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_mq_open {
128 struct audit_aux_data d;
134 struct audit_aux_data_mq_sendrecv {
135 struct audit_aux_data d;
138 unsigned int msg_prio;
139 struct timespec abs_timeout;
142 struct audit_aux_data_mq_notify {
143 struct audit_aux_data d;
145 struct sigevent notification;
148 struct audit_aux_data_execve {
149 struct audit_aux_data d;
152 struct mm_struct *mm;
155 struct audit_aux_data_fd_pair {
156 struct audit_aux_data d;
160 struct audit_aux_data_pids {
161 struct audit_aux_data d;
162 pid_t target_pid[AUDIT_AUX_PIDS];
163 uid_t target_auid[AUDIT_AUX_PIDS];
164 uid_t target_uid[AUDIT_AUX_PIDS];
165 unsigned int target_sessionid[AUDIT_AUX_PIDS];
166 u32 target_sid[AUDIT_AUX_PIDS];
167 char target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
171 struct audit_aux_data_bprm_fcaps {
172 struct audit_aux_data d;
173 struct audit_cap_data fcap;
174 unsigned int fcap_ver;
175 struct audit_cap_data old_pcap;
176 struct audit_cap_data new_pcap;
179 struct audit_aux_data_capset {
180 struct audit_aux_data d;
182 struct audit_cap_data cap;
185 struct audit_tree_refs {
186 struct audit_tree_refs *next;
187 struct audit_chunk *c[31];
190 /* The per-task audit context. */
191 struct audit_context {
192 int dummy; /* must be the first element */
193 int in_syscall; /* 1 if task is in a syscall */
194 enum audit_state state;
195 unsigned int serial; /* serial number for record */
196 struct timespec ctime; /* time of syscall entry */
197 int major; /* syscall number */
198 unsigned long argv[4]; /* syscall arguments */
199 int return_valid; /* return code is valid */
200 long return_code;/* syscall return code */
201 int auditable; /* 1 if record should be written */
203 struct audit_names names[AUDIT_NAMES];
204 char * filterkey; /* key for rule that triggered record */
206 struct audit_context *previous; /* For nested syscalls */
207 struct audit_aux_data *aux;
208 struct audit_aux_data *aux_pids;
209 struct sockaddr_storage *sockaddr;
211 /* Save things to print about task_struct */
213 uid_t uid, euid, suid, fsuid;
214 gid_t gid, egid, sgid, fsgid;
215 unsigned long personality;
221 unsigned int target_sessionid;
223 char target_comm[TASK_COMM_LEN];
225 struct audit_tree_refs *trees, *first_trees;
243 unsigned long qbytes;
247 struct mq_attr mqstat;
257 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
258 static inline int open_arg(int flags, int mask)
260 int n = ACC_MODE(flags);
261 if (flags & (O_TRUNC | O_CREAT))
262 n |= AUDIT_PERM_WRITE;
266 static int audit_match_perm(struct audit_context *ctx, int mask)
273 switch (audit_classify_syscall(ctx->arch, n)) {
275 if ((mask & AUDIT_PERM_WRITE) &&
276 audit_match_class(AUDIT_CLASS_WRITE, n))
278 if ((mask & AUDIT_PERM_READ) &&
279 audit_match_class(AUDIT_CLASS_READ, n))
281 if ((mask & AUDIT_PERM_ATTR) &&
282 audit_match_class(AUDIT_CLASS_CHATTR, n))
285 case 1: /* 32bit on biarch */
286 if ((mask & AUDIT_PERM_WRITE) &&
287 audit_match_class(AUDIT_CLASS_WRITE_32, n))
289 if ((mask & AUDIT_PERM_READ) &&
290 audit_match_class(AUDIT_CLASS_READ_32, n))
292 if ((mask & AUDIT_PERM_ATTR) &&
293 audit_match_class(AUDIT_CLASS_CHATTR_32, n))
297 return mask & ACC_MODE(ctx->argv[1]);
299 return mask & ACC_MODE(ctx->argv[2]);
300 case 4: /* socketcall */
301 return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
303 return mask & AUDIT_PERM_EXEC;
309 static int audit_match_filetype(struct audit_context *ctx, int which)
311 unsigned index = which & ~S_IFMT;
312 mode_t mode = which & S_IFMT;
317 if (index >= ctx->name_count)
319 if (ctx->names[index].ino == -1)
321 if ((ctx->names[index].mode ^ mode) & S_IFMT)
327 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
328 * ->first_trees points to its beginning, ->trees - to the current end of data.
329 * ->tree_count is the number of free entries in array pointed to by ->trees.
330 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
331 * "empty" becomes (p, p, 31) afterwards. We don't shrink the list (and seriously,
332 * it's going to remain 1-element for almost any setup) until we free context itself.
333 * References in it _are_ dropped - at the same time we free/drop aux stuff.
336 #ifdef CONFIG_AUDIT_TREE
337 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
339 struct audit_tree_refs *p = ctx->trees;
340 int left = ctx->tree_count;
342 p->c[--left] = chunk;
343 ctx->tree_count = left;
352 ctx->tree_count = 30;
358 static int grow_tree_refs(struct audit_context *ctx)
360 struct audit_tree_refs *p = ctx->trees;
361 ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
367 p->next = ctx->trees;
369 ctx->first_trees = ctx->trees;
370 ctx->tree_count = 31;
375 static void unroll_tree_refs(struct audit_context *ctx,
376 struct audit_tree_refs *p, int count)
378 #ifdef CONFIG_AUDIT_TREE
379 struct audit_tree_refs *q;
382 /* we started with empty chain */
383 p = ctx->first_trees;
385 /* if the very first allocation has failed, nothing to do */
390 for (q = p; q != ctx->trees; q = q->next, n = 31) {
392 audit_put_chunk(q->c[n]);
396 while (n-- > ctx->tree_count) {
397 audit_put_chunk(q->c[n]);
401 ctx->tree_count = count;
405 static void free_tree_refs(struct audit_context *ctx)
407 struct audit_tree_refs *p, *q;
408 for (p = ctx->first_trees; p; p = q) {
414 static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
416 #ifdef CONFIG_AUDIT_TREE
417 struct audit_tree_refs *p;
422 for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
423 for (n = 0; n < 31; n++)
424 if (audit_tree_match(p->c[n], tree))
429 for (n = ctx->tree_count; n < 31; n++)
430 if (audit_tree_match(p->c[n], tree))
437 /* Determine if any context name data matches a rule's watch data */
438 /* Compare a task_struct with an audit_rule. Return 1 on match, 0
440 static int audit_filter_rules(struct task_struct *tsk,
441 struct audit_krule *rule,
442 struct audit_context *ctx,
443 struct audit_names *name,
444 enum audit_state *state)
446 const struct cred *cred = get_task_cred(tsk);
447 int i, j, need_sid = 1;
450 for (i = 0; i < rule->field_count; i++) {
451 struct audit_field *f = &rule->fields[i];
456 result = audit_comparator(tsk->pid, f->op, f->val);
461 ctx->ppid = sys_getppid();
462 result = audit_comparator(ctx->ppid, f->op, f->val);
466 result = audit_comparator(cred->uid, f->op, f->val);
469 result = audit_comparator(cred->euid, f->op, f->val);
472 result = audit_comparator(cred->suid, f->op, f->val);
475 result = audit_comparator(cred->fsuid, f->op, f->val);
478 result = audit_comparator(cred->gid, f->op, f->val);
481 result = audit_comparator(cred->egid, f->op, f->val);
484 result = audit_comparator(cred->sgid, f->op, f->val);
487 result = audit_comparator(cred->fsgid, f->op, f->val);
490 result = audit_comparator(tsk->personality, f->op, f->val);
494 result = audit_comparator(ctx->arch, f->op, f->val);
498 if (ctx && ctx->return_valid)
499 result = audit_comparator(ctx->return_code, f->op, f->val);
502 if (ctx && ctx->return_valid) {
504 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
506 result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
511 result = audit_comparator(MAJOR(name->dev),
514 for (j = 0; j < ctx->name_count; j++) {
515 if (audit_comparator(MAJOR(ctx->names[j].dev), f->op, f->val)) {
524 result = audit_comparator(MINOR(name->dev),
527 for (j = 0; j < ctx->name_count; j++) {
528 if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
537 result = (name->ino == f->val);
539 for (j = 0; j < ctx->name_count; j++) {
540 if (audit_comparator(ctx->names[j].ino, f->op, f->val)) {
548 if (name && rule->watch->ino != (unsigned long)-1)
549 result = (name->dev == rule->watch->dev &&
550 name->ino == rule->watch->ino);
554 result = match_tree_refs(ctx, rule->tree);
559 result = audit_comparator(tsk->loginuid, f->op, f->val);
561 case AUDIT_SUBJ_USER:
562 case AUDIT_SUBJ_ROLE:
563 case AUDIT_SUBJ_TYPE:
566 /* NOTE: this may return negative values indicating
567 a temporary error. We simply treat this as a
568 match for now to avoid losing information that
569 may be wanted. An error message will also be
573 security_task_getsecid(tsk, &sid);
576 result = security_audit_rule_match(sid, f->type,
585 case AUDIT_OBJ_LEV_LOW:
586 case AUDIT_OBJ_LEV_HIGH:
587 /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
590 /* Find files that match */
592 result = security_audit_rule_match(
593 name->osid, f->type, f->op,
596 for (j = 0; j < ctx->name_count; j++) {
597 if (security_audit_rule_match(
606 /* Find ipc objects that match */
607 if (!ctx || ctx->type != AUDIT_IPC)
609 if (security_audit_rule_match(ctx->ipc.osid,
620 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
622 case AUDIT_FILTERKEY:
623 /* ignore this field for filtering */
627 result = audit_match_perm(ctx, f->val);
630 result = audit_match_filetype(ctx, f->val);
639 if (rule->filterkey && ctx)
640 ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
641 switch (rule->action) {
642 case AUDIT_NEVER: *state = AUDIT_DISABLED; break;
643 case AUDIT_ALWAYS: *state = AUDIT_RECORD_CONTEXT; break;
649 /* At process creation time, we can determine if system-call auditing is
650 * completely disabled for this task. Since we only have the task
651 * structure at this point, we can only check uid and gid.
653 static enum audit_state audit_filter_task(struct task_struct *tsk)
655 struct audit_entry *e;
656 enum audit_state state;
659 list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
660 if (audit_filter_rules(tsk, &e->rule, NULL, NULL, &state)) {
666 return AUDIT_BUILD_CONTEXT;
669 /* At syscall entry and exit time, this filter is called if the
670 * audit_state is not low enough that auditing cannot take place, but is
671 * also not high enough that we already know we have to write an audit
672 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
674 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
675 struct audit_context *ctx,
676 struct list_head *list)
678 struct audit_entry *e;
679 enum audit_state state;
681 if (audit_pid && tsk->tgid == audit_pid)
682 return AUDIT_DISABLED;
685 if (!list_empty(list)) {
686 int word = AUDIT_WORD(ctx->major);
687 int bit = AUDIT_BIT(ctx->major);
689 list_for_each_entry_rcu(e, list, list) {
690 if ((e->rule.mask[word] & bit) == bit &&
691 audit_filter_rules(tsk, &e->rule, ctx, NULL,
699 return AUDIT_BUILD_CONTEXT;
702 /* At syscall exit time, this filter is called if any audit_names[] have been
703 * collected during syscall processing. We only check rules in sublists at hash
704 * buckets applicable to the inode numbers in audit_names[].
705 * Regarding audit_state, same rules apply as for audit_filter_syscall().
707 enum audit_state audit_filter_inodes(struct task_struct *tsk,
708 struct audit_context *ctx)
711 struct audit_entry *e;
712 enum audit_state state;
714 if (audit_pid && tsk->tgid == audit_pid)
715 return AUDIT_DISABLED;
718 for (i = 0; i < ctx->name_count; i++) {
719 int word = AUDIT_WORD(ctx->major);
720 int bit = AUDIT_BIT(ctx->major);
721 struct audit_names *n = &ctx->names[i];
722 int h = audit_hash_ino((u32)n->ino);
723 struct list_head *list = &audit_inode_hash[h];
725 if (list_empty(list))
728 list_for_each_entry_rcu(e, list, list) {
729 if ((e->rule.mask[word] & bit) == bit &&
730 audit_filter_rules(tsk, &e->rule, ctx, n, &state)) {
737 return AUDIT_BUILD_CONTEXT;
740 void audit_set_auditable(struct audit_context *ctx)
745 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
749 struct audit_context *context = tsk->audit_context;
751 if (likely(!context))
753 context->return_valid = return_valid;
756 * we need to fix up the return code in the audit logs if the actual
757 * return codes are later going to be fixed up by the arch specific
760 * This is actually a test for:
761 * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
762 * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
764 * but is faster than a bunch of ||
766 if (unlikely(return_code <= -ERESTARTSYS) &&
767 (return_code >= -ERESTART_RESTARTBLOCK) &&
768 (return_code != -ENOIOCTLCMD))
769 context->return_code = -EINTR;
771 context->return_code = return_code;
773 if (context->in_syscall && !context->dummy && !context->auditable) {
774 enum audit_state state;
776 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
777 if (state == AUDIT_RECORD_CONTEXT) {
778 context->auditable = 1;
782 state = audit_filter_inodes(tsk, context);
783 if (state == AUDIT_RECORD_CONTEXT)
784 context->auditable = 1;
790 tsk->audit_context = NULL;
794 static inline void audit_free_names(struct audit_context *context)
799 if (context->auditable
800 ||context->put_count + context->ino_count != context->name_count) {
801 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
802 " name_count=%d put_count=%d"
803 " ino_count=%d [NOT freeing]\n",
805 context->serial, context->major, context->in_syscall,
806 context->name_count, context->put_count,
808 for (i = 0; i < context->name_count; i++) {
809 printk(KERN_ERR "names[%d] = %p = %s\n", i,
810 context->names[i].name,
811 context->names[i].name ?: "(null)");
818 context->put_count = 0;
819 context->ino_count = 0;
822 for (i = 0; i < context->name_count; i++) {
823 if (context->names[i].name && context->names[i].name_put)
824 __putname(context->names[i].name);
826 context->name_count = 0;
827 path_put(&context->pwd);
828 context->pwd.dentry = NULL;
829 context->pwd.mnt = NULL;
832 static inline void audit_free_aux(struct audit_context *context)
834 struct audit_aux_data *aux;
836 while ((aux = context->aux)) {
837 context->aux = aux->next;
840 while ((aux = context->aux_pids)) {
841 context->aux_pids = aux->next;
846 static inline void audit_zero_context(struct audit_context *context,
847 enum audit_state state)
849 memset(context, 0, sizeof(*context));
850 context->state = state;
853 static inline struct audit_context *audit_alloc_context(enum audit_state state)
855 struct audit_context *context;
857 if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
859 audit_zero_context(context, state);
864 * audit_alloc - allocate an audit context block for a task
867 * Filter on the task information and allocate a per-task audit context
868 * if necessary. Doing so turns on system call auditing for the
869 * specified task. This is called from copy_process, so no lock is
872 int audit_alloc(struct task_struct *tsk)
874 struct audit_context *context;
875 enum audit_state state;
877 if (likely(!audit_ever_enabled))
878 return 0; /* Return if not auditing. */
880 state = audit_filter_task(tsk);
881 if (likely(state == AUDIT_DISABLED))
884 if (!(context = audit_alloc_context(state))) {
885 audit_log_lost("out of memory in audit_alloc");
889 tsk->audit_context = context;
890 set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
894 static inline void audit_free_context(struct audit_context *context)
896 struct audit_context *previous;
900 previous = context->previous;
901 if (previous || (count && count < 10)) {
903 printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
904 " freeing multiple contexts (%d)\n",
905 context->serial, context->major,
906 context->name_count, count);
908 audit_free_names(context);
909 unroll_tree_refs(context, NULL, 0);
910 free_tree_refs(context);
911 audit_free_aux(context);
912 kfree(context->filterkey);
913 kfree(context->sockaddr);
918 printk(KERN_ERR "audit: freed %d contexts\n", count);
921 void audit_log_task_context(struct audit_buffer *ab)
928 security_task_getsecid(current, &sid);
932 error = security_secid_to_secctx(sid, &ctx, &len);
934 if (error != -EINVAL)
939 audit_log_format(ab, " subj=%s", ctx);
940 security_release_secctx(ctx, len);
944 audit_panic("error in audit_log_task_context");
948 EXPORT_SYMBOL(audit_log_task_context);
950 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
952 char name[sizeof(tsk->comm)];
953 struct mm_struct *mm = tsk->mm;
954 struct vm_area_struct *vma;
958 get_task_comm(name, tsk);
959 audit_log_format(ab, " comm=");
960 audit_log_untrustedstring(ab, name);
963 down_read(&mm->mmap_sem);
966 if ((vma->vm_flags & VM_EXECUTABLE) &&
968 audit_log_d_path(ab, "exe=",
969 &vma->vm_file->f_path);
974 up_read(&mm->mmap_sem);
976 audit_log_task_context(ab);
979 static int audit_log_pid_context(struct audit_context *context, pid_t pid,
980 uid_t auid, uid_t uid, unsigned int sessionid,
983 struct audit_buffer *ab;
988 ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
992 audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
994 if (security_secid_to_secctx(sid, &ctx, &len)) {
995 audit_log_format(ab, " obj=(none)");
998 audit_log_format(ab, " obj=%s", ctx);
999 security_release_secctx(ctx, len);
1001 audit_log_format(ab, " ocomm=");
1002 audit_log_untrustedstring(ab, comm);
1009 * to_send and len_sent accounting are very loose estimates. We aren't
1010 * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1011 * within about 500 bytes (next page boundry)
1013 * why snprintf? an int is up to 12 digits long. if we just assumed when
1014 * logging that a[%d]= was going to be 16 characters long we would be wasting
1015 * space in every audit message. In one 7500 byte message we can log up to
1016 * about 1000 min size arguments. That comes down to about 50% waste of space
1017 * if we didn't do the snprintf to find out how long arg_num_len was.
1019 static int audit_log_single_execve_arg(struct audit_context *context,
1020 struct audit_buffer **ab,
1023 const char __user *p,
1026 char arg_num_len_buf[12];
1027 const char __user *tmp_p = p;
1028 /* how many digits are in arg_num? 3 is the length of a=\n */
1029 size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 3;
1030 size_t len, len_left, to_send;
1031 size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1032 unsigned int i, has_cntl = 0, too_long = 0;
1035 /* strnlen_user includes the null we don't want to send */
1036 len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1039 * We just created this mm, if we can't find the strings
1040 * we just copied into it something is _very_ wrong. Similar
1041 * for strings that are too long, we should not have created
1044 if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1046 send_sig(SIGKILL, current, 0);
1050 /* walk the whole argument looking for non-ascii chars */
1052 if (len_left > MAX_EXECVE_AUDIT_LEN)
1053 to_send = MAX_EXECVE_AUDIT_LEN;
1056 ret = copy_from_user(buf, tmp_p, to_send);
1058 * There is no reason for this copy to be short. We just
1059 * copied them here, and the mm hasn't been exposed to user-
1064 send_sig(SIGKILL, current, 0);
1067 buf[to_send] = '\0';
1068 has_cntl = audit_string_contains_control(buf, to_send);
1071 * hex messages get logged as 2 bytes, so we can only
1072 * send half as much in each message
1074 max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1077 len_left -= to_send;
1079 } while (len_left > 0);
1083 if (len > max_execve_audit_len)
1086 /* rewalk the argument actually logging the message */
1087 for (i = 0; len_left > 0; i++) {
1090 if (len_left > max_execve_audit_len)
1091 to_send = max_execve_audit_len;
1095 /* do we have space left to send this argument in this ab? */
1096 room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1098 room_left -= (to_send * 2);
1100 room_left -= to_send;
1101 if (room_left < 0) {
1104 *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1110 * first record needs to say how long the original string was
1111 * so we can be sure nothing was lost.
1113 if ((i == 0) && (too_long))
1114 audit_log_format(*ab, "a%d_len=%zu ", arg_num,
1115 has_cntl ? 2*len : len);
1118 * normally arguments are small enough to fit and we already
1119 * filled buf above when we checked for control characters
1120 * so don't bother with another copy_from_user
1122 if (len >= max_execve_audit_len)
1123 ret = copy_from_user(buf, p, to_send);
1128 send_sig(SIGKILL, current, 0);
1131 buf[to_send] = '\0';
1133 /* actually log it */
1134 audit_log_format(*ab, "a%d", arg_num);
1136 audit_log_format(*ab, "[%d]", i);
1137 audit_log_format(*ab, "=");
1139 audit_log_n_hex(*ab, buf, to_send);
1141 audit_log_format(*ab, "\"%s\"", buf);
1142 audit_log_format(*ab, "\n");
1145 len_left -= to_send;
1146 *len_sent += arg_num_len;
1148 *len_sent += to_send * 2;
1150 *len_sent += to_send;
1152 /* include the null we didn't log */
1156 static void audit_log_execve_info(struct audit_context *context,
1157 struct audit_buffer **ab,
1158 struct audit_aux_data_execve *axi)
1161 size_t len, len_sent = 0;
1162 const char __user *p;
1165 if (axi->mm != current->mm)
1166 return; /* execve failed, no additional info */
1168 p = (const char __user *)axi->mm->arg_start;
1170 audit_log_format(*ab, "argc=%d ", axi->argc);
1173 * we need some kernel buffer to hold the userspace args. Just
1174 * allocate one big one rather than allocating one of the right size
1175 * for every single argument inside audit_log_single_execve_arg()
1176 * should be <8k allocation so should be pretty safe.
1178 buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1180 audit_panic("out of memory for argv string\n");
1184 for (i = 0; i < axi->argc; i++) {
1185 len = audit_log_single_execve_arg(context, ab, i,
1194 static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1198 audit_log_format(ab, " %s=", prefix);
1199 CAP_FOR_EACH_U32(i) {
1200 audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1204 static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1206 kernel_cap_t *perm = &name->fcap.permitted;
1207 kernel_cap_t *inh = &name->fcap.inheritable;
1210 if (!cap_isclear(*perm)) {
1211 audit_log_cap(ab, "cap_fp", perm);
1214 if (!cap_isclear(*inh)) {
1215 audit_log_cap(ab, "cap_fi", inh);
1220 audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1223 static void show_special(struct audit_context *context, int *call_panic)
1225 struct audit_buffer *ab;
1228 ab = audit_log_start(context, GFP_KERNEL, context->type);
1232 switch (context->type) {
1233 case AUDIT_SOCKETCALL: {
1234 int nargs = context->socketcall.nargs;
1235 audit_log_format(ab, "nargs=%d", nargs);
1236 for (i = 0; i < nargs; i++)
1237 audit_log_format(ab, " a%d=%lx", i,
1238 context->socketcall.args[i]);
1241 u32 osid = context->ipc.osid;
1243 audit_log_format(ab, "ouid=%u ogid=%u mode=%#o",
1244 context->ipc.uid, context->ipc.gid, context->ipc.mode);
1248 if (security_secid_to_secctx(osid, &ctx, &len)) {
1249 audit_log_format(ab, " osid=%u", osid);
1252 audit_log_format(ab, " obj=%s", ctx);
1253 security_release_secctx(ctx, len);
1256 if (context->ipc.has_perm) {
1258 ab = audit_log_start(context, GFP_KERNEL,
1259 AUDIT_IPC_SET_PERM);
1260 audit_log_format(ab,
1261 "qbytes=%lx ouid=%u ogid=%u mode=%#o",
1262 context->ipc.qbytes,
1263 context->ipc.perm_uid,
1264 context->ipc.perm_gid,
1265 context->ipc.perm_mode);
1270 case AUDIT_MQ_GETSETATTR: {
1271 struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1272 audit_log_format(ab,
1273 "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1275 context->mq_getsetattr.mqdes,
1276 attr->mq_flags, attr->mq_maxmsg,
1277 attr->mq_msgsize, attr->mq_curmsgs);
1283 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1285 const struct cred *cred;
1286 int i, call_panic = 0;
1287 struct audit_buffer *ab;
1288 struct audit_aux_data *aux;
1291 /* tsk == current */
1292 context->pid = tsk->pid;
1294 context->ppid = sys_getppid();
1295 cred = current_cred();
1296 context->uid = cred->uid;
1297 context->gid = cred->gid;
1298 context->euid = cred->euid;
1299 context->suid = cred->suid;
1300 context->fsuid = cred->fsuid;
1301 context->egid = cred->egid;
1302 context->sgid = cred->sgid;
1303 context->fsgid = cred->fsgid;
1304 context->personality = tsk->personality;
1306 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1308 return; /* audit_panic has been called */
1309 audit_log_format(ab, "arch=%x syscall=%d",
1310 context->arch, context->major);
1311 if (context->personality != PER_LINUX)
1312 audit_log_format(ab, " per=%lx", context->personality);
1313 if (context->return_valid)
1314 audit_log_format(ab, " success=%s exit=%ld",
1315 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1316 context->return_code);
1318 spin_lock_irq(&tsk->sighand->siglock);
1319 if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1320 tty = tsk->signal->tty->name;
1323 spin_unlock_irq(&tsk->sighand->siglock);
1325 audit_log_format(ab,
1326 " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1327 " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1328 " euid=%u suid=%u fsuid=%u"
1329 " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1334 context->name_count,
1340 context->euid, context->suid, context->fsuid,
1341 context->egid, context->sgid, context->fsgid, tty,
1345 audit_log_task_info(ab, tsk);
1346 if (context->filterkey) {
1347 audit_log_format(ab, " key=");
1348 audit_log_untrustedstring(ab, context->filterkey);
1350 audit_log_format(ab, " key=(null)");
1353 for (aux = context->aux; aux; aux = aux->next) {
1355 ab = audit_log_start(context, GFP_KERNEL, aux->type);
1357 continue; /* audit_panic has been called */
1359 switch (aux->type) {
1360 case AUDIT_MQ_OPEN: {
1361 struct audit_aux_data_mq_open *axi = (void *)aux;
1362 audit_log_format(ab,
1363 "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
1364 "mq_msgsize=%ld mq_curmsgs=%ld",
1365 axi->oflag, axi->mode, axi->attr.mq_flags,
1366 axi->attr.mq_maxmsg, axi->attr.mq_msgsize,
1367 axi->attr.mq_curmsgs);
1370 case AUDIT_MQ_SENDRECV: {
1371 struct audit_aux_data_mq_sendrecv *axi = (void *)aux;
1372 audit_log_format(ab,
1373 "mqdes=%d msg_len=%zd msg_prio=%u "
1374 "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1375 axi->mqdes, axi->msg_len, axi->msg_prio,
1376 axi->abs_timeout.tv_sec, axi->abs_timeout.tv_nsec);
1379 case AUDIT_MQ_NOTIFY: {
1380 struct audit_aux_data_mq_notify *axi = (void *)aux;
1381 audit_log_format(ab,
1382 "mqdes=%d sigev_signo=%d",
1384 axi->notification.sigev_signo);
1387 case AUDIT_EXECVE: {
1388 struct audit_aux_data_execve *axi = (void *)aux;
1389 audit_log_execve_info(context, &ab, axi);
1392 case AUDIT_FD_PAIR: {
1393 struct audit_aux_data_fd_pair *axs = (void *)aux;
1394 audit_log_format(ab, "fd0=%d fd1=%d", axs->fd[0], axs->fd[1]);
1397 case AUDIT_BPRM_FCAPS: {
1398 struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1399 audit_log_format(ab, "fver=%x", axs->fcap_ver);
1400 audit_log_cap(ab, "fp", &axs->fcap.permitted);
1401 audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1402 audit_log_format(ab, " fe=%d", axs->fcap.fE);
1403 audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1404 audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1405 audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1406 audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1407 audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1408 audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1411 case AUDIT_CAPSET: {
1412 struct audit_aux_data_capset *axs = (void *)aux;
1413 audit_log_format(ab, "pid=%d", axs->pid);
1414 audit_log_cap(ab, "cap_pi", &axs->cap.inheritable);
1415 audit_log_cap(ab, "cap_pp", &axs->cap.permitted);
1416 audit_log_cap(ab, "cap_pe", &axs->cap.effective);
1424 show_special(context, &call_panic);
1426 if (context->sockaddr_len) {
1427 ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1429 audit_log_format(ab, "saddr=");
1430 audit_log_n_hex(ab, (void *)context->sockaddr,
1431 context->sockaddr_len);
1436 for (aux = context->aux_pids; aux; aux = aux->next) {
1437 struct audit_aux_data_pids *axs = (void *)aux;
1439 for (i = 0; i < axs->pid_count; i++)
1440 if (audit_log_pid_context(context, axs->target_pid[i],
1441 axs->target_auid[i],
1443 axs->target_sessionid[i],
1445 axs->target_comm[i]))
1449 if (context->target_pid &&
1450 audit_log_pid_context(context, context->target_pid,
1451 context->target_auid, context->target_uid,
1452 context->target_sessionid,
1453 context->target_sid, context->target_comm))
1456 if (context->pwd.dentry && context->pwd.mnt) {
1457 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1459 audit_log_d_path(ab, "cwd=", &context->pwd);
1463 for (i = 0; i < context->name_count; i++) {
1464 struct audit_names *n = &context->names[i];
1466 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1468 continue; /* audit_panic has been called */
1470 audit_log_format(ab, "item=%d", i);
1473 switch(n->name_len) {
1474 case AUDIT_NAME_FULL:
1475 /* log the full path */
1476 audit_log_format(ab, " name=");
1477 audit_log_untrustedstring(ab, n->name);
1480 /* name was specified as a relative path and the
1481 * directory component is the cwd */
1482 audit_log_d_path(ab, " name=", &context->pwd);
1485 /* log the name's directory component */
1486 audit_log_format(ab, " name=");
1487 audit_log_n_untrustedstring(ab, n->name,
1491 audit_log_format(ab, " name=(null)");
1493 if (n->ino != (unsigned long)-1) {
1494 audit_log_format(ab, " inode=%lu"
1495 " dev=%02x:%02x mode=%#o"
1496 " ouid=%u ogid=%u rdev=%02x:%02x",
1509 if (security_secid_to_secctx(
1510 n->osid, &ctx, &len)) {
1511 audit_log_format(ab, " osid=%u", n->osid);
1514 audit_log_format(ab, " obj=%s", ctx);
1515 security_release_secctx(ctx, len);
1519 audit_log_fcaps(ab, n);
1524 /* Send end of event record to help user space know we are finished */
1525 ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1529 audit_panic("error converting sid to string");
1533 * audit_free - free a per-task audit context
1534 * @tsk: task whose audit context block to free
1536 * Called from copy_process and do_exit
1538 void audit_free(struct task_struct *tsk)
1540 struct audit_context *context;
1542 context = audit_get_context(tsk, 0, 0);
1543 if (likely(!context))
1546 /* Check for system calls that do not go through the exit
1547 * function (e.g., exit_group), then free context block.
1548 * We use GFP_ATOMIC here because we might be doing this
1549 * in the context of the idle thread */
1550 /* that can happen only if we are called from do_exit() */
1551 if (context->in_syscall && context->auditable)
1552 audit_log_exit(context, tsk);
1554 audit_free_context(context);
1558 * audit_syscall_entry - fill in an audit record at syscall entry
1559 * @arch: architecture type
1560 * @major: major syscall type (function)
1561 * @a1: additional syscall register 1
1562 * @a2: additional syscall register 2
1563 * @a3: additional syscall register 3
1564 * @a4: additional syscall register 4
1566 * Fill in audit context at syscall entry. This only happens if the
1567 * audit context was created when the task was created and the state or
1568 * filters demand the audit context be built. If the state from the
1569 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1570 * then the record will be written at syscall exit time (otherwise, it
1571 * will only be written if another part of the kernel requests that it
1574 void audit_syscall_entry(int arch, int major,
1575 unsigned long a1, unsigned long a2,
1576 unsigned long a3, unsigned long a4)
1578 struct task_struct *tsk = current;
1579 struct audit_context *context = tsk->audit_context;
1580 enum audit_state state;
1582 if (unlikely(!context))
1586 * This happens only on certain architectures that make system
1587 * calls in kernel_thread via the entry.S interface, instead of
1588 * with direct calls. (If you are porting to a new
1589 * architecture, hitting this condition can indicate that you
1590 * got the _exit/_leave calls backward in entry.S.)
1594 * ppc64 yes (see arch/powerpc/platforms/iseries/misc.S)
1596 * This also happens with vm86 emulation in a non-nested manner
1597 * (entries without exits), so this case must be caught.
1599 if (context->in_syscall) {
1600 struct audit_context *newctx;
1604 "audit(:%d) pid=%d in syscall=%d;"
1605 " entering syscall=%d\n",
1606 context->serial, tsk->pid, context->major, major);
1608 newctx = audit_alloc_context(context->state);
1610 newctx->previous = context;
1612 tsk->audit_context = newctx;
1614 /* If we can't alloc a new context, the best we
1615 * can do is to leak memory (any pending putname
1616 * will be lost). The only other alternative is
1617 * to abandon auditing. */
1618 audit_zero_context(context, context->state);
1621 BUG_ON(context->in_syscall || context->name_count);
1626 context->arch = arch;
1627 context->major = major;
1628 context->argv[0] = a1;
1629 context->argv[1] = a2;
1630 context->argv[2] = a3;
1631 context->argv[3] = a4;
1633 state = context->state;
1634 context->dummy = !audit_n_rules;
1635 if (!context->dummy && (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT))
1636 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1637 if (likely(state == AUDIT_DISABLED))
1640 context->serial = 0;
1641 context->ctime = CURRENT_TIME;
1642 context->in_syscall = 1;
1643 context->auditable = !!(state == AUDIT_RECORD_CONTEXT);
1647 void audit_finish_fork(struct task_struct *child)
1649 struct audit_context *ctx = current->audit_context;
1650 struct audit_context *p = child->audit_context;
1651 if (!p || !ctx || !ctx->auditable)
1653 p->arch = ctx->arch;
1654 p->major = ctx->major;
1655 memcpy(p->argv, ctx->argv, sizeof(ctx->argv));
1656 p->ctime = ctx->ctime;
1657 p->dummy = ctx->dummy;
1658 p->auditable = ctx->auditable;
1659 p->in_syscall = ctx->in_syscall;
1660 p->filterkey = kstrdup(ctx->filterkey, GFP_KERNEL);
1661 p->ppid = current->pid;
1665 * audit_syscall_exit - deallocate audit context after a system call
1666 * @valid: success/failure flag
1667 * @return_code: syscall return value
1669 * Tear down after system call. If the audit context has been marked as
1670 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1671 * filtering, or because some other part of the kernel write an audit
1672 * message), then write out the syscall information. In call cases,
1673 * free the names stored from getname().
1675 void audit_syscall_exit(int valid, long return_code)
1677 struct task_struct *tsk = current;
1678 struct audit_context *context;
1680 context = audit_get_context(tsk, valid, return_code);
1682 if (likely(!context))
1685 if (context->in_syscall && context->auditable)
1686 audit_log_exit(context, tsk);
1688 context->in_syscall = 0;
1689 context->auditable = 0;
1691 if (context->previous) {
1692 struct audit_context *new_context = context->previous;
1693 context->previous = NULL;
1694 audit_free_context(context);
1695 tsk->audit_context = new_context;
1697 audit_free_names(context);
1698 unroll_tree_refs(context, NULL, 0);
1699 audit_free_aux(context);
1700 context->aux = NULL;
1701 context->aux_pids = NULL;
1702 context->target_pid = 0;
1703 context->target_sid = 0;
1704 context->sockaddr_len = 0;
1706 kfree(context->filterkey);
1707 context->filterkey = NULL;
1708 tsk->audit_context = context;
1712 static inline void handle_one(const struct inode *inode)
1714 #ifdef CONFIG_AUDIT_TREE
1715 struct audit_context *context;
1716 struct audit_tree_refs *p;
1717 struct audit_chunk *chunk;
1719 if (likely(list_empty(&inode->inotify_watches)))
1721 context = current->audit_context;
1723 count = context->tree_count;
1725 chunk = audit_tree_lookup(inode);
1729 if (likely(put_tree_ref(context, chunk)))
1731 if (unlikely(!grow_tree_refs(context))) {
1732 printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1733 audit_set_auditable(context);
1734 audit_put_chunk(chunk);
1735 unroll_tree_refs(context, p, count);
1738 put_tree_ref(context, chunk);
1742 static void handle_path(const struct dentry *dentry)
1744 #ifdef CONFIG_AUDIT_TREE
1745 struct audit_context *context;
1746 struct audit_tree_refs *p;
1747 const struct dentry *d, *parent;
1748 struct audit_chunk *drop;
1752 context = current->audit_context;
1754 count = context->tree_count;
1759 seq = read_seqbegin(&rename_lock);
1761 struct inode *inode = d->d_inode;
1762 if (inode && unlikely(!list_empty(&inode->inotify_watches))) {
1763 struct audit_chunk *chunk;
1764 chunk = audit_tree_lookup(inode);
1766 if (unlikely(!put_tree_ref(context, chunk))) {
1772 parent = d->d_parent;
1777 if (unlikely(read_seqretry(&rename_lock, seq) || drop)) { /* in this order */
1780 /* just a race with rename */
1781 unroll_tree_refs(context, p, count);
1784 audit_put_chunk(drop);
1785 if (grow_tree_refs(context)) {
1786 /* OK, got more space */
1787 unroll_tree_refs(context, p, count);
1792 "out of memory, audit has lost a tree reference\n");
1793 unroll_tree_refs(context, p, count);
1794 audit_set_auditable(context);
1802 * audit_getname - add a name to the list
1803 * @name: name to add
1805 * Add a name to the list of audit names for this context.
1806 * Called from fs/namei.c:getname().
1808 void __audit_getname(const char *name)
1810 struct audit_context *context = current->audit_context;
1812 if (IS_ERR(name) || !name)
1815 if (!context->in_syscall) {
1816 #if AUDIT_DEBUG == 2
1817 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1818 __FILE__, __LINE__, context->serial, name);
1823 BUG_ON(context->name_count >= AUDIT_NAMES);
1824 context->names[context->name_count].name = name;
1825 context->names[context->name_count].name_len = AUDIT_NAME_FULL;
1826 context->names[context->name_count].name_put = 1;
1827 context->names[context->name_count].ino = (unsigned long)-1;
1828 context->names[context->name_count].osid = 0;
1829 ++context->name_count;
1830 if (!context->pwd.dentry) {
1831 read_lock(¤t->fs->lock);
1832 context->pwd = current->fs->pwd;
1833 path_get(¤t->fs->pwd);
1834 read_unlock(¤t->fs->lock);
1839 /* audit_putname - intercept a putname request
1840 * @name: name to intercept and delay for putname
1842 * If we have stored the name from getname in the audit context,
1843 * then we delay the putname until syscall exit.
1844 * Called from include/linux/fs.h:putname().
1846 void audit_putname(const char *name)
1848 struct audit_context *context = current->audit_context;
1851 if (!context->in_syscall) {
1852 #if AUDIT_DEBUG == 2
1853 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1854 __FILE__, __LINE__, context->serial, name);
1855 if (context->name_count) {
1857 for (i = 0; i < context->name_count; i++)
1858 printk(KERN_ERR "name[%d] = %p = %s\n", i,
1859 context->names[i].name,
1860 context->names[i].name ?: "(null)");
1867 ++context->put_count;
1868 if (context->put_count > context->name_count) {
1869 printk(KERN_ERR "%s:%d(:%d): major=%d"
1870 " in_syscall=%d putname(%p) name_count=%d"
1873 context->serial, context->major,
1874 context->in_syscall, name, context->name_count,
1875 context->put_count);
1882 static int audit_inc_name_count(struct audit_context *context,
1883 const struct inode *inode)
1885 if (context->name_count >= AUDIT_NAMES) {
1887 printk(KERN_DEBUG "name_count maxed, losing inode data: "
1888 "dev=%02x:%02x, inode=%lu\n",
1889 MAJOR(inode->i_sb->s_dev),
1890 MINOR(inode->i_sb->s_dev),
1894 printk(KERN_DEBUG "name_count maxed, losing inode data\n");
1897 context->name_count++;
1899 context->ino_count++;
1905 static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
1907 struct cpu_vfs_cap_data caps;
1910 memset(&name->fcap.permitted, 0, sizeof(kernel_cap_t));
1911 memset(&name->fcap.inheritable, 0, sizeof(kernel_cap_t));
1918 rc = get_vfs_caps_from_disk(dentry, &caps);
1922 name->fcap.permitted = caps.permitted;
1923 name->fcap.inheritable = caps.inheritable;
1924 name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
1925 name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
1931 /* Copy inode data into an audit_names. */
1932 static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
1933 const struct inode *inode)
1935 name->ino = inode->i_ino;
1936 name->dev = inode->i_sb->s_dev;
1937 name->mode = inode->i_mode;
1938 name->uid = inode->i_uid;
1939 name->gid = inode->i_gid;
1940 name->rdev = inode->i_rdev;
1941 security_inode_getsecid(inode, &name->osid);
1942 audit_copy_fcaps(name, dentry);
1946 * audit_inode - store the inode and device from a lookup
1947 * @name: name being audited
1948 * @dentry: dentry being audited
1950 * Called from fs/namei.c:path_lookup().
1952 void __audit_inode(const char *name, const struct dentry *dentry)
1955 struct audit_context *context = current->audit_context;
1956 const struct inode *inode = dentry->d_inode;
1958 if (!context->in_syscall)
1960 if (context->name_count
1961 && context->names[context->name_count-1].name
1962 && context->names[context->name_count-1].name == name)
1963 idx = context->name_count - 1;
1964 else if (context->name_count > 1
1965 && context->names[context->name_count-2].name
1966 && context->names[context->name_count-2].name == name)
1967 idx = context->name_count - 2;
1969 /* FIXME: how much do we care about inodes that have no
1970 * associated name? */
1971 if (audit_inc_name_count(context, inode))
1973 idx = context->name_count - 1;
1974 context->names[idx].name = NULL;
1976 handle_path(dentry);
1977 audit_copy_inode(&context->names[idx], dentry, inode);
1981 * audit_inode_child - collect inode info for created/removed objects
1982 * @dname: inode's dentry name
1983 * @dentry: dentry being audited
1984 * @parent: inode of dentry parent
1986 * For syscalls that create or remove filesystem objects, audit_inode
1987 * can only collect information for the filesystem object's parent.
1988 * This call updates the audit context with the child's information.
1989 * Syscalls that create a new filesystem object must be hooked after
1990 * the object is created. Syscalls that remove a filesystem object
1991 * must be hooked prior, in order to capture the target inode during
1992 * unsuccessful attempts.
1994 void __audit_inode_child(const char *dname, const struct dentry *dentry,
1995 const struct inode *parent)
1998 struct audit_context *context = current->audit_context;
1999 const char *found_parent = NULL, *found_child = NULL;
2000 const struct inode *inode = dentry->d_inode;
2003 if (!context->in_syscall)
2008 /* determine matching parent */
2012 /* parent is more likely, look for it first */
2013 for (idx = 0; idx < context->name_count; idx++) {
2014 struct audit_names *n = &context->names[idx];
2019 if (n->ino == parent->i_ino &&
2020 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2021 n->name_len = dirlen; /* update parent data in place */
2022 found_parent = n->name;
2027 /* no matching parent, look for matching child */
2028 for (idx = 0; idx < context->name_count; idx++) {
2029 struct audit_names *n = &context->names[idx];
2034 /* strcmp() is the more likely scenario */
2035 if (!strcmp(dname, n->name) ||
2036 !audit_compare_dname_path(dname, n->name, &dirlen)) {
2038 audit_copy_inode(n, NULL, inode);
2040 n->ino = (unsigned long)-1;
2041 found_child = n->name;
2047 if (!found_parent) {
2048 if (audit_inc_name_count(context, parent))
2050 idx = context->name_count - 1;
2051 context->names[idx].name = NULL;
2052 audit_copy_inode(&context->names[idx], NULL, parent);
2056 if (audit_inc_name_count(context, inode))
2058 idx = context->name_count - 1;
2060 /* Re-use the name belonging to the slot for a matching parent
2061 * directory. All names for this context are relinquished in
2062 * audit_free_names() */
2064 context->names[idx].name = found_parent;
2065 context->names[idx].name_len = AUDIT_NAME_FULL;
2066 /* don't call __putname() */
2067 context->names[idx].name_put = 0;
2069 context->names[idx].name = NULL;
2073 audit_copy_inode(&context->names[idx], NULL, inode);
2075 context->names[idx].ino = (unsigned long)-1;
2078 EXPORT_SYMBOL_GPL(__audit_inode_child);
2081 * auditsc_get_stamp - get local copies of audit_context values
2082 * @ctx: audit_context for the task
2083 * @t: timespec to store time recorded in the audit_context
2084 * @serial: serial value that is recorded in the audit_context
2086 * Also sets the context as auditable.
2088 int auditsc_get_stamp(struct audit_context *ctx,
2089 struct timespec *t, unsigned int *serial)
2091 if (!ctx->in_syscall)
2094 ctx->serial = audit_serial();
2095 t->tv_sec = ctx->ctime.tv_sec;
2096 t->tv_nsec = ctx->ctime.tv_nsec;
2097 *serial = ctx->serial;
2102 /* global counter which is incremented every time something logs in */
2103 static atomic_t session_id = ATOMIC_INIT(0);
2106 * audit_set_loginuid - set a task's audit_context loginuid
2107 * @task: task whose audit context is being modified
2108 * @loginuid: loginuid value
2112 * Called (set) from fs/proc/base.c::proc_loginuid_write().
2114 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
2116 unsigned int sessionid = atomic_inc_return(&session_id);
2117 struct audit_context *context = task->audit_context;
2119 if (context && context->in_syscall) {
2120 struct audit_buffer *ab;
2122 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2124 audit_log_format(ab, "login pid=%d uid=%u "
2125 "old auid=%u new auid=%u"
2126 " old ses=%u new ses=%u",
2127 task->pid, task_uid(task),
2128 task->loginuid, loginuid,
2129 task->sessionid, sessionid);
2133 task->sessionid = sessionid;
2134 task->loginuid = loginuid;
2139 * __audit_mq_open - record audit data for a POSIX MQ open
2142 * @u_attr: queue attributes
2144 * Returns 0 for success or NULL context or < 0 on error.
2146 int __audit_mq_open(int oflag, mode_t mode, struct mq_attr __user *u_attr)
2148 struct audit_aux_data_mq_open *ax;
2149 struct audit_context *context = current->audit_context;
2154 if (likely(!context))
2157 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2161 if (u_attr != NULL) {
2162 if (copy_from_user(&ax->attr, u_attr, sizeof(ax->attr))) {
2167 memset(&ax->attr, 0, sizeof(ax->attr));
2172 ax->d.type = AUDIT_MQ_OPEN;
2173 ax->d.next = context->aux;
2174 context->aux = (void *)ax;
2179 * __audit_mq_timedsend - record audit data for a POSIX MQ timed send
2180 * @mqdes: MQ descriptor
2181 * @msg_len: Message length
2182 * @msg_prio: Message priority
2183 * @u_abs_timeout: Message timeout in absolute time
2185 * Returns 0 for success or NULL context or < 0 on error.
2187 int __audit_mq_timedsend(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2188 const struct timespec __user *u_abs_timeout)
2190 struct audit_aux_data_mq_sendrecv *ax;
2191 struct audit_context *context = current->audit_context;
2196 if (likely(!context))
2199 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2203 if (u_abs_timeout != NULL) {
2204 if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
2209 memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
2212 ax->msg_len = msg_len;
2213 ax->msg_prio = msg_prio;
2215 ax->d.type = AUDIT_MQ_SENDRECV;
2216 ax->d.next = context->aux;
2217 context->aux = (void *)ax;
2222 * __audit_mq_timedreceive - record audit data for a POSIX MQ timed receive
2223 * @mqdes: MQ descriptor
2224 * @msg_len: Message length
2225 * @u_msg_prio: Message priority
2226 * @u_abs_timeout: Message timeout in absolute time
2228 * Returns 0 for success or NULL context or < 0 on error.
2230 int __audit_mq_timedreceive(mqd_t mqdes, size_t msg_len,
2231 unsigned int __user *u_msg_prio,
2232 const struct timespec __user *u_abs_timeout)
2234 struct audit_aux_data_mq_sendrecv *ax;
2235 struct audit_context *context = current->audit_context;
2240 if (likely(!context))
2243 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2247 if (u_msg_prio != NULL) {
2248 if (get_user(ax->msg_prio, u_msg_prio)) {
2255 if (u_abs_timeout != NULL) {
2256 if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
2261 memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
2264 ax->msg_len = msg_len;
2266 ax->d.type = AUDIT_MQ_SENDRECV;
2267 ax->d.next = context->aux;
2268 context->aux = (void *)ax;
2273 * __audit_mq_notify - record audit data for a POSIX MQ notify
2274 * @mqdes: MQ descriptor
2275 * @u_notification: Notification event
2277 * Returns 0 for success or NULL context or < 0 on error.
2280 int __audit_mq_notify(mqd_t mqdes, const struct sigevent __user *u_notification)
2282 struct audit_aux_data_mq_notify *ax;
2283 struct audit_context *context = current->audit_context;
2288 if (likely(!context))
2291 ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
2295 if (u_notification != NULL) {
2296 if (copy_from_user(&ax->notification, u_notification, sizeof(ax->notification))) {
2301 memset(&ax->notification, 0, sizeof(ax->notification));
2305 ax->d.type = AUDIT_MQ_NOTIFY;
2306 ax->d.next = context->aux;
2307 context->aux = (void *)ax;
2312 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2313 * @mqdes: MQ descriptor
2317 void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2319 struct audit_context *context = current->audit_context;
2320 context->mq_getsetattr.mqdes = mqdes;
2321 context->mq_getsetattr.mqstat = *mqstat;
2322 context->type = AUDIT_MQ_GETSETATTR;
2326 * audit_ipc_obj - record audit data for ipc object
2327 * @ipcp: ipc permissions
2330 void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2332 struct audit_context *context = current->audit_context;
2333 context->ipc.uid = ipcp->uid;
2334 context->ipc.gid = ipcp->gid;
2335 context->ipc.mode = ipcp->mode;
2336 context->ipc.has_perm = 0;
2337 security_ipc_getsecid(ipcp, &context->ipc.osid);
2338 context->type = AUDIT_IPC;
2342 * audit_ipc_set_perm - record audit data for new ipc permissions
2343 * @qbytes: msgq bytes
2344 * @uid: msgq user id
2345 * @gid: msgq group id
2346 * @mode: msgq mode (permissions)
2348 * Called only after audit_ipc_obj().
2350 void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
2352 struct audit_context *context = current->audit_context;
2354 context->ipc.qbytes = qbytes;
2355 context->ipc.perm_uid = uid;
2356 context->ipc.perm_gid = gid;
2357 context->ipc.perm_mode = mode;
2358 context->ipc.has_perm = 1;
2361 int audit_bprm(struct linux_binprm *bprm)
2363 struct audit_aux_data_execve *ax;
2364 struct audit_context *context = current->audit_context;
2366 if (likely(!audit_enabled || !context || context->dummy))
2369 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2373 ax->argc = bprm->argc;
2374 ax->envc = bprm->envc;
2376 ax->d.type = AUDIT_EXECVE;
2377 ax->d.next = context->aux;
2378 context->aux = (void *)ax;
2384 * audit_socketcall - record audit data for sys_socketcall
2385 * @nargs: number of args
2389 void audit_socketcall(int nargs, unsigned long *args)
2391 struct audit_context *context = current->audit_context;
2393 if (likely(!context || context->dummy))
2396 context->type = AUDIT_SOCKETCALL;
2397 context->socketcall.nargs = nargs;
2398 memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2402 * __audit_fd_pair - record audit data for pipe and socketpair
2403 * @fd1: the first file descriptor
2404 * @fd2: the second file descriptor
2406 * Returns 0 for success or NULL context or < 0 on error.
2408 int __audit_fd_pair(int fd1, int fd2)
2410 struct audit_context *context = current->audit_context;
2411 struct audit_aux_data_fd_pair *ax;
2413 if (likely(!context)) {
2417 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2425 ax->d.type = AUDIT_FD_PAIR;
2426 ax->d.next = context->aux;
2427 context->aux = (void *)ax;
2432 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2433 * @len: data length in user space
2434 * @a: data address in kernel space
2436 * Returns 0 for success or NULL context or < 0 on error.
2438 int audit_sockaddr(int len, void *a)
2440 struct audit_context *context = current->audit_context;
2442 if (likely(!context || context->dummy))
2445 if (!context->sockaddr) {
2446 void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2449 context->sockaddr = p;
2452 context->sockaddr_len = len;
2453 memcpy(context->sockaddr, a, len);
2457 void __audit_ptrace(struct task_struct *t)
2459 struct audit_context *context = current->audit_context;
2461 context->target_pid = t->pid;
2462 context->target_auid = audit_get_loginuid(t);
2463 context->target_uid = task_uid(t);
2464 context->target_sessionid = audit_get_sessionid(t);
2465 security_task_getsecid(t, &context->target_sid);
2466 memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2470 * audit_signal_info - record signal info for shutting down audit subsystem
2471 * @sig: signal value
2472 * @t: task being signaled
2474 * If the audit subsystem is being terminated, record the task (pid)
2475 * and uid that is doing that.
2477 int __audit_signal_info(int sig, struct task_struct *t)
2479 struct audit_aux_data_pids *axp;
2480 struct task_struct *tsk = current;
2481 struct audit_context *ctx = tsk->audit_context;
2482 uid_t uid = current_uid(), t_uid = task_uid(t);
2484 if (audit_pid && t->tgid == audit_pid) {
2485 if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2486 audit_sig_pid = tsk->pid;
2487 if (tsk->loginuid != -1)
2488 audit_sig_uid = tsk->loginuid;
2490 audit_sig_uid = uid;
2491 security_task_getsecid(tsk, &audit_sig_sid);
2493 if (!audit_signals || audit_dummy_context())
2497 /* optimize the common case by putting first signal recipient directly
2498 * in audit_context */
2499 if (!ctx->target_pid) {
2500 ctx->target_pid = t->tgid;
2501 ctx->target_auid = audit_get_loginuid(t);
2502 ctx->target_uid = t_uid;
2503 ctx->target_sessionid = audit_get_sessionid(t);
2504 security_task_getsecid(t, &ctx->target_sid);
2505 memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2509 axp = (void *)ctx->aux_pids;
2510 if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2511 axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2515 axp->d.type = AUDIT_OBJ_PID;
2516 axp->d.next = ctx->aux_pids;
2517 ctx->aux_pids = (void *)axp;
2519 BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2521 axp->target_pid[axp->pid_count] = t->tgid;
2522 axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2523 axp->target_uid[axp->pid_count] = t_uid;
2524 axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2525 security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2526 memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2533 * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2534 * @bprm: pointer to the bprm being processed
2535 * @new: the proposed new credentials
2536 * @old: the old credentials
2538 * Simply check if the proc already has the caps given by the file and if not
2539 * store the priv escalation info for later auditing at the end of the syscall
2543 int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2544 const struct cred *new, const struct cred *old)
2546 struct audit_aux_data_bprm_fcaps *ax;
2547 struct audit_context *context = current->audit_context;
2548 struct cpu_vfs_cap_data vcaps;
2549 struct dentry *dentry;
2551 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2555 ax->d.type = AUDIT_BPRM_FCAPS;
2556 ax->d.next = context->aux;
2557 context->aux = (void *)ax;
2559 dentry = dget(bprm->file->f_dentry);
2560 get_vfs_caps_from_disk(dentry, &vcaps);
2563 ax->fcap.permitted = vcaps.permitted;
2564 ax->fcap.inheritable = vcaps.inheritable;
2565 ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2566 ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2568 ax->old_pcap.permitted = old->cap_permitted;
2569 ax->old_pcap.inheritable = old->cap_inheritable;
2570 ax->old_pcap.effective = old->cap_effective;
2572 ax->new_pcap.permitted = new->cap_permitted;
2573 ax->new_pcap.inheritable = new->cap_inheritable;
2574 ax->new_pcap.effective = new->cap_effective;
2579 * __audit_log_capset - store information about the arguments to the capset syscall
2580 * @pid: target pid of the capset call
2581 * @new: the new credentials
2582 * @old: the old (current) credentials
2584 * Record the aguments userspace sent to sys_capset for later printing by the
2585 * audit system if applicable
2587 int __audit_log_capset(pid_t pid,
2588 const struct cred *new, const struct cred *old)
2590 struct audit_aux_data_capset *ax;
2591 struct audit_context *context = current->audit_context;
2593 if (likely(!audit_enabled || !context || context->dummy))
2596 ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2600 ax->d.type = AUDIT_CAPSET;
2601 ax->d.next = context->aux;
2602 context->aux = (void *)ax;
2605 ax->cap.effective = new->cap_effective;
2606 ax->cap.inheritable = new->cap_effective;
2607 ax->cap.permitted = new->cap_permitted;
2613 * audit_core_dumps - record information about processes that end abnormally
2614 * @signr: signal value
2616 * If a process ends with a core dump, something fishy is going on and we
2617 * should record the event for investigation.
2619 void audit_core_dumps(long signr)
2621 struct audit_buffer *ab;
2623 uid_t auid = audit_get_loginuid(current), uid;
2625 unsigned int sessionid = audit_get_sessionid(current);
2630 if (signr == SIGQUIT) /* don't care for those */
2633 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2634 current_uid_gid(&uid, &gid);
2635 audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2636 auid, uid, gid, sessionid);
2637 security_task_getsecid(current, &sid);
2642 if (security_secid_to_secctx(sid, &ctx, &len))
2643 audit_log_format(ab, " ssid=%u", sid);
2645 audit_log_format(ab, " subj=%s", ctx);
2646 security_release_secctx(ctx, len);
2649 audit_log_format(ab, " pid=%d comm=", current->pid);
2650 audit_log_untrustedstring(ab, current->comm);
2651 audit_log_format(ab, " sig=%ld", signr);