Linux kernel & device driver programming

Cross-Referenced Linux and Device Driver Code

[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ]
Version: [ 2.6.11.8 ] [ 2.6.25 ] [ 2.6.25.8 ] [ 2.6.31.13 ] Architecture: [ i386 ]
  1 /*
  2  *  NSA Security-Enhanced Linux (SELinux) security module
  3  *
  4  *  This file contains the SELinux hook function implementations.
  5  *
  6  *  Authors:  Stephen Smalley, <sds@epoch.ncsc.mil>
  7  *            Chris Vance, <cvance@nai.com>
  8  *            Wayne Salamon, <wsalamon@nai.com>
  9  *            James Morris <jmorris@redhat.com>
 10  *
 11  *  Copyright (C) 2001,2002 Networks Associates Technology, Inc.
 12  *  Copyright (C) 2003-2008 Red Hat, Inc., James Morris <jmorris@redhat.com>
 13  *                                         Eric Paris <eparis@redhat.com>
 14  *  Copyright (C) 2004-2005 Trusted Computer Solutions, Inc.
 15  *                          <dgoeddel@trustedcs.com>
 16  *  Copyright (C) 2006, 2007 Hewlett-Packard Development Company, L.P.
 17  *              Paul Moore <paul.moore@hp.com>
 18  *  Copyright (C) 2007 Hitachi Software Engineering Co., Ltd.
 19  *                     Yuichi Nakamura <ynakam@hitachisoft.jp>
 20  *
 21  *      This program is free software; you can redistribute it and/or modify
 22  *      it under the terms of the GNU General Public License version 2,
 23  *      as published by the Free Software Foundation.
 24  */
 25 
 26 #include <linux/init.h>
 27 #include <linux/kernel.h>
 28 #include <linux/tracehook.h>
 29 #include <linux/errno.h>
 30 #include <linux/sched.h>
 31 #include <linux/security.h>
 32 #include <linux/xattr.h>
 33 #include <linux/capability.h>
 34 #include <linux/unistd.h>
 35 #include <linux/mm.h>
 36 #include <linux/mman.h>
 37 #include <linux/slab.h>
 38 #include <linux/pagemap.h>
 39 #include <linux/swap.h>
 40 #include <linux/spinlock.h>
 41 #include <linux/syscalls.h>
 42 #include <linux/file.h>
 43 #include <linux/fdtable.h>
 44 #include <linux/namei.h>
 45 #include <linux/mount.h>
 46 #include <linux/proc_fs.h>
 47 #include <linux/netfilter_ipv4.h>
 48 #include <linux/netfilter_ipv6.h>
 49 #include <linux/tty.h>
 50 #include <net/icmp.h>
 51 #include <net/ip.h>             /* for local_port_range[] */
 52 #include <net/tcp.h>            /* struct or_callable used in sock_rcv_skb */
 53 #include <net/net_namespace.h>
 54 #include <net/netlabel.h>
 55 #include <linux/uaccess.h>
 56 #include <asm/ioctls.h>
 57 #include <asm/atomic.h>
 58 #include <linux/bitops.h>
 59 #include <linux/interrupt.h>
 60 #include <linux/netdevice.h>    /* for network interface checks */
 61 #include <linux/netlink.h>
 62 #include <linux/tcp.h>
 63 #include <linux/udp.h>
 64 #include <linux/dccp.h>
 65 #include <linux/quota.h>
 66 #include <linux/un.h>           /* for Unix socket types */
 67 #include <net/af_unix.h>        /* for Unix socket types */
 68 #include <linux/parser.h>
 69 #include <linux/nfs_mount.h>
 70 #include <net/ipv6.h>
 71 #include <linux/hugetlb.h>
 72 #include <linux/personality.h>
 73 #include <linux/sysctl.h>
 74 #include <linux/audit.h>
 75 #include <linux/string.h>
 76 #include <linux/selinux.h>
 77 #include <linux/mutex.h>
 78 #include <linux/posix-timers.h>
 79 
 80 #include "avc.h"
 81 #include "objsec.h"
 82 #include "netif.h"
 83 #include "netnode.h"
 84 #include "netport.h"
 85 #include "xfrm.h"
 86 #include "netlabel.h"
 87 #include "audit.h"
 88 
 89 #define XATTR_SELINUX_SUFFIX "selinux"
 90 #define XATTR_NAME_SELINUX XATTR_SECURITY_PREFIX XATTR_SELINUX_SUFFIX
 91 
 92 #define NUM_SEL_MNT_OPTS 5
 93 
 94 extern unsigned int policydb_loaded_version;
 95 extern int selinux_nlmsg_lookup(u16 sclass, u16 nlmsg_type, u32 *perm);
 96 extern struct security_operations *security_ops;
 97 
 98 /* SECMARK reference count */
 99 atomic_t selinux_secmark_refcount = ATOMIC_INIT(0);
100 
101 #ifdef CONFIG_SECURITY_SELINUX_DEVELOP
102 int selinux_enforcing;
103 
104 static int __init enforcing_setup(char *str)
105 {
106         unsigned long enforcing;
107         if (!strict_strtoul(str, 0, &enforcing))
108                 selinux_enforcing = enforcing ? 1 : 0;
109         return 1;
110 }
111 __setup("enforcing=", enforcing_setup);
112 #endif
113 
114 #ifdef CONFIG_SECURITY_SELINUX_BOOTPARAM
115 int selinux_enabled = CONFIG_SECURITY_SELINUX_BOOTPARAM_VALUE;
116 
117 static int __init selinux_enabled_setup(char *str)
118 {
119         unsigned long enabled;
120         if (!strict_strtoul(str, 0, &enabled))
121                 selinux_enabled = enabled ? 1 : 0;
122         return 1;
123 }
124 __setup("selinux=", selinux_enabled_setup);
125 #else
126 int selinux_enabled = 1;
127 #endif
128 
129 
130 /*
131  * Minimal support for a secondary security module,
132  * just to allow the use of the capability module.
133  */
134 static struct security_operations *secondary_ops;
135 
136 /* Lists of inode and superblock security structures initialized
137    before the policy was loaded. */
138 static LIST_HEAD(superblock_security_head);
139 static DEFINE_SPINLOCK(sb_security_lock);
140 
141 static struct kmem_cache *sel_inode_cache;
142 
143 /**
144  * selinux_secmark_enabled - Check to see if SECMARK is currently enabled
145  *
146  * Description:
147  * This function checks the SECMARK reference counter to see if any SECMARK
148  * targets are currently configured, if the reference counter is greater than
149  * zero SECMARK is considered to be enabled.  Returns true (1) if SECMARK is
150  * enabled, false (0) if SECMARK is disabled.
151  *
152  */
153 static int selinux_secmark_enabled(void)
154 {
155         return (atomic_read(&selinux_secmark_refcount) > 0);
156 }
157 
158 /*
159  * initialise the security for the init task
160  */
161 static void cred_init_security(void)
162 {
163         struct cred *cred = (struct cred *) current->real_cred;
164         struct task_security_struct *tsec;
165 
166         tsec = kzalloc(sizeof(struct task_security_struct), GFP_KERNEL);
167         if (!tsec)
168                 panic("SELinux:  Failed to initialize initial task.\n");
169 
170         tsec->osid = tsec->sid = SECINITSID_KERNEL;
171         cred->security = tsec;
172 }
173 
174 /*
175  * get the security ID of a set of credentials
176  */
177 static inline u32 cred_sid(const struct cred *cred)
178 {
179         const struct task_security_struct *tsec;
180 
181         tsec = cred->security;
182         return tsec->sid;
183 }
184 
185 /*
186  * get the objective security ID of a task
187  */
188 static inline u32 task_sid(const struct task_struct *task)
189 {
190         u32 sid;
191 
192         rcu_read_lock();
193         sid = cred_sid(__task_cred(task));
194         rcu_read_unlock();
195         return sid;
196 }
197 
198 /*
199  * get the subjective security ID of the current task
200  */
201 static inline u32 current_sid(void)
202 {
203         const struct task_security_struct *tsec = current_cred()->security;
204 
205         return tsec->sid;
206 }
207 
208 /* Allocate and free functions for each kind of security blob. */
209 
210 static int inode_alloc_security(struct inode *inode)
211 {
212         struct inode_security_struct *isec;
213         u32 sid = current_sid();
214 
215         isec = kmem_cache_zalloc(sel_inode_cache, GFP_NOFS);
216         if (!isec)
217                 return -ENOMEM;
218 
219         mutex_init(&isec->lock);
220         INIT_LIST_HEAD(&isec->list);
221         isec->inode = inode;
222         isec->sid = SECINITSID_UNLABELED;
223         isec->sclass = SECCLASS_FILE;
224         isec->task_sid = sid;
225         inode->i_security = isec;
226 
227         return 0;
228 }
229 
230 static void inode_free_security(struct inode *inode)
231 {
232         struct inode_security_struct *isec = inode->i_security;
233         struct superblock_security_struct *sbsec = inode->i_sb->s_security;
234 
235         spin_lock(&sbsec->isec_lock);
236         if (!list_empty(&isec->list))
237                 list_del_init(&isec->list);
238         spin_unlock(&sbsec->isec_lock);
239 
240         inode->i_security = NULL;
241         kmem_cache_free(sel_inode_cache, isec);
242 }
243 
244 static int file_alloc_security(struct file *file)
245 {
246         struct file_security_struct *fsec;
247         u32 sid = current_sid();
248 
249         fsec = kzalloc(sizeof(struct file_security_struct), GFP_KERNEL);
250         if (!fsec)
251                 return -ENOMEM;
252 
253         fsec->sid = sid;
254         fsec->fown_sid = sid;
255         file->f_security = fsec;
256 
257         return 0;
258 }
259 
260 static void file_free_security(struct file *file)
261 {
262         struct file_security_struct *fsec = file->f_security;
263         file->f_security = NULL;
264         kfree(fsec);
265 }
266 
267 static int superblock_alloc_security(struct super_block *sb)
268 {
269         struct superblock_security_struct *sbsec;
270 
271         sbsec = kzalloc(sizeof(struct superblock_security_struct), GFP_KERNEL);
272         if (!sbsec)
273                 return -ENOMEM;
274 
275         mutex_init(&sbsec->lock);
276         INIT_LIST_HEAD(&sbsec->list);
277         INIT_LIST_HEAD(&sbsec->isec_head);
278         spin_lock_init(&sbsec->isec_lock);
279         sbsec->sb = sb;
280         sbsec->sid = SECINITSID_UNLABELED;
281         sbsec->def_sid = SECINITSID_FILE;
282         sbsec->mntpoint_sid = SECINITSID_UNLABELED;
283         sb->s_security = sbsec;
284 
285         return 0;
286 }
287 
288 static void superblock_free_security(struct super_block *sb)
289 {
290         struct superblock_security_struct *sbsec = sb->s_security;
291 
292         spin_lock(&sb_security_lock);
293         if (!list_empty(&sbsec->list))
294                 list_del_init(&sbsec->list);
295         spin_unlock(&sb_security_lock);
296 
297         sb->s_security = NULL;
298         kfree(sbsec);
299 }
300 
301 static int sk_alloc_security(struct sock *sk, int family, gfp_t priority)
302 {
303         struct sk_security_struct *ssec;
304 
305         ssec = kzalloc(sizeof(*ssec), priority);
306         if (!ssec)
307                 return -ENOMEM;
308 
309         ssec->peer_sid = SECINITSID_UNLABELED;
310         ssec->sid = SECINITSID_UNLABELED;
311         sk->sk_security = ssec;
312 
313         selinux_netlbl_sk_security_reset(ssec);
314 
315         return 0;
316 }
317 
318 static void sk_free_security(struct sock *sk)
319 {
320         struct sk_security_struct *ssec = sk->sk_security;
321 
322         sk->sk_security = NULL;
323         selinux_netlbl_sk_security_free(ssec);
324         kfree(ssec);
325 }
326 
327 /* The security server must be initialized before
328    any labeling or access decisions can be provided. */
329 extern int ss_initialized;
330 
331 /* The file system's label must be initialized prior to use. */
332 
333 static char *labeling_behaviors[6] = {
334         "uses xattr",
335         "uses transition SIDs",
336         "uses task SIDs",
337         "uses genfs_contexts",
338         "not configured for labeling",
339         "uses mountpoint labeling",
340 };
341 
342 static int inode_doinit_with_dentry(struct inode *inode, struct dentry *opt_dentry);
343 
344 static inline int inode_doinit(struct inode *inode)
345 {
346         return inode_doinit_with_dentry(inode, NULL);
347 }
348 
349 enum {
350         Opt_error = -1,
351         Opt_context = 1,
352         Opt_fscontext = 2,
353         Opt_defcontext = 3,
354         Opt_rootcontext = 4,
355         Opt_labelsupport = 5,
356 };
357 
358 static const match_table_t tokens = {
359         {Opt_context, CONTEXT_STR "%s"},
360         {Opt_fscontext, FSCONTEXT_STR "%s"},
361         {Opt_defcontext, DEFCONTEXT_STR "%s"},
362         {Opt_rootcontext, ROOTCONTEXT_STR "%s"},
363         {Opt_labelsupport, LABELSUPP_STR},
364         {Opt_error, NULL},
365 };
366 
367 #define SEL_MOUNT_FAIL_MSG "SELinux:  duplicate or incompatible mount options\n"
368 
369 static int may_context_mount_sb_relabel(u32 sid,
370                         struct superblock_security_struct *sbsec,
371                         const struct cred *cred)
372 {
373         const struct task_security_struct *tsec = cred->security;
374         int rc;
375 
376         rc = avc_has_perm(tsec->sid, sbsec->sid, SECCLASS_FILESYSTEM,
377                           FILESYSTEM__RELABELFROM, NULL);
378         if (rc)
379                 return rc;
380 
381         rc = avc_has_perm(tsec->sid, sid, SECCLASS_FILESYSTEM,
382                           FILESYSTEM__RELABELTO, NULL);
383         return rc;
384 }
385 
386 static int may_context_mount_inode_relabel(u32 sid,
387                         struct superblock_security_struct *sbsec,
388                         const struct cred *cred)
389 {
390         const struct task_security_struct *tsec = cred->security;
391         int rc;
392         rc = avc_has_perm(tsec->sid, sbsec->sid, SECCLASS_FILESYSTEM,
393                           FILESYSTEM__RELABELFROM, NULL);
394         if (rc)
395                 return rc;
396 
397         rc = avc_has_perm(sid, sbsec->sid, SECCLASS_FILESYSTEM,
398                           FILESYSTEM__ASSOCIATE, NULL);
399         return rc;
400 }
401 
402 static int sb_finish_set_opts(struct super_block *sb)
403 {
404         struct superblock_security_struct *sbsec = sb->s_security;
405         struct dentry *root = sb->s_root;
406         struct inode *root_inode = root->d_inode;
407         int rc = 0;
408 
409         if (sbsec->behavior == SECURITY_FS_USE_XATTR) {
410                 /* Make sure that the xattr handler exists and that no
411                    error other than -ENODATA is returned by getxattr on
412                    the root directory.  -ENODATA is ok, as this may be
413                    the first boot of the SELinux kernel before we have
414                    assigned xattr values to the filesystem. */
415                 if (!root_inode->i_op->getxattr) {
416                         printk(KERN_WARNING "SELinux: (dev %s, type %s) has no "
417                                "xattr support\n", sb->s_id, sb->s_type->name);
418                         rc = -EOPNOTSUPP;
419                         goto out;
420                 }
421                 rc = root_inode->i_op->getxattr(root, XATTR_NAME_SELINUX, NULL, 0);
422                 if (rc < 0 && rc != -ENODATA) {
423                         if (rc == -EOPNOTSUPP)
424                                 printk(KERN_WARNING "SELinux: (dev %s, type "
425                                        "%s) has no security xattr handler\n",
426                                        sb->s_id, sb->s_type->name);
427                         else
428                                 printk(KERN_WARNING "SELinux: (dev %s, type "
429                                        "%s) getxattr errno %d\n", sb->s_id,
430                                        sb->s_type->name, -rc);
431                         goto out;
432                 }
433         }
434 
435         sbsec->flags |= (SE_SBINITIALIZED | SE_SBLABELSUPP);
436 
437         if (sbsec->behavior > ARRAY_SIZE(labeling_behaviors))
438                 printk(KERN_ERR "SELinux: initialized (dev %s, type %s), unknown behavior\n",
439                        sb->s_id, sb->s_type->name);
440         else
441                 printk(KERN_DEBUG "SELinux: initialized (dev %s, type %s), %s\n",
442                        sb->s_id, sb->s_type->name,
443                        labeling_behaviors[sbsec->behavior-1]);
444 
445         if (sbsec->behavior == SECURITY_FS_USE_GENFS ||
446             sbsec->behavior == SECURITY_FS_USE_MNTPOINT ||
447             sbsec->behavior == SECURITY_FS_USE_NONE ||
448             sbsec->behavior > ARRAY_SIZE(labeling_behaviors))
449                 sbsec->flags &= ~SE_SBLABELSUPP;
450 
451         /* Initialize the root inode. */
452         rc = inode_doinit_with_dentry(root_inode, root);
453 
454         /* Initialize any other inodes associated with the superblock, e.g.
455            inodes created prior to initial policy load or inodes created
456            during get_sb by a pseudo filesystem that directly
457            populates itself. */
458         spin_lock(&sbsec->isec_lock);
459 next_inode:
460         if (!list_empty(&sbsec->isec_head)) {
461                 struct inode_security_struct *isec =
462                                 list_entry(sbsec->isec_head.next,
463                                            struct inode_security_struct, list);
464                 struct inode *inode = isec->inode;
465                 spin_unlock(&sbsec->isec_lock);
466                 inode = igrab(inode);
467                 if (inode) {
468                         if (!IS_PRIVATE(inode))
469                                 inode_doinit(inode);
470                         iput(inode);
471                 }
472                 spin_lock(&sbsec->isec_lock);
473                 list_del_init(&isec->list);
474                 goto next_inode;
475         }
476         spin_unlock(&sbsec->isec_lock);
477 out:
478         return rc;
479 }
480 
481 /*
482  * This function should allow an FS to ask what it's mount security
483  * options were so it can use those later for submounts, displaying
484  * mount options, or whatever.
485  */
486 static int selinux_get_mnt_opts(const struct super_block *sb,
487                                 struct security_mnt_opts *opts)
488 {
489         int rc = 0, i;
490         struct superblock_security_struct *sbsec = sb->s_security;
491         char *context = NULL;
492         u32 len;
493         char tmp;
494 
495         security_init_mnt_opts(opts);
496 
497         if (!(sbsec->flags & SE_SBINITIALIZED))
498                 return -EINVAL;
499 
500         if (!ss_initialized)
501                 return -EINVAL;
502 
503         tmp = sbsec->flags & SE_MNTMASK;
504         /* count the number of mount options for this sb */
505         for (i = 0; i < 8; i++) {
506                 if (tmp & 0x01)
507                         opts->num_mnt_opts++;
508                 tmp >>= 1;
509         }
510         /* Check if the Label support flag is set */
511         if (sbsec->flags & SE_SBLABELSUPP)
512                 opts->num_mnt_opts++;
513 
514         opts->mnt_opts = kcalloc(opts->num_mnt_opts, sizeof(char *), GFP_ATOMIC);
515         if (!opts->mnt_opts) {
516                 rc = -ENOMEM;
517                 goto out_free;
518         }
519 
520         opts->mnt_opts_flags = kcalloc(opts->num_mnt_opts, sizeof(int), GFP_ATOMIC);
521         if (!opts->mnt_opts_flags) {
522                 rc = -ENOMEM;
523                 goto out_free;
524         }
525 
526         i = 0;
527         if (sbsec->flags & FSCONTEXT_MNT) {
528                 rc = security_sid_to_context(sbsec->sid, &context, &len);
529                 if (rc)
530                         goto out_free;
531                 opts->mnt_opts[i] = context;
532                 opts->mnt_opts_flags[i++] = FSCONTEXT_MNT;
533         }
534         if (sbsec->flags & CONTEXT_MNT) {
535                 rc = security_sid_to_context(sbsec->mntpoint_sid, &context, &len);
536                 if (rc)
537                         goto out_free;
538                 opts->mnt_opts[i] = context;
539                 opts->mnt_opts_flags[i++] = CONTEXT_MNT;
540         }
541         if (sbsec->flags & DEFCONTEXT_MNT) {
542                 rc = security_sid_to_context(sbsec->def_sid, &context, &len);
543                 if (rc)
544                         goto out_free;
545                 opts->mnt_opts[i] = context;
546                 opts->mnt_opts_flags[i++] = DEFCONTEXT_MNT;
547         }
548         if (sbsec->flags & ROOTCONTEXT_MNT) {
549                 struct inode *root = sbsec->sb->s_root->d_inode;
550                 struct inode_security_struct *isec = root->i_security;
551 
552                 rc = security_sid_to_context(isec->sid, &context, &len);
553                 if (rc)
554                         goto out_free;
555                 opts->mnt_opts[i] = context;
556                 opts->mnt_opts_flags[i++] = ROOTCONTEXT_MNT;
557         }
558         if (sbsec->flags & SE_SBLABELSUPP) {
559                 opts->mnt_opts[i] = NULL;
560                 opts->mnt_opts_flags[i++] = SE_SBLABELSUPP;
561         }
562 
563         BUG_ON(i != opts->num_mnt_opts);
564 
565         return 0;
566 
567 out_free:
568         security_free_mnt_opts(opts);
569         return rc;
570 }
571 
572 static int bad_option(struct superblock_security_struct *sbsec, char flag,
573                       u32 old_sid, u32 new_sid)
574 {
575         char mnt_flags = sbsec->flags & SE_MNTMASK;
576 
577         /* check if the old mount command had the same options */
578         if (sbsec->flags & SE_SBINITIALIZED)
579                 if (!(sbsec->flags & flag) ||
580                     (old_sid != new_sid))
581                         return 1;
582 
583         /* check if we were passed the same options twice,
584          * aka someone passed context=a,context=b
585          */
586         if (!(sbsec->flags & SE_SBINITIALIZED))
587                 if (mnt_flags & flag)
588                         return 1;
589         return 0;
590 }
591 
592 /*
593  * Allow filesystems with binary mount data to explicitly set mount point
594  * labeling information.
595  */
596 static int selinux_set_mnt_opts(struct super_block *sb,
597                                 struct security_mnt_opts *opts)
598 {
599         const struct cred *cred = current_cred();
600         int rc = 0, i;
601         struct superblock_security_struct *sbsec = sb->s_security;
602         const char *name = sb->s_type->name;
603         struct inode *inode = sbsec->sb->s_root->d_inode;
604         struct inode_security_struct *root_isec = inode->i_security;
605         u32 fscontext_sid = 0, context_sid = 0, rootcontext_sid = 0;
606         u32 defcontext_sid = 0;
607         char **mount_options = opts->mnt_opts;
608         int *flags = opts->mnt_opts_flags;
609         int num_opts = opts->num_mnt_opts;
610 
611         mutex_lock(&sbsec->lock);
612 
613         if (!ss_initialized) {
614                 if (!num_opts) {
615                         /* Defer initialization until selinux_complete_init,
616                            after the initial policy is loaded and the security
617                            server is ready to handle calls. */
618                         spin_lock(&sb_security_lock);
619                         if (list_empty(&sbsec->list))
620                                 list_add(&sbsec->list, &superblock_security_head);
621                         spin_unlock(&sb_security_lock);
622                         goto out;
623                 }
624                 rc = -EINVAL;
625                 printk(KERN_WARNING "SELinux: Unable to set superblock options "
626                         "before the security server is initialized\n");
627                 goto out;
628         }
629 
630         /*
631          * Binary mount data FS will come through this function twice.  Once
632          * from an explicit call and once from the generic calls from the vfs.
633          * Since the generic VFS calls will not contain any security mount data
634          * we need to skip the double mount verification.
635          *
636          * This does open a hole in which we will not notice if the first
637          * mount using this sb set explict options and a second mount using
638          * this sb does not set any security options.  (The first options
639          * will be used for both mounts)
640          */
641         if ((sbsec->flags & SE_SBINITIALIZED) && (sb->s_type->fs_flags & FS_BINARY_MOUNTDATA)
642             && (num_opts == 0))
643                 goto out;
644 
645         /*
646          * parse the mount options, check if they are valid sids.
647          * also check if someone is trying to mount the same sb more
648          * than once with different security options.
649          */
650         for (i = 0; i < num_opts; i++) {
651                 u32 sid;
652 
653                 if (flags[i] == SE_SBLABELSUPP)
654                         continue;
655                 rc = security_context_to_sid(mount_options[i],
656                                              strlen(mount_options[i]), &sid);
657                 if (rc) {
658                         printk(KERN_WARNING "SELinux: security_context_to_sid"
659                                "(%s) failed for (dev %s, type %s) errno=%d\n",
660                                mount_options[i], sb->s_id, name, rc);
661                         goto out;
662                 }
663                 switch (flags[i]) {
664                 case FSCONTEXT_MNT:
665                         fscontext_sid = sid;
666 
667                         if (bad_option(sbsec, FSCONTEXT_MNT, sbsec->sid,
668                                         fscontext_sid))
669                                 goto out_double_mount;
670 
671                         sbsec->flags |= FSCONTEXT_MNT;
672                         break;
673                 case CONTEXT_MNT:
674                         context_sid = sid;
675 
676                         if (bad_option(sbsec, CONTEXT_MNT, sbsec->mntpoint_sid,
677                                         context_sid))
678                                 goto out_double_mount;
679 
680                         sbsec->flags |= CONTEXT_MNT;
681                         break;
682                 case ROOTCONTEXT_MNT:
683                         rootcontext_sid = sid;
684 
685                         if (bad_option(sbsec, ROOTCONTEXT_MNT, root_isec->sid,
686                                         rootcontext_sid))
687                                 goto out_double_mount;
688 
689                         sbsec->flags |= ROOTCONTEXT_MNT;
690 
691                         break;
692                 case DEFCONTEXT_MNT:
693                         defcontext_sid = sid;
694 
695                         if (bad_option(sbsec, DEFCONTEXT_MNT, sbsec->def_sid,
696                                         defcontext_sid))
697                                 goto out_double_mount;
698 
699                         sbsec->flags |= DEFCONTEXT_MNT;
700 
701                         break;
702                 default:
703                         rc = -EINVAL;
704                         goto out;
705                 }
706         }
707 
708         if (sbsec->flags & SE_SBINITIALIZED) {
709                 /* previously mounted with options, but not on this attempt? */
710                 if ((sbsec->flags & SE_MNTMASK) && !num_opts)
711                         goto out_double_mount;
712                 rc = 0;
713                 goto out;
714         }
715 
716         if (strcmp(sb->s_type->name, "proc") == 0)
717                 sbsec->flags |= SE_SBPROC;
718 
719         /* Determine the labeling behavior to use for this filesystem type. */
720         rc = security_fs_use((sbsec->flags & SE_SBPROC) ? "proc" : sb->s_type->name, &sbsec->behavior, &sbsec->sid);
721         if (rc) {
722                 printk(KERN_WARNING "%s: security_fs_use(%s) returned %d\n",
723                        __func__, sb->s_type->name, rc);
724                 goto out;
725         }
726 
727         /* sets the context of the superblock for the fs being mounted. */
728         if (fscontext_sid) {
729                 rc = may_context_mount_sb_relabel(fscontext_sid, sbsec, cred);
730                 if (rc)
731                         goto out;
732 
733                 sbsec->sid = fscontext_sid;
734         }
735 
736         /*
737          * Switch to using mount point labeling behavior.
738          * sets the label used on all file below the mountpoint, and will set
739          * the superblock context if not already set.
740          */
741         if (context_sid) {
742                 if (!fscontext_sid) {
743                         rc = may_context_mount_sb_relabel(context_sid, sbsec,
744                                                           cred);
745                         if (rc)
746                                 goto out;
747                         sbsec->sid = context_sid;
748                 } else {
749                         rc = may_context_mount_inode_relabel(context_sid, sbsec,
750                                                              cred);
751                         if (rc)
752                                 goto out;
753                 }
754                 if (!rootcontext_sid)
755                         rootcontext_sid = context_sid;
756 
757                 sbsec->mntpoint_sid = context_sid;
758                 sbsec->behavior = SECURITY_FS_USE_MNTPOINT;
759         }
760 
761         if (rootcontext_sid) {
762                 rc = may_context_mount_inode_relabel(rootcontext_sid, sbsec,
763                                                      cred);
764                 if (rc)
765                         goto out;
766 
767                 root_isec->sid = rootcontext_sid;
768                 root_isec->initialized = 1;
769         }
770 
771         if (defcontext_sid) {
772                 if (sbsec->behavior != SECURITY_FS_USE_XATTR) {
773                         rc = -EINVAL;
774                         printk(KERN_WARNING "SELinux: defcontext option is "
775                                "invalid for this filesystem type\n");
776                         goto out;
777                 }
778 
779                 if (defcontext_sid != sbsec->def_sid) {
780                         rc = may_context_mount_inode_relabel(defcontext_sid,
781                                                              sbsec, cred);
782                         if (rc)
783                                 goto out;
784                 }
785 
786                 sbsec->def_sid = defcontext_sid;
787         }
788 
789         rc = sb_finish_set_opts(sb);
790 out:
791         mutex_unlock(&sbsec->lock);
792         return rc;
793 out_double_mount:
794         rc = -EINVAL;
795         printk(KERN_WARNING "SELinux: mount invalid.  Same superblock, different "
796                "security settings for (dev %s, type %s)\n", sb->s_id, name);
797         goto out;
798 }
799 
800 static void selinux_sb_clone_mnt_opts(const struct super_block *oldsb,
801                                         struct super_block *newsb)
802 {
803         const struct superblock_security_struct *oldsbsec = oldsb->s_security;
804         struct superblock_security_struct *newsbsec = newsb->s_security;
805 
806         int set_fscontext =     (oldsbsec->flags & FSCONTEXT_MNT);
807         int set_context =       (oldsbsec->flags & CONTEXT_MNT);
808         int set_rootcontext =   (oldsbsec->flags & ROOTCONTEXT_MNT);
809 
810         /*
811          * if the parent was able to be mounted it clearly had no special lsm
812          * mount options.  thus we can safely put this sb on the list and deal
813          * with it later
814          */
815         if (!ss_initialized) {
816                 spin_lock(&sb_security_lock);
817                 if (list_empty(&newsbsec->list))
818                         list_add(&newsbsec->list, &superblock_security_head);
819                 spin_unlock(&sb_security_lock);
820                 return;
821         }
822 
823         /* how can we clone if the old one wasn't set up?? */
824         BUG_ON(!(oldsbsec->flags & SE_SBINITIALIZED));
825 
826         /* if fs is reusing a sb, just let its options stand... */
827         if (newsbsec->flags & SE_SBINITIALIZED)
828                 return;
829 
830         mutex_lock(&newsbsec->lock);
831 
832         newsbsec->flags = oldsbsec->flags;
833 
834         newsbsec->sid = oldsbsec->sid;
835         newsbsec->def_sid = oldsbsec->def_sid;
836         newsbsec->behavior = oldsbsec->behavior;
837 
838         if (set_context) {
839                 u32 sid = oldsbsec->mntpoint_sid;
840 
841                 if (!set_fscontext)
842                         newsbsec->sid = sid;
843                 if (!set_rootcontext) {
844                         struct inode *newinode = newsb->s_root->d_inode;
845                         struct inode_security_struct *newisec = newinode->i_security;
846                         newisec->sid = sid;
847                 }
848                 newsbsec->mntpoint_sid = sid;
849         }
850         if (set_rootcontext) {
851                 const struct inode *oldinode = oldsb->s_root->d_inode;
852                 const struct inode_security_struct *oldisec = oldinode->i_security;
853                 struct inode *newinode = newsb->s_root->d_inode;
854                 struct inode_security_struct *newisec = newinode->i_security;
855 
856                 newisec->sid = oldisec->sid;
857         }
858 
859         sb_finish_set_opts(newsb);
860         mutex_unlock(&newsbsec->lock);
861 }
862 
863 static int selinux_parse_opts_str(char *options,
864                                   struct security_mnt_opts *opts)
865 {
866         char *p;
867         char *context = NULL, *defcontext = NULL;
868         char *fscontext = NULL, *rootcontext = NULL;
869         int rc, num_mnt_opts = 0;
870 
871         opts->num_mnt_opts = 0;
872 
873         /* Standard string-based options. */
874         while ((p = strsep(&options, "|")) != NULL) {
875                 int token;
876                 substring_t args[MAX_OPT_ARGS];
877 
878                 if (!*p)
879                         continue;
880 
881                 token = match_token(p, tokens, args);
882 
883                 switch (token) {
884                 case Opt_context:
885                         if (context || defcontext) {
886                                 rc = -EINVAL;
887                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
888                                 goto out_err;
889                         }
890                         context = match_strdup(&args[0]);
891                         if (!context) {
892                                 rc = -ENOMEM;
893                                 goto out_err;
894                         }
895                         break;
896 
897                 case Opt_fscontext:
898                         if (fscontext) {
899                                 rc = -EINVAL;
900                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
901                                 goto out_err;
902                         }
903                         fscontext = match_strdup(&args[0]);
904                         if (!fscontext) {
905                                 rc = -ENOMEM;
906                                 goto out_err;
907                         }
908                         break;
909 
910                 case Opt_rootcontext:
911                         if (rootcontext) {
912                                 rc = -EINVAL;
913                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
914                                 goto out_err;
915                         }
916                         rootcontext = match_strdup(&args[0]);
917                         if (!rootcontext) {
918                                 rc = -ENOMEM;
919                                 goto out_err;
920                         }
921                         break;
922 
923                 case Opt_defcontext:
924                         if (context || defcontext) {
925                                 rc = -EINVAL;
926                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
927                                 goto out_err;
928                         }
929                         defcontext = match_strdup(&args[0]);
930                         if (!defcontext) {
931                                 rc = -ENOMEM;
932                                 goto out_err;
933                         }
934                         break;
935                 case Opt_labelsupport:
936                         break;
937                 default:
938                         rc = -EINVAL;
939                         printk(KERN_WARNING "SELinux:  unknown mount option\n");
940                         goto out_err;
941 
942                 }
943         }
944 
945         rc = -ENOMEM;
946         opts->mnt_opts = kcalloc(NUM_SEL_MNT_OPTS, sizeof(char *), GFP_ATOMIC);
947         if (!opts->mnt_opts)
948                 goto out_err;
949 
950         opts->mnt_opts_flags = kcalloc(NUM_SEL_MNT_OPTS, sizeof(int), GFP_ATOMIC);
951         if (!opts->mnt_opts_flags) {
952                 kfree(opts->mnt_opts);
953                 goto out_err;
954         }
955 
956         if (fscontext) {
957                 opts->mnt_opts[num_mnt_opts] = fscontext;
958                 opts->mnt_opts_flags[num_mnt_opts++] = FSCONTEXT_MNT;
959         }
960         if (context) {
961                 opts->mnt_opts[num_mnt_opts] = context;
962                 opts->mnt_opts_flags[num_mnt_opts++] = CONTEXT_MNT;
963         }
964         if (rootcontext) {
965                 opts->mnt_opts[num_mnt_opts] = rootcontext;
966                 opts->mnt_opts_flags[num_mnt_opts++] = ROOTCONTEXT_MNT;
967         }
968         if (defcontext) {
969                 opts->mnt_opts[num_mnt_opts] = defcontext;
970                 opts->mnt_opts_flags[num_mnt_opts++] = DEFCONTEXT_MNT;
971         }
972 
973         opts->num_mnt_opts = num_mnt_opts;
974         return 0;
975 
976 out_err:
977         kfree(context);
978         kfree(defcontext);
979         kfree(fscontext);
980         kfree(rootcontext);
981         return rc;
982 }
983 /*
984  * string mount options parsing and call set the sbsec
985  */
986 static int superblock_doinit(struct super_block *sb, void *data)
987 {
988         int rc = 0;
989         char *options = data;
990         struct security_mnt_opts opts;
991 
992         security_init_mnt_opts(&opts);
993 
994         if (!data)
995                 goto out;
996 
997         BUG_ON(sb->s_type->fs_flags & FS_BINARY_MOUNTDATA);
998 
999         rc = selinux_parse_opts_str(options, &opts);
1000         if (rc)
1001                 goto out_err;
1002 
1003 out:
1004         rc = selinux_set_mnt_opts(sb, &opts);
1005 
1006 out_err:
1007         security_free_mnt_opts(&opts);
1008         return rc;
1009 }
1010 
1011 static void selinux_write_opts(struct seq_file *m,
1012                                struct security_mnt_opts *opts)
1013 {
1014         int i;
1015         char *prefix;
1016 
1017         for (i = 0; i < opts->num_mnt_opts; i++) {
1018                 char *has_comma;
1019 
1020                 if (opts->mnt_opts[i])
1021                         has_comma = strchr(opts->mnt_opts[i], ',');
1022                 else
1023                         has_comma = NULL;
1024 
1025                 switch (opts->mnt_opts_flags[i]) {
1026                 case CONTEXT_MNT:
1027                         prefix = CONTEXT_STR;
1028                         break;
1029                 case FSCONTEXT_MNT:
1030                         prefix = FSCONTEXT_STR;
1031                         break;
1032                 case ROOTCONTEXT_MNT:
1033                         prefix = ROOTCONTEXT_STR;
1034                         break;
1035                 case DEFCONTEXT_MNT:
1036                         prefix = DEFCONTEXT_STR;
1037                         break;
1038                 case SE_SBLABELSUPP:
1039                         seq_putc(m, ',');
1040                         seq_puts(m, LABELSUPP_STR);
1041                         continue;
1042                 default:
1043                         BUG();
1044                 };
1045                 /* we need a comma before each option */
1046                 seq_putc(m, ',');
1047                 seq_puts(m, prefix);
1048                 if (has_comma)
1049                         seq_putc(m, '\"');
1050                 seq_puts(m, opts->mnt_opts[i]);
1051                 if (has_comma)
1052                         seq_putc(m, '\"');
1053         }
1054 }
1055 
1056 static int selinux_sb_show_options(struct seq_file *m, struct super_block *sb)
1057 {
1058         struct security_mnt_opts opts;
1059         int rc;
1060 
1061         rc = selinux_get_mnt_opts(sb, &opts);
1062         if (rc) {
1063                 /* before policy load we may get EINVAL, don't show anything */
1064                 if (rc == -EINVAL)
1065                         rc = 0;
1066                 return rc;
1067         }
1068 
1069         selinux_write_opts(m, &opts);
1070 
1071         security_free_mnt_opts(&opts);
1072 
1073         return rc;
1074 }
1075 
1076 static inline u16 inode_mode_to_security_class(umode_t mode)
1077 {
1078         switch (mode & S_IFMT) {
1079         case S_IFSOCK:
1080                 return SECCLASS_SOCK_FILE;
1081         case S_IFLNK:
1082                 return SECCLASS_LNK_FILE;
1083         case S_IFREG:
1084                 return SECCLASS_FILE;
1085         case S_IFBLK:
1086                 return SECCLASS_BLK_FILE;
1087         case S_IFDIR:
1088                 return SECCLASS_DIR;
1089         case S_IFCHR:
1090                 return SECCLASS_CHR_FILE;
1091         case S_IFIFO:
1092                 return SECCLASS_FIFO_FILE;
1093 
1094         }
1095 
1096         return SECCLASS_FILE;
1097 }
1098 
1099 static inline int default_protocol_stream(int protocol)
1100 {
1101         return (protocol == IPPROTO_IP || protocol == IPPROTO_TCP);
1102 }
1103 
1104 static inline int default_protocol_dgram(int protocol)
1105 {
1106         return (protocol == IPPROTO_IP || protocol == IPPROTO_UDP);
1107 }
1108 
1109 static inline u16 socket_type_to_security_class(int family, int type, int protocol)
1110 {
1111         switch (family) {
1112         case PF_UNIX:
1113                 switch (type) {
1114                 case SOCK_STREAM:
1115                 case SOCK_SEQPACKET:
1116                         return SECCLASS_UNIX_STREAM_SOCKET;
1117                 case SOCK_DGRAM:
1118                         return SECCLASS_UNIX_DGRAM_SOCKET;
1119                 }
1120                 break;
1121         case PF_INET:
1122         case PF_INET6:
1123                 switch (type) {
1124                 case SOCK_STREAM:
1125                         if (default_protocol_stream(protocol))
1126                                 return SECCLASS_TCP_SOCKET;
1127                         else
1128                                 return SECCLASS_RAWIP_SOCKET;
1129                 case SOCK_DGRAM:
1130                         if (default_protocol_dgram(protocol))
1131                                 return SECCLASS_UDP_SOCKET;
1132                         else
1133                                 return SECCLASS_RAWIP_SOCKET;
1134                 case SOCK_DCCP:
1135                         return SECCLASS_DCCP_SOCKET;
1136                 default:
1137                         return SECCLASS_RAWIP_SOCKET;
1138                 }
1139                 break;
1140         case PF_NETLINK:
1141                 switch (protocol) {
1142                 case NETLINK_ROUTE:
1143                         return SECCLASS_NETLINK_ROUTE_SOCKET;
1144                 case NETLINK_FIREWALL:
1145                         return SECCLASS_NETLINK_FIREWALL_SOCKET;
1146                 case NETLINK_INET_DIAG:
1147                         return SECCLASS_NETLINK_TCPDIAG_SOCKET;
1148                 case NETLINK_NFLOG:
1149                         return SECCLASS_NETLINK_NFLOG_SOCKET;
1150                 case NETLINK_XFRM:
1151                         return SECCLASS_NETLINK_XFRM_SOCKET;
1152                 case NETLINK_SELINUX:
1153                         return SECCLASS_NETLINK_SELINUX_SOCKET;
1154                 case NETLINK_AUDIT:
1155                         return SECCLASS_NETLINK_AUDIT_SOCKET;
1156                 case NETLINK_IP6_FW:
1157                         return SECCLASS_NETLINK_IP6FW_SOCKET;
1158                 case NETLINK_DNRTMSG:
1159                         return SECCLASS_NETLINK_DNRT_SOCKET;
1160                 case NETLINK_KOBJECT_UEVENT:
1161                         return SECCLASS_NETLINK_KOBJECT_UEVENT_SOCKET;
1162                 default:
1163                         return SECCLASS_NETLINK_SOCKET;
1164                 }
1165         case PF_PACKET:
1166                 return SECCLASS_PACKET_SOCKET;
1167         case PF_KEY:
1168                 return SECCLASS_KEY_SOCKET;
1169         case PF_APPLETALK:
1170                 return SECCLASS_APPLETALK_SOCKET;
1171         }
1172 
1173         return SECCLASS_SOCKET;
1174 }
1175 
1176 #ifdef CONFIG_PROC_FS
1177 static int selinux_proc_get_sid(struct proc_dir_entry *de,
1178                                 u16 tclass,
1179                                 u32 *sid)
1180 {
1181         int buflen, rc;
1182         char *buffer, *path, *end;
1183 
1184         buffer = (char *)__get_free_page(GFP_KERNEL);
1185         if (!buffer)
1186                 return -ENOMEM;
1187 
1188         buflen = PAGE_SIZE;
1189         end = buffer+buflen;
1190         *--end = '\0';
1191         buflen--;
1192         path = end-1;
1193         *path = '/';
1194         while (de && de != de->parent) {
1195                 buflen -= de->namelen + 1;
1196                 if (buflen < 0)
1197                         break;
1198                 end -= de->namelen;
1199                 memcpy(end, de->name, de->namelen);
1200                 *--end = '/';
1201                 path = end;
1202                 de = de->parent;
1203         }
1204         rc = security_genfs_sid("proc", path, tclass, sid);
1205         free_page((unsigned long)buffer);
1206         return rc;
1207 }
1208 #else
1209 static int selinux_proc_get_sid(struct proc_dir_entry *de,
1210                                 u16 tclass,
1211                                 u32 *sid)
1212 {
1213         return -EINVAL;
1214 }
1215 #endif
1216 
1217 /* The inode's security attributes must be initialized before first use. */
1218 static int inode_doinit_with_dentry(struct inode *inode, struct dentry *opt_dentry)
1219 {
1220         struct superblock_security_struct *sbsec = NULL;
1221         struct inode_security_struct *isec = inode->i_security;
1222         u32 sid;
1223         struct dentry *dentry;
1224 #define INITCONTEXTLEN 255
1225         char *context = NULL;
1226         unsigned len = 0;
1227         int rc = 0;
1228 
1229         if (isec->initialized)
1230                 goto out;
1231 
1232         mutex_lock(&isec->lock);
1233         if (isec->initialized)
1234                 goto out_unlock;
1235 
1236         sbsec = inode->i_sb->s_security;
1237         if (!(sbsec->flags & SE_SBINITIALIZED)) {
1238                 /* Defer initialization until selinux_complete_init,
1239                    after the initial policy is loaded and the security
1240                    server is ready to handle calls. */
1241                 spin_lock(&sbsec->isec_lock);
1242                 if (list_empty(&isec->list))
1243                         list_add(&isec->list, &sbsec->isec_head);
1244                 spin_unlock(&sbsec->isec_lock);
1245                 goto out_unlock;
1246         }
1247 
1248         switch (sbsec->behavior) {
1249         case SECURITY_FS_USE_XATTR:
1250                 if (!inode->i_op->getxattr) {
1251                         isec->sid = sbsec->def_sid;
1252                         break;
1253                 }
1254 
1255                 /* Need a dentry, since the xattr API requires one.
1256                    Life would be simpler if we could just pass the inode. */
1257                 if (opt_dentry) {
1258                         /* Called from d_instantiate or d_splice_alias. */
1259                         dentry = dget(opt_dentry);
1260                 } else {
1261                         /* Called from selinux_complete_init, try to find a dentry. */
1262                         dentry = d_find_alias(inode);
1263                 }
1264                 if (!dentry) {
1265                         /*
1266                          * this is can be hit on boot when a file is accessed
1267                          * before the policy is loaded.  When we load policy we
1268                          * may find inodes that have no dentry on the
1269                          * sbsec->isec_head list.  No reason to complain as these
1270                          * will get fixed up the next time we go through
1271                          * inode_doinit with a dentry, before these inodes could
1272                          * be used again by userspace.
1273                          */
1274                         goto out_unlock;
1275                 }
1276 
1277                 len = INITCONTEXTLEN;
1278                 context = kmalloc(len+1, GFP_NOFS);
1279                 if (!context) {
1280                         rc = -ENOMEM;
1281                         dput(dentry);
1282                         goto out_unlock;
1283                 }
1284                 context[len] = '\0';
1285                 rc = inode->i_op->getxattr(dentry, XATTR_NAME_SELINUX,
1286                                            context, len);
1287                 if (rc == -ERANGE) {
1288                         kfree(context);
1289 
1290                         /* Need a larger buffer.  Query for the right size. */
1291                         rc = inode->i_op->getxattr(dentry, XATTR_NAME_SELINUX,
1292                                                    NULL, 0);
1293                         if (rc < 0) {
1294                                 dput(dentry);
1295                                 goto out_unlock;
1296                         }
1297                         len = rc;
1298                         context = kmalloc(len+1, GFP_NOFS);
1299                         if (!context) {
1300                                 rc = -ENOMEM;
1301                                 dput(dentry);
1302                                 goto out_unlock;
1303                         }
1304                         context[len] = '\0';
1305                         rc = inode->i_op->getxattr(dentry,
1306                                                    XATTR_NAME_SELINUX,
1307                                                    context, len);
1308                 }
1309                 dput(dentry);
1310                 if (rc < 0) {
1311                         if (rc != -ENODATA) {
1312                                 printk(KERN_WARNING "SELinux: %s:  getxattr returned "
1313                                        "%d for dev=%s ino=%ld\n", __func__,
1314                                        -rc, inode->i_sb->s_id, inode->i_ino);
1315                                 kfree(context);
1316                                 goto out_unlock;
1317                         }
1318                         /* Map ENODATA to the default file SID */
1319                         sid = sbsec->def_sid;
1320                         rc = 0;
1321                 } else {
1322                         rc = security_context_to_sid_default(context, rc, &sid,
1323                                                              sbsec->def_sid,
1324                                                              GFP_NOFS);
1325                         if (rc) {
1326                                 char *dev = inode->i_sb->s_id;
1327                                 unsigned long ino = inode->i_ino;
1328 
1329                                 if (rc == -EINVAL) {
1330                                         if (printk_ratelimit())
1331                                                 printk(KERN_NOTICE "SELinux: inode=%lu on dev=%s was found to have an invalid "
1332                                                         "context=%s.  This indicates you may need to relabel the inode or the "
1333                                                         "filesystem in question.\n", ino, dev, context);
1334                                 } else {
1335                                         printk(KERN_WARNING "SELinux: %s:  context_to_sid(%s) "
1336                                                "returned %d for dev=%s ino=%ld\n",
1337                                                __func__, context, -rc, dev, ino);
1338                                 }
1339                                 kfree(context);
1340                                 /* Leave with the unlabeled SID */
1341                                 rc = 0;
1342                                 break;
1343                         }
1344                 }
1345                 kfree(context);
1346                 isec->sid = sid;
1347                 break;
1348         case SECURITY_FS_USE_TASK:
1349                 isec->sid = isec->task_sid;
1350                 break;
1351         case SECURITY_FS_USE_TRANS:
1352                 /* Default to the fs SID. */
1353                 isec->sid = sbsec->sid;
1354 
1355                 /* Try to obtain a transition SID. */
1356                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
1357                 rc = security_transition_sid(isec->task_sid,
1358                                              sbsec->sid,
1359                                              isec->sclass,
1360                                              &sid);
1361                 if (rc)
1362                         goto out_unlock;
1363                 isec->sid = sid;
1364                 break;
1365         case SECURITY_FS_USE_MNTPOINT:
1366                 isec->sid = sbsec->mntpoint_sid;
1367                 break;
1368         default:
1369                 /* Default to the fs superblock SID. */
1370                 isec->sid = sbsec->sid;
1371 
1372                 if ((sbsec->flags & SE_SBPROC) && !S_ISLNK(inode->i_mode)) {
1373                         struct proc_inode *proci = PROC_I(inode);
1374                         if (proci->pde) {
1375                                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
1376                                 rc = selinux_proc_get_sid(proci->pde,
1377                                                           isec->sclass,
1378                                                           &sid);
1379                                 if (rc)
1380                                         goto out_unlock;
1381                                 isec->sid = sid;
1382                         }
1383                 }
1384                 break;
1385         }
1386 
1387         isec->initialized = 1;
1388 
1389 out_unlock:
1390         mutex_unlock(&isec->lock);
1391 out:
1392         if (isec->sclass == SECCLASS_FILE)
1393                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
1394         return rc;
1395 }
1396 
1397 /* Convert a Linux signal to an access vector. */
1398 static inline u32 signal_to_av(int sig)
1399 {
1400         u32 perm = 0;
1401 
1402         switch (sig) {
1403         case SIGCHLD:
1404                 /* Commonly granted from child to parent. */
1405                 perm = PROCESS__SIGCHLD;
1406                 break;
1407         case SIGKILL:
1408                 /* Cannot be caught or ignored */
1409                 perm = PROCESS__SIGKILL;
1410                 break;
1411         case SIGSTOP:
1412                 /* Cannot be caught or ignored */
1413                 perm = PROCESS__SIGSTOP;
1414                 break;
1415         default:
1416                 /* All other signals. */
1417                 perm = PROCESS__SIGNAL;
1418                 break;
1419         }
1420 
1421         return perm;
1422 }
1423 
1424 /*
1425  * Check permission between a pair of credentials
1426  * fork check, ptrace check, etc.
1427  */
1428 static int cred_has_perm(const struct cred *actor,
1429                          const struct cred *target,
1430                          u32 perms)
1431 {
1432         u32 asid = cred_sid(actor), tsid = cred_sid(target);
1433 
1434         return avc_has_perm(asid, tsid, SECCLASS_PROCESS, perms, NULL);
1435 }
1436 
1437 /*
1438  * Check permission between a pair of tasks, e.g. signal checks,
1439  * fork check, ptrace check, etc.
1440  * tsk1 is the actor and tsk2 is the target
1441  * - this uses the default subjective creds of tsk1
1442  */
1443 static int task_has_perm(const struct task_struct *tsk1,
1444                          const struct task_struct *tsk2,
1445                          u32 perms)
1446 {
1447         const struct task_security_struct *__tsec1, *__tsec2;
1448         u32 sid1, sid2;
1449 
1450         rcu_read_lock();
1451         __tsec1 = __task_cred(tsk1)->security;  sid1 = __tsec1->sid;
1452         __tsec2 = __task_cred(tsk2)->security;  sid2 = __tsec2->sid;
1453         rcu_read_unlock();
1454         return avc_has_perm(sid1, sid2, SECCLASS_PROCESS, perms, NULL);
1455 }
1456 
1457 /*
1458  * Check permission between current and another task, e.g. signal checks,
1459  * fork check, ptrace check, etc.
1460  * current is the actor and tsk2 is the target
1461  * - this uses current's subjective creds
1462  */
1463 static int current_has_perm(const struct task_struct *tsk,
1464                             u32 perms)
1465 {
1466         u32 sid, tsid;
1467 
1468         sid = current_sid();
1469         tsid = task_sid(tsk);
1470         return avc_has_perm(sid, tsid, SECCLASS_PROCESS, perms, NULL);
1471 }
1472 
1473 #if CAP_LAST_CAP > 63
1474 #error Fix SELinux to handle capabilities > 63.
1475 #endif
1476 
1477 /* Check whether a task is allowed to use a capability. */
1478 static int task_has_capability(struct task_struct *tsk,
1479                                const struct cred *cred,
1480                                int cap, int audit)
1481 {
1482         struct avc_audit_data ad;
1483         struct av_decision avd;
1484         u16 sclass;
1485         u32 sid = cred_sid(cred);
1486         u32 av = CAP_TO_MASK(cap);
1487         int rc;
1488 
1489         AVC_AUDIT_DATA_INIT(&ad, CAP);
1490         ad.tsk = tsk;
1491         ad.u.cap = cap;
1492 
1493         switch (CAP_TO_INDEX(cap)) {
1494         case 0:
1495                 sclass = SECCLASS_CAPABILITY;
1496                 break;
1497         case 1:
1498                 sclass = SECCLASS_CAPABILITY2;
1499                 break;
1500         default:
1501                 printk(KERN_ERR
1502                        "SELinux:  out of range capability %d\n", cap);
1503                 BUG();
1504         }
1505 
1506         rc = avc_has_perm_noaudit(sid, sid, sclass, av, 0, &avd);
1507         if (audit == SECURITY_CAP_AUDIT)
1508                 avc_audit(sid, sid, sclass, av, &avd, rc, &ad);
1509         return rc;
1510 }
1511 
1512 /* Check whether a task is allowed to use a system operation. */
1513 static int task_has_system(struct task_struct *tsk,
1514                            u32 perms)
1515 {
1516         u32 sid = task_sid(tsk);
1517 
1518         return avc_has_perm(sid, SECINITSID_KERNEL,
1519                             SECCLASS_SYSTEM, perms, NULL);
1520 }
1521 
1522 /* Check whether a task has a particular permission to an inode.
1523    The 'adp' parameter is optional and allows other audit
1524    data to be passed (e.g. the dentry). */
1525 static int inode_has_perm(const struct cred *cred,
1526                           struct inode *inode,
1527                           u32 perms,
1528                           struct avc_audit_data *adp)
1529 {
1530         struct inode_security_struct *isec;
1531         struct avc_audit_data ad;
1532         u32 sid;
1533 
1534         if (unlikely(IS_PRIVATE(inode)))
1535                 return 0;
1536 
1537         sid = cred_sid(cred);
1538         isec = inode->i_security;
1539 
1540         if (!adp) {
1541                 adp = &ad;
1542                 AVC_AUDIT_DATA_INIT(&ad, FS);
1543                 ad.u.fs.inode = inode;
1544         }
1545 
1546         return avc_has_perm(sid, isec->sid, isec->sclass, perms, adp);
1547 }
1548 
1549 /* Same as inode_has_perm, but pass explicit audit data containing
1550    the dentry to help the auditing code to more easily generate the
1551    pathname if needed. */
1552 static inline int dentry_has_perm(const struct cred *cred,
1553                                   struct vfsmount *mnt,
1554                                   struct dentry *dentry,
1555                                   u32 av)
1556 {
1557         struct inode *inode = dentry->d_inode;
1558         struct avc_audit_data ad;
1559 
1560         AVC_AUDIT_DATA_INIT(&ad, FS);
1561         ad.u.fs.path.mnt = mnt;
1562         ad.u.fs.path.dentry = dentry;
1563         return inode_has_perm(cred, inode, av, &ad);
1564 }
1565 
1566 /* Check whether a task can use an open file descriptor to
1567    access an inode in a given way.  Check access to the
1568    descriptor itself, and then use dentry_has_perm to
1569    check a particular permission to the file.
1570    Access to the descriptor is implicitly granted if it
1571    has the same SID as the process.  If av is zero, then
1572    access to the file is not checked, e.g. for cases
1573    where only the descriptor is affected like seek. */
1574 static int file_has_perm(const struct cred *cred,
1575                          struct file *file,
1576                          u32 av)
1577 {
1578         struct file_security_struct *fsec = file->f_security;
1579         struct inode *inode = file->f_path.dentry->d_inode;
1580         struct avc_audit_data ad;
1581         u32 sid = cred_sid(cred);
1582         int rc;
1583 
1584         AVC_AUDIT_DATA_INIT(&ad, FS);
1585         ad.u.fs.path = file->f_path;
1586 
1587         if (sid != fsec->sid) {
1588                 rc = avc_has_perm(sid, fsec->sid,
1589                                   SECCLASS_FD,
1590                                   FD__USE,
1591                                   &ad);
1592                 if (rc)
1593                         goto out;
1594         }
1595 
1596         /* av is zero if only checking access to the descriptor. */
1597         rc = 0;
1598         if (av)
1599                 rc = inode_has_perm(cred, inode, av, &ad);
1600 
1601 out:
1602         return rc;
1603 }
1604 
1605 /* Check whether a task can create a file. */
1606 static int may_create(struct inode *dir,
1607                       struct dentry *dentry,
1608                       u16 tclass)
1609 {
1610         const struct cred *cred = current_cred();
1611         const struct task_security_struct *tsec = cred->security;
1612         struct inode_security_struct *dsec;
1613         struct superblock_security_struct *sbsec;
1614         u32 sid, newsid;
1615         struct avc_audit_data ad;
1616         int rc;
1617 
1618         dsec = dir->i_security;
1619         sbsec = dir->i_sb->s_security;
1620 
1621         sid = tsec->sid;
1622         newsid = tsec->create_sid;
1623 
1624         AVC_AUDIT_DATA_INIT(&ad, FS);
1625         ad.u.fs.path.dentry = dentry;
1626 
1627         rc = avc_has_perm(sid, dsec->sid, SECCLASS_DIR,
1628                           DIR__ADD_NAME | DIR__SEARCH,
1629                           &ad);
1630         if (rc)
1631                 return rc;
1632 
1633         if (!newsid || !(sbsec->flags & SE_SBLABELSUPP)) {
1634                 rc = security_transition_sid(sid, dsec->sid, tclass, &newsid);
1635                 if (rc)
1636                         return rc;
1637         }
1638 
1639         rc = avc_has_perm(sid, newsid, tclass, FILE__CREATE, &ad);
1640         if (rc)
1641                 return rc;
1642 
1643         return avc_has_perm(newsid, sbsec->sid,
1644                             SECCLASS_FILESYSTEM,
1645                             FILESYSTEM__ASSOCIATE, &ad);
1646 }
1647 
1648 /* Check whether a task can create a key. */
1649 static int may_create_key(u32 ksid,
1650                           struct task_struct *ctx)
1651 {
1652         u32 sid = task_sid(ctx);
1653 
1654         return avc_has_perm(sid, ksid, SECCLASS_KEY, KEY__CREATE, NULL);
1655 }
1656 
1657 #define MAY_LINK        0
1658 #define MAY_UNLINK      1
1659 #define MAY_RMDIR       2
1660 
1661 /* Check whether a task can link, unlink, or rmdir a file/directory. */
1662 static int may_link(struct inode *dir,
1663                     struct dentry *dentry,
1664                     int kind)
1665 
1666 {
1667         struct inode_security_struct *dsec, *isec;
1668         struct avc_audit_data ad;
1669         u32 sid = current_sid();
1670         u32 av;
1671         int rc;
1672 
1673         dsec = dir->i_security;
1674         isec = dentry->d_inode->i_security;
1675 
1676         AVC_AUDIT_DATA_INIT(&ad, FS);
1677         ad.u.fs.path.dentry = dentry;
1678 
1679         av = DIR__SEARCH;
1680         av |= (kind ? DIR__REMOVE_NAME : DIR__ADD_NAME);
1681         rc = avc_has_perm(sid, dsec->sid, SECCLASS_DIR, av, &ad);
1682         if (rc)
1683                 return rc;
1684 
1685         switch (kind) {
1686         case MAY_LINK:
1687                 av = FILE__LINK;
1688                 break;
1689         case MAY_UNLINK:
1690                 av = FILE__UNLINK;
1691                 break;
1692         case MAY_RMDIR:
1693                 av = DIR__RMDIR;
1694                 break;
1695         default:
1696                 printk(KERN_WARNING "SELinux: %s:  unrecognized kind %d\n",
1697                         __func__, kind);
1698                 return 0;
1699         }
1700 
1701         rc = avc_has_perm(sid, isec->sid, isec->sclass, av, &ad);
1702         return rc;
1703 }
1704 
1705 static inline int may_rename(struct inode *old_dir,
1706                              struct dentry *old_dentry,
1707                              struct inode *new_dir,
1708                              struct dentry *new_dentry)
1709 {
1710         struct inode_security_struct *old_dsec, *new_dsec, *old_isec, *new_isec;
1711         struct avc_audit_data ad;
1712         u32 sid = current_sid();
1713         u32 av;
1714         int old_is_dir, new_is_dir;
1715         int rc;
1716 
1717         old_dsec = old_dir->i_security;
1718         old_isec = old_dentry->d_inode->i_security;
1719         old_is_dir = S_ISDIR(old_dentry->d_inode->i_mode);
1720         new_dsec = new_dir->i_security;
1721 
1722         AVC_AUDIT_DATA_INIT(&ad, FS);
1723 
1724         ad.u.fs.path.dentry = old_dentry;
1725         rc = avc_has_perm(sid, old_dsec->sid, SECCLASS_DIR,
1726                           DIR__REMOVE_NAME | DIR__SEARCH, &ad);
1727         if (rc)
1728                 return rc;
1729         rc = avc_has_perm(sid, old_isec->sid,
1730                           old_isec->sclass, FILE__RENAME, &ad);
1731         if (rc)
1732                 return rc;
1733         if (old_is_dir && new_dir != old_dir) {
1734                 rc = avc_has_perm(sid, old_isec->sid,
1735                                   old_isec->sclass, DIR__REPARENT, &ad);
1736                 if (rc)
1737                         return rc;
1738         }
1739 
1740         ad.u.fs.path.dentry = new_dentry;
1741         av = DIR__ADD_NAME | DIR__SEARCH;
1742         if (new_dentry->d_inode)
1743                 av |= DIR__REMOVE_NAME;
1744         rc = avc_has_perm(sid, new_dsec->sid, SECCLASS_DIR, av, &ad);
1745         if (rc)
1746                 return rc;
1747         if (new_dentry->d_inode) {
1748                 new_isec = new_dentry->d_inode->i_security;
1749                 new_is_dir = S_ISDIR(new_dentry->d_inode->i_mode);
1750                 rc = avc_has_perm(sid, new_isec->sid,
1751                                   new_isec->sclass,
1752                                   (new_is_dir ? DIR__RMDIR : FILE__UNLINK), &ad);
1753                 if (rc)
1754                         return rc;
1755         }
1756 
1757         return 0;
1758 }
1759 
1760 /* Check whether a task can perform a filesystem operation. */
1761 static int superblock_has_perm(const struct cred *cred,
1762                                struct super_block *sb,
1763                                u32 perms,
1764                                struct avc_audit_data *ad)
1765 {
1766         struct superblock_security_struct *sbsec;
1767         u32 sid = cred_sid(cred);
1768 
1769         sbsec = sb->s_security;
1770         return avc_has_perm(sid, sbsec->sid, SECCLASS_FILESYSTEM, perms, ad);
1771 }
1772 
1773 /* Convert a Linux mode and permission mask to an access vector. */
1774 static inline u32 file_mask_to_av(int mode, int mask)
1775 {
1776         u32 av = 0;
1777 
1778         if ((mode & S_IFMT) != S_IFDIR) {
1779                 if (mask & MAY_EXEC)
1780                         av |= FILE__EXECUTE;
1781                 if (mask & MAY_READ)
1782                         av |= FILE__READ;
1783 
1784                 if (mask & MAY_APPEND)
1785                         av |= FILE__APPEND;
1786                 else if (mask & MAY_WRITE)
1787                         av |= FILE__WRITE;
1788 
1789         } else {
1790                 if (mask & MAY_EXEC)
1791                         av |= DIR__SEARCH;
1792                 if (mask & MAY_WRITE)
1793                         av |= DIR__WRITE;
1794                 if (mask & MAY_READ)
1795                         av |= DIR__READ;
1796         }
1797 
1798         return av;
1799 }
1800 
1801 /* Convert a Linux file to an access vector. */
1802 static inline u32 file_to_av(struct file *file)
1803 {
1804         u32 av = 0;
1805 
1806         if (file->f_mode & FMODE_READ)
1807                 av |= FILE__READ;
1808         if (file->f_mode & FMODE_WRITE) {
1809                 if (file->f_flags & O_APPEND)
1810                         av |= FILE__APPEND;
1811                 else
1812                         av |= FILE__WRITE;
1813         }
1814         if (!av) {
1815                 /*
1816                  * Special file opened with flags 3 for ioctl-only use.
1817                  */
1818                 av = FILE__IOCTL;
1819         }
1820 
1821         return av;
1822 }
1823 
1824 /*
1825  * Convert a file to an access vector and include the correct open
1826  * open permission.
1827  */
1828 static inline u32 open_file_to_av(struct file *file)
1829 {
1830         u32 av = file_to_av(file);
1831 
1832         if (selinux_policycap_openperm) {
1833                 mode_t mode = file->f_path.dentry->d_inode->i_mode;
1834                 /*
1835                  * lnk files and socks do not really have an 'open'
1836                  */
1837                 if (S_ISREG(mode))
1838                         av |= FILE__OPEN;
1839                 else if (S_ISCHR(mode))
1840                         av |= CHR_FILE__OPEN;
1841                 else if (S_ISBLK(mode))
1842                         av |= BLK_FILE__OPEN;
1843                 else if (S_ISFIFO(mode))
1844                         av |= FIFO_FILE__OPEN;
1845                 else if (S_ISDIR(mode))
1846                         av |= DIR__OPEN;
1847                 else if (S_ISSOCK(mode))
1848                         av |= SOCK_FILE__OPEN;
1849                 else
1850                         printk(KERN_ERR "SELinux: WARNING: inside %s with "
1851                                 "unknown mode:%o\n", __func__, mode);
1852         }
1853         return av;
1854 }
1855 
1856 /* Hook functions begin here. */
1857 
1858 static int selinux_ptrace_may_access(struct task_struct *child,
1859                                      unsigned int mode)
1860 {
1861         int rc;
1862 
1863         rc = cap_ptrace_may_access(child, mode);
1864         if (rc)
1865                 return rc;
1866 
1867         if (mode == PTRACE_MODE_READ) {
1868                 u32 sid = current_sid();
1869                 u32 csid = task_sid(child);
1870                 return avc_has_perm(sid, csid, SECCLASS_FILE, FILE__READ, NULL);
1871         }
1872 
1873         return current_has_perm(child, PROCESS__PTRACE);
1874 }
1875 
1876 static int selinux_ptrace_traceme(struct task_struct *parent)
1877 {
1878         int rc;
1879 
1880         rc = cap_ptrace_traceme(parent);
1881         if (rc)
1882                 return rc;
1883 
1884         return task_has_perm(parent, current, PROCESS__PTRACE);
1885 }
1886 
1887 static int selinux_capget(struct task_struct *target, kernel_cap_t *effective,
1888                           kernel_cap_t *inheritable, kernel_cap_t *permitted)
1889 {
1890         int error;
1891 
1892         error = current_has_perm(target, PROCESS__GETCAP);
1893         if (error)
1894                 return error;
1895 
1896         return cap_capget(target, effective, inheritable, permitted);
1897 }
1898 
1899 static int selinux_capset(struct cred *new, const struct cred *old,
1900                           const kernel_cap_t *effective,
1901                           const kernel_cap_t *inheritable,
1902                           const kernel_cap_t *permitted)
1903 {
1904         int error;
1905 
1906         error = cap_capset(new, old,
1907                                       effective, inheritable, permitted);
1908         if (error)
1909                 return error;
1910 
1911         return cred_has_perm(old, new, PROCESS__SETCAP);
1912 }
1913 
1914 /*
1915  * (This comment used to live with the selinux_task_setuid hook,
1916  * which was removed).
1917  *
1918  * Since setuid only affects the current process, and since the SELinux
1919  * controls are not based on the Linux identity attributes, SELinux does not
1920  * need to control this operation.  However, SELinux does control the use of
1921  * the CAP_SETUID and CAP_SETGID capabilities using the capable hook.
1922  */
1923 
1924 static int selinux_capable(struct task_struct *tsk, const struct cred *cred,
1925                            int cap, int audit)
1926 {
1927         int rc;
1928 
1929         rc = cap_capable(tsk, cred, cap, audit);
1930         if (rc)
1931                 return rc;
1932 
1933         return task_has_capability(tsk, cred, cap, audit);
1934 }
1935 
1936 static int selinux_sysctl_get_sid(ctl_table *table, u16 tclass, u32 *sid)
1937 {
1938         int buflen, rc;
1939         char *buffer, *path, *end;
1940 
1941         rc = -ENOMEM;
1942         buffer = (char *)__get_free_page(GFP_KERNEL);
1943         if (!buffer)
1944                 goto out;
1945 
1946         buflen = PAGE_SIZE;
1947         end = buffer+buflen;
1948         *--end = '\0';
1949         buflen--;
1950         path = end-1;
1951         *path = '/';
1952         while (table) {
1953                 const char *name = table->procname;
1954                 size_t namelen = strlen(name);
1955                 buflen -= namelen + 1;
1956                 if (buflen < 0)
1957                         goto out_free;
1958                 end -= namelen;
1959                 memcpy(end, name, namelen);
1960                 *--end = '/';
1961                 path = end;
1962                 table = table->parent;
1963         }
1964         buflen -= 4;
1965         if (buflen < 0)
1966                 goto out_free;
1967         end -= 4;
1968         memcpy(end, "/sys", 4);
1969         path = end;
1970         rc = security_genfs_sid("proc", path, tclass, sid);
1971 out_free:
1972         free_page((unsigned long)buffer);
1973 out:
1974         return rc;
1975 }
1976 
1977 static int selinux_sysctl(ctl_table *table, int op)
1978 {
1979         int error = 0;
1980         u32 av;
1981         u32 tsid, sid;
1982         int rc;
1983 
1984         sid = current_sid();
1985 
1986         rc = selinux_sysctl_get_sid(table, (op == 0001) ?
1987                                     SECCLASS_DIR : SECCLASS_FILE, &tsid);
1988         if (rc) {
1989                 /* Default to the well-defined sysctl SID. */
1990                 tsid = SECINITSID_SYSCTL;
1991         }
1992 
1993         /* The op values are "defined" in sysctl.c, thereby creating
1994          * a bad coupling between this module and sysctl.c */
1995         if (op == 001) {
1996                 error = avc_has_perm(sid, tsid,
1997                                      SECCLASS_DIR, DIR__SEARCH, NULL);
1998         } else {
1999                 av = 0;
2000                 if (op & 004)
2001                         av |= FILE__READ;
2002                 if (op & 002)
2003                         av |= FILE__WRITE;
2004                 if (av)
2005                         error = avc_has_perm(sid, tsid,
2006                                              SECCLASS_FILE, av, NULL);
2007         }
2008 
2009         return error;
2010 }
2011 
2012 static int selinux_quotactl(int cmds, int type, int id, struct super_block *sb)
2013 {
2014         const struct cred *cred = current_cred();
2015         int rc = 0;
2016 
2017         if (!sb)
2018                 return 0;
2019 
2020         switch (cmds) {
2021         case Q_SYNC:
2022         case Q_QUOTAON:
2023         case Q_QUOTAOFF:
2024         case Q_SETINFO:
2025         case Q_SETQUOTA:
2026                 rc = superblock_has_perm(cred, sb, FILESYSTEM__QUOTAMOD, NULL);
2027                 break;
2028         case Q_GETFMT:
2029         case Q_GETINFO:
2030         case Q_GETQUOTA:
2031                 rc = superblock_has_perm(cred, sb, FILESYSTEM__QUOTAGET, NULL);
2032                 break;
2033         default:
2034                 rc = 0;  /* let the kernel handle invalid cmds */
2035                 break;
2036         }
2037         return rc;
2038 }
2039 
2040 static int selinux_quota_on(struct dentry *dentry)
2041 {
2042         const struct cred *cred = current_cred();
2043 
2044         return dentry_has_perm(cred, NULL, dentry, FILE__QUOTAON);
2045 }
2046 
2047 static int selinux_syslog(int type)
2048 {
2049         int rc;
2050 
2051         rc = cap_syslog(type);
2052         if (rc)
2053                 return rc;
2054 
2055         switch (type) {
2056         case 3:         /* Read last kernel messages */
2057         case 10:        /* Return size of the log buffer */
2058                 rc = task_has_system(current, SYSTEM__SYSLOG_READ);
2059                 break;
2060         case 6:         /* Disable logging to console */
2061         case 7:         /* Enable logging to console */
2062         case 8:         /* Set level of messages printed to console */
2063                 rc = task_has_system(current, SYSTEM__SYSLOG_CONSOLE);
2064                 break;
2065         case 0:         /* Close log */
2066         case 1:         /* Open log */
2067         case 2:         /* Read from log */
2068         case 4:         /* Read/clear last kernel messages */
2069         case 5:         /* Clear ring buffer */
2070         default:
2071                 rc = task_has_system(current, SYSTEM__SYSLOG_MOD);
2072                 break;
2073         }
2074         return rc;
2075 }
2076 
2077 /*
2078  * Check that a process has enough memory to allocate a new virtual
2079  * mapping. 0 means there is enough memory for the allocation to
2080  * succeed and -ENOMEM implies there is not.
2081  *
2082  * Do not audit the selinux permission check, as this is applied to all
2083  * processes that allocate mappings.
2084  */
2085 static int selinux_vm_enough_memory(struct mm_struct *mm, long pages)
2086 {
2087         int rc, cap_sys_admin = 0;
2088 
2089         rc = selinux_capable(current, current_cred(), CAP_SYS_ADMIN,
2090                              SECURITY_CAP_NOAUDIT);
2091         if (rc == 0)
2092                 cap_sys_admin = 1;
2093 
2094         return __vm_enough_memory(mm, pages, cap_sys_admin);
2095 }
2096 
2097 /* binprm security operations */
2098 
2099 static int selinux_bprm_set_creds(struct linux_binprm *bprm)
2100 {
2101         const struct task_security_struct *old_tsec;
2102         struct task_security_struct *new_tsec;
2103         struct inode_security_struct *isec;
2104         struct avc_audit_data ad;
2105         struct inode *inode = bprm->file->f_path.dentry->d_inode;
2106         int rc;
2107 
2108         rc = cap_bprm_set_creds(bprm);
2109         if (rc)
2110                 return rc;
2111 
2112         /* SELinux context only depends on initial program or script and not
2113          * the script interpreter */
2114         if (bprm->cred_prepared)
2115                 return 0;
2116 
2117         old_tsec = current_security();
2118         new_tsec = bprm->cred->security;
2119         isec = inode->i_security;
2120 
2121         /* Default to the current task SID. */
2122         new_tsec->sid = old_tsec->sid;
2123         new_tsec->osid = old_tsec->sid;
2124 
2125         /* Reset fs, key, and sock SIDs on execve. */
2126         new_tsec->create_sid = 0;
2127         new_tsec->keycreate_sid = 0;
2128         new_tsec->sockcreate_sid = 0;
2129 
2130         if (old_tsec->exec_sid) {
2131                 new_tsec->sid = old_tsec->exec_sid;
2132                 /* Reset exec SID on execve. */
2133                 new_tsec->exec_sid = 0;
2134         } else {
2135                 /* Check for a default transition on this program. */
2136                 rc = security_transition_sid(old_tsec->sid, isec->sid,
2137                                              SECCLASS_PROCESS, &new_tsec->sid);
2138                 if (rc)
2139                         return rc;
2140         }
2141 
2142         AVC_AUDIT_DATA_INIT(&ad, FS);
2143         ad.u.fs.path = bprm->file->f_path;
2144 
2145         if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
2146                 new_tsec->sid = old_tsec->sid;
2147 
2148         if (new_tsec->sid == old_tsec->sid) {
2149                 rc = avc_has_perm(old_tsec->sid, isec->sid,
2150                                   SECCLASS_FILE, FILE__EXECUTE_NO_TRANS, &ad);
2151                 if (rc)
2152                         return rc;
2153         } else {
2154                 /* Check permissions for the transition. */
2155                 rc = avc_has_perm(old_tsec->sid, new_tsec->sid,
2156                                   SECCLASS_PROCESS, PROCESS__TRANSITION, &ad);
2157                 if (rc)
2158                         return rc;
2159 
2160                 rc = avc_has_perm(new_tsec->sid, isec->sid,
2161                                   SECCLASS_FILE, FILE__ENTRYPOINT, &ad);
2162                 if (rc)
2163                         return rc;
2164 
2165                 /* Check for shared state */
2166                 if (bprm->unsafe & LSM_UNSAFE_SHARE) {
2167                         rc = avc_has_perm(old_tsec->sid, new_tsec->sid,
2168                                           SECCLASS_PROCESS, PROCESS__SHARE,
2169                                           NULL);
2170                         if (rc)
2171                                 return -EPERM;
2172                 }
2173 
2174                 /* Make sure that anyone attempting to ptrace over a task that
2175                  * changes its SID has the appropriate permit */
2176                 if (bprm->unsafe &
2177                     (LSM_UNSAFE_PTRACE | LSM_UNSAFE_PTRACE_CAP)) {
2178                         struct task_struct *tracer;
2179                         struct task_security_struct *sec;
2180                         u32 ptsid = 0;
2181 
2182                         rcu_read_lock();
2183                         tracer = tracehook_tracer_task(current);
2184                         if (likely(tracer != NULL)) {
2185                                 sec = __task_cred(tracer)->security;
2186                                 ptsid = sec->sid;
2187                         }
2188                         rcu_read_unlock();
2189 
2190                         if (ptsid != 0) {
2191                                 rc = avc_has_perm(ptsid, new_tsec->sid,
2192                                                   SECCLASS_PROCESS,
2193                                                   PROCESS__PTRACE, NULL);
2194                                 if (rc)
2195                                         return -EPERM;
2196                         }
2197                 }
2198 
2199                 /* Clear any possibly unsafe personality bits on exec: */
2200                 bprm->per_clear |= PER_CLEAR_ON_SETID;
2201         }
2202 
2203         return 0;
2204 }
2205 
2206 static int selinux_bprm_secureexec(struct linux_binprm *bprm)
2207 {
2208         const struct cred *cred = current_cred();
2209         const struct task_security_struct *tsec = cred->security;
2210         u32 sid, osid;
2211         int atsecure = 0;
2212 
2213         sid = tsec->sid;
2214         osid = tsec->osid;
2215 
2216         if (osid != sid) {
2217                 /* Enable secure mode for SIDs transitions unless
2218                    the noatsecure permission is granted between
2219                    the two SIDs, i.e. ahp returns 0. */
2220                 atsecure = avc_has_perm(osid, sid,
2221                                         SECCLASS_PROCESS,
2222                                         PROCESS__NOATSECURE, NULL);
2223         }
2224 
2225         return (atsecure || cap_bprm_secureexec(bprm));
2226 }
2227 
2228 extern struct vfsmount *selinuxfs_mount;
2229 extern struct dentry *selinux_null;
2230 
2231 /* Derived from fs/exec.c:flush_old_files. */
2232 static inline void flush_unauthorized_files(const struct cred *cred,
2233                                             struct files_struct *files)
2234 {
2235         struct avc_audit_data ad;
2236         struct file *file, *devnull = NULL;
2237         struct tty_struct *tty;
2238         struct fdtable *fdt;
2239         long j = -1;
2240         int drop_tty = 0;
2241 
2242         tty = get_current_tty();
2243         if (tty) {
2244                 file_list_lock();
2245                 if (!list_empty(&tty->tty_files)) {
2246                         struct inode *inode;
2247 
2248                         /* Revalidate access to controlling tty.
2249                            Use inode_has_perm on the tty inode directly rather
2250                            than using file_has_perm, as this particular open
2251                            file may belong to another process and we are only
2252                            interested in the inode-based check here. */
2253                         file = list_first_entry(&tty->tty_files, struct file, f_u.fu_list);
2254                         inode = file->f_path.dentry->d_inode;
2255                         if (inode_has_perm(cred, inode,
2256                                            FILE__READ | FILE__WRITE, NULL)) {
2257                                 drop_tty = 1;
2258                         }
2259                 }
2260                 file_list_unlock();
2261                 tty_kref_put(tty);
2262         }
2263         /* Reset controlling tty. */
2264         if (drop_tty)
2265                 no_tty();
2266 
2267         /* Revalidate access to inherited open files. */
2268 
2269         AVC_AUDIT_DATA_INIT(&ad, FS);
2270 
2271         spin_lock(&files->file_lock);
2272         for (;;) {
2273                 unsigned long set, i;
2274                 int fd;
2275 
2276                 j++;
2277                 i = j * __NFDBITS;
2278                 fdt = files_fdtable(files);
2279                 if (i >= fdt->max_fds)
2280                         break;
2281                 set = fdt->open_fds->fds_bits[j];
2282                 if (!set)
2283                         continue;
2284                 spin_unlock(&files->file_lock);
2285                 for ( ; set ; i++, set >>= 1) {
2286                         if (set & 1) {
2287                                 file = fget(i);
2288                                 if (!file)
2289                                         continue;
2290                                 if (file_has_perm(cred,
2291                                                   file,
2292                                                   file_to_av(file))) {
2293                                         sys_close(i);
2294                                         fd = get_unused_fd();
2295                                         if (fd != i) {
2296                                                 if (fd >= 0)
2297                                                         put_unused_fd(fd);
2298                                                 fput(file);
2299                                                 continue;
2300                                         }
2301                                         if (devnull) {
2302                                                 get_file(devnull);
2303                                         } else {
2304                                                 devnull = dentry_open(
2305                                                         dget(selinux_null),
2306                                                         mntget(selinuxfs_mount),
2307                                                         O_RDWR, cred);
2308                                                 if (IS_ERR(devnull)) {
2309                                                         devnull = NULL;
2310                                                         put_unused_fd(fd);
2311                                                         fput(file);
2312                                                         continue;
2313                                                 }
2314                                         }
2315                                         fd_install(fd, devnull);
2316                                 }
2317                                 fput(file);
2318                         }
2319                 }
2320                 spin_lock(&files->file_lock);
2321 
2322         }
2323         spin_unlock(&files->file_lock);
2324 }
2325 
2326 /*
2327  * Prepare a process for imminent new credential changes due to exec
2328  */
2329 static void selinux_bprm_committing_creds(struct linux_binprm *bprm)
2330 {
2331         struct task_security_struct *new_tsec;
2332         struct rlimit *rlim, *initrlim;
2333         int rc, i;
2334 
2335         new_tsec = bprm->cred->security;
2336         if (new_tsec->sid == new_tsec->osid)
2337                 return;
2338 
2339         /* Close files for which the new task SID is not authorized. */
2340         flush_unauthorized_files(bprm->cred, current->files);
2341 
2342         /* Always clear parent death signal on SID transitions. */
2343         current->pdeath_signal = 0;
2344 
2345         /* Check whether the new SID can inherit resource limits from the old
2346          * SID.  If not, reset all soft limits to the lower of the current
2347          * task's hard limit and the init task's soft limit.
2348          *
2349          * Note that the setting of hard limits (even to lower them) can be
2350          * controlled by the setrlimit check.  The inclusion of the init task's
2351          * soft limit into the computation is to avoid resetting soft limits
2352          * higher than the default soft limit for cases where the default is
2353          * lower than the hard limit, e.g. RLIMIT_CORE or RLIMIT_STACK.
2354          */
2355         rc = avc_has_perm(new_tsec->osid, new_tsec->sid, SECCLASS_PROCESS,
2356                           PROCESS__RLIMITINH, NULL);
2357         if (rc) {
2358                 for (i = 0; i < RLIM_NLIMITS; i++) {
2359                         rlim = current->signal->rlim + i;
2360                         initrlim = init_task.signal->rlim + i;
2361                         rlim->rlim_cur = min(rlim->rlim_max, initrlim->rlim_cur);
2362                 }
2363                 update_rlimit_cpu(current->signal->rlim[RLIMIT_CPU].rlim_cur);
2364         }
2365 }
2366 
2367 /*
2368  * Clean up the process immediately after the installation of new credentials
2369  * due to exec
2370  */
2371 static void selinux_bprm_committed_creds(struct linux_binprm *bprm)
2372 {
2373         const struct task_security_struct *tsec = current_security();
2374         struct itimerval itimer;
2375         u32 osid, sid;
2376         int rc, i;
2377 
2378         osid = tsec->osid;
2379         sid = tsec->sid;
2380 
2381         if (sid == osid)
2382                 return;
2383 
2384         /* Check whether the new SID can inherit signal state from the old SID.
2385          * If not, clear itimers to avoid subsequent signal generation and
2386          * flush and unblock signals.
2387          *
2388          * This must occur _after_ the task SID has been updated so that any
2389          * kill done after the flush will be checked against the new SID.
2390          */
2391         rc = avc_has_perm(osid, sid, SECCLASS_PROCESS, PROCESS__SIGINH, NULL);
2392         if (rc) {
2393                 memset(&itimer, 0, sizeof itimer);
2394                 for (i = 0; i < 3; i++)
2395                         do_setitimer(i, &itimer, NULL);
2396                 spin_lock_irq(&current->sighand->siglock);
2397                 if (!(current->signal->flags & SIGNAL_GROUP_EXIT)) {
2398                         __flush_signals(current);
2399                         flush_signal_handlers(current, 1);
2400                         sigemptyset(&current->blocked);
2401                 }
2402                 spin_unlock_irq(&current->sighand->siglock);
2403         }
2404 
2405         /* Wake up the parent if it is waiting so that it can recheck
2406          * wait permission to the new task SID. */
2407         read_lock(&tasklist_lock);
2408         wake_up_interruptible(&current->real_parent->signal->wait_chldexit);
2409         read_unlock(&tasklist_lock);
2410 }
2411 
2412 /* superblock security operations */
2413 
2414 static int selinux_sb_alloc_security(struct super_block *sb)
2415 {
2416         return superblock_alloc_security(sb);
2417 }
2418 
2419 static void selinux_sb_free_security(struct super_block *sb)
2420 {
2421         superblock_free_security(sb);
2422 }
2423 
2424 static inline int match_prefix(char *prefix, int plen, char *option, int olen)
2425 {
2426         if (plen > olen)
2427                 return 0;
2428 
2429         return !memcmp(prefix, option, plen);
2430 }
2431 
2432 static inline int selinux_option(char *option, int len)
2433 {
2434         return (match_prefix(CONTEXT_STR, sizeof(CONTEXT_STR)-1, option, len) ||
2435                 match_prefix(FSCONTEXT_STR, sizeof(FSCONTEXT_STR)-1, option, len) ||
2436                 match_prefix(DEFCONTEXT_STR, sizeof(DEFCONTEXT_STR)-1, option, len) ||
2437                 match_prefix(ROOTCONTEXT_STR, sizeof(ROOTCONTEXT_STR)-1, option, len) ||
2438                 match_prefix(LABELSUPP_STR, sizeof(LABELSUPP_STR)-1, option, len));
2439 }
2440 
2441 static inline void take_option(char **to, char *from, int *first, int len)
2442 {
2443         if (!*first) {
2444                 **to = ',';
2445                 *to += 1;
2446         } else
2447                 *first = 0;
2448         memcpy(*to, from, len);
2449         *to += len;
2450 }
2451 
2452 static inline void take_selinux_option(char **to, char *from, int *first,
2453                                        int len)
2454 {
2455         int current_size = 0;
2456 
2457         if (!*first) {
2458                 **to = '|';
2459                 *to += 1;
2460         } else
2461                 *first = 0;
2462 
2463         while (current_size < len) {
2464                 if (*from != '"') {
2465                         **to = *from;
2466                         *to += 1;
2467                 }
2468                 from += 1;
2469                 current_size += 1;
2470         }
2471 }
2472 
2473 static int selinux_sb_copy_data(char *orig, char *copy)
2474 {
2475         int fnosec, fsec, rc = 0;
2476         char *in_save, *in_curr, *in_end;
2477         char *sec_curr, *nosec_save, *nosec;
2478         int open_quote = 0;
2479 
2480         in_curr = orig;
2481         sec_curr = copy;
2482 
2483         nosec = (char *)get_zeroed_page(GFP_KERNEL);
2484         if (!nosec) {
2485                 rc = -ENOMEM;
2486                 goto out;
2487         }
2488 
2489         nosec_save = nosec;
2490         fnosec = fsec = 1;
2491         in_save = in_end = orig;
2492 
2493         do {
2494                 if (*in_end == '"')
2495                         open_quote = !open_quote;
2496                 if ((*in_end == ',' && open_quote == 0) ||
2497                                 *in_end == '\0') {
2498                         int len = in_end - in_curr;
2499 
2500                         if (selinux_option(in_curr, len))
2501                                 take_selinux_option(&sec_curr, in_curr, &fsec, len);
2502                         else
2503                                 take_option(&nosec, in_curr, &fnosec, len);
2504 
2505                         in_curr = in_end + 1;
2506                 }
2507         } while (*in_end++);
2508 
2509         strcpy(in_save, nosec_save);
2510         free_page((unsigned long)nosec_save);
2511 out:
2512         return rc;
2513 }
2514 
2515 static int selinux_sb_kern_mount(struct super_block *sb, int flags, void *data)
2516 {
2517         const struct cred *cred = current_cred();
2518         struct avc_audit_data ad;
2519         int rc;
2520 
2521         rc = superblock_doinit(sb, data);
2522         if (rc)
2523                 return rc;
2524 
2525         /* Allow all mounts performed by the kernel */
2526         if (flags & MS_KERNMOUNT)
2527                 return 0;
2528 
2529         AVC_AUDIT_DATA_INIT(&ad, FS);
2530         ad.u.fs.path.dentry = sb->s_root;
2531         return superblock_has_perm(cred, sb, FILESYSTEM__MOUNT, &ad);
2532 }
2533 
2534 static int selinux_sb_statfs(struct dentry *dentry)
2535 {
2536         const struct cred *cred = current_cred();
2537         struct avc_audit_data ad;
2538 
2539         AVC_AUDIT_DATA_INIT(&ad, FS);
2540         ad.u.fs.path.dentry = dentry->d_sb->s_root;
2541         return superblock_has_perm(cred, dentry->d_sb, FILESYSTEM__GETATTR, &ad);
2542 }
2543 
2544 static int selinux_mount(char *dev_name,
2545                          struct path *path,
2546                          char *type,
2547                          unsigned long flags,
2548                          void *data)
2549 {
2550         const struct cred *cred = current_cred();
2551 
2552         if (flags & MS_REMOUNT)
2553                 return superblock_has_perm(cred, path->mnt->mnt_sb,
2554                                            FILESYSTEM__REMOUNT, NULL);
2555         else
2556                 return dentry_has_perm(cred, path->mnt, path->dentry,
2557                                        FILE__MOUNTON);
2558 }
2559 
2560 static int selinux_umount(struct vfsmount *mnt, int flags)
2561 {
2562         const struct cred *cred = current_cred();
2563 
2564         return superblock_has_perm(cred, mnt->mnt_sb,
2565                                    FILESYSTEM__UNMOUNT, NULL);
2566 }
2567 
2568 /* inode security operations */
2569 
2570 static int selinux_inode_alloc_security(struct inode *inode)
2571 {
2572         return inode_alloc_security(inode);
2573 }
2574 
2575 static void selinux_inode_free_security(struct inode *inode)
2576 {
2577         inode_free_security(inode);
2578 }
2579 
2580 static int selinux_inode_init_security(struct inode *inode, struct inode *dir,
2581                                        char **name, void **value,
2582                                        size_t *len)
2583 {
2584         const struct cred *cred = current_cred();
2585         const struct task_security_struct *tsec = cred->security;
2586         struct inode_security_struct *dsec;
2587         struct superblock_security_struct *sbsec;
2588         u32 sid, newsid, clen;
2589         int rc;
2590         char *namep = NULL, *context;
2591 
2592         dsec = dir->i_security;
2593         sbsec = dir->i_sb->s_security;
2594 
2595         sid = tsec->sid;
2596         newsid = tsec->create_sid;
2597 
2598         if (!newsid || !(sbsec->flags & SE_SBLABELSUPP)) {
2599                 rc = security_transition_sid(sid, dsec->sid,
2600                                              inode_mode_to_security_class(inode->i_mode),
2601                                              &newsid);
2602                 if (rc) {
2603                         printk(KERN_WARNING "%s:  "
2604                                "security_transition_sid failed, rc=%d (dev=%s "
2605                                "ino=%ld)\n",
2606                                __func__,
2607                                -rc, inode->i_sb->s_id, inode->i_ino);
2608                         return rc;
2609                 }
2610         }
2611 
2612         /* Possibly defer initialization to selinux_complete_init. */
2613         if (sbsec->flags & SE_SBINITIALIZED) {
2614                 struct inode_security_struct *isec = inode->i_security;
2615                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
2616                 isec->sid = newsid;
2617                 isec->initialized = 1;
2618         }
2619 
2620         if (!ss_initialized || !(sbsec->flags & SE_SBLABELSUPP))
2621                 return -EOPNOTSUPP;
2622 
2623         if (name) {
2624                 namep = kstrdup(XATTR_SELINUX_SUFFIX, GFP_NOFS);
2625                 if (!namep)
2626                         return -ENOMEM;
2627                 *name = namep;
2628         }
2629 
2630         if (value && len) {
2631                 rc = security_sid_to_context_force(newsid, &context, &clen);
2632                 if (rc) {
2633                         kfree(namep);
2634                         return rc;
2635                 }
2636                 *value = context;
2637                 *len = clen;
2638         }
2639 
2640         return 0;
2641 }
2642 
2643 static int selinux_inode_create(struct inode *dir, struct dentry *dentry, int mask)
2644 {
2645         return may_create(dir, dentry, SECCLASS_FILE);
2646 }
2647 
2648 static int selinux_inode_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry)
2649 {
2650         return may_link(dir, old_dentry, MAY_LINK);
2651 }
2652 
2653 static int selinux_inode_unlink(struct inode *dir, struct dentry *dentry)
2654 {
2655         return may_link(dir, dentry, MAY_UNLINK);
2656 }
2657 
2658 static int selinux_inode_symlink(struct inode *dir, struct dentry *dentry, const char *name)
2659 {
2660         return may_create(dir, dentry, SECCLASS_LNK_FILE);
2661 }
2662 
2663 static int selinux_inode_mkdir(struct inode *dir, struct dentry *dentry, int mask)
2664 {
2665         return may_create(dir, dentry, SECCLASS_DIR);
2666 }
2667 
2668 static int selinux_inode_rmdir(struct inode *dir, struct dentry *dentry)
2669 {
2670         return may_link(dir, dentry, MAY_RMDIR);
2671 }
2672 
2673 static int selinux_inode_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
2674 {
2675         return may_create(dir, dentry, inode_mode_to_security_class(mode));
2676 }
2677 
2678 static int selinux_inode_rename(struct inode *old_inode, struct dentry *old_dentry,
2679                                 struct inode *new_inode, struct dentry *new_dentry)
2680 {
2681         return may_rename(old_inode, old_dentry, new_inode, new_dentry);
2682 }
2683 
2684 static int selinux_inode_readlink(struct dentry *dentry)
2685 {
2686         const struct cred *cred = current_cred();
2687 
2688         return dentry_has_perm(cred, NULL, dentry, FILE__READ);
2689 }
2690 
2691 static int selinux_inode_follow_link(struct dentry *dentry, struct nameidata *nameidata)
2692 {
2693         const struct cred *cred = current_cred();
2694 
2695         return dentry_has_perm(cred, NULL, dentry, FILE__READ);
2696 }
2697 
2698 static int selinux_inode_permission(struct inode *inode, int mask)
2699 {
2700         const struct cred *cred = current_cred();
2701 
2702         if (!mask) {
2703                 /* No permission to check.  Existence test. */
2704                 return 0;
2705         }
2706 
2707         return inode_has_perm(cred, inode,
2708                               file_mask_to_av(inode->i_mode, mask), NULL);
2709 }
2710 
2711 static int selinux_inode_setattr(struct dentry *dentry, struct iattr *iattr)
2712 {
2713         const struct cred *cred = current_cred();
2714 
2715         if (iattr->ia_valid & ATTR_FORCE)
2716                 return 0;
2717 
2718         if (iattr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID |
2719                                ATTR_ATIME_SET | ATTR_MTIME_SET))
2720                 return dentry_has_perm(cred, NULL, dentry, FILE__SETATTR);
2721 
2722         return dentry_has_perm(cred, NULL, dentry, FILE__WRITE);
2723 }
2724 
2725 static int selinux_inode_getattr(struct vfsmount *mnt, struct dentry *dentry)
2726 {
2727         const struct cred *cred = current_cred();
2728 
2729         return dentry_has_perm(cred, mnt, dentry, FILE__GETATTR);
2730 }
2731 
2732 static int selinux_inode_setotherxattr(struct dentry *dentry, const char *name)
2733 {
2734         const struct cred *cred = current_cred();
2735 
2736         if (!strncmp(name, XATTR_SECURITY_PREFIX,
2737                      sizeof XATTR_SECURITY_PREFIX - 1)) {
2738                 if (!strcmp(name, XATTR_NAME_CAPS)) {
2739                         if (!capable(CAP_SETFCAP))
2740                                 return -EPERM;
2741                 } else if (!capable(CAP_SYS_ADMIN)) {
2742                         /* A different attribute in the security namespace.
2743                            Restrict to administrator. */
2744                         return -EPERM;
2745                 }
2746         }
2747 
2748         /* Not an attribute we recognize, so just check the
2749            ordinary setattr permission. */
2750         return dentry_has_perm(cred, NULL, dentry, FILE__SETATTR);
2751 }
2752 
2753 static int selinux_inode_setxattr(struct dentry *dentry, const char *name,
2754                                   const void *value, size_t size, int flags)
2755 {
2756         struct inode *inode = dentry->d_inode;
2757         struct inode_security_struct *isec = inode->i_security;
2758         struct superblock_security_struct *sbsec;
2759         struct avc_audit_data ad;
2760         u32 newsid, sid = current_sid();
2761         int rc = 0;
2762 
2763         if (strcmp(name, XATTR_NAME_SELINUX))
2764                 return selinux_inode_setotherxattr(dentry, name);
2765 
2766         sbsec = inode->i_sb->s_security;
2767         if (!(sbsec->flags & SE_SBLABELSUPP))
2768                 return -EOPNOTSUPP;
2769 
2770         if (!is_owner_or_cap(inode))
2771                 return -EPERM;
2772 
2773         AVC_AUDIT_DATA_INIT(&ad, FS);
2774         ad.u.fs.path.dentry = dentry;
2775 
2776         rc = avc_has_perm(sid, isec->sid, isec->sclass,
2777                           FILE__RELABELFROM, &ad);
2778         if (rc)
2779                 return rc;
2780 
2781         rc = security_context_to_sid(value, size, &newsid);
2782         if (rc == -EINVAL) {
2783                 if (!capable(CAP_MAC_ADMIN))
2784                         return rc;
2785                 rc = security_context_to_sid_force(value, size, &newsid);
2786         }
2787         if (rc)
2788                 return rc;
2789 
2790         rc = avc_has_perm(sid, newsid, isec->sclass,
2791                           FILE__RELABELTO, &ad);
2792         if (rc)
2793                 return rc;
2794 
2795         rc = security_validate_transition(isec->sid, newsid, sid,
2796                                           isec->sclass);
2797         if (rc)
2798                 return rc;
2799 
2800         return avc_has_perm(newsid,
2801                             sbsec->sid,
2802                             SECCLASS_FILESYSTEM,
2803                             FILESYSTEM__ASSOCIATE,
2804                             &ad);
2805 }
2806 
2807 static void selinux_inode_post_setxattr(struct dentry *dentry, const char *name,
2808                                         const void *value, size_t size,
2809                                         int flags)
2810 {
2811         struct inode *inode = dentry->d_inode;
2812         struct inode_security_struct *isec = inode->i_security;
2813         u32 newsid;
2814         int rc;
2815 
2816         if (strcmp(name, XATTR_NAME_SELINUX)) {
2817                 /* Not an attribute we recognize, so nothing to do. */
2818                 return;
2819         }
2820 
2821         rc = security_context_to_sid_force(value, size, &newsid);
2822         if (rc) {
2823                 printk(KERN_ERR "SELinux:  unable to map context to SID"
2824                        "for (%s, %lu), rc=%d\n",
2825                        inode->i_sb->s_id, inode->i_ino, -rc);
2826                 return;
2827         }
2828 
2829         isec->sid = newsid;
2830         return;
2831 }
2832 
2833 static int selinux_inode_getxattr(struct dentry *dentry, const char *name)
2834 {
2835         const struct cred *cred = current_cred();
2836 
2837         return dentry_has_perm(cred, NULL, dentry, FILE__GETATTR);
2838 }
2839 
2840 static int selinux_inode_listxattr(struct dentry *dentry)
2841 {
2842         const struct cred *cred = current_cred();
2843 
2844         return dentry_has_perm(cred, NULL, dentry, FILE__GETATTR);
2845 }
2846 
2847 static int selinux_inode_removexattr(struct dentry *dentry, const char *name)
2848 {
2849         if (strcmp(name, XATTR_NAME_SELINUX))
2850                 return selinux_inode_setotherxattr(dentry, name);
2851 
2852         /* No one is allowed to remove a SELinux security label.
2853            You can change the label, but all data must be labeled. */
2854         return -EACCES;
2855 }
2856 
2857 /*
2858  * Copy the inode security context value to the user.
2859  *
2860  * Permission check is handled by selinux_inode_getxattr hook.
2861  */
2862 static int selinux_inode_getsecurity(const struct inode *inode, const char *name, void **buffer, bool alloc)
2863 {
2864         u32 size;
2865         int error;
2866         char *context = NULL;
2867         struct inode_security_struct *isec = inode->i_security;
2868 
2869         if (strcmp(name, XATTR_SELINUX_SUFFIX))
2870                 return -EOPNOTSUPP;
2871 
2872         /*
2873          * If the caller has CAP_MAC_ADMIN, then get the raw context
2874          * value even if it is not defined by current policy; otherwise,
2875          * use the in-core value under current policy.
2876          * Use the non-auditing forms of the permission checks since
2877          * getxattr may be called by unprivileged processes commonly
2878          * and lack of permission just means that we fall back to the
2879          * in-core context value, not a denial.
2880          */
2881         error = selinux_capable(current, current_cred(), CAP_MAC_ADMIN,
2882                                 SECURITY_CAP_NOAUDIT);
2883         if (!error)
2884                 error = security_sid_to_context_force(isec->sid, &context,
2885                                                       &size);
2886         else
2887                 error = security_sid_to_context(isec->sid, &context, &size);
2888         if (error)
2889                 return error;
2890         error = size;
2891         if (alloc) {
2892                 *buffer = context;
2893                 goto out_nofree;
2894         }
2895         kfree(context);
2896 out_nofree:
2897         return error;
2898 }
2899 
2900 static int selinux_inode_setsecurity(struct inode *inode, const char *name,
2901                                      const void *value, size_t size, int flags)
2902 {
2903         struct inode_security_struct *isec = inode->i_security;
2904         u32 newsid;
2905         int rc;
2906 
2907         if (strcmp(name, XATTR_SELINUX_SUFFIX))
2908                 return -EOPNOTSUPP;
2909 
2910         if (!value || !size)
2911                 return -EACCES;
2912 
2913         rc = security_context_to_sid((void *)value, size, &newsid);
2914         if (rc)
2915                 return rc;
2916 
2917         isec->sid = newsid;
2918         return 0;
2919 }
2920 
2921 static int selinux_inode_listsecurity(struct inode *inode, char *buffer, size_t buffer_size)
2922 {
2923         const int len = sizeof(XATTR_NAME_SELINUX);
2924         if (buffer && len <= buffer_size)
2925                 memcpy(buffer, XATTR_NAME_SELINUX, len);
2926         return len;
2927 }
2928 
2929 static void selinux_inode_getsecid(const struct inode *inode, u32 *secid)
2930 {
2931         struct inode_security_struct *isec = inode->i_security;
2932         *secid = isec->sid;
2933 }
2934 
2935 /* file security operations */
2936 
2937 static int selinux_revalidate_file_permission(struct file *file, int mask)
2938 {
2939         const struct cred *cred = current_cred();
2940         struct inode *inode = file->f_path.dentry->d_inode;
2941 
2942         if (!mask) {
2943                 /* No permission to check.  Existence test. */
2944                 return 0;
2945         }
2946 
2947         /* file_mask_to_av won't add FILE__WRITE if MAY_APPEND is set */
2948         if ((file->f_flags & O_APPEND) && (mask & MAY_WRITE))
2949                 mask |= MAY_APPEND;
2950 
2951         return file_has_perm(cred, file,
2952                              file_mask_to_av(inode->i_mode, mask));
2953 }
2954 
2955 static int selinux_file_permission(struct file *file, int mask)
2956 {
2957         if (!mask)
2958                 /* No permission to check.  Existence test. */
2959                 return 0;
2960 
2961         return selinux_revalidate_file_permission(file, mask);
2962 }
2963 
2964 static int selinux_file_alloc_security(struct file *file)
2965 {
2966         return file_alloc_security(file);
2967 }
2968 
2969 static void selinux_file_free_security(struct file *file)
2970 {
2971         file_free_security(file);
2972 }
2973 
2974 static int selinux_file_ioctl(struct file *file, unsigned int cmd,
2975                               unsigned long arg)
2976 {
2977         const struct cred *cred = current_cred();
2978         u32 av = 0;
2979 
2980         if (_IOC_DIR(cmd) & _IOC_WRITE)
2981                 av |= FILE__WRITE;
2982         if (_IOC_DIR(cmd) & _IOC_READ)
2983                 av |= FILE__READ;
2984         if (!av)
2985                 av = FILE__IOCTL;
2986 
2987         return file_has_perm(cred, file, av);
2988 }
2989 
2990 static int file_map_prot_check(struct file *file, unsigned long prot, int shared)
2991 {
2992         const struct cred *cred = current_cred();
2993         int rc = 0;
2994 
2995 #ifndef CONFIG_PPC32
2996         if ((prot & PROT_EXEC) && (!file || (!shared && (prot & PROT_WRITE)))) {
2997                 /*
2998                  * We are making executable an anonymous mapping or a
2999                  * private file mapping that will also be writable.
3000                  * This has an additional check.
3001                  */
3002                 rc = cred_has_perm(cred, cred, PROCESS__EXECMEM);
3003                 if (rc)
3004                         goto error;
3005         }
3006 #endif
3007 
3008         if (file) {
3009                 /* read access is always possible with a mapping */
3010                 u32 av = FILE__READ;
3011 
3012                 /* write access only matters if the mapping is shared */
3013                 if (shared && (prot & PROT_WRITE))
3014                         av |= FILE__WRITE;
3015 
3016                 if (prot & PROT_EXEC)
3017                         av |= FILE__EXECUTE;
3018 
3019                 return file_has_perm(cred, file, av);
3020         }
3021 
3022 error:
3023         return rc;
3024 }
3025 
3026 static int selinux_file_mmap(struct file *file, unsigned long reqprot,
3027                              unsigned long prot, unsigned long flags,
3028                              unsigned long addr, unsigned long addr_only)
3029 {
3030         int rc = 0;
3031         u32 sid = current_sid();
3032 
3033         /*
3034          * notice that we are intentionally putting the SELinux check before
3035          * the secondary cap_file_mmap check.  This is such a likely attempt
3036          * at bad behaviour/exploit that we always want to get the AVC, even
3037          * if DAC would have also denied the operation.
3038          */
3039         if (addr < CONFIG_LSM_MMAP_MIN_ADDR) {
3040                 rc = avc_has_perm(sid, sid, SECCLASS_MEMPROTECT,
3041                                   MEMPROTECT__MMAP_ZERO, NULL);
3042                 if (rc)
3043                         return rc;
3044         }
3045 
3046         /* do DAC check on address space usage */
3047         rc = cap_file_mmap(file, reqprot, prot, flags, addr, addr_only);
3048         if (rc || addr_only)
3049                 return rc;
3050 
3051         if (selinux_checkreqprot)
3052                 prot = reqprot;
3053 
3054         return file_map_prot_check(file, prot,
3055                                    (flags & MAP_TYPE) == MAP_SHARED);
3056 }
3057 
3058 static int selinux_file_mprotect(struct vm_area_struct *vma,
3059                                  unsigned long reqprot,
3060                                  unsigned long prot)
3061 {
3062         const struct cred *cred = current_cred();
3063 
3064         if (selinux_checkreqprot)
3065                 prot = reqprot;
3066 
3067 #ifndef CONFIG_PPC32
3068         if ((prot & PROT_EXEC) && !(vma->vm_flags & VM_EXEC)) {
3069                 int rc = 0;
3070                 if (vma->vm_start >= vma->vm_mm->start_brk &&
3071                     vma->vm_end <= vma->vm_mm->brk) {
3072                         rc = cred_has_perm(cred, cred, PROCESS__EXECHEAP);
3073                 } else if (!vma->vm_file &&
3074                            vma->vm_start <= vma->vm_mm->start_stack &&
3075                            vma->vm_end >= vma->vm_mm->start_stack) {
3076                         rc = current_has_perm(current, PROCESS__EXECSTACK);
3077                 } else if (vma->vm_file && vma->anon_vma) {
3078                         /*
3079                          * We are making executable a file mapping that has
3080                          * had some COW done. Since pages might have been
3081                          * written, check ability to execute the possibly
3082                          * modified content.  This typically should only
3083                          * occur for text relocations.
3084                          */
3085                         rc = file_has_perm(cred, vma->vm_file, FILE__EXECMOD);
3086                 }
3087                 if (rc)
3088                         return rc;
3089         }
3090 #endif
3091 
3092         return file_map_prot_check(vma->vm_file, prot, vma->vm_flags&VM_SHARED);
3093 }
3094 
3095 static int selinux_file_lock(struct file *file, unsigned int cmd)
3096 {
3097         const struct cred *cred = current_cred();
3098 
3099         return file_has_perm(cred, file, FILE__LOCK);
3100 }
3101 
3102 static int selinux_file_fcntl(struct file *file, unsigned int cmd,
3103                               unsigned long arg)
3104 {
3105         const struct cred *cred = current_cred();
3106         int err = 0;
3107 
3108         switch (cmd) {
3109         case F_SETFL:
3110                 if (!file->f_path.dentry || !file->f_path.dentry->d_inode) {
3111                         err = -EINVAL;
3112                         break;
3113                 }
3114 
3115                 if ((file->f_flags & O_APPEND) && !(arg & O_APPEND)) {
3116                         err = file_has_perm(cred, file, FILE__WRITE);
3117                         break;
3118                 }
3119                 /* fall through */
3120         case F_SETOWN:
3121         case F_SETSIG:
3122         case F_GETFL:
3123         case F_GETOWN:
3124         case F_GETSIG:
3125                 /* Just check FD__USE permission */
3126                 err = file_has_perm(cred, file, 0);
3127                 break;
3128         case F_GETLK:
3129         case F_SETLK:
3130         case F_SETLKW:
3131 #if BITS_PER_LONG == 32
3132         case F_GETLK64:
3133         case F_SETLK64:
3134         case F_SETLKW64:
3135 #endif
3136                 if (!file->f_path.dentry || !file->f_path.dentry->d_inode) {
3137                         err = -EINVAL;
3138                         break;
3139                 }
3140                 err = file_has_perm(cred, file, FILE__LOCK);
3141                 break;
3142         }
3143 
3144         return err;
3145 }
3146 
3147 static int selinux_file_set_fowner(struct file *file)
3148 {
3149         struct file_security_struct *fsec;
3150 
3151         fsec = file->f_security;
3152         fsec->fown_sid = current_sid();
3153 
3154         return 0;
3155 }
3156 
3157 static int selinux_file_send_sigiotask(struct task_struct *tsk,
3158                                        struct fown_struct *fown, int signum)
3159 {
3160         struct file *file;
3161         u32 sid = task_sid(tsk);
3162         u32 perm;
3163         struct file_security_struct *fsec;
3164 
3165         /* struct fown_struct is never outside the context of a struct file */
3166         file = container_of(fown, struct file, f_owner);
3167 
3168         fsec = file->f_security;
3169 
3170         if (!signum)
3171                 perm = signal_to_av(SIGIO); /* as per send_sigio_to_task */
3172         else
3173                 perm = signal_to_av(signum);
3174 
3175         return avc_has_perm(fsec->fown_sid, sid,
3176                             SECCLASS_PROCESS, perm, NULL);
3177 }
3178 
3179 static int selinux_file_receive(struct file *file)
3180 {
3181         const struct cred *cred = current_cred();
3182 
3183         return file_has_perm(cred, file, file_to_av(file));
3184 }
3185 
3186 static int selinux_dentry_open(struct file *file, const struct cred *cred)
3187 {
3188         struct file_security_struct *fsec;
3189         struct inode *inode;
3190         struct inode_security_struct *isec;
3191 
3192         inode = file->f_path.dentry->d_inode;
3193         fsec = file->f_security;
3194         isec = inode->i_security;
3195         /*
3196          * Save inode label and policy sequence number
3197          * at open-time so that selinux_file_permission
3198          * can determine whether revalidation is necessary.
3199          * Task label is already saved in the file security
3200          * struct as its SID.
3201          */
3202         fsec->isid = isec->sid;
3203         fsec->pseqno = avc_policy_seqno();
3204         /*
3205          * Since the inode label or policy seqno may have changed
3206          * between the selinux_inode_permission check and the saving
3207          * of state above, recheck that access is still permitted.
3208          * Otherwise, access might never be revalidated against the
3209          * new inode label or new policy.
3210          * This check is not redundant - do not remove.
3211          */
3212         return inode_has_perm(cred, inode, open_file_to_av(file), NULL);
3213 }
3214 
3215 /* task security operations */
3216 
3217 static int selinux_task_create(unsigned long clone_flags)
3218 {
3219         return current_has_perm(current, PROCESS__FORK);
3220 }
3221 
3222 /*
3223  * detach and free the LSM part of a set of credentials
3224  */
3225 static void selinux_cred_free(struct cred *cred)
3226 {
3227         struct task_security_struct *tsec = cred->security;
3228         cred->security = NULL;
3229         kfree(tsec);
3230 }
3231 
3232 /*
3233  * prepare a new set of credentials for modification
3234  */
3235 static int selinux_cred_prepare(struct cred *new, const struct cred *old,
3236                                 gfp_t gfp)
3237 {
3238         const struct task_security_struct *old_tsec;
3239         struct task_security_struct *tsec;
3240 
3241         old_tsec = old->security;
3242 
3243         tsec = kmemdup(old_tsec, sizeof(struct task_security_struct), gfp);
3244         if (!tsec)
3245                 return -ENOMEM;
3246 
3247         new->security = tsec;
3248         return 0;
3249 }
3250 
3251 /*
3252  * set the security data for a kernel service
3253  * - all the creation contexts are set to unlabelled
3254  */
3255 static int selinux_kernel_act_as(struct cred *new, u32 secid)
3256 {
3257         struct task_security_struct *tsec = new->security;
3258         u32 sid = current_sid();
3259         int ret;
3260 
3261         ret = avc_has_perm(sid, secid,
3262                            SECCLASS_KERNEL_SERVICE,
3263                            KERNEL_SERVICE__USE_AS_OVERRIDE,
3264                            NULL);
3265         if (ret == 0) {
3266                 tsec->sid = secid;
3267                 tsec->create_sid = 0;
3268                 tsec->keycreate_sid = 0;
3269                 tsec->sockcreate_sid = 0;
3270         }
3271         return ret;
3272 }
3273 
3274 /*
3275  * set the file creation context in a security record to the same as the
3276  * objective context of the specified inode
3277  */
3278 static int selinux_kernel_create_files_as(struct cred *new, struct inode *inode)
3279 {
3280         struct inode_security_struct *isec = inode->i_security;
3281         struct task_security_struct *tsec = new->security;
3282         u32 sid = current_sid();
3283         int ret;
3284 
3285         ret = avc_has_perm(sid, isec->sid,
3286                            SECCLASS_KERNEL_SERVICE,
3287                            KERNEL_SERVICE__CREATE_FILES_AS,
3288                            NULL);
3289 
3290         if (ret == 0)
3291                 tsec->create_sid = isec->sid;
3292         return 0;
3293 }
3294 
3295 static int selinux_task_setpgid(struct task_struct *p, pid_t pgid)
3296 {
3297         return current_has_perm(p, PROCESS__SETPGID);
3298 }
3299 
3300 static int selinux_task_getpgid(struct task_struct *p)
3301 {
3302         return current_has_perm(p, PROCESS__GETPGID);
3303 }
3304 
3305 static int selinux_task_getsid(struct task_struct *p)
3306 {
3307         return current_has_perm(p, PROCESS__GETSESSION);
3308 }
3309 
3310 static void selinux_task_getsecid(struct task_struct *p, u32 *secid)
3311 {
3312         *secid = task_sid(p);
3313 }
3314 
3315 static int selinux_task_setnice(struct task_struct *p, int nice)
3316 {
3317         int rc;
3318 
3319         rc = cap_task_setnice(p, nice);
3320         if (rc)
3321                 return rc;
3322 
3323         return current_has_perm(p, PROCESS__SETSCHED);
3324 }
3325 
3326 static int selinux_task_setioprio(struct task_struct *p, int ioprio)
3327 {
3328         int rc;
3329 
3330         rc = cap_task_setioprio(p, ioprio);
3331         if (rc)
3332                 return rc;
3333 
3334         return current_has_perm(p, PROCESS__SETSCHED);
3335 }
3336 
3337 static int selinux_task_getioprio(struct task_struct *p)
3338 {
3339         return current_has_perm(p, PROCESS__GETSCHED);
3340 }
3341 
3342 static int selinux_task_setrlimit(unsigned int resource, struct rlimit *new_rlim)
3343 {
3344         struct rlimit *old_rlim = current->signal->rlim + resource;
3345 
3346         /* Control the ability to change the hard limit (whether
3347            lowering or raising it), so that the hard limit can
3348            later be used as a safe reset point for the soft limit
3349            upon context transitions.  See selinux_bprm_committing_creds. */
3350         if (old_rlim->rlim_max != new_rlim->rlim_max)
3351                 return current_has_perm(current, PROCESS__SETRLIMIT);
3352 
3353         return 0;
3354 }
3355 
3356 static int selinux_task_setscheduler(struct task_struct *p, int policy, struct sched_param *lp)
3357 {
3358         int rc;
3359 
3360         rc = cap_task_setscheduler(p, policy, lp);
3361         if (rc)
3362                 return rc;
3363 
3364         return current_has_perm(p, PROCESS__SETSCHED);
3365 }
3366 
3367 static int selinux_task_getscheduler(struct task_struct *p)
3368 {
3369         return current_has_perm(p, PROCESS__GETSCHED);
3370 }
3371 
3372 static int selinux_task_movememory(struct task_struct *p)
3373 {
3374         return current_has_perm(p, PROCESS__SETSCHED);
3375 }
3376 
3377 static int selinux_task_kill(struct task_struct *p, struct siginfo *info,
3378                                 int sig, u32 secid)
3379 {
3380         u32 perm;
3381         int rc;
3382 
3383         if (!sig)
3384                 perm = PROCESS__SIGNULL; /* null signal; existence test */
3385         else
3386                 perm = signal_to_av(sig);
3387         if (secid)
3388                 rc = avc_has_perm(secid, task_sid(p),
3389                                   SECCLASS_PROCESS, perm, NULL);
3390         else
3391                 rc = current_has_perm(p, perm);
3392         return rc;
3393 }
3394 
3395 static int selinux_task_wait(struct task_struct *p)
3396 {
3397         return task_has_perm(p, current, PROCESS__SIGCHLD);
3398 }
3399 
3400 static void selinux_task_to_inode(struct task_struct *p,
3401                                   struct inode *inode)
3402 {
3403         struct inode_security_struct *isec = inode->i_security;
3404         u32 sid = task_sid(p);
3405 
3406         isec->sid = sid;
3407         isec->initialized = 1;
3408 }
3409 
3410 /* Returns error only if unable to parse addresses */
3411 static int selinux_parse_skb_ipv4(struct sk_buff *skb,
3412                         struct avc_audit_data *ad, u8 *proto)
3413 {
3414         int offset, ihlen, ret = -EINVAL;
3415         struct iphdr _iph, *ih;
3416 
3417         offset = skb_network_offset(skb);
3418         ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
3419         if (ih == NULL)
3420                 goto out;
3421 
3422         ihlen = ih->ihl * 4;
3423         if (ihlen < sizeof(_iph))
3424                 goto out;
3425 
3426         ad->u.net.v4info.saddr = ih->saddr;
3427         ad->u.net.v4info.daddr = ih->daddr;
3428         ret = 0;
3429 
3430         if (proto)
3431                 *proto = ih->protocol;
3432 
3433         switch (ih->protocol) {
3434         case IPPROTO_TCP: {
3435                 struct tcphdr _tcph, *th;
3436 
3437                 if (ntohs(ih->frag_off) & IP_OFFSET)
3438                         break;
3439 
3440                 offset += ihlen;
3441                 th = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph);
3442                 if (th == NULL)
3443                         break;
3444 
3445                 ad->u.net.sport = th->source;
3446                 ad->u.net.dport = th->dest;
3447                 break;
3448         }
3449 
3450         case IPPROTO_UDP: {
3451                 struct udphdr _udph, *uh;
3452 
3453                 if (ntohs(ih->frag_off) & IP_OFFSET)
3454                         break;
3455 
3456                 offset += ihlen;
3457                 uh = skb_header_pointer(skb, offset, sizeof(_udph), &_udph);
3458                 if (uh == NULL)
3459                         break;
3460 
3461                 ad->u.net.sport = uh->source;
3462                 ad->u.net.dport = uh->dest;
3463                 break;
3464         }
3465 
3466         case IPPROTO_DCCP: {
3467                 struct dccp_hdr _dccph, *dh;
3468 
3469                 if (ntohs(ih->frag_off) & IP_OFFSET)
3470                         break;
3471 
3472                 offset += ihlen;
3473                 dh = skb_header_pointer(skb, offset, sizeof(_dccph), &_dccph);
3474                 if (dh == NULL)
3475                         break;
3476 
3477                 ad->u.net.sport = dh->dccph_sport;
3478                 ad->u.net.dport = dh->dccph_dport;
3479                 break;
3480         }
3481 
3482         default:
3483                 break;
3484         }
3485 out:
3486         return ret;
3487 }
3488 
3489 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
3490 
3491 /* Returns error only if unable to parse addresses */
3492 static int selinux_parse_skb_ipv6(struct sk_buff *skb,
3493                         struct avc_audit_data *ad, u8 *proto)
3494 {
3495         u8 nexthdr;
3496         int ret = -EINVAL, offset;
3497         struct ipv6hdr _ipv6h, *ip6;
3498 
3499         offset = skb_network_offset(skb);
3500         ip6 = skb_header_pointer(skb, offset, sizeof(_ipv6h), &_ipv6h);
3501         if (ip6 == NULL)
3502                 goto out;
3503 
3504         ipv6_addr_copy(&ad->u.net.v6info.saddr, &ip6->saddr);
3505         ipv6_addr_copy(&ad->u.net.v6info.daddr, &ip6->daddr);
3506         ret = 0;
3507 
3508         nexthdr = ip6->nexthdr;
3509         offset += sizeof(_ipv6h);
3510         offset = ipv6_skip_exthdr(skb, offset, &nexthdr);
3511         if (offset < 0)
3512                 goto out;
3513 
3514         if (proto)
3515                 *proto = nexthdr;
3516 
3517         switch (nexthdr) {
3518         case IPPROTO_TCP: {
3519                 struct tcphdr _tcph, *th;
3520 
3521                 th = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph);
3522                 if (th == NULL)
3523                         break;
3524 
3525                 ad->u.net.sport = th->source;
3526                 ad->u.net.dport = th->dest;
3527                 break;
3528         }
3529 
3530         case IPPROTO_UDP: {
3531                 struct udphdr _udph, *uh;
3532 
3533                 uh = skb_header_pointer(skb, offset, sizeof(_udph), &_udph);
3534                 if (uh == NULL)
3535                         break;
3536 
3537                 ad->u.net.sport = uh->source;
3538                 ad->u.net.dport = uh->dest;
3539                 break;
3540         }
3541 
3542         case IPPROTO_DCCP: {
3543                 struct dccp_hdr _dccph, *dh;
3544 
3545                 dh = skb_header_pointer(skb, offset, sizeof(_dccph), &_dccph);
3546                 if (dh == NULL)
3547                         break;
3548 
3549                 ad->u.net.sport = dh->dccph_sport;
3550                 ad->u.net.dport = dh->dccph_dport;
3551                 break;
3552         }
3553 
3554         /* includes fragments */
3555         default:
3556                 break;
3557         }
3558 out:
3559         return ret;
3560 }
3561 
3562 #endif /* IPV6 */
3563 
3564 static int selinux_parse_skb(struct sk_buff *skb, struct avc_audit_data *ad,
3565                              char **_addrp, int src, u8 *proto)
3566 {
3567         char *addrp;
3568         int ret;
3569 
3570         switch (ad->u.net.family) {
3571         case PF_INET:
3572                 ret = selinux_parse_skb_ipv4(skb, ad, proto);
3573                 if (ret)
3574                         goto parse_error;
3575                 addrp = (char *)(src ? &ad->u.net.v4info.saddr :
3576                                        &ad->u.net.v4info.daddr);
3577                 goto okay;
3578 
3579 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
3580         case PF_INET6:
3581                 ret = selinux_parse_skb_ipv6(skb, ad, proto);
3582                 if (ret)
3583                         goto parse_error;
3584                 addrp = (char *)(src ? &ad->u.net.v6info.saddr :
3585                                        &ad->u.net.v6info.daddr);
3586                 goto okay;
3587 #endif  /* IPV6 */
3588         default:
3589                 addrp = NULL;
3590                 goto okay;
3591         }
3592 
3593 parse_error:
3594         printk(KERN_WARNING
3595                "SELinux: failure in selinux_parse_skb(),"
3596                " unable to parse packet\n");
3597         return ret;
3598 
3599 okay:
3600         if (_addrp)
3601                 *_addrp = addrp;
3602         return 0;
3603 }
3604 
3605 /**
3606  * selinux_skb_peerlbl_sid - Determine the peer label of a packet
3607  * @skb: the packet
3608  * @family: protocol family
3609  * @sid: the packet's peer label SID
3610  *
3611  * Description:
3612  * Check the various different forms of network peer labeling and determine
3613  * the peer label/SID for the packet; most of the magic actually occurs in
3614  * the security server function security_net_peersid_cmp().  The function
3615  * returns zero if the value in @sid is valid (although it may be SECSID_NULL)
3616  * or -EACCES if @sid is invalid due to inconsistencies with the different
3617  * peer labels.
3618  *
3619  */
3620 static int selinux_skb_peerlbl_sid(struct sk_buff *skb, u16 family, u32 *sid)
3621 {
3622         int err;
3623         u32 xfrm_sid;
3624         u32 nlbl_sid;
3625         u32 nlbl_type;
3626 
3627         selinux_skb_xfrm_sid(skb, &xfrm_sid);
3628         selinux_netlbl_skbuff_getsid(skb, family, &nlbl_type, &nlbl_sid);
3629 
3630         err = security_net_peersid_resolve(nlbl_sid, nlbl_type, xfrm_sid, sid);
3631         if (unlikely(err)) {
3632                 printk(KERN_WARNING
3633                        "SELinux: failure in selinux_skb_peerlbl_sid(),"
3634                        " unable to determine packet's peer label\n");
3635                 return -EACCES;
3636         }
3637 
3638         return 0;
3639 }
3640 
3641 /* socket security operations */
3642 static int socket_has_perm(struct task_struct *task, struct socket *sock,
3643                            u32 perms)
3644 {
3645         struct inode_security_struct *isec;
3646         struct avc_audit_data ad;
3647         u32 sid;
3648         int err = 0;
3649 
3650         isec = SOCK_INODE(sock)->i_security;
3651 
3652         if (isec->sid == SECINITSID_KERNEL)
3653                 goto out;
3654         sid = task_sid(task);
3655 
3656         AVC_AUDIT_DATA_INIT(&ad, NET);
3657         ad.u.net.sk = sock->sk;
3658         err = avc_has_perm(sid, isec->sid, isec->sclass, perms, &ad);
3659 
3660 out:
3661         return err;
3662 }
3663 
3664 static int selinux_socket_create(int family, int type,
3665                                  int protocol, int kern)
3666 {
3667         const struct cred *cred = current_cred();
3668         const struct task_security_struct *tsec = cred->security;
3669         u32 sid, newsid;
3670         u16 secclass;
3671         int err = 0;
3672 
3673         if (kern)
3674                 goto out;
3675 
3676         sid = tsec->sid;
3677         newsid = tsec->sockcreate_sid ?: sid;
3678 
3679         secclass = socket_type_to_security_class(family, type, protocol);
3680         err = avc_has_perm(sid, newsid, secclass, SOCKET__CREATE, NULL);
3681 
3682 out:
3683         return err;
3684 }
3685 
3686 static int selinux_socket_post_create(struct socket *sock, int family,
3687                                       int type, int protocol, int kern)
3688 {
3689         const struct cred *cred = current_cred();
3690         const struct task_security_struct *tsec = cred->security;
3691         struct inode_security_struct *isec;
3692         struct sk_security_struct *sksec;
3693         u32 sid, newsid;
3694         int err = 0;
3695 
3696         sid = tsec->sid;
3697         newsid = tsec->sockcreate_sid;
3698 
3699         isec = SOCK_INODE(sock)->i_security;
3700 
3701         if (kern)
3702                 isec->sid = SECINITSID_KERNEL;
3703         else if (newsid)
3704                 isec->sid = newsid;
3705         else
3706                 isec->sid = sid;
3707 
3708         isec->sclass = socket_type_to_security_class(family, type, protocol);
3709         isec->initialized = 1;
3710 
3711         if (sock->sk) {
3712                 sksec = sock->sk->sk_security;
3713                 sksec->sid = isec->sid;
3714                 sksec->sclass = isec->sclass;
3715                 err = selinux_netlbl_socket_post_create(sock->sk, family);
3716         }
3717 
3718         return err;
3719 }
3720 
3721 /* Range of port numbers used to automatically bind.
3722    Need to determine whether we should perform a name_bind
3723    permission check between the socket and the port number. */
3724 
3725 static int selinux_socket_bind(struct socket *sock, struct sockaddr *address, int addrlen)
3726 {
3727         u16 family;
3728         int err;
3729 
3730         err = socket_has_perm(current, sock, SOCKET__BIND);
3731         if (err)
3732                 goto out;
3733 
3734         /*
3735          * If PF_INET or PF_INET6, check name_bind permission for the port.
3736          * Multiple address binding for SCTP is not supported yet: we just
3737          * check the first address now.
3738          */
3739         family = sock->sk->sk_family;
3740         if (family == PF_INET || family == PF_INET6) {
3741                 char *addrp;
3742                 struct inode_security_struct *isec;
3743                 struct avc_audit_data ad;
3744                 struct sockaddr_in *addr4 = NULL;
3745                 struct sockaddr_in6 *addr6 = NULL;
3746                 unsigned short snum;
3747                 struct sock *sk = sock->sk;
3748                 u32 sid, node_perm;
3749 
3750                 isec = SOCK_INODE(sock)->i_security;
3751 
3752                 if (family == PF_INET) {
3753                         addr4 = (struct sockaddr_in *)address;
3754                         snum = ntohs(addr4->sin_port);
3755                         addrp = (char *)&addr4->sin_addr.s_addr;
3756                 } else {
3757                         addr6 = (struct sockaddr_in6 *)address;
3758                         snum = ntohs(addr6->sin6_port);
3759                         addrp = (char *)&addr6->sin6_addr.s6_addr;
3760                 }
3761 
3762                 if (snum) {
3763                         int low, high;
3764 
3765                         inet_get_local_port_range(&low, &high);
3766 
3767                         if (snum < max(PROT_SOCK, low) || snum > high) {
3768                                 err = sel_netport_sid(sk->sk_protocol,
3769                                                       snum, &sid);
3770                                 if (err)
3771                                         goto out;
3772                                 AVC_AUDIT_DATA_INIT(&ad, NET);
3773                                 ad.u.net.sport = htons(snum);
3774                                 ad.u.net.family = family;
3775                                 err = avc_has_perm(isec->sid, sid,
3776                                                    isec->sclass,
3777                                                    SOCKET__