Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / security / smack / smack_lsm.c
diff --git a/kernel/security/smack/smack_lsm.c b/kernel/security/smack/smack_lsm.c
new file mode 100644 (file)
index 0000000..b644757
--- /dev/null
@@ -0,0 +1,4494 @@
+/*
+ *  Simplified MAC Kernel (smack) security module
+ *
+ *  This file contains the smack hook function implementations.
+ *
+ *  Authors:
+ *     Casey Schaufler <casey@schaufler-ca.com>
+ *     Jarkko Sakkinen <jarkko.sakkinen@intel.com>
+ *
+ *  Copyright (C) 2007 Casey Schaufler <casey@schaufler-ca.com>
+ *  Copyright (C) 2009 Hewlett-Packard Development Company, L.P.
+ *                Paul Moore <paul@paul-moore.com>
+ *  Copyright (C) 2010 Nokia Corporation
+ *  Copyright (C) 2011 Intel Corporation.
+ *
+ *     This program is free software; you can redistribute it and/or modify
+ *     it under the terms of the GNU General Public License version 2,
+ *      as published by the Free Software Foundation.
+ */
+
+#include <linux/xattr.h>
+#include <linux/pagemap.h>
+#include <linux/mount.h>
+#include <linux/stat.h>
+#include <linux/kd.h>
+#include <asm/ioctls.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/udp.h>
+#include <linux/dccp.h>
+#include <linux/slab.h>
+#include <linux/mutex.h>
+#include <linux/pipe_fs_i.h>
+#include <net/cipso_ipv4.h>
+#include <net/ip.h>
+#include <net/ipv6.h>
+#include <linux/audit.h>
+#include <linux/magic.h>
+#include <linux/dcache.h>
+#include <linux/personality.h>
+#include <linux/msg.h>
+#include <linux/shm.h>
+#include <linux/binfmts.h>
+#include "smack.h"
+
+#define TRANS_TRUE     "TRUE"
+#define TRANS_TRUE_SIZE        4
+
+#define SMK_CONNECTING 0
+#define SMK_RECEIVING  1
+#define SMK_SENDING    2
+
+#if IS_ENABLED(CONFIG_IPV6) && !defined(CONFIG_SECURITY_SMACK_NETFILTER)
+LIST_HEAD(smk_ipv6_port_list);
+#endif /* CONFIG_IPV6 && !CONFIG_SECURITY_SMACK_NETFILTER */
+static struct kmem_cache *smack_inode_cache;
+int smack_enabled;
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static char *smk_bu_mess[] = {
+       "Bringup Error",        /* Unused */
+       "Bringup",              /* SMACK_BRINGUP_ALLOW */
+       "Unconfined Subject",   /* SMACK_UNCONFINED_SUBJECT */
+       "Unconfined Object",    /* SMACK_UNCONFINED_OBJECT */
+};
+
+static void smk_bu_mode(int mode, char *s)
+{
+       int i = 0;
+
+       if (mode & MAY_READ)
+               s[i++] = 'r';
+       if (mode & MAY_WRITE)
+               s[i++] = 'w';
+       if (mode & MAY_EXEC)
+               s[i++] = 'x';
+       if (mode & MAY_APPEND)
+               s[i++] = 'a';
+       if (mode & MAY_TRANSMUTE)
+               s[i++] = 't';
+       if (mode & MAY_LOCK)
+               s[i++] = 'l';
+       if (i == 0)
+               s[i++] = '-';
+       s[i] = '\0';
+}
+#endif
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static int smk_bu_note(char *note, struct smack_known *sskp,
+                      struct smack_known *oskp, int mode, int rc)
+{
+       char acc[SMK_NUM_ACCESS_TYPE + 1];
+
+       if (rc <= 0)
+               return rc;
+       if (rc > SMACK_UNCONFINED_OBJECT)
+               rc = 0;
+
+       smk_bu_mode(mode, acc);
+       pr_info("Smack %s: (%s %s %s) %s\n", smk_bu_mess[rc],
+               sskp->smk_known, oskp->smk_known, acc, note);
+       return 0;
+}
+#else
+#define smk_bu_note(note, sskp, oskp, mode, RC) (RC)
+#endif
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static int smk_bu_current(char *note, struct smack_known *oskp,
+                         int mode, int rc)
+{
+       struct task_smack *tsp = current_security();
+       char acc[SMK_NUM_ACCESS_TYPE + 1];
+
+       if (rc <= 0)
+               return rc;
+       if (rc > SMACK_UNCONFINED_OBJECT)
+               rc = 0;
+
+       smk_bu_mode(mode, acc);
+       pr_info("Smack %s: (%s %s %s) %s %s\n", smk_bu_mess[rc],
+               tsp->smk_task->smk_known, oskp->smk_known,
+               acc, current->comm, note);
+       return 0;
+}
+#else
+#define smk_bu_current(note, oskp, mode, RC) (RC)
+#endif
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static int smk_bu_task(struct task_struct *otp, int mode, int rc)
+{
+       struct task_smack *tsp = current_security();
+       struct smack_known *smk_task = smk_of_task_struct(otp);
+       char acc[SMK_NUM_ACCESS_TYPE + 1];
+
+       if (rc <= 0)
+               return rc;
+       if (rc > SMACK_UNCONFINED_OBJECT)
+               rc = 0;
+
+       smk_bu_mode(mode, acc);
+       pr_info("Smack %s: (%s %s %s) %s to %s\n", smk_bu_mess[rc],
+               tsp->smk_task->smk_known, smk_task->smk_known, acc,
+               current->comm, otp->comm);
+       return 0;
+}
+#else
+#define smk_bu_task(otp, mode, RC) (RC)
+#endif
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static int smk_bu_inode(struct inode *inode, int mode, int rc)
+{
+       struct task_smack *tsp = current_security();
+       struct inode_smack *isp = inode->i_security;
+       char acc[SMK_NUM_ACCESS_TYPE + 1];
+
+       if (isp->smk_flags & SMK_INODE_IMPURE)
+               pr_info("Smack Unconfined Corruption: inode=(%s %ld) %s\n",
+                       inode->i_sb->s_id, inode->i_ino, current->comm);
+
+       if (rc <= 0)
+               return rc;
+       if (rc > SMACK_UNCONFINED_OBJECT)
+               rc = 0;
+       if (rc == SMACK_UNCONFINED_SUBJECT &&
+           (mode & (MAY_WRITE | MAY_APPEND)))
+               isp->smk_flags |= SMK_INODE_IMPURE;
+
+       smk_bu_mode(mode, acc);
+
+       pr_info("Smack %s: (%s %s %s) inode=(%s %ld) %s\n", smk_bu_mess[rc],
+               tsp->smk_task->smk_known, isp->smk_inode->smk_known, acc,
+               inode->i_sb->s_id, inode->i_ino, current->comm);
+       return 0;
+}
+#else
+#define smk_bu_inode(inode, mode, RC) (RC)
+#endif
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static int smk_bu_file(struct file *file, int mode, int rc)
+{
+       struct task_smack *tsp = current_security();
+       struct smack_known *sskp = tsp->smk_task;
+       struct inode *inode = file_inode(file);
+       struct inode_smack *isp = inode->i_security;
+       char acc[SMK_NUM_ACCESS_TYPE + 1];
+
+       if (isp->smk_flags & SMK_INODE_IMPURE)
+               pr_info("Smack Unconfined Corruption: inode=(%s %ld) %s\n",
+                       inode->i_sb->s_id, inode->i_ino, current->comm);
+
+       if (rc <= 0)
+               return rc;
+       if (rc > SMACK_UNCONFINED_OBJECT)
+               rc = 0;
+
+       smk_bu_mode(mode, acc);
+       pr_info("Smack %s: (%s %s %s) file=(%s %ld %pD) %s\n", smk_bu_mess[rc],
+               sskp->smk_known, smk_of_inode(inode)->smk_known, acc,
+               inode->i_sb->s_id, inode->i_ino, file,
+               current->comm);
+       return 0;
+}
+#else
+#define smk_bu_file(file, mode, RC) (RC)
+#endif
+
+#ifdef CONFIG_SECURITY_SMACK_BRINGUP
+static int smk_bu_credfile(const struct cred *cred, struct file *file,
+                               int mode, int rc)
+{
+       struct task_smack *tsp = cred->security;
+       struct smack_known *sskp = tsp->smk_task;
+       struct inode *inode = file->f_inode;
+       struct inode_smack *isp = inode->i_security;
+       char acc[SMK_NUM_ACCESS_TYPE + 1];
+
+       if (isp->smk_flags & SMK_INODE_IMPURE)
+               pr_info("Smack Unconfined Corruption: inode=(%s %ld) %s\n",
+                       inode->i_sb->s_id, inode->i_ino, current->comm);
+
+       if (rc <= 0)
+               return rc;
+       if (rc > SMACK_UNCONFINED_OBJECT)
+               rc = 0;
+
+       smk_bu_mode(mode, acc);
+       pr_info("Smack %s: (%s %s %s) file=(%s %ld %pD) %s\n", smk_bu_mess[rc],
+               sskp->smk_known, smk_of_inode(inode)->smk_known, acc,
+               inode->i_sb->s_id, inode->i_ino, file,
+               current->comm);
+       return 0;
+}
+#else
+#define smk_bu_credfile(cred, file, mode, RC) (RC)
+#endif
+
+/**
+ * smk_fetch - Fetch the smack label from a file.
+ * @name: type of the label (attribute)
+ * @ip: a pointer to the inode
+ * @dp: a pointer to the dentry
+ *
+ * Returns a pointer to the master list entry for the Smack label
+ * or NULL if there was no label to fetch.
+ */
+static struct smack_known *smk_fetch(const char *name, struct inode *ip,
+                                       struct dentry *dp)
+{
+       int rc;
+       char *buffer;
+       struct smack_known *skp = NULL;
+
+       if (ip->i_op->getxattr == NULL)
+               return NULL;
+
+       buffer = kzalloc(SMK_LONGLABEL, GFP_KERNEL);
+       if (buffer == NULL)
+               return NULL;
+
+       rc = ip->i_op->getxattr(dp, name, buffer, SMK_LONGLABEL);
+       if (rc > 0)
+               skp = smk_import_entry(buffer, rc);
+
+       kfree(buffer);
+
+       return skp;
+}
+
+/**
+ * new_inode_smack - allocate an inode security blob
+ * @skp: a pointer to the Smack label entry to use in the blob
+ *
+ * Returns the new blob or NULL if there's no memory available
+ */
+struct inode_smack *new_inode_smack(struct smack_known *skp)
+{
+       struct inode_smack *isp;
+
+       isp = kmem_cache_zalloc(smack_inode_cache, GFP_NOFS);
+       if (isp == NULL)
+               return NULL;
+
+       isp->smk_inode = skp;
+       isp->smk_flags = 0;
+       mutex_init(&isp->smk_lock);
+
+       return isp;
+}
+
+/**
+ * new_task_smack - allocate a task security blob
+ * @task: a pointer to the Smack label for the running task
+ * @forked: a pointer to the Smack label for the forked task
+ * @gfp: type of the memory for the allocation
+ *
+ * Returns the new blob or NULL if there's no memory available
+ */
+static struct task_smack *new_task_smack(struct smack_known *task,
+                                       struct smack_known *forked, gfp_t gfp)
+{
+       struct task_smack *tsp;
+
+       tsp = kzalloc(sizeof(struct task_smack), gfp);
+       if (tsp == NULL)
+               return NULL;
+
+       tsp->smk_task = task;
+       tsp->smk_forked = forked;
+       INIT_LIST_HEAD(&tsp->smk_rules);
+       mutex_init(&tsp->smk_rules_lock);
+
+       return tsp;
+}
+
+/**
+ * smk_copy_rules - copy a rule set
+ * @nhead: new rules header pointer
+ * @ohead: old rules header pointer
+ * @gfp: type of the memory for the allocation
+ *
+ * Returns 0 on success, -ENOMEM on error
+ */
+static int smk_copy_rules(struct list_head *nhead, struct list_head *ohead,
+                               gfp_t gfp)
+{
+       struct smack_rule *nrp;
+       struct smack_rule *orp;
+       int rc = 0;
+
+       INIT_LIST_HEAD(nhead);
+
+       list_for_each_entry_rcu(orp, ohead, list) {
+               nrp = kzalloc(sizeof(struct smack_rule), gfp);
+               if (nrp == NULL) {
+                       rc = -ENOMEM;
+                       break;
+               }
+               *nrp = *orp;
+               list_add_rcu(&nrp->list, nhead);
+       }
+       return rc;
+}
+
+/**
+ * smk_ptrace_mode - helper function for converting PTRACE_MODE_* into MAY_*
+ * @mode - input mode in form of PTRACE_MODE_*
+ *
+ * Returns a converted MAY_* mode usable by smack rules
+ */
+static inline unsigned int smk_ptrace_mode(unsigned int mode)
+{
+       switch (mode) {
+       case PTRACE_MODE_READ:
+               return MAY_READ;
+       case PTRACE_MODE_ATTACH:
+               return MAY_READWRITE;
+       }
+
+       return 0;
+}
+
+/**
+ * smk_ptrace_rule_check - helper for ptrace access
+ * @tracer: tracer process
+ * @tracee_known: label entry of the process that's about to be traced
+ * @mode: ptrace attachment mode (PTRACE_MODE_*)
+ * @func: name of the function that called us, used for audit
+ *
+ * Returns 0 on access granted, -error on error
+ */
+static int smk_ptrace_rule_check(struct task_struct *tracer,
+                                struct smack_known *tracee_known,
+                                unsigned int mode, const char *func)
+{
+       int rc;
+       struct smk_audit_info ad, *saip = NULL;
+       struct task_smack *tsp;
+       struct smack_known *tracer_known;
+
+       if ((mode & PTRACE_MODE_NOAUDIT) == 0) {
+               smk_ad_init(&ad, func, LSM_AUDIT_DATA_TASK);
+               smk_ad_setfield_u_tsk(&ad, tracer);
+               saip = &ad;
+       }
+
+       rcu_read_lock();
+       tsp = __task_cred(tracer)->security;
+       tracer_known = smk_of_task(tsp);
+
+       if ((mode & PTRACE_MODE_ATTACH) &&
+           (smack_ptrace_rule == SMACK_PTRACE_EXACT ||
+            smack_ptrace_rule == SMACK_PTRACE_DRACONIAN)) {
+               if (tracer_known->smk_known == tracee_known->smk_known)
+                       rc = 0;
+               else if (smack_ptrace_rule == SMACK_PTRACE_DRACONIAN)
+                       rc = -EACCES;
+               else if (capable(CAP_SYS_PTRACE))
+                       rc = 0;
+               else
+                       rc = -EACCES;
+
+               if (saip)
+                       smack_log(tracer_known->smk_known,
+                                 tracee_known->smk_known,
+                                 0, rc, saip);
+
+               rcu_read_unlock();
+               return rc;
+       }
+
+       /* In case of rule==SMACK_PTRACE_DEFAULT or mode==PTRACE_MODE_READ */
+       rc = smk_tskacc(tsp, tracee_known, smk_ptrace_mode(mode), saip);
+
+       rcu_read_unlock();
+       return rc;
+}
+
+/*
+ * LSM hooks.
+ * We he, that is fun!
+ */
+
+/**
+ * smack_ptrace_access_check - Smack approval on PTRACE_ATTACH
+ * @ctp: child task pointer
+ * @mode: ptrace attachment mode (PTRACE_MODE_*)
+ *
+ * Returns 0 if access is OK, an error code otherwise
+ *
+ * Do the capability checks.
+ */
+static int smack_ptrace_access_check(struct task_struct *ctp, unsigned int mode)
+{
+       int rc;
+       struct smack_known *skp;
+
+       rc = cap_ptrace_access_check(ctp, mode);
+       if (rc != 0)
+               return rc;
+
+       skp = smk_of_task_struct(ctp);
+
+       rc = smk_ptrace_rule_check(current, skp, mode, __func__);
+       return rc;
+}
+
+/**
+ * smack_ptrace_traceme - Smack approval on PTRACE_TRACEME
+ * @ptp: parent task pointer
+ *
+ * Returns 0 if access is OK, an error code otherwise
+ *
+ * Do the capability checks, and require PTRACE_MODE_ATTACH.
+ */
+static int smack_ptrace_traceme(struct task_struct *ptp)
+{
+       int rc;
+       struct smack_known *skp;
+
+       rc = cap_ptrace_traceme(ptp);
+       if (rc != 0)
+               return rc;
+
+       skp = smk_of_task(current_security());
+
+       rc = smk_ptrace_rule_check(ptp, skp, PTRACE_MODE_ATTACH, __func__);
+       return rc;
+}
+
+/**
+ * smack_syslog - Smack approval on syslog
+ * @type: message type
+ *
+ * Returns 0 on success, error code otherwise.
+ */
+static int smack_syslog(int typefrom_file)
+{
+       int rc = 0;
+       struct smack_known *skp = smk_of_current();
+
+       if (smack_privileged(CAP_MAC_OVERRIDE))
+               return 0;
+
+       if (smack_syslog_label != NULL && smack_syslog_label != skp)
+               rc = -EACCES;
+
+       return rc;
+}
+
+
+/*
+ * Superblock Hooks.
+ */
+
+/**
+ * smack_sb_alloc_security - allocate a superblock blob
+ * @sb: the superblock getting the blob
+ *
+ * Returns 0 on success or -ENOMEM on error.
+ */
+static int smack_sb_alloc_security(struct super_block *sb)
+{
+       struct superblock_smack *sbsp;
+
+       sbsp = kzalloc(sizeof(struct superblock_smack), GFP_KERNEL);
+
+       if (sbsp == NULL)
+               return -ENOMEM;
+
+       sbsp->smk_root = &smack_known_floor;
+       sbsp->smk_default = &smack_known_floor;
+       sbsp->smk_floor = &smack_known_floor;
+       sbsp->smk_hat = &smack_known_hat;
+       /*
+        * smk_initialized will be zero from kzalloc.
+        */
+       sb->s_security = sbsp;
+
+       return 0;
+}
+
+/**
+ * smack_sb_free_security - free a superblock blob
+ * @sb: the superblock getting the blob
+ *
+ */
+static void smack_sb_free_security(struct super_block *sb)
+{
+       kfree(sb->s_security);
+       sb->s_security = NULL;
+}
+
+/**
+ * smack_sb_copy_data - copy mount options data for processing
+ * @orig: where to start
+ * @smackopts: mount options string
+ *
+ * Returns 0 on success or -ENOMEM on error.
+ *
+ * Copy the Smack specific mount options out of the mount
+ * options list.
+ */
+static int smack_sb_copy_data(char *orig, char *smackopts)
+{
+       char *cp, *commap, *otheropts, *dp;
+
+       otheropts = (char *)get_zeroed_page(GFP_KERNEL);
+       if (otheropts == NULL)
+               return -ENOMEM;
+
+       for (cp = orig, commap = orig; commap != NULL; cp = commap + 1) {
+               if (strstr(cp, SMK_FSDEFAULT) == cp)
+                       dp = smackopts;
+               else if (strstr(cp, SMK_FSFLOOR) == cp)
+                       dp = smackopts;
+               else if (strstr(cp, SMK_FSHAT) == cp)
+                       dp = smackopts;
+               else if (strstr(cp, SMK_FSROOT) == cp)
+                       dp = smackopts;
+               else if (strstr(cp, SMK_FSTRANS) == cp)
+                       dp = smackopts;
+               else
+                       dp = otheropts;
+
+               commap = strchr(cp, ',');
+               if (commap != NULL)
+                       *commap = '\0';
+
+               if (*dp != '\0')
+                       strcat(dp, ",");
+               strcat(dp, cp);
+       }
+
+       strcpy(orig, otheropts);
+       free_page((unsigned long)otheropts);
+
+       return 0;
+}
+
+/**
+ * smack_sb_kern_mount - Smack specific mount processing
+ * @sb: the file system superblock
+ * @flags: the mount flags
+ * @data: the smack mount options
+ *
+ * Returns 0 on success, an error code on failure
+ */
+static int smack_sb_kern_mount(struct super_block *sb, int flags, void *data)
+{
+       struct dentry *root = sb->s_root;
+       struct inode *inode = d_backing_inode(root);
+       struct superblock_smack *sp = sb->s_security;
+       struct inode_smack *isp;
+       struct smack_known *skp;
+       char *op;
+       char *commap;
+       int transmute = 0;
+       int specified = 0;
+
+       if (sp->smk_initialized)
+               return 0;
+
+       sp->smk_initialized = 1;
+
+       for (op = data; op != NULL; op = commap) {
+               commap = strchr(op, ',');
+               if (commap != NULL)
+                       *commap++ = '\0';
+
+               if (strncmp(op, SMK_FSHAT, strlen(SMK_FSHAT)) == 0) {
+                       op += strlen(SMK_FSHAT);
+                       skp = smk_import_entry(op, 0);
+                       if (skp != NULL) {
+                               sp->smk_hat = skp;
+                               specified = 1;
+                       }
+               } else if (strncmp(op, SMK_FSFLOOR, strlen(SMK_FSFLOOR)) == 0) {
+                       op += strlen(SMK_FSFLOOR);
+                       skp = smk_import_entry(op, 0);
+                       if (skp != NULL) {
+                               sp->smk_floor = skp;
+                               specified = 1;
+                       }
+               } else if (strncmp(op, SMK_FSDEFAULT,
+                                  strlen(SMK_FSDEFAULT)) == 0) {
+                       op += strlen(SMK_FSDEFAULT);
+                       skp = smk_import_entry(op, 0);
+                       if (skp != NULL) {
+                               sp->smk_default = skp;
+                               specified = 1;
+                       }
+               } else if (strncmp(op, SMK_FSROOT, strlen(SMK_FSROOT)) == 0) {
+                       op += strlen(SMK_FSROOT);
+                       skp = smk_import_entry(op, 0);
+                       if (skp != NULL) {
+                               sp->smk_root = skp;
+                               specified = 1;
+                       }
+               } else if (strncmp(op, SMK_FSTRANS, strlen(SMK_FSTRANS)) == 0) {
+                       op += strlen(SMK_FSTRANS);
+                       skp = smk_import_entry(op, 0);
+                       if (skp != NULL) {
+                               sp->smk_root = skp;
+                               transmute = 1;
+                               specified = 1;
+                       }
+               }
+       }
+
+       if (!smack_privileged(CAP_MAC_ADMIN)) {
+               /*
+                * Unprivileged mounts don't get to specify Smack values.
+                */
+               if (specified)
+                       return -EPERM;
+               /*
+                * Unprivileged mounts get root and default from the caller.
+                */
+               skp = smk_of_current();
+               sp->smk_root = skp;
+               sp->smk_default = skp;
+       }
+       /*
+        * Initialize the root inode.
+        */
+       isp = inode->i_security;
+       if (isp == NULL) {
+               isp = new_inode_smack(sp->smk_root);
+               if (isp == NULL)
+                       return -ENOMEM;
+               inode->i_security = isp;
+       } else
+               isp->smk_inode = sp->smk_root;
+
+       if (transmute)
+               isp->smk_flags |= SMK_INODE_TRANSMUTE;
+
+       return 0;
+}
+
+/**
+ * smack_sb_statfs - Smack check on statfs
+ * @dentry: identifies the file system in question
+ *
+ * Returns 0 if current can read the floor of the filesystem,
+ * and error code otherwise
+ */
+static int smack_sb_statfs(struct dentry *dentry)
+{
+       struct superblock_smack *sbp = dentry->d_sb->s_security;
+       int rc;
+       struct smk_audit_info ad;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       rc = smk_curacc(sbp->smk_floor, MAY_READ, &ad);
+       rc = smk_bu_current("statfs", sbp->smk_floor, MAY_READ, rc);
+       return rc;
+}
+
+/*
+ * BPRM hooks
+ */
+
+/**
+ * smack_bprm_set_creds - set creds for exec
+ * @bprm: the exec information
+ *
+ * Returns 0 if it gets a blob, -EPERM if exec forbidden and -ENOMEM otherwise
+ */
+static int smack_bprm_set_creds(struct linux_binprm *bprm)
+{
+       struct inode *inode = file_inode(bprm->file);
+       struct task_smack *bsp = bprm->cred->security;
+       struct inode_smack *isp;
+       int rc;
+
+       rc = cap_bprm_set_creds(bprm);
+       if (rc != 0)
+               return rc;
+
+       if (bprm->cred_prepared)
+               return 0;
+
+       isp = inode->i_security;
+       if (isp->smk_task == NULL || isp->smk_task == bsp->smk_task)
+               return 0;
+
+       if (bprm->unsafe & (LSM_UNSAFE_PTRACE | LSM_UNSAFE_PTRACE_CAP)) {
+               struct task_struct *tracer;
+               rc = 0;
+
+               rcu_read_lock();
+               tracer = ptrace_parent(current);
+               if (likely(tracer != NULL))
+                       rc = smk_ptrace_rule_check(tracer,
+                                                  isp->smk_task,
+                                                  PTRACE_MODE_ATTACH,
+                                                  __func__);
+               rcu_read_unlock();
+
+               if (rc != 0)
+                       return rc;
+       } else if (bprm->unsafe)
+               return -EPERM;
+
+       bsp->smk_task = isp->smk_task;
+       bprm->per_clear |= PER_CLEAR_ON_SETID;
+
+       return 0;
+}
+
+/**
+ * smack_bprm_committing_creds - Prepare to install the new credentials
+ * from bprm.
+ *
+ * @bprm: binprm for exec
+ */
+static void smack_bprm_committing_creds(struct linux_binprm *bprm)
+{
+       struct task_smack *bsp = bprm->cred->security;
+
+       if (bsp->smk_task != bsp->smk_forked)
+               current->pdeath_signal = 0;
+}
+
+/**
+ * smack_bprm_secureexec - Return the decision to use secureexec.
+ * @bprm: binprm for exec
+ *
+ * Returns 0 on success.
+ */
+static int smack_bprm_secureexec(struct linux_binprm *bprm)
+{
+       struct task_smack *tsp = current_security();
+       int ret = cap_bprm_secureexec(bprm);
+
+       if (!ret && (tsp->smk_task != tsp->smk_forked))
+               ret = 1;
+
+       return ret;
+}
+
+/*
+ * Inode hooks
+ */
+
+/**
+ * smack_inode_alloc_security - allocate an inode blob
+ * @inode: the inode in need of a blob
+ *
+ * Returns 0 if it gets a blob, -ENOMEM otherwise
+ */
+static int smack_inode_alloc_security(struct inode *inode)
+{
+       struct smack_known *skp = smk_of_current();
+
+       inode->i_security = new_inode_smack(skp);
+       if (inode->i_security == NULL)
+               return -ENOMEM;
+       return 0;
+}
+
+/**
+ * smack_inode_free_security - free an inode blob
+ * @inode: the inode with a blob
+ *
+ * Clears the blob pointer in inode
+ */
+static void smack_inode_free_security(struct inode *inode)
+{
+       kmem_cache_free(smack_inode_cache, inode->i_security);
+       inode->i_security = NULL;
+}
+
+/**
+ * smack_inode_init_security - copy out the smack from an inode
+ * @inode: the newly created inode
+ * @dir: containing directory object
+ * @qstr: unused
+ * @name: where to put the attribute name
+ * @value: where to put the attribute value
+ * @len: where to put the length of the attribute
+ *
+ * Returns 0 if it all works out, -ENOMEM if there's no memory
+ */
+static int smack_inode_init_security(struct inode *inode, struct inode *dir,
+                                    const struct qstr *qstr, const char **name,
+                                    void **value, size_t *len)
+{
+       struct inode_smack *issp = inode->i_security;
+       struct smack_known *skp = smk_of_current();
+       struct smack_known *isp = smk_of_inode(inode);
+       struct smack_known *dsp = smk_of_inode(dir);
+       int may;
+
+       if (name)
+               *name = XATTR_SMACK_SUFFIX;
+
+       if (value && len) {
+               rcu_read_lock();
+               may = smk_access_entry(skp->smk_known, dsp->smk_known,
+                                      &skp->smk_rules);
+               rcu_read_unlock();
+
+               /*
+                * If the access rule allows transmutation and
+                * the directory requests transmutation then
+                * by all means transmute.
+                * Mark the inode as changed.
+                */
+               if (may > 0 && ((may & MAY_TRANSMUTE) != 0) &&
+                   smk_inode_transmutable(dir)) {
+                       isp = dsp;
+                       issp->smk_flags |= SMK_INODE_CHANGED;
+               }
+
+               *value = kstrdup(isp->smk_known, GFP_NOFS);
+               if (*value == NULL)
+                       return -ENOMEM;
+
+               *len = strlen(isp->smk_known);
+       }
+
+       return 0;
+}
+
+/**
+ * smack_inode_link - Smack check on link
+ * @old_dentry: the existing object
+ * @dir: unused
+ * @new_dentry: the new object
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_link(struct dentry *old_dentry, struct inode *dir,
+                           struct dentry *new_dentry)
+{
+       struct smack_known *isp;
+       struct smk_audit_info ad;
+       int rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, old_dentry);
+
+       isp = smk_of_inode(d_backing_inode(old_dentry));
+       rc = smk_curacc(isp, MAY_WRITE, &ad);
+       rc = smk_bu_inode(d_backing_inode(old_dentry), MAY_WRITE, rc);
+
+       if (rc == 0 && d_is_positive(new_dentry)) {
+               isp = smk_of_inode(d_backing_inode(new_dentry));
+               smk_ad_setfield_u_fs_path_dentry(&ad, new_dentry);
+               rc = smk_curacc(isp, MAY_WRITE, &ad);
+               rc = smk_bu_inode(d_backing_inode(new_dentry), MAY_WRITE, rc);
+       }
+
+       return rc;
+}
+
+/**
+ * smack_inode_unlink - Smack check on inode deletion
+ * @dir: containing directory object
+ * @dentry: file to unlink
+ *
+ * Returns 0 if current can write the containing directory
+ * and the object, error code otherwise
+ */
+static int smack_inode_unlink(struct inode *dir, struct dentry *dentry)
+{
+       struct inode *ip = d_backing_inode(dentry);
+       struct smk_audit_info ad;
+       int rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       /*
+        * You need write access to the thing you're unlinking
+        */
+       rc = smk_curacc(smk_of_inode(ip), MAY_WRITE, &ad);
+       rc = smk_bu_inode(ip, MAY_WRITE, rc);
+       if (rc == 0) {
+               /*
+                * You also need write access to the containing directory
+                */
+               smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_INODE);
+               smk_ad_setfield_u_fs_inode(&ad, dir);
+               rc = smk_curacc(smk_of_inode(dir), MAY_WRITE, &ad);
+               rc = smk_bu_inode(dir, MAY_WRITE, rc);
+       }
+       return rc;
+}
+
+/**
+ * smack_inode_rmdir - Smack check on directory deletion
+ * @dir: containing directory object
+ * @dentry: directory to unlink
+ *
+ * Returns 0 if current can write the containing directory
+ * and the directory, error code otherwise
+ */
+static int smack_inode_rmdir(struct inode *dir, struct dentry *dentry)
+{
+       struct smk_audit_info ad;
+       int rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       /*
+        * You need write access to the thing you're removing
+        */
+       rc = smk_curacc(smk_of_inode(d_backing_inode(dentry)), MAY_WRITE, &ad);
+       rc = smk_bu_inode(d_backing_inode(dentry), MAY_WRITE, rc);
+       if (rc == 0) {
+               /*
+                * You also need write access to the containing directory
+                */
+               smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_INODE);
+               smk_ad_setfield_u_fs_inode(&ad, dir);
+               rc = smk_curacc(smk_of_inode(dir), MAY_WRITE, &ad);
+               rc = smk_bu_inode(dir, MAY_WRITE, rc);
+       }
+
+       return rc;
+}
+
+/**
+ * smack_inode_rename - Smack check on rename
+ * @old_inode: unused
+ * @old_dentry: the old object
+ * @new_inode: unused
+ * @new_dentry: the new object
+ *
+ * Read and write access is required on both the old and
+ * new directories.
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_rename(struct inode *old_inode,
+                             struct dentry *old_dentry,
+                             struct inode *new_inode,
+                             struct dentry *new_dentry)
+{
+       int rc;
+       struct smack_known *isp;
+       struct smk_audit_info ad;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, old_dentry);
+
+       isp = smk_of_inode(d_backing_inode(old_dentry));
+       rc = smk_curacc(isp, MAY_READWRITE, &ad);
+       rc = smk_bu_inode(d_backing_inode(old_dentry), MAY_READWRITE, rc);
+
+       if (rc == 0 && d_is_positive(new_dentry)) {
+               isp = smk_of_inode(d_backing_inode(new_dentry));
+               smk_ad_setfield_u_fs_path_dentry(&ad, new_dentry);
+               rc = smk_curacc(isp, MAY_READWRITE, &ad);
+               rc = smk_bu_inode(d_backing_inode(new_dentry), MAY_READWRITE, rc);
+       }
+       return rc;
+}
+
+/**
+ * smack_inode_permission - Smack version of permission()
+ * @inode: the inode in question
+ * @mask: the access requested
+ *
+ * This is the important Smack hook.
+ *
+ * Returns 0 if access is permitted, -EACCES otherwise
+ */
+static int smack_inode_permission(struct inode *inode, int mask)
+{
+       struct smk_audit_info ad;
+       int no_block = mask & MAY_NOT_BLOCK;
+       int rc;
+
+       mask &= (MAY_READ|MAY_WRITE|MAY_EXEC|MAY_APPEND);
+       /*
+        * No permission to check. Existence test. Yup, it's there.
+        */
+       if (mask == 0)
+               return 0;
+
+       /* May be droppable after audit */
+       if (no_block)
+               return -ECHILD;
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_INODE);
+       smk_ad_setfield_u_fs_inode(&ad, inode);
+       rc = smk_curacc(smk_of_inode(inode), mask, &ad);
+       rc = smk_bu_inode(inode, mask, rc);
+       return rc;
+}
+
+/**
+ * smack_inode_setattr - Smack check for setting attributes
+ * @dentry: the object
+ * @iattr: for the force flag
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_setattr(struct dentry *dentry, struct iattr *iattr)
+{
+       struct smk_audit_info ad;
+       int rc;
+
+       /*
+        * Need to allow for clearing the setuid bit.
+        */
+       if (iattr->ia_valid & ATTR_FORCE)
+               return 0;
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       rc = smk_curacc(smk_of_inode(d_backing_inode(dentry)), MAY_WRITE, &ad);
+       rc = smk_bu_inode(d_backing_inode(dentry), MAY_WRITE, rc);
+       return rc;
+}
+
+/**
+ * smack_inode_getattr - Smack check for getting attributes
+ * @mnt: vfsmount of the object
+ * @dentry: the object
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_getattr(const struct path *path)
+{
+       struct smk_audit_info ad;
+       struct inode *inode = d_backing_inode(path->dentry);
+       int rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+       smk_ad_setfield_u_fs_path(&ad, *path);
+       rc = smk_curacc(smk_of_inode(inode), MAY_READ, &ad);
+       rc = smk_bu_inode(inode, MAY_READ, rc);
+       return rc;
+}
+
+/**
+ * smack_inode_setxattr - Smack check for setting xattrs
+ * @dentry: the object
+ * @name: name of the attribute
+ * @value: value of the attribute
+ * @size: size of the value
+ * @flags: unused
+ *
+ * This protects the Smack attribute explicitly.
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_setxattr(struct dentry *dentry, const char *name,
+                               const void *value, size_t size, int flags)
+{
+       struct smk_audit_info ad;
+       struct smack_known *skp;
+       int check_priv = 0;
+       int check_import = 0;
+       int check_star = 0;
+       int rc = 0;
+
+       /*
+        * Check label validity here so import won't fail in post_setxattr
+        */
+       if (strcmp(name, XATTR_NAME_SMACK) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKIPIN) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKIPOUT) == 0) {
+               check_priv = 1;
+               check_import = 1;
+       } else if (strcmp(name, XATTR_NAME_SMACKEXEC) == 0 ||
+                  strcmp(name, XATTR_NAME_SMACKMMAP) == 0) {
+               check_priv = 1;
+               check_import = 1;
+               check_star = 1;
+       } else if (strcmp(name, XATTR_NAME_SMACKTRANSMUTE) == 0) {
+               check_priv = 1;
+               if (size != TRANS_TRUE_SIZE ||
+                   strncmp(value, TRANS_TRUE, TRANS_TRUE_SIZE) != 0)
+                       rc = -EINVAL;
+       } else
+               rc = cap_inode_setxattr(dentry, name, value, size, flags);
+
+       if (check_priv && !smack_privileged(CAP_MAC_ADMIN))
+               rc = -EPERM;
+
+       if (rc == 0 && check_import) {
+               skp = size ? smk_import_entry(value, size) : NULL;
+               if (skp == NULL || (check_star &&
+                   (skp == &smack_known_star || skp == &smack_known_web)))
+                       rc = -EINVAL;
+       }
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       if (rc == 0) {
+               rc = smk_curacc(smk_of_inode(d_backing_inode(dentry)), MAY_WRITE, &ad);
+               rc = smk_bu_inode(d_backing_inode(dentry), MAY_WRITE, rc);
+       }
+
+       return rc;
+}
+
+/**
+ * smack_inode_post_setxattr - Apply the Smack update approved above
+ * @dentry: object
+ * @name: attribute name
+ * @value: attribute value
+ * @size: attribute size
+ * @flags: unused
+ *
+ * Set the pointer in the inode blob to the entry found
+ * in the master label list.
+ */
+static void smack_inode_post_setxattr(struct dentry *dentry, const char *name,
+                                     const void *value, size_t size, int flags)
+{
+       struct smack_known *skp;
+       struct inode_smack *isp = d_backing_inode(dentry)->i_security;
+
+       if (strcmp(name, XATTR_NAME_SMACKTRANSMUTE) == 0) {
+               isp->smk_flags |= SMK_INODE_TRANSMUTE;
+               return;
+       }
+
+       if (strcmp(name, XATTR_NAME_SMACK) == 0) {
+               skp = smk_import_entry(value, size);
+               if (skp != NULL)
+                       isp->smk_inode = skp;
+               else
+                       isp->smk_inode = &smack_known_invalid;
+       } else if (strcmp(name, XATTR_NAME_SMACKEXEC) == 0) {
+               skp = smk_import_entry(value, size);
+               if (skp != NULL)
+                       isp->smk_task = skp;
+               else
+                       isp->smk_task = &smack_known_invalid;
+       } else if (strcmp(name, XATTR_NAME_SMACKMMAP) == 0) {
+               skp = smk_import_entry(value, size);
+               if (skp != NULL)
+                       isp->smk_mmap = skp;
+               else
+                       isp->smk_mmap = &smack_known_invalid;
+       }
+
+       return;
+}
+
+/**
+ * smack_inode_getxattr - Smack check on getxattr
+ * @dentry: the object
+ * @name: unused
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_getxattr(struct dentry *dentry, const char *name)
+{
+       struct smk_audit_info ad;
+       int rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       rc = smk_curacc(smk_of_inode(d_backing_inode(dentry)), MAY_READ, &ad);
+       rc = smk_bu_inode(d_backing_inode(dentry), MAY_READ, rc);
+       return rc;
+}
+
+/**
+ * smack_inode_removexattr - Smack check on removexattr
+ * @dentry: the object
+ * @name: name of the attribute
+ *
+ * Removing the Smack attribute requires CAP_MAC_ADMIN
+ *
+ * Returns 0 if access is permitted, an error code otherwise
+ */
+static int smack_inode_removexattr(struct dentry *dentry, const char *name)
+{
+       struct inode_smack *isp;
+       struct smk_audit_info ad;
+       int rc = 0;
+
+       if (strcmp(name, XATTR_NAME_SMACK) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKIPIN) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKIPOUT) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKEXEC) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKTRANSMUTE) == 0 ||
+           strcmp(name, XATTR_NAME_SMACKMMAP) == 0) {
+               if (!smack_privileged(CAP_MAC_ADMIN))
+                       rc = -EPERM;
+       } else
+               rc = cap_inode_removexattr(dentry, name);
+
+       if (rc != 0)
+               return rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_DENTRY);
+       smk_ad_setfield_u_fs_path_dentry(&ad, dentry);
+
+       rc = smk_curacc(smk_of_inode(d_backing_inode(dentry)), MAY_WRITE, &ad);
+       rc = smk_bu_inode(d_backing_inode(dentry), MAY_WRITE, rc);
+       if (rc != 0)
+               return rc;
+
+       isp = d_backing_inode(dentry)->i_security;
+       /*
+        * Don't do anything special for these.
+        *      XATTR_NAME_SMACKIPIN
+        *      XATTR_NAME_SMACKIPOUT
+        *      XATTR_NAME_SMACKEXEC
+        */
+       if (strcmp(name, XATTR_NAME_SMACK) == 0)
+               isp->smk_task = NULL;
+       else if (strcmp(name, XATTR_NAME_SMACKMMAP) == 0)
+               isp->smk_mmap = NULL;
+       else if (strcmp(name, XATTR_NAME_SMACKTRANSMUTE) == 0)
+               isp->smk_flags &= ~SMK_INODE_TRANSMUTE;
+
+       return 0;
+}
+
+/**
+ * smack_inode_getsecurity - get smack xattrs
+ * @inode: the object
+ * @name: attribute name
+ * @buffer: where to put the result
+ * @alloc: unused
+ *
+ * Returns the size of the attribute or an error code
+ */
+static int smack_inode_getsecurity(const struct inode *inode,
+                                  const char *name, void **buffer,
+                                  bool alloc)
+{
+       struct socket_smack *ssp;
+       struct socket *sock;
+       struct super_block *sbp;
+       struct inode *ip = (struct inode *)inode;
+       struct smack_known *isp;
+       int ilen;
+       int rc = 0;
+
+       if (strcmp(name, XATTR_SMACK_SUFFIX) == 0) {
+               isp = smk_of_inode(inode);
+               ilen = strlen(isp->smk_known);
+               *buffer = isp->smk_known;
+               return ilen;
+       }
+
+       /*
+        * The rest of the Smack xattrs are only on sockets.
+        */
+       sbp = ip->i_sb;
+       if (sbp->s_magic != SOCKFS_MAGIC)
+               return -EOPNOTSUPP;
+
+       sock = SOCKET_I(ip);
+       if (sock == NULL || sock->sk == NULL)
+               return -EOPNOTSUPP;
+
+       ssp = sock->sk->sk_security;
+
+       if (strcmp(name, XATTR_SMACK_IPIN) == 0)
+               isp = ssp->smk_in;
+       else if (strcmp(name, XATTR_SMACK_IPOUT) == 0)
+               isp = ssp->smk_out;
+       else
+               return -EOPNOTSUPP;
+
+       ilen = strlen(isp->smk_known);
+       if (rc == 0) {
+               *buffer = isp->smk_known;
+               rc = ilen;
+       }
+
+       return rc;
+}
+
+
+/**
+ * smack_inode_listsecurity - list the Smack attributes
+ * @inode: the object
+ * @buffer: where they go
+ * @buffer_size: size of buffer
+ *
+ * Returns 0 on success, -EINVAL otherwise
+ */
+static int smack_inode_listsecurity(struct inode *inode, char *buffer,
+                                   size_t buffer_size)
+{
+       int len = sizeof(XATTR_NAME_SMACK);
+
+       if (buffer != NULL && len <= buffer_size)
+               memcpy(buffer, XATTR_NAME_SMACK, len);
+
+       return len;
+}
+
+/**
+ * smack_inode_getsecid - Extract inode's security id
+ * @inode: inode to extract the info from
+ * @secid: where result will be saved
+ */
+static void smack_inode_getsecid(const struct inode *inode, u32 *secid)
+{
+       struct inode_smack *isp = inode->i_security;
+
+       *secid = isp->smk_inode->smk_secid;
+}
+
+/*
+ * File Hooks
+ */
+
+/**
+ * smack_file_permission - Smack check on file operations
+ * @file: unused
+ * @mask: unused
+ *
+ * Returns 0
+ *
+ * Should access checks be done on each read or write?
+ * UNICOS and SELinux say yes.
+ * Trusted Solaris, Trusted Irix, and just about everyone else says no.
+ *
+ * I'll say no for now. Smack does not do the frequent
+ * label changing that SELinux does.
+ */
+static int smack_file_permission(struct file *file, int mask)
+{
+       return 0;
+}
+
+/**
+ * smack_file_alloc_security - assign a file security blob
+ * @file: the object
+ *
+ * The security blob for a file is a pointer to the master
+ * label list, so no allocation is done.
+ *
+ * f_security is the owner security information. It
+ * isn't used on file access checks, it's for send_sigio.
+ *
+ * Returns 0
+ */
+static int smack_file_alloc_security(struct file *file)
+{
+       struct smack_known *skp = smk_of_current();
+
+       file->f_security = skp;
+       return 0;
+}
+
+/**
+ * smack_file_free_security - clear a file security blob
+ * @file: the object
+ *
+ * The security blob for a file is a pointer to the master
+ * label list, so no memory is freed.
+ */
+static void smack_file_free_security(struct file *file)
+{
+       file->f_security = NULL;
+}
+
+/**
+ * smack_file_ioctl - Smack check on ioctls
+ * @file: the object
+ * @cmd: what to do
+ * @arg: unused
+ *
+ * Relies heavily on the correct use of the ioctl command conventions.
+ *
+ * Returns 0 if allowed, error code otherwise
+ */
+static int smack_file_ioctl(struct file *file, unsigned int cmd,
+                           unsigned long arg)
+{
+       int rc = 0;
+       struct smk_audit_info ad;
+       struct inode *inode = file_inode(file);
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+       smk_ad_setfield_u_fs_path(&ad, file->f_path);
+
+       if (_IOC_DIR(cmd) & _IOC_WRITE) {
+               rc = smk_curacc(smk_of_inode(inode), MAY_WRITE, &ad);
+               rc = smk_bu_file(file, MAY_WRITE, rc);
+       }
+
+       if (rc == 0 && (_IOC_DIR(cmd) & _IOC_READ)) {
+               rc = smk_curacc(smk_of_inode(inode), MAY_READ, &ad);
+               rc = smk_bu_file(file, MAY_READ, rc);
+       }
+
+       return rc;
+}
+
+/**
+ * smack_file_lock - Smack check on file locking
+ * @file: the object
+ * @cmd: unused
+ *
+ * Returns 0 if current has lock access, error code otherwise
+ */
+static int smack_file_lock(struct file *file, unsigned int cmd)
+{
+       struct smk_audit_info ad;
+       int rc;
+       struct inode *inode = file_inode(file);
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+       smk_ad_setfield_u_fs_path(&ad, file->f_path);
+       rc = smk_curacc(smk_of_inode(inode), MAY_LOCK, &ad);
+       rc = smk_bu_file(file, MAY_LOCK, rc);
+       return rc;
+}
+
+/**
+ * smack_file_fcntl - Smack check on fcntl
+ * @file: the object
+ * @cmd: what action to check
+ * @arg: unused
+ *
+ * Generally these operations are harmless.
+ * File locking operations present an obvious mechanism
+ * for passing information, so they require write access.
+ *
+ * Returns 0 if current has access, error code otherwise
+ */
+static int smack_file_fcntl(struct file *file, unsigned int cmd,
+                           unsigned long arg)
+{
+       struct smk_audit_info ad;
+       int rc = 0;
+       struct inode *inode = file_inode(file);
+
+       switch (cmd) {
+       case F_GETLK:
+               break;
+       case F_SETLK:
+       case F_SETLKW:
+               smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+               smk_ad_setfield_u_fs_path(&ad, file->f_path);
+               rc = smk_curacc(smk_of_inode(inode), MAY_LOCK, &ad);
+               rc = smk_bu_file(file, MAY_LOCK, rc);
+               break;
+       case F_SETOWN:
+       case F_SETSIG:
+               smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+               smk_ad_setfield_u_fs_path(&ad, file->f_path);
+               rc = smk_curacc(smk_of_inode(inode), MAY_WRITE, &ad);
+               rc = smk_bu_file(file, MAY_WRITE, rc);
+               break;
+       default:
+               break;
+       }
+
+       return rc;
+}
+
+/**
+ * smack_mmap_file :
+ * Check permissions for a mmap operation.  The @file may be NULL, e.g.
+ * if mapping anonymous memory.
+ * @file contains the file structure for file to map (may be NULL).
+ * @reqprot contains the protection requested by the application.
+ * @prot contains the protection that will be applied by the kernel.
+ * @flags contains the operational flags.
+ * Return 0 if permission is granted.
+ */
+static int smack_mmap_file(struct file *file,
+                          unsigned long reqprot, unsigned long prot,
+                          unsigned long flags)
+{
+       struct smack_known *skp;
+       struct smack_known *mkp;
+       struct smack_rule *srp;
+       struct task_smack *tsp;
+       struct smack_known *okp;
+       struct inode_smack *isp;
+       int may;
+       int mmay;
+       int tmay;
+       int rc;
+
+       if (file == NULL)
+               return 0;
+
+       isp = file_inode(file)->i_security;
+       if (isp->smk_mmap == NULL)
+               return 0;
+       mkp = isp->smk_mmap;
+
+       tsp = current_security();
+       skp = smk_of_current();
+       rc = 0;
+
+       rcu_read_lock();
+       /*
+        * For each Smack rule associated with the subject
+        * label verify that the SMACK64MMAP also has access
+        * to that rule's object label.
+        */
+       list_for_each_entry_rcu(srp, &skp->smk_rules, list) {
+               okp = srp->smk_object;
+               /*
+                * Matching labels always allows access.
+                */
+               if (mkp->smk_known == okp->smk_known)
+                       continue;
+               /*
+                * If there is a matching local rule take
+                * that into account as well.
+                */
+               may = smk_access_entry(srp->smk_subject->smk_known,
+                                      okp->smk_known,
+                                      &tsp->smk_rules);
+               if (may == -ENOENT)
+                       may = srp->smk_access;
+               else
+                       may &= srp->smk_access;
+               /*
+                * If may is zero the SMACK64MMAP subject can't
+                * possibly have less access.
+                */
+               if (may == 0)
+                       continue;
+
+               /*
+                * Fetch the global list entry.
+                * If there isn't one a SMACK64MMAP subject
+                * can't have as much access as current.
+                */
+               mmay = smk_access_entry(mkp->smk_known, okp->smk_known,
+                                       &mkp->smk_rules);
+               if (mmay == -ENOENT) {
+                       rc = -EACCES;
+                       break;
+               }
+               /*
+                * If there is a local entry it modifies the
+                * potential access, too.
+                */
+               tmay = smk_access_entry(mkp->smk_known, okp->smk_known,
+                                       &tsp->smk_rules);
+               if (tmay != -ENOENT)
+                       mmay &= tmay;
+
+               /*
+                * If there is any access available to current that is
+                * not available to a SMACK64MMAP subject
+                * deny access.
+                */
+               if ((may | mmay) != mmay) {
+                       rc = -EACCES;
+                       break;
+               }
+       }
+
+       rcu_read_unlock();
+
+       return rc;
+}
+
+/**
+ * smack_file_set_fowner - set the file security blob value
+ * @file: object in question
+ *
+ */
+static void smack_file_set_fowner(struct file *file)
+{
+       file->f_security = smk_of_current();
+}
+
+/**
+ * smack_file_send_sigiotask - Smack on sigio
+ * @tsk: The target task
+ * @fown: the object the signal come from
+ * @signum: unused
+ *
+ * Allow a privileged task to get signals even if it shouldn't
+ *
+ * Returns 0 if a subject with the object's smack could
+ * write to the task, an error code otherwise.
+ */
+static int smack_file_send_sigiotask(struct task_struct *tsk,
+                                    struct fown_struct *fown, int signum)
+{
+       struct smack_known *skp;
+       struct smack_known *tkp = smk_of_task(tsk->cred->security);
+       struct file *file;
+       int rc;
+       struct smk_audit_info ad;
+
+       /*
+        * struct fown_struct is never outside the context of a struct file
+        */
+       file = container_of(fown, struct file, f_owner);
+
+       /* we don't log here as rc can be overriden */
+       skp = file->f_security;
+       rc = smk_access(skp, tkp, MAY_WRITE, NULL);
+       rc = smk_bu_note("sigiotask", skp, tkp, MAY_WRITE, rc);
+       if (rc != 0 && has_capability(tsk, CAP_MAC_OVERRIDE))
+               rc = 0;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_TASK);
+       smk_ad_setfield_u_tsk(&ad, tsk);
+       smack_log(skp->smk_known, tkp->smk_known, MAY_WRITE, rc, &ad);
+       return rc;
+}
+
+/**
+ * smack_file_receive - Smack file receive check
+ * @file: the object
+ *
+ * Returns 0 if current has access, error code otherwise
+ */
+static int smack_file_receive(struct file *file)
+{
+       int rc;
+       int may = 0;
+       struct smk_audit_info ad;
+       struct inode *inode = file_inode(file);
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+       smk_ad_setfield_u_fs_path(&ad, file->f_path);
+       /*
+        * This code relies on bitmasks.
+        */
+       if (file->f_mode & FMODE_READ)
+               may = MAY_READ;
+       if (file->f_mode & FMODE_WRITE)
+               may |= MAY_WRITE;
+
+       rc = smk_curacc(smk_of_inode(inode), may, &ad);
+       rc = smk_bu_file(file, may, rc);
+       return rc;
+}
+
+/**
+ * smack_file_open - Smack dentry open processing
+ * @file: the object
+ * @cred: task credential
+ *
+ * Set the security blob in the file structure.
+ * Allow the open only if the task has read access. There are
+ * many read operations (e.g. fstat) that you can do with an
+ * fd even if you have the file open write-only.
+ *
+ * Returns 0
+ */
+static int smack_file_open(struct file *file, const struct cred *cred)
+{
+       struct task_smack *tsp = cred->security;
+       struct inode *inode = file_inode(file);
+       struct smk_audit_info ad;
+       int rc;
+
+       if (smack_privileged(CAP_MAC_OVERRIDE))
+               return 0;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_PATH);
+       smk_ad_setfield_u_fs_path(&ad, file->f_path);
+       rc = smk_access(tsp->smk_task, smk_of_inode(inode), MAY_READ, &ad);
+       rc = smk_bu_credfile(cred, file, MAY_READ, rc);
+
+       return rc;
+}
+
+/*
+ * Task hooks
+ */
+
+/**
+ * smack_cred_alloc_blank - "allocate" blank task-level security credentials
+ * @new: the new credentials
+ * @gfp: the atomicity of any memory allocations
+ *
+ * Prepare a blank set of credentials for modification.  This must allocate all
+ * the memory the LSM module might require such that cred_transfer() can
+ * complete without error.
+ */
+static int smack_cred_alloc_blank(struct cred *cred, gfp_t gfp)
+{
+       struct task_smack *tsp;
+
+       tsp = new_task_smack(NULL, NULL, gfp);
+       if (tsp == NULL)
+               return -ENOMEM;
+
+       cred->security = tsp;
+
+       return 0;
+}
+
+
+/**
+ * smack_cred_free - "free" task-level security credentials
+ * @cred: the credentials in question
+ *
+ */
+static void smack_cred_free(struct cred *cred)
+{
+       struct task_smack *tsp = cred->security;
+       struct smack_rule *rp;
+       struct list_head *l;
+       struct list_head *n;
+
+       if (tsp == NULL)
+               return;
+       cred->security = NULL;
+
+       list_for_each_safe(l, n, &tsp->smk_rules) {
+               rp = list_entry(l, struct smack_rule, list);
+               list_del(&rp->list);
+               kfree(rp);
+       }
+       kfree(tsp);
+}
+
+/**
+ * smack_cred_prepare - prepare new set of credentials for modification
+ * @new: the new credentials
+ * @old: the original credentials
+ * @gfp: the atomicity of any memory allocations
+ *
+ * Prepare a new set of credentials for modification.
+ */
+static int smack_cred_prepare(struct cred *new, const struct cred *old,
+                             gfp_t gfp)
+{
+       struct task_smack *old_tsp = old->security;
+       struct task_smack *new_tsp;
+       int rc;
+
+       new_tsp = new_task_smack(old_tsp->smk_task, old_tsp->smk_task, gfp);
+       if (new_tsp == NULL)
+               return -ENOMEM;
+
+       rc = smk_copy_rules(&new_tsp->smk_rules, &old_tsp->smk_rules, gfp);
+       if (rc != 0)
+               return rc;
+
+       new->security = new_tsp;
+       return 0;
+}
+
+/**
+ * smack_cred_transfer - Transfer the old credentials to the new credentials
+ * @new: the new credentials
+ * @old: the original credentials
+ *
+ * Fill in a set of blank credentials from another set of credentials.
+ */
+static void smack_cred_transfer(struct cred *new, const struct cred *old)
+{
+       struct task_smack *old_tsp = old->security;
+       struct task_smack *new_tsp = new->security;
+
+       new_tsp->smk_task = old_tsp->smk_task;
+       new_tsp->smk_forked = old_tsp->smk_task;
+       mutex_init(&new_tsp->smk_rules_lock);
+       INIT_LIST_HEAD(&new_tsp->smk_rules);
+
+
+       /* cbs copy rule list */
+}
+
+/**
+ * smack_kernel_act_as - Set the subjective context in a set of credentials
+ * @new: points to the set of credentials to be modified.
+ * @secid: specifies the security ID to be set
+ *
+ * Set the security data for a kernel service.
+ */
+static int smack_kernel_act_as(struct cred *new, u32 secid)
+{
+       struct task_smack *new_tsp = new->security;
+       struct smack_known *skp = smack_from_secid(secid);
+
+       if (skp == NULL)
+               return -EINVAL;
+
+       new_tsp->smk_task = skp;
+       return 0;
+}
+
+/**
+ * smack_kernel_create_files_as - Set the file creation label in a set of creds
+ * @new: points to the set of credentials to be modified
+ * @inode: points to the inode to use as a reference
+ *
+ * Set the file creation context in a set of credentials to the same
+ * as the objective context of the specified inode
+ */
+static int smack_kernel_create_files_as(struct cred *new,
+                                       struct inode *inode)
+{
+       struct inode_smack *isp = inode->i_security;
+       struct task_smack *tsp = new->security;
+
+       tsp->smk_forked = isp->smk_inode;
+       tsp->smk_task = tsp->smk_forked;
+       return 0;
+}
+
+/**
+ * smk_curacc_on_task - helper to log task related access
+ * @p: the task object
+ * @access: the access requested
+ * @caller: name of the calling function for audit
+ *
+ * Return 0 if access is permitted
+ */
+static int smk_curacc_on_task(struct task_struct *p, int access,
+                               const char *caller)
+{
+       struct smk_audit_info ad;
+       struct smack_known *skp = smk_of_task_struct(p);
+       int rc;
+
+       smk_ad_init(&ad, caller, LSM_AUDIT_DATA_TASK);
+       smk_ad_setfield_u_tsk(&ad, p);
+       rc = smk_curacc(skp, access, &ad);
+       rc = smk_bu_task(p, access, rc);
+       return rc;
+}
+
+/**
+ * smack_task_setpgid - Smack check on setting pgid
+ * @p: the task object
+ * @pgid: unused
+ *
+ * Return 0 if write access is permitted
+ */
+static int smack_task_setpgid(struct task_struct *p, pid_t pgid)
+{
+       return smk_curacc_on_task(p, MAY_WRITE, __func__);
+}
+
+/**
+ * smack_task_getpgid - Smack access check for getpgid
+ * @p: the object task
+ *
+ * Returns 0 if current can read the object task, error code otherwise
+ */
+static int smack_task_getpgid(struct task_struct *p)
+{
+       return smk_curacc_on_task(p, MAY_READ, __func__);
+}
+
+/**
+ * smack_task_getsid - Smack access check for getsid
+ * @p: the object task
+ *
+ * Returns 0 if current can read the object task, error code otherwise
+ */
+static int smack_task_getsid(struct task_struct *p)
+{
+       return smk_curacc_on_task(p, MAY_READ, __func__);
+}
+
+/**
+ * smack_task_getsecid - get the secid of the task
+ * @p: the object task
+ * @secid: where to put the result
+ *
+ * Sets the secid to contain a u32 version of the smack label.
+ */
+static void smack_task_getsecid(struct task_struct *p, u32 *secid)
+{
+       struct smack_known *skp = smk_of_task_struct(p);
+
+       *secid = skp->smk_secid;
+}
+
+/**
+ * smack_task_setnice - Smack check on setting nice
+ * @p: the task object
+ * @nice: unused
+ *
+ * Return 0 if write access is permitted
+ */
+static int smack_task_setnice(struct task_struct *p, int nice)
+{
+       int rc;
+
+       rc = cap_task_setnice(p, nice);
+       if (rc == 0)
+               rc = smk_curacc_on_task(p, MAY_WRITE, __func__);
+       return rc;
+}
+
+/**
+ * smack_task_setioprio - Smack check on setting ioprio
+ * @p: the task object
+ * @ioprio: unused
+ *
+ * Return 0 if write access is permitted
+ */
+static int smack_task_setioprio(struct task_struct *p, int ioprio)
+{
+       int rc;
+
+       rc = cap_task_setioprio(p, ioprio);
+       if (rc == 0)
+               rc = smk_curacc_on_task(p, MAY_WRITE, __func__);
+       return rc;
+}
+
+/**
+ * smack_task_getioprio - Smack check on reading ioprio
+ * @p: the task object
+ *
+ * Return 0 if read access is permitted
+ */
+static int smack_task_getioprio(struct task_struct *p)
+{
+       return smk_curacc_on_task(p, MAY_READ, __func__);
+}
+
+/**
+ * smack_task_setscheduler - Smack check on setting scheduler
+ * @p: the task object
+ * @policy: unused
+ * @lp: unused
+ *
+ * Return 0 if read access is permitted
+ */
+static int smack_task_setscheduler(struct task_struct *p)
+{
+       int rc;
+
+       rc = cap_task_setscheduler(p);
+       if (rc == 0)
+               rc = smk_curacc_on_task(p, MAY_WRITE, __func__);
+       return rc;
+}
+
+/**
+ * smack_task_getscheduler - Smack check on reading scheduler
+ * @p: the task object
+ *
+ * Return 0 if read access is permitted
+ */
+static int smack_task_getscheduler(struct task_struct *p)
+{
+       return smk_curacc_on_task(p, MAY_READ, __func__);
+}
+
+/**
+ * smack_task_movememory - Smack check on moving memory
+ * @p: the task object
+ *
+ * Return 0 if write access is permitted
+ */
+static int smack_task_movememory(struct task_struct *p)
+{
+       return smk_curacc_on_task(p, MAY_WRITE, __func__);
+}
+
+/**
+ * smack_task_kill - Smack check on signal delivery
+ * @p: the task object
+ * @info: unused
+ * @sig: unused
+ * @secid: identifies the smack to use in lieu of current's
+ *
+ * Return 0 if write access is permitted
+ *
+ * The secid behavior is an artifact of an SELinux hack
+ * in the USB code. Someday it may go away.
+ */
+static int smack_task_kill(struct task_struct *p, struct siginfo *info,
+                          int sig, u32 secid)
+{
+       struct smk_audit_info ad;
+       struct smack_known *skp;
+       struct smack_known *tkp = smk_of_task_struct(p);
+       int rc;
+
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_TASK);
+       smk_ad_setfield_u_tsk(&ad, p);
+       /*
+        * Sending a signal requires that the sender
+        * can write the receiver.
+        */
+       if (secid == 0) {
+               rc = smk_curacc(tkp, MAY_WRITE, &ad);
+               rc = smk_bu_task(p, MAY_WRITE, rc);
+               return rc;
+       }
+       /*
+        * If the secid isn't 0 we're dealing with some USB IO
+        * specific behavior. This is not clean. For one thing
+        * we can't take privilege into account.
+        */
+       skp = smack_from_secid(secid);
+       rc = smk_access(skp, tkp, MAY_WRITE, &ad);
+       rc = smk_bu_note("USB signal", skp, tkp, MAY_WRITE, rc);
+       return rc;
+}
+
+/**
+ * smack_task_wait - Smack access check for waiting
+ * @p: task to wait for
+ *
+ * Returns 0
+ */
+static int smack_task_wait(struct task_struct *p)
+{
+       /*
+        * Allow the operation to succeed.
+        * Zombies are bad.
+        * In userless environments (e.g. phones) programs
+        * get marked with SMACK64EXEC and even if the parent
+        * and child shouldn't be talking the parent still
+        * may expect to know when the child exits.
+        */
+       return 0;
+}
+
+/**
+ * smack_task_to_inode - copy task smack into the inode blob
+ * @p: task to copy from
+ * @inode: inode to copy to
+ *
+ * Sets the smack pointer in the inode security blob
+ */
+static void smack_task_to_inode(struct task_struct *p, struct inode *inode)
+{
+       struct inode_smack *isp = inode->i_security;
+       struct smack_known *skp = smk_of_task_struct(p);
+
+       isp->smk_inode = skp;
+}
+
+/*
+ * Socket hooks.
+ */
+
+/**
+ * smack_sk_alloc_security - Allocate a socket blob
+ * @sk: the socket
+ * @family: unused
+ * @gfp_flags: memory allocation flags
+ *
+ * Assign Smack pointers to current
+ *
+ * Returns 0 on success, -ENOMEM is there's no memory
+ */
+static int smack_sk_alloc_security(struct sock *sk, int family, gfp_t gfp_flags)
+{
+       struct smack_known *skp = smk_of_current();
+       struct socket_smack *ssp;
+
+       ssp = kzalloc(sizeof(struct socket_smack), gfp_flags);
+       if (ssp == NULL)
+               return -ENOMEM;
+
+       ssp->smk_in = skp;
+       ssp->smk_out = skp;
+       ssp->smk_packet = NULL;
+
+       sk->sk_security = ssp;
+
+       return 0;
+}
+
+/**
+ * smack_sk_free_security - Free a socket blob
+ * @sk: the socket
+ *
+ * Clears the blob pointer
+ */
+static void smack_sk_free_security(struct sock *sk)
+{
+       kfree(sk->sk_security);
+}
+
+/**
+* smack_host_label - check host based restrictions
+* @sip: the object end
+*
+* looks for host based access restrictions
+*
+* This version will only be appropriate for really small sets of single label
+* hosts.  The caller is responsible for ensuring that the RCU read lock is
+* taken before calling this function.
+*
+* Returns the label of the far end or NULL if it's not special.
+*/
+static struct smack_known *smack_host_label(struct sockaddr_in *sip)
+{
+       struct smk_netlbladdr *snp;
+       struct in_addr *siap = &sip->sin_addr;
+
+       if (siap->s_addr == 0)
+               return NULL;
+
+       list_for_each_entry_rcu(snp, &smk_netlbladdr_list, list)
+               /*
+               * we break after finding the first match because
+               * the list is sorted from longest to shortest mask
+               * so we have found the most specific match
+               */
+               if ((&snp->smk_host.sin_addr)->s_addr ==
+                   (siap->s_addr & (&snp->smk_mask)->s_addr)) {
+                       /* we have found the special CIPSO option */
+                       if (snp->smk_label == &smack_cipso_option)
+                               return NULL;
+                       return snp->smk_label;
+               }
+
+       return NULL;
+}
+
+/**
+ * smack_netlabel - Set the secattr on a socket
+ * @sk: the socket
+ * @labeled: socket label scheme
+ *
+ * Convert the outbound smack value (smk_out) to a
+ * secattr and attach it to the socket.
+ *
+ * Returns 0 on success or an error code
+ */
+static int smack_netlabel(struct sock *sk, int labeled)
+{
+       struct smack_known *skp;
+       struct socket_smack *ssp = sk->sk_security;
+       int rc = 0;
+
+       /*
+        * Usually the netlabel code will handle changing the
+        * packet labeling based on the label.
+        * The case of a single label host is different, because
+        * a single label host should never get a labeled packet
+        * even though the label is usually associated with a packet
+        * label.
+        */
+       local_bh_disable();
+       bh_lock_sock_nested(sk);
+
+       if (ssp->smk_out == smack_net_ambient ||
+           labeled == SMACK_UNLABELED_SOCKET)
+               netlbl_sock_delattr(sk);
+       else {
+               skp = ssp->smk_out;
+               rc = netlbl_sock_setattr(sk, sk->sk_family, &skp->smk_netlabel);
+       }
+
+       bh_unlock_sock(sk);
+       local_bh_enable();
+
+       return rc;
+}
+
+/**
+ * smack_netlbel_send - Set the secattr on a socket and perform access checks
+ * @sk: the socket
+ * @sap: the destination address
+ *
+ * Set the correct secattr for the given socket based on the destination
+ * address and perform any outbound access checks needed.
+ *
+ * Returns 0 on success or an error code.
+ *
+ */
+static int smack_netlabel_send(struct sock *sk, struct sockaddr_in *sap)
+{
+       struct smack_known *skp;
+       int rc;
+       int sk_lbl;
+       struct smack_known *hkp;
+       struct socket_smack *ssp = sk->sk_security;
+       struct smk_audit_info ad;
+
+       rcu_read_lock();
+       hkp = smack_host_label(sap);
+       if (hkp != NULL) {
+#ifdef CONFIG_AUDIT
+               struct lsm_network_audit net;
+
+               smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+               ad.a.u.net->family = sap->sin_family;
+               ad.a.u.net->dport = sap->sin_port;
+               ad.a.u.net->v4info.daddr = sap->sin_addr.s_addr;
+#endif
+               sk_lbl = SMACK_UNLABELED_SOCKET;
+               skp = ssp->smk_out;
+               rc = smk_access(skp, hkp, MAY_WRITE, &ad);
+               rc = smk_bu_note("IPv4 host check", skp, hkp, MAY_WRITE, rc);
+       } else {
+               sk_lbl = SMACK_CIPSO_SOCKET;
+               rc = 0;
+       }
+       rcu_read_unlock();
+       if (rc != 0)
+               return rc;
+
+       return smack_netlabel(sk, sk_lbl);
+}
+
+#if IS_ENABLED(CONFIG_IPV6) && !defined(CONFIG_SECURITY_SMACK_NETFILTER)
+/**
+ * smk_ipv6_port_label - Smack port access table management
+ * @sock: socket
+ * @address: address
+ *
+ * Create or update the port list entry
+ */
+static void smk_ipv6_port_label(struct socket *sock, struct sockaddr *address)
+{
+       struct sock *sk = sock->sk;
+       struct sockaddr_in6 *addr6;
+       struct socket_smack *ssp = sock->sk->sk_security;
+       struct smk_port_label *spp;
+       unsigned short port = 0;
+
+       if (address == NULL) {
+               /*
+                * This operation is changing the Smack information
+                * on the bound socket. Take the changes to the port
+                * as well.
+                */
+               list_for_each_entry(spp, &smk_ipv6_port_list, list) {
+                       if (sk != spp->smk_sock)
+                               continue;
+                       spp->smk_in = ssp->smk_in;
+                       spp->smk_out = ssp->smk_out;
+                       return;
+               }
+               /*
+                * A NULL address is only used for updating existing
+                * bound entries. If there isn't one, it's OK.
+                */
+               return;
+       }
+
+       addr6 = (struct sockaddr_in6 *)address;
+       port = ntohs(addr6->sin6_port);
+       /*
+        * This is a special case that is safely ignored.
+        */
+       if (port == 0)
+               return;
+
+       /*
+        * Look for an existing port list entry.
+        * This is an indication that a port is getting reused.
+        */
+       list_for_each_entry(spp, &smk_ipv6_port_list, list) {
+               if (spp->smk_port != port)
+                       continue;
+               spp->smk_port = port;
+               spp->smk_sock = sk;
+               spp->smk_in = ssp->smk_in;
+               spp->smk_out = ssp->smk_out;
+               return;
+       }
+
+       /*
+        * A new port entry is required.
+        */
+       spp = kzalloc(sizeof(*spp), GFP_KERNEL);
+       if (spp == NULL)
+               return;
+
+       spp->smk_port = port;
+       spp->smk_sock = sk;
+       spp->smk_in = ssp->smk_in;
+       spp->smk_out = ssp->smk_out;
+
+       list_add(&spp->list, &smk_ipv6_port_list);
+       return;
+}
+
+/**
+ * smk_ipv6_port_check - check Smack port access
+ * @sock: socket
+ * @address: address
+ *
+ * Create or update the port list entry
+ */
+static int smk_ipv6_port_check(struct sock *sk, struct sockaddr_in6 *address,
+                               int act)
+{
+       __be16 *bep;
+       __be32 *be32p;
+       struct smk_port_label *spp;
+       struct socket_smack *ssp = sk->sk_security;
+       struct smack_known *skp;
+       unsigned short port = 0;
+       struct smack_known *object;
+       struct smk_audit_info ad;
+       int rc;
+#ifdef CONFIG_AUDIT
+       struct lsm_network_audit net;
+#endif
+
+       if (act == SMK_RECEIVING) {
+               skp = smack_net_ambient;
+               object = ssp->smk_in;
+       } else {
+               skp = ssp->smk_out;
+               object = smack_net_ambient;
+       }
+
+       /*
+        * Get the IP address and port from the address.
+        */
+       port = ntohs(address->sin6_port);
+       bep = (__be16 *)(&address->sin6_addr);
+       be32p = (__be32 *)(&address->sin6_addr);
+
+       /*
+        * It's remote, so port lookup does no good.
+        */
+       if (be32p[0] || be32p[1] || be32p[2] || bep[6] || ntohs(bep[7]) != 1)
+               goto auditout;
+
+       /*
+        * It's local so the send check has to have passed.
+        */
+       if (act == SMK_RECEIVING) {
+               skp = &smack_known_web;
+               goto auditout;
+       }
+
+       list_for_each_entry(spp, &smk_ipv6_port_list, list) {
+               if (spp->smk_port != port)
+                       continue;
+               object = spp->smk_in;
+               if (act == SMK_CONNECTING)
+                       ssp->smk_packet = spp->smk_out;
+               break;
+       }
+
+auditout:
+
+#ifdef CONFIG_AUDIT
+       smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+       ad.a.u.net->family = sk->sk_family;
+       ad.a.u.net->dport = port;
+       if (act == SMK_RECEIVING)
+               ad.a.u.net->v6info.saddr = address->sin6_addr;
+       else
+               ad.a.u.net->v6info.daddr = address->sin6_addr;
+#endif
+       rc = smk_access(skp, object, MAY_WRITE, &ad);
+       rc = smk_bu_note("IPv6 port check", skp, object, MAY_WRITE, rc);
+       return rc;
+}
+#endif /* CONFIG_IPV6 && !CONFIG_SECURITY_SMACK_NETFILTER */
+
+/**
+ * smack_inode_setsecurity - set smack xattrs
+ * @inode: the object
+ * @name: attribute name
+ * @value: attribute value
+ * @size: size of the attribute
+ * @flags: unused
+ *
+ * Sets the named attribute in the appropriate blob
+ *
+ * Returns 0 on success, or an error code
+ */
+static int smack_inode_setsecurity(struct inode *inode, const char *name,
+                                  const void *value, size_t size, int flags)
+{
+       struct smack_known *skp;
+       struct inode_smack *nsp = inode->i_security;
+       struct socket_smack *ssp;
+       struct socket *sock;
+       int rc = 0;
+
+       if (value == NULL || size > SMK_LONGLABEL || size == 0)
+               return -EINVAL;
+
+       skp = smk_import_entry(value, size);
+       if (skp == NULL)
+               return -EINVAL;
+
+       if (strcmp(name, XATTR_SMACK_SUFFIX) == 0) {
+               nsp->smk_inode = skp;
+               nsp->smk_flags |= SMK_INODE_INSTANT;
+               return 0;
+       }
+       /*
+        * The rest of the Smack xattrs are only on sockets.
+        */
+       if (inode->i_sb->s_magic != SOCKFS_MAGIC)
+               return -EOPNOTSUPP;
+
+       sock = SOCKET_I(inode);
+       if (sock == NULL || sock->sk == NULL)
+               return -EOPNOTSUPP;
+
+       ssp = sock->sk->sk_security;
+
+       if (strcmp(name, XATTR_SMACK_IPIN) == 0)
+               ssp->smk_in = skp;
+       else if (strcmp(name, XATTR_SMACK_IPOUT) == 0) {
+               ssp->smk_out = skp;
+               if (sock->sk->sk_family == PF_INET) {
+                       rc = smack_netlabel(sock->sk, SMACK_CIPSO_SOCKET);
+                       if (rc != 0)
+                               printk(KERN_WARNING
+                                       "Smack: \"%s\" netlbl error %d.\n",
+                                       __func__, -rc);
+               }
+       } else
+               return -EOPNOTSUPP;
+
+#if IS_ENABLED(CONFIG_IPV6) && !defined(CONFIG_SECURITY_SMACK_NETFILTER)
+       if (sock->sk->sk_family == PF_INET6)
+               smk_ipv6_port_label(sock, NULL);
+#endif /* CONFIG_IPV6 && !CONFIG_SECURITY_SMACK_NETFILTER */
+
+       return 0;
+}
+
+/**
+ * smack_socket_post_create - finish socket setup
+ * @sock: the socket
+ * @family: protocol family
+ * @type: unused
+ * @protocol: unused
+ * @kern: unused
+ *
+ * Sets the netlabel information on the socket
+ *
+ * Returns 0 on success, and error code otherwise
+ */
+static int smack_socket_post_create(struct socket *sock, int family,
+                                   int type, int protocol, int kern)
+{
+       struct socket_smack *ssp;
+
+       if (sock->sk == NULL)
+               return 0;
+
+       /*
+        * Sockets created by kernel threads receive web label.
+        */
+       if (unlikely(current->flags & PF_KTHREAD)) {
+               ssp = sock->sk->sk_security;
+               ssp->smk_in = &smack_known_web;
+               ssp->smk_out = &smack_known_web;
+       }
+
+       if (family != PF_INET)
+               return 0;
+       /*
+        * Set the outbound netlbl.
+        */
+       return smack_netlabel(sock->sk, SMACK_CIPSO_SOCKET);
+}
+
+#ifndef CONFIG_SECURITY_SMACK_NETFILTER
+/**
+ * smack_socket_bind - record port binding information.
+ * @sock: the socket
+ * @address: the port address
+ * @addrlen: size of the address
+ *
+ * Records the label bound to a port.
+ *
+ * Returns 0
+ */
+static int smack_socket_bind(struct socket *sock, struct sockaddr *address,
+                               int addrlen)
+{
+#if IS_ENABLED(CONFIG_IPV6)
+       if (sock->sk != NULL && sock->sk->sk_family == PF_INET6)
+               smk_ipv6_port_label(sock, address);
+#endif
+
+       return 0;
+}
+#endif /* !CONFIG_SECURITY_SMACK_NETFILTER */
+
+/**
+ * smack_socket_connect - connect access check
+ * @sock: the socket
+ * @sap: the other end
+ * @addrlen: size of sap
+ *
+ * Verifies that a connection may be possible
+ *
+ * Returns 0 on success, and error code otherwise
+ */
+static int smack_socket_connect(struct socket *sock, struct sockaddr *sap,
+                               int addrlen)
+{
+       int rc = 0;
+
+       if (sock->sk == NULL)
+               return 0;
+
+       switch (sock->sk->sk_family) {
+       case PF_INET:
+               if (addrlen < sizeof(struct sockaddr_in))
+                       return -EINVAL;
+               rc = smack_netlabel_send(sock->sk, (struct sockaddr_in *)sap);
+               break;
+       case PF_INET6:
+               if (addrlen < sizeof(struct sockaddr_in6))
+                       return -EINVAL;
+#if IS_ENABLED(CONFIG_IPV6) && !defined(CONFIG_SECURITY_SMACK_NETFILTER)
+               rc = smk_ipv6_port_check(sock->sk, (struct sockaddr_in6 *)sap,
+                                               SMK_CONNECTING);
+#endif /* CONFIG_IPV6 && !CONFIG_SECURITY_SMACK_NETFILTER */
+               break;
+       }
+       return rc;
+}
+
+/**
+ * smack_flags_to_may - convert S_ to MAY_ values
+ * @flags: the S_ value
+ *
+ * Returns the equivalent MAY_ value
+ */
+static int smack_flags_to_may(int flags)
+{
+       int may = 0;
+
+       if (flags & S_IRUGO)
+               may |= MAY_READ;
+       if (flags & S_IWUGO)
+               may |= MAY_WRITE;
+       if (flags & S_IXUGO)
+               may |= MAY_EXEC;
+
+       return may;
+}
+
+/**
+ * smack_msg_msg_alloc_security - Set the security blob for msg_msg
+ * @msg: the object
+ *
+ * Returns 0
+ */
+static int smack_msg_msg_alloc_security(struct msg_msg *msg)
+{
+       struct smack_known *skp = smk_of_current();
+
+       msg->security = skp;
+       return 0;
+}
+
+/**
+ * smack_msg_msg_free_security - Clear the security blob for msg_msg
+ * @msg: the object
+ *
+ * Clears the blob pointer
+ */
+static void smack_msg_msg_free_security(struct msg_msg *msg)
+{
+       msg->security = NULL;
+}
+
+/**
+ * smack_of_shm - the smack pointer for the shm
+ * @shp: the object
+ *
+ * Returns a pointer to the smack value
+ */
+static struct smack_known *smack_of_shm(struct shmid_kernel *shp)
+{
+       return (struct smack_known *)shp->shm_perm.security;
+}
+
+/**
+ * smack_shm_alloc_security - Set the security blob for shm
+ * @shp: the object
+ *
+ * Returns 0
+ */
+static int smack_shm_alloc_security(struct shmid_kernel *shp)
+{
+       struct kern_ipc_perm *isp = &shp->shm_perm;
+       struct smack_known *skp = smk_of_current();
+
+       isp->security = skp;
+       return 0;
+}
+
+/**
+ * smack_shm_free_security - Clear the security blob for shm
+ * @shp: the object
+ *
+ * Clears the blob pointer
+ */
+static void smack_shm_free_security(struct shmid_kernel *shp)
+{
+       struct kern_ipc_perm *isp = &shp->shm_perm;
+
+       isp->security = NULL;
+}
+
+/**
+ * smk_curacc_shm : check if current has access on shm
+ * @shp : the object
+ * @access : access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smk_curacc_shm(struct shmid_kernel *shp, int access)
+{
+       struct smack_known *ssp = smack_of_shm(shp);
+       struct smk_audit_info ad;
+       int rc;
+
+#ifdef CONFIG_AUDIT
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_IPC);
+       ad.a.u.ipc_id = shp->shm_perm.id;
+#endif
+       rc = smk_curacc(ssp, access, &ad);
+       rc = smk_bu_current("shm", ssp, access, rc);
+       return rc;
+}
+
+/**
+ * smack_shm_associate - Smack access check for shm
+ * @shp: the object
+ * @shmflg: access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_shm_associate(struct shmid_kernel *shp, int shmflg)
+{
+       int may;
+
+       may = smack_flags_to_may(shmflg);
+       return smk_curacc_shm(shp, may);
+}
+
+/**
+ * smack_shm_shmctl - Smack access check for shm
+ * @shp: the object
+ * @cmd: what it wants to do
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_shm_shmctl(struct shmid_kernel *shp, int cmd)
+{
+       int may;
+
+       switch (cmd) {
+       case IPC_STAT:
+       case SHM_STAT:
+               may = MAY_READ;
+               break;
+       case IPC_SET:
+       case SHM_LOCK:
+       case SHM_UNLOCK:
+       case IPC_RMID:
+               may = MAY_READWRITE;
+               break;
+       case IPC_INFO:
+       case SHM_INFO:
+               /*
+                * System level information.
+                */
+               return 0;
+       default:
+               return -EINVAL;
+       }
+       return smk_curacc_shm(shp, may);
+}
+
+/**
+ * smack_shm_shmat - Smack access for shmat
+ * @shp: the object
+ * @shmaddr: unused
+ * @shmflg: access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_shm_shmat(struct shmid_kernel *shp, char __user *shmaddr,
+                          int shmflg)
+{
+       int may;
+
+       may = smack_flags_to_may(shmflg);
+       return smk_curacc_shm(shp, may);
+}
+
+/**
+ * smack_of_sem - the smack pointer for the sem
+ * @sma: the object
+ *
+ * Returns a pointer to the smack value
+ */
+static struct smack_known *smack_of_sem(struct sem_array *sma)
+{
+       return (struct smack_known *)sma->sem_perm.security;
+}
+
+/**
+ * smack_sem_alloc_security - Set the security blob for sem
+ * @sma: the object
+ *
+ * Returns 0
+ */
+static int smack_sem_alloc_security(struct sem_array *sma)
+{
+       struct kern_ipc_perm *isp = &sma->sem_perm;
+       struct smack_known *skp = smk_of_current();
+
+       isp->security = skp;
+       return 0;
+}
+
+/**
+ * smack_sem_free_security - Clear the security blob for sem
+ * @sma: the object
+ *
+ * Clears the blob pointer
+ */
+static void smack_sem_free_security(struct sem_array *sma)
+{
+       struct kern_ipc_perm *isp = &sma->sem_perm;
+
+       isp->security = NULL;
+}
+
+/**
+ * smk_curacc_sem : check if current has access on sem
+ * @sma : the object
+ * @access : access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smk_curacc_sem(struct sem_array *sma, int access)
+{
+       struct smack_known *ssp = smack_of_sem(sma);
+       struct smk_audit_info ad;
+       int rc;
+
+#ifdef CONFIG_AUDIT
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_IPC);
+       ad.a.u.ipc_id = sma->sem_perm.id;
+#endif
+       rc = smk_curacc(ssp, access, &ad);
+       rc = smk_bu_current("sem", ssp, access, rc);
+       return rc;
+}
+
+/**
+ * smack_sem_associate - Smack access check for sem
+ * @sma: the object
+ * @semflg: access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_sem_associate(struct sem_array *sma, int semflg)
+{
+       int may;
+
+       may = smack_flags_to_may(semflg);
+       return smk_curacc_sem(sma, may);
+}
+
+/**
+ * smack_sem_shmctl - Smack access check for sem
+ * @sma: the object
+ * @cmd: what it wants to do
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_sem_semctl(struct sem_array *sma, int cmd)
+{
+       int may;
+
+       switch (cmd) {
+       case GETPID:
+       case GETNCNT:
+       case GETZCNT:
+       case GETVAL:
+       case GETALL:
+       case IPC_STAT:
+       case SEM_STAT:
+               may = MAY_READ;
+               break;
+       case SETVAL:
+       case SETALL:
+       case IPC_RMID:
+       case IPC_SET:
+               may = MAY_READWRITE;
+               break;
+       case IPC_INFO:
+       case SEM_INFO:
+               /*
+                * System level information
+                */
+               return 0;
+       default:
+               return -EINVAL;
+       }
+
+       return smk_curacc_sem(sma, may);
+}
+
+/**
+ * smack_sem_semop - Smack checks of semaphore operations
+ * @sma: the object
+ * @sops: unused
+ * @nsops: unused
+ * @alter: unused
+ *
+ * Treated as read and write in all cases.
+ *
+ * Returns 0 if access is allowed, error code otherwise
+ */
+static int smack_sem_semop(struct sem_array *sma, struct sembuf *sops,
+                          unsigned nsops, int alter)
+{
+       return smk_curacc_sem(sma, MAY_READWRITE);
+}
+
+/**
+ * smack_msg_alloc_security - Set the security blob for msg
+ * @msq: the object
+ *
+ * Returns 0
+ */
+static int smack_msg_queue_alloc_security(struct msg_queue *msq)
+{
+       struct kern_ipc_perm *kisp = &msq->q_perm;
+       struct smack_known *skp = smk_of_current();
+
+       kisp->security = skp;
+       return 0;
+}
+
+/**
+ * smack_msg_free_security - Clear the security blob for msg
+ * @msq: the object
+ *
+ * Clears the blob pointer
+ */
+static void smack_msg_queue_free_security(struct msg_queue *msq)
+{
+       struct kern_ipc_perm *kisp = &msq->q_perm;
+
+       kisp->security = NULL;
+}
+
+/**
+ * smack_of_msq - the smack pointer for the msq
+ * @msq: the object
+ *
+ * Returns a pointer to the smack label entry
+ */
+static struct smack_known *smack_of_msq(struct msg_queue *msq)
+{
+       return (struct smack_known *)msq->q_perm.security;
+}
+
+/**
+ * smk_curacc_msq : helper to check if current has access on msq
+ * @msq : the msq
+ * @access : access requested
+ *
+ * return 0 if current has access, error otherwise
+ */
+static int smk_curacc_msq(struct msg_queue *msq, int access)
+{
+       struct smack_known *msp = smack_of_msq(msq);
+       struct smk_audit_info ad;
+       int rc;
+
+#ifdef CONFIG_AUDIT
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_IPC);
+       ad.a.u.ipc_id = msq->q_perm.id;
+#endif
+       rc = smk_curacc(msp, access, &ad);
+       rc = smk_bu_current("msq", msp, access, rc);
+       return rc;
+}
+
+/**
+ * smack_msg_queue_associate - Smack access check for msg_queue
+ * @msq: the object
+ * @msqflg: access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_msg_queue_associate(struct msg_queue *msq, int msqflg)
+{
+       int may;
+
+       may = smack_flags_to_may(msqflg);
+       return smk_curacc_msq(msq, may);
+}
+
+/**
+ * smack_msg_queue_msgctl - Smack access check for msg_queue
+ * @msq: the object
+ * @cmd: what it wants to do
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_msg_queue_msgctl(struct msg_queue *msq, int cmd)
+{
+       int may;
+
+       switch (cmd) {
+       case IPC_STAT:
+       case MSG_STAT:
+               may = MAY_READ;
+               break;
+       case IPC_SET:
+       case IPC_RMID:
+               may = MAY_READWRITE;
+               break;
+       case IPC_INFO:
+       case MSG_INFO:
+               /*
+                * System level information
+                */
+               return 0;
+       default:
+               return -EINVAL;
+       }
+
+       return smk_curacc_msq(msq, may);
+}
+
+/**
+ * smack_msg_queue_msgsnd - Smack access check for msg_queue
+ * @msq: the object
+ * @msg: unused
+ * @msqflg: access requested
+ *
+ * Returns 0 if current has the requested access, error code otherwise
+ */
+static int smack_msg_queue_msgsnd(struct msg_queue *msq, struct msg_msg *msg,
+                                 int msqflg)
+{
+       int may;
+
+       may = smack_flags_to_may(msqflg);
+       return smk_curacc_msq(msq, may);
+}
+
+/**
+ * smack_msg_queue_msgsnd - Smack access check for msg_queue
+ * @msq: the object
+ * @msg: unused
+ * @target: unused
+ * @type: unused
+ * @mode: unused
+ *
+ * Returns 0 if current has read and write access, error code otherwise
+ */
+static int smack_msg_queue_msgrcv(struct msg_queue *msq, struct msg_msg *msg,
+                       struct task_struct *target, long type, int mode)
+{
+       return smk_curacc_msq(msq, MAY_READWRITE);
+}
+
+/**
+ * smack_ipc_permission - Smack access for ipc_permission()
+ * @ipp: the object permissions
+ * @flag: access requested
+ *
+ * Returns 0 if current has read and write access, error code otherwise
+ */
+static int smack_ipc_permission(struct kern_ipc_perm *ipp, short flag)
+{
+       struct smack_known *iskp = ipp->security;
+       int may = smack_flags_to_may(flag);
+       struct smk_audit_info ad;
+       int rc;
+
+#ifdef CONFIG_AUDIT
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_IPC);
+       ad.a.u.ipc_id = ipp->id;
+#endif
+       rc = smk_curacc(iskp, may, &ad);
+       rc = smk_bu_current("svipc", iskp, may, rc);
+       return rc;
+}
+
+/**
+ * smack_ipc_getsecid - Extract smack security id
+ * @ipp: the object permissions
+ * @secid: where result will be saved
+ */
+static void smack_ipc_getsecid(struct kern_ipc_perm *ipp, u32 *secid)
+{
+       struct smack_known *iskp = ipp->security;
+
+       *secid = iskp->smk_secid;
+}
+
+/**
+ * smack_d_instantiate - Make sure the blob is correct on an inode
+ * @opt_dentry: dentry where inode will be attached
+ * @inode: the object
+ *
+ * Set the inode's security blob if it hasn't been done already.
+ */
+static void smack_d_instantiate(struct dentry *opt_dentry, struct inode *inode)
+{
+       struct super_block *sbp;
+       struct superblock_smack *sbsp;
+       struct inode_smack *isp;
+       struct smack_known *skp;
+       struct smack_known *ckp = smk_of_current();
+       struct smack_known *final;
+       char trattr[TRANS_TRUE_SIZE];
+       int transflag = 0;
+       int rc;
+       struct dentry *dp;
+
+       if (inode == NULL)
+               return;
+
+       isp = inode->i_security;
+
+       mutex_lock(&isp->smk_lock);
+       /*
+        * If the inode is already instantiated
+        * take the quick way out
+        */
+       if (isp->smk_flags & SMK_INODE_INSTANT)
+               goto unlockandout;
+
+       sbp = inode->i_sb;
+       sbsp = sbp->s_security;
+       /*
+        * We're going to use the superblock default label
+        * if there's no label on the file.
+        */
+       final = sbsp->smk_default;
+
+       /*
+        * If this is the root inode the superblock
+        * may be in the process of initialization.
+        * If that is the case use the root value out
+        * of the superblock.
+        */
+       if (opt_dentry->d_parent == opt_dentry) {
+               switch (sbp->s_magic) {
+               case CGROUP_SUPER_MAGIC:
+                       /*
+                        * The cgroup filesystem is never mounted,
+                        * so there's no opportunity to set the mount
+                        * options.
+                        */
+                       sbsp->smk_root = &smack_known_star;
+                       sbsp->smk_default = &smack_known_star;
+                       isp->smk_inode = sbsp->smk_root;
+                       break;
+               case TMPFS_MAGIC:
+                       /*
+                        * What about shmem/tmpfs anonymous files with dentry
+                        * obtained from d_alloc_pseudo()?
+                        */
+                       isp->smk_inode = smk_of_current();
+                       break;
+               default:
+                       isp->smk_inode = sbsp->smk_root;
+                       break;
+               }
+               isp->smk_flags |= SMK_INODE_INSTANT;
+               goto unlockandout;
+       }
+
+       /*
+        * This is pretty hackish.
+        * Casey says that we shouldn't have to do
+        * file system specific code, but it does help
+        * with keeping it simple.
+        */
+       switch (sbp->s_magic) {
+       case SMACK_MAGIC:
+       case PIPEFS_MAGIC:
+       case SOCKFS_MAGIC:
+       case CGROUP_SUPER_MAGIC:
+               /*
+                * Casey says that it's a little embarrassing
+                * that the smack file system doesn't do
+                * extended attributes.
+                *
+                * Casey says pipes are easy (?)
+                *
+                * Socket access is controlled by the socket
+                * structures associated with the task involved.
+                *
+                * Cgroupfs is special
+                */
+               final = &smack_known_star;
+               break;
+       case DEVPTS_SUPER_MAGIC:
+               /*
+                * devpts seems content with the label of the task.
+                * Programs that change smack have to treat the
+                * pty with respect.
+                */
+               final = ckp;
+               break;
+       case PROC_SUPER_MAGIC:
+               /*
+                * Casey says procfs appears not to care.
+                * The superblock default suffices.
+                */
+               break;
+       case TMPFS_MAGIC:
+               /*
+                * Device labels should come from the filesystem,
+                * but watch out, because they're volitile,
+                * getting recreated on every reboot.
+                */
+               final = &smack_known_star;
+               /*
+                * No break.
+                *
+                * If a smack value has been set we want to use it,
+                * but since tmpfs isn't giving us the opportunity
+                * to set mount options simulate setting the
+                * superblock default.
+                */
+       default:
+               /*
+                * This isn't an understood special case.
+                * Get the value from the xattr.
+                */
+
+               /*
+                * UNIX domain sockets use lower level socket data.
+                */
+               if (S_ISSOCK(inode->i_mode)) {
+                       final = &smack_known_star;
+                       break;
+               }
+               /*
+                * No xattr support means, alas, no SMACK label.
+                * Use the aforeapplied default.
+                * It would be curious if the label of the task
+                * does not match that assigned.
+                */
+               if (inode->i_op->getxattr == NULL)
+                       break;
+               /*
+                * Get the dentry for xattr.
+                */
+               dp = dget(opt_dentry);
+               skp = smk_fetch(XATTR_NAME_SMACK, inode, dp);
+               if (skp != NULL)
+                       final = skp;
+
+               /*
+                * Transmuting directory
+                */
+               if (S_ISDIR(inode->i_mode)) {
+                       /*
+                        * If this is a new directory and the label was
+                        * transmuted when the inode was initialized
+                        * set the transmute attribute on the directory
+                        * and mark the inode.
+                        *
+                        * If there is a transmute attribute on the
+                        * directory mark the inode.
+                        */
+                       if (isp->smk_flags & SMK_INODE_CHANGED) {
+                               isp->smk_flags &= ~SMK_INODE_CHANGED;
+                               rc = inode->i_op->setxattr(dp,
+                                       XATTR_NAME_SMACKTRANSMUTE,
+                                       TRANS_TRUE, TRANS_TRUE_SIZE,
+                                       0);
+                       } else {
+                               rc = inode->i_op->getxattr(dp,
+                                       XATTR_NAME_SMACKTRANSMUTE, trattr,
+                                       TRANS_TRUE_SIZE);
+                               if (rc >= 0 && strncmp(trattr, TRANS_TRUE,
+                                                      TRANS_TRUE_SIZE) != 0)
+                                       rc = -EINVAL;
+                       }
+                       if (rc >= 0)
+                               transflag = SMK_INODE_TRANSMUTE;
+               }
+               /*
+                * Don't let the exec or mmap label be "*" or "@".
+                */
+               skp = smk_fetch(XATTR_NAME_SMACKEXEC, inode, dp);
+               if (skp == &smack_known_star || skp == &smack_known_web)
+                       skp = NULL;
+               isp->smk_task = skp;
+               skp = smk_fetch(XATTR_NAME_SMACKMMAP, inode, dp);
+               if (skp == &smack_known_star || skp == &smack_known_web)
+                       skp = NULL;
+               isp->smk_mmap = skp;
+
+               dput(dp);
+               break;
+       }
+
+       if (final == NULL)
+               isp->smk_inode = ckp;
+       else
+               isp->smk_inode = final;
+
+       isp->smk_flags |= (SMK_INODE_INSTANT | transflag);
+
+unlockandout:
+       mutex_unlock(&isp->smk_lock);
+       return;
+}
+
+/**
+ * smack_getprocattr - Smack process attribute access
+ * @p: the object task
+ * @name: the name of the attribute in /proc/.../attr
+ * @value: where to put the result
+ *
+ * Places a copy of the task Smack into value
+ *
+ * Returns the length of the smack label or an error code
+ */
+static int smack_getprocattr(struct task_struct *p, char *name, char **value)
+{
+       struct smack_known *skp = smk_of_task_struct(p);
+       char *cp;
+       int slen;
+
+       if (strcmp(name, "current") != 0)
+               return -EINVAL;
+
+       cp = kstrdup(skp->smk_known, GFP_KERNEL);
+       if (cp == NULL)
+               return -ENOMEM;
+
+       slen = strlen(cp);
+       *value = cp;
+       return slen;
+}
+
+/**
+ * smack_setprocattr - Smack process attribute setting
+ * @p: the object task
+ * @name: the name of the attribute in /proc/.../attr
+ * @value: the value to set
+ * @size: the size of the value
+ *
+ * Sets the Smack value of the task. Only setting self
+ * is permitted and only with privilege
+ *
+ * Returns the length of the smack label or an error code
+ */
+static int smack_setprocattr(struct task_struct *p, char *name,
+                            void *value, size_t size)
+{
+       struct task_smack *tsp;
+       struct cred *new;
+       struct smack_known *skp;
+
+       /*
+        * Changing another process' Smack value is too dangerous
+        * and supports no sane use case.
+        */
+       if (p != current)
+               return -EPERM;
+
+       if (!smack_privileged(CAP_MAC_ADMIN))
+               return -EPERM;
+
+       if (value == NULL || size == 0 || size >= SMK_LONGLABEL)
+               return -EINVAL;
+
+       if (strcmp(name, "current") != 0)
+               return -EINVAL;
+
+       skp = smk_import_entry(value, size);
+       if (skp == NULL)
+               return -EINVAL;
+
+       /*
+        * No process is ever allowed the web ("@") label.
+        */
+       if (skp == &smack_known_web)
+               return -EPERM;
+
+       new = prepare_creds();
+       if (new == NULL)
+               return -ENOMEM;
+
+       tsp = new->security;
+       tsp->smk_task = skp;
+
+       commit_creds(new);
+       return size;
+}
+
+/**
+ * smack_unix_stream_connect - Smack access on UDS
+ * @sock: one sock
+ * @other: the other sock
+ * @newsk: unused
+ *
+ * Return 0 if a subject with the smack of sock could access
+ * an object with the smack of other, otherwise an error code
+ */
+static int smack_unix_stream_connect(struct sock *sock,
+                                    struct sock *other, struct sock *newsk)
+{
+       struct smack_known *skp;
+       struct smack_known *okp;
+       struct socket_smack *ssp = sock->sk_security;
+       struct socket_smack *osp = other->sk_security;
+       struct socket_smack *nsp = newsk->sk_security;
+       struct smk_audit_info ad;
+       int rc = 0;
+#ifdef CONFIG_AUDIT
+       struct lsm_network_audit net;
+#endif
+
+       if (!smack_privileged(CAP_MAC_OVERRIDE)) {
+               skp = ssp->smk_out;
+               okp = osp->smk_in;
+#ifdef CONFIG_AUDIT
+               smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+               smk_ad_setfield_u_net_sk(&ad, other);
+#endif
+               rc = smk_access(skp, okp, MAY_WRITE, &ad);
+               rc = smk_bu_note("UDS connect", skp, okp, MAY_WRITE, rc);
+               if (rc == 0) {
+                       okp = osp->smk_out;
+                       skp = ssp->smk_in;
+                       rc = smk_access(okp, skp, MAY_WRITE, &ad);
+                       rc = smk_bu_note("UDS connect", okp, skp,
+                                               MAY_WRITE, rc);
+               }
+       }
+
+       /*
+        * Cross reference the peer labels for SO_PEERSEC.
+        */
+       if (rc == 0) {
+               nsp->smk_packet = ssp->smk_out;
+               ssp->smk_packet = osp->smk_out;
+       }
+
+       return rc;
+}
+
+/**
+ * smack_unix_may_send - Smack access on UDS
+ * @sock: one socket
+ * @other: the other socket
+ *
+ * Return 0 if a subject with the smack of sock could access
+ * an object with the smack of other, otherwise an error code
+ */
+static int smack_unix_may_send(struct socket *sock, struct socket *other)
+{
+       struct socket_smack *ssp = sock->sk->sk_security;
+       struct socket_smack *osp = other->sk->sk_security;
+       struct smk_audit_info ad;
+       int rc;
+
+#ifdef CONFIG_AUDIT
+       struct lsm_network_audit net;
+
+       smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+       smk_ad_setfield_u_net_sk(&ad, other->sk);
+#endif
+
+       if (smack_privileged(CAP_MAC_OVERRIDE))
+               return 0;
+
+       rc = smk_access(ssp->smk_out, osp->smk_in, MAY_WRITE, &ad);
+       rc = smk_bu_note("UDS send", ssp->smk_out, osp->smk_in, MAY_WRITE, rc);
+       return rc;
+}
+
+/**
+ * smack_socket_sendmsg - Smack check based on destination host
+ * @sock: the socket
+ * @msg: the message
+ * @size: the size of the message
+ *
+ * Return 0 if the current subject can write to the destination host.
+ * For IPv4 this is only a question if the destination is a single label host.
+ * For IPv6 this is a check against the label of the port.
+ */
+static int smack_socket_sendmsg(struct socket *sock, struct msghdr *msg,
+                               int size)
+{
+       struct sockaddr_in *sip = (struct sockaddr_in *) msg->msg_name;
+#if IS_ENABLED(CONFIG_IPV6) && !defined(CONFIG_SECURITY_SMACK_NETFILTER)
+       struct sockaddr_in6 *sap = (struct sockaddr_in6 *) msg->msg_name;
+#endif /* CONFIG_IPV6 && !CONFIG_SECURITY_SMACK_NETFILTER */
+       int rc = 0;
+
+       /*
+        * Perfectly reasonable for this to be NULL
+        */
+       if (sip == NULL)
+               return 0;
+
+       switch (sip->sin_family) {
+       case AF_INET:
+               rc = smack_netlabel_send(sock->sk, sip);
+               break;
+       case AF_INET6:
+#if IS_ENABLED(CONFIG_IPV6) && !defined(CONFIG_SECURITY_SMACK_NETFILTER)
+               rc = smk_ipv6_port_check(sock->sk, sap, SMK_SENDING);
+#endif /* CONFIG_IPV6 && !CONFIG_SECURITY_SMACK_NETFILTER */
+               break;
+       }
+       return rc;
+}
+
+/**
+ * smack_from_secattr - Convert a netlabel attr.mls.lvl/attr.mls.cat pair to smack
+ * @sap: netlabel secattr
+ * @ssp: socket security information
+ *
+ * Returns a pointer to a Smack label entry found on the label list.
+ */
+static struct smack_known *smack_from_secattr(struct netlbl_lsm_secattr *sap,
+                                               struct socket_smack *ssp)
+{
+       struct smack_known *skp;
+       int found = 0;
+       int acat;
+       int kcat;
+
+       if ((sap->flags & NETLBL_SECATTR_MLS_LVL) != 0) {
+               /*
+                * Looks like a CIPSO packet.
+                * If there are flags but no level netlabel isn't
+                * behaving the way we expect it to.
+                *
+                * Look it up in the label table
+                * Without guidance regarding the smack value
+                * for the packet fall back on the network
+                * ambient value.
+                */
+               rcu_read_lock();
+               list_for_each_entry(skp, &smack_known_list, list) {
+                       if (sap->attr.mls.lvl != skp->smk_netlabel.attr.mls.lvl)
+                               continue;
+                       /*
+                        * Compare the catsets. Use the netlbl APIs.
+                        */
+                       if ((sap->flags & NETLBL_SECATTR_MLS_CAT) == 0) {
+                               if ((skp->smk_netlabel.flags &
+                                    NETLBL_SECATTR_MLS_CAT) == 0)
+                                       found = 1;
+                               break;
+                       }
+                       for (acat = -1, kcat = -1; acat == kcat; ) {
+                               acat = netlbl_catmap_walk(sap->attr.mls.cat,
+                                                         acat + 1);
+                               kcat = netlbl_catmap_walk(
+                                       skp->smk_netlabel.attr.mls.cat,
+                                       kcat + 1);
+                               if (acat < 0 || kcat < 0)
+                                       break;
+                       }
+                       if (acat == kcat) {
+                               found = 1;
+                               break;
+                       }
+               }
+               rcu_read_unlock();
+
+               if (found)
+                       return skp;
+
+               if (ssp != NULL && ssp->smk_in == &smack_known_star)
+                       return &smack_known_web;
+               return &smack_known_star;
+       }
+       if ((sap->flags & NETLBL_SECATTR_SECID) != 0) {
+               /*
+                * Looks like a fallback, which gives us a secid.
+                */
+               skp = smack_from_secid(sap->attr.secid);
+               /*
+                * This has got to be a bug because it is
+                * impossible to specify a fallback without
+                * specifying the label, which will ensure
+                * it has a secid, and the only way to get a
+                * secid is from a fallback.
+                */
+               BUG_ON(skp == NULL);
+               return skp;
+       }
+       /*
+        * Without guidance regarding the smack value
+        * for the packet fall back on the network
+        * ambient value.
+        */
+       return smack_net_ambient;
+}
+
+#if IS_ENABLED(CONFIG_IPV6)
+static int smk_skb_to_addr_ipv6(struct sk_buff *skb, struct sockaddr_in6 *sip)
+{
+       u8 nexthdr;
+       int offset;
+       int proto = -EINVAL;
+       struct ipv6hdr _ipv6h;
+       struct ipv6hdr *ip6;
+       __be16 frag_off;
+       struct tcphdr _tcph, *th;
+       struct udphdr _udph, *uh;
+       struct dccp_hdr _dccph, *dh;
+
+       sip->sin6_port = 0;
+
+       offset = skb_network_offset(skb);
+       ip6 = skb_header_pointer(skb, offset, sizeof(_ipv6h), &_ipv6h);
+       if (ip6 == NULL)
+               return -EINVAL;
+       sip->sin6_addr = ip6->saddr;
+
+       nexthdr = ip6->nexthdr;
+       offset += sizeof(_ipv6h);
+       offset = ipv6_skip_exthdr(skb, offset, &nexthdr, &frag_off);
+       if (offset < 0)
+               return -EINVAL;
+
+       proto = nexthdr;
+       switch (proto) {
+       case IPPROTO_TCP:
+               th = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph);
+               if (th != NULL)
+                       sip->sin6_port = th->source;
+               break;
+       case IPPROTO_UDP:
+               uh = skb_header_pointer(skb, offset, sizeof(_udph), &_udph);
+               if (uh != NULL)
+                       sip->sin6_port = uh->source;
+               break;
+       case IPPROTO_DCCP:
+               dh = skb_header_pointer(skb, offset, sizeof(_dccph), &_dccph);
+               if (dh != NULL)
+                       sip->sin6_port = dh->dccph_sport;
+               break;
+       }
+       return proto;
+}
+#endif /* CONFIG_IPV6 */
+
+/**
+ * smack_socket_sock_rcv_skb - Smack packet delivery access check
+ * @sk: socket
+ * @skb: packet
+ *
+ * Returns 0 if the packet should be delivered, an error code otherwise
+ */
+static int smack_socket_sock_rcv_skb(struct sock *sk, struct sk_buff *skb)
+{
+       struct netlbl_lsm_secattr secattr;
+       struct socket_smack *ssp = sk->sk_security;
+       struct smack_known *skp = NULL;
+       int rc = 0;
+       struct smk_audit_info ad;
+#ifdef CONFIG_AUDIT
+       struct lsm_network_audit net;
+#endif
+#if IS_ENABLED(CONFIG_IPV6)
+       struct sockaddr_in6 sadd;
+       int proto;
+#endif /* CONFIG_IPV6 */
+
+       switch (sk->sk_family) {
+       case PF_INET:
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+               /*
+                * If there is a secmark use it rather than the CIPSO label.
+                * If there is no secmark fall back to CIPSO.
+                * The secmark is assumed to reflect policy better.
+                */
+               if (skb && skb->secmark != 0) {
+                       skp = smack_from_secid(skb->secmark);
+                       goto access_check;
+               }
+#endif /* CONFIG_SECURITY_SMACK_NETFILTER */
+               /*
+                * Translate what netlabel gave us.
+                */
+               netlbl_secattr_init(&secattr);
+
+               rc = netlbl_skbuff_getattr(skb, sk->sk_family, &secattr);
+               if (rc == 0)
+                       skp = smack_from_secattr(&secattr, ssp);
+               else
+                       skp = smack_net_ambient;
+
+               netlbl_secattr_destroy(&secattr);
+
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+access_check:
+#endif
+#ifdef CONFIG_AUDIT
+               smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+               ad.a.u.net->family = sk->sk_family;
+               ad.a.u.net->netif = skb->skb_iif;
+               ipv4_skb_to_auditdata(skb, &ad.a, NULL);
+#endif
+               /*
+                * Receiving a packet requires that the other end
+                * be able to write here. Read access is not required.
+                * This is the simplist possible security model
+                * for networking.
+                */
+               rc = smk_access(skp, ssp->smk_in, MAY_WRITE, &ad);
+               rc = smk_bu_note("IPv4 delivery", skp, ssp->smk_in,
+                                       MAY_WRITE, rc);
+               if (rc != 0)
+                       netlbl_skbuff_err(skb, rc, 0);
+               break;
+#if IS_ENABLED(CONFIG_IPV6)
+       case PF_INET6:
+               proto = smk_skb_to_addr_ipv6(skb, &sadd);
+               if (proto != IPPROTO_UDP && proto != IPPROTO_TCP)
+                       break;
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+               if (skb && skb->secmark != 0)
+                       skp = smack_from_secid(skb->secmark);
+               else
+                       skp = smack_net_ambient;
+#ifdef CONFIG_AUDIT
+               smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+               ad.a.u.net->family = sk->sk_family;
+               ad.a.u.net->netif = skb->skb_iif;
+               ipv6_skb_to_auditdata(skb, &ad.a, NULL);
+#endif /* CONFIG_AUDIT */
+               rc = smk_access(skp, ssp->smk_in, MAY_WRITE, &ad);
+               rc = smk_bu_note("IPv6 delivery", skp, ssp->smk_in,
+                                       MAY_WRITE, rc);
+#else /* CONFIG_SECURITY_SMACK_NETFILTER */
+               rc = smk_ipv6_port_check(sk, &sadd, SMK_RECEIVING);
+#endif /* CONFIG_SECURITY_SMACK_NETFILTER */
+               break;
+#endif /* CONFIG_IPV6 */
+       }
+
+       return rc;
+}
+
+/**
+ * smack_socket_getpeersec_stream - pull in packet label
+ * @sock: the socket
+ * @optval: user's destination
+ * @optlen: size thereof
+ * @len: max thereof
+ *
+ * returns zero on success, an error code otherwise
+ */
+static int smack_socket_getpeersec_stream(struct socket *sock,
+                                         char __user *optval,
+                                         int __user *optlen, unsigned len)
+{
+       struct socket_smack *ssp;
+       char *rcp = "";
+       int slen = 1;
+       int rc = 0;
+
+       ssp = sock->sk->sk_security;
+       if (ssp->smk_packet != NULL) {
+               rcp = ssp->smk_packet->smk_known;
+               slen = strlen(rcp) + 1;
+       }
+
+       if (slen > len)
+               rc = -ERANGE;
+       else if (copy_to_user(optval, rcp, slen) != 0)
+               rc = -EFAULT;
+
+       if (put_user(slen, optlen) != 0)
+               rc = -EFAULT;
+
+       return rc;
+}
+
+
+/**
+ * smack_socket_getpeersec_dgram - pull in packet label
+ * @sock: the peer socket
+ * @skb: packet data
+ * @secid: pointer to where to put the secid of the packet
+ *
+ * Sets the netlabel socket state on sk from parent
+ */
+static int smack_socket_getpeersec_dgram(struct socket *sock,
+                                        struct sk_buff *skb, u32 *secid)
+
+{
+       struct netlbl_lsm_secattr secattr;
+       struct socket_smack *ssp = NULL;
+       struct smack_known *skp;
+       int family = PF_UNSPEC;
+       u32 s = 0;      /* 0 is the invalid secid */
+       int rc;
+
+       if (skb != NULL) {
+               if (skb->protocol == htons(ETH_P_IP))
+                       family = PF_INET;
+#if IS_ENABLED(CONFIG_IPV6)
+               else if (skb->protocol == htons(ETH_P_IPV6))
+                       family = PF_INET6;
+#endif /* CONFIG_IPV6 */
+       }
+       if (family == PF_UNSPEC && sock != NULL)
+               family = sock->sk->sk_family;
+
+       switch (family) {
+       case PF_UNIX:
+               ssp = sock->sk->sk_security;
+               s = ssp->smk_out->smk_secid;
+               break;
+       case PF_INET:
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+               s = skb->secmark;
+               if (s != 0)
+                       break;
+#endif
+               /*
+                * Translate what netlabel gave us.
+                */
+               if (sock != NULL && sock->sk != NULL)
+                       ssp = sock->sk->sk_security;
+               netlbl_secattr_init(&secattr);
+               rc = netlbl_skbuff_getattr(skb, family, &secattr);
+               if (rc == 0) {
+                       skp = smack_from_secattr(&secattr, ssp);
+                       s = skp->smk_secid;
+               }
+               netlbl_secattr_destroy(&secattr);
+               break;
+#if IS_ENABLED(CONFIG_IPV6)
+       case PF_INET6:
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+               s = skb->secmark;
+#endif /* CONFIG_SECURITY_SMACK_NETFILTER */
+               break;
+#endif /* CONFIG_IPV6 */
+       }
+       *secid = s;
+       if (s == 0)
+               return -EINVAL;
+       return 0;
+}
+
+/**
+ * smack_sock_graft - Initialize a newly created socket with an existing sock
+ * @sk: child sock
+ * @parent: parent socket
+ *
+ * Set the smk_{in,out} state of an existing sock based on the process that
+ * is creating the new socket.
+ */
+static void smack_sock_graft(struct sock *sk, struct socket *parent)
+{
+       struct socket_smack *ssp;
+       struct smack_known *skp = smk_of_current();
+
+       if (sk == NULL ||
+           (sk->sk_family != PF_INET && sk->sk_family != PF_INET6))
+               return;
+
+       ssp = sk->sk_security;
+       ssp->smk_in = skp;
+       ssp->smk_out = skp;
+       /* cssp->smk_packet is already set in smack_inet_csk_clone() */
+}
+
+/**
+ * smack_inet_conn_request - Smack access check on connect
+ * @sk: socket involved
+ * @skb: packet
+ * @req: unused
+ *
+ * Returns 0 if a task with the packet label could write to
+ * the socket, otherwise an error code
+ */
+static int smack_inet_conn_request(struct sock *sk, struct sk_buff *skb,
+                                  struct request_sock *req)
+{
+       u16 family = sk->sk_family;
+       struct smack_known *skp;
+       struct socket_smack *ssp = sk->sk_security;
+       struct netlbl_lsm_secattr secattr;
+       struct sockaddr_in addr;
+       struct iphdr *hdr;
+       struct smack_known *hskp;
+       int rc;
+       struct smk_audit_info ad;
+#ifdef CONFIG_AUDIT
+       struct lsm_network_audit net;
+#endif
+
+#if IS_ENABLED(CONFIG_IPV6)
+       if (family == PF_INET6) {
+               /*
+                * Handle mapped IPv4 packets arriving
+                * via IPv6 sockets. Don't set up netlabel
+                * processing on IPv6.
+                */
+               if (skb->protocol == htons(ETH_P_IP))
+                       family = PF_INET;
+               else
+                       return 0;
+       }
+#endif /* CONFIG_IPV6 */
+
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+       /*
+        * If there is a secmark use it rather than the CIPSO label.
+        * If there is no secmark fall back to CIPSO.
+        * The secmark is assumed to reflect policy better.
+        */
+       if (skb && skb->secmark != 0) {
+               skp = smack_from_secid(skb->secmark);
+               goto access_check;
+       }
+#endif /* CONFIG_SECURITY_SMACK_NETFILTER */
+
+       netlbl_secattr_init(&secattr);
+       rc = netlbl_skbuff_getattr(skb, family, &secattr);
+       if (rc == 0)
+               skp = smack_from_secattr(&secattr, ssp);
+       else
+               skp = &smack_known_huh;
+       netlbl_secattr_destroy(&secattr);
+
+#ifdef CONFIG_SECURITY_SMACK_NETFILTER
+access_check:
+#endif
+
+#ifdef CONFIG_AUDIT
+       smk_ad_init_net(&ad, __func__, LSM_AUDIT_DATA_NET, &net);
+       ad.a.u.net->family = family;
+       ad.a.u.net->netif = skb->skb_iif;
+       ipv4_skb_to_auditdata(skb, &ad.a, NULL);
+#endif
+       /*
+        * Receiving a packet requires that the other end be able to write
+        * here. Read access is not required.
+        */
+       rc = smk_access(skp, ssp->smk_in, MAY_WRITE, &ad);
+       rc = smk_bu_note("IPv4 connect", skp, ssp->smk_in, MAY_WRITE, rc);
+       if (rc != 0)
+               return rc;
+
+       /*
+        * Save the peer's label in the request_sock so we can later setup
+        * smk_packet in the child socket so that SO_PEERCRED can report it.
+        */
+       req->peer_secid = skp->smk_secid;
+
+       /*
+        * We need to decide if we want to label the incoming connection here
+        * if we do we only need to label the request_sock and the stack will
+        * propagate the wire-label to the sock when it is created.
+        */
+       hdr = ip_hdr(skb);
+       addr.sin_addr.s_addr = hdr->saddr;
+       rcu_read_lock();
+       hskp = smack_host_label(&addr);
+       rcu_read_unlock();
+
+       if (hskp == NULL)
+               rc = netlbl_req_setattr(req, &skp->smk_netlabel);
+       else
+               netlbl_req_delattr(req);
+
+       return rc;
+}
+
+/**
+ * smack_inet_csk_clone - Copy the connection information to the new socket
+ * @sk: the new socket
+ * @req: the connection's request_sock
+ *
+ * Transfer the connection's peer label to the newly created socket.
+ */
+static void smack_inet_csk_clone(struct sock *sk,
+                                const struct request_sock *req)
+{
+       struct socket_smack *ssp = sk->sk_security;
+       struct smack_known *skp;
+
+       if (req->peer_secid != 0) {
+               skp = smack_from_secid(req->peer_secid);
+               ssp->smk_packet = skp;
+       } else
+               ssp->smk_packet = NULL;
+}
+
+/*
+ * Key management security hooks
+ *
+ * Casey has not tested key support very heavily.
+ * The permission check is most likely too restrictive.
+ * If you care about keys please have a look.
+ */
+#ifdef CONFIG_KEYS
+
+/**
+ * smack_key_alloc - Set the key security blob
+ * @key: object
+ * @cred: the credentials to use
+ * @flags: unused
+ *
+ * No allocation required
+ *
+ * Returns 0
+ */
+static int smack_key_alloc(struct key *key, const struct cred *cred,
+                          unsigned long flags)
+{
+       struct smack_known *skp = smk_of_task(cred->security);
+
+       key->security = skp;
+       return 0;
+}
+
+/**
+ * smack_key_free - Clear the key security blob
+ * @key: the object
+ *
+ * Clear the blob pointer
+ */
+static void smack_key_free(struct key *key)
+{
+       key->security = NULL;
+}
+
+/**
+ * smack_key_permission - Smack access on a key
+ * @key_ref: gets to the object
+ * @cred: the credentials to use
+ * @perm: requested key permissions
+ *
+ * Return 0 if the task has read and write to the object,
+ * an error code otherwise
+ */
+static int smack_key_permission(key_ref_t key_ref,
+                               const struct cred *cred, unsigned perm)
+{
+       struct key *keyp;
+       struct smk_audit_info ad;
+       struct smack_known *tkp = smk_of_task(cred->security);
+       int request = 0;
+       int rc;
+
+       keyp = key_ref_to_ptr(key_ref);
+       if (keyp == NULL)
+               return -EINVAL;
+       /*
+        * If the key hasn't been initialized give it access so that
+        * it may do so.
+        */
+       if (keyp->security == NULL)
+               return 0;
+       /*
+        * This should not occur
+        */
+       if (tkp == NULL)
+               return -EACCES;
+#ifdef CONFIG_AUDIT
+       smk_ad_init(&ad, __func__, LSM_AUDIT_DATA_KEY);
+       ad.a.u.key_struct.key = keyp->serial;
+       ad.a.u.key_struct.key_desc = keyp->description;
+#endif
+       if (perm & KEY_NEED_READ)
+               request = MAY_READ;
+       if (perm & (KEY_NEED_WRITE | KEY_NEED_LINK | KEY_NEED_SETATTR))
+               request = MAY_WRITE;
+       rc = smk_access(tkp, keyp->security, request, &ad);
+       rc = smk_bu_note("key access", tkp, keyp->security, request, rc);
+       return rc;
+}
+
+/*
+ * smack_key_getsecurity - Smack label tagging the key
+ * @key points to the key to be queried
+ * @_buffer points to a pointer that should be set to point to the
+ * resulting string (if no label or an error occurs).
+ * Return the length of the string (including terminating NUL) or -ve if
+ * an error.
+ * May also return 0 (and a NULL buffer pointer) if there is no label.
+ */
+static int smack_key_getsecurity(struct key *key, char **_buffer)
+{
+       struct smack_known *skp = key->security;
+       size_t length;
+       char *copy;
+
+       if (key->security == NULL) {
+               *_buffer = NULL;
+               return 0;
+       }
+
+       copy = kstrdup(skp->smk_known, GFP_KERNEL);
+       if (copy == NULL)
+               return -ENOMEM;
+       length = strlen(copy) + 1;
+
+       *_buffer = copy;
+       return length;
+}
+
+#endif /* CONFIG_KEYS */
+
+/*
+ * Smack Audit hooks
+ *
+ * Audit requires a unique representation of each Smack specific
+ * rule. This unique representation is used to distinguish the
+ * object to be audited from remaining kernel objects and also
+ * works as a glue between the audit hooks.
+ *
+ * Since repository entries are added but never deleted, we'll use
+ * the smack_known label address related to the given audit rule as
+ * the needed unique representation. This also better fits the smack
+ * model where nearly everything is a label.
+ */
+#ifdef CONFIG_AUDIT
+
+/**
+ * smack_audit_rule_init - Initialize a smack audit rule
+ * @field: audit rule fields given from user-space (audit.h)
+ * @op: required testing operator (=, !=, >, <, ...)
+ * @rulestr: smack label to be audited
+ * @vrule: pointer to save our own audit rule representation
+ *
+ * Prepare to audit cases where (@field @op @rulestr) is true.
+ * The label to be audited is created if necessay.
+ */
+static int smack_audit_rule_init(u32 field, u32 op, char *rulestr, void **vrule)
+{
+       struct smack_known *skp;
+       char **rule = (char **)vrule;
+       *rule = NULL;
+
+       if (field != AUDIT_SUBJ_USER && field != AUDIT_OBJ_USER)
+               return -EINVAL;
+
+       if (op != Audit_equal && op != Audit_not_equal)
+               return -EINVAL;
+
+       skp = smk_import_entry(rulestr, 0);
+       if (skp)
+               *rule = skp->smk_known;
+
+       return 0;
+}
+
+/**
+ * smack_audit_rule_known - Distinguish Smack audit rules
+ * @krule: rule of interest, in Audit kernel representation format
+ *
+ * This is used to filter Smack rules from remaining Audit ones.
+ * If it's proved that this rule belongs to us, the
+ * audit_rule_match hook will be called to do the final judgement.
+ */
+static int smack_audit_rule_known(struct audit_krule *krule)
+{
+       struct audit_field *f;
+       int i;
+
+       for (i = 0; i < krule->field_count; i++) {
+               f = &krule->fields[i];
+
+               if (f->type == AUDIT_SUBJ_USER || f->type == AUDIT_OBJ_USER)
+                       return 1;
+       }
+
+       return 0;
+}
+
+/**
+ * smack_audit_rule_match - Audit given object ?
+ * @secid: security id for identifying the object to test
+ * @field: audit rule flags given from user-space
+ * @op: required testing operator
+ * @vrule: smack internal rule presentation
+ * @actx: audit context associated with the check
+ *
+ * The core Audit hook. It's used to take the decision of
+ * whether to audit or not to audit a given object.
+ */
+static int smack_audit_rule_match(u32 secid, u32 field, u32 op, void *vrule,
+                                 struct audit_context *actx)
+{
+       struct smack_known *skp;
+       char *rule = vrule;
+
+       if (unlikely(!rule)) {
+               WARN_ONCE(1, "Smack: missing rule\n");
+               return -ENOENT;
+       }
+
+       if (field != AUDIT_SUBJ_USER && field != AUDIT_OBJ_USER)
+               return 0;
+
+       skp = smack_from_secid(secid);
+
+       /*
+        * No need to do string comparisons. If a match occurs,
+        * both pointers will point to the same smack_known
+        * label.
+        */
+       if (op == Audit_equal)
+               return (rule == skp->smk_known);
+       if (op == Audit_not_equal)
+               return (rule != skp->smk_known);
+
+       return 0;
+}
+
+/**
+ * smack_audit_rule_free - free smack rule representation
+ * @vrule: rule to be freed.
+ *
+ * No memory was allocated.
+ */
+static void smack_audit_rule_free(void *vrule)
+{
+       /* No-op */
+}
+
+#endif /* CONFIG_AUDIT */
+
+/**
+ * smack_ismaclabel - check if xattr @name references a smack MAC label
+ * @name: Full xattr name to check.
+ */
+static int smack_ismaclabel(const char *name)
+{
+       return (strcmp(name, XATTR_SMACK_SUFFIX) == 0);
+}
+
+
+/**
+ * smack_secid_to_secctx - return the smack label for a secid
+ * @secid: incoming integer
+ * @secdata: destination
+ * @seclen: how long it is
+ *
+ * Exists for networking code.
+ */
+static int smack_secid_to_secctx(u32 secid, char **secdata, u32 *seclen)
+{
+       struct smack_known *skp = smack_from_secid(secid);
+
+       if (secdata)
+               *secdata = skp->smk_known;
+       *seclen = strlen(skp->smk_known);
+       return 0;
+}
+
+/**
+ * smack_secctx_to_secid - return the secid for a smack label
+ * @secdata: smack label
+ * @seclen: how long result is
+ * @secid: outgoing integer
+ *
+ * Exists for audit and networking code.
+ */
+static int smack_secctx_to_secid(const char *secdata, u32 seclen, u32 *secid)
+{
+       struct smack_known *skp = smk_find_entry(secdata);
+
+       if (skp)
+               *secid = skp->smk_secid;
+       else
+               *secid = 0;
+       return 0;
+}
+
+/**
+ * smack_release_secctx - don't do anything.
+ * @secdata: unused
+ * @seclen: unused
+ *
+ * Exists to make sure nothing gets done, and properly
+ */
+static void smack_release_secctx(char *secdata, u32 seclen)
+{
+}
+
+static int smack_inode_notifysecctx(struct inode *inode, void *ctx, u32 ctxlen)
+{
+       return smack_inode_setsecurity(inode, XATTR_SMACK_SUFFIX, ctx, ctxlen, 0);
+}
+
+static int smack_inode_setsecctx(struct dentry *dentry, void *ctx, u32 ctxlen)
+{
+       return __vfs_setxattr_noperm(dentry, XATTR_NAME_SMACK, ctx, ctxlen, 0);
+}
+
+static int smack_inode_getsecctx(struct inode *inode, void **ctx, u32 *ctxlen)
+{
+       int len = 0;
+       len = smack_inode_getsecurity(inode, XATTR_SMACK_SUFFIX, ctx, true);
+
+       if (len < 0)
+               return len;
+       *ctxlen = len;
+       return 0;
+}
+
+struct security_operations smack_ops = {
+       .name =                         "smack",
+
+       .ptrace_access_check =          smack_ptrace_access_check,
+       .ptrace_traceme =               smack_ptrace_traceme,
+       .syslog =                       smack_syslog,
+
+       .sb_alloc_security =            smack_sb_alloc_security,
+       .sb_free_security =             smack_sb_free_security,
+       .sb_copy_data =                 smack_sb_copy_data,
+       .sb_kern_mount =                smack_sb_kern_mount,
+       .sb_statfs =                    smack_sb_statfs,
+
+       .bprm_set_creds =               smack_bprm_set_creds,
+       .bprm_committing_creds =        smack_bprm_committing_creds,
+       .bprm_secureexec =              smack_bprm_secureexec,
+
+       .inode_alloc_security =         smack_inode_alloc_security,
+       .inode_free_security =          smack_inode_free_security,
+       .inode_init_security =          smack_inode_init_security,
+       .inode_link =                   smack_inode_link,
+       .inode_unlink =                 smack_inode_unlink,
+       .inode_rmdir =                  smack_inode_rmdir,
+       .inode_rename =                 smack_inode_rename,
+       .inode_permission =             smack_inode_permission,
+       .inode_setattr =                smack_inode_setattr,
+       .inode_getattr =                smack_inode_getattr,
+       .inode_setxattr =               smack_inode_setxattr,
+       .inode_post_setxattr =          smack_inode_post_setxattr,
+       .inode_getxattr =               smack_inode_getxattr,
+       .inode_removexattr =            smack_inode_removexattr,
+       .inode_getsecurity =            smack_inode_getsecurity,
+       .inode_setsecurity =            smack_inode_setsecurity,
+       .inode_listsecurity =           smack_inode_listsecurity,
+       .inode_getsecid =               smack_inode_getsecid,
+
+       .file_permission =              smack_file_permission,
+       .file_alloc_security =          smack_file_alloc_security,
+       .file_free_security =           smack_file_free_security,
+       .file_ioctl =                   smack_file_ioctl,
+       .file_lock =                    smack_file_lock,
+       .file_fcntl =                   smack_file_fcntl,
+       .mmap_file =                    smack_mmap_file,
+       .mmap_addr =                    cap_mmap_addr,
+       .file_set_fowner =              smack_file_set_fowner,
+       .file_send_sigiotask =          smack_file_send_sigiotask,
+       .file_receive =                 smack_file_receive,
+
+       .file_open =                    smack_file_open,
+
+       .cred_alloc_blank =             smack_cred_alloc_blank,
+       .cred_free =                    smack_cred_free,
+       .cred_prepare =                 smack_cred_prepare,
+       .cred_transfer =                smack_cred_transfer,
+       .kernel_act_as =                smack_kernel_act_as,
+       .kernel_create_files_as =       smack_kernel_create_files_as,
+       .task_setpgid =                 smack_task_setpgid,
+       .task_getpgid =                 smack_task_getpgid,
+       .task_getsid =                  smack_task_getsid,
+       .task_getsecid =                smack_task_getsecid,
+       .task_setnice =                 smack_task_setnice,
+       .task_setioprio =               smack_task_setioprio,
+       .task_getioprio =               smack_task_getioprio,
+       .task_setscheduler =            smack_task_setscheduler,
+       .task_getscheduler =            smack_task_getscheduler,
+       .task_movememory =              smack_task_movememory,
+       .task_kill =                    smack_task_kill,
+       .task_wait =                    smack_task_wait,
+       .task_to_inode =                smack_task_to_inode,
+
+       .ipc_permission =               smack_ipc_permission,
+       .ipc_getsecid =                 smack_ipc_getsecid,
+
+       .msg_msg_alloc_security =       smack_msg_msg_alloc_security,
+       .msg_msg_free_security =        smack_msg_msg_free_security,
+
+       .msg_queue_alloc_security =     smack_msg_queue_alloc_security,
+       .msg_queue_free_security =      smack_msg_queue_free_security,
+       .msg_queue_associate =          smack_msg_queue_associate,
+       .msg_queue_msgctl =             smack_msg_queue_msgctl,
+       .msg_queue_msgsnd =             smack_msg_queue_msgsnd,
+       .msg_queue_msgrcv =             smack_msg_queue_msgrcv,
+
+       .shm_alloc_security =           smack_shm_alloc_security,
+       .shm_free_security =            smack_shm_free_security,
+       .shm_associate =                smack_shm_associate,
+       .shm_shmctl =                   smack_shm_shmctl,
+       .shm_shmat =                    smack_shm_shmat,
+
+       .sem_alloc_security =           smack_sem_alloc_security,
+       .sem_free_security =            smack_sem_free_security,
+       .sem_associate =                smack_sem_associate,
+       .sem_semctl =                   smack_sem_semctl,
+       .sem_semop =                    smack_sem_semop,
+
+       .d_instantiate =                smack_d_instantiate,
+
+       .getprocattr =                  smack_getprocattr,
+       .setprocattr =                  smack_setprocattr,
+
+       .unix_stream_connect =          smack_unix_stream_connect,
+       .unix_may_send =                smack_unix_may_send,
+
+       .socket_post_create =           smack_socket_post_create,
+#ifndef CONFIG_SECURITY_SMACK_NETFILTER
+       .socket_bind =                  smack_socket_bind,
+#endif /* CONFIG_SECURITY_SMACK_NETFILTER */
+       .socket_connect =               smack_socket_connect,
+       .socket_sendmsg =               smack_socket_sendmsg,
+       .socket_sock_rcv_skb =          smack_socket_sock_rcv_skb,
+       .socket_getpeersec_stream =     smack_socket_getpeersec_stream,
+       .socket_getpeersec_dgram =      smack_socket_getpeersec_dgram,
+       .sk_alloc_security =            smack_sk_alloc_security,
+       .sk_free_security =             smack_sk_free_security,
+       .sock_graft =                   smack_sock_graft,
+       .inet_conn_request =            smack_inet_conn_request,
+       .inet_csk_clone =               smack_inet_csk_clone,
+
+ /* key management security hooks */
+#ifdef CONFIG_KEYS
+       .key_alloc =                    smack_key_alloc,
+       .key_free =                     smack_key_free,
+       .key_permission =               smack_key_permission,
+       .key_getsecurity =              smack_key_getsecurity,
+#endif /* CONFIG_KEYS */
+
+ /* Audit hooks */
+#ifdef CONFIG_AUDIT
+       .audit_rule_init =              smack_audit_rule_init,
+       .audit_rule_known =             smack_audit_rule_known,
+       .audit_rule_match =             smack_audit_rule_match,
+       .audit_rule_free =              smack_audit_rule_free,
+#endif /* CONFIG_AUDIT */
+
+       .ismaclabel =                   smack_ismaclabel,
+       .secid_to_secctx =              smack_secid_to_secctx,
+       .secctx_to_secid =              smack_secctx_to_secid,
+       .release_secctx =               smack_release_secctx,
+       .inode_notifysecctx =           smack_inode_notifysecctx,
+       .inode_setsecctx =              smack_inode_setsecctx,
+       .inode_getsecctx =              smack_inode_getsecctx,
+};
+
+
+static __init void init_smack_known_list(void)
+{
+       /*
+        * Initialize rule list locks
+        */
+       mutex_init(&smack_known_huh.smk_rules_lock);
+       mutex_init(&smack_known_hat.smk_rules_lock);
+       mutex_init(&smack_known_floor.smk_rules_lock);
+       mutex_init(&smack_known_star.smk_rules_lock);
+       mutex_init(&smack_known_invalid.smk_rules_lock);
+       mutex_init(&smack_known_web.smk_rules_lock);
+       /*
+        * Initialize rule lists
+        */
+       INIT_LIST_HEAD(&smack_known_huh.smk_rules);
+       INIT_LIST_HEAD(&smack_known_hat.smk_rules);
+       INIT_LIST_HEAD(&smack_known_star.smk_rules);
+       INIT_LIST_HEAD(&smack_known_floor.smk_rules);
+       INIT_LIST_HEAD(&smack_known_invalid.smk_rules);
+       INIT_LIST_HEAD(&smack_known_web.smk_rules);
+       /*
+        * Create the known labels list
+        */
+       smk_insert_entry(&smack_known_huh);
+       smk_insert_entry(&smack_known_hat);
+       smk_insert_entry(&smack_known_star);
+       smk_insert_entry(&smack_known_floor);
+       smk_insert_entry(&smack_known_invalid);
+       smk_insert_entry(&smack_known_web);
+}
+
+/**
+ * smack_init - initialize the smack system
+ *
+ * Returns 0
+ */
+static __init int smack_init(void)
+{
+       struct cred *cred;
+       struct task_smack *tsp;
+
+       if (!security_module_enable(&smack_ops))
+               return 0;
+
+       smack_enabled = 1;
+
+       smack_inode_cache = KMEM_CACHE(inode_smack, 0);
+       if (!smack_inode_cache)
+               return -ENOMEM;
+
+       tsp = new_task_smack(&smack_known_floor, &smack_known_floor,
+                               GFP_KERNEL);
+       if (tsp == NULL) {
+               kmem_cache_destroy(smack_inode_cache);
+               return -ENOMEM;
+       }
+
+       printk(KERN_INFO "Smack:  Initializing.\n");
+
+       /*
+        * Set the security state for the initial task.
+        */
+       cred = (struct cred *) current->cred;
+       cred->security = tsp;
+
+       /* initialize the smack_known_list */
+       init_smack_known_list();
+
+       /*
+        * Register with LSM
+        */
+       if (register_security(&smack_ops))
+               panic("smack: Unable to register with kernel.\n");
+
+       return 0;
+}
+
+/*
+ * Smack requires early initialization in order to label
+ * all processes and objects when they are created.
+ */
+security_initcall(smack_init);