kernel/auditsc.c

   1 /* auditsc.c -- System-call auditing support
   2  * Handles all system-call specific auditing features.
   3  *
   4  * Copyright 2003-2004 Red Hat Inc., Durham, North Carolina.
   5  * All Rights Reserved.
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published by
   9  * the Free Software Foundation; either version 2 of the License, or
  10  * (at your option) any later version.
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License
  18  * along with this program; if not, write to the Free Software
  19  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  20  *
  21  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
  22  *
  23  * Many of the ideas implemented here are from Stephen C. Tweedie,
  24  * especially the idea of avoiding a copy by using getname.
  25  *
  26  * The method for actual interception of syscall entry and exit (not in
  27  * this file -- see entry.S) is based on a GPL'd patch written by
  28  * okir@suse.de and Copyright 2003 SuSE Linux AG.
  29  *
  30  */
  31
  32 #include <linux/init.h>
  33 #include <asm/atomic.h>
  34 #include <asm/types.h>
  35 #include <linux/mm.h>
  36 #include <linux/module.h>
  37 #include <linux/mount.h>
  38 #include <linux/socket.h>
  39 #include <linux/audit.h>
  40 #include <linux/personality.h>
  41 #include <linux/time.h>
  42 #include <asm/unistd.h>
  43
  44 /* 0 = no checking
  45    1 = put_count checking
  46    2 = verbose put_count checking
  47 */
  48 #define AUDIT_DEBUG 0
  49
  50 /* No syscall auditing will take place unless audit_enabled != 0. */
  51 extern int audit_enabled;
  52
  53 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
  54  * for saving names from getname(). */
  55 #define AUDIT_NAMES    20
  56
  57 /* AUDIT_NAMES_RESERVED is the number of slots we reserve in the
  58  * audit_context from being used for nameless inodes from
  59  * path_lookup. */
  60 #define AUDIT_NAMES_RESERVED 7
  61
  62 /* At task start time, the audit_state is set in the audit_context using
  63    a per-task filter.  At syscall entry, the audit_state is augmented by
  64    the syscall filter. */
  65 enum audit_state {
  66         AUDIT_DISABLED,         /* Do not create per-task audit_context.
  67                                  * No syscall-specific audit records can
  68                                  * be generated. */
  69         AUDIT_SETUP_CONTEXT,    /* Create the per-task audit_context,
  70                                  * but don't necessarily fill it in at
  71                                  * syscall entry time (i.e., filter
  72                                  * instead). */
  73         AUDIT_BUILD_CONTEXT,    /* Create the per-task audit_context,
  74                                  * and always fill it in at syscall
  75                                  * entry time.  This makes a full
  76                                  * syscall record available if some
  77                                  * other part of the kernel decides it
  78                                  * should be recorded. */
  79         AUDIT_RECORD_CONTEXT    /* Create the per-task audit_context,
  80                                  * always fill it in at syscall entry
  81                                  * time, and always write out the audit
  82                                  * record at syscall exit time.  */
  83 };
  84
  85 /* When fs/namei.c:getname() is called, we store the pointer in name and
  86  * we don't let putname() free it (instead we free all of the saved
  87  * pointers at syscall exit time).
  88  *
  89  * Further, in fs/namei.c:path_lookup() we store the inode and device. */
  90 struct audit_names {
  91         const char      *name;
  92         unsigned long   ino;
  93         dev_t           dev;
  94         umode_t         mode;
  95         uid_t           uid;
  96         gid_t           gid;
  97         dev_t           rdev;
  98 };
  99
 100 struct audit_aux_data {
 101         struct audit_aux_data   *next;
 102         int                     type;
 103 };
 104
 105 #define AUDIT_AUX_IPCPERM       0
 106
 107 struct audit_aux_data_ipcctl {
 108         struct audit_aux_data   d;
 109         struct ipc_perm         p;
 110         unsigned long           qbytes;
 111         uid_t                   uid;
 112         gid_t                   gid;
 113         mode_t                  mode;
 114 };
 115
 116 struct audit_aux_data_socketcall {
 117         struct audit_aux_data   d;
 118         int                     nargs;
 119         unsigned long           args[0];
 120 };
 121
 122 struct audit_aux_data_sockaddr {
 123         struct audit_aux_data   d;
 124         int                     len;
 125         char                    a[0];
 126 };
 127
 128 struct audit_aux_data_path {
 129         struct audit_aux_data   d;
 130         struct dentry           *dentry;
 131         struct vfsmount         *mnt;
 132 };
 133
 134 /* The per-task audit context. */
 135 struct audit_context {
 136         int                 in_syscall; /* 1 if task is in a syscall */
 137         enum audit_state    state;
 138         unsigned int        serial;     /* serial number for record */
 139         struct timespec     ctime;      /* time of syscall entry */
 140         uid_t               loginuid;   /* login uid (identity) */
 141         int                 major;      /* syscall number */
 142         unsigned long       argv[4];    /* syscall arguments */
 143         int                 return_valid; /* return code is valid */
 144         long                return_code;/* syscall return code */
 145         int                 auditable;  /* 1 if record should be written */
 146         int                 name_count;
 147         struct audit_names  names[AUDIT_NAMES];
 148         struct dentry *     pwd;
 149         struct vfsmount *   pwdmnt;
 150         struct audit_context *previous; /* For nested syscalls */
 151         struct audit_aux_data *aux;
 152
 153                                 /* Save things to print about task_struct */
 154         pid_t               pid;
 155         uid_t               uid, euid, suid, fsuid;
 156         gid_t               gid, egid, sgid, fsgid;
 157         unsigned long       personality;
 158         int                 arch;
 159
 160 #if AUDIT_DEBUG
 161         int                 put_count;
 162         int                 ino_count;
 163 #endif
 164 };
 165
 166                                 /* Public API */
 167 /* There are three lists of rules -- one to search at task creation
 168  * time, one to search at syscall entry time, and another to search at
 169  * syscall exit time. */
 170 static LIST_HEAD(audit_tsklist);
 171 static LIST_HEAD(audit_entlist);
 172 static LIST_HEAD(audit_extlist);
 173
 174 struct audit_entry {
 175         struct list_head  list;
 176         struct rcu_head   rcu;
 177         struct audit_rule rule;
 178 };
 179
 180 extern int audit_pid;
 181
 182 /* Check to see if two rules are identical.  It is called from
 183  * audit_del_rule during AUDIT_DEL. */
 184 static int audit_compare_rule(struct audit_rule *a, struct audit_rule *b)
 185 {
 186         int i;
 187
 188         if (a->flags != b->flags)
 189                 return 1;
 190
 191         if (a->action != b->action)
 192                 return 1;
 193
 194         if (a->field_count != b->field_count)
 195                 return 1;
 196
 197         for (i = 0; i < a->field_count; i++) {
 198                 if (a->fields[i] != b->fields[i]
 199                     || a->values[i] != b->values[i])
 200                         return 1;
 201         }
 202
 203         for (i = 0; i < AUDIT_BITMASK_SIZE; i++)
 204                 if (a->mask[i] != b->mask[i])
 205                         return 1;
 206
 207         return 0;
 208 }
 209
 210 /* Note that audit_add_rule and audit_del_rule are called via
 211  * audit_receive() in audit.c, and are protected by
 212  * audit_netlink_sem. */
 213 static inline int audit_add_rule(struct audit_entry *entry,
 214                                  struct list_head *list)
 215 {
 216         if (entry->rule.flags & AUDIT_PREPEND) {
 217                 entry->rule.flags &= ~AUDIT_PREPEND;
 218                 list_add_rcu(&entry->list, list);
 219         } else {
 220                 list_add_tail_rcu(&entry->list, list);
 221         }
 222         return 0;
 223 }
 224
 225 static void audit_free_rule(struct rcu_head *head)
 226 {
 227         struct audit_entry *e = container_of(head, struct audit_entry, rcu);
 228         kfree(e);
 229 }
 230
 231 /* Note that audit_add_rule and audit_del_rule are called via
 232  * audit_receive() in audit.c, and are protected by
 233  * audit_netlink_sem. */
 234 static inline int audit_del_rule(struct audit_rule *rule,
 235                                  struct list_head *list)
 236 {
 237         struct audit_entry  *e;
 238
 239         /* Do not use the _rcu iterator here, since this is the only
 240          * deletion routine. */
 241         list_for_each_entry(e, list, list) {
 242                 if (!audit_compare_rule(rule, &e->rule)) {
 243                         list_del_rcu(&e->list);
 244                         call_rcu(&e->rcu, audit_free_rule);
 245                         return 0;
 246                 }
 247         }
 248         return -EFAULT;         /* No matching rule */
 249 }
 250
 251 /* Copy rule from user-space to kernel-space.  Called during
 252  * AUDIT_ADD. */
 253 static int audit_copy_rule(struct audit_rule *d, struct audit_rule *s)
 254 {
 255         int i;
 256
 257         if (s->action != AUDIT_NEVER
 258             && s->action != AUDIT_POSSIBLE
 259             && s->action != AUDIT_ALWAYS)
 260                 return -1;
 261         if (s->field_count < 0 || s->field_count > AUDIT_MAX_FIELDS)
 262                 return -1;
 263
 264         d->flags        = s->flags;
 265         d->action       = s->action;
 266         d->field_count  = s->field_count;
 267         for (i = 0; i < d->field_count; i++) {
 268                 d->fields[i] = s->fields[i];
 269                 d->values[i] = s->values[i];
 270         }
 271         for (i = 0; i < AUDIT_BITMASK_SIZE; i++) d->mask[i] = s->mask[i];
 272         return 0;
 273 }
 274
 275 int audit_receive_filter(int type, int pid, int uid, int seq, void *data,
 276                                                         uid_t loginuid)
 277 {
 278         u32                flags;
 279         struct audit_entry *entry;
 280         int                err = 0;
 281
 282         switch (type) {
 283         case AUDIT_LIST:
 284                 /* The *_rcu iterators not needed here because we are
 285                    always called with audit_netlink_sem held. */
 286                 list_for_each_entry(entry, &audit_tsklist, list)
 287                         audit_send_reply(pid, seq, AUDIT_LIST, 0, 1,
 288                                          &entry->rule, sizeof(entry->rule));
 289                 list_for_each_entry(entry, &audit_entlist, list)
 290                         audit_send_reply(pid, seq, AUDIT_LIST, 0, 1,
 291                                          &entry->rule, sizeof(entry->rule));
 292                 list_for_each_entry(entry, &audit_extlist, list)
 293                         audit_send_reply(pid, seq, AUDIT_LIST, 0, 1,
 294                                          &entry->rule, sizeof(entry->rule));
 295                 audit_send_reply(pid, seq, AUDIT_LIST, 1, 1, NULL, 0);
 296                 break;
 297         case AUDIT_ADD:
 298                 if (!(entry = kmalloc(sizeof(*entry), GFP_KERNEL)))
 299                         return -ENOMEM;
 300                 if (audit_copy_rule(&entry->rule, data)) {
 301                         kfree(entry);
 302                         return -EINVAL;
 303                 }
 304                 flags = entry->rule.flags;
 305                 if (!err && (flags & AUDIT_PER_TASK))
 306                         err = audit_add_rule(entry, &audit_tsklist);
 307                 if (!err && (flags & AUDIT_AT_ENTRY))
 308                         err = audit_add_rule(entry, &audit_entlist);
 309                 if (!err && (flags & AUDIT_AT_EXIT))
 310                         err = audit_add_rule(entry, &audit_extlist);
 311                 audit_log(NULL, AUDIT_CONFIG_CHANGE,
 312                                 "auid=%u added an audit rule\n", loginuid);
 313                 break;
 314         case AUDIT_DEL:
 315                 flags =((struct audit_rule *)data)->flags;
 316                 if (!err && (flags & AUDIT_PER_TASK))
 317                         err = audit_del_rule(data, &audit_tsklist);
 318                 if (!err && (flags & AUDIT_AT_ENTRY))
 319                         err = audit_del_rule(data, &audit_entlist);
 320                 if (!err && (flags & AUDIT_AT_EXIT))
 321                         err = audit_del_rule(data, &audit_extlist);
 322                 audit_log(NULL, AUDIT_CONFIG_CHANGE,
 323                                 "auid=%u removed an audit rule\n", loginuid);
 324                 break;
 325         default:
 326                 return -EINVAL;
 327         }
 328
 329         return err;
 330 }
 331
 332 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
 333  * otherwise. */
 334 static int audit_filter_rules(struct task_struct *tsk,
 335                               struct audit_rule *rule,
 336                               struct audit_context *ctx,
 337                               enum audit_state *state)
 338 {
 339         int i, j;
 340
 341         for (i = 0; i < rule->field_count; i++) {
 342                 u32 field  = rule->fields[i] & ~AUDIT_NEGATE;
 343                 u32 value  = rule->values[i];
 344                 int result = 0;
 345
 346                 switch (field) {
 347                 case AUDIT_PID:
 348                         result = (tsk->pid == value);
 349                         break;
 350                 case AUDIT_UID:
 351                         result = (tsk->uid == value);
 352                         break;
 353                 case AUDIT_EUID:
 354                         result = (tsk->euid == value);
 355                         break;
 356                 case AUDIT_SUID:
 357                         result = (tsk->suid == value);
 358                         break;
 359                 case AUDIT_FSUID:
 360                         result = (tsk->fsuid == value);
 361                         break;
 362                 case AUDIT_GID:
 363                         result = (tsk->gid == value);
 364                         break;
 365                 case AUDIT_EGID:
 366                         result = (tsk->egid == value);
 367                         break;
 368                 case AUDIT_SGID:
 369                         result = (tsk->sgid == value);
 370                         break;
 371                 case AUDIT_FSGID:
 372                         result = (tsk->fsgid == value);
 373                         break;
 374                 case AUDIT_PERS:
 375                         result = (tsk->personality == value);
 376                         break;
 377                 case AUDIT_ARCH:
 378                         if (ctx)
 379                                 result = (ctx->arch == value);
 380                         break;
 381
 382                 case AUDIT_EXIT:
 383                         if (ctx && ctx->return_valid)
 384                                 result = (ctx->return_code == value);
 385                         break;
 386                 case AUDIT_SUCCESS:
 387                         if (ctx && ctx->return_valid)
 388                                 result = (ctx->return_valid == AUDITSC_SUCCESS);
 389                         break;
 390                 case AUDIT_DEVMAJOR:
 391                         if (ctx) {
 392                                 for (j = 0; j < ctx->name_count; j++) {
 393                                         if (MAJOR(ctx->names[j].dev)==value) {
 394                                                 ++result;
 395                                                 break;
 396                                         }
 397                                 }
 398                         }
 399                         break;
 400                 case AUDIT_DEVMINOR:
 401                         if (ctx) {
 402                                 for (j = 0; j < ctx->name_count; j++) {
 403                                         if (MINOR(ctx->names[j].dev)==value) {
 404                                                 ++result;
 405                                                 break;
 406                                         }
 407                                 }
 408                         }
 409                         break;
 410                 case AUDIT_INODE:
 411                         if (ctx) {
 412                                 for (j = 0; j < ctx->name_count; j++) {
 413                                         if (ctx->names[j].ino == value) {
 414                                                 ++result;
 415                                                 break;
 416                                         }
 417                                 }
 418                         }
 419                         break;
 420                 case AUDIT_LOGINUID:
 421                         result = 0;
 422                         if (ctx)
 423                                 result = (ctx->loginuid == value);
 424                         break;
 425                 case AUDIT_ARG0:
 426                 case AUDIT_ARG1:
 427                 case AUDIT_ARG2:
 428                 case AUDIT_ARG3:
 429                         if (ctx)
 430                                 result = (ctx->argv[field-AUDIT_ARG0]==value);
 431                         break;
 432                 }
 433
 434                 if (rule->fields[i] & AUDIT_NEGATE)
 435                         result = !result;
 436                 if (!result)
 437                         return 0;
 438         }
 439         switch (rule->action) {
 440         case AUDIT_NEVER:    *state = AUDIT_DISABLED;       break;
 441         case AUDIT_POSSIBLE: *state = AUDIT_BUILD_CONTEXT;  break;
 442         case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
 443         }
 444         return 1;
 445 }
 446
 447 /* At process creation time, we can determine if system-call auditing is
 448  * completely disabled for this task.  Since we only have the task
 449  * structure at this point, we can only check uid and gid.
 450  */
 451 static enum audit_state audit_filter_task(struct task_struct *tsk)
 452 {
 453         struct audit_entry *e;
 454         enum audit_state   state;
 455
 456         rcu_read_lock();
 457         list_for_each_entry_rcu(e, &audit_tsklist, list) {
 458                 if (audit_filter_rules(tsk, &e->rule, NULL, &state)) {
 459                         rcu_read_unlock();
 460                         return state;
 461                 }
 462         }
 463         rcu_read_unlock();
 464         return AUDIT_BUILD_CONTEXT;
 465 }
 466
 467 /* At syscall entry and exit time, this filter is called if the
 468  * audit_state is not low enough that auditing cannot take place, but is
 469  * also not high enough that we already know we have to write an audit
 470  * record (i.e., the state is AUDIT_SETUP_CONTEXT or  AUDIT_BUILD_CONTEXT).
 471  */
 472 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
 473                                              struct audit_context *ctx,
 474                                              struct list_head *list)
 475 {
 476         struct audit_entry *e;
 477         enum audit_state   state;
 478         int                word = AUDIT_WORD(ctx->major);
 479         int                bit  = AUDIT_BIT(ctx->major);
 480
 481         rcu_read_lock();
 482         list_for_each_entry_rcu(e, list, list) {
 483                 if ((e->rule.mask[word] & bit) == bit
 484                     && audit_filter_rules(tsk, &e->rule, ctx, &state)) {
 485                         rcu_read_unlock();
 486                         return state;
 487                 }
 488         }
 489         rcu_read_unlock();
 490         return AUDIT_BUILD_CONTEXT;
 491 }
 492
 493 /* This should be called with task_lock() held. */
 494 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
 495                                                       int return_valid,
 496                                                       int return_code)
 497 {
 498         struct audit_context *context = tsk->audit_context;
 499
 500         if (likely(!context))
 501                 return NULL;
 502         context->return_valid = return_valid;
 503         context->return_code  = return_code;
 504
 505         if (context->in_syscall && !context->auditable) {
 506                 enum audit_state state;
 507                 state = audit_filter_syscall(tsk, context, &audit_extlist);
 508                 if (state == AUDIT_RECORD_CONTEXT)
 509                         context->auditable = 1;
 510         }
 511
 512         context->pid = tsk->pid;
 513         context->uid = tsk->uid;
 514         context->gid = tsk->gid;
 515         context->euid = tsk->euid;
 516         context->suid = tsk->suid;
 517         context->fsuid = tsk->fsuid;
 518         context->egid = tsk->egid;
 519         context->sgid = tsk->sgid;
 520         context->fsgid = tsk->fsgid;
 521         context->personality = tsk->personality;
 522         tsk->audit_context = NULL;
 523         return context;
 524 }
 525
 526 static inline void audit_free_names(struct audit_context *context)
 527 {
 528         int i;
 529
 530 #if AUDIT_DEBUG == 2
 531         if (context->auditable
 532             ||context->put_count + context->ino_count != context->name_count) {
 533                 printk(KERN_ERR "audit.c:%d(:%d): major=%d in_syscall=%d"
 534                        " name_count=%d put_count=%d"
 535                        " ino_count=%d [NOT freeing]\n",
 536                        __LINE__,
 537                        context->serial, context->major, context->in_syscall,
 538                        context->name_count, context->put_count,
 539                        context->ino_count);
 540                 for (i = 0; i < context->name_count; i++)
 541                         printk(KERN_ERR "names[%d] = %p = %s\n", i,
 542                                context->names[i].name,
 543                                context->names[i].name);
 544                 dump_stack();
 545                 return;
 546         }
 547 #endif
 548 #if AUDIT_DEBUG
 549         context->put_count  = 0;
 550         context->ino_count  = 0;
 551 #endif
 552
 553         for (i = 0; i < context->name_count; i++)
 554                 if (context->names[i].name)
 555                         __putname(context->names[i].name);
 556         context->name_count = 0;
 557         if (context->pwd)
 558                 dput(context->pwd);
 559         if (context->pwdmnt)
 560                 mntput(context->pwdmnt);
 561         context->pwd = NULL;
 562         context->pwdmnt = NULL;
 563 }
 564
 565 static inline void audit_free_aux(struct audit_context *context)
 566 {
 567         struct audit_aux_data *aux;
 568
 569         while ((aux = context->aux)) {
 570                 if (aux->type == AUDIT_AVC_PATH) {
 571                         struct audit_aux_data_path *axi = (void *)aux;
 572                         dput(axi->dentry);
 573                         mntput(axi->mnt);
 574                 }
 575                 context->aux = aux->next;
 576                 kfree(aux);
 577         }
 578 }
 579
 580 static inline void audit_zero_context(struct audit_context *context,
 581                                       enum audit_state state)
 582 {
 583         uid_t loginuid = context->loginuid;
 584
 585         memset(context, 0, sizeof(*context));
 586         context->state      = state;
 587         context->loginuid   = loginuid;
 588 }
 589
 590 static inline struct audit_context *audit_alloc_context(enum audit_state state)
 591 {
 592         struct audit_context *context;
 593
 594         if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
 595                 return NULL;
 596         audit_zero_context(context, state);
 597         return context;
 598 }
 599
 600 /* Filter on the task information and allocate a per-task audit context
 601  * if necessary.  Doing so turns on system call auditing for the
 602  * specified task.  This is called from copy_process, so no lock is
 603  * needed. */
 604 int audit_alloc(struct task_struct *tsk)
 605 {
 606         struct audit_context *context;
 607         enum audit_state     state;
 608
 609         if (likely(!audit_enabled))
 610                 return 0; /* Return if not auditing. */
 611
 612         state = audit_filter_task(tsk);
 613         if (likely(state == AUDIT_DISABLED))
 614                 return 0;
 615
 616         if (!(context = audit_alloc_context(state))) {
 617                 audit_log_lost("out of memory in audit_alloc");
 618                 return -ENOMEM;
 619         }
 620
 621                                 /* Preserve login uid */
 622         context->loginuid = -1;
 623         if (current->audit_context)
 624                 context->loginuid = current->audit_context->loginuid;
 625
 626         tsk->audit_context  = context;
 627         set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
 628         return 0;
 629 }
 630
 631 static inline void audit_free_context(struct audit_context *context)
 632 {
 633         struct audit_context *previous;
 634         int                  count = 0;
 635
 636         do {
 637                 previous = context->previous;
 638                 if (previous || (count &&  count < 10)) {
 639                         ++count;
 640                         printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
 641                                " freeing multiple contexts (%d)\n",
 642                                context->serial, context->major,
 643                                context->name_count, count);
 644                 }
 645                 audit_free_names(context);
 646                 audit_free_aux(context);
 647                 kfree(context);
 648                 context  = previous;
 649         } while (context);
 650         if (count >= 10)
 651                 printk(KERN_ERR "audit: freed %d contexts\n", count);
 652 }
 653
 654 static void audit_log_task_info(struct audit_buffer *ab)
 655 {
 656         char name[sizeof(current->comm)];
 657         struct mm_struct *mm = current->mm;
 658         struct vm_area_struct *vma;
 659
 660         get_task_comm(name, current);
 661         audit_log_format(ab, " comm=");
 662         audit_log_untrustedstring(ab, name);
 663
 664         if (!mm)
 665                 return;
 666
 667         down_read(&mm->mmap_sem);
 668         vma = mm->mmap;
 669         while (vma) {
 670                 if ((vma->vm_flags & VM_EXECUTABLE) &&
 671                     vma->vm_file) {
 672                         audit_log_d_path(ab, "exe=",
 673                                          vma->vm_file->f_dentry,
 674                                          vma->vm_file->f_vfsmnt);
 675                         break;
 676                 }
 677                 vma = vma->vm_next;
 678         }
 679         up_read(&mm->mmap_sem);
 680 }
 681
 682 static void audit_log_exit(struct audit_context *context)
 683 {
 684         int i;
 685         struct audit_buffer *ab;
 686         struct audit_aux_data *aux;
 687
 688         ab = audit_log_start(context, AUDIT_SYSCALL);
 689         if (!ab)
 690                 return;         /* audit_panic has been called */
 691         audit_log_format(ab, "arch=%x syscall=%d",
 692                          context->arch, context->major);
 693         if (context->personality != PER_LINUX)
 694                 audit_log_format(ab, " per=%lx", context->personality);
 695         if (context->return_valid)
 696                 audit_log_format(ab, " success=%s exit=%ld",
 697                                  (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
 698                                  context->return_code);
 699         audit_log_format(ab,
 700                   " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
 701                   " pid=%d auid=%u uid=%u gid=%u"
 702                   " euid=%u suid=%u fsuid=%u"
 703                   " egid=%u sgid=%u fsgid=%u",
 704                   context->argv[0],
 705                   context->argv[1],
 706                   context->argv[2],
 707                   context->argv[3],
 708                   context->name_count,
 709                   context->pid,
 710                   context->loginuid,
 711                   context->uid,
 712                   context->gid,
 713                   context->euid, context->suid, context->fsuid,
 714                   context->egid, context->sgid, context->fsgid);
 715         audit_log_task_info(ab);
 716         audit_log_end(ab);
 717
 718         for (aux = context->aux; aux; aux = aux->next) {
 719
 720                 ab = audit_log_start(context, aux->type);
 721                 if (!ab)
 722                         continue; /* audit_panic has been called */
 723
 724                 switch (aux->type) {
 725                 case AUDIT_IPC: {
 726                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 727                         audit_log_format(ab,
 728                                          " qbytes=%lx iuid=%u igid=%u mode=%x",
 729                                          axi->qbytes, axi->uid, axi->gid, axi->mode);
 730                         break; }
 731
 732                 case AUDIT_SOCKETCALL: {
 733                         int i;
 734                         struct audit_aux_data_socketcall *axs = (void *)aux;
 735                         audit_log_format(ab, "nargs=%d", axs->nargs);
 736                         for (i=0; i<axs->nargs; i++)
 737                                 audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
 738                         break; }
 739
 740                 case AUDIT_SOCKADDR: {
 741                         struct audit_aux_data_sockaddr *axs = (void *)aux;
 742
 743                         audit_log_format(ab, "saddr=");
 744                         audit_log_hex(ab, axs->a, axs->len);
 745                         break; }
 746
 747                 case AUDIT_AVC_PATH: {
 748                         struct audit_aux_data_path *axi = (void *)aux;
 749                         audit_log_d_path(ab, "path=", axi->dentry, axi->mnt);
 750                         break; }
 751
 752                 }
 753                 audit_log_end(ab);
 754         }
 755
 756         if (context->pwd && context->pwdmnt) {
 757                 ab = audit_log_start(context, AUDIT_CWD);
 758                 if (ab) {
 759                         audit_log_d_path(ab, "cwd=", context->pwd, context->pwdmnt);
 760                         audit_log_end(ab);
 761                 }
 762         }
 763         for (i = 0; i < context->name_count; i++) {
 764                 ab = audit_log_start(context, AUDIT_PATH);
 765                 if (!ab)
 766                         continue; /* audit_panic has been called */
 767
 768                 audit_log_format(ab, "item=%d", i);
 769                 if (context->names[i].name) {
 770                         audit_log_format(ab, " name=");
 771                         audit_log_untrustedstring(ab, context->names[i].name);
 772                 }
 773                 if (context->names[i].ino != (unsigned long)-1)
 774                         audit_log_format(ab, " inode=%lu dev=%02x:%02x mode=%#o"
 775                                              " ouid=%u ogid=%u rdev=%02x:%02x",
 776                                          context->names[i].ino,
 777                                          MAJOR(context->names[i].dev),
 778                                          MINOR(context->names[i].dev),
 779                                          context->names[i].mode,
 780                                          context->names[i].uid,
 781                                          context->names[i].gid,
 782                                          MAJOR(context->names[i].rdev),
 783                                          MINOR(context->names[i].rdev));
 784                 audit_log_end(ab);
 785         }
 786 }
 787
 788 /* Free a per-task audit context.  Called from copy_process and
 789  * __put_task_struct. */
 790 void audit_free(struct task_struct *tsk)
 791 {
 792         struct audit_context *context;
 793
 794         task_lock(tsk);
 795         context = audit_get_context(tsk, 0, 0);
 796         task_unlock(tsk);
 797
 798         if (likely(!context))
 799                 return;
 800
 801         /* Check for system calls that do not go through the exit
 802          * function (e.g., exit_group), then free context block. */
 803         if (context->in_syscall && context->auditable && context->pid != audit_pid)
 804                 audit_log_exit(context);
 805
 806         audit_free_context(context);
 807 }
 808
 809 /* Fill in audit context at syscall entry.  This only happens if the
 810  * audit context was created when the task was created and the state or
 811  * filters demand the audit context be built.  If the state from the
 812  * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
 813  * then the record will be written at syscall exit time (otherwise, it
 814  * will only be written if another part of the kernel requests that it
 815  * be written). */
 816 void audit_syscall_entry(struct task_struct *tsk, int arch, int major,
 817                          unsigned long a1, unsigned long a2,
 818                          unsigned long a3, unsigned long a4)
 819 {
 820         struct audit_context *context = tsk->audit_context;
 821         enum audit_state     state;
 822
 823         BUG_ON(!context);
 824
 825         /* This happens only on certain architectures that make system
 826          * calls in kernel_thread via the entry.S interface, instead of
 827          * with direct calls.  (If you are porting to a new
 828          * architecture, hitting this condition can indicate that you
 829          * got the _exit/_leave calls backward in entry.S.)
 830          *
 831          * i386     no
 832          * x86_64   no
 833          * ppc64    yes (see arch/ppc64/kernel/misc.S)
 834          *
 835          * This also happens with vm86 emulation in a non-nested manner
 836          * (entries without exits), so this case must be caught.
 837          */
 838         if (context->in_syscall) {
 839                 struct audit_context *newctx;
 840
 841 #if defined(__NR_vm86) && defined(__NR_vm86old)
 842                 /* vm86 mode should only be entered once */
 843                 if (major == __NR_vm86 || major == __NR_vm86old)
 844                         return;
 845 #endif
 846 #if AUDIT_DEBUG
 847                 printk(KERN_ERR
 848                        "audit(:%d) pid=%d in syscall=%d;"
 849                        " entering syscall=%d\n",
 850                        context->serial, tsk->pid, context->major, major);
 851 #endif
 852                 newctx = audit_alloc_context(context->state);
 853                 if (newctx) {
 854                         newctx->previous   = context;
 855                         context            = newctx;
 856                         tsk->audit_context = newctx;
 857                 } else  {
 858                         /* If we can't alloc a new context, the best we
 859                          * can do is to leak memory (any pending putname
 860                          * will be lost).  The only other alternative is
 861                          * to abandon auditing. */
 862                         audit_zero_context(context, context->state);
 863                 }
 864         }
 865         BUG_ON(context->in_syscall || context->name_count);
 866
 867         if (!audit_enabled)
 868                 return;
 869
 870         context->arch       = arch;
 871         context->major      = major;
 872         context->argv[0]    = a1;
 873         context->argv[1]    = a2;
 874         context->argv[2]    = a3;
 875         context->argv[3]    = a4;
 876
 877         state = context->state;
 878         if (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT)
 879                 state = audit_filter_syscall(tsk, context, &audit_entlist);
 880         if (likely(state == AUDIT_DISABLED))
 881                 return;
 882
 883         context->serial     = audit_serial();
 884         context->ctime      = CURRENT_TIME;
 885         context->in_syscall = 1;
 886         context->auditable  = !!(state == AUDIT_RECORD_CONTEXT);
 887 }
 888
 889 /* Tear down after system call.  If the audit context has been marked as
 890  * auditable (either because of the AUDIT_RECORD_CONTEXT state from
 891  * filtering, or because some other part of the kernel write an audit
 892  * message), then write out the syscall information.  In call cases,
 893  * free the names stored from getname(). */
 894 void audit_syscall_exit(struct task_struct *tsk, int valid, long return_code)
 895 {
 896         struct audit_context *context;
 897
 898         get_task_struct(tsk);
 899         task_lock(tsk);
 900         context = audit_get_context(tsk, valid, return_code);
 901         task_unlock(tsk);
 902
 903         /* Not having a context here is ok, since the parent may have
 904          * called __put_task_struct. */
 905         if (likely(!context))
 906                 return;
 907
 908         if (context->in_syscall && context->auditable && context->pid != audit_pid)
 909                 audit_log_exit(context);
 910
 911         context->in_syscall = 0;
 912         context->auditable  = 0;
 913
 914         if (context->previous) {
 915                 struct audit_context *new_context = context->previous;
 916                 context->previous  = NULL;
 917                 audit_free_context(context);
 918                 tsk->audit_context = new_context;
 919         } else {
 920                 audit_free_names(context);
 921                 audit_free_aux(context);
 922                 audit_zero_context(context, context->state);
 923                 tsk->audit_context = context;
 924         }
 925         put_task_struct(tsk);
 926 }
 927
 928 /* Add a name to the list.  Called from fs/namei.c:getname(). */
 929 void audit_getname(const char *name)
 930 {
 931         struct audit_context *context = current->audit_context;
 932
 933         if (!context || IS_ERR(name) || !name)
 934                 return;
 935
 936         if (!context->in_syscall) {
 937 #if AUDIT_DEBUG == 2
 938                 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
 939                        __FILE__, __LINE__, context->serial, name);
 940                 dump_stack();
 941 #endif
 942                 return;
 943         }
 944         BUG_ON(context->name_count >= AUDIT_NAMES);
 945         context->names[context->name_count].name = name;
 946         context->names[context->name_count].ino  = (unsigned long)-1;
 947         ++context->name_count;
 948         if (!context->pwd) {
 949                 read_lock(&current->fs->lock);
 950                 context->pwd = dget(current->fs->pwd);
 951                 context->pwdmnt = mntget(current->fs->pwdmnt);
 952                 read_unlock(&current->fs->lock);
 953         }
 954
 955 }
 956
 957 /* Intercept a putname request.  Called from
 958  * include/linux/fs.h:putname().  If we have stored the name from
 959  * getname in the audit context, then we delay the putname until syscall
 960  * exit. */
 961 void audit_putname(const char *name)
 962 {
 963         struct audit_context *context = current->audit_context;
 964
 965         BUG_ON(!context);
 966         if (!context->in_syscall) {
 967 #if AUDIT_DEBUG == 2
 968                 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
 969                        __FILE__, __LINE__, context->serial, name);
 970                 if (context->name_count) {
 971                         int i;
 972                         for (i = 0; i < context->name_count; i++)
 973                                 printk(KERN_ERR "name[%d] = %p = %s\n", i,
 974                                        context->names[i].name,
 975                                        context->names[i].name);
 976                 }
 977 #endif
 978                 __putname(name);
 979         }
 980 #if AUDIT_DEBUG
 981         else {
 982                 ++context->put_count;
 983                 if (context->put_count > context->name_count) {
 984                         printk(KERN_ERR "%s:%d(:%d): major=%d"
 985                                " in_syscall=%d putname(%p) name_count=%d"
 986                                " put_count=%d\n",
 987                                __FILE__, __LINE__,
 988                                context->serial, context->major,
 989                                context->in_syscall, name, context->name_count,
 990                                context->put_count);
 991                         dump_stack();
 992                 }
 993         }
 994 #endif
 995 }
 996
 997 /* Store the inode and device from a lookup.  Called from
 998  * fs/namei.c:path_lookup(). */
 999 void audit_inode(const char *name, const struct inode *inode)
1000 {
1001         int idx;
1002         struct audit_context *context = current->audit_context;
1003
1004         if (!context->in_syscall)
1005                 return;
1006         if (context->name_count
1007             && context->names[context->name_count-1].name
1008             && context->names[context->name_count-1].name == name)
1009                 idx = context->name_count - 1;
1010         else if (context->name_count > 1
1011                  && context->names[context->name_count-2].name
1012                  && context->names[context->name_count-2].name == name)
1013                 idx = context->name_count - 2;
1014         else {
1015                 /* FIXME: how much do we care about inodes that have no
1016                  * associated name? */
1017                 if (context->name_count >= AUDIT_NAMES - AUDIT_NAMES_RESERVED)
1018                         return;
1019                 idx = context->name_count++;
1020                 context->names[idx].name = NULL;
1021 #if AUDIT_DEBUG
1022                 ++context->ino_count;
1023 #endif
1024         }
1025         context->names[idx].ino  = inode->i_ino;
1026         context->names[idx].dev  = inode->i_sb->s_dev;
1027         context->names[idx].mode = inode->i_mode;
1028         context->names[idx].uid  = inode->i_uid;
1029         context->names[idx].gid  = inode->i_gid;
1030         context->names[idx].rdev = inode->i_rdev;
1031 }
1032
1033 void auditsc_get_stamp(struct audit_context *ctx,
1034                        struct timespec *t, unsigned int *serial)
1035 {
1036         t->tv_sec  = ctx->ctime.tv_sec;
1037         t->tv_nsec = ctx->ctime.tv_nsec;
1038         *serial    = ctx->serial;
1039         ctx->auditable = 1;
1040 }
1041
1042 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
1043 {
1044         if (task->audit_context) {
1045                 struct audit_buffer *ab;
1046
1047                 ab = audit_log_start(NULL, AUDIT_LOGIN);
1048                 if (ab) {
1049                         audit_log_format(ab, "login pid=%d uid=%u "
1050                                 "old auid=%u new auid=%u",
1051                                 task->pid, task->uid,
1052                                 task->audit_context->loginuid, loginuid);
1053                         audit_log_end(ab);
1054                 }
1055                 task->audit_context->loginuid = loginuid;
1056         }
1057         return 0;
1058 }
1059
1060 uid_t audit_get_loginuid(struct audit_context *ctx)
1061 {
1062         return ctx ? ctx->loginuid : -1;
1063 }
1064
1065 int audit_ipc_perms(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
1066 {
1067         struct audit_aux_data_ipcctl *ax;
1068         struct audit_context *context = current->audit_context;
1069
1070         if (likely(!context))
1071                 return 0;
1072
1073         ax = kmalloc(sizeof(*ax), GFP_KERNEL);
1074         if (!ax)
1075                 return -ENOMEM;
1076
1077         ax->qbytes = qbytes;
1078         ax->uid = uid;
1079         ax->gid = gid;
1080         ax->mode = mode;
1081
1082         ax->d.type = AUDIT_IPC;
1083         ax->d.next = context->aux;
1084         context->aux = (void *)ax;
1085         return 0;
1086 }
1087
1088 int audit_socketcall(int nargs, unsigned long *args)
1089 {
1090         struct audit_aux_data_socketcall *ax;
1091         struct audit_context *context = current->audit_context;
1092
1093         if (likely(!context))
1094                 return 0;
1095
1096         ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
1097         if (!ax)
1098                 return -ENOMEM;
1099
1100         ax->nargs = nargs;
1101         memcpy(ax->args, args, nargs * sizeof(unsigned long));
1102
1103         ax->d.type = AUDIT_SOCKETCALL;
1104         ax->d.next = context->aux;
1105         context->aux = (void *)ax;
1106         return 0;
1107 }
1108
1109 int audit_sockaddr(int len, void *a)
1110 {
1111         struct audit_aux_data_sockaddr *ax;
1112         struct audit_context *context = current->audit_context;
1113
1114         if (likely(!context))
1115                 return 0;
1116
1117         ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
1118         if (!ax)
1119                 return -ENOMEM;
1120
1121         ax->len = len;
1122         memcpy(ax->a, a, len);
1123
1124         ax->d.type = AUDIT_SOCKADDR;
1125         ax->d.next = context->aux;
1126         context->aux = (void *)ax;
1127         return 0;
1128 }
1129
1130 int audit_avc_path(struct dentry *dentry, struct vfsmount *mnt)
1131 {
1132         struct audit_aux_data_path *ax;
1133         struct audit_context *context = current->audit_context;
1134
1135         if (likely(!context))
1136                 return 0;
1137
1138         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1139         if (!ax)
1140                 return -ENOMEM;
1141
1142         ax->dentry = dget(dentry);
1143         ax->mnt = mntget(mnt);
1144
1145         ax->d.type = AUDIT_AVC_PATH;
1146         ax->d.next = context->aux;
1147         context->aux = (void *)ax;
1148         return 0;
1149 }
1150
1151 void audit_signal_info(int sig, struct task_struct *t)
1152 {
1153         extern pid_t audit_sig_pid;
1154         extern uid_t audit_sig_uid;
1155
1156         if (unlikely(audit_pid && t->pid == audit_pid)) {
1157                 if (sig == SIGTERM || sig == SIGHUP) {
1158                         struct audit_context *ctx = current->audit_context;
1159                         audit_sig_pid = current->pid;
1160                         if (ctx)
1161                                 audit_sig_uid = ctx->loginuid;
1162                         else
1163                                 audit_sig_uid = current->uid;
1164                 }
1165         }
1166 }
1167