]> nv-tegra.nvidia Code Review - linux-3.10.git/blobdiff - kernel/auditfilter.c
reboot: checkpatch.pl the new kernel/reboot.c file
[linux-3.10.git] / kernel / auditfilter.c
index 03a6919103d4df8368cda70a3a151b593f77548d..38c475cc55ab3e477152bd1876334bd033322b75 100644 (file)
@@ -27,8 +27,8 @@
 #include <linux/namei.h>
 #include <linux/netlink.h>
 #include <linux/sched.h>
-#include <linux/inotify.h>
-#include <linux/selinux.h>
+#include <linux/slab.h>
+#include <linux/security.h>
 #include "audit.h"
 
 /*
  *             Synchronizes writes and blocking reads of audit's filterlist
  *             data.  Rcu is used to traverse the filterlist and access
  *             contents of structs audit_entry, audit_watch and opaque
- *             selinux rules during filtering.  If modified, these structures
+ *             LSM rules during filtering.  If modified, these structures
  *             must be copied and replace their counterparts in the filterlist.
  *             An audit_parent struct is not accessed during filtering, so may
  *             be written directly provided audit_filter_mutex is held.
  */
 
-/*
- * Reference counting:
- *
- * audit_parent: lifetime is from audit_init_parent() to receipt of an IN_IGNORED
- *     event.  Each audit_watch holds a reference to its associated parent.
- *
- * audit_watch: if added to lists, lifetime is from audit_init_watch() to
- *     audit_remove_watch().  Additionally, an audit_watch may exist
- *     temporarily to assist in searching existing filter data.  Each
- *     audit_krule holds a reference to its associated watch.
- */
-
-struct audit_parent {
-       struct list_head        ilist;  /* entry in inotify registration list */
-       struct list_head        watches; /* associated watches */
-       struct inotify_watch    wdata;  /* inotify watch data */
-       unsigned                flags;  /* status flags */
-};
-
-/*
- * audit_parent status flags:
- *
- * AUDIT_PARENT_INVALID - set anytime rules/watches are auto-removed due to
- * a filesystem event to ensure we're adding audit watches to a valid parent.
- * Technically not needed for IN_DELETE_SELF or IN_UNMOUNT events, as we cannot
- * receive them while we have nameidata, but must be used for IN_MOVE_SELF which
- * we can receive while holding nameidata.
- */
-#define AUDIT_PARENT_INVALID   0x001
-
 /* Audit filter lists, defined in <linux/audit.h> */
 struct list_head audit_filter_list[AUDIT_NR_FILTERS] = {
        LIST_HEAD_INIT(audit_filter_list[0]),
@@ -86,114 +56,42 @@ struct list_head audit_filter_list[AUDIT_NR_FILTERS] = {
 #error Fix audit_filter_list initialiser
 #endif
 };
+static struct list_head audit_rules_list[AUDIT_NR_FILTERS] = {
+       LIST_HEAD_INIT(audit_rules_list[0]),
+       LIST_HEAD_INIT(audit_rules_list[1]),
+       LIST_HEAD_INIT(audit_rules_list[2]),
+       LIST_HEAD_INIT(audit_rules_list[3]),
+       LIST_HEAD_INIT(audit_rules_list[4]),
+       LIST_HEAD_INIT(audit_rules_list[5]),
+};
 
-static DEFINE_MUTEX(audit_filter_mutex);
-
-/* Inotify handle */
-extern struct inotify_handle *audit_ih;
-
-/* Inotify events we care about. */
-#define AUDIT_IN_WATCH IN_MOVE|IN_CREATE|IN_DELETE|IN_DELETE_SELF|IN_MOVE_SELF
-
-void audit_free_parent(struct inotify_watch *i_watch)
-{
-       struct audit_parent *parent;
-
-       parent = container_of(i_watch, struct audit_parent, wdata);
-       WARN_ON(!list_empty(&parent->watches));
-       kfree(parent);
-}
-
-static inline void audit_get_watch(struct audit_watch *watch)
-{
-       atomic_inc(&watch->count);
-}
-
-static void audit_put_watch(struct audit_watch *watch)
-{
-       if (atomic_dec_and_test(&watch->count)) {
-               WARN_ON(watch->parent);
-               WARN_ON(!list_empty(&watch->rules));
-               kfree(watch->path);
-               kfree(watch);
-       }
-}
-
-static void audit_remove_watch(struct audit_watch *watch)
-{
-       list_del(&watch->wlist);
-       put_inotify_watch(&watch->parent->wdata);
-       watch->parent = NULL;
-       audit_put_watch(watch); /* match initial get */
-}
+DEFINE_MUTEX(audit_filter_mutex);
 
 static inline void audit_free_rule(struct audit_entry *e)
 {
        int i;
+       struct audit_krule *erule = &e->rule;
 
        /* some rules don't have associated watches */
-       if (e->rule.watch)
-               audit_put_watch(e->rule.watch);
-       if (e->rule.fields)
-               for (i = 0; i < e->rule.field_count; i++) {
-                       struct audit_field *f = &e->rule.fields[i];
-                       kfree(f->se_str);
-                       selinux_audit_rule_free(f->se_rule);
+       if (erule->watch)
+               audit_put_watch(erule->watch);
+       if (erule->fields)
+               for (i = 0; i < erule->field_count; i++) {
+                       struct audit_field *f = &erule->fields[i];
+                       kfree(f->lsm_str);
+                       security_audit_rule_free(f->lsm_rule);
                }
-       kfree(e->rule.fields);
+       kfree(erule->fields);
+       kfree(erule->filterkey);
        kfree(e);
 }
 
-static inline void audit_free_rule_rcu(struct rcu_head *head)
+void audit_free_rule_rcu(struct rcu_head *head)
 {
        struct audit_entry *e = container_of(head, struct audit_entry, rcu);
        audit_free_rule(e);
 }
 
-/* Initialize a parent watch entry. */
-static struct audit_parent *audit_init_parent(struct nameidata *ndp)
-{
-       struct audit_parent *parent;
-       s32 wd;
-
-       parent = kzalloc(sizeof(*parent), GFP_KERNEL);
-       if (unlikely(!parent))
-               return ERR_PTR(-ENOMEM);
-
-       INIT_LIST_HEAD(&parent->watches);
-       parent->flags = 0;
-
-       inotify_init_watch(&parent->wdata);
-       /* grab a ref so inotify watch hangs around until we take audit_filter_mutex */
-       get_inotify_watch(&parent->wdata);
-       wd = inotify_add_watch(audit_ih, &parent->wdata, ndp->dentry->d_inode,
-                              AUDIT_IN_WATCH);
-       if (wd < 0) {
-               audit_free_parent(&parent->wdata);
-               return ERR_PTR(wd);
-       }
-
-       return parent;
-}
-
-/* Initialize a watch entry. */
-static struct audit_watch *audit_init_watch(char *path)
-{
-       struct audit_watch *watch;
-
-       watch = kzalloc(sizeof(*watch), GFP_KERNEL);
-       if (unlikely(!watch))
-               return ERR_PTR(-ENOMEM);
-
-       INIT_LIST_HEAD(&watch->rules);
-       atomic_set(&watch->count, 1);
-       watch->path = path;
-       watch->dev = (dev_t)-1;
-       watch->ino = (unsigned long)-1;
-
-       return watch;
-}
-
 /* Initialize an audit filterlist entry. */
 static inline struct audit_entry *audit_init_entry(u32 field_count)
 {
@@ -216,7 +114,7 @@ static inline struct audit_entry *audit_init_entry(u32 field_count)
 
 /* Unpack a filter field's string representation from user-space
  * buffer. */
-static char *audit_unpack_string(void **bufp, size_t *remain, size_t len)
+char *audit_unpack_string(void **bufp, size_t *remain, size_t len)
 {
        char *str;
 
@@ -246,37 +144,84 @@ static inline int audit_to_inode(struct audit_krule *krule,
                                 struct audit_field *f)
 {
        if (krule->listnr != AUDIT_FILTER_EXIT ||
-           krule->watch || krule->inode_f)
+           krule->watch || krule->inode_f || krule->tree ||
+           (f->op != Audit_equal && f->op != Audit_not_equal))
                return -EINVAL;
 
        krule->inode_f = f;
        return 0;
 }
 
-/* Translate a watch string to kernel respresentation. */
-static int audit_to_watch(struct audit_krule *krule, char *path, int len,
-                         u32 op)
+static __u32 *classes[AUDIT_SYSCALL_CLASSES];
+
+int __init audit_register_class(int class, unsigned *list)
 {
-       struct audit_watch *watch;
+       __u32 *p = kzalloc(AUDIT_BITMASK_SIZE * sizeof(__u32), GFP_KERNEL);
+       if (!p)
+               return -ENOMEM;
+       while (*list != ~0U) {
+               unsigned n = *list++;
+               if (n >= AUDIT_BITMASK_SIZE * 32 - AUDIT_SYSCALL_CLASSES) {
+                       kfree(p);
+                       return -EINVAL;
+               }
+               p[AUDIT_WORD(n)] |= AUDIT_BIT(n);
+       }
+       if (class >= AUDIT_SYSCALL_CLASSES || classes[class]) {
+               kfree(p);
+               return -EINVAL;
+       }
+       classes[class] = p;
+       return 0;
+}
 
-       if (!audit_ih)
-               return -EOPNOTSUPP;
+int audit_match_class(int class, unsigned syscall)
+{
+       if (unlikely(syscall >= AUDIT_BITMASK_SIZE * 32))
+               return 0;
+       if (unlikely(class >= AUDIT_SYSCALL_CLASSES || !classes[class]))
+               return 0;
+       return classes[class][AUDIT_WORD(syscall)] & AUDIT_BIT(syscall);
+}
 
-       if (path[0] != '/' || path[len-1] == '/' ||
-           krule->listnr != AUDIT_FILTER_EXIT ||
-           op & ~AUDIT_EQUAL ||
-           krule->inode_f || krule->watch) /* 1 inode # per rule, for hash */
-               return -EINVAL;
+#ifdef CONFIG_AUDITSYSCALL
+static inline int audit_match_class_bits(int class, u32 *mask)
+{
+       int i;
 
-       watch = audit_init_watch(path);
-       if (unlikely(IS_ERR(watch)))
-               return PTR_ERR(watch);
+       if (classes[class]) {
+               for (i = 0; i < AUDIT_BITMASK_SIZE; i++)
+                       if (mask[i] & classes[class][i])
+                               return 0;
+       }
+       return 1;
+}
 
-       audit_get_watch(watch);
-       krule->watch = watch;
+static int audit_match_signal(struct audit_entry *entry)
+{
+       struct audit_field *arch = entry->rule.arch_f;
+
+       if (!arch) {
+               /* When arch is unspecified, we must check both masks on biarch
+                * as syscall number alone is ambiguous. */
+               return (audit_match_class_bits(AUDIT_CLASS_SIGNAL,
+                                              entry->rule.mask) &&
+                       audit_match_class_bits(AUDIT_CLASS_SIGNAL_32,
+                                              entry->rule.mask));
+       }
 
-       return 0;
+       switch(audit_classify_arch(arch->val)) {
+       case 0: /* native */
+               return (audit_match_class_bits(AUDIT_CLASS_SIGNAL,
+                                              entry->rule.mask));
+       case 1: /* 32bit on biarch */
+               return (audit_match_class_bits(AUDIT_CLASS_SIGNAL_32,
+                                              entry->rule.mask));
+       default:
+               return 1;
+       }
 }
+#endif
 
 /* Common user-space to kernel rule translation. */
 static inline struct audit_entry *audit_to_entry_common(struct audit_rule *rule)
@@ -290,13 +235,15 @@ static inline struct audit_entry *audit_to_entry_common(struct audit_rule *rule)
        switch(listnr) {
        default:
                goto exit_err;
-       case AUDIT_FILTER_USER:
-       case AUDIT_FILTER_TYPE:
 #ifdef CONFIG_AUDITSYSCALL
        case AUDIT_FILTER_ENTRY:
+               if (rule->action == AUDIT_ALWAYS)
+                       goto exit_err;
        case AUDIT_FILTER_EXIT:
        case AUDIT_FILTER_TASK:
 #endif
+       case AUDIT_FILTER_USER:
+       case AUDIT_FILTER_TYPE:
                ;
        }
        if (unlikely(rule->action == AUDIT_POSSIBLE)) {
@@ -321,83 +268,125 @@ static inline struct audit_entry *audit_to_entry_common(struct audit_rule *rule)
        for (i = 0; i < AUDIT_BITMASK_SIZE; i++)
                entry->rule.mask[i] = rule->mask[i];
 
+       for (i = 0; i < AUDIT_SYSCALL_CLASSES; i++) {
+               int bit = AUDIT_BITMASK_SIZE * 32 - i - 1;
+               __u32 *p = &entry->rule.mask[AUDIT_WORD(bit)];
+               __u32 *class;
+
+               if (!(*p & AUDIT_BIT(bit)))
+                       continue;
+               *p &= ~AUDIT_BIT(bit);
+               class = classes[i];
+               if (class) {
+                       int j;
+                       for (j = 0; j < AUDIT_BITMASK_SIZE; j++)
+                               entry->rule.mask[j] |= class[j];
+               }
+       }
+
        return entry;
 
 exit_err:
        return ERR_PTR(err);
 }
 
-/* Translate struct audit_rule to kernel's rule respresentation.
- * Exists for backward compatibility with userspace. */
-static struct audit_entry *audit_rule_to_entry(struct audit_rule *rule)
+static u32 audit_ops[] =
 {
-       struct audit_entry *entry;
-       struct audit_field *f;
-       int err = 0;
-       int i;
-
-       entry = audit_to_entry_common(rule);
-       if (IS_ERR(entry))
-               goto exit_nofree;
-
-       for (i = 0; i < rule->field_count; i++) {
-               struct audit_field *f = &entry->rule.fields[i];
-
-               f->op = rule->fields[i] & (AUDIT_NEGATE|AUDIT_OPERATORS);
-               f->type = rule->fields[i] & ~(AUDIT_NEGATE|AUDIT_OPERATORS);
-               f->val = rule->values[i];
-
-               err = -EINVAL;
-               if (f->type & AUDIT_UNUSED_BITS)
-                       goto exit_free;
-
-               switch(f->type) {
-               case AUDIT_SE_USER:
-               case AUDIT_SE_ROLE:
-               case AUDIT_SE_TYPE:
-               case AUDIT_SE_SEN:
-               case AUDIT_SE_CLR:
-               case AUDIT_WATCH:
-                       goto exit_free;
-               case AUDIT_INODE:
-                       err = audit_to_inode(&entry->rule, f);
-                       if (err)
-                               goto exit_free;
-                       break;
-               }
-
-               entry->rule.vers_ops = (f->op & AUDIT_OPERATORS) ? 2 : 1;
-
-               /* Support for legacy operators where
-                * AUDIT_NEGATE bit signifies != and otherwise assumes == */
-               if (f->op & AUDIT_NEGATE)
-                       f->op = AUDIT_NOT_EQUAL;
-               else if (!f->op)
-                       f->op = AUDIT_EQUAL;
-               else if (f->op == AUDIT_OPERATORS) {
-                       err = -EINVAL;
-                       goto exit_free;
-               }
-       }
+       [Audit_equal] = AUDIT_EQUAL,
+       [Audit_not_equal] = AUDIT_NOT_EQUAL,
+       [Audit_bitmask] = AUDIT_BIT_MASK,
+       [Audit_bittest] = AUDIT_BIT_TEST,
+       [Audit_lt] = AUDIT_LESS_THAN,
+       [Audit_gt] = AUDIT_GREATER_THAN,
+       [Audit_le] = AUDIT_LESS_THAN_OR_EQUAL,
+       [Audit_ge] = AUDIT_GREATER_THAN_OR_EQUAL,
+};
 
-       f = entry->rule.inode_f;
-       if (f) {
-               switch(f->op) {
-               case AUDIT_NOT_EQUAL:
-                       entry->rule.inode_f = NULL;
-               case AUDIT_EQUAL:
-                       break;
-               default:
-                       goto exit_free;
-               }
-       }
+static u32 audit_to_op(u32 op)
+{
+       u32 n;
+       for (n = Audit_equal; n < Audit_bad && audit_ops[n] != op; n++)
+               ;
+       return n;
+}
 
-exit_nofree:
-       return entry;
+/* check if an audit field is valid */
+static int audit_field_valid(struct audit_entry *entry, struct audit_field *f)
+{
+       switch(f->type) {
+       case AUDIT_MSGTYPE:
+               if (entry->rule.listnr != AUDIT_FILTER_TYPE &&
+                   entry->rule.listnr != AUDIT_FILTER_USER)
+                       return -EINVAL;
+               break;
+       };
 
-exit_free:
-       audit_free_rule(entry);
-       return ERR_PTR(err);
+       switch(f->type) {
+       default:
+               return -EINVAL;
+       case AUDIT_UID:
+       case AUDIT_EUID:
+       case AUDIT_SUID:
+       case AUDIT_FSUID:
+       case AUDIT_LOGINUID:
+       case AUDIT_OBJ_UID:
+       case AUDIT_GID:
+       case AUDIT_EGID:
+       case AUDIT_SGID:
+       case AUDIT_FSGID:
+       case AUDIT_OBJ_GID:
+       case AUDIT_PID:
+       case AUDIT_PERS:
+       case AUDIT_MSGTYPE:
+       case AUDIT_PPID:
+       case AUDIT_DEVMAJOR:
+       case AUDIT_DEVMINOR:
+       case AUDIT_EXIT:
+       case AUDIT_SUCCESS:
+               /* bit ops are only useful on syscall args */
+               if (f->op == Audit_bitmask || f->op == Audit_bittest)
+                       return -EINVAL;
+               break;
+       case AUDIT_ARG0:
+       case AUDIT_ARG1:
+       case AUDIT_ARG2:
+       case AUDIT_ARG3:
+       case AUDIT_SUBJ_USER:
+       case AUDIT_SUBJ_ROLE:
+       case AUDIT_SUBJ_TYPE:
+       case AUDIT_SUBJ_SEN:
+       case AUDIT_SUBJ_CLR:
+       case AUDIT_OBJ_USER:
+       case AUDIT_OBJ_ROLE:
+       case AUDIT_OBJ_TYPE:
+       case AUDIT_OBJ_LEV_LOW:
+       case AUDIT_OBJ_LEV_HIGH:
+       case AUDIT_WATCH:
+       case AUDIT_DIR:
+       case AUDIT_FILTERKEY:
+               break;
+       case AUDIT_LOGINUID_SET:
+               if ((f->val != 0) && (f->val != 1))
+                       return -EINVAL;
+       /* FALL THROUGH */
+       case AUDIT_ARCH:
+               if (f->op != Audit_not_equal && f->op != Audit_equal)
+                       return -EINVAL;
+               break;
+       case AUDIT_PERM:
+               if (f->val & ~15)
+                       return -EINVAL;
+               break;
+       case AUDIT_FILETYPE:
+               if (f->val & ~S_IFMT)
+                       return -EINVAL;
+               break;
+       case AUDIT_FIELD_COMPARE:
+               if (f->val > AUDIT_MAX_FIELD_COMPARE)
+                       return -EINVAL;
+               break;
+       };
+       return 0;
 }
 
 /* Translate struct audit_rule_data to kernel's rule respresentation. */
@@ -406,7 +395,6 @@ static struct audit_entry *audit_data_to_entry(struct audit_rule_data *data,
 {
        int err = 0;
        struct audit_entry *entry;
-       struct audit_field *f;
        void *bufp;
        size_t remain = datasz - sizeof(struct audit_rule_data);
        int i;
@@ -422,32 +410,73 @@ static struct audit_entry *audit_data_to_entry(struct audit_rule_data *data,
                struct audit_field *f = &entry->rule.fields[i];
 
                err = -EINVAL;
-               if (!(data->fieldflags[i] & AUDIT_OPERATORS) ||
-                   data->fieldflags[i] & ~AUDIT_OPERATORS)
+
+               f->op = audit_to_op(data->fieldflags[i]);
+               if (f->op == Audit_bad)
                        goto exit_free;
 
-               f->op = data->fieldflags[i] & AUDIT_OPERATORS;
                f->type = data->fields[i];
                f->val = data->values[i];
-               f->se_str = NULL;
-               f->se_rule = NULL;
-               switch(f->type) {
-               case AUDIT_SE_USER:
-               case AUDIT_SE_ROLE:
-               case AUDIT_SE_TYPE:
-               case AUDIT_SE_SEN:
-               case AUDIT_SE_CLR:
+               f->uid = INVALID_UID;
+               f->gid = INVALID_GID;
+               f->lsm_str = NULL;
+               f->lsm_rule = NULL;
+
+               /* Support legacy tests for a valid loginuid */
+               if ((f->type == AUDIT_LOGINUID) && (f->val == AUDIT_UID_UNSET)) {
+                       f->type = AUDIT_LOGINUID_SET;
+                       f->val = 0;
+               }
+
+               err = audit_field_valid(entry, f);
+               if (err)
+                       goto exit_free;
+
+               err = -EINVAL;
+               switch (f->type) {
+               case AUDIT_LOGINUID:
+               case AUDIT_UID:
+               case AUDIT_EUID:
+               case AUDIT_SUID:
+               case AUDIT_FSUID:
+               case AUDIT_OBJ_UID:
+                       f->uid = make_kuid(current_user_ns(), f->val);
+                       if (!uid_valid(f->uid))
+                               goto exit_free;
+                       break;
+               case AUDIT_GID:
+               case AUDIT_EGID:
+               case AUDIT_SGID:
+               case AUDIT_FSGID:
+               case AUDIT_OBJ_GID:
+                       f->gid = make_kgid(current_user_ns(), f->val);
+                       if (!gid_valid(f->gid))
+                               goto exit_free;
+                       break;
+               case AUDIT_ARCH:
+                       entry->rule.arch_f = f;
+                       break;
+               case AUDIT_SUBJ_USER:
+               case AUDIT_SUBJ_ROLE:
+               case AUDIT_SUBJ_TYPE:
+               case AUDIT_SUBJ_SEN:
+               case AUDIT_SUBJ_CLR:
+               case AUDIT_OBJ_USER:
+               case AUDIT_OBJ_ROLE:
+               case AUDIT_OBJ_TYPE:
+               case AUDIT_OBJ_LEV_LOW:
+               case AUDIT_OBJ_LEV_HIGH:
                        str = audit_unpack_string(&bufp, &remain, f->val);
                        if (IS_ERR(str))
                                goto exit_free;
                        entry->rule.buflen += f->val;
 
-                       err = selinux_audit_rule_init(f->type, f->op, str,
-                                                     &f->se_rule);
+                       err = security_audit_rule_init(f->type, f->op, str,
+                                                      (void **)&f->lsm_rule);
                        /* Keep currently invalid fields around in case they
                         * become valid after a policy reload. */
                        if (err == -EINVAL) {
-                               printk(KERN_WARNING "audit rule for selinux "
+                               printk(KERN_WARNING "audit rule for LSM "
                                       "\'%s\' is invalid\n",  str);
                                err = 0;
                        }
@@ -455,7 +484,7 @@ static struct audit_entry *audit_data_to_entry(struct audit_rule_data *data,
                                kfree(str);
                                goto exit_free;
                        } else
-                               f->se_str = str;
+                               f->lsm_str = str;
                        break;
                case AUDIT_WATCH:
                        str = audit_unpack_string(&bufp, &remain, f->val);
@@ -469,36 +498,51 @@ static struct audit_entry *audit_data_to_entry(struct audit_rule_data *data,
                                goto exit_free;
                        }
                        break;
+               case AUDIT_DIR:
+                       str = audit_unpack_string(&bufp, &remain, f->val);
+                       if (IS_ERR(str))
+                               goto exit_free;
+                       entry->rule.buflen += f->val;
+
+                       err = audit_make_tree(&entry->rule, str, f->op);
+                       kfree(str);
+                       if (err)
+                               goto exit_free;
+                       break;
                case AUDIT_INODE:
                        err = audit_to_inode(&entry->rule, f);
                        if (err)
                                goto exit_free;
                        break;
-               }
-       }
-
-       f = entry->rule.inode_f;
-       if (f) {
-               switch(f->op) {
-               case AUDIT_NOT_EQUAL:
-                       entry->rule.inode_f = NULL;
-               case AUDIT_EQUAL:
+               case AUDIT_FILTERKEY:
+                       if (entry->rule.filterkey || f->val > AUDIT_MAX_KEY_LEN)
+                               goto exit_free;
+                       str = audit_unpack_string(&bufp, &remain, f->val);
+                       if (IS_ERR(str))
+                               goto exit_free;
+                       entry->rule.buflen += f->val;
+                       entry->rule.filterkey = str;
                        break;
-               default:
-                       goto exit_free;
                }
        }
 
+       if (entry->rule.inode_f && entry->rule.inode_f->op == Audit_not_equal)
+               entry->rule.inode_f = NULL;
+
 exit_nofree:
        return entry;
 
 exit_free:
+       if (entry->rule.watch)
+               audit_put_watch(entry->rule.watch); /* matches initial get */
+       if (entry->rule.tree)
+               audit_put_tree(entry->rule.tree); /* that's the temporary one */
        audit_free_rule(entry);
        return ERR_PTR(err);
 }
 
 /* Pack a filter field's string representation into data block. */
-static inline size_t audit_pack_string(void **bufp, char *str)
+static inline size_t audit_pack_string(void **bufp, const char *str)
 {
        size_t len = strlen(str);
 
@@ -508,37 +552,6 @@ static inline size_t audit_pack_string(void **bufp, char *str)
        return len;
 }
 
-/* Translate kernel rule respresentation to struct audit_rule.
- * Exists for backward compatibility with userspace. */
-static struct audit_rule *audit_krule_to_rule(struct audit_krule *krule)
-{
-       struct audit_rule *rule;
-       int i;
-
-       rule = kmalloc(sizeof(*rule), GFP_KERNEL);
-       if (unlikely(!rule))
-               return NULL;
-       memset(rule, 0, sizeof(*rule));
-
-       rule->flags = krule->flags | krule->listnr;
-       rule->action = krule->action;
-       rule->field_count = krule->field_count;
-       for (i = 0; i < rule->field_count; i++) {
-               rule->values[i] = krule->fields[i].val;
-               rule->fields[i] = krule->fields[i].type;
-
-               if (krule->vers_ops == 1) {
-                       if (krule->fields[i].op & AUDIT_NOT_EQUAL)
-                               rule->fields[i] |= AUDIT_NEGATE;
-               } else {
-                       rule->fields[i] |= krule->fields[i].op;
-               }
-       }
-       for (i = 0; i < AUDIT_BITMASK_SIZE; i++) rule->mask[i] = krule->mask[i];
-
-       return rule;
-}
-
 /* Translate kernel rule respresentation to struct audit_rule_data. */
 static struct audit_rule_data *audit_krule_to_data(struct audit_krule *krule)
 {
@@ -559,19 +572,34 @@ static struct audit_rule_data *audit_krule_to_data(struct audit_krule *krule)
                struct audit_field *f = &krule->fields[i];
 
                data->fields[i] = f->type;
-               data->fieldflags[i] = f->op;
+               data->fieldflags[i] = audit_ops[f->op];
                switch(f->type) {
-               case AUDIT_SE_USER:
-               case AUDIT_SE_ROLE:
-               case AUDIT_SE_TYPE:
-               case AUDIT_SE_SEN:
-               case AUDIT_SE_CLR:
+               case AUDIT_SUBJ_USER:
+               case AUDIT_SUBJ_ROLE:
+               case AUDIT_SUBJ_TYPE:
+               case AUDIT_SUBJ_SEN:
+               case AUDIT_SUBJ_CLR:
+               case AUDIT_OBJ_USER:
+               case AUDIT_OBJ_ROLE:
+               case AUDIT_OBJ_TYPE:
+               case AUDIT_OBJ_LEV_LOW:
+               case AUDIT_OBJ_LEV_HIGH:
                        data->buflen += data->values[i] =
-                               audit_pack_string(&bufp, f->se_str);
+                               audit_pack_string(&bufp, f->lsm_str);
                        break;
                case AUDIT_WATCH:
                        data->buflen += data->values[i] =
-                               audit_pack_string(&bufp, krule->watch->path);
+                               audit_pack_string(&bufp,
+                                                 audit_watch_path(krule->watch));
+                       break;
+               case AUDIT_DIR:
+                       data->buflen += data->values[i] =
+                               audit_pack_string(&bufp,
+                                                 audit_tree_path(krule->tree));
+                       break;
+               case AUDIT_FILTERKEY:
+                       data->buflen += data->values[i] =
+                               audit_pack_string(&bufp, krule->filterkey);
                        break;
                default:
                        data->values[i] = f->val;
@@ -600,16 +628,49 @@ static int audit_compare_rule(struct audit_krule *a, struct audit_krule *b)
                        return 1;
 
                switch(a->fields[i].type) {
-               case AUDIT_SE_USER:
-               case AUDIT_SE_ROLE:
-               case AUDIT_SE_TYPE:
-               case AUDIT_SE_SEN:
-               case AUDIT_SE_CLR:
-                       if (strcmp(a->fields[i].se_str, b->fields[i].se_str))
+               case AUDIT_SUBJ_USER:
+               case AUDIT_SUBJ_ROLE:
+               case AUDIT_SUBJ_TYPE:
+               case AUDIT_SUBJ_SEN:
+               case AUDIT_SUBJ_CLR:
+               case AUDIT_OBJ_USER:
+               case AUDIT_OBJ_ROLE:
+               case AUDIT_OBJ_TYPE:
+               case AUDIT_OBJ_LEV_LOW:
+               case AUDIT_OBJ_LEV_HIGH:
+                       if (strcmp(a->fields[i].lsm_str, b->fields[i].lsm_str))
                                return 1;
                        break;
                case AUDIT_WATCH:
-                       if (strcmp(a->watch->path, b->watch->path))
+                       if (strcmp(audit_watch_path(a->watch),
+                                  audit_watch_path(b->watch)))
+                               return 1;
+                       break;
+               case AUDIT_DIR:
+                       if (strcmp(audit_tree_path(a->tree),
+                                  audit_tree_path(b->tree)))
+                               return 1;
+                       break;
+               case AUDIT_FILTERKEY:
+                       /* both filterkeys exist based on above type compare */
+                       if (strcmp(a->filterkey, b->filterkey))
+                               return 1;
+                       break;
+               case AUDIT_UID:
+               case AUDIT_EUID:
+               case AUDIT_SUID:
+               case AUDIT_FSUID:
+               case AUDIT_LOGINUID:
+               case AUDIT_OBJ_UID:
+                       if (!uid_eq(a->fields[i].uid, b->fields[i].uid))
+                               return 1;
+                       break;
+               case AUDIT_GID:
+               case AUDIT_EGID:
+               case AUDIT_SGID:
+               case AUDIT_FSGID:
+               case AUDIT_OBJ_GID:
+                       if (!gid_eq(a->fields[i].gid, b->fields[i].gid))
                                return 1;
                        break;
                default:
@@ -625,54 +686,28 @@ static int audit_compare_rule(struct audit_krule *a, struct audit_krule *b)
        return 0;
 }
 
-/* Duplicate the given audit watch.  The new watch's rules list is initialized
- * to an empty list and wlist is undefined. */
-static struct audit_watch *audit_dupe_watch(struct audit_watch *old)
-{
-       char *path;
-       struct audit_watch *new;
-
-       path = kstrdup(old->path, GFP_KERNEL);
-       if (unlikely(!path))
-               return ERR_PTR(-ENOMEM);
-
-       new = audit_init_watch(path);
-       if (unlikely(IS_ERR(new))) {
-               kfree(path);
-               goto out;
-       }
-
-       new->dev = old->dev;
-       new->ino = old->ino;
-       get_inotify_watch(&old->parent->wdata);
-       new->parent = old->parent;
-
-out:
-       return new;
-}
-
-/* Duplicate selinux field information.  The se_rule is opaque, so must be
+/* Duplicate LSM field information.  The lsm_rule is opaque, so must be
  * re-initialized. */
-static inline int audit_dupe_selinux_field(struct audit_field *df,
+static inline int audit_dupe_lsm_field(struct audit_field *df,
                                           struct audit_field *sf)
 {
        int ret = 0;
-       char *se_str;
+       char *lsm_str;
 
-       /* our own copy of se_str */
-       se_str = kstrdup(sf->se_str, GFP_KERNEL);
-       if (unlikely(IS_ERR(se_str)))
-           return -ENOMEM;
-       df->se_str = se_str;
+       /* our own copy of lsm_str */
+       lsm_str = kstrdup(sf->lsm_str, GFP_KERNEL);
+       if (unlikely(!lsm_str))
+               return -ENOMEM;
+       df->lsm_str = lsm_str;
 
-       /* our own (refreshed) copy of se_rule */
-       ret = selinux_audit_rule_init(df->type, df->op, df->se_str,
-                                     &df->se_rule);
+       /* our own (refreshed) copy of lsm_rule */
+       ret = security_audit_rule_init(df->type, df->op, df->lsm_str,
+                                      (void **)&df->lsm_rule);
        /* Keep currently invalid fields around in case they
         * become valid after a policy reload. */
        if (ret == -EINVAL) {
-               printk(KERN_WARNING "audit rule for selinux \'%s\' is "
-                      "invalid\n", df->se_str);
+               printk(KERN_WARNING "audit rule for LSM \'%s\' is "
+                      "invalid\n", df->lsm_str);
                ret = 0;
        }
 
@@ -680,17 +715,17 @@ static inline int audit_dupe_selinux_field(struct audit_field *df,
 }
 
 /* Duplicate an audit rule.  This will be a deep copy with the exception
- * of the watch - that pointer is carried over.  The selinux specific fields
+ * of the watch - that pointer is carried over.  The LSM specific fields
  * will be updated in the copy.  The point is to be able to replace the old
  * rule with the new rule in the filterlist, then free the old rule.
  * The rlist element is undefined; list manipulations are handled apart from
  * the initial copy. */
-static struct audit_entry *audit_dupe_rule(struct audit_krule *old,
-                                          struct audit_watch *watch)
+struct audit_entry *audit_dupe_rule(struct audit_krule *old)
 {
        u32 fcount = old->field_count;
        struct audit_entry *entry;
        struct audit_krule *new;
+       char *fk;
        int i, err = 0;
 
        entry = audit_init_entry(fcount);
@@ -704,23 +739,44 @@ static struct audit_entry *audit_dupe_rule(struct audit_krule *old,
        new->action = old->action;
        for (i = 0; i < AUDIT_BITMASK_SIZE; i++)
                new->mask[i] = old->mask[i];
+       new->prio = old->prio;
        new->buflen = old->buflen;
        new->inode_f = old->inode_f;
-       new->watch = NULL;
        new->field_count = old->field_count;
+
+       /*
+        * note that we are OK with not refcounting here; audit_match_tree()
+        * never dereferences tree and we can't get false positives there
+        * since we'd have to have rule gone from the list *and* removed
+        * before the chunks found by lookup had been allocated, i.e. before
+        * the beginning of list scan.
+        */
+       new->tree = old->tree;
        memcpy(new->fields, old->fields, sizeof(struct audit_field) * fcount);
 
-       /* deep copy this information, updating the se_rule fields, because
+       /* deep copy this information, updating the lsm_rule fields, because
         * the originals will all be freed when the old rule is freed. */
        for (i = 0; i < fcount; i++) {
                switch (new->fields[i].type) {
-               case AUDIT_SE_USER:
-               case AUDIT_SE_ROLE:
-               case AUDIT_SE_TYPE:
-               case AUDIT_SE_SEN:
-               case AUDIT_SE_CLR:
-                       err = audit_dupe_selinux_field(&new->fields[i],
+               case AUDIT_SUBJ_USER:
+               case AUDIT_SUBJ_ROLE:
+               case AUDIT_SUBJ_TYPE:
+               case AUDIT_SUBJ_SEN:
+               case AUDIT_SUBJ_CLR:
+               case AUDIT_OBJ_USER:
+               case AUDIT_OBJ_ROLE:
+               case AUDIT_OBJ_TYPE:
+               case AUDIT_OBJ_LEV_LOW:
+               case AUDIT_OBJ_LEV_HIGH:
+                       err = audit_dupe_lsm_field(&new->fields[i],
                                                       &old->fields[i]);
+                       break;
+               case AUDIT_FILTERKEY:
+                       fk = kstrdup(old->filterkey, GFP_KERNEL);
+                       if (unlikely(!fk))
+                               err = -ENOMEM;
+                       else
+                               new->filterkey = fk;
                }
                if (err) {
                        audit_free_rule(entry);
@@ -728,128 +784,27 @@ static struct audit_entry *audit_dupe_rule(struct audit_krule *old,
                }
        }
 
-       if (watch) {
-               audit_get_watch(watch);
-               new->watch = watch;
+       if (old->watch) {
+               audit_get_watch(old->watch);
+               new->watch = old->watch;
        }
 
        return entry;
 }
 
-/* Update inode info in audit rules based on filesystem event. */
-static void audit_update_watch(struct audit_parent *parent,
-                              const char *dname, dev_t dev,
-                              unsigned long ino, unsigned invalidating)
-{
-       struct audit_watch *owatch, *nwatch, *nextw;
-       struct audit_krule *r, *nextr;
-       struct audit_entry *oentry, *nentry;
-       struct audit_buffer *ab;
-
-       mutex_lock(&audit_filter_mutex);
-       list_for_each_entry_safe(owatch, nextw, &parent->watches, wlist) {
-               if (audit_compare_dname_path(dname, owatch->path))
-                       continue;
-
-               /* If the update involves invalidating rules, do the inode-based
-                * filtering now, so we don't omit records. */
-               if (invalidating &&
-                   audit_filter_inodes(current, current->audit_context) == AUDIT_RECORD_CONTEXT)
-                       audit_set_auditable(current->audit_context);
-
-               nwatch = audit_dupe_watch(owatch);
-               if (unlikely(IS_ERR(nwatch))) {
-                       mutex_unlock(&audit_filter_mutex);
-                       audit_panic("error updating watch, skipping");
-                       return;
-               }
-               nwatch->dev = dev;
-               nwatch->ino = ino;
-
-               list_for_each_entry_safe(r, nextr, &owatch->rules, rlist) {
-
-                       oentry = container_of(r, struct audit_entry, rule);
-                       list_del(&oentry->rule.rlist);
-                       list_del_rcu(&oentry->list);
-
-                       nentry = audit_dupe_rule(&oentry->rule, nwatch);
-                       if (unlikely(IS_ERR(nentry)))
-                               audit_panic("error updating watch, removing");
-                       else {
-                               int h = audit_hash_ino((u32)ino);
-                               list_add(&nentry->rule.rlist, &nwatch->rules);
-                               list_add_rcu(&nentry->list, &audit_inode_hash[h]);
-                       }
-
-                       call_rcu(&oentry->rcu, audit_free_rule_rcu);
-               }
-
-               ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE);
-               audit_log_format(ab, "audit updated rules specifying watch=");
-               audit_log_untrustedstring(ab, owatch->path);
-               audit_log_format(ab, " with dev=%u ino=%lu\n", dev, ino);
-               audit_log_end(ab);
-
-               audit_remove_watch(owatch);
-               goto add_watch_to_parent; /* event applies to a single watch */
-       }
-       mutex_unlock(&audit_filter_mutex);
-       return;
-
-add_watch_to_parent:
-       list_add(&nwatch->wlist, &parent->watches);
-       mutex_unlock(&audit_filter_mutex);
-       return;
-}
-
-/* Remove all watches & rules associated with a parent that is going away. */
-static void audit_remove_parent_watches(struct audit_parent *parent)
-{
-       struct audit_watch *w, *nextw;
-       struct audit_krule *r, *nextr;
-       struct audit_entry *e;
-
-       mutex_lock(&audit_filter_mutex);
-       parent->flags |= AUDIT_PARENT_INVALID;
-       list_for_each_entry_safe(w, nextw, &parent->watches, wlist) {
-               list_for_each_entry_safe(r, nextr, &w->rules, rlist) {
-                       e = container_of(r, struct audit_entry, rule);
-                       list_del(&r->rlist);
-                       list_del_rcu(&e->list);
-                       call_rcu(&e->rcu, audit_free_rule_rcu);
-
-                       audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                                "audit implicitly removed rule from list=%d\n",
-                                 AUDIT_FILTER_EXIT);
-               }
-               audit_remove_watch(w);
-       }
-       mutex_unlock(&audit_filter_mutex);
-}
-
-/* Unregister inotify watches for parents on in_list.
- * Generates an IN_IGNORED event. */
-static void audit_inotify_unregister(struct list_head *in_list)
-{
-       struct audit_parent *p, *n;
-
-       list_for_each_entry_safe(p, n, in_list, ilist) {
-               list_del(&p->ilist);
-               inotify_rm_watch(audit_ih, &p->wdata);
-               /* the put matching the get in audit_do_del_rule() */
-               put_inotify_watch(&p->wdata);
-       }
-}
-
 /* Find an existing audit rule.
  * Caller must hold audit_filter_mutex to prevent stale rule data. */
 static struct audit_entry *audit_find_rule(struct audit_entry *entry,
-                                          struct list_head *list)
+                                          struct list_head **p)
 {
        struct audit_entry *e, *found = NULL;
+       struct list_head *list;
        int h;
 
-       if (entry->rule.watch) {
+       if (entry->rule.inode_f) {
+               h = audit_hash_ino(entry->rule.inode_f->val);
+               *p = list = &audit_inode_hash[h];
+       } else if (entry->rule.watch) {
                /* we don't know the inode number, so must walk entire hash */
                for (h = 0; h < AUDIT_INODE_BUCKETS; h++) {
                        list = &audit_inode_hash[h];
@@ -860,6 +815,8 @@ static struct audit_entry *audit_find_rule(struct audit_entry *entry,
                                }
                }
                goto out;
+       } else {
+               *p = list = &audit_filter_list[entry->rule.listnr];
        }
 
        list_for_each_entry(e, list, list)
@@ -872,325 +829,155 @@ out:
        return found;
 }
 
-/* Get path information necessary for adding watches. */
-static int audit_get_nd(char *path, struct nameidata **ndp,
-                       struct nameidata **ndw)
-{
-       struct nameidata *ndparent, *ndwatch;
-       int err;
-
-       ndparent = kmalloc(sizeof(*ndparent), GFP_KERNEL);
-       if (unlikely(!ndparent))
-               return -ENOMEM;
-
-       ndwatch = kmalloc(sizeof(*ndwatch), GFP_KERNEL);
-       if (unlikely(!ndwatch)) {
-               kfree(ndparent);
-               return -ENOMEM;
-       }
-
-       err = path_lookup(path, LOOKUP_PARENT, ndparent);
-       if (err) {
-               kfree(ndparent);
-               kfree(ndwatch);
-               return err;
-       }
-
-       err = path_lookup(path, 0, ndwatch);
-       if (err) {
-               kfree(ndwatch);
-               ndwatch = NULL;
-       }
-
-       *ndp = ndparent;
-       *ndw = ndwatch;
-
-       return 0;
-}
-
-/* Release resources used for watch path information. */
-static void audit_put_nd(struct nameidata *ndp, struct nameidata *ndw)
-{
-       if (ndp) {
-               path_release(ndp);
-               kfree(ndp);
-       }
-       if (ndw) {
-               path_release(ndw);
-               kfree(ndw);
-       }
-}
-
-/* Associate the given rule with an existing parent inotify_watch.
- * Caller must hold audit_filter_mutex. */
-static void audit_add_to_parent(struct audit_krule *krule,
-                               struct audit_parent *parent)
-{
-       struct audit_watch *w, *watch = krule->watch;
-       int watch_found = 0;
-
-       list_for_each_entry(w, &parent->watches, wlist) {
-               if (strcmp(watch->path, w->path))
-                       continue;
-
-               watch_found = 1;
-
-               /* put krule's and initial refs to temporary watch */
-               audit_put_watch(watch);
-               audit_put_watch(watch);
-
-               audit_get_watch(w);
-               krule->watch = watch = w;
-               break;
-       }
-
-       if (!watch_found) {
-               get_inotify_watch(&parent->wdata);
-               watch->parent = parent;
-
-               list_add(&watch->wlist, &parent->watches);
-       }
-       list_add(&krule->rlist, &watch->rules);
-}
-
-/* Find a matching watch entry, or add this one.
- * Caller must hold audit_filter_mutex. */
-static int audit_add_watch(struct audit_krule *krule, struct nameidata *ndp,
-                          struct nameidata *ndw)
-{
-       struct audit_watch *watch = krule->watch;
-       struct inotify_watch *i_watch;
-       struct audit_parent *parent;
-       int ret = 0;
-
-       /* update watch filter fields */
-       if (ndw) {
-               watch->dev = ndw->dentry->d_inode->i_sb->s_dev;
-               watch->ino = ndw->dentry->d_inode->i_ino;
-       }
-
-       /* The audit_filter_mutex must not be held during inotify calls because
-        * we hold it during inotify event callback processing.  If an existing
-        * inotify watch is found, inotify_find_watch() grabs a reference before
-        * returning.
-        */
-       mutex_unlock(&audit_filter_mutex);
-
-       if (inotify_find_watch(audit_ih, ndp->dentry->d_inode, &i_watch) < 0) {
-               parent = audit_init_parent(ndp);
-               if (IS_ERR(parent)) {
-                       /* caller expects mutex locked */
-                       mutex_lock(&audit_filter_mutex);
-                       return PTR_ERR(parent);
-               }
-       } else
-               parent = container_of(i_watch, struct audit_parent, wdata);
-
-       mutex_lock(&audit_filter_mutex);
-
-       /* parent was moved before we took audit_filter_mutex */
-       if (parent->flags & AUDIT_PARENT_INVALID)
-               ret = -ENOENT;
-       else
-               audit_add_to_parent(krule, parent);
-
-       /* match get in audit_init_parent or inotify_find_watch */
-       put_inotify_watch(&parent->wdata);
-       return ret;
-}
+static u64 prio_low = ~0ULL/2;
+static u64 prio_high = ~0ULL/2 - 1;
 
 /* Add rule to given filterlist if not a duplicate. */
-static inline int audit_add_rule(struct audit_entry *entry,
-                                struct list_head *list)
+static inline int audit_add_rule(struct audit_entry *entry)
 {
        struct audit_entry *e;
-       struct audit_field *inode_f = entry->rule.inode_f;
        struct audit_watch *watch = entry->rule.watch;
-       struct nameidata *ndp, *ndw;
-       int h, err, putnd_needed = 0;
+       struct audit_tree *tree = entry->rule.tree;
+       struct list_head *list;
+       int err;
+#ifdef CONFIG_AUDITSYSCALL
+       int dont_count = 0;
 
-       if (inode_f) {
-               h = audit_hash_ino(inode_f->val);
-               list = &audit_inode_hash[h];
-       }
+       /* If either of these, don't count towards total */
+       if (entry->rule.listnr == AUDIT_FILTER_USER ||
+               entry->rule.listnr == AUDIT_FILTER_TYPE)
+               dont_count = 1;
+#endif
 
        mutex_lock(&audit_filter_mutex);
-       e = audit_find_rule(entry, list);
-       mutex_unlock(&audit_filter_mutex);
+       e = audit_find_rule(entry, &list);
        if (e) {
+               mutex_unlock(&audit_filter_mutex);
                err = -EEXIST;
+               /* normally audit_add_tree_rule() will free it on failure */
+               if (tree)
+                       audit_put_tree(tree);
                goto error;
        }
 
-       /* Avoid calling path_lookup under audit_filter_mutex. */
        if (watch) {
-               err = audit_get_nd(watch->path, &ndp, &ndw);
-               if (err)
+               /* audit_filter_mutex is dropped and re-taken during this call */
+               err = audit_add_watch(&entry->rule, &list);
+               if (err) {
+                       mutex_unlock(&audit_filter_mutex);
                        goto error;
-               putnd_needed = 1;
+               }
        }
-
-       mutex_lock(&audit_filter_mutex);
-       if (watch) {
-               /* audit_filter_mutex is dropped and re-taken during this call */
-               err = audit_add_watch(&entry->rule, ndp, ndw);
+       if (tree) {
+               err = audit_add_tree_rule(&entry->rule);
                if (err) {
                        mutex_unlock(&audit_filter_mutex);
                        goto error;
                }
-               h = audit_hash_ino((u32)watch->ino);
-               list = &audit_inode_hash[h];
+       }
+
+       entry->rule.prio = ~0ULL;
+       if (entry->rule.listnr == AUDIT_FILTER_EXIT) {
+               if (entry->rule.flags & AUDIT_FILTER_PREPEND)
+                       entry->rule.prio = ++prio_high;
+               else
+                       entry->rule.prio = --prio_low;
        }
 
        if (entry->rule.flags & AUDIT_FILTER_PREPEND) {
+               list_add(&entry->rule.list,
+                        &audit_rules_list[entry->rule.listnr]);
                list_add_rcu(&entry->list, list);
+               entry->rule.flags &= ~AUDIT_FILTER_PREPEND;
        } else {
+               list_add_tail(&entry->rule.list,
+                             &audit_rules_list[entry->rule.listnr]);
                list_add_tail_rcu(&entry->list, list);
        }
-       mutex_unlock(&audit_filter_mutex);
+#ifdef CONFIG_AUDITSYSCALL
+       if (!dont_count)
+               audit_n_rules++;
 
-       if (putnd_needed)
-               audit_put_nd(ndp, ndw);
+       if (!audit_match_signal(entry))
+               audit_signals++;
+#endif
+       mutex_unlock(&audit_filter_mutex);
 
        return 0;
 
 error:
-       if (putnd_needed)
-               audit_put_nd(ndp, ndw);
        if (watch)
                audit_put_watch(watch); /* tmp watch, matches initial get */
        return err;
 }
 
 /* Remove an existing rule from filterlist. */
-static inline int audit_del_rule(struct audit_entry *entry,
-                                struct list_head *list)
+static inline int audit_del_rule(struct audit_entry *entry)
 {
        struct audit_entry  *e;
-       struct audit_field *inode_f = entry->rule.inode_f;
-       struct audit_watch *watch, *tmp_watch = entry->rule.watch;
-       LIST_HEAD(inotify_list);
-       int h, ret = 0;
-
-       if (inode_f) {
-               h = audit_hash_ino(inode_f->val);
-               list = &audit_inode_hash[h];
-       }
+       struct audit_watch *watch = entry->rule.watch;
+       struct audit_tree *tree = entry->rule.tree;
+       struct list_head *list;
+       int ret = 0;
+#ifdef CONFIG_AUDITSYSCALL
+       int dont_count = 0;
+
+       /* If either of these, don't count towards total */
+       if (entry->rule.listnr == AUDIT_FILTER_USER ||
+               entry->rule.listnr == AUDIT_FILTER_TYPE)
+               dont_count = 1;
+#endif
 
        mutex_lock(&audit_filter_mutex);
-       e = audit_find_rule(entry, list);
+       e = audit_find_rule(entry, &list);
        if (!e) {
                mutex_unlock(&audit_filter_mutex);
                ret = -ENOENT;
                goto out;
        }
 
-       watch = e->rule.watch;
-       if (watch) {
-               struct audit_parent *parent = watch->parent;
-
-               list_del(&e->rule.rlist);
-
-               if (list_empty(&watch->rules)) {
-                       audit_remove_watch(watch);
+       if (e->rule.watch)
+               audit_remove_watch_rule(&e->rule);
 
-                       if (list_empty(&parent->watches)) {
-                               /* Put parent on the inotify un-registration
-                                * list.  Grab a reference before releasing
-                                * audit_filter_mutex, to be released in
-                                * audit_inotify_unregister(). */
-                               list_add(&parent->ilist, &inotify_list);
-                               get_inotify_watch(&parent->wdata);
-                       }
-               }
-       }
+       if (e->rule.tree)
+               audit_remove_tree_rule(&e->rule);
 
        list_del_rcu(&e->list);
+       list_del(&e->rule.list);
        call_rcu(&e->rcu, audit_free_rule_rcu);
 
-       mutex_unlock(&audit_filter_mutex);
+#ifdef CONFIG_AUDITSYSCALL
+       if (!dont_count)
+               audit_n_rules--;
 
-       if (!list_empty(&inotify_list))
-               audit_inotify_unregister(&inotify_list);
+       if (!audit_match_signal(entry))
+               audit_signals--;
+#endif
+       mutex_unlock(&audit_filter_mutex);
 
 out:
-       if (tmp_watch)
-               audit_put_watch(tmp_watch); /* match initial get */
+       if (watch)
+               audit_put_watch(watch); /* match initial get */
+       if (tree)
+               audit_put_tree(tree);   /* that's the temporary one */
 
        return ret;
 }
 
-/* List rules using struct audit_rule.  Exists for backward
- * compatibility with userspace. */
-static void audit_list(int pid, int seq, struct sk_buff_head *q)
-{
-       struct sk_buff *skb;
-       struct audit_entry *entry;
-       int i;
-
-       /* This is a blocking read, so use audit_filter_mutex instead of rcu
-        * iterator to sync with list writers. */
-       for (i=0; i<AUDIT_NR_FILTERS; i++) {
-               list_for_each_entry(entry, &audit_filter_list[i], list) {
-                       struct audit_rule *rule;
-
-                       rule = audit_krule_to_rule(&entry->rule);
-                       if (unlikely(!rule))
-                               break;
-                       skb = audit_make_reply(pid, seq, AUDIT_LIST, 0, 1,
-                                        rule, sizeof(*rule));
-                       if (skb)
-                               skb_queue_tail(q, skb);
-                       kfree(rule);
-               }
-       }
-       for (i = 0; i < AUDIT_INODE_BUCKETS; i++) {
-               list_for_each_entry(entry, &audit_inode_hash[i], list) {
-                       struct audit_rule *rule;
-
-                       rule = audit_krule_to_rule(&entry->rule);
-                       if (unlikely(!rule))
-                               break;
-                       skb = audit_make_reply(pid, seq, AUDIT_LIST, 0, 1,
-                                        rule, sizeof(*rule));
-                       if (skb)
-                               skb_queue_tail(q, skb);
-                       kfree(rule);
-               }
-       }
-       skb = audit_make_reply(pid, seq, AUDIT_LIST, 1, 1, NULL, 0);
-       if (skb)
-               skb_queue_tail(q, skb);
-}
-
 /* List rules using struct audit_rule_data. */
 static void audit_list_rules(int pid, int seq, struct sk_buff_head *q)
 {
        struct sk_buff *skb;
-       struct audit_entry *e;
+       struct audit_krule *r;
        int i;
 
        /* This is a blocking read, so use audit_filter_mutex instead of rcu
         * iterator to sync with list writers. */
        for (i=0; i<AUDIT_NR_FILTERS; i++) {
-               list_for_each_entry(e, &audit_filter_list[i], list) {
-                       struct audit_rule_data *data;
-
-                       data = audit_krule_to_data(&e->rule);
-                       if (unlikely(!data))
-                               break;
-                       skb = audit_make_reply(pid, seq, AUDIT_LIST_RULES, 0, 1,
-                                        data, sizeof(*data) + data->buflen);
-                       if (skb)
-                               skb_queue_tail(q, skb);
-                       kfree(data);
-               }
-       }
-       for (i=0; i< AUDIT_INODE_BUCKETS; i++) {
-               list_for_each_entry(e, &audit_inode_hash[i], list) {
+               list_for_each_entry(r, &audit_rules_list[i], list) {
                        struct audit_rule_data *data;
 
-                       data = audit_krule_to_data(&e->rule);
+                       data = audit_krule_to_data(r);
                        if (unlikely(!data))
                                break;
                        skb = audit_make_reply(pid, seq, AUDIT_LIST_RULES, 0, 1,
@@ -1205,19 +992,37 @@ static void audit_list_rules(int pid, int seq, struct sk_buff_head *q)
                skb_queue_tail(q, skb);
 }
 
+/* Log rule additions and removals */
+static void audit_log_rule_change(char *action, struct audit_krule *rule, int res)
+{
+       struct audit_buffer *ab;
+       uid_t loginuid = from_kuid(&init_user_ns, audit_get_loginuid(current));
+       u32 sessionid = audit_get_sessionid(current);
+
+       if (!audit_enabled)
+               return;
+
+       ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE);
+       if (!ab)
+               return;
+       audit_log_format(ab, "auid=%u ses=%u" ,loginuid, sessionid);
+       audit_log_task_context(ab);
+       audit_log_format(ab, " op=");
+       audit_log_string(ab, action);
+       audit_log_key(ab, rule->filterkey);
+       audit_log_format(ab, " list=%d res=%d", rule->listnr, res);
+       audit_log_end(ab);
+}
+
 /**
  * audit_receive_filter - apply all rules to the specified message type
  * @type: audit message type
  * @pid: target pid for netlink audit messages
- * @uid: target uid for netlink audit messages
  * @seq: netlink audit message sequence (serial) number
  * @data: payload data
  * @datasz: size of payload data
- * @loginuid: loginuid of sender
- * @sid: SE Linux Security ID of sender
  */
-int audit_receive_filter(int type, int pid, int uid, int seq, void *data,
-                        size_t datasz, uid_t loginuid, u32 sid)
+int audit_receive_filter(int type, int pid, int seq, void *data, size_t datasz)
 {
        struct task_struct *tsk;
        struct audit_netlink_list *dest;
@@ -1225,14 +1030,13 @@ int audit_receive_filter(int type, int pid, int uid, int seq, void *data,
        struct audit_entry *entry;
 
        switch (type) {
-       case AUDIT_LIST:
        case AUDIT_LIST_RULES:
                /* We can't just spew out the rules here because we might fill
                 * the available socket buffer space and deadlock waiting for
                 * auditctl to read from it... which isn't ever going to
                 * happen if we're actually running in the context of auditctl
                 * trying to _send_ the stuff */
-                
+
                dest = kmalloc(sizeof(struct audit_netlink_list), GFP_KERNEL);
                if (!dest)
                        return -ENOMEM;
@@ -1240,10 +1044,7 @@ int audit_receive_filter(int type, int pid, int uid, int seq, void *data,
                skb_queue_head_init(&dest->q);
 
                mutex_lock(&audit_filter_mutex);
-               if (type == AUDIT_LIST)
-                       audit_list(pid, seq, &dest->q);
-               else
-                       audit_list_rules(pid, seq, &dest->q);
+               audit_list_rules(pid, seq, &dest->q);
                mutex_unlock(&audit_filter_mutex);
 
                tsk = kthread_run(audit_send_list, dest, "audit_send_list");
@@ -1253,69 +1054,23 @@ int audit_receive_filter(int type, int pid, int uid, int seq, void *data,
                        err = PTR_ERR(tsk);
                }
                break;
-       case AUDIT_ADD:
        case AUDIT_ADD_RULE:
-               if (type == AUDIT_ADD)
-                       entry = audit_rule_to_entry(data);
-               else
-                       entry = audit_data_to_entry(data, datasz);
+               entry = audit_data_to_entry(data, datasz);
                if (IS_ERR(entry))
                        return PTR_ERR(entry);
 
-               err = audit_add_rule(entry,
-                                    &audit_filter_list[entry->rule.listnr]);
-
-               if (sid) {
-                       char *ctx = NULL;
-                       u32 len;
-                       if (selinux_ctxid_to_string(sid, &ctx, &len)) {
-                               /* Maybe call audit_panic? */
-                               audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                                "auid=%u ssid=%u add rule to list=%d res=%d",
-                                loginuid, sid, entry->rule.listnr, !err);
-                       } else
-                               audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                                "auid=%u subj=%s add rule to list=%d res=%d",
-                                loginuid, ctx, entry->rule.listnr, !err);
-                       kfree(ctx);
-               } else
-                       audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                               "auid=%u add rule to list=%d res=%d",
-                               loginuid, entry->rule.listnr, !err);
-
+               err = audit_add_rule(entry);
+               audit_log_rule_change("add rule", &entry->rule, !err);
                if (err)
                        audit_free_rule(entry);
                break;
-       case AUDIT_DEL:
        case AUDIT_DEL_RULE:
-               if (type == AUDIT_DEL)
-                       entry = audit_rule_to_entry(data);
-               else
-                       entry = audit_data_to_entry(data, datasz);
+               entry = audit_data_to_entry(data, datasz);
                if (IS_ERR(entry))
                        return PTR_ERR(entry);
 
-               err = audit_del_rule(entry,
-                                    &audit_filter_list[entry->rule.listnr]);
-
-               if (sid) {
-                       char *ctx = NULL;
-                       u32 len;
-                       if (selinux_ctxid_to_string(sid, &ctx, &len)) {
-                               /* Maybe call audit_panic? */
-                               audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                                       "auid=%u ssid=%u remove rule from list=%d res=%d",
-                                        loginuid, sid, entry->rule.listnr, !err);
-                       } else
-                               audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                                       "auid=%u subj=%s remove rule from list=%d res=%d",
-                                        loginuid, ctx, entry->rule.listnr, !err);
-                       kfree(ctx);
-               } else
-                       audit_log(NULL, GFP_KERNEL, AUDIT_CONFIG_CHANGE,
-                               "auid=%u remove rule from list=%d res=%d",
-                               loginuid, entry->rule.listnr, !err);
-
+               err = audit_del_rule(entry);
+               audit_log_rule_change("remove rule", &entry->rule, !err);
                audit_free_rule(entry);
                break;
        default:
@@ -1325,62 +1080,135 @@ int audit_receive_filter(int type, int pid, int uid, int seq, void *data,
        return err;
 }
 
-int audit_comparator(const u32 left, const u32 op, const u32 right)
+int audit_comparator(u32 left, u32 op, u32 right)
 {
        switch (op) {
-       case AUDIT_EQUAL:
+       case Audit_equal:
                return (left == right);
-       case AUDIT_NOT_EQUAL:
+       case Audit_not_equal:
                return (left != right);
-       case AUDIT_LESS_THAN:
+       case Audit_lt:
                return (left < right);
-       case AUDIT_LESS_THAN_OR_EQUAL:
+       case Audit_le:
                return (left <= right);
-       case AUDIT_GREATER_THAN:
+       case Audit_gt:
                return (left > right);
-       case AUDIT_GREATER_THAN_OR_EQUAL:
+       case Audit_ge:
                return (left >= right);
+       case Audit_bitmask:
+               return (left & right);
+       case Audit_bittest:
+               return ((left & right) == right);
+       default:
+               BUG();
+               return 0;
        }
-       BUG();
-       return 0;
 }
 
-/* Compare given dentry name with last component in given path,
- * return of 0 indicates a match. */
-int audit_compare_dname_path(const char *dname, const char *path)
+int audit_uid_comparator(kuid_t left, u32 op, kuid_t right)
 {
-       int dlen, plen;
-       const char *p;
+       switch (op) {
+       case Audit_equal:
+               return uid_eq(left, right);
+       case Audit_not_equal:
+               return !uid_eq(left, right);
+       case Audit_lt:
+               return uid_lt(left, right);
+       case Audit_le:
+               return uid_lte(left, right);
+       case Audit_gt:
+               return uid_gt(left, right);
+       case Audit_ge:
+               return uid_gte(left, right);
+       case Audit_bitmask:
+       case Audit_bittest:
+       default:
+               BUG();
+               return 0;
+       }
+}
 
-       if (!dname || !path)
-               return 1;
+int audit_gid_comparator(kgid_t left, u32 op, kgid_t right)
+{
+       switch (op) {
+       case Audit_equal:
+               return gid_eq(left, right);
+       case Audit_not_equal:
+               return !gid_eq(left, right);
+       case Audit_lt:
+               return gid_lt(left, right);
+       case Audit_le:
+               return gid_lte(left, right);
+       case Audit_gt:
+               return gid_gt(left, right);
+       case Audit_ge:
+               return gid_gte(left, right);
+       case Audit_bitmask:
+       case Audit_bittest:
+       default:
+               BUG();
+               return 0;
+       }
+}
+
+/**
+ * parent_len - find the length of the parent portion of a pathname
+ * @path: pathname of which to determine length
+ */
+int parent_len(const char *path)
+{
+       int plen;
+       const char *p;
 
-       dlen = strlen(dname);
        plen = strlen(path);
-       if (plen < dlen)
-               return 1;
+
+       if (plen == 0)
+               return plen;
 
        /* disregard trailing slashes */
        p = path + plen - 1;
        while ((*p == '/') && (p > path))
                p--;
 
-       /* find last path component */
-       p = p - dlen + 1;
-       if (p < path)
+       /* walk backward until we find the next slash or hit beginning */
+       while ((*p != '/') && (p > path))
+               p--;
+
+       /* did we find a slash? Then increment to include it in path */
+       if (*p == '/')
+               p++;
+
+       return p - path;
+}
+
+/**
+ * audit_compare_dname_path - compare given dentry name with last component in
+ *                           given path. Return of 0 indicates a match.
+ * @dname:     dentry name that we're comparing
+ * @path:      full pathname that we're comparing
+ * @parentlen: length of the parent if known. Passing in AUDIT_NAME_FULL
+ *             here indicates that we must compute this value.
+ */
+int audit_compare_dname_path(const char *dname, const char *path, int parentlen)
+{
+       int dlen, pathlen;
+       const char *p;
+
+       dlen = strlen(dname);
+       pathlen = strlen(path);
+       if (pathlen < dlen)
                return 1;
-       else if (p > path) {
-               if (*--p != '/')
-                       return 1;
-               else
-                       p++;
-       }
+
+       parentlen = parentlen == AUDIT_NAME_FULL ? parent_len(path) : parentlen;
+       if (pathlen - parentlen != dlen)
+               return 1;
+
+       p = path + parentlen;
 
        return strncmp(p, dname, dlen);
 }
 
-static int audit_filter_user_rules(struct netlink_skb_parms *cb,
-                                  struct audit_krule *rule,
+static int audit_filter_user_rules(struct audit_krule *rule, int type,
                                   enum audit_state *state)
 {
        int i;
@@ -1388,19 +1216,42 @@ static int audit_filter_user_rules(struct netlink_skb_parms *cb,
        for (i = 0; i < rule->field_count; i++) {
                struct audit_field *f = &rule->fields[i];
                int result = 0;
+               u32 sid;
 
                switch (f->type) {
                case AUDIT_PID:
-                       result = audit_comparator(cb->creds.pid, f->op, f->val);
+                       result = audit_comparator(task_pid_vnr(current), f->op, f->val);
                        break;
                case AUDIT_UID:
-                       result = audit_comparator(cb->creds.uid, f->op, f->val);
+                       result = audit_uid_comparator(current_uid(), f->op, f->uid);
                        break;
                case AUDIT_GID:
-                       result = audit_comparator(cb->creds.gid, f->op, f->val);
+                       result = audit_gid_comparator(current_gid(), f->op, f->gid);
                        break;
                case AUDIT_LOGINUID:
-                       result = audit_comparator(cb->loginuid, f->op, f->val);
+                       result = audit_uid_comparator(audit_get_loginuid(current),
+                                                 f->op, f->uid);
+                       break;
+               case AUDIT_LOGINUID_SET:
+                       result = audit_comparator(audit_loginuid_set(current),
+                                                 f->op, f->val);
+                       break;
+               case AUDIT_MSGTYPE:
+                       result = audit_comparator(type, f->op, f->val);
+                       break;
+               case AUDIT_SUBJ_USER:
+               case AUDIT_SUBJ_ROLE:
+               case AUDIT_SUBJ_TYPE:
+               case AUDIT_SUBJ_SEN:
+               case AUDIT_SUBJ_CLR:
+                       if (f->lsm_rule) {
+                               security_task_getsecid(current, &sid);
+                               result = security_audit_rule_match(sid,
+                                                                  f->type,
+                                                                  f->op,
+                                                                  f->lsm_rule,
+                                                                  NULL);
+                       }
                        break;
                }
 
@@ -1414,15 +1265,15 @@ static int audit_filter_user_rules(struct netlink_skb_parms *cb,
        return 1;
 }
 
-int audit_filter_user(struct netlink_skb_parms *cb, int type)
+int audit_filter_user(int type)
 {
+       enum audit_state state = AUDIT_DISABLED;
        struct audit_entry *e;
-       enum audit_state   state;
        int ret = 1;
 
        rcu_read_lock();
        list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_USER], list) {
-               if (audit_filter_user_rules(cb, &e->rule, &state)) {
+               if (audit_filter_user_rules(&e->rule, type, &state)) {
                        if (state == AUDIT_DISABLED)
                                ret = 0;
                        break;
@@ -1437,7 +1288,7 @@ int audit_filter_type(int type)
 {
        struct audit_entry *e;
        int result = 0;
-       
+
        rcu_read_lock();
        if (list_empty(&audit_filter_list[AUDIT_FILTER_TYPE]))
                goto unlock_and_return;
@@ -1461,94 +1312,57 @@ unlock_and_return:
        return result;
 }
 
-/* Check to see if the rule contains any selinux fields.  Returns 1 if there
-   are selinux fields specified in the rule, 0 otherwise. */
-static inline int audit_rule_has_selinux(struct audit_krule *rule)
+static int update_lsm_rule(struct audit_krule *r)
 {
-       int i;
+       struct audit_entry *entry = container_of(r, struct audit_entry, rule);
+       struct audit_entry *nentry;
+       int err = 0;
 
-       for (i = 0; i < rule->field_count; i++) {
-               struct audit_field *f = &rule->fields[i];
-               switch (f->type) {
-               case AUDIT_SE_USER:
-               case AUDIT_SE_ROLE:
-               case AUDIT_SE_TYPE:
-               case AUDIT_SE_SEN:
-               case AUDIT_SE_CLR:
-                       return 1;
-               }
+       if (!security_audit_rule_known(r))
+               return 0;
+
+       nentry = audit_dupe_rule(r);
+       if (IS_ERR(nentry)) {
+               /* save the first error encountered for the
+                * return value */
+               err = PTR_ERR(nentry);
+               audit_panic("error updating LSM filters");
+               if (r->watch)
+                       list_del(&r->rlist);
+               list_del_rcu(&entry->list);
+               list_del(&r->list);
+       } else {
+               if (r->watch || r->tree)
+                       list_replace_init(&r->rlist, &nentry->rule.rlist);
+               list_replace_rcu(&entry->list, &nentry->list);
+               list_replace(&r->list, &nentry->rule.list);
        }
+       call_rcu(&entry->rcu, audit_free_rule_rcu);
 
-       return 0;
+       return err;
 }
 
-/* This function will re-initialize the se_rule field of all applicable rules.
- * It will traverse the filter lists serarching for rules that contain selinux
+/* This function will re-initialize the lsm_rule field of all applicable rules.
+ * It will traverse the filter lists serarching for rules that contain LSM
  * specific filter fields.  When such a rule is found, it is copied, the
- * selinux field is re-initialized, and the old rule is replaced with the
+ * LSM field is re-initialized, and the old rule is replaced with the
  * updated rule. */
-int selinux_audit_rule_update(void)
+int audit_update_lsm_rules(void)
 {
-       struct audit_entry *entry, *n, *nentry;
-       struct audit_watch *watch;
+       struct audit_krule *r, *n;
        int i, err = 0;
 
        /* audit_filter_mutex synchronizes the writers */
        mutex_lock(&audit_filter_mutex);
 
        for (i = 0; i < AUDIT_NR_FILTERS; i++) {
-               list_for_each_entry_safe(entry, n, &audit_filter_list[i], list) {
-                       if (!audit_rule_has_selinux(&entry->rule))
-                               continue;
-
-                       watch = entry->rule.watch;
-                       nentry = audit_dupe_rule(&entry->rule, watch);
-                       if (unlikely(IS_ERR(nentry))) {
-                               /* save the first error encountered for the
-                                * return value */
-                               if (!err)
-                                       err = PTR_ERR(nentry);
-                               audit_panic("error updating selinux filters");
-                               if (watch)
-                                       list_del(&entry->rule.rlist);
-                               list_del_rcu(&entry->list);
-                       } else {
-                               if (watch) {
-                                       list_add(&nentry->rule.rlist,
-                                                &watch->rules);
-                                       list_del(&entry->rule.rlist);
-                               }
-                               list_replace_rcu(&entry->list, &nentry->list);
-                       }
-                       call_rcu(&entry->rcu, audit_free_rule_rcu);
+               list_for_each_entry_safe(r, n, &audit_rules_list[i], list) {
+                       int res = update_lsm_rule(r);
+                       if (!err)
+                               err = res;
                }
        }
-
        mutex_unlock(&audit_filter_mutex);
 
        return err;
 }
-
-/* Update watch data in audit rules based on inotify events. */
-void audit_handle_ievent(struct inotify_watch *i_watch, u32 wd, u32 mask,
-                        u32 cookie, const char *dname, struct inode *inode)
-{
-       struct audit_parent *parent;
-
-       parent = container_of(i_watch, struct audit_parent, wdata);
-
-       if (mask & (IN_CREATE|IN_MOVED_TO) && inode)
-               audit_update_watch(parent, dname, inode->i_sb->s_dev,
-                                  inode->i_ino, 0);
-       else if (mask & (IN_DELETE|IN_MOVED_FROM))
-               audit_update_watch(parent, dname, (dev_t)-1, (unsigned long)-1, 1);
-       /* inotify automatically removes the watch and sends IN_IGNORED */
-       else if (mask & (IN_DELETE_SELF|IN_UNMOUNT))
-               audit_remove_parent_watches(parent);
-       /* inotify does not remove the watch, so remove it manually */
-       else if(mask & IN_MOVE_SELF) {
-               audit_remove_parent_watches(parent);
-               inotify_remove_watch_locked(audit_ih, i_watch);
-       } else if (mask & IN_IGNORED)
-               put_inotify_watch(i_watch);
-}