tty: Make get_current_tty use a kref
[kernel.git] / security / selinux / hooks.c
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
79 #include "avc.h"
80 #include "objsec.h"
81 #include "netif.h"
82 #include "netnode.h"
83 #include "netport.h"
84 #include "xfrm.h"
85 #include "netlabel.h"
86 #include "audit.h"
87
88 #define XATTR_SELINUX_SUFFIX "selinux"
89 #define XATTR_NAME_SELINUX XATTR_SECURITY_PREFIX XATTR_SELINUX_SUFFIX
90
91 #define NUM_SEL_MNT_OPTS 4
92
93 extern unsigned int policydb_loaded_version;
94 extern int selinux_nlmsg_lookup(u16 sclass, u16 nlmsg_type, u32 *perm);
95 extern int selinux_compat_net;
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 /* Allocate and free functions for each kind of security blob. */
159
160 static int task_alloc_security(struct task_struct *task)
161 {
162         struct task_security_struct *tsec;
163
164         tsec = kzalloc(sizeof(struct task_security_struct), GFP_KERNEL);
165         if (!tsec)
166                 return -ENOMEM;
167
168         tsec->osid = tsec->sid = SECINITSID_UNLABELED;
169         task->security = tsec;
170
171         return 0;
172 }
173
174 static void task_free_security(struct task_struct *task)
175 {
176         struct task_security_struct *tsec = task->security;
177         task->security = NULL;
178         kfree(tsec);
179 }
180
181 static int inode_alloc_security(struct inode *inode)
182 {
183         struct task_security_struct *tsec = current->security;
184         struct inode_security_struct *isec;
185
186         isec = kmem_cache_zalloc(sel_inode_cache, GFP_NOFS);
187         if (!isec)
188                 return -ENOMEM;
189
190         mutex_init(&isec->lock);
191         INIT_LIST_HEAD(&isec->list);
192         isec->inode = inode;
193         isec->sid = SECINITSID_UNLABELED;
194         isec->sclass = SECCLASS_FILE;
195         isec->task_sid = tsec->sid;
196         inode->i_security = isec;
197
198         return 0;
199 }
200
201 static void inode_free_security(struct inode *inode)
202 {
203         struct inode_security_struct *isec = inode->i_security;
204         struct superblock_security_struct *sbsec = inode->i_sb->s_security;
205
206         spin_lock(&sbsec->isec_lock);
207         if (!list_empty(&isec->list))
208                 list_del_init(&isec->list);
209         spin_unlock(&sbsec->isec_lock);
210
211         inode->i_security = NULL;
212         kmem_cache_free(sel_inode_cache, isec);
213 }
214
215 static int file_alloc_security(struct file *file)
216 {
217         struct task_security_struct *tsec = current->security;
218         struct file_security_struct *fsec;
219
220         fsec = kzalloc(sizeof(struct file_security_struct), GFP_KERNEL);
221         if (!fsec)
222                 return -ENOMEM;
223
224         fsec->sid = tsec->sid;
225         fsec->fown_sid = tsec->sid;
226         file->f_security = fsec;
227
228         return 0;
229 }
230
231 static void file_free_security(struct file *file)
232 {
233         struct file_security_struct *fsec = file->f_security;
234         file->f_security = NULL;
235         kfree(fsec);
236 }
237
238 static int superblock_alloc_security(struct super_block *sb)
239 {
240         struct superblock_security_struct *sbsec;
241
242         sbsec = kzalloc(sizeof(struct superblock_security_struct), GFP_KERNEL);
243         if (!sbsec)
244                 return -ENOMEM;
245
246         mutex_init(&sbsec->lock);
247         INIT_LIST_HEAD(&sbsec->list);
248         INIT_LIST_HEAD(&sbsec->isec_head);
249         spin_lock_init(&sbsec->isec_lock);
250         sbsec->sb = sb;
251         sbsec->sid = SECINITSID_UNLABELED;
252         sbsec->def_sid = SECINITSID_FILE;
253         sbsec->mntpoint_sid = SECINITSID_UNLABELED;
254         sb->s_security = sbsec;
255
256         return 0;
257 }
258
259 static void superblock_free_security(struct super_block *sb)
260 {
261         struct superblock_security_struct *sbsec = sb->s_security;
262
263         spin_lock(&sb_security_lock);
264         if (!list_empty(&sbsec->list))
265                 list_del_init(&sbsec->list);
266         spin_unlock(&sb_security_lock);
267
268         sb->s_security = NULL;
269         kfree(sbsec);
270 }
271
272 static int sk_alloc_security(struct sock *sk, int family, gfp_t priority)
273 {
274         struct sk_security_struct *ssec;
275
276         ssec = kzalloc(sizeof(*ssec), priority);
277         if (!ssec)
278                 return -ENOMEM;
279
280         ssec->peer_sid = SECINITSID_UNLABELED;
281         ssec->sid = SECINITSID_UNLABELED;
282         sk->sk_security = ssec;
283
284         selinux_netlbl_sk_security_reset(ssec, family);
285
286         return 0;
287 }
288
289 static void sk_free_security(struct sock *sk)
290 {
291         struct sk_security_struct *ssec = sk->sk_security;
292
293         sk->sk_security = NULL;
294         kfree(ssec);
295 }
296
297 /* The security server must be initialized before
298    any labeling or access decisions can be provided. */
299 extern int ss_initialized;
300
301 /* The file system's label must be initialized prior to use. */
302
303 static char *labeling_behaviors[6] = {
304         "uses xattr",
305         "uses transition SIDs",
306         "uses task SIDs",
307         "uses genfs_contexts",
308         "not configured for labeling",
309         "uses mountpoint labeling",
310 };
311
312 static int inode_doinit_with_dentry(struct inode *inode, struct dentry *opt_dentry);
313
314 static inline int inode_doinit(struct inode *inode)
315 {
316         return inode_doinit_with_dentry(inode, NULL);
317 }
318
319 enum {
320         Opt_error = -1,
321         Opt_context = 1,
322         Opt_fscontext = 2,
323         Opt_defcontext = 3,
324         Opt_rootcontext = 4,
325 };
326
327 static match_table_t tokens = {
328         {Opt_context, CONTEXT_STR "%s"},
329         {Opt_fscontext, FSCONTEXT_STR "%s"},
330         {Opt_defcontext, DEFCONTEXT_STR "%s"},
331         {Opt_rootcontext, ROOTCONTEXT_STR "%s"},
332         {Opt_error, NULL},
333 };
334
335 #define SEL_MOUNT_FAIL_MSG "SELinux:  duplicate or incompatible mount options\n"
336
337 static int may_context_mount_sb_relabel(u32 sid,
338                         struct superblock_security_struct *sbsec,
339                         struct task_security_struct *tsec)
340 {
341         int rc;
342
343         rc = avc_has_perm(tsec->sid, sbsec->sid, SECCLASS_FILESYSTEM,
344                           FILESYSTEM__RELABELFROM, NULL);
345         if (rc)
346                 return rc;
347
348         rc = avc_has_perm(tsec->sid, sid, SECCLASS_FILESYSTEM,
349                           FILESYSTEM__RELABELTO, NULL);
350         return rc;
351 }
352
353 static int may_context_mount_inode_relabel(u32 sid,
354                         struct superblock_security_struct *sbsec,
355                         struct task_security_struct *tsec)
356 {
357         int rc;
358         rc = avc_has_perm(tsec->sid, sbsec->sid, SECCLASS_FILESYSTEM,
359                           FILESYSTEM__RELABELFROM, NULL);
360         if (rc)
361                 return rc;
362
363         rc = avc_has_perm(sid, sbsec->sid, SECCLASS_FILESYSTEM,
364                           FILESYSTEM__ASSOCIATE, NULL);
365         return rc;
366 }
367
368 static int sb_finish_set_opts(struct super_block *sb)
369 {
370         struct superblock_security_struct *sbsec = sb->s_security;
371         struct dentry *root = sb->s_root;
372         struct inode *root_inode = root->d_inode;
373         int rc = 0;
374
375         if (sbsec->behavior == SECURITY_FS_USE_XATTR) {
376                 /* Make sure that the xattr handler exists and that no
377                    error other than -ENODATA is returned by getxattr on
378                    the root directory.  -ENODATA is ok, as this may be
379                    the first boot of the SELinux kernel before we have
380                    assigned xattr values to the filesystem. */
381                 if (!root_inode->i_op->getxattr) {
382                         printk(KERN_WARNING "SELinux: (dev %s, type %s) has no "
383                                "xattr support\n", sb->s_id, sb->s_type->name);
384                         rc = -EOPNOTSUPP;
385                         goto out;
386                 }
387                 rc = root_inode->i_op->getxattr(root, XATTR_NAME_SELINUX, NULL, 0);
388                 if (rc < 0 && rc != -ENODATA) {
389                         if (rc == -EOPNOTSUPP)
390                                 printk(KERN_WARNING "SELinux: (dev %s, type "
391                                        "%s) has no security xattr handler\n",
392                                        sb->s_id, sb->s_type->name);
393                         else
394                                 printk(KERN_WARNING "SELinux: (dev %s, type "
395                                        "%s) getxattr errno %d\n", sb->s_id,
396                                        sb->s_type->name, -rc);
397                         goto out;
398                 }
399         }
400
401         sbsec->initialized = 1;
402
403         if (sbsec->behavior > ARRAY_SIZE(labeling_behaviors))
404                 printk(KERN_ERR "SELinux: initialized (dev %s, type %s), unknown behavior\n",
405                        sb->s_id, sb->s_type->name);
406         else
407                 printk(KERN_DEBUG "SELinux: initialized (dev %s, type %s), %s\n",
408                        sb->s_id, sb->s_type->name,
409                        labeling_behaviors[sbsec->behavior-1]);
410
411         /* Initialize the root inode. */
412         rc = inode_doinit_with_dentry(root_inode, root);
413
414         /* Initialize any other inodes associated with the superblock, e.g.
415            inodes created prior to initial policy load or inodes created
416            during get_sb by a pseudo filesystem that directly
417            populates itself. */
418         spin_lock(&sbsec->isec_lock);
419 next_inode:
420         if (!list_empty(&sbsec->isec_head)) {
421                 struct inode_security_struct *isec =
422                                 list_entry(sbsec->isec_head.next,
423                                            struct inode_security_struct, list);
424                 struct inode *inode = isec->inode;
425                 spin_unlock(&sbsec->isec_lock);
426                 inode = igrab(inode);
427                 if (inode) {
428                         if (!IS_PRIVATE(inode))
429                                 inode_doinit(inode);
430                         iput(inode);
431                 }
432                 spin_lock(&sbsec->isec_lock);
433                 list_del_init(&isec->list);
434                 goto next_inode;
435         }
436         spin_unlock(&sbsec->isec_lock);
437 out:
438         return rc;
439 }
440
441 /*
442  * This function should allow an FS to ask what it's mount security
443  * options were so it can use those later for submounts, displaying
444  * mount options, or whatever.
445  */
446 static int selinux_get_mnt_opts(const struct super_block *sb,
447                                 struct security_mnt_opts *opts)
448 {
449         int rc = 0, i;
450         struct superblock_security_struct *sbsec = sb->s_security;
451         char *context = NULL;
452         u32 len;
453         char tmp;
454
455         security_init_mnt_opts(opts);
456
457         if (!sbsec->initialized)
458                 return -EINVAL;
459
460         if (!ss_initialized)
461                 return -EINVAL;
462
463         /*
464          * if we ever use sbsec flags for anything other than tracking mount
465          * settings this is going to need a mask
466          */
467         tmp = sbsec->flags;
468         /* count the number of mount options for this sb */
469         for (i = 0; i < 8; i++) {
470                 if (tmp & 0x01)
471                         opts->num_mnt_opts++;
472                 tmp >>= 1;
473         }
474
475         opts->mnt_opts = kcalloc(opts->num_mnt_opts, sizeof(char *), GFP_ATOMIC);
476         if (!opts->mnt_opts) {
477                 rc = -ENOMEM;
478                 goto out_free;
479         }
480
481         opts->mnt_opts_flags = kcalloc(opts->num_mnt_opts, sizeof(int), GFP_ATOMIC);
482         if (!opts->mnt_opts_flags) {
483                 rc = -ENOMEM;
484                 goto out_free;
485         }
486
487         i = 0;
488         if (sbsec->flags & FSCONTEXT_MNT) {
489                 rc = security_sid_to_context(sbsec->sid, &context, &len);
490                 if (rc)
491                         goto out_free;
492                 opts->mnt_opts[i] = context;
493                 opts->mnt_opts_flags[i++] = FSCONTEXT_MNT;
494         }
495         if (sbsec->flags & CONTEXT_MNT) {
496                 rc = security_sid_to_context(sbsec->mntpoint_sid, &context, &len);
497                 if (rc)
498                         goto out_free;
499                 opts->mnt_opts[i] = context;
500                 opts->mnt_opts_flags[i++] = CONTEXT_MNT;
501         }
502         if (sbsec->flags & DEFCONTEXT_MNT) {
503                 rc = security_sid_to_context(sbsec->def_sid, &context, &len);
504                 if (rc)
505                         goto out_free;
506                 opts->mnt_opts[i] = context;
507                 opts->mnt_opts_flags[i++] = DEFCONTEXT_MNT;
508         }
509         if (sbsec->flags & ROOTCONTEXT_MNT) {
510                 struct inode *root = sbsec->sb->s_root->d_inode;
511                 struct inode_security_struct *isec = root->i_security;
512
513                 rc = security_sid_to_context(isec->sid, &context, &len);
514                 if (rc)
515                         goto out_free;
516                 opts->mnt_opts[i] = context;
517                 opts->mnt_opts_flags[i++] = ROOTCONTEXT_MNT;
518         }
519
520         BUG_ON(i != opts->num_mnt_opts);
521
522         return 0;
523
524 out_free:
525         security_free_mnt_opts(opts);
526         return rc;
527 }
528
529 static int bad_option(struct superblock_security_struct *sbsec, char flag,
530                       u32 old_sid, u32 new_sid)
531 {
532         /* check if the old mount command had the same options */
533         if (sbsec->initialized)
534                 if (!(sbsec->flags & flag) ||
535                     (old_sid != new_sid))
536                         return 1;
537
538         /* check if we were passed the same options twice,
539          * aka someone passed context=a,context=b
540          */
541         if (!sbsec->initialized)
542                 if (sbsec->flags & flag)
543                         return 1;
544         return 0;
545 }
546
547 /*
548  * Allow filesystems with binary mount data to explicitly set mount point
549  * labeling information.
550  */
551 static int selinux_set_mnt_opts(struct super_block *sb,
552                                 struct security_mnt_opts *opts)
553 {
554         int rc = 0, i;
555         struct task_security_struct *tsec = current->security;
556         struct superblock_security_struct *sbsec = sb->s_security;
557         const char *name = sb->s_type->name;
558         struct inode *inode = sbsec->sb->s_root->d_inode;
559         struct inode_security_struct *root_isec = inode->i_security;
560         u32 fscontext_sid = 0, context_sid = 0, rootcontext_sid = 0;
561         u32 defcontext_sid = 0;
562         char **mount_options = opts->mnt_opts;
563         int *flags = opts->mnt_opts_flags;
564         int num_opts = opts->num_mnt_opts;
565
566         mutex_lock(&sbsec->lock);
567
568         if (!ss_initialized) {
569                 if (!num_opts) {
570                         /* Defer initialization until selinux_complete_init,
571                            after the initial policy is loaded and the security
572                            server is ready to handle calls. */
573                         spin_lock(&sb_security_lock);
574                         if (list_empty(&sbsec->list))
575                                 list_add(&sbsec->list, &superblock_security_head);
576                         spin_unlock(&sb_security_lock);
577                         goto out;
578                 }
579                 rc = -EINVAL;
580                 printk(KERN_WARNING "SELinux: Unable to set superblock options "
581                         "before the security server is initialized\n");
582                 goto out;
583         }
584
585         /*
586          * Binary mount data FS will come through this function twice.  Once
587          * from an explicit call and once from the generic calls from the vfs.
588          * Since the generic VFS calls will not contain any security mount data
589          * we need to skip the double mount verification.
590          *
591          * This does open a hole in which we will not notice if the first
592          * mount using this sb set explict options and a second mount using
593          * this sb does not set any security options.  (The first options
594          * will be used for both mounts)
595          */
596         if (sbsec->initialized && (sb->s_type->fs_flags & FS_BINARY_MOUNTDATA)
597             && (num_opts == 0))
598                 goto out;
599
600         /*
601          * parse the mount options, check if they are valid sids.
602          * also check if someone is trying to mount the same sb more
603          * than once with different security options.
604          */
605         for (i = 0; i < num_opts; i++) {
606                 u32 sid;
607                 rc = security_context_to_sid(mount_options[i],
608                                              strlen(mount_options[i]), &sid);
609                 if (rc) {
610                         printk(KERN_WARNING "SELinux: security_context_to_sid"
611                                "(%s) failed for (dev %s, type %s) errno=%d\n",
612                                mount_options[i], sb->s_id, name, rc);
613                         goto out;
614                 }
615                 switch (flags[i]) {
616                 case FSCONTEXT_MNT:
617                         fscontext_sid = sid;
618
619                         if (bad_option(sbsec, FSCONTEXT_MNT, sbsec->sid,
620                                         fscontext_sid))
621                                 goto out_double_mount;
622
623                         sbsec->flags |= FSCONTEXT_MNT;
624                         break;
625                 case CONTEXT_MNT:
626                         context_sid = sid;
627
628                         if (bad_option(sbsec, CONTEXT_MNT, sbsec->mntpoint_sid,
629                                         context_sid))
630                                 goto out_double_mount;
631
632                         sbsec->flags |= CONTEXT_MNT;
633                         break;
634                 case ROOTCONTEXT_MNT:
635                         rootcontext_sid = sid;
636
637                         if (bad_option(sbsec, ROOTCONTEXT_MNT, root_isec->sid,
638                                         rootcontext_sid))
639                                 goto out_double_mount;
640
641                         sbsec->flags |= ROOTCONTEXT_MNT;
642
643                         break;
644                 case DEFCONTEXT_MNT:
645                         defcontext_sid = sid;
646
647                         if (bad_option(sbsec, DEFCONTEXT_MNT, sbsec->def_sid,
648                                         defcontext_sid))
649                                 goto out_double_mount;
650
651                         sbsec->flags |= DEFCONTEXT_MNT;
652
653                         break;
654                 default:
655                         rc = -EINVAL;
656                         goto out;
657                 }
658         }
659
660         if (sbsec->initialized) {
661                 /* previously mounted with options, but not on this attempt? */
662                 if (sbsec->flags && !num_opts)
663                         goto out_double_mount;
664                 rc = 0;
665                 goto out;
666         }
667
668         if (strcmp(sb->s_type->name, "proc") == 0)
669                 sbsec->proc = 1;
670
671         /* Determine the labeling behavior to use for this filesystem type. */
672         rc = security_fs_use(sb->s_type->name, &sbsec->behavior, &sbsec->sid);
673         if (rc) {
674                 printk(KERN_WARNING "%s: security_fs_use(%s) returned %d\n",
675                        __func__, sb->s_type->name, rc);
676                 goto out;
677         }
678
679         /* sets the context of the superblock for the fs being mounted. */
680         if (fscontext_sid) {
681
682                 rc = may_context_mount_sb_relabel(fscontext_sid, sbsec, tsec);
683                 if (rc)
684                         goto out;
685
686                 sbsec->sid = fscontext_sid;
687         }
688
689         /*
690          * Switch to using mount point labeling behavior.
691          * sets the label used on all file below the mountpoint, and will set
692          * the superblock context if not already set.
693          */
694         if (context_sid) {
695                 if (!fscontext_sid) {
696                         rc = may_context_mount_sb_relabel(context_sid, sbsec, tsec);
697                         if (rc)
698                                 goto out;
699                         sbsec->sid = context_sid;
700                 } else {
701                         rc = may_context_mount_inode_relabel(context_sid, sbsec, tsec);
702                         if (rc)
703                                 goto out;
704                 }
705                 if (!rootcontext_sid)
706                         rootcontext_sid = context_sid;
707
708                 sbsec->mntpoint_sid = context_sid;
709                 sbsec->behavior = SECURITY_FS_USE_MNTPOINT;
710         }
711
712         if (rootcontext_sid) {
713                 rc = may_context_mount_inode_relabel(rootcontext_sid, sbsec, tsec);
714                 if (rc)
715                         goto out;
716
717                 root_isec->sid = rootcontext_sid;
718                 root_isec->initialized = 1;
719         }
720
721         if (defcontext_sid) {
722                 if (sbsec->behavior != SECURITY_FS_USE_XATTR) {
723                         rc = -EINVAL;
724                         printk(KERN_WARNING "SELinux: defcontext option is "
725                                "invalid for this filesystem type\n");
726                         goto out;
727                 }
728
729                 if (defcontext_sid != sbsec->def_sid) {
730                         rc = may_context_mount_inode_relabel(defcontext_sid,
731                                                              sbsec, tsec);
732                         if (rc)
733                                 goto out;
734                 }
735
736                 sbsec->def_sid = defcontext_sid;
737         }
738
739         rc = sb_finish_set_opts(sb);
740 out:
741         mutex_unlock(&sbsec->lock);
742         return rc;
743 out_double_mount:
744         rc = -EINVAL;
745         printk(KERN_WARNING "SELinux: mount invalid.  Same superblock, different "
746                "security settings for (dev %s, type %s)\n", sb->s_id, name);
747         goto out;
748 }
749
750 static void selinux_sb_clone_mnt_opts(const struct super_block *oldsb,
751                                         struct super_block *newsb)
752 {
753         const struct superblock_security_struct *oldsbsec = oldsb->s_security;
754         struct superblock_security_struct *newsbsec = newsb->s_security;
755
756         int set_fscontext =     (oldsbsec->flags & FSCONTEXT_MNT);
757         int set_context =       (oldsbsec->flags & CONTEXT_MNT);
758         int set_rootcontext =   (oldsbsec->flags & ROOTCONTEXT_MNT);
759
760         /*
761          * if the parent was able to be mounted it clearly had no special lsm
762          * mount options.  thus we can safely put this sb on the list and deal
763          * with it later
764          */
765         if (!ss_initialized) {
766                 spin_lock(&sb_security_lock);
767                 if (list_empty(&newsbsec->list))
768                         list_add(&newsbsec->list, &superblock_security_head);
769                 spin_unlock(&sb_security_lock);
770                 return;
771         }
772
773         /* how can we clone if the old one wasn't set up?? */
774         BUG_ON(!oldsbsec->initialized);
775
776         /* if fs is reusing a sb, just let its options stand... */
777         if (newsbsec->initialized)
778                 return;
779
780         mutex_lock(&newsbsec->lock);
781
782         newsbsec->flags = oldsbsec->flags;
783
784         newsbsec->sid = oldsbsec->sid;
785         newsbsec->def_sid = oldsbsec->def_sid;
786         newsbsec->behavior = oldsbsec->behavior;
787
788         if (set_context) {
789                 u32 sid = oldsbsec->mntpoint_sid;
790
791                 if (!set_fscontext)
792                         newsbsec->sid = sid;
793                 if (!set_rootcontext) {
794                         struct inode *newinode = newsb->s_root->d_inode;
795                         struct inode_security_struct *newisec = newinode->i_security;
796                         newisec->sid = sid;
797                 }
798                 newsbsec->mntpoint_sid = sid;
799         }
800         if (set_rootcontext) {
801                 const struct inode *oldinode = oldsb->s_root->d_inode;
802                 const struct inode_security_struct *oldisec = oldinode->i_security;
803                 struct inode *newinode = newsb->s_root->d_inode;
804                 struct inode_security_struct *newisec = newinode->i_security;
805
806                 newisec->sid = oldisec->sid;
807         }
808
809         sb_finish_set_opts(newsb);
810         mutex_unlock(&newsbsec->lock);
811 }
812
813 static int selinux_parse_opts_str(char *options,
814                                   struct security_mnt_opts *opts)
815 {
816         char *p;
817         char *context = NULL, *defcontext = NULL;
818         char *fscontext = NULL, *rootcontext = NULL;
819         int rc, num_mnt_opts = 0;
820
821         opts->num_mnt_opts = 0;
822
823         /* Standard string-based options. */
824         while ((p = strsep(&options, "|")) != NULL) {
825                 int token;
826                 substring_t args[MAX_OPT_ARGS];
827
828                 if (!*p)
829                         continue;
830
831                 token = match_token(p, tokens, args);
832
833                 switch (token) {
834                 case Opt_context:
835                         if (context || defcontext) {
836                                 rc = -EINVAL;
837                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
838                                 goto out_err;
839                         }
840                         context = match_strdup(&args[0]);
841                         if (!context) {
842                                 rc = -ENOMEM;
843                                 goto out_err;
844                         }
845                         break;
846
847                 case Opt_fscontext:
848                         if (fscontext) {
849                                 rc = -EINVAL;
850                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
851                                 goto out_err;
852                         }
853                         fscontext = match_strdup(&args[0]);
854                         if (!fscontext) {
855                                 rc = -ENOMEM;
856                                 goto out_err;
857                         }
858                         break;
859
860                 case Opt_rootcontext:
861                         if (rootcontext) {
862                                 rc = -EINVAL;
863                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
864                                 goto out_err;
865                         }
866                         rootcontext = match_strdup(&args[0]);
867                         if (!rootcontext) {
868                                 rc = -ENOMEM;
869                                 goto out_err;
870                         }
871                         break;
872
873                 case Opt_defcontext:
874                         if (context || defcontext) {
875                                 rc = -EINVAL;
876                                 printk(KERN_WARNING SEL_MOUNT_FAIL_MSG);
877                                 goto out_err;
878                         }
879                         defcontext = match_strdup(&args[0]);
880                         if (!defcontext) {
881                                 rc = -ENOMEM;
882                                 goto out_err;
883                         }
884                         break;
885
886                 default:
887                         rc = -EINVAL;
888                         printk(KERN_WARNING "SELinux:  unknown mount option\n");
889                         goto out_err;
890
891                 }
892         }
893
894         rc = -ENOMEM;
895         opts->mnt_opts = kcalloc(NUM_SEL_MNT_OPTS, sizeof(char *), GFP_ATOMIC);
896         if (!opts->mnt_opts)
897                 goto out_err;
898
899         opts->mnt_opts_flags = kcalloc(NUM_SEL_MNT_OPTS, sizeof(int), GFP_ATOMIC);
900         if (!opts->mnt_opts_flags) {
901                 kfree(opts->mnt_opts);
902                 goto out_err;
903         }
904
905         if (fscontext) {
906                 opts->mnt_opts[num_mnt_opts] = fscontext;
907                 opts->mnt_opts_flags[num_mnt_opts++] = FSCONTEXT_MNT;
908         }
909         if (context) {
910                 opts->mnt_opts[num_mnt_opts] = context;
911                 opts->mnt_opts_flags[num_mnt_opts++] = CONTEXT_MNT;
912         }
913         if (rootcontext) {
914                 opts->mnt_opts[num_mnt_opts] = rootcontext;
915                 opts->mnt_opts_flags[num_mnt_opts++] = ROOTCONTEXT_MNT;
916         }
917         if (defcontext) {
918                 opts->mnt_opts[num_mnt_opts] = defcontext;
919                 opts->mnt_opts_flags[num_mnt_opts++] = DEFCONTEXT_MNT;
920         }
921
922         opts->num_mnt_opts = num_mnt_opts;
923         return 0;
924
925 out_err:
926         kfree(context);
927         kfree(defcontext);
928         kfree(fscontext);
929         kfree(rootcontext);
930         return rc;
931 }
932 /*
933  * string mount options parsing and call set the sbsec
934  */
935 static int superblock_doinit(struct super_block *sb, void *data)
936 {
937         int rc = 0;
938         char *options = data;
939         struct security_mnt_opts opts;
940
941         security_init_mnt_opts(&opts);
942
943         if (!data)
944                 goto out;
945
946         BUG_ON(sb->s_type->fs_flags & FS_BINARY_MOUNTDATA);
947
948         rc = selinux_parse_opts_str(options, &opts);
949         if (rc)
950                 goto out_err;
951
952 out:
953         rc = selinux_set_mnt_opts(sb, &opts);
954
955 out_err:
956         security_free_mnt_opts(&opts);
957         return rc;
958 }
959
960 static void selinux_write_opts(struct seq_file *m,
961                                struct security_mnt_opts *opts)
962 {
963         int i;
964         char *prefix;
965
966         for (i = 0; i < opts->num_mnt_opts; i++) {
967                 char *has_comma = strchr(opts->mnt_opts[i], ',');
968
969                 switch (opts->mnt_opts_flags[i]) {
970                 case CONTEXT_MNT:
971                         prefix = CONTEXT_STR;
972                         break;
973                 case FSCONTEXT_MNT:
974                         prefix = FSCONTEXT_STR;
975                         break;
976                 case ROOTCONTEXT_MNT:
977                         prefix = ROOTCONTEXT_STR;
978                         break;
979                 case DEFCONTEXT_MNT:
980                         prefix = DEFCONTEXT_STR;
981                         break;
982                 default:
983                         BUG();
984                 };
985                 /* we need a comma before each option */
986                 seq_putc(m, ',');
987                 seq_puts(m, prefix);
988                 if (has_comma)
989                         seq_putc(m, '\"');
990                 seq_puts(m, opts->mnt_opts[i]);
991                 if (has_comma)
992                         seq_putc(m, '\"');
993         }
994 }
995
996 static int selinux_sb_show_options(struct seq_file *m, struct super_block *sb)
997 {
998         struct security_mnt_opts opts;
999         int rc;
1000
1001         rc = selinux_get_mnt_opts(sb, &opts);
1002         if (rc) {
1003                 /* before policy load we may get EINVAL, don't show anything */
1004                 if (rc == -EINVAL)
1005                         rc = 0;
1006                 return rc;
1007         }
1008
1009         selinux_write_opts(m, &opts);
1010
1011         security_free_mnt_opts(&opts);
1012
1013         return rc;
1014 }
1015
1016 static inline u16 inode_mode_to_security_class(umode_t mode)
1017 {
1018         switch (mode & S_IFMT) {
1019         case S_IFSOCK:
1020                 return SECCLASS_SOCK_FILE;
1021         case S_IFLNK:
1022                 return SECCLASS_LNK_FILE;
1023         case S_IFREG:
1024                 return SECCLASS_FILE;
1025         case S_IFBLK:
1026                 return SECCLASS_BLK_FILE;
1027         case S_IFDIR:
1028                 return SECCLASS_DIR;
1029         case S_IFCHR:
1030                 return SECCLASS_CHR_FILE;
1031         case S_IFIFO:
1032                 return SECCLASS_FIFO_FILE;
1033
1034         }
1035
1036         return SECCLASS_FILE;
1037 }
1038
1039 static inline int default_protocol_stream(int protocol)
1040 {
1041         return (protocol == IPPROTO_IP || protocol == IPPROTO_TCP);
1042 }
1043
1044 static inline int default_protocol_dgram(int protocol)
1045 {
1046         return (protocol == IPPROTO_IP || protocol == IPPROTO_UDP);
1047 }
1048
1049 static inline u16 socket_type_to_security_class(int family, int type, int protocol)
1050 {
1051         switch (family) {
1052         case PF_UNIX:
1053                 switch (type) {
1054                 case SOCK_STREAM:
1055                 case SOCK_SEQPACKET:
1056                         return SECCLASS_UNIX_STREAM_SOCKET;
1057                 case SOCK_DGRAM:
1058                         return SECCLASS_UNIX_DGRAM_SOCKET;
1059                 }
1060                 break;
1061         case PF_INET:
1062         case PF_INET6:
1063                 switch (type) {
1064                 case SOCK_STREAM:
1065                         if (default_protocol_stream(protocol))
1066                                 return SECCLASS_TCP_SOCKET;
1067                         else
1068                                 return SECCLASS_RAWIP_SOCKET;
1069                 case SOCK_DGRAM:
1070                         if (default_protocol_dgram(protocol))
1071                                 return SECCLASS_UDP_SOCKET;
1072                         else
1073                                 return SECCLASS_RAWIP_SOCKET;
1074                 case SOCK_DCCP:
1075                         return SECCLASS_DCCP_SOCKET;
1076                 default:
1077                         return SECCLASS_RAWIP_SOCKET;
1078                 }
1079                 break;
1080         case PF_NETLINK:
1081                 switch (protocol) {
1082                 case NETLINK_ROUTE:
1083                         return SECCLASS_NETLINK_ROUTE_SOCKET;
1084                 case NETLINK_FIREWALL:
1085                         return SECCLASS_NETLINK_FIREWALL_SOCKET;
1086                 case NETLINK_INET_DIAG:
1087                         return SECCLASS_NETLINK_TCPDIAG_SOCKET;
1088                 case NETLINK_NFLOG:
1089                         return SECCLASS_NETLINK_NFLOG_SOCKET;
1090                 case NETLINK_XFRM:
1091                         return SECCLASS_NETLINK_XFRM_SOCKET;
1092                 case NETLINK_SELINUX:
1093                         return SECCLASS_NETLINK_SELINUX_SOCKET;
1094                 case NETLINK_AUDIT:
1095                         return SECCLASS_NETLINK_AUDIT_SOCKET;
1096                 case NETLINK_IP6_FW:
1097                         return SECCLASS_NETLINK_IP6FW_SOCKET;
1098                 case NETLINK_DNRTMSG:
1099                         return SECCLASS_NETLINK_DNRT_SOCKET;
1100                 case NETLINK_KOBJECT_UEVENT:
1101                         return SECCLASS_NETLINK_KOBJECT_UEVENT_SOCKET;
1102                 default:
1103                         return SECCLASS_NETLINK_SOCKET;
1104                 }
1105         case PF_PACKET:
1106                 return SECCLASS_PACKET_SOCKET;
1107         case PF_KEY:
1108                 return SECCLASS_KEY_SOCKET;
1109         case PF_APPLETALK:
1110                 return SECCLASS_APPLETALK_SOCKET;
1111         }
1112
1113         return SECCLASS_SOCKET;
1114 }
1115
1116 #ifdef CONFIG_PROC_FS
1117 static int selinux_proc_get_sid(struct proc_dir_entry *de,
1118                                 u16 tclass,
1119                                 u32 *sid)
1120 {
1121         int buflen, rc;
1122         char *buffer, *path, *end;
1123
1124         buffer = (char *)__get_free_page(GFP_KERNEL);
1125         if (!buffer)
1126                 return -ENOMEM;
1127
1128         buflen = PAGE_SIZE;
1129         end = buffer+buflen;
1130         *--end = '\0';
1131         buflen--;
1132         path = end-1;
1133         *path = '/';
1134         while (de && de != de->parent) {
1135                 buflen -= de->namelen + 1;
1136                 if (buflen < 0)
1137                         break;
1138                 end -= de->namelen;
1139                 memcpy(end, de->name, de->namelen);
1140                 *--end = '/';
1141                 path = end;
1142                 de = de->parent;
1143         }
1144         rc = security_genfs_sid("proc", path, tclass, sid);
1145         free_page((unsigned long)buffer);
1146         return rc;
1147 }
1148 #else
1149 static int selinux_proc_get_sid(struct proc_dir_entry *de,
1150                                 u16 tclass,
1151                                 u32 *sid)
1152 {
1153         return -EINVAL;
1154 }
1155 #endif
1156
1157 /* The inode's security attributes must be initialized before first use. */
1158 static int inode_doinit_with_dentry(struct inode *inode, struct dentry *opt_dentry)
1159 {
1160         struct superblock_security_struct *sbsec = NULL;
1161         struct inode_security_struct *isec = inode->i_security;
1162         u32 sid;
1163         struct dentry *dentry;
1164 #define INITCONTEXTLEN 255
1165         char *context = NULL;
1166         unsigned len = 0;
1167         int rc = 0;
1168
1169         if (isec->initialized)
1170                 goto out;
1171
1172         mutex_lock(&isec->lock);
1173         if (isec->initialized)
1174                 goto out_unlock;
1175
1176         sbsec = inode->i_sb->s_security;
1177         if (!sbsec->initialized) {
1178                 /* Defer initialization until selinux_complete_init,
1179                    after the initial policy is loaded and the security
1180                    server is ready to handle calls. */
1181                 spin_lock(&sbsec->isec_lock);
1182                 if (list_empty(&isec->list))
1183                         list_add(&isec->list, &sbsec->isec_head);
1184                 spin_unlock(&sbsec->isec_lock);
1185                 goto out_unlock;
1186         }
1187
1188         switch (sbsec->behavior) {
1189         case SECURITY_FS_USE_XATTR:
1190                 if (!inode->i_op->getxattr) {
1191                         isec->sid = sbsec->def_sid;
1192                         break;
1193                 }
1194
1195                 /* Need a dentry, since the xattr API requires one.
1196                    Life would be simpler if we could just pass the inode. */
1197                 if (opt_dentry) {
1198                         /* Called from d_instantiate or d_splice_alias. */
1199                         dentry = dget(opt_dentry);
1200                 } else {
1201                         /* Called from selinux_complete_init, try to find a dentry. */
1202                         dentry = d_find_alias(inode);
1203                 }
1204                 if (!dentry) {
1205                         printk(KERN_WARNING "SELinux: %s:  no dentry for dev=%s "
1206                                "ino=%ld\n", __func__, inode->i_sb->s_id,
1207                                inode->i_ino);
1208                         goto out_unlock;
1209                 }
1210
1211                 len = INITCONTEXTLEN;
1212                 context = kmalloc(len, GFP_NOFS);
1213                 if (!context) {
1214                         rc = -ENOMEM;
1215                         dput(dentry);
1216                         goto out_unlock;
1217                 }
1218                 rc = inode->i_op->getxattr(dentry, XATTR_NAME_SELINUX,
1219                                            context, len);
1220                 if (rc == -ERANGE) {
1221                         /* Need a larger buffer.  Query for the right size. */
1222                         rc = inode->i_op->getxattr(dentry, XATTR_NAME_SELINUX,
1223                                                    NULL, 0);
1224                         if (rc < 0) {
1225                                 dput(dentry);
1226                                 goto out_unlock;
1227                         }
1228                         kfree(context);
1229                         len = rc;
1230                         context = kmalloc(len, GFP_NOFS);
1231                         if (!context) {
1232                                 rc = -ENOMEM;
1233                                 dput(dentry);
1234                                 goto out_unlock;
1235                         }
1236                         rc = inode->i_op->getxattr(dentry,
1237                                                    XATTR_NAME_SELINUX,
1238                                                    context, len);
1239                 }
1240                 dput(dentry);
1241                 if (rc < 0) {
1242                         if (rc != -ENODATA) {
1243                                 printk(KERN_WARNING "SELinux: %s:  getxattr returned "
1244                                        "%d for dev=%s ino=%ld\n", __func__,
1245                                        -rc, inode->i_sb->s_id, inode->i_ino);
1246                                 kfree(context);
1247                                 goto out_unlock;
1248                         }
1249                         /* Map ENODATA to the default file SID */
1250                         sid = sbsec->def_sid;
1251                         rc = 0;
1252                 } else {
1253                         rc = security_context_to_sid_default(context, rc, &sid,
1254                                                              sbsec->def_sid,
1255                                                              GFP_NOFS);
1256                         if (rc) {
1257                                 printk(KERN_WARNING "SELinux: %s:  context_to_sid(%s) "
1258                                        "returned %d for dev=%s ino=%ld\n",
1259                                        __func__, context, -rc,
1260                                        inode->i_sb->s_id, inode->i_ino);
1261                                 kfree(context);
1262                                 /* Leave with the unlabeled SID */
1263                                 rc = 0;
1264                                 break;
1265                         }
1266                 }
1267                 kfree(context);
1268                 isec->sid = sid;
1269                 break;
1270         case SECURITY_FS_USE_TASK:
1271                 isec->sid = isec->task_sid;
1272                 break;
1273         case SECURITY_FS_USE_TRANS:
1274                 /* Default to the fs SID. */
1275                 isec->sid = sbsec->sid;
1276
1277                 /* Try to obtain a transition SID. */
1278                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
1279                 rc = security_transition_sid(isec->task_sid,
1280                                              sbsec->sid,
1281                                              isec->sclass,
1282                                              &sid);
1283                 if (rc)
1284                         goto out_unlock;
1285                 isec->sid = sid;
1286                 break;
1287         case SECURITY_FS_USE_MNTPOINT:
1288                 isec->sid = sbsec->mntpoint_sid;
1289                 break;
1290         default:
1291                 /* Default to the fs superblock SID. */
1292                 isec->sid = sbsec->sid;
1293
1294                 if (sbsec->proc && !S_ISLNK(inode->i_mode)) {
1295                         struct proc_inode *proci = PROC_I(inode);
1296                         if (proci->pde) {
1297                                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
1298                                 rc = selinux_proc_get_sid(proci->pde,
1299                                                           isec->sclass,
1300                                                           &sid);
1301                                 if (rc)
1302                                         goto out_unlock;
1303                                 isec->sid = sid;
1304                         }
1305                 }
1306                 break;
1307         }
1308
1309         isec->initialized = 1;
1310
1311 out_unlock:
1312         mutex_unlock(&isec->lock);
1313 out:
1314         if (isec->sclass == SECCLASS_FILE)
1315                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
1316         return rc;
1317 }
1318
1319 /* Convert a Linux signal to an access vector. */
1320 static inline u32 signal_to_av(int sig)
1321 {
1322         u32 perm = 0;
1323
1324         switch (sig) {
1325         case SIGCHLD:
1326                 /* Commonly granted from child to parent. */
1327                 perm = PROCESS__SIGCHLD;
1328                 break;
1329         case SIGKILL:
1330                 /* Cannot be caught or ignored */
1331                 perm = PROCESS__SIGKILL;
1332                 break;
1333         case SIGSTOP:
1334                 /* Cannot be caught or ignored */
1335                 perm = PROCESS__SIGSTOP;
1336                 break;
1337         default:
1338                 /* All other signals. */
1339                 perm = PROCESS__SIGNAL;
1340                 break;
1341         }
1342
1343         return perm;
1344 }
1345
1346 /* Check permission betweeen a pair of tasks, e.g. signal checks,
1347    fork check, ptrace check, etc. */
1348 static int task_has_perm(struct task_struct *tsk1,
1349                          struct task_struct *tsk2,
1350                          u32 perms)
1351 {
1352         struct task_security_struct *tsec1, *tsec2;
1353
1354         tsec1 = tsk1->security;
1355         tsec2 = tsk2->security;
1356         return avc_has_perm(tsec1->sid, tsec2->sid,
1357                             SECCLASS_PROCESS, perms, NULL);
1358 }
1359
1360 #if CAP_LAST_CAP > 63
1361 #error Fix SELinux to handle capabilities > 63.
1362 #endif
1363
1364 /* Check whether a task is allowed to use a capability. */
1365 static int task_has_capability(struct task_struct *tsk,
1366                                int cap)
1367 {
1368         struct task_security_struct *tsec;
1369         struct avc_audit_data ad;
1370         u16 sclass;
1371         u32 av = CAP_TO_MASK(cap);
1372
1373         tsec = tsk->security;
1374
1375         AVC_AUDIT_DATA_INIT(&ad, CAP);
1376         ad.tsk = tsk;
1377         ad.u.cap = cap;
1378
1379         switch (CAP_TO_INDEX(cap)) {
1380         case 0:
1381                 sclass = SECCLASS_CAPABILITY;
1382                 break;
1383         case 1:
1384                 sclass = SECCLASS_CAPABILITY2;
1385                 break;
1386         default:
1387                 printk(KERN_ERR
1388                        "SELinux:  out of range capability %d\n", cap);
1389                 BUG();
1390         }
1391         return avc_has_perm(tsec->sid, tsec->sid, sclass, av, &ad);
1392 }
1393
1394 /* Check whether a task is allowed to use a system operation. */
1395 static int task_has_system(struct task_struct *tsk,
1396                            u32 perms)
1397 {
1398         struct task_security_struct *tsec;
1399
1400         tsec = tsk->security;
1401
1402         return avc_has_perm(tsec->sid, SECINITSID_KERNEL,
1403                             SECCLASS_SYSTEM, perms, NULL);
1404 }
1405
1406 /* Check whether a task has a particular permission to an inode.
1407    The 'adp' parameter is optional and allows other audit
1408    data to be passed (e.g. the dentry). */
1409 static int inode_has_perm(struct task_struct *tsk,
1410                           struct inode *inode,
1411                           u32 perms,
1412                           struct avc_audit_data *adp)
1413 {
1414         struct task_security_struct *tsec;
1415         struct inode_security_struct *isec;
1416         struct avc_audit_data ad;
1417
1418         if (unlikely(IS_PRIVATE(inode)))
1419                 return 0;
1420
1421         tsec = tsk->security;
1422         isec = inode->i_security;
1423
1424         if (!adp) {
1425                 adp = &ad;
1426                 AVC_AUDIT_DATA_INIT(&ad, FS);
1427                 ad.u.fs.inode = inode;
1428         }
1429
1430         return avc_has_perm(tsec->sid, isec->sid, isec->sclass, perms, adp);
1431 }
1432
1433 /* Same as inode_has_perm, but pass explicit audit data containing
1434    the dentry to help the auditing code to more easily generate the
1435    pathname if needed. */
1436 static inline int dentry_has_perm(struct task_struct *tsk,
1437                                   struct vfsmount *mnt,
1438                                   struct dentry *dentry,
1439                                   u32 av)
1440 {
1441         struct inode *inode = dentry->d_inode;
1442         struct avc_audit_data ad;
1443         AVC_AUDIT_DATA_INIT(&ad, FS);
1444         ad.u.fs.path.mnt = mnt;
1445         ad.u.fs.path.dentry = dentry;
1446         return inode_has_perm(tsk, inode, av, &ad);
1447 }
1448
1449 /* Check whether a task can use an open file descriptor to
1450    access an inode in a given way.  Check access to the
1451    descriptor itself, and then use dentry_has_perm to
1452    check a particular permission to the file.
1453    Access to the descriptor is implicitly granted if it
1454    has the same SID as the process.  If av is zero, then
1455    access to the file is not checked, e.g. for cases
1456    where only the descriptor is affected like seek. */
1457 static int file_has_perm(struct task_struct *tsk,
1458                                 struct file *file,
1459                                 u32 av)
1460 {
1461         struct task_security_struct *tsec = tsk->security;
1462         struct file_security_struct *fsec = file->f_security;
1463         struct inode *inode = file->f_path.dentry->d_inode;
1464         struct avc_audit_data ad;
1465         int rc;
1466
1467         AVC_AUDIT_DATA_INIT(&ad, FS);
1468         ad.u.fs.path = file->f_path;
1469
1470         if (tsec->sid != fsec->sid) {
1471                 rc = avc_has_perm(tsec->sid, fsec->sid,
1472                                   SECCLASS_FD,
1473                                   FD__USE,
1474                                   &ad);
1475                 if (rc)
1476                         return rc;
1477         }
1478
1479         /* av is zero if only checking access to the descriptor. */
1480         if (av)
1481                 return inode_has_perm(tsk, inode, av, &ad);
1482
1483         return 0;
1484 }
1485
1486 /* Check whether a task can create a file. */
1487 static int may_create(struct inode *dir,
1488                       struct dentry *dentry,
1489                       u16 tclass)
1490 {
1491         struct task_security_struct *tsec;
1492         struct inode_security_struct *dsec;
1493         struct superblock_security_struct *sbsec;
1494         u32 newsid;
1495         struct avc_audit_data ad;
1496         int rc;
1497
1498         tsec = current->security;
1499         dsec = dir->i_security;
1500         sbsec = dir->i_sb->s_security;
1501
1502         AVC_AUDIT_DATA_INIT(&ad, FS);
1503         ad.u.fs.path.dentry = dentry;
1504
1505         rc = avc_has_perm(tsec->sid, dsec->sid, SECCLASS_DIR,
1506                           DIR__ADD_NAME | DIR__SEARCH,
1507                           &ad);
1508         if (rc)
1509                 return rc;
1510
1511         if (tsec->create_sid && sbsec->behavior != SECURITY_FS_USE_MNTPOINT) {
1512                 newsid = tsec->create_sid;
1513         } else {
1514                 rc = security_transition_sid(tsec->sid, dsec->sid, tclass,
1515                                              &newsid);
1516                 if (rc)
1517                         return rc;
1518         }
1519
1520         rc = avc_has_perm(tsec->sid, newsid, tclass, FILE__CREATE, &ad);
1521         if (rc)
1522                 return rc;
1523
1524         return avc_has_perm(newsid, sbsec->sid,
1525                             SECCLASS_FILESYSTEM,
1526                             FILESYSTEM__ASSOCIATE, &ad);
1527 }
1528
1529 /* Check whether a task can create a key. */
1530 static int may_create_key(u32 ksid,
1531                           struct task_struct *ctx)
1532 {
1533         struct task_security_struct *tsec;
1534
1535         tsec = ctx->security;
1536
1537         return avc_has_perm(tsec->sid, ksid, SECCLASS_KEY, KEY__CREATE, NULL);
1538 }
1539
1540 #define MAY_LINK        0
1541 #define MAY_UNLINK      1
1542 #define MAY_RMDIR       2
1543
1544 /* Check whether a task can link, unlink, or rmdir a file/directory. */
1545 static int may_link(struct inode *dir,
1546                     struct dentry *dentry,
1547                     int kind)
1548
1549 {
1550         struct task_security_struct *tsec;
1551         struct inode_security_struct *dsec, *isec;
1552         struct avc_audit_data ad;
1553         u32 av;
1554         int rc;
1555
1556         tsec = current->security;
1557         dsec = dir->i_security;
1558         isec = dentry->d_inode->i_security;
1559
1560         AVC_AUDIT_DATA_INIT(&ad, FS);
1561         ad.u.fs.path.dentry = dentry;
1562
1563         av = DIR__SEARCH;
1564         av |= (kind ? DIR__REMOVE_NAME : DIR__ADD_NAME);
1565         rc = avc_has_perm(tsec->sid, dsec->sid, SECCLASS_DIR, av, &ad);
1566         if (rc)
1567                 return rc;
1568
1569         switch (kind) {
1570         case MAY_LINK:
1571                 av = FILE__LINK;
1572                 break;
1573         case MAY_UNLINK:
1574                 av = FILE__UNLINK;
1575                 break;
1576         case MAY_RMDIR:
1577                 av = DIR__RMDIR;
1578                 break;
1579         default:
1580                 printk(KERN_WARNING "SELinux: %s:  unrecognized kind %d\n",
1581                         __func__, kind);
1582                 return 0;
1583         }
1584
1585         rc = avc_has_perm(tsec->sid, isec->sid, isec->sclass, av, &ad);
1586         return rc;
1587 }
1588
1589 static inline int may_rename(struct inode *old_dir,
1590                              struct dentry *old_dentry,
1591                              struct inode *new_dir,
1592                              struct dentry *new_dentry)
1593 {
1594         struct task_security_struct *tsec;
1595         struct inode_security_struct *old_dsec, *new_dsec, *old_isec, *new_isec;
1596         struct avc_audit_data ad;
1597         u32 av;
1598         int old_is_dir, new_is_dir;
1599         int rc;
1600
1601         tsec = current->security;
1602         old_dsec = old_dir->i_security;
1603         old_isec = old_dentry->d_inode->i_security;
1604         old_is_dir = S_ISDIR(old_dentry->d_inode->i_mode);
1605         new_dsec = new_dir->i_security;
1606
1607         AVC_AUDIT_DATA_INIT(&ad, FS);
1608
1609         ad.u.fs.path.dentry = old_dentry;
1610         rc = avc_has_perm(tsec->sid, old_dsec->sid, SECCLASS_DIR,
1611                           DIR__REMOVE_NAME | DIR__SEARCH, &ad);
1612         if (rc)
1613                 return rc;
1614         rc = avc_has_perm(tsec->sid, old_isec->sid,
1615                           old_isec->sclass, FILE__RENAME, &ad);
1616         if (rc)
1617                 return rc;
1618         if (old_is_dir && new_dir != old_dir) {
1619                 rc = avc_has_perm(tsec->sid, old_isec->sid,
1620                                   old_isec->sclass, DIR__REPARENT, &ad);
1621                 if (rc)
1622                         return rc;
1623         }
1624
1625         ad.u.fs.path.dentry = new_dentry;
1626         av = DIR__ADD_NAME | DIR__SEARCH;
1627         if (new_dentry->d_inode)
1628                 av |= DIR__REMOVE_NAME;
1629         rc = avc_has_perm(tsec->sid, new_dsec->sid, SECCLASS_DIR, av, &ad);
1630         if (rc)
1631                 return rc;
1632         if (new_dentry->d_inode) {
1633                 new_isec = new_dentry->d_inode->i_security;
1634                 new_is_dir = S_ISDIR(new_dentry->d_inode->i_mode);
1635                 rc = avc_has_perm(tsec->sid, new_isec->sid,
1636                                   new_isec->sclass,
1637                                   (new_is_dir ? DIR__RMDIR : FILE__UNLINK), &ad);
1638                 if (rc)
1639                         return rc;
1640         }
1641
1642         return 0;
1643 }
1644
1645 /* Check whether a task can perform a filesystem operation. */
1646 static int superblock_has_perm(struct task_struct *tsk,
1647                                struct super_block *sb,
1648                                u32 perms,
1649                                struct avc_audit_data *ad)
1650 {
1651         struct task_security_struct *tsec;
1652         struct superblock_security_struct *sbsec;
1653
1654         tsec = tsk->security;
1655         sbsec = sb->s_security;
1656         return avc_has_perm(tsec->sid, sbsec->sid, SECCLASS_FILESYSTEM,
1657                             perms, ad);
1658 }
1659
1660 /* Convert a Linux mode and permission mask to an access vector. */
1661 static inline u32 file_mask_to_av(int mode, int mask)
1662 {
1663         u32 av = 0;
1664
1665         if ((mode & S_IFMT) != S_IFDIR) {
1666                 if (mask & MAY_EXEC)
1667                         av |= FILE__EXECUTE;
1668                 if (mask & MAY_READ)
1669                         av |= FILE__READ;
1670
1671                 if (mask & MAY_APPEND)
1672                         av |= FILE__APPEND;
1673                 else if (mask & MAY_WRITE)
1674                         av |= FILE__WRITE;
1675
1676         } else {
1677                 if (mask & MAY_EXEC)
1678                         av |= DIR__SEARCH;
1679                 if (mask & MAY_WRITE)
1680                         av |= DIR__WRITE;
1681                 if (mask & MAY_READ)
1682                         av |= DIR__READ;
1683         }
1684
1685         return av;
1686 }
1687
1688 /*
1689  * Convert a file mask to an access vector and include the correct open
1690  * open permission.
1691  */
1692 static inline u32 open_file_mask_to_av(int mode, int mask)
1693 {
1694         u32 av = file_mask_to_av(mode, mask);
1695
1696         if (selinux_policycap_openperm) {
1697                 /*
1698                  * lnk files and socks do not really have an 'open'
1699                  */
1700                 if (S_ISREG(mode))
1701                         av |= FILE__OPEN;
1702                 else if (S_ISCHR(mode))
1703                         av |= CHR_FILE__OPEN;
1704                 else if (S_ISBLK(mode))
1705                         av |= BLK_FILE__OPEN;
1706                 else if (S_ISFIFO(mode))
1707                         av |= FIFO_FILE__OPEN;
1708                 else if (S_ISDIR(mode))
1709                         av |= DIR__OPEN;
1710                 else
1711                         printk(KERN_ERR "SELinux: WARNING: inside %s with "
1712                                 "unknown mode:%x\n", __func__, mode);
1713         }
1714         return av;
1715 }
1716
1717 /* Convert a Linux file to an access vector. */
1718 static inline u32 file_to_av(struct file *file)
1719 {
1720         u32 av = 0;
1721
1722         if (file->f_mode & FMODE_READ)
1723                 av |= FILE__READ;
1724         if (file->f_mode & FMODE_WRITE) {
1725                 if (file->f_flags & O_APPEND)
1726                         av |= FILE__APPEND;
1727                 else
1728                         av |= FILE__WRITE;
1729         }
1730         if (!av) {
1731                 /*
1732                  * Special file opened with flags 3 for ioctl-only use.
1733                  */
1734                 av = FILE__IOCTL;
1735         }
1736
1737         return av;
1738 }
1739
1740 /* Hook functions begin here. */
1741
1742 static int selinux_ptrace_may_access(struct task_struct *child,
1743                                      unsigned int mode)
1744 {
1745         int rc;
1746
1747         rc = secondary_ops->ptrace_may_access(child, mode);
1748         if (rc)
1749                 return rc;
1750
1751         if (mode == PTRACE_MODE_READ) {
1752                 struct task_security_struct *tsec = current->security;
1753                 struct task_security_struct *csec = child->security;
1754                 return avc_has_perm(tsec->sid, csec->sid,
1755                                     SECCLASS_FILE, FILE__READ, NULL);
1756         }
1757
1758         return task_has_perm(current, child, PROCESS__PTRACE);
1759 }
1760
1761 static int selinux_ptrace_traceme(struct task_struct *parent)
1762 {
1763         int rc;
1764
1765         rc = secondary_ops->ptrace_traceme(parent);
1766         if (rc)
1767                 return rc;
1768
1769         return task_has_perm(parent, current, PROCESS__PTRACE);
1770 }
1771
1772 static int selinux_capget(struct task_struct *target, kernel_cap_t *effective,
1773                           kernel_cap_t *inheritable, kernel_cap_t *permitted)
1774 {
1775         int error;
1776
1777         error = task_has_perm(current, target, PROCESS__GETCAP);
1778         if (error)
1779                 return error;
1780
1781         return secondary_ops->capget(target, effective, inheritable, permitted);
1782 }
1783
1784 static int selinux_capset_check(struct task_struct *target, kernel_cap_t *effective,
1785                                 kernel_cap_t *inheritable, kernel_cap_t *permitted)
1786 {
1787         int error;
1788
1789         error = secondary_ops->capset_check(target, effective, inheritable, permitted);
1790         if (error)
1791                 return error;
1792
1793         return task_has_perm(current, target, PROCESS__SETCAP);
1794 }
1795
1796 static void selinux_capset_set(struct task_struct *target, kernel_cap_t *effective,
1797                                kernel_cap_t *inheritable, kernel_cap_t *permitted)
1798 {
1799         secondary_ops->capset_set(target, effective, inheritable, permitted);
1800 }
1801
1802 static int selinux_capable(struct task_struct *tsk, int cap)
1803 {
1804         int rc;
1805
1806         rc = secondary_ops->capable(tsk, cap);
1807         if (rc)
1808                 return rc;
1809
1810         return task_has_capability(tsk, cap);
1811 }
1812
1813 static int selinux_sysctl_get_sid(ctl_table *table, u16 tclass, u32 *sid)
1814 {
1815         int buflen, rc;
1816         char *buffer, *path, *end;
1817
1818         rc = -ENOMEM;
1819         buffer = (char *)__get_free_page(GFP_KERNEL);
1820         if (!buffer)
1821                 goto out;
1822
1823         buflen = PAGE_SIZE;
1824         end = buffer+buflen;
1825         *--end = '\0';
1826         buflen--;
1827         path = end-1;
1828         *path = '/';
1829         while (table) {
1830                 const char *name = table->procname;
1831                 size_t namelen = strlen(name);
1832                 buflen -= namelen + 1;
1833                 if (buflen < 0)
1834                         goto out_free;
1835                 end -= namelen;
1836                 memcpy(end, name, namelen);
1837                 *--end = '/';
1838                 path = end;
1839                 table = table->parent;
1840         }
1841         buflen -= 4;
1842         if (buflen < 0)
1843                 goto out_free;
1844         end -= 4;
1845         memcpy(end, "/sys", 4);
1846         path = end;
1847         rc = security_genfs_sid("proc", path, tclass, sid);
1848 out_free:
1849         free_page((unsigned long)buffer);
1850 out:
1851         return rc;
1852 }
1853
1854 static int selinux_sysctl(ctl_table *table, int op)
1855 {
1856         int error = 0;
1857         u32 av;
1858         struct task_security_struct *tsec;
1859         u32 tsid;
1860         int rc;
1861
1862         rc = secondary_ops->sysctl(table, op);
1863         if (rc)
1864                 return rc;
1865
1866         tsec = current->security;
1867
1868         rc = selinux_sysctl_get_sid(table, (op == 0001) ?
1869                                     SECCLASS_DIR : SECCLASS_FILE, &tsid);
1870         if (rc) {
1871                 /* Default to the well-defined sysctl SID. */
1872                 tsid = SECINITSID_SYSCTL;
1873         }
1874
1875         /* The op values are "defined" in sysctl.c, thereby creating
1876          * a bad coupling between this module and sysctl.c */
1877         if (op == 001) {
1878                 error = avc_has_perm(tsec->sid, tsid,
1879                                      SECCLASS_DIR, DIR__SEARCH, NULL);
1880         } else {
1881                 av = 0;
1882                 if (op & 004)
1883                         av |= FILE__READ;
1884                 if (op & 002)
1885                         av |= FILE__WRITE;
1886                 if (av)
1887                         error = avc_has_perm(tsec->sid, tsid,
1888                                              SECCLASS_FILE, av, NULL);
1889         }
1890
1891         return error;
1892 }
1893
1894 static int selinux_quotactl(int cmds, int type, int id, struct super_block *sb)
1895 {
1896         int rc = 0;
1897
1898         if (!sb)
1899                 return 0;
1900
1901         switch (cmds) {
1902         case Q_SYNC:
1903         case Q_QUOTAON:
1904         case Q_QUOTAOFF:
1905         case Q_SETINFO:
1906         case Q_SETQUOTA:
1907                 rc = superblock_has_perm(current, sb, FILESYSTEM__QUOTAMOD,
1908                                          NULL);
1909                 break;
1910         case Q_GETFMT:
1911         case Q_GETINFO:
1912         case Q_GETQUOTA:
1913                 rc = superblock_has_perm(current, sb, FILESYSTEM__QUOTAGET,
1914                                          NULL);
1915                 break;
1916         default:
1917                 rc = 0;  /* let the kernel handle invalid cmds */
1918                 break;
1919         }
1920         return rc;
1921 }
1922
1923 static int selinux_quota_on(struct dentry *dentry)
1924 {
1925         return dentry_has_perm(current, NULL, dentry, FILE__QUOTAON);
1926 }
1927
1928 static int selinux_syslog(int type)
1929 {
1930         int rc;
1931
1932         rc = secondary_ops->syslog(type);
1933         if (rc)
1934                 return rc;
1935
1936         switch (type) {
1937         case 3:         /* Read last kernel messages */
1938         case 10:        /* Return size of the log buffer */
1939                 rc = task_has_system(current, SYSTEM__SYSLOG_READ);
1940                 break;
1941         case 6:         /* Disable logging to console */
1942         case 7:         /* Enable logging to console */
1943         case 8:         /* Set level of messages printed to console */
1944                 rc = task_has_system(current, SYSTEM__SYSLOG_CONSOLE);
1945                 break;
1946         case 0:         /* Close log */
1947         case 1:         /* Open log */
1948         case 2:         /* Read from log */
1949         case 4:         /* Read/clear last kernel messages */
1950         case 5:         /* Clear ring buffer */
1951         default:
1952                 rc = task_has_system(current, SYSTEM__SYSLOG_MOD);
1953                 break;
1954         }
1955         return rc;
1956 }
1957
1958 /*
1959  * Check that a process has enough memory to allocate a new virtual
1960  * mapping. 0 means there is enough memory for the allocation to
1961  * succeed and -ENOMEM implies there is not.
1962  *
1963  * Note that secondary_ops->capable and task_has_perm_noaudit return 0
1964  * if the capability is granted, but __vm_enough_memory requires 1 if
1965  * the capability is granted.
1966  *
1967  * Do not audit the selinux permission check, as this is applied to all
1968  * processes that allocate mappings.
1969  */
1970 static int selinux_vm_enough_memory(struct mm_struct *mm, long pages)
1971 {
1972         int rc, cap_sys_admin = 0;
1973         struct task_security_struct *tsec = current->security;
1974
1975         rc = secondary_ops->capable(current, CAP_SYS_ADMIN);
1976         if (rc == 0)
1977                 rc = avc_has_perm_noaudit(tsec->sid, tsec->sid,
1978                                           SECCLASS_CAPABILITY,
1979                                           CAP_TO_MASK(CAP_SYS_ADMIN),
1980                                           0,
1981                                           NULL);
1982
1983         if (rc == 0)
1984                 cap_sys_admin = 1;
1985
1986         return __vm_enough_memory(mm, pages, cap_sys_admin);
1987 }
1988
1989 /* binprm security operations */
1990
1991 static int selinux_bprm_alloc_security(struct linux_binprm *bprm)
1992 {
1993         struct bprm_security_struct *bsec;
1994
1995         bsec = kzalloc(sizeof(struct bprm_security_struct), GFP_KERNEL);
1996         if (!bsec)
1997                 return -ENOMEM;
1998
1999         bsec->sid = SECINITSID_UNLABELED;
2000         bsec->set = 0;
2001
2002         bprm->security = bsec;
2003         return 0;
2004 }
2005
2006 static int selinux_bprm_set_security(struct linux_binprm *bprm)
2007 {
2008         struct task_security_struct *tsec;
2009         struct inode *inode = bprm->file->f_path.dentry->d_inode;
2010         struct inode_security_struct *isec;
2011         struct bprm_security_struct *bsec;
2012         u32 newsid;
2013         struct avc_audit_data ad;
2014         int rc;
2015
2016         rc = secondary_ops->bprm_set_security(bprm);
2017         if (rc)
2018                 return rc;
2019
2020         bsec = bprm->security;
2021
2022         if (bsec->set)
2023                 return 0;
2024
2025         tsec = current->security;
2026         isec = inode->i_security;
2027
2028         /* Default to the current task SID. */
2029         bsec->sid = tsec->sid;
2030
2031         /* Reset fs, key, and sock SIDs on execve. */
2032         tsec->create_sid = 0;
2033         tsec->keycreate_sid = 0;
2034         tsec->sockcreate_sid = 0;
2035
2036         if (tsec->exec_sid) {
2037                 newsid = tsec->exec_sid;
2038                 /* Reset exec SID on execve. */
2039                 tsec->exec_sid = 0;
2040         } else {
2041                 /* Check for a default transition on this program. */
2042                 rc = security_transition_sid(tsec->sid, isec->sid,
2043                                              SECCLASS_PROCESS, &newsid);
2044                 if (rc)
2045                         return rc;
2046         }
2047
2048         AVC_AUDIT_DATA_INIT(&ad, FS);
2049         ad.u.fs.path = bprm->file->f_path;
2050
2051         if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
2052                 newsid = tsec->sid;
2053
2054         if (tsec->sid == newsid) {
2055                 rc = avc_has_perm(tsec->sid, isec->sid,
2056                                   SECCLASS_FILE, FILE__EXECUTE_NO_TRANS, &ad);
2057                 if (rc)
2058                         return rc;
2059         } else {
2060                 /* Check permissions for the transition. */
2061                 rc = avc_has_perm(tsec->sid, newsid,
2062                                   SECCLASS_PROCESS, PROCESS__TRANSITION, &ad);
2063                 if (rc)
2064                         return rc;
2065
2066                 rc = avc_has_perm(newsid, isec->sid,
2067                                   SECCLASS_FILE, FILE__ENTRYPOINT, &ad);
2068                 if (rc)
2069                         return rc;
2070
2071                 /* Clear any possibly unsafe personality bits on exec: */
2072                 current->personality &= ~PER_CLEAR_ON_SETID;
2073
2074                 /* Set the security field to the new SID. */
2075                 bsec->sid = newsid;
2076         }
2077
2078         bsec->set = 1;
2079         return 0;
2080 }
2081
2082 static int selinux_bprm_check_security(struct linux_binprm *bprm)
2083 {
2084         return secondary_ops->bprm_check_security(bprm);
2085 }
2086
2087
2088 static int selinux_bprm_secureexec(struct linux_binprm *bprm)
2089 {
2090         struct task_security_struct *tsec = current->security;
2091         int atsecure = 0;
2092
2093         if (tsec->osid != tsec->sid) {
2094                 /* Enable secure mode for SIDs transitions unless
2095                    the noatsecure permission is granted between
2096                    the two SIDs, i.e. ahp returns 0. */
2097                 atsecure = avc_has_perm(tsec->osid, tsec->sid,
2098                                          SECCLASS_PROCESS,
2099                                          PROCESS__NOATSECURE, NULL);
2100         }
2101
2102         return (atsecure || secondary_ops->bprm_secureexec(bprm));
2103 }
2104
2105 static void selinux_bprm_free_security(struct linux_binprm *bprm)
2106 {
2107         kfree(bprm->security);
2108         bprm->security = NULL;
2109 }
2110
2111 extern struct vfsmount *selinuxfs_mount;
2112 extern struct dentry *selinux_null;
2113
2114 /* Derived from fs/exec.c:flush_old_files. */
2115 static inline void flush_unauthorized_files(struct files_struct *files)
2116 {
2117         struct avc_audit_data ad;
2118         struct file *file, *devnull = NULL;
2119         struct tty_struct *tty;
2120         struct fdtable *fdt;
2121         long j = -1;
2122         int drop_tty = 0;
2123
2124         mutex_lock(&tty_mutex);
2125         tty = get_current_tty();
2126         mutex_unlock(&tty_mutex);
2127         if (tty) {
2128                 file_list_lock();
2129                 file = list_entry(tty->tty_files.next, typeof(*file), f_u.fu_list);
2130                 if (file) {
2131                         /* Revalidate access to controlling tty.
2132                            Use inode_has_perm on the tty inode directly rather
2133                            than using file_has_perm, as this particular open
2134                            file may belong to another process and we are only
2135                            interested in the inode-based check here. */
2136                         struct inode *inode = file->f_path.dentry->d_inode;
2137                         if (inode_has_perm(current, inode,
2138                                            FILE__READ | FILE__WRITE, NULL)) {
2139                                 drop_tty = 1;
2140                         }
2141                 }
2142                 file_list_unlock();
2143                 tty_kref_put(tty);
2144         }
2145         /* Reset controlling tty. */
2146         if (drop_tty)
2147                 no_tty();
2148
2149         /* Revalidate access to inherited open files. */
2150
2151         AVC_AUDIT_DATA_INIT(&ad, FS);
2152
2153         spin_lock(&files->file_lock);
2154         for (;;) {
2155                 unsigned long set, i;
2156                 int fd;
2157
2158                 j++;
2159                 i = j * __NFDBITS;
2160                 fdt = files_fdtable(files);
2161                 if (i >= fdt->max_fds)
2162                         break;
2163                 set = fdt->open_fds->fds_bits[j];
2164                 if (!set)
2165                         continue;
2166                 spin_unlock(&files->file_lock);
2167                 for ( ; set ; i++, set >>= 1) {
2168                         if (set & 1) {
2169                                 file = fget(i);
2170                                 if (!file)
2171                                         continue;
2172                                 if (file_has_perm(current,
2173                                                   file,
2174                                                   file_to_av(file))) {
2175                                         sys_close(i);
2176                                         fd = get_unused_fd();
2177                                         if (fd != i) {
2178                                                 if (fd >= 0)
2179                                                         put_unused_fd(fd);
2180                                                 fput(file);
2181                                                 continue;
2182                                         }
2183                                         if (devnull) {
2184                                                 get_file(devnull);
2185                                         } else {
2186                                                 devnull = dentry_open(dget(selinux_null), mntget(selinuxfs_mount), O_RDWR);
2187                                                 if (IS_ERR(devnull)) {
2188                                                         devnull = NULL;
2189                                                         put_unused_fd(fd);
2190                                                         fput(file);
2191                                                         continue;
2192                                                 }
2193                                         }
2194                                         fd_install(fd, devnull);
2195                                 }
2196                                 fput(file);
2197                         }
2198                 }
2199                 spin_lock(&files->file_lock);
2200
2201         }
2202         spin_unlock(&files->file_lock);
2203 }
2204
2205 static void selinux_bprm_apply_creds(struct linux_binprm *bprm, int unsafe)
2206 {
2207         struct task_security_struct *tsec;
2208         struct bprm_security_struct *bsec;
2209         u32 sid;
2210         int rc;
2211
2212         secondary_ops->bprm_apply_creds(bprm, unsafe);
2213
2214         tsec = current->security;
2215
2216         bsec = bprm->security;
2217         sid = bsec->sid;
2218
2219         tsec->osid = tsec->sid;
2220         bsec->unsafe = 0;
2221         if (tsec->sid != sid) {
2222                 /* Check for shared state.  If not ok, leave SID
2223                    unchanged and kill. */
2224                 if (unsafe & LSM_UNSAFE_SHARE) {
2225                         rc = avc_has_perm(tsec->sid, sid, SECCLASS_PROCESS,
2226                                         PROCESS__SHARE, NULL);
2227                         if (rc) {
2228                                 bsec->unsafe = 1;
2229                                 return;
2230                         }
2231                 }
2232
2233                 /* Check for ptracing, and update the task SID if ok.
2234                    Otherwise, leave SID unchanged and kill. */
2235                 if (unsafe & (LSM_UNSAFE_PTRACE | LSM_UNSAFE_PTRACE_CAP)) {
2236                         struct task_struct *tracer;
2237                         struct task_security_struct *sec;
2238                         u32 ptsid = 0;
2239
2240                         rcu_read_lock();
2241                         tracer = tracehook_tracer_task(current);
2242                         if (likely(tracer != NULL)) {
2243                                 sec = tracer->security;
2244                                 ptsid = sec->sid;
2245                         }
2246                         rcu_read_unlock();
2247
2248                         if (ptsid != 0) {
2249                                 rc = avc_has_perm(ptsid, sid, SECCLASS_PROCESS,
2250                                                   PROCESS__PTRACE, NULL);
2251                                 if (rc) {
2252                                         bsec->unsafe = 1;
2253                                         return;
2254                                 }
2255                         }
2256                 }
2257                 tsec->sid = sid;
2258         }
2259 }
2260
2261 /*
2262  * called after apply_creds without the task lock held
2263  */
2264 static void selinux_bprm_post_apply_creds(struct linux_binprm *bprm)
2265 {
2266         struct task_security_struct *tsec;
2267         struct rlimit *rlim, *initrlim;
2268         struct itimerval itimer;
2269         struct bprm_security_struct *bsec;
2270         int rc, i;
2271
2272         tsec = current->security;
2273         bsec = bprm->security;
2274
2275         if (bsec->unsafe) {
2276                 force_sig_specific(SIGKILL, current);
2277                 return;
2278         }
2279         if (tsec->osid == tsec->sid)
2280                 return;
2281
2282         /* Close files for which the new task SID is not authorized. */
2283         flush_unauthorized_files(current->files);
2284
2285         /* Check whether the new SID can inherit signal state
2286            from the old SID.  If not, clear itimers to avoid
2287            subsequent signal generation and flush and unblock
2288            signals. This must occur _after_ the task SID has
2289           been updated so that any kill done after the flush
2290           will be checked against the new SID. */
2291         rc = avc_has_perm(tsec->osid, tsec->sid, SECCLASS_PROCESS,
2292                           PROCESS__SIGINH, NULL);
2293         if (rc) {
2294                 memset(&itimer, 0, sizeof itimer);
2295                 for (i = 0; i < 3; i++)
2296                         do_setitimer(i, &itimer, NULL);
2297                 flush_signals(current);
2298                 spin_lock_irq(&current->sighand->siglock);
2299                 flush_signal_handlers(current, 1);
2300                 sigemptyset(&current->blocked);
2301                 recalc_sigpending();
2302                 spin_unlock_irq(&current->sighand->siglock);
2303         }
2304
2305         /* Always clear parent death signal on SID transitions. */
2306         current->pdeath_signal = 0;
2307
2308         /* Check whether the new SID can inherit resource limits
2309            from the old SID.  If not, reset all soft limits to
2310            the lower of the current task's hard limit and the init
2311            task's soft limit.  Note that the setting of hard limits
2312            (even to lower them) can be controlled by the setrlimit
2313            check. The inclusion of the init task's soft limit into
2314            the computation is to avoid resetting soft limits higher
2315            than the default soft limit for cases where the default
2316            is lower than the hard limit, e.g. RLIMIT_CORE or
2317            RLIMIT_STACK.*/
2318         rc = avc_has_perm(tsec->osid, tsec->sid, SECCLASS_PROCESS,
2319                           PROCESS__RLIMITINH, NULL);
2320         if (rc) {
2321                 for (i = 0; i < RLIM_NLIMITS; i++) {
2322                         rlim = current->signal->rlim + i;
2323                         initrlim = init_task.signal->rlim+i;
2324                         rlim->rlim_cur = min(rlim->rlim_max, initrlim->rlim_cur);
2325                 }
2326                 if (current->signal->rlim[RLIMIT_CPU].rlim_cur != RLIM_INFINITY) {
2327                         /*
2328                          * This will cause RLIMIT_CPU calculations
2329                          * to be refigured.
2330                          */
2331                         current->it_prof_expires = jiffies_to_cputime(1);
2332                 }
2333         }
2334
2335         /* Wake up the parent if it is waiting so that it can
2336            recheck wait permission to the new task SID. */
2337         wake_up_interruptible(&current->parent->signal->wait_chldexit);
2338 }
2339
2340 /* superblock security operations */
2341
2342 static int selinux_sb_alloc_security(struct super_block *sb)
2343 {
2344         return superblock_alloc_security(sb);
2345 }
2346
2347 static void selinux_sb_free_security(struct super_block *sb)
2348 {
2349         superblock_free_security(sb);
2350 }
2351
2352 static inline int match_prefix(char *prefix, int plen, char *option, int olen)
2353 {
2354         if (plen > olen)
2355                 return 0;
2356
2357         return !memcmp(prefix, option, plen);
2358 }
2359
2360 static inline int selinux_option(char *option, int len)
2361 {
2362         return (match_prefix(CONTEXT_STR, sizeof(CONTEXT_STR)-1, option, len) ||
2363                 match_prefix(FSCONTEXT_STR, sizeof(FSCONTEXT_STR)-1, option, len) ||
2364                 match_prefix(DEFCONTEXT_STR, sizeof(DEFCONTEXT_STR)-1, option, len) ||
2365                 match_prefix(ROOTCONTEXT_STR, sizeof(ROOTCONTEXT_STR)-1, option, len));
2366 }
2367
2368 static inline void take_option(char **to, char *from, int *first, int len)
2369 {
2370         if (!*first) {
2371                 **to = ',';
2372                 *to += 1;
2373         } else
2374                 *first = 0;
2375         memcpy(*to, from, len);
2376         *to += len;
2377 }
2378
2379 static inline void take_selinux_option(char **to, char *from, int *first,
2380                                        int len)
2381 {
2382         int current_size = 0;
2383
2384         if (!*first) {
2385                 **to = '|';
2386                 *to += 1;
2387         } else
2388                 *first = 0;
2389
2390         while (current_size < len) {
2391                 if (*from != '"') {
2392                         **to = *from;
2393                         *to += 1;
2394                 }
2395                 from += 1;
2396                 current_size += 1;
2397         }
2398 }
2399
2400 static int selinux_sb_copy_data(char *orig, char *copy)
2401 {
2402         int fnosec, fsec, rc = 0;
2403         char *in_save, *in_curr, *in_end;
2404         char *sec_curr, *nosec_save, *nosec;
2405         int open_quote = 0;
2406
2407         in_curr = orig;
2408         sec_curr = copy;
2409
2410         nosec = (char *)get_zeroed_page(GFP_KERNEL);
2411         if (!nosec) {
2412                 rc = -ENOMEM;
2413                 goto out;
2414         }
2415
2416         nosec_save = nosec;
2417         fnosec = fsec = 1;
2418         in_save = in_end = orig;
2419
2420         do {
2421                 if (*in_end == '"')
2422                         open_quote = !open_quote;
2423                 if ((*in_end == ',' && open_quote == 0) ||
2424                                 *in_end == '\0') {
2425                         int len = in_end - in_curr;
2426
2427                         if (selinux_option(in_curr, len))
2428                                 take_selinux_option(&sec_curr, in_curr, &fsec, len);
2429                         else
2430                                 take_option(&nosec, in_curr, &fnosec, len);
2431
2432                         in_curr = in_end + 1;
2433                 }
2434         } while (*in_end++);
2435
2436         strcpy(in_save, nosec_save);
2437         free_page((unsigned long)nosec_save);
2438 out:
2439         return rc;
2440 }
2441
2442 static int selinux_sb_kern_mount(struct super_block *sb, void *data)
2443 {
2444         struct avc_audit_data ad;
2445         int rc;
2446
2447         rc = superblock_doinit(sb, data);
2448         if (rc)
2449                 return rc;
2450
2451         AVC_AUDIT_DATA_INIT(&ad, FS);
2452         ad.u.fs.path.dentry = sb->s_root;
2453         return superblock_has_perm(current, sb, FILESYSTEM__MOUNT, &ad);
2454 }
2455
2456 static int selinux_sb_statfs(struct dentry *dentry)
2457 {
2458         struct avc_audit_data ad;
2459
2460         AVC_AUDIT_DATA_INIT(&ad, FS);
2461         ad.u.fs.path.dentry = dentry->d_sb->s_root;
2462         return superblock_has_perm(current, dentry->d_sb, FILESYSTEM__GETATTR, &ad);
2463 }
2464
2465 static int selinux_mount(char *dev_name,
2466                          struct path *path,
2467                          char *type,
2468                          unsigned long flags,
2469                          void *data)
2470 {
2471         int rc;
2472
2473         rc = secondary_ops->sb_mount(dev_name, path, type, flags, data);
2474         if (rc)
2475                 return rc;
2476
2477         if (flags & MS_REMOUNT)
2478                 return superblock_has_perm(current, path->mnt->mnt_sb,
2479                                            FILESYSTEM__REMOUNT, NULL);
2480         else
2481                 return dentry_has_perm(current, path->mnt, path->dentry,
2482                                        FILE__MOUNTON);
2483 }
2484
2485 static int selinux_umount(struct vfsmount *mnt, int flags)
2486 {
2487         int rc;
2488
2489         rc = secondary_ops->sb_umount(mnt, flags);
2490         if (rc)
2491                 return rc;
2492
2493         return superblock_has_perm(current, mnt->mnt_sb,
2494                                    FILESYSTEM__UNMOUNT, NULL);
2495 }
2496
2497 /* inode security operations */
2498
2499 static int selinux_inode_alloc_security(struct inode *inode)
2500 {
2501         return inode_alloc_security(inode);
2502 }
2503
2504 static void selinux_inode_free_security(struct inode *inode)
2505 {
2506         inode_free_security(inode);
2507 }
2508
2509 static int selinux_inode_init_security(struct inode *inode, struct inode *dir,
2510                                        char **name, void **value,
2511                                        size_t *len)
2512 {
2513         struct task_security_struct *tsec;
2514         struct inode_security_struct *dsec;
2515         struct superblock_security_struct *sbsec;
2516         u32 newsid, clen;
2517         int rc;
2518         char *namep = NULL, *context;
2519
2520         tsec = current->security;
2521         dsec = dir->i_security;
2522         sbsec = dir->i_sb->s_security;
2523
2524         if (tsec->create_sid && sbsec->behavior != SECURITY_FS_USE_MNTPOINT) {
2525                 newsid = tsec->create_sid;
2526         } else {
2527                 rc = security_transition_sid(tsec->sid, dsec->sid,
2528                                              inode_mode_to_security_class(inode->i_mode),
2529                                              &newsid);
2530                 if (rc) {
2531                         printk(KERN_WARNING "%s:  "
2532                                "security_transition_sid failed, rc=%d (dev=%s "
2533                                "ino=%ld)\n",
2534                                __func__,
2535                                -rc, inode->i_sb->s_id, inode->i_ino);
2536                         return rc;
2537                 }
2538         }
2539
2540         /* Possibly defer initialization to selinux_complete_init. */
2541         if (sbsec->initialized) {
2542                 struct inode_security_struct *isec = inode->i_security;
2543                 isec->sclass = inode_mode_to_security_class(inode->i_mode);
2544                 isec->sid = newsid;
2545                 isec->initialized = 1;
2546         }
2547
2548         if (!ss_initialized || sbsec->behavior == SECURITY_FS_USE_MNTPOINT)
2549                 return -EOPNOTSUPP;
2550
2551         if (name) {
2552                 namep = kstrdup(XATTR_SELINUX_SUFFIX, GFP_NOFS);
2553                 if (!namep)
2554                         return -ENOMEM;
2555                 *name = namep;
2556         }
2557
2558         if (value && len) {
2559                 rc = security_sid_to_context_force(newsid, &context, &clen);
2560                 if (rc) {
2561                         kfree(namep);
2562                         return rc;
2563                 }
2564                 *value = context;
2565                 *len = clen;
2566         }
2567
2568         return 0;
2569 }
2570
2571 static int selinux_inode_create(struct inode *dir, struct dentry *dentry, int mask)
2572 {
2573         return may_create(dir, dentry, SECCLASS_FILE);
2574 }
2575
2576 static int selinux_inode_link(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry)
2577 {
2578         int rc;
2579
2580         rc = secondary_ops->inode_link(old_dentry, dir, new_dentry);
2581         if (rc)
2582                 return rc;
2583         return may_link(dir, old_dentry, MAY_LINK);
2584 }
2585
2586 static int selinux_inode_unlink(struct inode *dir, struct dentry *dentry)
2587 {
2588         int rc;
2589
2590         rc = secondary_ops->inode_unlink(dir, dentry);
2591         if (rc)
2592                 return rc;
2593         return may_link(dir, dentry, MAY_UNLINK);
2594 }
2595
2596 static int selinux_inode_symlink(struct inode *dir, struct dentry *dentry, const char *name)
2597 {
2598         return may_create(dir, dentry, SECCLASS_LNK_FILE);
2599 }
2600
2601 static int selinux_inode_mkdir(struct inode *dir, struct dentry *dentry, int mask)
2602 {
2603         return may_create(dir, dentry, SECCLASS_DIR);
2604 }
2605
2606 static int selinux_inode_rmdir(struct inode *dir, struct dentry *dentry)
2607 {
2608         return may_link(dir, dentry, MAY_RMDIR);
2609 }
2610
2611 static int selinux_inode_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
2612 {
2613         int rc;
2614
2615         rc = secondary_ops->inode_mknod(dir, dentry, mode, dev);
2616         if (rc)
2617                 return rc;
2618
2619         return may_create(dir, dentry, inode_mode_to_security_class(mode));
2620 }
2621
2622 static int selinux_inode_rename(struct inode *old_inode, struct dentry *old_dentry,
2623                                 struct inode *new_inode, struct dentry *new_dentry)
2624 {
2625         return may_rename(old_inode, old_dentry, new_inode, new_dentry);
2626 }
2627
2628 static int selinux_inode_readlink(struct dentry *dentry)
2629 {
2630         return dentry_has_perm(current, NULL, dentry, FILE__READ);
2631 }
2632
2633 static int selinux_inode_follow_link(struct dentry *dentry, struct nameidata *nameidata)
2634 {
2635         int rc;
2636
2637         rc = secondary_ops->inode_follow_link(dentry, nameidata);
2638         if (rc)
2639                 return rc;
2640         return dentry_has_perm(current, NULL, dentry, FILE__READ);
2641 }
2642
2643 static int selinux_inode_permission(struct inode *inode, int mask)
2644 {
2645         int rc;
2646
2647         rc = secondary_ops->inode_permission(inode, mask);
2648         if (rc)
2649                 return rc;
2650
2651         if (!mask) {
2652                 /* No permission to check.  Existence test. */
2653                 return 0;
2654         }
2655
2656         return inode_has_perm(current, inode,
2657                                open_file_mask_to_av(inode->i_mode, mask), NULL);
2658 }
2659
2660 static int selinux_inode_setattr(struct dentry *dentry, struct iattr *iattr)
2661 {
2662         int rc;
2663
2664         rc = secondary_ops->inode_setattr(dentry, iattr);
2665         if (rc)
2666                 return rc;
2667
2668         if (iattr->ia_valid & ATTR_FORCE)
2669                 return 0;
2670
2671         if (iattr->ia_valid & (ATTR_MODE | ATTR_UID | ATTR_GID |
2672                                ATTR_ATIME_SET | ATTR_MTIME_SET))
2673                 return dentry_has_perm(current, NULL, dentry, FILE__SETATTR);
2674
2675         return dentry_has_perm(current, NULL, dentry, FILE__WRITE);
2676 }
2677
2678 static int selinux_inode_getattr(struct vfsmount *mnt, struct dentry *dentry)
2679 {
2680         return dentry_has_perm(current, mnt, dentry, FILE__GETATTR);
2681 }
2682
2683 static int selinux_inode_setotherxattr(struct dentry *dentry, const char *name)
2684 {
2685         if (!strncmp(name, XATTR_SECURITY_PREFIX,
2686                      sizeof XATTR_SECURITY_PREFIX - 1)) {
2687                 if (!strcmp(name, XATTR_NAME_CAPS)) {
2688                         if (!capable(CAP_SETFCAP))
2689                                 return -EPERM;
2690                 } else if (!capable(CAP_SYS_ADMIN)) {
2691                         /* A different attribute in the security namespace.
2692                            Restrict to administrator. */
2693                         return -EPERM;
2694                 }
2695         }
2696
2697         /* Not an attribute we recognize, so just check the
2698            ordinary setattr permission. */
2699         return dentry_has_perm(current, NULL, dentry, FILE__SETATTR);
2700 }
2701
2702 static int selinux_inode_setxattr(struct dentry *dentry, const char *name,
2703                                   const void *value, size_t size, int flags)
2704 {
2705         struct task_security_struct *tsec = current->security;
2706         struct inode *inode = dentry->d_inode;
2707         struct inode_security_struct *isec = inode->i_security;
2708         struct superblock_security_struct *sbsec;
2709         struct avc_audit_data ad;
2710         u32 newsid;
2711         int rc = 0;
2712
2713         if (strcmp(name, XATTR_NAME_SELINUX))
2714                 return selinux_inode_setotherxattr(dentry, name);
2715
2716         sbsec = inode->i_sb->s_security;
2717         if (sbsec->behavior == SECURITY_FS_USE_MNTPOINT)
2718                 return -EOPNOTSUPP;
2719
2720         if (!is_owner_or_cap(inode))
2721                 return -EPERM;
2722
2723         AVC_AUDIT_DATA_INIT(&ad, FS);
2724         ad.u.fs.path.dentry = dentry;
2725
2726         rc = avc_has_perm(tsec->sid, isec->sid, isec->sclass,
2727                           FILE__RELABELFROM, &ad);
2728         if (rc)
2729                 return rc;
2730
2731         rc = security_context_to_sid(value, size, &newsid);
2732         if (rc == -EINVAL) {
2733                 if (!capable(CAP_MAC_ADMIN))
2734                         return rc;
2735                 rc = security_context_to_sid_force(value, size, &newsid);
2736         }
2737         if (rc)
2738                 return rc;
2739
2740         rc = avc_has_perm(tsec->sid, newsid, isec->sclass,
2741                           FILE__RELABELTO, &ad);
2742         if (rc)
2743                 return rc;
2744
2745         rc = security_validate_transition(isec->sid, newsid, tsec->sid,
2746                                           isec->sclass);
2747         if (rc)
2748                 return rc;
2749
2750         return avc_has_perm(newsid,
2751                             sbsec->sid,
2752                             SECCLASS_FILESYSTEM,
2753                             FILESYSTEM__ASSOCIATE,
2754                             &ad);
2755 }
2756
2757 static void selinux_inode_post_setxattr(struct dentry *dentry, const char *name,
2758                                         const void *value, size_t size,
2759                                         int flags)
2760 {
2761         struct inode *inode = dentry->d_inode;
2762         struct inode_security_struct *isec = inode->i_security;
2763         u32 newsid;
2764         int rc;
2765
2766         if (strcmp(name, XATTR_NAME_SELINUX)) {
2767                 /* Not an attribute we recognize, so nothing to do. */
2768                 return;
2769         }
2770
2771         rc = security_context_to_sid_force(value, size, &newsid);
2772         if (rc) {
2773                 printk(KERN_ERR "SELinux:  unable to map context to SID"
2774                        "for (%s, %lu), rc=%d\n",
2775                        inode->i_sb->s_id, inode->i_ino, -rc);
2776                 return;
2777         }
2778
2779         isec->sid = newsid;
2780         return;
2781 }
2782
2783 static int selinux_inode_getxattr(struct dentry *dentry, const char *name)
2784 {
2785         return dentry_has_perm(current, NULL, dentry, FILE__GETATTR);
2786 }
2787
2788 static int selinux_inode_listxattr(struct dentry *dentry)
2789 {
2790         return dentry_has_perm(current, NULL, dentry, FILE__GETATTR);
2791 }
2792
2793 static int selinux_inode_removexattr(struct dentry *dentry, const char *name)
2794 {
2795         if (strcmp(name, XATTR_NAME_SELINUX))
2796                 return selinux_inode_setotherxattr(dentry, name);
2797
2798         /* No one is allowed to remove a SELinux security label.
2799            You can change the label, but all data must be labeled. */
2800         return -EACCES;
2801 }
2802
2803 /*
2804  * Copy the inode security context value to the user.
2805  *
2806  * Permission check is handled by selinux_inode_getxattr hook.
2807  */
2808 static int selinux_inode_getsecurity(const struct inode *inode, const char *name, void **buffer, bool alloc)
2809 {
2810         u32 size;
2811         int error;
2812         char *context = NULL;
2813         struct task_security_struct *tsec = current->security;
2814         struct inode_security_struct *isec = inode->i_security;
2815
2816         if (strcmp(name, XATTR_SELINUX_SUFFIX))
2817                 return -EOPNOTSUPP;
2818
2819         /*
2820          * If the caller has CAP_MAC_ADMIN, then get the raw context
2821          * value even if it is not defined by current policy; otherwise,
2822          * use the in-core value under current policy.
2823          * Use the non-auditing forms of the permission checks since
2824          * getxattr may be called by unprivileged processes commonly
2825          * and lack of permission just means that we fall back to the
2826          * in-core context value, not a denial.
2827          */
2828         error = secondary_ops->capable(current, CAP_MAC_ADMIN);
2829         if (!error)
2830                 error = avc_has_perm_noaudit(tsec->sid, tsec->sid,
2831                                              SECCLASS_CAPABILITY2,
2832                                              CAPABILITY2__MAC_ADMIN,
2833                                              0,
2834                                              NULL);
2835         if (!error)
2836                 error = security_sid_to_context_force(isec->sid, &context,
2837                                                       &size);
2838         else
2839                 error = security_sid_to_context(isec->sid, &context, &size);
2840         if (error)
2841                 return error;
2842         error = size;
2843         if (alloc) {
2844                 *buffer = context;
2845                 goto out_nofree;
2846         }
2847         kfree(context);
2848 out_nofree:
2849         return error;
2850 }
2851
2852 static int selinux_inode_setsecurity(struct inode *inode, const char *name,
2853                                      const void *value, size_t size, int flags)
2854 {
2855         struct inode_security_struct *isec = inode->i_security;
2856         u32 newsid;
2857         int rc;
2858
2859         if (strcmp(name, XATTR_SELINUX_SUFFIX))
2860                 return -EOPNOTSUPP;
2861
2862         if (!value || !size)
2863                 return -EACCES;
2864
2865         rc = security_context_to_sid((void *)value, size, &newsid);
2866         if (rc)
2867                 return rc;
2868
2869         isec->sid = newsid;
2870         return 0;
2871 }
2872
2873 static int selinux_inode_listsecurity(struct inode *inode, char *buffer, size_t buffer_size)
2874 {
2875         const int len = sizeof(XATTR_NAME_SELINUX);
2876         if (buffer && len <= buffer_size)
2877                 memcpy(buffer, XATTR_NAME_SELINUX, len);
2878         return len;
2879 }
2880
2881 static int selinux_inode_need_killpriv(struct dentry *dentry)
2882 {
2883         return secondary_ops->inode_need_killpriv(dentry);
2884 }
2885
2886 static int selinux_inode_killpriv(struct dentry *dentry)
2887 {
2888         return secondary_ops->inode_killpriv(dentry);
2889 }
2890
2891 static void selinux_inode_getsecid(const struct inode *inode, u32 *secid)
2892 {
2893         struct inode_security_struct *isec = inode->i_security;
2894         *secid = isec->sid;
2895 }
2896
2897 /* file security operations */
2898
2899 static int selinux_revalidate_file_permission(struct file *file, int mask)
2900 {
2901         int rc;
2902         struct inode *inode = file->f_path.dentry->d_inode;
2903
2904         if (!mask) {
2905                 /* No permission to check.  Existence test. */
2906                 return 0;
2907         }
2908
2909         /* file_mask_to_av won't add FILE__WRITE if MAY_APPEND is set */
2910         if ((file->f_flags & O_APPEND) && (mask & MAY_WRITE))
2911                 mask |= MAY_APPEND;
2912
2913         rc = file_has_perm(current, file,
2914                            file_mask_to_av(inode->i_mode, mask));
2915         if (rc)
2916                 return rc;
2917
2918         return selinux_netlbl_inode_permission(inode, mask);
2919 }
2920
2921 static int selinux_file_permission(struct file *file, int mask)
2922 {
2923         struct inode *inode = file->f_path.dentry->d_inode;
2924         struct task_security_struct *tsec = current->security;
2925         struct file_security_struct *fsec = file->f_security;
2926         struct inode_security_struct *isec = inode->i_security;
2927
2928         if (!mask) {
2929                 /* No permission to check.  Existence test. */
2930                 return 0;
2931         }
2932
2933         if (tsec->sid == fsec->sid && fsec->isid == isec->sid
2934             && fsec->pseqno == avc_policy_seqno())
2935                 return selinux_netlbl_inode_permission(inode, mask);
2936
2937         return selinux_revalidate_file_permission(file, mask);
2938 }
2939
2940 static int selinux_file_alloc_security(struct file *file)
2941 {
2942         return file_alloc_security(file);
2943 }
2944
2945 static void selinux_file_free_security(struct file *file)
2946 {
2947         file_free_security(file);
2948 }
2949
2950 static int selinux_file_ioctl(struct file *file, unsigned int cmd,
2951                               unsigned long arg)
2952 {
2953         u32 av = 0;
2954
2955         if (_IOC_DIR(cmd) & _IOC_WRITE)
2956                 av |= FILE__WRITE;
2957         if (_IOC_DIR(cmd) & _IOC_READ)
2958                 av |= FILE__READ;
2959         if (!av)
2960                 av = FILE__IOCTL;
2961
2962         return file_has_perm(current, file, av);
2963 }
2964
2965 static int file_map_prot_check(struct file *file, unsigned long prot, int shared)
2966 {
2967 #ifndef CONFIG_PPC32
2968         if ((prot & PROT_EXEC) && (!file || (!shared && (prot & PROT_WRITE)))) {
2969                 /*
2970                  * We are making executable an anonymous mapping or a
2971                  * private file mapping that will also be writable.
2972                  * This has an additional check.
2973                  */
2974                 int rc = task_has_perm(current, current, PROCESS__EXECMEM);
2975                 if (rc)
2976                         return rc;
2977         }
2978 #endif
2979
2980         if (file) {
2981                 /* read access is always possible with a mapping */
2982                 u32 av = FILE__READ;
2983
2984                 /* write access only matters if the mapping is shared */
2985                 if (shared && (prot & PROT_WRITE))
2986                         av |= FILE__WRITE;
2987
2988                 if (prot & PROT_EXEC)
2989                         av |= FILE__EXECUTE;
2990
2991                 return file_has_perm(current, file, av);
2992         }
2993         return 0;
2994 }
2995
2996 static int selinux_file_mmap(struct file *file, unsigned long reqprot,
2997                              unsigned long prot, unsigned long flags,
2998                              unsigned long addr, unsigned long addr_only)
2999 {
3000         int rc = 0;
3001         u32 sid = ((struct task_security_struct *)(current->security))->sid;
3002
3003         if (addr < mmap_min_addr)
3004                 rc = avc_has_perm(sid, sid, SECCLASS_MEMPROTECT,
3005                                   MEMPROTECT__MMAP_ZERO, NULL);
3006         if (rc || addr_only)
3007                 return rc;
3008
3009         if (selinux_checkreqprot)
3010                 prot = reqprot;
3011
3012         return file_map_prot_check(file, prot,
3013                                    (flags & MAP_TYPE) == MAP_SHARED);
3014 }
3015
3016 static int selinux_file_mprotect(struct vm_area_struct *vma,
3017                                  unsigned long reqprot,
3018                                  unsigned long prot)
3019 {
3020         int rc;
3021
3022         rc = secondary_ops->file_mprotect(vma, reqprot, prot);
3023         if (rc)
3024                 return rc;
3025
3026         if (selinux_checkreqprot)
3027                 prot = reqprot;
3028
3029 #ifndef CONFIG_PPC32
3030         if ((prot & PROT_EXEC) && !(vma->vm_flags & VM_EXEC)) {
3031                 rc = 0;
3032                 if (vma->vm_start >= vma->vm_mm->start_brk &&
3033                     vma->vm_end <= vma->vm_mm->brk) {
3034                         rc = task_has_perm(current, current,
3035                                            PROCESS__EXECHEAP);
3036                 } else if (!vma->vm_file &&
3037                            vma->vm_start <= vma->vm_mm->start_stack &&
3038                            vma->vm_end >= vma->vm_mm->start_stack) {
3039                         rc = task_has_perm(current, current, PROCESS__EXECSTACK);
3040                 } else if (vma->vm_file && vma->anon_vma) {
3041                         /*
3042                          * We are making executable a file mapping that has
3043                          * had some COW done. Since pages might have been
3044                          * written, check ability to execute the possibly
3045                          * modified content.  This typically should only
3046                          * occur for text relocations.
3047                          */
3048                         rc = file_has_perm(current, vma->vm_file,
3049                                            FILE__EXECMOD);
3050                 }
3051                 if (rc)
3052                         return rc;
3053         }
3054 #endif
3055
3056         return file_map_prot_check(vma->vm_file, prot, vma->vm_flags&VM_SHARED);
3057 }
3058
3059 static int selinux_file_lock(struct file *file, unsigned int cmd)
3060 {
3061         return file_has_perm(current, file, FILE__LOCK);
3062 }
3063
3064 static int selinux_file_fcntl(struct file *file, unsigned int cmd,
3065                               unsigned long arg)
3066 {
3067         int err = 0;
3068
3069         switch (cmd) {
3070         case F_SETFL:
3071                 if (!file->f_path.dentry || !file->f_path.dentry->d_inode) {
3072                         err = -EINVAL;
3073                         break;
3074                 }
3075
3076                 if ((file->f_flags & O_APPEND) && !(arg & O_APPEND)) {
3077                         err = file_has_perm(current, file, FILE__WRITE);
3078                         break;
3079                 }
3080                 /* fall through */
3081         case F_SETOWN:
3082         case F_SETSIG:
3083         case F_GETFL:
3084         case F_GETOWN:
3085         case F_GETSIG:
3086                 /* Just check FD__USE permission */
3087                 err = file_has_perm(current, file, 0);
3088                 break;
3089         case F_GETLK:
3090         case F_SETLK:
3091         case F_SETLKW:
3092 #if BITS_PER_LONG == 32
3093         case F_GETLK64:
3094         case F_SETLK64:
3095         case F_SETLKW64:
3096 #endif
3097                 if (!file->f_path.dentry || !file->f_path.dentry->d_inode) {
3098                         err = -EINVAL;
3099                         break;
3100                 }
3101                 err = file_has_perm(current, file, FILE__LOCK);
3102                 break;
3103         }
3104
3105         return err;
3106 }
3107
3108 static int selinux_file_set_fowner(struct file *file)
3109 {
3110         struct task_security_struct *tsec;
3111         struct file_security_struct *fsec;
3112
3113         tsec = current->security;
3114         fsec = file->f_security;
3115         fsec->fown_sid = tsec->sid;
3116
3117         return 0;
3118 }
3119
3120 static int selinux_file_send_sigiotask(struct task_struct *tsk,
3121                                        struct fown_struct *fown, int signum)
3122 {
3123         struct file *file;
3124         u32 perm;
3125         struct task_security_struct *tsec;
3126         struct file_security_struct *fsec;
3127
3128         /* struct fown_struct is never outside the context of a struct file */
3129         file = container_of(fown, struct file, f_owner);
3130
3131         tsec = tsk->security;
3132         fsec = file->f_security;
3133
3134         if (!signum)
3135                 perm = signal_to_av(SIGIO); /* as per send_sigio_to_task */
3136         else
3137                 perm = signal_to_av(signum);
3138
3139         return avc_has_perm(fsec->fown_sid, tsec->sid,
3140                             SECCLASS_PROCESS, perm, NULL);
3141 }
3142
3143 static int selinux_file_receive(struct file *file)
3144 {
3145         return file_has_perm(current, file, file_to_av(file));
3146 }
3147
3148 static int selinux_dentry_open(struct file *file)
3149 {
3150         struct file_security_struct *fsec;
3151         struct inode *inode;
3152         struct inode_security_struct *isec;
3153         inode = file->f_path.dentry->d_inode;
3154         fsec = file->f_security;
3155         isec = inode->i_security;
3156         /*
3157          * Save inode label and policy sequence number
3158          * at open-time so that selinux_file_permission
3159          * can determine whether revalidation is necessary.
3160          * Task label is already saved in the file security
3161          * struct as its SID.
3162          */
3163         fsec->isid = isec->sid;
3164         fsec->pseqno = avc_policy_seqno();
3165         /*
3166          * Since the inode label or policy seqno may have changed
3167          * between the selinux_inode_permission check and the saving
3168          * of state above, recheck that access is still permitted.
3169          * Otherwise, access might never be revalidated against the
3170          * new inode label or new policy.
3171          * This check is not redundant - do not remove.
3172          */
3173         return inode_has_perm(current, inode, file_to_av(file), NULL);
3174 }
3175
3176 /* task security operations */
3177
3178 static int selinux_task_create(unsigned long clone_flags)
3179 {
3180         int rc;
3181
3182         rc = secondary_ops->task_create(clone_flags);
3183         if (rc)
3184                 return rc;
3185
3186         return task_has_perm(current, current, PROCESS__FORK);
3187 }
3188
3189 static int selinux_task_alloc_security(struct task_struct *tsk)
3190 {
3191         struct task_security_struct *tsec1, *tsec2;
3192         int rc;
3193
3194         tsec1 = current->security;
3195
3196         rc = task_alloc_security(tsk);
3197         if (rc)
3198                 return rc;
3199         tsec2 = tsk->security;
3200
3201         tsec2->osid = tsec1->osid;
3202         tsec2->sid = tsec1->sid;
3203
3204         /* Retain the exec, fs, key, and sock SIDs across fork */
3205         tsec2->exec_sid = tsec1->exec_sid;
3206         tsec2->create_sid = tsec1->create_sid;
3207         tsec2->keycreate_sid = tsec1->keycreate_sid;
3208         tsec2->sockcreate_sid = tsec1->sockcreate_sid;
3209
3210         return 0;
3211 }
3212
3213 static void selinux_task_free_security(struct task_struct *tsk)
3214 {
3215         task_free_security(tsk);
3216 }
3217
3218 static int selinux_task_setuid(uid_t id0, uid_t id1, uid_t id2, int flags)
3219 {
3220         /* Since setuid only affects the current process, and
3221            since the SELinux controls are not based on the Linux
3222            identity attributes, SELinux does not need to control
3223            this operation.  However, SELinux does control the use
3224            of the CAP_SETUID and CAP_SETGID capabilities using the
3225            capable hook. */
3226         return 0;
3227 }
3228
3229 static int selinux_task_post_setuid(uid_t id0, uid_t id1, uid_t id2, int flags)
3230 {
3231         return secondary_ops->task_post_setuid(id0, id1, id2, flags);
3232 }
3233
3234 static int selinux_task_setgid(gid_t id0, gid_t id1, gid_t id2, int flags)
3235 {
3236         /* See the comment for setuid above. */
3237         return 0;
3238 }
3239
3240 static int selinux_task_setpgid(struct task_struct *p, pid_t pgid)
3241 {
3242         return task_has_perm(current, p, PROCESS__SETPGID);
3243 }
3244
3245 static int selinux_task_getpgid(struct task_struct *p)
3246 {
3247         return task_has_perm(current, p, PROCESS__GETPGID);
3248 }
3249
3250 static int selinux_task_getsid(struct task_struct *p)
3251 {
3252         return task_has_perm(current, p, PROCESS__GETSESSION);
3253 }
3254
3255 static void selinux_task_getsecid(struct task_struct *p, u32 *secid)
3256 {
3257         struct task_security_struct *tsec = p->security;
3258         *secid = tsec->sid;
3259 }
3260
3261 static int selinux_task_setgroups(struct group_info *group_info)
3262 {
3263         /* See the comment for setuid above. */
3264         return 0;
3265 }
3266
3267 static int selinux_task_setnice(struct task_struct *p, int nice)
3268 {
3269         int rc;
3270
3271         rc = secondary_ops->task_setnice(p, nice);
3272         if (rc)
3273                 return rc;
3274
3275         return task_has_perm(current, p, PROCESS__SETSCHED);
3276 }
3277
3278 static int selinux_task_setioprio(struct task_struct *p, int ioprio)
3279 {
3280         int rc;
3281
3282         rc = secondary_ops->task_setioprio(p, ioprio);
3283         if (rc)
3284                 return rc;
3285
3286         return task_has_perm(current, p, PROCESS__SETSCHED);
3287 }
3288
3289 static int selinux_task_getioprio(struct task_struct *p)
3290 {
3291         return task_has_perm(current, p, PROCESS__GETSCHED);
3292 }
3293
3294 static int selinux_task_setrlimit(unsigned int resource, struct rlimit *new_rlim)
3295 {
3296         struct rlimit *old_rlim = current->signal->rlim + resource;
3297         int rc;
3298
3299         rc = secondary_ops->task_setrlimit(resource, new_rlim);
3300         if (rc)
3301                 return rc;
3302
3303         /* Control the ability to change the hard limit (whether
3304            lowering or raising it), so that the hard limit can
3305            later be used as a safe reset point for the soft limit
3306            upon context transitions. See selinux_bprm_apply_creds. */
3307         if (old_rlim->rlim_max != new_rlim->rlim_max)
3308                 return task_has_perm(current, current, PROCESS__SETRLIMIT);
3309
3310         return 0;
3311 }
3312
3313 static int selinux_task_setscheduler(struct task_struct *p, int policy, struct sched_param *lp)
3314 {
3315         int rc;
3316
3317         rc = secondary_ops->task_setscheduler(p, policy, lp);
3318         if (rc)
3319                 return rc;
3320
3321         return task_has_perm(current, p, PROCESS__SETSCHED);
3322 }
3323
3324 static int selinux_task_getscheduler(struct task_struct *p)
3325 {
3326         return task_has_perm(current, p, PROCESS__GETSCHED);
3327 }
3328
3329 static int selinux_task_movememory(struct task_struct *p)
3330 {
3331         return task_has_perm(current, p, PROCESS__SETSCHED);
3332 }
3333
3334 static int selinux_task_kill(struct task_struct *p, struct siginfo *info,
3335                                 int sig, u32 secid)
3336 {
3337         u32 perm;
3338         int rc;
3339         struct task_security_struct *tsec;
3340
3341         rc = secondary_ops->task_kill(p, info, sig, secid);
3342         if (rc)
3343                 return rc;
3344
3345         if (!sig)
3346                 perm = PROCESS__SIGNULL; /* null signal; existence test */
3347         else
3348                 perm = signal_to_av(sig);
3349         tsec = p->security;
3350         if (secid)
3351                 rc = avc_has_perm(secid, tsec->sid, SECCLASS_PROCESS, perm, NULL);
3352         else
3353                 rc = task_has_perm(current, p, perm);
3354         return rc;
3355 }
3356
3357 static int selinux_task_prctl(int option,
3358                               unsigned long arg2,
3359                               unsigned long arg3,
3360                               unsigned long arg4,
3361                               unsigned long arg5,
3362                               long *rc_p)
3363 {
3364         /* The current prctl operations do not appear to require
3365            any SELinux controls since they merely observe or modify
3366            the state of the current process. */
3367         return secondary_ops->task_prctl(option, arg2, arg3, arg4, arg5, rc_p);
3368 }
3369
3370 static int selinux_task_wait(struct task_struct *p)
3371 {
3372         return task_has_perm(p, current, PROCESS__SIGCHLD);
3373 }
3374
3375 static void selinux_task_reparent_to_init(struct task_struct *p)
3376 {
3377         struct task_security_struct *tsec;
3378
3379         secondary_ops->task_reparent_to_init(p);
3380
3381         tsec = p->security;
3382         tsec->osid = tsec->sid;
3383         tsec->sid = SECINITSID_KERNEL;
3384         return;
3385 }
3386
3387 static void selinux_task_to_inode(struct task_struct *p,
3388                                   struct inode *inode)
3389 {
3390         struct task_security_struct *tsec = p->security;
3391         struct inode_security_struct *isec = inode->i_security;
3392
3393         isec->sid = tsec->sid;
3394         isec->initialized = 1;
3395         return;
3396 }
3397
3398 /* Returns error only if unable to parse addresses */
3399 static int selinux_parse_skb_ipv4(struct sk_buff *skb,
3400                         struct avc_audit_data *ad, u8 *proto)
3401 {
3402         int offset, ihlen, ret = -EINVAL;
3403         struct iphdr _iph, *ih;
3404
3405         offset = skb_network_offset(skb);
3406         ih = skb_header_pointer(skb, offset, sizeof(_iph), &_iph);
3407         if (ih == NULL)
3408                 goto out;
3409
3410         ihlen = ih->ihl * 4;
3411         if (ihlen < sizeof(_iph))
3412                 goto out;
3413
3414         ad->u.net.v4info.saddr = ih->saddr;
3415         ad->u.net.v4info.daddr = ih->daddr;
3416         ret = 0;
3417
3418         if (proto)
3419                 *proto = ih->protocol;
3420
3421         switch (ih->protocol) {
3422         case IPPROTO_TCP: {
3423                 struct tcphdr _tcph, *th;
3424
3425                 if (ntohs(ih->frag_off) & IP_OFFSET)
3426                         break;
3427
3428                 offset += ihlen;
3429                 th = skb_header_pointer(skb, offset, sizeof(_tcph), &_tcph);
3430                 if (th == NULL)
3431                         break;
3432
3433                 ad->u.net.sport =