futex: Fix uninterruptible loop due to gate_area
[linux-2.6.git] / kernel / auditsc.c
index 496c3dd..ce4b054 100644 (file)
 
 #include <linux/init.h>
 #include <asm/types.h>
-#include <asm/atomic.h>
+#include <linux/atomic.h>
 #include <linux/fs.h>
 #include <linux/namei.h>
 #include <linux/mm.h>
 #include <linux/module.h>
+#include <linux/slab.h>
 #include <linux/mount.h>
 #include <linux/socket.h>
 #include <linux/mqueue.h>
@@ -64,7 +65,8 @@
 #include <linux/binfmts.h>
 #include <linux/highmem.h>
 #include <linux/syscalls.h>
-#include <linux/inotify.h>
+#include <linux/capability.h>
+#include <linux/fs_struct.h>
 
 #include "audit.h"
 
@@ -84,6 +86,15 @@ int audit_n_rules;
 /* determines whether we collect data for signals sent */
 int audit_signals;
 
+struct audit_cap_data {
+       kernel_cap_t            permitted;
+       kernel_cap_t            inheritable;
+       union {
+               unsigned int    fE;             /* effective bit of a file capability */
+               kernel_cap_t    effective;      /* effective set of a process */
+       };
+};
+
 /* When fs/namei.c:getname() is called, we store the pointer in name and
  * we don't let putname() free it (instead we free all of the saved
  * pointers at syscall exit time).
@@ -100,6 +111,8 @@ struct audit_names {
        gid_t           gid;
        dev_t           rdev;
        u32             osid;
+       struct audit_cap_data fcap;
+       unsigned int    fcap_ver;
 };
 
 struct audit_aux_data {
@@ -112,43 +125,6 @@ struct audit_aux_data {
 /* Number of target pids per aux struct. */
 #define AUDIT_AUX_PIDS 16
 
-struct audit_aux_data_mq_open {
-       struct audit_aux_data   d;
-       int                     oflag;
-       mode_t                  mode;
-       struct mq_attr          attr;
-};
-
-struct audit_aux_data_mq_sendrecv {
-       struct audit_aux_data   d;
-       mqd_t                   mqdes;
-       size_t                  msg_len;
-       unsigned int            msg_prio;
-       struct timespec         abs_timeout;
-};
-
-struct audit_aux_data_mq_notify {
-       struct audit_aux_data   d;
-       mqd_t                   mqdes;
-       struct sigevent         notification;
-};
-
-struct audit_aux_data_mq_getsetattr {
-       struct audit_aux_data   d;
-       mqd_t                   mqdes;
-       struct mq_attr          mqstat;
-};
-
-struct audit_aux_data_ipcctl {
-       struct audit_aux_data   d;
-       struct ipc_perm         p;
-       unsigned long           qbytes;
-       uid_t                   uid;
-       gid_t                   gid;
-       mode_t                  mode;
-       u32                     osid;
-};
-
 struct audit_aux_data_execve {
        struct audit_aux_data   d;
        int argc;
@@ -156,23 +132,6 @@ struct audit_aux_data_execve {
        struct mm_struct *mm;
 };
 
-struct audit_aux_data_socketcall {
-       struct audit_aux_data   d;
-       int                     nargs;
-       unsigned long           args[0];
-};
-
-struct audit_aux_data_sockaddr {
-       struct audit_aux_data   d;
-       int                     len;
-       char                    a[0];
-};
-
-struct audit_aux_data_fd_pair {
-       struct  audit_aux_data d;
-       int     fd[2];
-};
-
 struct audit_aux_data_pids {
        struct audit_aux_data   d;
        pid_t                   target_pid[AUDIT_AUX_PIDS];
@@ -184,6 +143,20 @@ struct audit_aux_data_pids {
        int                     pid_count;
 };
 
+struct audit_aux_data_bprm_fcaps {
+       struct audit_aux_data   d;
+       struct audit_cap_data   fcap;
+       unsigned int            fcap_ver;
+       struct audit_cap_data   old_pcap;
+       struct audit_cap_data   new_pcap;
+};
+
+struct audit_aux_data_capset {
+       struct audit_aux_data   d;
+       pid_t                   pid;
+       struct audit_cap_data   cap;
+};
+
 struct audit_tree_refs {
        struct audit_tree_refs *next;
        struct audit_chunk *c[31];
@@ -193,14 +166,14 @@ struct audit_tree_refs {
 struct audit_context {
        int                 dummy;      /* must be the first element */
        int                 in_syscall; /* 1 if task is in a syscall */
-       enum audit_state    state;
+       enum audit_state    state, current_state;
        unsigned int        serial;     /* serial number for record */
-       struct timespec     ctime;      /* time of syscall entry */
        int                 major;      /* syscall number */
+       struct timespec     ctime;      /* time of syscall entry */
        unsigned long       argv[4];    /* syscall arguments */
-       int                 return_valid; /* return code is valid */
        long                return_code;/* syscall return code */
-       int                 auditable;  /* 1 if record should be written */
+       u64                 prio;
+       int                 return_valid; /* return code is valid */
        int                 name_count;
        struct audit_names  names[AUDIT_NAMES];
        char *              filterkey;  /* key for rule that triggered record */
@@ -208,7 +181,8 @@ struct audit_context {
        struct audit_context *previous; /* For nested syscalls */
        struct audit_aux_data *aux;
        struct audit_aux_data *aux_pids;
-
+       struct sockaddr_storage *sockaddr;
+       size_t sockaddr_len;
                                /* Save things to print about task_struct */
        pid_t               pid, ppid;
        uid_t               uid, euid, suid, fsuid;
@@ -224,15 +198,62 @@ struct audit_context {
        char                target_comm[TASK_COMM_LEN];
 
        struct audit_tree_refs *trees, *first_trees;
+       struct list_head killed_trees;
        int tree_count;
 
+       int type;
+       union {
+               struct {
+                       int nargs;
+                       long args[6];
+               } socketcall;
+               struct {
+                       uid_t                   uid;
+                       gid_t                   gid;
+                       mode_t                  mode;
+                       u32                     osid;
+                       int                     has_perm;
+                       uid_t                   perm_uid;
+                       gid_t                   perm_gid;
+                       mode_t                  perm_mode;
+                       unsigned long           qbytes;
+               } ipc;
+               struct {
+                       mqd_t                   mqdes;
+                       struct mq_attr          mqstat;
+               } mq_getsetattr;
+               struct {
+                       mqd_t                   mqdes;
+                       int                     sigev_signo;
+               } mq_notify;
+               struct {
+                       mqd_t                   mqdes;
+                       size_t                  msg_len;
+                       unsigned int            msg_prio;
+                       struct timespec         abs_timeout;
+               } mq_sendrecv;
+               struct {
+                       int                     oflag;
+                       mode_t                  mode;
+                       struct mq_attr          attr;
+               } mq_open;
+               struct {
+                       pid_t                   pid;
+                       struct audit_cap_data   cap;
+               } capset;
+               struct {
+                       int                     fd;
+                       int                     flags;
+               } mmap;
+       };
+       int fds[2];
+
 #if AUDIT_DEBUG
        int                 put_count;
        int                 ino_count;
 #endif
 };
 
-#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
 static inline int open_arg(int flags, int mask)
 {
        int n = ACC_MODE(flags);
@@ -243,7 +264,11 @@ static inline int open_arg(int flags, int mask)
 
 static int audit_match_perm(struct audit_context *ctx, int mask)
 {
-       unsigned n = ctx->major;
+       unsigned n;
+       if (unlikely(!ctx))
+               return 0;
+       n = ctx->major;
+
        switch (audit_classify_syscall(ctx->arch, n)) {
        case 0: /* native */
                if ((mask & AUDIT_PERM_WRITE) &&
@@ -284,6 +309,10 @@ static int audit_match_filetype(struct audit_context *ctx, int which)
 {
        unsigned index = which & ~S_IFMT;
        mode_t mode = which & S_IFMT;
+
+       if (unlikely(!ctx))
+               return 0;
+
        if (index >= ctx->name_count)
                return 0;
        if (ctx->names[index].ino == -1)
@@ -304,6 +333,14 @@ static int audit_match_filetype(struct audit_context *ctx, int which)
  */
 
 #ifdef CONFIG_AUDIT_TREE
+static void audit_set_auditable(struct audit_context *ctx)
+{
+       if (!ctx->prio) {
+               ctx->prio = 1;
+               ctx->current_state = AUDIT_RECORD_CONTEXT;
+       }
+}
+
 static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
 {
        struct audit_tree_refs *p = ctx->trees;
@@ -406,16 +443,25 @@ static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
 
 /* Determine if any context name data matches a rule's watch data */
 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
- * otherwise. */
+ * otherwise.
+ *
+ * If task_creation is true, this is an explicit indication that we are
+ * filtering a task rule at task creation time.  This and tsk == current are
+ * the only situations where tsk->cred may be accessed without an rcu read lock.
+ */
 static int audit_filter_rules(struct task_struct *tsk,
                              struct audit_krule *rule,
                              struct audit_context *ctx,
                              struct audit_names *name,
-                             enum audit_state *state)
+                             enum audit_state *state,
+                             bool task_creation)
 {
+       const struct cred *cred;
        int i, j, need_sid = 1;
        u32 sid;
 
+       cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
+
        for (i = 0; i < rule->field_count; i++) {
                struct audit_field *f = &rule->fields[i];
                int result = 0;
@@ -432,28 +478,28 @@ static int audit_filter_rules(struct task_struct *tsk,
                        }
                        break;
                case AUDIT_UID:
-                       result = audit_comparator(tsk->uid, f->op, f->val);
+                       result = audit_comparator(cred->uid, f->op, f->val);
                        break;
                case AUDIT_EUID:
-                       result = audit_comparator(tsk->euid, f->op, f->val);
+                       result = audit_comparator(cred->euid, f->op, f->val);
                        break;
                case AUDIT_SUID:
-                       result = audit_comparator(tsk->suid, f->op, f->val);
+                       result = audit_comparator(cred->suid, f->op, f->val);
                        break;
                case AUDIT_FSUID:
-                       result = audit_comparator(tsk->fsuid, f->op, f->val);
+                       result = audit_comparator(cred->fsuid, f->op, f->val);
                        break;
                case AUDIT_GID:
-                       result = audit_comparator(tsk->gid, f->op, f->val);
+                       result = audit_comparator(cred->gid, f->op, f->val);
                        break;
                case AUDIT_EGID:
-                       result = audit_comparator(tsk->egid, f->op, f->val);
+                       result = audit_comparator(cred->egid, f->op, f->val);
                        break;
                case AUDIT_SGID:
-                       result = audit_comparator(tsk->sgid, f->op, f->val);
+                       result = audit_comparator(cred->sgid, f->op, f->val);
                        break;
                case AUDIT_FSGID:
-                       result = audit_comparator(tsk->fsgid, f->op, f->val);
+                       result = audit_comparator(cred->fsgid, f->op, f->val);
                        break;
                case AUDIT_PERS:
                        result = audit_comparator(tsk->personality, f->op, f->val);
@@ -514,9 +560,8 @@ static int audit_filter_rules(struct task_struct *tsk,
                        }
                        break;
                case AUDIT_WATCH:
-                       if (name && rule->watch->ino != (unsigned long)-1)
-                               result = (name->dev == rule->watch->dev &&
-                                         name->ino == rule->watch->ino);
+                       if (name)
+                               result = audit_watch_compare(rule->watch, name->ino, name->dev);
                        break;
                case AUDIT_DIR:
                        if (ctx)
@@ -573,19 +618,12 @@ static int audit_filter_rules(struct task_struct *tsk,
                                        }
                                }
                                /* Find ipc objects that match */
-                               if (ctx) {
-                                       struct audit_aux_data *aux;
-                                       for (aux = ctx->aux; aux;
-                                            aux = aux->next) {
-                                               if (aux->type == AUDIT_IPC) {
-                                                       struct audit_aux_data_ipcctl *axi = (void *)aux;
-                                                       if (security_audit_rule_match(axi->osid, f->type, f->op, f->lsm_rule, ctx)) {
-                                                               ++result;
-                                                               break;
-                                                       }
-                                               }
-                                       }
-                               }
+                               if (!ctx || ctx->type != AUDIT_IPC)
+                                       break;
+                               if (security_audit_rule_match(ctx->ipc.osid,
+                                                             f->type, f->op,
+                                                             f->lsm_rule, ctx))
+                                       ++result;
                        }
                        break;
                case AUDIT_ARG0:
@@ -610,8 +648,16 @@ static int audit_filter_rules(struct task_struct *tsk,
                if (!result)
                        return 0;
        }
-       if (rule->filterkey && ctx)
-               ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
+
+       if (ctx) {
+               if (rule->prio <= ctx->prio)
+                       return 0;
+               if (rule->filterkey) {
+                       kfree(ctx->filterkey);
+                       ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
+               }
+               ctx->prio = rule->prio;
+       }
        switch (rule->action) {
        case AUDIT_NEVER:    *state = AUDIT_DISABLED;       break;
        case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
@@ -623,14 +669,17 @@ static int audit_filter_rules(struct task_struct *tsk,
  * completely disabled for this task.  Since we only have the task
  * structure at this point, we can only check uid and gid.
  */
-static enum audit_state audit_filter_task(struct task_struct *tsk)
+static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
 {
        struct audit_entry *e;
        enum audit_state   state;
 
        rcu_read_lock();
        list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
-               if (audit_filter_rules(tsk, &e->rule, NULL, NULL, &state)) {
+               if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
+                                      &state, true)) {
+                       if (state == AUDIT_RECORD_CONTEXT)
+                               *key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
                        rcu_read_unlock();
                        return state;
                }
@@ -662,8 +711,9 @@ static enum audit_state audit_filter_syscall(struct task_struct *tsk,
                list_for_each_entry_rcu(e, list, list) {
                        if ((e->rule.mask[word] & bit) == bit &&
                            audit_filter_rules(tsk, &e->rule, ctx, NULL,
-                                              &state)) {
+                                              &state, false)) {
                                rcu_read_unlock();
+                               ctx->current_state = state;
                                return state;
                        }
                }
@@ -677,15 +727,14 @@ static enum audit_state audit_filter_syscall(struct task_struct *tsk,
  * buckets applicable to the inode numbers in audit_names[].
  * Regarding audit_state, same rules apply as for audit_filter_syscall().
  */
-enum audit_state audit_filter_inodes(struct task_struct *tsk,
-                                    struct audit_context *ctx)
+void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
 {
        int i;
        struct audit_entry *e;
        enum audit_state state;
 
        if (audit_pid && tsk->tgid == audit_pid)
-               return AUDIT_DISABLED;
+               return;
 
        rcu_read_lock();
        for (i = 0; i < ctx->name_count; i++) {
@@ -700,24 +749,20 @@ enum audit_state audit_filter_inodes(struct task_struct *tsk,
 
                list_for_each_entry_rcu(e, list, list) {
                        if ((e->rule.mask[word] & bit) == bit &&
-                           audit_filter_rules(tsk, &e->rule, ctx, n, &state)) {
+                           audit_filter_rules(tsk, &e->rule, ctx, n,
+                                              &state, false)) {
                                rcu_read_unlock();
-                               return state;
+                               ctx->current_state = state;
+                               return;
                        }
                }
        }
        rcu_read_unlock();
-       return AUDIT_BUILD_CONTEXT;
-}
-
-void audit_set_auditable(struct audit_context *ctx)
-{
-       ctx->auditable = 1;
 }
 
 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
                                                      int return_valid,
-                                                     int return_code)
+                                                     long return_code)
 {
        struct audit_context *context = tsk->audit_context;
 
@@ -743,23 +788,11 @@ static inline struct audit_context *audit_get_context(struct task_struct *tsk,
        else
                context->return_code  = return_code;
 
-       if (context->in_syscall && !context->dummy && !context->auditable) {
-               enum audit_state state;
-
-               state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
-               if (state == AUDIT_RECORD_CONTEXT) {
-                       context->auditable = 1;
-                       goto get_context;
-               }
-
-               state = audit_filter_inodes(tsk, context);
-               if (state == AUDIT_RECORD_CONTEXT)
-                       context->auditable = 1;
-
+       if (context->in_syscall && !context->dummy) {
+               audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
+               audit_filter_inodes(tsk, context);
        }
 
-get_context:
-
        tsk->audit_context = NULL;
        return context;
 }
@@ -769,8 +802,7 @@ static inline void audit_free_names(struct audit_context *context)
        int i;
 
 #if AUDIT_DEBUG == 2
-       if (context->auditable
-           ||context->put_count + context->ino_count != context->name_count) {
+       if (context->put_count + context->ino_count != context->name_count) {
                printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
                       " name_count=%d put_count=%d"
                       " ino_count=%d [NOT freeing]\n",
@@ -821,6 +853,7 @@ static inline void audit_zero_context(struct audit_context *context,
 {
        memset(context, 0, sizeof(*context));
        context->state      = state;
+       context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
 }
 
 static inline struct audit_context *audit_alloc_context(enum audit_state state)
@@ -830,6 +863,7 @@ static inline struct audit_context *audit_alloc_context(enum audit_state state)
        if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
                return NULL;
        audit_zero_context(context, state);
+       INIT_LIST_HEAD(&context->killed_trees);
        return context;
 }
 
@@ -846,18 +880,21 @@ int audit_alloc(struct task_struct *tsk)
 {
        struct audit_context *context;
        enum audit_state     state;
+       char *key = NULL;
 
        if (likely(!audit_ever_enabled))
                return 0; /* Return if not auditing. */
 
-       state = audit_filter_task(tsk);
+       state = audit_filter_task(tsk, &key);
        if (likely(state == AUDIT_DISABLED))
                return 0;
 
        if (!(context = audit_alloc_context(state))) {
+               kfree(key);
                audit_log_lost("out of memory in audit_alloc");
                return -ENOMEM;
        }
+       context->filterkey = key;
 
        tsk->audit_context  = context;
        set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
@@ -883,6 +920,7 @@ static inline void audit_free_context(struct audit_context *context)
                free_tree_refs(context);
                audit_free_aux(context);
                kfree(context->filterkey);
+               kfree(context->sockaddr);
                kfree(context);
                context  = previous;
        } while (context);
@@ -980,7 +1018,7 @@ static int audit_log_pid_context(struct audit_context *context, pid_t pid,
 /*
  * to_send and len_sent accounting are very loose estimates.  We aren't
  * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
- * within about 500 bytes (next page boundry)
+ * within about 500 bytes (next page boundary)
  *
  * why snprintf?  an int is up to 12 digits long.  if we just assumed when
  * logging that a[%d]= was going to be 16 characters long we would be wasting
@@ -997,8 +1035,8 @@ static int audit_log_single_execve_arg(struct audit_context *context,
 {
        char arg_num_len_buf[12];
        const char __user *tmp_p = p;
-       /* how many digits are in arg_num? 3 is the length of a=\n */
-       size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 3;
+       /* how many digits are in arg_num? 5 is the length of ' a=""' */
+       size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
        size_t len, len_left, to_send;
        size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
        unsigned int i, has_cntl = 0, too_long = 0;
@@ -1083,7 +1121,7 @@ static int audit_log_single_execve_arg(struct audit_context *context,
                 * so we can be sure nothing was lost.
                 */
                if ((i == 0) && (too_long))
-                       audit_log_format(*ab, "a%d_len=%zu ", arg_num,
+                       audit_log_format(*ab, " a%d_len=%zu", arg_num,
                                         has_cntl ? 2*len : len);
 
                /*
@@ -1103,15 +1141,14 @@ static int audit_log_single_execve_arg(struct audit_context *context,
                buf[to_send] = '\0';
 
                /* actually log it */
-               audit_log_format(*ab, "a%d", arg_num);
+               audit_log_format(*ab, " a%d", arg_num);
                if (too_long)
                        audit_log_format(*ab, "[%d]", i);
                audit_log_format(*ab, "=");
                if (has_cntl)
                        audit_log_n_hex(*ab, buf, to_send);
                else
-                       audit_log_format(*ab, "\"%s\"", buf);
-               audit_log_format(*ab, "\n");
+                       audit_log_string(*ab, buf);
 
                p += to_send;
                len_left -= to_send;
@@ -1139,7 +1176,7 @@ static void audit_log_execve_info(struct audit_context *context,
 
        p = (const char __user *)axi->mm->arg_start;
 
-       audit_log_format(*ab, "argc=%d ", axi->argc);
+       audit_log_format(*ab, "argc=%d", axi->argc);
 
        /*
         * we need some kernel buffer to hold the userspace args.  Just
@@ -1163,8 +1200,133 @@ static void audit_log_execve_info(struct audit_context *context,
        kfree(buf);
 }
 
+static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
+{
+       int i;
+
+       audit_log_format(ab, " %s=", prefix);
+       CAP_FOR_EACH_U32(i) {
+               audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
+       }
+}
+
+static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
+{
+       kernel_cap_t *perm = &name->fcap.permitted;
+       kernel_cap_t *inh = &name->fcap.inheritable;
+       int log = 0;
+
+       if (!cap_isclear(*perm)) {
+               audit_log_cap(ab, "cap_fp", perm);
+               log = 1;
+       }
+       if (!cap_isclear(*inh)) {
+               audit_log_cap(ab, "cap_fi", inh);
+               log = 1;
+       }
+
+       if (log)
+               audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
+}
+
+static void show_special(struct audit_context *context, int *call_panic)
+{
+       struct audit_buffer *ab;
+       int i;
+
+       ab = audit_log_start(context, GFP_KERNEL, context->type);
+       if (!ab)
+               return;
+
+       switch (context->type) {
+       case AUDIT_SOCKETCALL: {
+               int nargs = context->socketcall.nargs;
+               audit_log_format(ab, "nargs=%d", nargs);
+               for (i = 0; i < nargs; i++)
+                       audit_log_format(ab, " a%d=%lx", i,
+                               context->socketcall.args[i]);
+               break; }
+       case AUDIT_IPC: {
+               u32 osid = context->ipc.osid;
+
+               audit_log_format(ab, "ouid=%u ogid=%u mode=%#o",
+                        context->ipc.uid, context->ipc.gid, context->ipc.mode);
+               if (osid) {
+                       char *ctx = NULL;
+                       u32 len;
+                       if (security_secid_to_secctx(osid, &ctx, &len)) {
+                               audit_log_format(ab, " osid=%u", osid);
+                               *call_panic = 1;
+                       } else {
+                               audit_log_format(ab, " obj=%s", ctx);
+                               security_release_secctx(ctx, len);
+                       }
+               }
+               if (context->ipc.has_perm) {
+                       audit_log_end(ab);
+                       ab = audit_log_start(context, GFP_KERNEL,
+                                            AUDIT_IPC_SET_PERM);
+                       audit_log_format(ab,
+                               "qbytes=%lx ouid=%u ogid=%u mode=%#o",
+                               context->ipc.qbytes,
+                               context->ipc.perm_uid,
+                               context->ipc.perm_gid,
+                               context->ipc.perm_mode);
+                       if (!ab)
+                               return;
+               }
+               break; }
+       case AUDIT_MQ_OPEN: {
+               audit_log_format(ab,
+                       "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
+                       "mq_msgsize=%ld mq_curmsgs=%ld",
+                       context->mq_open.oflag, context->mq_open.mode,
+                       context->mq_open.attr.mq_flags,
+                       context->mq_open.attr.mq_maxmsg,
+                       context->mq_open.attr.mq_msgsize,
+                       context->mq_open.attr.mq_curmsgs);
+               break; }
+       case AUDIT_MQ_SENDRECV: {
+               audit_log_format(ab,
+                       "mqdes=%d msg_len=%zd msg_prio=%u "
+                       "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
+                       context->mq_sendrecv.mqdes,
+                       context->mq_sendrecv.msg_len,
+                       context->mq_sendrecv.msg_prio,
+                       context->mq_sendrecv.abs_timeout.tv_sec,
+                       context->mq_sendrecv.abs_timeout.tv_nsec);
+               break; }
+       case AUDIT_MQ_NOTIFY: {
+               audit_log_format(ab, "mqdes=%d sigev_signo=%d",
+                               context->mq_notify.mqdes,
+                               context->mq_notify.sigev_signo);
+               break; }
+       case AUDIT_MQ_GETSETATTR: {
+               struct mq_attr *attr = &context->mq_getsetattr.mqstat;
+               audit_log_format(ab,
+                       "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
+                       "mq_curmsgs=%ld ",
+                       context->mq_getsetattr.mqdes,
+                       attr->mq_flags, attr->mq_maxmsg,
+                       attr->mq_msgsize, attr->mq_curmsgs);
+               break; }
+       case AUDIT_CAPSET: {
+               audit_log_format(ab, "pid=%d", context->capset.pid);
+               audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
+               audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
+               audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
+               break; }
+       case AUDIT_MMAP: {
+               audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
+                                context->mmap.flags);
+               break; }
+       }
+       audit_log_end(ab);
+}
+
 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
 {
+       const struct cred *cred;
        int i, call_panic = 0;
        struct audit_buffer *ab;
        struct audit_aux_data *aux;
@@ -1174,14 +1336,15 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
        context->pid = tsk->pid;
        if (!context->ppid)
                context->ppid = sys_getppid();
-       context->uid = tsk->uid;
-       context->gid = tsk->gid;
-       context->euid = tsk->euid;
-       context->suid = tsk->suid;
-       context->fsuid = tsk->fsuid;
-       context->egid = tsk->egid;
-       context->sgid = tsk->sgid;
-       context->fsgid = tsk->fsgid;
+       cred = current_cred();
+       context->uid   = cred->uid;
+       context->gid   = cred->gid;
+       context->euid  = cred->euid;
+       context->suid  = cred->suid;
+       context->fsuid = cred->fsuid;
+       context->egid  = cred->egid;
+       context->sgid  = cred->sgid;
+       context->fsgid = cred->fsgid;
        context->personality = tsk->personality;
 
        ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
@@ -1196,13 +1359,13 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
                                 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
                                 context->return_code);
 
-       mutex_lock(&tty_mutex);
-       read_lock(&tasklist_lock);
+       spin_lock_irq(&tsk->sighand->siglock);
        if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
                tty = tsk->signal->tty->name;
        else
                tty = "(none)";
-       read_unlock(&tasklist_lock);
+       spin_unlock_irq(&tsk->sighand->siglock);
+
        audit_log_format(ab,
                  " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
                  " ppid=%d pid=%d auid=%u uid=%u gid=%u"
@@ -1222,14 +1385,9 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
                  context->egid, context->sgid, context->fsgid, tty,
                  tsk->sessionid);
 
-       mutex_unlock(&tty_mutex);
 
        audit_log_task_info(ab, tsk);
-       if (context->filterkey) {
-               audit_log_format(ab, " key=");
-               audit_log_untrustedstring(ab, context->filterkey);
-       } else
-               audit_log_format(ab, " key=(null)");
+       audit_log_key(ab, context->filterkey);
        audit_log_end(ab);
 
        for (aux = context->aux; aux; aux = aux->next) {
@@ -1239,96 +1397,50 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
                        continue; /* audit_panic has been called */
 
                switch (aux->type) {
-               case AUDIT_MQ_OPEN: {
-                       struct audit_aux_data_mq_open *axi = (void *)aux;
-                       audit_log_format(ab,
-                               "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
-                               "mq_msgsize=%ld mq_curmsgs=%ld",
-                               axi->oflag, axi->mode, axi->attr.mq_flags,
-                               axi->attr.mq_maxmsg, axi->attr.mq_msgsize,
-                               axi->attr.mq_curmsgs);
-                       break; }
-
-               case AUDIT_MQ_SENDRECV: {
-                       struct audit_aux_data_mq_sendrecv *axi = (void *)aux;
-                       audit_log_format(ab,
-                               "mqdes=%d msg_len=%zd msg_prio=%u "
-                               "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
-                               axi->mqdes, axi->msg_len, axi->msg_prio,
-                               axi->abs_timeout.tv_sec, axi->abs_timeout.tv_nsec);
-                       break; }
-
-               case AUDIT_MQ_NOTIFY: {
-                       struct audit_aux_data_mq_notify *axi = (void *)aux;
-                       audit_log_format(ab,
-                               "mqdes=%d sigev_signo=%d",
-                               axi->mqdes,
-                               axi->notification.sigev_signo);
-                       break; }
-
-               case AUDIT_MQ_GETSETATTR: {
-                       struct audit_aux_data_mq_getsetattr *axi = (void *)aux;
-                       audit_log_format(ab,
-                               "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
-                               "mq_curmsgs=%ld ",
-                               axi->mqdes,
-                               axi->mqstat.mq_flags, axi->mqstat.mq_maxmsg,
-                               axi->mqstat.mq_msgsize, axi->mqstat.mq_curmsgs);
-                       break; }
-
-               case AUDIT_IPC: {
-                       struct audit_aux_data_ipcctl *axi = (void *)aux;
-                       audit_log_format(ab, 
-                                "ouid=%u ogid=%u mode=%#o",
-                                axi->uid, axi->gid, axi->mode);
-                       if (axi->osid != 0) {
-                               char *ctx = NULL;
-                               u32 len;
-                               if (security_secid_to_secctx(
-                                               axi->osid, &ctx, &len)) {
-                                       audit_log_format(ab, " osid=%u",
-                                                       axi->osid);
-                                       call_panic = 1;
-                               } else {
-                                       audit_log_format(ab, " obj=%s", ctx);
-                                       security_release_secctx(ctx, len);
-                               }
-                       }
-                       break; }
-
-               case AUDIT_IPC_SET_PERM: {
-                       struct audit_aux_data_ipcctl *axi = (void *)aux;
-                       audit_log_format(ab,
-                               "qbytes=%lx ouid=%u ogid=%u mode=%#o",
-                               axi->qbytes, axi->uid, axi->gid, axi->mode);
-                       break; }
 
                case AUDIT_EXECVE: {
                        struct audit_aux_data_execve *axi = (void *)aux;
                        audit_log_execve_info(context, &ab, axi);
                        break; }
 
-               case AUDIT_SOCKETCALL: {
-                       struct audit_aux_data_socketcall *axs = (void *)aux;
-                       audit_log_format(ab, "nargs=%d", axs->nargs);
-                       for (i=0; i<axs->nargs; i++)
-                               audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
+               case AUDIT_BPRM_FCAPS: {
+                       struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
+                       audit_log_format(ab, "fver=%x", axs->fcap_ver);
+                       audit_log_cap(ab, "fp", &axs->fcap.permitted);
+                       audit_log_cap(ab, "fi", &axs->fcap.inheritable);
+                       audit_log_format(ab, " fe=%d", axs->fcap.fE);
+                       audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
+                       audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
+                       audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
+                       audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
+                       audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
+                       audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
                        break; }
 
-               case AUDIT_SOCKADDR: {
-                       struct audit_aux_data_sockaddr *axs = (void *)aux;
+               }
+               audit_log_end(ab);
+       }
 
-                       audit_log_format(ab, "saddr=");
-                       audit_log_n_hex(ab, axs->a, axs->len);
-                       break; }
+       if (context->type)
+               show_special(context, &call_panic);
 
-               case AUDIT_FD_PAIR: {
-                       struct audit_aux_data_fd_pair *axs = (void *)aux;
-                       audit_log_format(ab, "fd0=%d fd1=%d", axs->fd[0], axs->fd[1]);
-                       break; }
+       if (context->fds[0] >= 0) {
+               ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
+               if (ab) {
+                       audit_log_format(ab, "fd0=%d fd1=%d",
+                                       context->fds[0], context->fds[1]);
+                       audit_log_end(ab);
+               }
+       }
 
+       if (context->sockaddr_len) {
+               ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
+               if (ab) {
+                       audit_log_format(ab, "saddr=");
+                       audit_log_n_hex(ab, (void *)context->sockaddr,
+                                       context->sockaddr_len);
+                       audit_log_end(ab);
                }
-               audit_log_end(ab);
        }
 
        for (aux = context->aux_pids; aux; aux = aux->next) {
@@ -1377,7 +1489,7 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
                        case 0:
                                /* name was specified as a relative path and the
                                 * directory component is the cwd */
-                               audit_log_d_path(ab, " name=", &context->pwd);
+                               audit_log_d_path(ab, "name=", &context->pwd);
                                break;
                        default:
                                /* log the name's directory component */
@@ -1414,6 +1526,8 @@ static void audit_log_exit(struct audit_context *context, struct task_struct *ts
                        }
                }
 
+               audit_log_fcaps(ab, n);
+
                audit_log_end(ab);
        }
 
@@ -1444,15 +1558,16 @@ void audit_free(struct task_struct *tsk)
         * We use GFP_ATOMIC here because we might be doing this
         * in the context of the idle thread */
        /* that can happen only if we are called from do_exit() */
-       if (context->in_syscall && context->auditable)
+       if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
                audit_log_exit(context, tsk);
+       if (!list_empty(&context->killed_trees))
+               audit_kill_trees(&context->killed_trees);
 
        audit_free_context(context);
 }
 
 /**
  * audit_syscall_entry - fill in an audit record at syscall entry
- * @tsk: task being audited
  * @arch: architecture type
  * @major: major syscall type (function)
  * @a1: additional syscall register 1
@@ -1529,21 +1644,42 @@ void audit_syscall_entry(int arch, int major,
 
        state = context->state;
        context->dummy = !audit_n_rules;
-       if (!context->dummy && (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT))
+       if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
+               context->prio = 0;
                state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
+       }
        if (likely(state == AUDIT_DISABLED))
                return;
 
        context->serial     = 0;
        context->ctime      = CURRENT_TIME;
        context->in_syscall = 1;
-       context->auditable  = !!(state == AUDIT_RECORD_CONTEXT);
+       context->current_state  = state;
        context->ppid       = 0;
 }
 
+void audit_finish_fork(struct task_struct *child)
+{
+       struct audit_context *ctx = current->audit_context;
+       struct audit_context *p = child->audit_context;
+       if (!p || !ctx)
+               return;
+       if (!ctx->in_syscall || ctx->current_state != AUDIT_RECORD_CONTEXT)
+               return;
+       p->arch = ctx->arch;
+       p->major = ctx->major;
+       memcpy(p->argv, ctx->argv, sizeof(ctx->argv));
+       p->ctime = ctx->ctime;
+       p->dummy = ctx->dummy;
+       p->in_syscall = ctx->in_syscall;
+       p->filterkey = kstrdup(ctx->filterkey, GFP_KERNEL);
+       p->ppid = current->pid;
+       p->prio = ctx->prio;
+       p->current_state = ctx->current_state;
+}
+
 /**
  * audit_syscall_exit - deallocate audit context after a system call
- * @tsk: task being audited
  * @valid: success/failure flag
  * @return_code: syscall return value
  *
@@ -1563,11 +1699,14 @@ void audit_syscall_exit(int valid, long return_code)
        if (likely(!context))
                return;
 
-       if (context->in_syscall && context->auditable)
+       if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
                audit_log_exit(context, tsk);
 
        context->in_syscall = 0;
-       context->auditable  = 0;
+       context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
+
+       if (!list_empty(&context->killed_trees))
+               audit_kill_trees(&context->killed_trees);
 
        if (context->previous) {
                struct audit_context *new_context = context->previous;
@@ -1582,8 +1721,13 @@ void audit_syscall_exit(int valid, long return_code)
                context->aux_pids = NULL;
                context->target_pid = 0;
                context->target_sid = 0;
-               kfree(context->filterkey);
-               context->filterkey = NULL;
+               context->sockaddr_len = 0;
+               context->type = 0;
+               context->fds[0] = -1;
+               if (context->state != AUDIT_RECORD_CONTEXT) {
+                       kfree(context->filterkey);
+                       context->filterkey = NULL;
+               }
                tsk->audit_context = context;
        }
 }
@@ -1595,7 +1739,7 @@ static inline void handle_one(const struct inode *inode)
        struct audit_tree_refs *p;
        struct audit_chunk *chunk;
        int count;
-       if (likely(list_empty(&inode->inotify_watches)))
+       if (likely(hlist_empty(&inode->i_fsnotify_marks)))
                return;
        context = current->audit_context;
        p = context->trees;
@@ -1638,7 +1782,7 @@ retry:
        seq = read_seqbegin(&rename_lock);
        for(;;) {
                struct inode *inode = d->d_inode;
-               if (inode && unlikely(!list_empty(&inode->inotify_watches))) {
+               if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
                        struct audit_chunk *chunk;
                        chunk = audit_tree_lookup(inode);
                        if (chunk) {
@@ -1706,13 +1850,8 @@ void __audit_getname(const char *name)
        context->names[context->name_count].ino  = (unsigned long)-1;
        context->names[context->name_count].osid = 0;
        ++context->name_count;
-       if (!context->pwd.dentry) {
-               read_lock(&current->fs->lock);
-               context->pwd = current->fs->pwd;
-               path_get(&current->fs->pwd);
-               read_unlock(&current->fs->lock);
-       }
-
+       if (!context->pwd.dentry)
+               get_fs_pwd(current->fs, &context->pwd);
 }
 
 /* audit_putname - intercept a putname request
@@ -1763,7 +1902,7 @@ static int audit_inc_name_count(struct audit_context *context,
 {
        if (context->name_count >= AUDIT_NAMES) {
                if (inode)
-                       printk(KERN_DEBUG "name_count maxed, losing inode data: "
+                       printk(KERN_DEBUG "audit: name_count maxed, losing inode data: "
                               "dev=%02x:%02x, inode=%lu\n",
                               MAJOR(inode->i_sb->s_dev),
                               MINOR(inode->i_sb->s_dev),
@@ -1780,8 +1919,36 @@ static int audit_inc_name_count(struct audit_context *context,
        return 0;
 }
 
+
+static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
+{
+       struct cpu_vfs_cap_data caps;
+       int rc;
+
+       memset(&name->fcap.permitted, 0, sizeof(kernel_cap_t));
+       memset(&name->fcap.inheritable, 0, sizeof(kernel_cap_t));
+       name->fcap.fE = 0;
+       name->fcap_ver = 0;
+
+       if (!dentry)
+               return 0;
+
+       rc = get_vfs_caps_from_disk(dentry, &caps);
+       if (rc)
+               return rc;
+
+       name->fcap.permitted = caps.permitted;
+       name->fcap.inheritable = caps.inheritable;
+       name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
+       name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
+
+       return 0;
+}
+
+
 /* Copy inode data into an audit_names. */
-static void audit_copy_inode(struct audit_names *name, const struct inode *inode)
+static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
+                            const struct inode *inode)
 {
        name->ino   = inode->i_ino;
        name->dev   = inode->i_sb->s_dev;
@@ -1790,6 +1957,7 @@ static void audit_copy_inode(struct audit_names *name, const struct inode *inode
        name->gid   = inode->i_gid;
        name->rdev  = inode->i_rdev;
        security_inode_getsecid(inode, &name->osid);
+       audit_copy_fcaps(name, dentry);
 }
 
 /**
@@ -1824,12 +1992,11 @@ void __audit_inode(const char *name, const struct dentry *dentry)
                context->names[idx].name = NULL;
        }
        handle_path(dentry);
-       audit_copy_inode(&context->names[idx], inode);
+       audit_copy_inode(&context->names[idx], dentry, inode);
 }
 
 /**
  * audit_inode_child - collect inode info for created/removed objects
- * @dname: inode's dentry name
  * @dentry: dentry being audited
  * @parent: inode of dentry parent
  *
@@ -1841,13 +2008,14 @@ void __audit_inode(const char *name, const struct dentry *dentry)
  * must be hooked prior, in order to capture the target inode during
  * unsuccessful attempts.
  */
-void __audit_inode_child(const char *dname, const struct dentry *dentry,
+void __audit_inode_child(const struct dentry *dentry,
                         const struct inode *parent)
 {
        int idx;
        struct audit_context *context = current->audit_context;
        const char *found_parent = NULL, *found_child = NULL;
        const struct inode *inode = dentry->d_inode;
+       const char *dname = dentry->d_name.name;
        int dirlen = 0;
 
        if (!context->in_syscall)
@@ -1855,9 +2023,6 @@ void __audit_inode_child(const char *dname, const struct dentry *dentry,
 
        if (inode)
                handle_one(inode);
-       /* determine matching parent */
-       if (!dname)
-               goto add_names;
 
        /* parent is more likely, look for it first */
        for (idx = 0; idx < context->name_count; idx++) {
@@ -1885,7 +2050,7 @@ void __audit_inode_child(const char *dname, const struct dentry *dentry,
                if (!strcmp(dname, n->name) ||
                     !audit_compare_dname_path(dname, n->name, &dirlen)) {
                        if (inode)
-                               audit_copy_inode(n, inode);
+                               audit_copy_inode(n, NULL, inode);
                        else
                                n->ino = (unsigned long)-1;
                        found_child = n->name;
@@ -1899,7 +2064,7 @@ add_names:
                        return;
                idx = context->name_count - 1;
                context->names[idx].name = NULL;
-               audit_copy_inode(&context->names[idx], parent);
+               audit_copy_inode(&context->names[idx], NULL, parent);
        }
 
        if (!found_child) {
@@ -1920,7 +2085,7 @@ add_names:
                }
 
                if (inode)
-                       audit_copy_inode(&context->names[idx], inode);
+                       audit_copy_inode(&context->names[idx], NULL, inode);
                else
                        context->names[idx].ino = (unsigned long)-1;
        }
@@ -1935,15 +2100,21 @@ EXPORT_SYMBOL_GPL(__audit_inode_child);
  *
  * Also sets the context as auditable.
  */
-void auditsc_get_stamp(struct audit_context *ctx,
+int auditsc_get_stamp(struct audit_context *ctx,
                       struct timespec *t, unsigned int *serial)
 {
+       if (!ctx->in_syscall)
+               return 0;
        if (!ctx->serial)
                ctx->serial = audit_serial();
        t->tv_sec  = ctx->ctime.tv_sec;
        t->tv_nsec = ctx->ctime.tv_nsec;
        *serial    = ctx->serial;
-       ctx->auditable = 1;
+       if (!ctx->prio) {
+               ctx->prio = 1;
+               ctx->current_state = AUDIT_RECORD_CONTEXT;
+       }
+       return 1;
 }
 
 /* global counter which is incremented every time something logs in */
@@ -1971,7 +2142,7 @@ int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
                        audit_log_format(ab, "login pid=%d uid=%u "
                                "old auid=%u new auid=%u"
                                " old ses=%u new ses=%u",
-                               task->pid, task->uid,
+                               task->pid, task_uid(task),
                                task->loginuid, loginuid,
                                task->sessionid, sessionid);
                        audit_log_end(ab);
@@ -1986,173 +2157,68 @@ int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
  * __audit_mq_open - record audit data for a POSIX MQ open
  * @oflag: open flag
  * @mode: mode bits
- * @u_attr: queue attributes
+ * @attr: queue attributes
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
-int __audit_mq_open(int oflag, mode_t mode, struct mq_attr __user *u_attr)
+void __audit_mq_open(int oflag, mode_t mode, struct mq_attr *attr)
 {
-       struct audit_aux_data_mq_open *ax;
        struct audit_context *context = current->audit_context;
 
-       if (!audit_enabled)
-               return 0;
-
-       if (likely(!context))
-               return 0;
-
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       if (u_attr != NULL) {
-               if (copy_from_user(&ax->attr, u_attr, sizeof(ax->attr))) {
-                       kfree(ax);
-                       return -EFAULT;
-               }
-       } else
-               memset(&ax->attr, 0, sizeof(ax->attr));
+       if (attr)
+               memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
+       else
+               memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
 
-       ax->oflag = oflag;
-       ax->mode = mode;
+       context->mq_open.oflag = oflag;
+       context->mq_open.mode = mode;
 
-       ax->d.type = AUDIT_MQ_OPEN;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->type = AUDIT_MQ_OPEN;
 }
 
 /**
- * __audit_mq_timedsend - record audit data for a POSIX MQ timed send
+ * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
  * @mqdes: MQ descriptor
  * @msg_len: Message length
  * @msg_prio: Message priority
- * @u_abs_timeout: Message timeout in absolute time
+ * @abs_timeout: Message timeout in absolute time
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
-int __audit_mq_timedsend(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
-                       const struct timespec __user *u_abs_timeout)
+void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
+                       const struct timespec *abs_timeout)
 {
-       struct audit_aux_data_mq_sendrecv *ax;
        struct audit_context *context = current->audit_context;
+       struct timespec *p = &context->mq_sendrecv.abs_timeout;
 
-       if (!audit_enabled)
-               return 0;
-
-       if (likely(!context))
-               return 0;
-
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       if (u_abs_timeout != NULL) {
-               if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
-                       kfree(ax);
-                       return -EFAULT;
-               }
-       } else
-               memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
-
-       ax->mqdes = mqdes;
-       ax->msg_len = msg_len;
-       ax->msg_prio = msg_prio;
-
-       ax->d.type = AUDIT_MQ_SENDRECV;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
-}
-
-/**
- * __audit_mq_timedreceive - record audit data for a POSIX MQ timed receive
- * @mqdes: MQ descriptor
- * @msg_len: Message length
- * @u_msg_prio: Message priority
- * @u_abs_timeout: Message timeout in absolute time
- *
- * Returns 0 for success or NULL context or < 0 on error.
- */
-int __audit_mq_timedreceive(mqd_t mqdes, size_t msg_len,
-                               unsigned int __user *u_msg_prio,
-                               const struct timespec __user *u_abs_timeout)
-{
-       struct audit_aux_data_mq_sendrecv *ax;
-       struct audit_context *context = current->audit_context;
-
-       if (!audit_enabled)
-               return 0;
-
-       if (likely(!context))
-               return 0;
-
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       if (u_msg_prio != NULL) {
-               if (get_user(ax->msg_prio, u_msg_prio)) {
-                       kfree(ax);
-                       return -EFAULT;
-               }
-       } else
-               ax->msg_prio = 0;
-
-       if (u_abs_timeout != NULL) {
-               if (copy_from_user(&ax->abs_timeout, u_abs_timeout, sizeof(ax->abs_timeout))) {
-                       kfree(ax);
-                       return -EFAULT;
-               }
-       } else
-               memset(&ax->abs_timeout, 0, sizeof(ax->abs_timeout));
+       if (abs_timeout)
+               memcpy(p, abs_timeout, sizeof(struct timespec));
+       else
+               memset(p, 0, sizeof(struct timespec));
 
-       ax->mqdes = mqdes;
-       ax->msg_len = msg_len;
+       context->mq_sendrecv.mqdes = mqdes;
+       context->mq_sendrecv.msg_len = msg_len;
+       context->mq_sendrecv.msg_prio = msg_prio;
 
-       ax->d.type = AUDIT_MQ_SENDRECV;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->type = AUDIT_MQ_SENDRECV;
 }
 
 /**
  * __audit_mq_notify - record audit data for a POSIX MQ notify
  * @mqdes: MQ descriptor
- * @u_notification: Notification event
+ * @notification: Notification event
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
 
-int __audit_mq_notify(mqd_t mqdes, const struct sigevent __user *u_notification)
+void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
 {
-       struct audit_aux_data_mq_notify *ax;
        struct audit_context *context = current->audit_context;
 
-       if (!audit_enabled)
-               return 0;
-
-       if (likely(!context))
-               return 0;
-
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       if (u_notification != NULL) {
-               if (copy_from_user(&ax->notification, u_notification, sizeof(ax->notification))) {
-                       kfree(ax);
-                       return -EFAULT;
-               }
-       } else
-               memset(&ax->notification, 0, sizeof(ax->notification));
-
-       ax->mqdes = mqdes;
+       if (notification)
+               context->mq_notify.sigev_signo = notification->sigev_signo;
+       else
+               context->mq_notify.sigev_signo = 0;
 
-       ax->d.type = AUDIT_MQ_NOTIFY;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->mq_notify.mqdes = mqdes;
+       context->type = AUDIT_MQ_NOTIFY;
 }
 
 /**
@@ -2160,55 +2226,29 @@ int __audit_mq_notify(mqd_t mqdes, const struct sigevent __user *u_notification)
  * @mqdes: MQ descriptor
  * @mqstat: MQ flags
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
-int __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
+void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
 {
-       struct audit_aux_data_mq_getsetattr *ax;
        struct audit_context *context = current->audit_context;
-
-       if (!audit_enabled)
-               return 0;
-
-       if (likely(!context))
-               return 0;
-
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       ax->mqdes = mqdes;
-       ax->mqstat = *mqstat;
-
-       ax->d.type = AUDIT_MQ_GETSETATTR;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->mq_getsetattr.mqdes = mqdes;
+       context->mq_getsetattr.mqstat = *mqstat;
+       context->type = AUDIT_MQ_GETSETATTR;
 }
 
 /**
  * audit_ipc_obj - record audit data for ipc object
  * @ipcp: ipc permissions
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
-int __audit_ipc_obj(struct kern_ipc_perm *ipcp)
+void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
 {
-       struct audit_aux_data_ipcctl *ax;
        struct audit_context *context = current->audit_context;
-
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       ax->uid = ipcp->uid;
-       ax->gid = ipcp->gid;
-       ax->mode = ipcp->mode;
-       security_ipc_getsecid(ipcp, &ax->osid);
-       ax->d.type = AUDIT_IPC;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->ipc.uid = ipcp->uid;
+       context->ipc.gid = ipcp->gid;
+       context->ipc.mode = ipcp->mode;
+       context->ipc.has_perm = 0;
+       security_ipc_getsecid(ipcp, &context->ipc.osid);
+       context->type = AUDIT_IPC;
 }
 
 /**
@@ -2218,26 +2258,17 @@ int __audit_ipc_obj(struct kern_ipc_perm *ipcp)
  * @gid: msgq group id
  * @mode: msgq mode (permissions)
  *
- * Returns 0 for success or NULL context or < 0 on error.
+ * Called only after audit_ipc_obj().
  */
-int __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
+void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
 {
-       struct audit_aux_data_ipcctl *ax;
        struct audit_context *context = current->audit_context;
 
-       ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
-       if (!ax)
-               return -ENOMEM;
-
-       ax->qbytes = qbytes;
-       ax->uid = uid;
-       ax->gid = gid;
-       ax->mode = mode;
-
-       ax->d.type = AUDIT_IPC_SET_PERM;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->ipc.qbytes = qbytes;
+       context->ipc.perm_uid = uid;
+       context->ipc.perm_gid = gid;
+       context->ipc.perm_mode = mode;
+       context->ipc.has_perm = 1;
 }
 
 int audit_bprm(struct linux_binprm *bprm)
@@ -2267,27 +2298,17 @@ int audit_bprm(struct linux_binprm *bprm)
  * @nargs: number of args
  * @args: args array
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
-int audit_socketcall(int nargs, unsigned long *args)
+void audit_socketcall(int nargs, unsigned long *args)
 {
-       struct audit_aux_data_socketcall *ax;
        struct audit_context *context = current->audit_context;
 
        if (likely(!context || context->dummy))
-               return 0;
-
-       ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
-       if (!ax)
-               return -ENOMEM;
-
-       ax->nargs = nargs;
-       memcpy(ax->args, args, nargs * sizeof(unsigned long));
+               return;
 
-       ax->d.type = AUDIT_SOCKETCALL;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->type = AUDIT_SOCKETCALL;
+       context->socketcall.nargs = nargs;
+       memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
 }
 
 /**
@@ -2295,29 +2316,12 @@ int audit_socketcall(int nargs, unsigned long *args)
  * @fd1: the first file descriptor
  * @fd2: the second file descriptor
  *
- * Returns 0 for success or NULL context or < 0 on error.
  */
-int __audit_fd_pair(int fd1, int fd2)
+void __audit_fd_pair(int fd1, int fd2)
 {
        struct audit_context *context = current->audit_context;
-       struct audit_aux_data_fd_pair *ax;
-
-       if (likely(!context)) {
-               return 0;
-       }
-
-       ax = kmalloc(sizeof(*ax), GFP_KERNEL);
-       if (!ax) {
-               return -ENOMEM;
-       }
-
-       ax->fd[0] = fd1;
-       ax->fd[1] = fd2;
-
-       ax->d.type = AUDIT_FD_PAIR;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
-       return 0;
+       context->fds[0] = fd1;
+       context->fds[1] = fd2;
 }
 
 /**
@@ -2329,22 +2333,20 @@ int __audit_fd_pair(int fd1, int fd2)
  */
 int audit_sockaddr(int len, void *a)
 {
-       struct audit_aux_data_sockaddr *ax;
        struct audit_context *context = current->audit_context;
 
        if (likely(!context || context->dummy))
                return 0;
 
-       ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
-       if (!ax)
-               return -ENOMEM;
-
-       ax->len = len;
-       memcpy(ax->a, a, len);
+       if (!context->sockaddr) {
+               void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
+               if (!p)
+                       return -ENOMEM;
+               context->sockaddr = p;
+       }
 
-       ax->d.type = AUDIT_SOCKADDR;
-       ax->d.next = context->aux;
-       context->aux = (void *)ax;
+       context->sockaddr_len = len;
+       memcpy(context->sockaddr, a, len);
        return 0;
 }
 
@@ -2354,7 +2356,7 @@ void __audit_ptrace(struct task_struct *t)
 
        context->target_pid = t->pid;
        context->target_auid = audit_get_loginuid(t);
-       context->target_uid = t->uid;
+       context->target_uid = task_uid(t);
        context->target_sessionid = audit_get_sessionid(t);
        security_task_getsecid(t, &context->target_sid);
        memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
@@ -2373,6 +2375,7 @@ int __audit_signal_info(int sig, struct task_struct *t)
        struct audit_aux_data_pids *axp;
        struct task_struct *tsk = current;
        struct audit_context *ctx = tsk->audit_context;
+       uid_t uid = current_uid(), t_uid = task_uid(t);
 
        if (audit_pid && t->tgid == audit_pid) {
                if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
@@ -2380,7 +2383,7 @@ int __audit_signal_info(int sig, struct task_struct *t)
                        if (tsk->loginuid != -1)
                                audit_sig_uid = tsk->loginuid;
                        else
-                               audit_sig_uid = tsk->uid;
+                               audit_sig_uid = uid;
                        security_task_getsecid(tsk, &audit_sig_sid);
                }
                if (!audit_signals || audit_dummy_context())
@@ -2392,7 +2395,7 @@ int __audit_signal_info(int sig, struct task_struct *t)
        if (!ctx->target_pid) {
                ctx->target_pid = t->tgid;
                ctx->target_auid = audit_get_loginuid(t);
-               ctx->target_uid = t->uid;
+               ctx->target_uid = t_uid;
                ctx->target_sessionid = audit_get_sessionid(t);
                security_task_getsecid(t, &ctx->target_sid);
                memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
@@ -2413,7 +2416,7 @@ int __audit_signal_info(int sig, struct task_struct *t)
 
        axp->target_pid[axp->pid_count] = t->tgid;
        axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
-       axp->target_uid[axp->pid_count] = t->uid;
+       axp->target_uid[axp->pid_count] = t_uid;
        axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
        security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
        memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
@@ -2423,6 +2426,80 @@ int __audit_signal_info(int sig, struct task_struct *t)
 }
 
 /**
+ * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
+ * @bprm: pointer to the bprm being processed
+ * @new: the proposed new credentials
+ * @old: the old credentials
+ *
+ * Simply check if the proc already has the caps given by the file and if not
+ * store the priv escalation info for later auditing at the end of the syscall
+ *
+ * -Eric
+ */
+int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
+                          const struct cred *new, const struct cred *old)
+{
+       struct audit_aux_data_bprm_fcaps *ax;
+       struct audit_context *context = current->audit_context;
+       struct cpu_vfs_cap_data vcaps;
+       struct dentry *dentry;
+
+       ax = kmalloc(sizeof(*ax), GFP_KERNEL);
+       if (!ax)
+               return -ENOMEM;
+
+       ax->d.type = AUDIT_BPRM_FCAPS;
+       ax->d.next = context->aux;
+       context->aux = (void *)ax;
+
+       dentry = dget(bprm->file->f_dentry);
+       get_vfs_caps_from_disk(dentry, &vcaps);
+       dput(dentry);
+
+       ax->fcap.permitted = vcaps.permitted;
+       ax->fcap.inheritable = vcaps.inheritable;
+       ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
+       ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
+
+       ax->old_pcap.permitted   = old->cap_permitted;
+       ax->old_pcap.inheritable = old->cap_inheritable;
+       ax->old_pcap.effective   = old->cap_effective;
+
+       ax->new_pcap.permitted   = new->cap_permitted;
+       ax->new_pcap.inheritable = new->cap_inheritable;
+       ax->new_pcap.effective   = new->cap_effective;
+       return 0;
+}
+
+/**
+ * __audit_log_capset - store information about the arguments to the capset syscall
+ * @pid: target pid of the capset call
+ * @new: the new credentials
+ * @old: the old (current) credentials
+ *
+ * Record the aguments userspace sent to sys_capset for later printing by the
+ * audit system if applicable
+ */
+void __audit_log_capset(pid_t pid,
+                      const struct cred *new, const struct cred *old)
+{
+       struct audit_context *context = current->audit_context;
+       context->capset.pid = pid;
+       context->capset.cap.effective   = new->cap_effective;
+       context->capset.cap.inheritable = new->cap_effective;
+       context->capset.cap.permitted   = new->cap_permitted;
+       context->type = AUDIT_CAPSET;
+}
+
+void __audit_mmap_fd(int fd, int flags)
+{
+       struct audit_context *context = current->audit_context;
+       context->mmap.fd = fd;
+       context->mmap.flags = flags;
+       context->type = AUDIT_MMAP;
+}
+
+/**
  * audit_core_dumps - record information about processes that end abnormally
  * @signr: signal value
  *
@@ -2433,7 +2510,8 @@ void audit_core_dumps(long signr)
 {
        struct audit_buffer *ab;
        u32 sid;
-       uid_t auid = audit_get_loginuid(current);
+       uid_t auid = audit_get_loginuid(current), uid;
+       gid_t gid;
        unsigned int sessionid = audit_get_sessionid(current);
 
        if (!audit_enabled)
@@ -2443,8 +2521,9 @@ void audit_core_dumps(long signr)
                return;
 
        ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
+       current_uid_gid(&uid, &gid);
        audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
-                       auid, current->uid, current->gid, sessionid);
+                        auid, uid, gid, sessionid);
        security_task_getsecid(current, &sid);
        if (sid) {
                char *ctx = NULL;
@@ -2462,3 +2541,11 @@ void audit_core_dumps(long signr)
        audit_log_format(ab, " sig=%ld", signr);
        audit_log_end(ab);
 }
+
+struct list_head *audit_killed_trees(void)
+{
+       struct audit_context *ctx = current->audit_context;
+       if (likely(!ctx || !ctx->in_syscall))
+               return NULL;
+       return &ctx->killed_trees;
+}