void audit_log_task_context(struct audit_buffer *ab)
650 len = security_getprocattr(current, "current", NULL, 0);
657 ctx = kmalloc(len, GFP_KERNEL);
661 len = security_getprocattr(current, "current", ctx, len);
665 audit_log_format(ab, " subj=%s", ctx);
670 audit_panic("error in audit_log_task_context");
674 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
676 char name[sizeof(tsk->comm)];
677 struct mm_struct *mm = tsk->mm;
678 struct vm_area_struct *vma;
682 get_task_comm(name, tsk);
683 audit_log_format(ab, " comm=");
684 audit_log_untrustedstring(ab, name);
687 down_read(&mm->mmap_sem);
690 if ((vma->vm_flags & VM_EXECUTABLE) &&
692 audit_log_d_path(ab, "exe=",
693 vma->vm_file->f_dentry,
694 vma->vm_file->f_vfsmnt);
699 up_read(&mm->mmap_sem);
701 audit_log_task_context(ab);
704 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
706 int i, call_panic = 0;
707 struct audit_buffer *ab;
708 struct audit_aux_data *aux;
713 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
715 return; /* audit_panic has been called */
716 audit_log_format(ab, "arch=%x syscall=%d",
717 context->arch, context->major);
718 if (context->personality != PER_LINUX)
719 audit_log_format(ab, " per=%lx", context->personality);
720 if (context->return_valid)
721 audit_log_format(ab, " success=%s exit=%ld",
722 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
723 context->return_code);
724 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
725 tty = tsk->signal->tty->name;
729 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
730 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
731 " euid=%u suid=%u fsuid=%u"
732 " egid=%u sgid=%u fsgid=%u tty=%s",
743 context->euid, context->suid, context->fsuid,
744 context->egid, context->sgid, context->fsgid, tty);
745 audit_log_task_info(ab, tsk);
746 if (context->filterkey) {
747 audit_log_format(ab, " key=");
748 audit_log_untrustedstring(ab, context->filterkey);
750 audit_log_format(ab, " key=(null)");
753 for (aux = context->aux; aux; aux = aux->next) {
755 ab = audit_log_start(context, GFP_KERNEL, aux->type);
757 continue; /* audit_panic has been called */
760 case AUDIT_MQ_OPEN: {
761 struct audit_aux_data_mq_open *axi = (void *)aux;
763 "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
764 "mq_msgsize=%ld mq_curmsgs=%ld",
765 axi->oflag, axi->mode, axi->attr.mq_flags,
766 axi->attr.mq_maxmsg, axi->attr.mq_msgsize,
767 axi->attr.mq_curmsgs);
770 case AUDIT_MQ_SENDRECV: {
771 struct audit_aux_data_mq_sendrecv *axi = (void *)aux;
773 "mqdes=%d msg_len=%zd msg_prio=%u "
774 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
775 axi->mqdes, axi->msg_len, axi->msg_prio,
776 axi->abs_timeout.tv_sec, axi->abs_timeout.tv_nsec);
779 case AUDIT_MQ_NOTIFY: {
780 struct audit_aux_data_mq_notify *axi = (void *)aux;
782 "mqdes=%d sigev_signo=%d",
784 axi->notification.sigev_signo);
787 case AUDIT_MQ_GETSETATTR: {
788 struct audit_aux_data_mq_getsetattr *axi = (void *)aux;
790 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
793 axi->mqstat.mq_flags, axi->mqstat.mq_maxmsg,
794 axi->mqstat.mq_msgsize, axi->mqstat.mq_curmsgs);
798 struct audit_aux_data_ipcctl *axi = (void *)aux;
800 "ouid=%u ogid=%u mode=%x",
801 axi->uid, axi->gid, axi->mode);
802 if (axi->osid != 0) {
805 if (selinux_ctxid_to_string(
806 axi->osid, &ctx, &len)) {
807 audit_log_format(ab, " osid=%u",
811 audit_log_format(ab, " obj=%s", ctx);
816 case AUDIT_IPC_SET_PERM: {
817 struct audit_aux_data_ipcctl *axi = (void *)aux;
819 "qbytes=%lx ouid=%u ogid=%u mode=%x",
820 axi->qbytes, axi->uid, axi->gid, axi->mode);
824 struct audit_aux_data_execve *axi = (void *)aux;
827 for (i = 0, p = axi->mem; i < axi->argc; i++) {
828 audit_log_format(ab, "a%d=", i);
829 p = audit_log_untrustedstring(ab, p);
830 audit_log_format(ab, "\n");
834 case AUDIT_SOCKETCALL: {
836 struct audit_aux_data_socketcall *axs = (void *)aux;
837 audit_log_format(ab, "nargs=%d", axs->nargs);
838 for (i=0; i<axs->nargs; i++)
839 audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
842 case AUDIT_SOCKADDR: {
843 struct audit_aux_data_sockaddr *axs = (void *)aux;
845 audit_log_format(ab, "saddr=");
846 audit_log_hex(ab, axs->a, axs->len);
849 case AUDIT_AVC_PATH: {
850 struct audit_aux_data_path *axi = (void *)aux;
851 audit_log_d_path(ab, "path=", axi->dentry, axi->mnt);
858 if (context->pwd && context->pwdmnt) {
859 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
861 audit_log_d_path(ab, "cwd=", context->pwd, context->pwdmnt);
865 for (i = 0; i < context->name_count; i++) {
866 struct audit_names *n = &context->names[i];
868 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
870 continue; /* audit_panic has been called */
872 audit_log_format(ab, "item=%d", i);
875 switch(n->name_len) {
876 case AUDIT_NAME_FULL:
877 /* log the full path */
878 audit_log_format(ab, " name=");
879 audit_log_untrustedstring(ab, n->name);
882 /* name was specified as a relative path and the
883 * directory component is the cwd */
884 audit_log_d_path(ab, " name=", context->pwd,
888 /* log the name's directory component */
889 audit_log_format(ab, " name=");
890 audit_log_n_untrustedstring(ab, n->name_len,
894 audit_log_format(ab, " name=(null)");
896 if (n->ino != (unsigned long)-1) {
897 audit_log_format(ab, " inode=%lu"
898 " dev=%02x:%02x mode=%#o"
899 " ouid=%u ogid=%u rdev=%02x:%02x",
912 if (selinux_ctxid_to_string(
913 n->osid, &ctx, &len)) {
914 audit_log_format(ab, " osid=%u", n->osid);
917 audit_log_format(ab, " obj=%s", ctx);
924 audit_panic("error converting sid to string");
928 * audit_free - free a per-task audit context
929 * @tsk: task whose audit context block to free
931 * Called from copy_process and do_exit
933 void audit_free(struct task_struct *tsk)
935 struct audit_context *context;
937 context = audit_get_context(tsk, 0, 0);
938 if (likely(!context))
941 /* Check for system calls that do not go through the exit
942 * function (e.g., exit_group), then free context block.
943 * We use GFP_ATOMIC here because we might be doing this
944 * in the context of the idle thread */
945 /* that can happen only if we are called from do_exit() */
946 if (context->in_syscall && context->auditable)
947 audit_log_exit(context, tsk);
949 audit_free_context(context);
953 * audit_syscall_entry - fill in an audit record at syscall entry
954 * @tsk: task being audited
955 * @arch: architecture type
956 * @major: major syscall type (function)
957 * @a1: additional syscall register 1
958 * @a2: additional syscall register 2
959 * @a3: additional syscall register 3
960 * @a4: additional syscall register 4
962 * Fill in audit context at syscall entry. This only happens if the
963 * audit context was created when the task was created and the state or
964 * filters demand the audit context be built. If the state from the
965 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
966 * then the record will be written at syscall exit time (otherwise, it
967 * will only be written if another part of the kernel requests that it
970 void audit_syscall_entry(int arch, int major,
971 unsigned long a1, unsigned long a2,
972 unsigned long a3, unsigned long a4)
974 struct task_struct *tsk = current;
975 struct audit_context *context = tsk->audit_context;
976 enum audit_state state;
981 * This happens only on certain architectures that make system
982 * calls in kernel_thread via the entry.S interface, instead of
983 * with direct calls. (If you are porting to a new
984 * architecture, hitting this condition can indicate that you
985 * got the _exit/_leave calls backward in entry.S.)
989 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
991 * This also happens with vm86 emulation in a non-nested manner
992 * (entries without exits), so this case must be caught.
994 if (context->in_syscall) {
995 struct audit_context *newctx;
999 "audit(:%d) pid=%d in syscall=%d;"
1000 " entering syscall=%d\n",
1001 context->serial, tsk->pid, context->major, major);
1003 newctx = audit_alloc_context(context->state);
1005 newctx->previous = context;
1007 tsk->audit_context = newctx;
1009 /* If we can't alloc a new context, the best we
1010 * can do is to leak memory (any pending putname
1011 * will be lost). The only other alternative is
1012 * to abandon auditing. */
1013 audit_zero_context(context, context->state);
1016 BUG_ON(context->in_syscall || context->name_count);
1021 context->arch = arch;
1022 context->major = major;
1023 context->argv[0] = a1;
1024 context->argv[1] = a2;
1025 context->argv[2] = a3;
1026 context->argv[3] = a4;
1028 state = context->state;
1029 if (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT)
1030 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1031 if (likely(state == AUDIT_DISABLED))
1034 context->serial = 0;
1035 context->ctime = CURRENT_TIME;
1036 context->in_syscall = 1;
1037 context->auditable = !!(state == AUDIT_RECORD_CONTEXT);
1041 * audit_syscall_exit - deallocate audit context after a system call
1042 * @tsk: task being audited
1043 * @valid: success/failure flag
1044 * @return_code: syscall return value
1046 * Tear down after system call. If the audit context has been marked as
1047 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1048 * filtering, or because some other part of the kernel write an audit
1049 * message), then write out the syscall information. In call cases,
1050 * free the names stored from getname().
1052 void audit_syscall_exit(int valid, long return_code)
1054 struct task_struct *tsk = current;
1055 struct audit_context *context;
1057 context = audit_get_context(tsk, valid, return_code);
1059 if (likely(!context))
1062 if (context->in_syscall && context->auditable)
1063 audit_log_exit(context, tsk);
1065 context->in_syscall = 0;
1066 context->auditable = 0;
1068 if (context->previous) {
1069 struct audit_context *new_context = context->previous;
1070 context->previous = NULL;
1071 audit_free_context(context);
1072 tsk->audit_context = new_context;
1074 audit_free_names(context);
1075 audit_free_aux(context);
1076 kfree(context->filterkey);
1077 context->filterkey = NULL;
1078 tsk->audit_context = context;
1083 * audit_getname - add a name to the list
1084 * @name: name to add
1086 * Add a name to the list of audit names for this context.
1087 * Called from fs/namei.c:getname().
1089 void __audit_getname(const char *name)
1091 struct audit_context *context = current->audit_context;
1093 if (IS_ERR(name) || !name)
1096 if (!context->in_syscall) {
1097 #if AUDIT_DEBUG == 2
1098 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1099 __FILE__, __LINE__, context->serial, name);
1104 BUG_ON(context->name_count >= AUDIT_NAMES);
1105 context->names[context->name_count].name = name;
1106 context->names[context->name_count].name_len = AUDIT_NAME_FULL;
1107 context->names[context->name_count].name_put = 1;
1108 context->names[context->name_count].ino = (unsigned long)-1;
1109 ++context->name_count;
1110 if (!context->pwd) {
1111 read_lock(¤t->fs->lock);
1112 context->pwd = dget(current->fs->pwd);
1113 context->pwdmnt = mntget(current->fs->pwdmnt);
1114 read_unlock(¤t->fs->lock);
1119 /* audit_putname - intercept a putname request
1120 * @name: name to intercept and delay for putname
1122 * If we have stored the name from getname in the audit context,
1123 * then we delay the putname until syscall exit.
1124 * Called from include/linux/fs.h:putname().
1126 void audit_putname(const char *name)
1128 struct audit_context *context = current->audit_context;
1131 if (!context->in_syscall) {
1132 #if AUDIT_DEBUG == 2
1133 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1134 __FILE__, __LINE__, context->serial, name);
1135 if (context->name_count) {
1137 for (i = 0; i < context->name_count; i++)
1138 printk(KERN_ERR "name[%d] = %p = %s\n", i,
1139 context->names[i].name,
1140 context->names[i].name ?: "(null)");
1147 ++context->put_count;
1148 if (context->put_count > context->name_count) {
1149 printk(KERN_ERR "%s:%d(:%d): major=%d"
1150 " in_syscall=%d putname(%p) name_count=%d"
1153 context->serial, context->major,
1154 context->in_syscall, name, context->name_count,
1155 context->put_count);
1162 static void audit_inode_context(int idx, const struct inode *inode)
1164 struct audit_context *context = current->audit_context;
1166 selinux_get_inode_sid(inode, &context->names[idx].osid);
1171 * audit_inode - store the inode and device from a lookup
1172 * @name: name being audited
1173 * @inode: inode being audited
1175 * Called from fs/namei.c:path_lookup().
1177 void __audit_inode(const char *name, const struct inode *inode)
1180 struct audit_context *context = current->audit_context;
1182 if (!context->in_syscall)
1184 if (context->name_count
1185 && context->names[context->name_count-1].name
1186 && context->names[context->name_count-1].name == name)
1187 idx = context->name_count - 1;
1188 else if (context->name_count > 1
1189 && context->names[context->name_count-2].name
1190 && context->names[context->name_count-2].name == name)
1191 idx = context->name_count - 2;
1193 /* FIXME: how much do we care about inodes that have no
1194 * associated name? */
1195 if (context->name_count >= AUDIT_NAMES - AUDIT_NAMES_RESERVED)
1197 idx = context->name_count++;
1198 context->names[idx].name = NULL;
1200 ++context->ino_count;
1203 context->names[idx].ino = inode->i_ino;
1204 context->names[idx].dev = inode->i_sb->s_dev;
1205 context->names[idx].mode = inode->i_mode;
1206 context->names[idx].uid = inode->i_uid;
1207 context->names[idx].gid = inode->i_gid;
1208 context->names[idx].rdev = inode->i_rdev;
1209 audit_inode_context(idx, inode);
1213 * audit_inode_child - collect inode info for created/removed objects
1214 * @dname: inode's dentry name
1215 * @inode: inode being audited
1216 * @pino: inode number of dentry parent
1218 * For syscalls that create or remove filesystem objects, audit_inode
1219 * can only collect information for the filesystem object's parent.
1220 * This call updates the audit context with the child's information.
1221 * Syscalls that create a new filesystem object must be hooked after
1222 * the object is created. Syscalls that remove a filesystem object
1223 * must be hooked prior, in order to capture the target inode during
1224 * unsuccessful attempts.
1226 void __audit_inode_child(const char *dname, const struct inode *inode,
1230 struct audit_context *context = current->audit_context;
1231 const char *found_name = NULL;
1234 if (!context->in_syscall)
1237 /* determine matching parent */
1239 goto update_context;
1240 for (idx = 0; idx < context->name_count; idx++)
1241 if (context->names[idx].ino == pino) {
1242 const char *name = context->names[idx].name;
1247 if (audit_compare_dname_path(dname, name, &dirlen) == 0) {
1248 context->names[idx].name_len = dirlen;
1255 idx = context->name_count++;
1257 context->ino_count++;
1259 /* Re-use the name belonging to the slot for a matching parent directory.
1260 * All names for this context are relinquished in audit_free_names() */
1261 context->names[idx].name = found_name;
1262 context->names[idx].name_len = AUDIT_NAME_FULL;
1263 context->names[idx].name_put = 0; /* don't call __putname() */
1266 context->names[idx].ino = inode->i_ino;
1267 context->names[idx].dev = inode->i_sb->s_dev;
1268 context->names[idx].mode = inode->i_mode;
1269 context->names[idx].uid = inode->i_uid;
1270 context->names[idx].gid = inode->i_gid;
1271 context->names[idx].rdev = inode->i_rdev;
1272 audit_inode_context(idx, inode);
1274 context->names[idx].ino = (unsigned long)-1;
1278 * auditsc_get_stamp - get local copies of audit_context values
1279 * @ctx: audit_context for the task
1280 * @t: timespec to store time recorded in the audit_context
1281 * @serial: serial value that is recorded in the audit_context
1283 * Also sets the context as auditable.
1285 void auditsc_get_stamp(struct audit_context *ctx,
1286 struct timespec *t, unsigned int *serial)
1289 ctx->serial = audit_serial();
1290 t->tv_sec = ctx->ctime.tv_sec;
1291 t->tv_nsec = ctx->ctime.tv_nsec;
1292 *serial = ctx->serial;
1297 * audit_set_loginuid - set a task's audit_context loginuid
1298 * @task: task whose audit context is being modified
1299 * @loginuid: loginuid value
1303 * Called (set) from fs/proc/base.c::proc_loginuid_write().
1305 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
1307 struct audit_context *context = task->audit_context;
1310 /* Only log if audit is enabled */
1311 if (context->in_syscall) {
1312 struct audit_buffer *ab;
1314 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1316 audit_log_format(ab, "login pid=%d uid=%u "
1317 "old auid=%u new auid=%u",
1318 task->pid, task->uid,
1319 context->loginuid, loginuid);
1323 context->loginuid = loginuid;
1329 * audit_get_loginuid - get the loginuid for an audit_context
1330 * @ctx: the audit_context
1332 * Returns the context's loginuid or -1 if @ctx is NULL.
1334 uid_t audit_get_loginuid(struct audit_context *ctx)
1336 return ctx ? ctx->loginuid : -1;
1340 * __audit_mq_open - record audit data for a POSIX MQ open
1343 * @u_attr: queue attributes
1345 * Returns 0 for success or NULL context or < 0 on error.
1347 int __audit_mq_open(int oflag, mode_t mode, struct mq_attr __user *u_attr)
1349 struct audit_aux_data_mq_open *ax;
1350 struct audit_context *context = current->audit_context;
1355 if (likely(!context))
1358 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1362 if (u_attr != NULL) {
1363 if (copy_from_user(&ax->attr, u_attr, sizeof(ax->attr))) {
1368 memset(&ax->attr, 0, sizeof(ax->attr));
1373 ax->d.type = AUDIT_MQ_OPEN;
1374 ax->d.next = context->aux;
1375 context->aux = (void *)ax;
1380 * __audit_mq_timedsend - record audit data for a POSIX MQ timed send
1381 * @mqdes: MQ descriptor
1382 * @msg_len: Message length
1383 * @msg_prio: Message priority
1384 * @u_abs_timeout: Message timeout in absolute time
1386 * Returns 0 for success or NULL context or < 0 on error.
1388 int __audit_mq_timedsend(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
1389 const struct timespec __user *u_abs_timeout)
1391 struct audit_aux_data_mq_sendrecv *ax;
1392 struct audit_context *context = current->audit_context;
1397 if (likely(!context))
1400 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1404 if (u_abs_timeout != NULL) {
1405 if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
1410 memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
1413 ax->msg_len = msg_len;
1414 ax->msg_prio = msg_prio;
1416 ax->d.type = AUDIT_MQ_SENDRECV;
1417 ax->d.next = context->aux;
1418 context->aux = (void *)ax;
1423 * __audit_mq_timedreceive - record audit data for a POSIX MQ timed receive
1424 * @mqdes: MQ descriptor
1425 * @msg_len: Message length
1426 * @u_msg_prio: Message priority
1427 * @u_abs_timeout: Message timeout in absolute time
1429 * Returns 0 for success or NULL context or < 0 on error.
1431 int __audit_mq_timedreceive(mqd_t mqdes, size_t msg_len,
1432 unsigned int __user *u_msg_prio,
1433 const struct timespec __user *u_abs_timeout)
1435 struct audit_aux_data_mq_sendrecv *ax;
1436 struct audit_context *context = current->audit_context;
1441 if (likely(!context))
1444 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1448 if (u_msg_prio != NULL) {
1449 if (get_user(ax->msg_prio, u_msg_prio)) {
1456 if (u_abs_timeout != NULL) {
1457 if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
1462 memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
1465 ax->msg_len = msg_len;
1467 ax->d.type = AUDIT_MQ_SENDRECV;
1468 ax->d.next = context->aux;
1469 context->aux = (void *)ax;
1474 * __audit_mq_notify - record audit data for a POSIX MQ notify
1475 * @mqdes: MQ descriptor
1476 * @u_notification: Notification event
1478 * Returns 0 for success or NULL context or < 0 on error.
1481 int __audit_mq_notify(mqd_t mqdes, const struct sigevent __user *u_notification)
1483 struct audit_aux_data_mq_notify *ax;
1484 struct audit_context *context = current->audit_context;
1489 if (likely(!context))
1492 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1496 if (u_notification != NULL) {
1497 if (copy_from_user(&ax->notification, u_notification, sizeof(ax->notification))) {
1502 memset(&ax->notification, 0, sizeof(ax->notification));
1506 ax->d.type = AUDIT_MQ_NOTIFY;
1507 ax->d.next = context->aux;
1508 context->aux = (void *)ax;
1513 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
1514 * @mqdes: MQ descriptor
1517 * Returns 0 for success or NULL context or < 0 on error.
1519 int __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
1521 struct audit_aux_data_mq_getsetattr *ax;
1522 struct audit_context *context = current->audit_context;
1527 if (likely(!context))
1530 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1535 ax->mqstat = *mqstat;
1537 ax->d.type = AUDIT_MQ_GETSETATTR;
1538 ax->d.next = context->aux;
1539 context->aux = (void *)ax;
1544 * audit_ipc_obj - record audit data for ipc object
1545 * @ipcp: ipc permissions
1547 * Returns 0 for success or NULL context or < 0 on error.
1549 int __audit_ipc_obj(struct kern_ipc_perm *ipcp)
1551 struct audit_aux_data_ipcctl *ax;
1552 struct audit_context *context = current->audit_context;
1554 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1558 ax->uid = ipcp->uid;
1559 ax->gid = ipcp->gid;
1560 ax->mode = ipcp->mode;
1561 selinux_get_ipc_sid(ipcp, &ax->osid);
1563 ax->d.type = AUDIT_IPC;
1564 ax->d.next = context->aux;
1565 context->aux = (void *)ax;
1570 * audit_ipc_set_perm - record audit data for new ipc permissions
1571 * @qbytes: msgq bytes
1572 * @uid: msgq user id
1573 * @gid: msgq group id
1574 * @mode: msgq mode (permissions)
1576 * Returns 0 for success or NULL context or < 0 on error.
1578 int __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
1580 struct audit_aux_data_ipcctl *ax;
1581 struct audit_context *context = current->audit_context;
1583 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1587 ax->qbytes = qbytes;
1592 ax->d.type = AUDIT_IPC_SET_PERM;
1593 ax->d.next = context->aux;
1594 context->aux = (void *)ax;
1598 int audit_bprm(struct linux_binprm *bprm)
1600 struct audit_aux_data_execve *ax;
1601 struct audit_context *context = current->audit_context;
1602 unsigned long p, next;
1605 if (likely(!audit_enabled || !context))
1608 ax = kmalloc(sizeof(*ax) + PAGE_SIZE * MAX_ARG_PAGES - bprm->p,
1613 ax->argc = bprm->argc;
1614 ax->envc = bprm->envc;
1615 for (p = bprm->p, to = ax->mem; p < MAX_ARG_PAGES*PAGE_SIZE; p = next) {
1616 struct page *page = bprm->page[p / PAGE_SIZE];
1617 void *kaddr = kmap(page);
1618 next = (p + PAGE_SIZE) & ~(PAGE_SIZE - 1);
1619 memcpy(to, kaddr + (p & (PAGE_SIZE - 1)), next - p);
1624 ax->d.type = AUDIT_EXECVE;
1625 ax->d.next = context->aux;
1626 context->aux = (void *)ax;
1632 * audit_socketcall - record audit data for sys_socketcall
1633 * @nargs: number of args
1636 * Returns 0 for success or NULL context or < 0 on error.
1638 int audit_socketcall(int nargs, unsigned long *args)
1640 struct audit_aux_data_socketcall *ax;
1641 struct audit_context *context = current->audit_context;
1643 if (likely(!context))
1646 ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
1651 memcpy(ax->args, args, nargs * sizeof(unsigned long));
1653 ax->d.type = AUDIT_SOCKETCALL;
1654 ax->d.next = context->aux;
1655 context->aux = (void *)ax;
1660 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
1661 * @len: data length in user space
1662 * @a: data address in kernel space
1664 * Returns 0 for success or NULL context or < 0 on error.
1666 int audit_sockaddr(int len, void *a)
1668 struct audit_aux_data_sockaddr *ax;
1669 struct audit_context *context = current->audit_context;
1671 if (likely(!context))
1674 ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
1679 memcpy(ax->a, a, len);
1681 ax->d.type = AUDIT_SOCKADDR;
1682 ax->d.next = context->aux;
1683 context->aux = (void *)ax;
1688 * audit_avc_path - record the granting or denial of permissions
1689 * @dentry: dentry to record
1690 * @mnt: mnt to record
1692 * Returns 0 for success or NULL context or < 0 on error.
1694 * Called from security/selinux/avc.c::avc_audit()
1696 int audit_avc_path(struct dentry *dentry, struct vfsmount *mnt)
1698 struct audit_aux_data_path *ax;
1699 struct audit_context *context = current->audit_context;
1701 if (likely(!context))
1704 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1708 ax->dentry = dget(dentry);
1709 ax->mnt = mntget(mnt);
1711 ax->d.type = AUDIT_AVC_PATH;
1712 ax->d.next = context->aux;
1713 context->aux = (void *)ax;
1718 * audit_signal_info - record signal info for shutting down audit subsystem
1719 * @sig: signal value
1720 * @t: task being signaled
1722 * If the audit subsystem is being terminated, record the task (pid)
1723 * and uid that is doing that.
1725 void __audit_signal_info(int sig, struct task_struct *t)
1727 extern pid_t audit_sig_pid;
1728 extern uid_t audit_sig_uid;
1729 extern u32 audit_sig_sid;
1731 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1) {
1732 struct task_struct *tsk = current;
1733 struct audit_context *ctx = tsk->audit_context;
1734 audit_sig_pid = tsk->pid;
1736 audit_sig_uid = ctx->loginuid;
1738 audit_sig_uid = tsk->uid;
1739 selinux_get_task_sid(tsk, &audit_sig_sid);