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  * Copyright 2005 Hewlett-Packard Development Company, L.P.
   6  * Copyright (C) 2005 IBM Corporation
   7  * All Rights Reserved.
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  22  *
  23  * Written by Rickard E. (Rik) Faith <faith@redhat.com>
  24  *
  25  * Many of the ideas implemented here are from Stephen C. Tweedie,
  26  * especially the idea of avoiding a copy by using getname.
  27  *
  28  * The method for actual interception of syscall entry and exit (not in
  29  * this file -- see entry.S) is based on a GPL'd patch written by
  30  * okir@suse.de and Copyright 2003 SuSE Linux AG.
  31  *
  32  * The support of additional filter rules compares (>, <, >=, <=) was
  33  * added by Dustin Kirkland <dustin.kirkland@us.ibm.com>, 2005.
  34  *
  35  * Modified by Amy Griffis <amy.griffis@hp.com> to collect additional
  36  * filesystem information.
  37  *
  38  * Subject and object context labeling support added by <danjones@us.ibm.com>
  39  * and <dustin.kirkland@us.ibm.com> for LSPP certification compliance.
  40  */
  41
  42 #include <linux/init.h>
  43 #include <asm/types.h>
  44 #include <asm/atomic.h>
  45 #include <asm/types.h>
  46 #include <linux/fs.h>
  47 #include <linux/namei.h>
  48 #include <linux/mm.h>
  49 #include <linux/module.h>
  50 #include <linux/mount.h>
  51 #include <linux/socket.h>
  52 #include <linux/audit.h>
  53 #include <linux/personality.h>
  54 #include <linux/time.h>
  55 #include <linux/netlink.h>
  56 #include <linux/compiler.h>
  57 #include <asm/unistd.h>
  58 #include <linux/security.h>
  59 #include <linux/list.h>
  60 #include <linux/tty.h>
  61
  62 #include "audit.h"
  63
  64 extern struct list_head audit_filter_list[];
  65
  66 /* No syscall auditing will take place unless audit_enabled != 0. */
  67 extern int audit_enabled;
  68
  69 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
  70  * for saving names from getname(). */
  71 #define AUDIT_NAMES    20
  72
  73 /* AUDIT_NAMES_RESERVED is the number of slots we reserve in the
  74  * audit_context from being used for nameless inodes from
  75  * path_lookup. */
  76 #define AUDIT_NAMES_RESERVED 7
  77
  78 /* When fs/namei.c:getname() is called, we store the pointer in name and
  79  * we don't let putname() free it (instead we free all of the saved
  80  * pointers at syscall exit time).
  81  *
  82  * Further, in fs/namei.c:path_lookup() we store the inode and device. */
  83 struct audit_names {
  84         const char      *name;
  85         unsigned long   ino;
  86         unsigned long   pino;
  87         dev_t           dev;
  88         umode_t         mode;
  89         uid_t           uid;
  90         gid_t           gid;
  91         dev_t           rdev;
  92         char            *ctx;
  93 };
  94
  95 struct audit_aux_data {
  96         struct audit_aux_data   *next;
  97         int                     type;
  98 };
  99
 100 #define AUDIT_AUX_IPCPERM       0
 101
 102 struct audit_aux_data_ipcctl {
 103         struct audit_aux_data   d;
 104         struct ipc_perm         p;
 105         unsigned long           qbytes;
 106         uid_t                   uid;
 107         gid_t                   gid;
 108         mode_t                  mode;
 109         char                    *ctx;
 110 };
 111
 112 struct audit_aux_data_socketcall {
 113         struct audit_aux_data   d;
 114         int                     nargs;
 115         unsigned long           args[0];
 116 };
 117
 118 struct audit_aux_data_sockaddr {
 119         struct audit_aux_data   d;
 120         int                     len;
 121         char                    a[0];
 122 };
 123
 124 struct audit_aux_data_path {
 125         struct audit_aux_data   d;
 126         struct dentry           *dentry;
 127         struct vfsmount         *mnt;
 128 };
 129
 130 /* The per-task audit context. */
 131 struct audit_context {
 132         int                 in_syscall; /* 1 if task is in a syscall */
 133         enum audit_state    state;
 134         unsigned int        serial;     /* serial number for record */
 135         struct timespec     ctime;      /* time of syscall entry */
 136         uid_t               loginuid;   /* login uid (identity) */
 137         int                 major;      /* syscall number */
 138         unsigned long       argv[4];    /* syscall arguments */
 139         int                 return_valid; /* return code is valid */
 140         long                return_code;/* syscall return code */
 141         int                 auditable;  /* 1 if record should be written */
 142         int                 name_count;
 143         struct audit_names  names[AUDIT_NAMES];
 144         struct dentry *     pwd;
 145         struct vfsmount *   pwdmnt;
 146         struct audit_context *previous; /* For nested syscalls */
 147         struct audit_aux_data *aux;
 148
 149                                 /* Save things to print about task_struct */
 150         pid_t               pid;
 151         uid_t               uid, euid, suid, fsuid;
 152         gid_t               gid, egid, sgid, fsgid;
 153         unsigned long       personality;
 154         int                 arch;
 155
 156 #if AUDIT_DEBUG
 157         int                 put_count;
 158         int                 ino_count;
 159 #endif
 160 };
 161
 162
 163 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
 164  * otherwise. */
 165 static int audit_filter_rules(struct task_struct *tsk,
 166                               struct audit_krule *rule,
 167                               struct audit_context *ctx,
 168                               enum audit_state *state)
 169 {
 170         int i, j;
 171
 172         for (i = 0; i < rule->field_count; i++) {
 173                 struct audit_field *f = &rule->fields[i];
 174                 int result = 0;
 175
 176                 switch (f->type) {
 177                 case AUDIT_PID:
 178                         result = audit_comparator(tsk->pid, f->op, f->val);
 179                         break;
 180                 case AUDIT_UID:
 181                         result = audit_comparator(tsk->uid, f->op, f->val);
 182                         break;
 183                 case AUDIT_EUID:
 184                         result = audit_comparator(tsk->euid, f->op, f->val);
 185                         break;
 186                 case AUDIT_SUID:
 187                         result = audit_comparator(tsk->suid, f->op, f->val);
 188                         break;
 189                 case AUDIT_FSUID:
 190                         result = audit_comparator(tsk->fsuid, f->op, f->val);
 191                         break;
 192                 case AUDIT_GID:
 193                         result = audit_comparator(tsk->gid, f->op, f->val);
 194                         break;
 195                 case AUDIT_EGID:
 196                         result = audit_comparator(tsk->egid, f->op, f->val);
 197                         break;
 198                 case AUDIT_SGID:
 199                         result = audit_comparator(tsk->sgid, f->op, f->val);
 200                         break;
 201                 case AUDIT_FSGID:
 202                         result = audit_comparator(tsk->fsgid, f->op, f->val);
 203                         break;
 204                 case AUDIT_PERS:
 205                         result = audit_comparator(tsk->personality, f->op, f->val);
 206                         break;
 207                 case AUDIT_ARCH:
 208                         if (ctx)
 209                                 result = audit_comparator(ctx->arch, f->op, f->val);
 210                         break;
 211
 212                 case AUDIT_EXIT:
 213                         if (ctx && ctx->return_valid)
 214                                 result = audit_comparator(ctx->return_code, f->op, f->val);
 215                         break;
 216                 case AUDIT_SUCCESS:
 217                         if (ctx && ctx->return_valid) {
 218                                 if (f->val)
 219                                         result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
 220                                 else
 221                                         result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
 222                         }
 223                         break;
 224                 case AUDIT_DEVMAJOR:
 225                         if (ctx) {
 226                                 for (j = 0; j < ctx->name_count; j++) {
 227                                         if (audit_comparator(MAJOR(ctx->names[j].dev),  f->op, f->val)) {
 228                                                 ++result;
 229                                                 break;
 230                                         }
 231                                 }
 232                         }
 233                         break;
 234                 case AUDIT_DEVMINOR:
 235                         if (ctx) {
 236                                 for (j = 0; j < ctx->name_count; j++) {
 237                                         if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
 238                                                 ++result;
 239                                                 break;
 240                                         }
 241                                 }
 242                         }
 243                         break;
 244                 case AUDIT_INODE:
 245                         if (ctx) {
 246                                 for (j = 0; j < ctx->name_count; j++) {
 247                                         if (audit_comparator(ctx->names[j].ino, f->op, f->val) ||
 248                                             audit_comparator(ctx->names[j].pino, f->op, f->val)) {
 249                                                 ++result;
 250                                                 break;
 251                                         }
 252                                 }
 253                         }
 254                         break;
 255                 case AUDIT_LOGINUID:
 256                         result = 0;
 257                         if (ctx)
 258                                 result = audit_comparator(ctx->loginuid, f->op, f->val);
 259                         break;
 260                 case AUDIT_ARG0:
 261                 case AUDIT_ARG1:
 262                 case AUDIT_ARG2:
 263                 case AUDIT_ARG3:
 264                         if (ctx)
 265                                 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
 266                         break;
 267                 }
 268
 269                 if (!result)
 270                         return 0;
 271         }
 272         switch (rule->action) {
 273         case AUDIT_NEVER:    *state = AUDIT_DISABLED;       break;
 274         case AUDIT_POSSIBLE: *state = AUDIT_BUILD_CONTEXT;  break;
 275         case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
 276         }
 277         return 1;
 278 }
 279
 280 /* At process creation time, we can determine if system-call auditing is
 281  * completely disabled for this task.  Since we only have the task
 282  * structure at this point, we can only check uid and gid.
 283  */
 284 static enum audit_state audit_filter_task(struct task_struct *tsk)
 285 {
 286         struct audit_entry *e;
 287         enum audit_state   state;
 288
 289         rcu_read_lock();
 290         list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
 291                 if (audit_filter_rules(tsk, &e->rule, NULL, &state)) {
 292                         rcu_read_unlock();
 293                         return state;
 294                 }
 295         }
 296         rcu_read_unlock();
 297         return AUDIT_BUILD_CONTEXT;
 298 }
 299
 300 /* At syscall entry and exit time, this filter is called if the
 301  * audit_state is not low enough that auditing cannot take place, but is
 302  * also not high enough that we already know we have to write an audit
 303  * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
 304  */
 305 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
 306                                              struct audit_context *ctx,
 307                                              struct list_head *list)
 308 {
 309         struct audit_entry *e;
 310         enum audit_state state;
 311
 312         if (audit_pid && tsk->tgid == audit_pid)
 313                 return AUDIT_DISABLED;
 314
 315         rcu_read_lock();
 316         if (!list_empty(list)) {
 317                 int word = AUDIT_WORD(ctx->major);
 318                 int bit  = AUDIT_BIT(ctx->major);
 319
 320                 list_for_each_entry_rcu(e, list, list) {
 321                         if ((e->rule.mask[word] & bit) == bit
 322                                         && audit_filter_rules(tsk, &e->rule, ctx, &state)) {
 323                                 rcu_read_unlock();
 324                                 return state;
 325                         }
 326                 }
 327         }
 328         rcu_read_unlock();
 329         return AUDIT_BUILD_CONTEXT;
 330 }
 331
 332 /* This should be called with task_lock() held. */
 333 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
 334                                                       int return_valid,
 335                                                       int return_code)
 336 {
 337         struct audit_context *context = tsk->audit_context;
 338
 339         if (likely(!context))
 340                 return NULL;
 341         context->return_valid = return_valid;
 342         context->return_code  = return_code;
 343
 344         if (context->in_syscall && !context->auditable) {
 345                 enum audit_state state;
 346                 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
 347                 if (state == AUDIT_RECORD_CONTEXT)
 348                         context->auditable = 1;
 349         }
 350
 351         context->pid = tsk->pid;
 352         context->uid = tsk->uid;
 353         context->gid = tsk->gid;
 354         context->euid = tsk->euid;
 355         context->suid = tsk->suid;
 356         context->fsuid = tsk->fsuid;
 357         context->egid = tsk->egid;
 358         context->sgid = tsk->sgid;
 359         context->fsgid = tsk->fsgid;
 360         context->personality = tsk->personality;
 361         tsk->audit_context = NULL;
 362         return context;
 363 }
 364
 365 static inline void audit_free_names(struct audit_context *context)
 366 {
 367         int i;
 368
 369 #if AUDIT_DEBUG == 2
 370         if (context->auditable
 371             ||context->put_count + context->ino_count != context->name_count) {
 372                 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
 373                        " name_count=%d put_count=%d"
 374                        " ino_count=%d [NOT freeing]\n",
 375                        __FILE__, __LINE__,
 376                        context->serial, context->major, context->in_syscall,
 377                        context->name_count, context->put_count,
 378                        context->ino_count);
 379                 for (i = 0; i < context->name_count; i++) {
 380                         printk(KERN_ERR "names[%d] = %p = %s\n", i,
 381                                context->names[i].name,
 382                                context->names[i].name ?: "(null)");
 383                 }
 384                 dump_stack();
 385                 return;
 386         }
 387 #endif
 388 #if AUDIT_DEBUG
 389         context->put_count  = 0;
 390         context->ino_count  = 0;
 391 #endif
 392
 393         for (i = 0; i < context->name_count; i++) {
 394                 char *p = context->names[i].ctx;
 395                 context->names[i].ctx = NULL;
 396                 kfree(p);
 397                 if (context->names[i].name)
 398                         __putname(context->names[i].name);
 399         }
 400         context->name_count = 0;
 401         if (context->pwd)
 402                 dput(context->pwd);
 403         if (context->pwdmnt)
 404                 mntput(context->pwdmnt);
 405         context->pwd = NULL;
 406         context->pwdmnt = NULL;
 407 }
 408
 409 static inline void audit_free_aux(struct audit_context *context)
 410 {
 411         struct audit_aux_data *aux;
 412
 413         while ((aux = context->aux)) {
 414                 if (aux->type == AUDIT_AVC_PATH) {
 415                         struct audit_aux_data_path *axi = (void *)aux;
 416                         dput(axi->dentry);
 417                         mntput(axi->mnt);
 418                 }
 419                 if ( aux->type == AUDIT_IPC ) {
 420                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 421                         if (axi->ctx)
 422                                 kfree(axi->ctx);
 423                 }
 424
 425                 context->aux = aux->next;
 426                 kfree(aux);
 427         }
 428 }
 429
 430 static inline void audit_zero_context(struct audit_context *context,
 431                                       enum audit_state state)
 432 {
 433         uid_t loginuid = context->loginuid;
 434
 435         memset(context, 0, sizeof(*context));
 436         context->state      = state;
 437         context->loginuid   = loginuid;
 438 }
 439
 440 static inline struct audit_context *audit_alloc_context(enum audit_state state)
 441 {
 442         struct audit_context *context;
 443
 444         if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
 445                 return NULL;
 446         audit_zero_context(context, state);
 447         return context;
 448 }
 449
 450 /**
 451  * audit_alloc - allocate an audit context block for a task
 452  * @tsk: task
 453  *
 454  * Filter on the task information and allocate a per-task audit context
 455  * if necessary.  Doing so turns on system call auditing for the
 456  * specified task.  This is called from copy_process, so no lock is
 457  * needed.
 458  */
 459 int audit_alloc(struct task_struct *tsk)
 460 {
 461         struct audit_context *context;
 462         enum audit_state     state;
 463
 464         if (likely(!audit_enabled))
 465                 return 0; /* Return if not auditing. */
 466
 467         state = audit_filter_task(tsk);
 468         if (likely(state == AUDIT_DISABLED))
 469                 return 0;
 470
 471         if (!(context = audit_alloc_context(state))) {
 472                 audit_log_lost("out of memory in audit_alloc");
 473                 return -ENOMEM;
 474         }
 475
 476                                 /* Preserve login uid */
 477         context->loginuid = -1;
 478         if (current->audit_context)
 479                 context->loginuid = current->audit_context->loginuid;
 480
 481         tsk->audit_context  = context;
 482         set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
 483         return 0;
 484 }
 485
 486 static inline void audit_free_context(struct audit_context *context)
 487 {
 488         struct audit_context *previous;
 489         int                  count = 0;
 490
 491         do {
 492                 previous = context->previous;
 493                 if (previous || (count &&  count < 10)) {
 494                         ++count;
 495                         printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
 496                                " freeing multiple contexts (%d)\n",
 497                                context->serial, context->major,
 498                                context->name_count, count);
 499                 }
 500                 audit_free_names(context);
 501                 audit_free_aux(context);
 502                 kfree(context);
 503                 context  = previous;
 504         } while (context);
 505         if (count >= 10)
 506                 printk(KERN_ERR "audit: freed %d contexts\n", count);
 507 }
 508
 509 static void audit_log_task_context(struct audit_buffer *ab, gfp_t gfp_mask)
 510 {
 511         char *ctx = NULL;
 512         ssize_t len = 0;
 513
 514         len = security_getprocattr(current, "current", NULL, 0);
 515         if (len < 0) {
 516                 if (len != -EINVAL)
 517                         goto error_path;
 518                 return;
 519         }
 520
 521         ctx = kmalloc(len, gfp_mask);
 522         if (!ctx)
 523                 goto error_path;
 524
 525         len = security_getprocattr(current, "current", ctx, len);
 526         if (len < 0 )
 527                 goto error_path;
 528
 529         audit_log_format(ab, " subj=%s", ctx);
 530         return;
 531
 532 error_path:
 533         if (ctx)
 534                 kfree(ctx);
 535         audit_panic("error in audit_log_task_context");
 536         return;
 537 }
 538
 539 static void audit_log_task_info(struct audit_buffer *ab, gfp_t gfp_mask)
 540 {
 541         char name[sizeof(current->comm)];
 542         struct mm_struct *mm = current->mm;
 543         struct vm_area_struct *vma;
 544
 545         get_task_comm(name, current);
 546         audit_log_format(ab, " comm=");
 547         audit_log_untrustedstring(ab, name);
 548
 549         if (!mm)
 550                 return;
 551
 552         /*
 553          * this is brittle; all callers that pass GFP_ATOMIC will have
 554          * NULL current->mm and we won't get here.
 555          */
 556         down_read(&mm->mmap_sem);
 557         vma = mm->mmap;
 558         while (vma) {
 559                 if ((vma->vm_flags & VM_EXECUTABLE) &&
 560                     vma->vm_file) {
 561                         audit_log_d_path(ab, "exe=",
 562                                          vma->vm_file->f_dentry,
 563                                          vma->vm_file->f_vfsmnt);
 564                         break;
 565                 }
 566                 vma = vma->vm_next;
 567         }
 568         up_read(&mm->mmap_sem);
 569         audit_log_task_context(ab, gfp_mask);
 570 }
 571
 572 static void audit_log_exit(struct audit_context *context, gfp_t gfp_mask)
 573 {
 574         int i;
 575         struct audit_buffer *ab;
 576         struct audit_aux_data *aux;
 577         const char *tty;
 578
 579         ab = audit_log_start(context, gfp_mask, AUDIT_SYSCALL);
 580         if (!ab)
 581                 return;         /* audit_panic has been called */
 582         audit_log_format(ab, "arch=%x syscall=%d",
 583                          context->arch, context->major);
 584         if (context->personality != PER_LINUX)
 585                 audit_log_format(ab, " per=%lx", context->personality);
 586         if (context->return_valid)
 587                 audit_log_format(ab, " success=%s exit=%ld",
 588                                  (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
 589                                  context->return_code);
 590         if (current->signal->tty && current->signal->tty->name)
 591                 tty = current->signal->tty->name;
 592         else
 593                 tty = "(none)";
 594         audit_log_format(ab,
 595                   " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
 596                   " pid=%d auid=%u uid=%u gid=%u"
 597                   " euid=%u suid=%u fsuid=%u"
 598                   " egid=%u sgid=%u fsgid=%u tty=%s",
 599                   context->argv[0],
 600                   context->argv[1],
 601                   context->argv[2],
 602                   context->argv[3],
 603                   context->name_count,
 604                   context->pid,
 605                   context->loginuid,
 606                   context->uid,
 607                   context->gid,
 608                   context->euid, context->suid, context->fsuid,
 609                   context->egid, context->sgid, context->fsgid, tty);
 610         audit_log_task_info(ab, gfp_mask);
 611         audit_log_end(ab);
 612
 613         for (aux = context->aux; aux; aux = aux->next) {
 614
 615                 ab = audit_log_start(context, gfp_mask, aux->type);
 616                 if (!ab)
 617                         continue; /* audit_panic has been called */
 618
 619                 switch (aux->type) {
 620                 case AUDIT_IPC: {
 621                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 622                         audit_log_format(ab,
 623                                          " qbytes=%lx iuid=%u igid=%u mode=%x obj=%s",
 624                                          axi->qbytes, axi->uid, axi->gid, axi->mode, axi->ctx);
 625                         break; }
 626
 627                 case AUDIT_SOCKETCALL: {
 628                         int i;
 629                         struct audit_aux_data_socketcall *axs = (void *)aux;
 630                         audit_log_format(ab, "nargs=%d", axs->nargs);
 631                         for (i=0; i<axs->nargs; i++)
 632                                 audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
 633                         break; }
 634
 635                 case AUDIT_SOCKADDR: {
 636                         struct audit_aux_data_sockaddr *axs = (void *)aux;
 637
 638                         audit_log_format(ab, "saddr=");
 639                         audit_log_hex(ab, axs->a, axs->len);
 640                         break; }
 641
 642                 case AUDIT_AVC_PATH: {
 643                         struct audit_aux_data_path *axi = (void *)aux;
 644                         audit_log_d_path(ab, "path=", axi->dentry, axi->mnt);
 645                         break; }
 646
 647                 }
 648                 audit_log_end(ab);
 649         }
 650
 651         if (context->pwd && context->pwdmnt) {
 652                 ab = audit_log_start(context, gfp_mask, AUDIT_CWD);
 653                 if (ab) {
 654                         audit_log_d_path(ab, "cwd=", context->pwd, context->pwdmnt);
 655                         audit_log_end(ab);
 656                 }
 657         }
 658         for (i = 0; i < context->name_count; i++) {
 659                 unsigned long ino  = context->names[i].ino;
 660                 unsigned long pino = context->names[i].pino;
 661
 662                 ab = audit_log_start(context, gfp_mask, AUDIT_PATH);
 663                 if (!ab)
 664                         continue; /* audit_panic has been called */
 665
 666                 audit_log_format(ab, "item=%d", i);
 667
 668                 audit_log_format(ab, " name=");
 669                 if (context->names[i].name)
 670                         audit_log_untrustedstring(ab, context->names[i].name);
 671                 else
 672                         audit_log_format(ab, "(null)");
 673
 674                 if (pino != (unsigned long)-1)
 675                         audit_log_format(ab, " parent=%lu",  pino);
 676                 if (ino != (unsigned long)-1)
 677                         audit_log_format(ab, " inode=%lu",  ino);
 678                 if ((pino != (unsigned long)-1) || (ino != (unsigned long)-1))
 679                         audit_log_format(ab, " dev=%02x:%02x mode=%#o"
 680                                          " ouid=%u ogid=%u rdev=%02x:%02x",
 681                                          MAJOR(context->names[i].dev),
 682                                          MINOR(context->names[i].dev),
 683                                          context->names[i].mode,
 684                                          context->names[i].uid,
 685                                          context->names[i].gid,
 686                                          MAJOR(context->names[i].rdev),
 687                                          MINOR(context->names[i].rdev));
 688                 if (context->names[i].ctx) {
 689                         audit_log_format(ab, " obj=%s",
 690                                         context->names[i].ctx);
 691                 }
 692
 693                 audit_log_end(ab);
 694         }
 695 }
 696
 697 /**
 698  * audit_free - free a per-task audit context
 699  * @tsk: task whose audit context block to free
 700  *
 701  * Called from copy_process and __put_task_struct.
 702  */
 703 void audit_free(struct task_struct *tsk)
 704 {
 705         struct audit_context *context;
 706
 707         /*
 708          * No need to lock the task - when we execute audit_free()
 709          * then the task has no external references anymore, and
 710          * we are tearing it down. (The locking also confuses
 711          * DEBUG_LOCKDEP - this freeing may occur in softirq
 712          * contexts as well, via RCU.)
 713          */
 714         context = audit_get_context(tsk, 0, 0);
 715         if (likely(!context))
 716                 return;
 717
 718         /* Check for system calls that do not go through the exit
 719          * function (e.g., exit_group), then free context block.
 720          * We use GFP_ATOMIC here because we might be doing this
 721          * in the context of the idle thread */
 722         if (context->in_syscall && context->auditable)
 723                 audit_log_exit(context, GFP_ATOMIC);
 724
 725         audit_free_context(context);
 726 }
 727
 728 /**
 729  * audit_syscall_entry - fill in an audit record at syscall entry
 730  * @tsk: task being audited
 731  * @arch: architecture type
 732  * @major: major syscall type (function)
 733  * @a1: additional syscall register 1
 734  * @a2: additional syscall register 2
 735  * @a3: additional syscall register 3
 736  * @a4: additional syscall register 4
 737  *
 738  * Fill in audit context at syscall entry.  This only happens if the
 739  * audit context was created when the task was created and the state or
 740  * filters demand the audit context be built.  If the state from the
 741  * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
 742  * then the record will be written at syscall exit time (otherwise, it
 743  * will only be written if another part of the kernel requests that it
 744  * be written).
 745  */
 746 void audit_syscall_entry(struct task_struct *tsk, int arch, int major,
 747                          unsigned long a1, unsigned long a2,
 748                          unsigned long a3, unsigned long a4)
 749 {
 750         struct audit_context *context = tsk->audit_context;
 751         enum audit_state     state;
 752
 753         BUG_ON(!context);
 754
 755         /*
 756          * This happens only on certain architectures that make system
 757          * calls in kernel_thread via the entry.S interface, instead of
 758          * with direct calls.  (If you are porting to a new
 759          * architecture, hitting this condition can indicate that you
 760          * got the _exit/_leave calls backward in entry.S.)
 761          *
 762          * i386     no
 763          * x86_64   no
 764          * ppc64    yes (see arch/powerpc/platforms/iseries/misc.S)
 765          *
 766          * This also happens with vm86 emulation in a non-nested manner
 767          * (entries without exits), so this case must be caught.
 768          */
 769         if (context->in_syscall) {
 770                 struct audit_context *newctx;
 771
 772 #if AUDIT_DEBUG
 773                 printk(KERN_ERR
 774                        "audit(:%d) pid=%d in syscall=%d;"
 775                        " entering syscall=%d\n",
 776                        context->serial, tsk->pid, context->major, major);
 777 #endif
 778                 newctx = audit_alloc_context(context->state);
 779                 if (newctx) {
 780                         newctx->previous   = context;
 781                         context            = newctx;
 782                         tsk->audit_context = newctx;
 783                 } else  {
 784                         /* If we can't alloc a new context, the best we
 785                          * can do is to leak memory (any pending putname
 786                          * will be lost).  The only other alternative is
 787                          * to abandon auditing. */
 788                         audit_zero_context(context, context->state);
 789                 }
 790         }
 791         BUG_ON(context->in_syscall || context->name_count);
 792
 793         if (!audit_enabled)
 794                 return;
 795
 796         context->arch       = arch;
 797         context->major      = major;
 798         context->argv[0]    = a1;
 799         context->argv[1]    = a2;
 800         context->argv[2]    = a3;
 801         context->argv[3]    = a4;
 802
 803         state = context->state;
 804         if (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT)
 805                 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
 806         if (likely(state == AUDIT_DISABLED))
 807                 return;
 808
 809         context->serial     = 0;
 810         context->ctime      = CURRENT_TIME;
 811         context->in_syscall = 1;
 812         context->auditable  = !!(state == AUDIT_RECORD_CONTEXT);
 813 }
 814
 815 /**
 816  * audit_syscall_exit - deallocate audit context after a system call
 817  * @tsk: task being audited
 818  * @valid: success/failure flag
 819  * @return_code: syscall return value
 820  *
 821  * Tear down after system call.  If the audit context has been marked as
 822  * auditable (either because of the AUDIT_RECORD_CONTEXT state from
 823  * filtering, or because some other part of the kernel write an audit
 824  * message), then write out the syscall information.  In call cases,
 825  * free the names stored from getname().
 826  */
 827 void audit_syscall_exit(struct task_struct *tsk, int valid, long return_code)
 828 {
 829         struct audit_context *context;
 830
 831         get_task_struct(tsk);
 832         task_lock(tsk);
 833         context = audit_get_context(tsk, valid, return_code);
 834         task_unlock(tsk);
 835
 836         /* Not having a context here is ok, since the parent may have
 837          * called __put_task_struct. */
 838         if (likely(!context))
 839                 goto out;
 840
 841         if (context->in_syscall && context->auditable)
 842                 audit_log_exit(context, GFP_KERNEL);
 843
 844         context->in_syscall = 0;
 845         context->auditable  = 0;
 846
 847         if (context->previous) {
 848                 struct audit_context *new_context = context->previous;
 849                 context->previous  = NULL;
 850                 audit_free_context(context);
 851                 tsk->audit_context = new_context;
 852         } else {
 853                 audit_free_names(context);
 854                 audit_free_aux(context);
 855                 tsk->audit_context = context;
 856         }
 857  out:
 858         put_task_struct(tsk);
 859 }
 860
 861 /**
 862  * audit_getname - add a name to the list
 863  * @name: name to add
 864  *
 865  * Add a name to the list of audit names for this context.
 866  * Called from fs/namei.c:getname().
 867  */
 868 void audit_getname(const char *name)
 869 {
 870         struct audit_context *context = current->audit_context;
 871
 872         if (!context || IS_ERR(name) || !name)
 873                 return;
 874
 875         if (!context->in_syscall) {
 876 #if AUDIT_DEBUG == 2
 877                 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
 878                        __FILE__, __LINE__, context->serial, name);
 879                 dump_stack();
 880 #endif
 881                 return;
 882         }
 883         BUG_ON(context->name_count >= AUDIT_NAMES);
 884         context->names[context->name_count].name = name;
 885         context->names[context->name_count].ino  = (unsigned long)-1;
 886         ++context->name_count;
 887         if (!context->pwd) {
 888                 read_lock(&current->fs->lock);
 889                 context->pwd = dget(current->fs->pwd);
 890                 context->pwdmnt = mntget(current->fs->pwdmnt);
 891                 read_unlock(&current->fs->lock);
 892         }
 893
 894 }
 895
 896 /* audit_putname - intercept a putname request
 897  * @name: name to intercept and delay for putname
 898  *
 899  * If we have stored the name from getname in the audit context,
 900  * then we delay the putname until syscall exit.
 901  * Called from include/linux/fs.h:putname().
 902  */
 903 void audit_putname(const char *name)
 904 {
 905         struct audit_context *context = current->audit_context;
 906
 907         BUG_ON(!context);
 908         if (!context->in_syscall) {
 909 #if AUDIT_DEBUG == 2
 910                 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
 911                        __FILE__, __LINE__, context->serial, name);
 912                 if (context->name_count) {
 913                         int i;
 914                         for (i = 0; i < context->name_count; i++)
 915                                 printk(KERN_ERR "name[%d] = %p = %s\n", i,
 916                                        context->names[i].name,
 917                                        context->names[i].name ?: "(null)");
 918                 }
 919 #endif
 920                 __putname(name);
 921         }
 922 #if AUDIT_DEBUG
 923         else {
 924                 ++context->put_count;
 925                 if (context->put_count > context->name_count) {
 926                         printk(KERN_ERR "%s:%d(:%d): major=%d"
 927                                " in_syscall=%d putname(%p) name_count=%d"
 928                                " put_count=%d\n",
 929                                __FILE__, __LINE__,
 930                                context->serial, context->major,
 931                                context->in_syscall, name, context->name_count,
 932                                context->put_count);
 933                         dump_stack();
 934                 }
 935         }
 936 #endif
 937 }
 938
 939 void audit_inode_context(int idx, const struct inode *inode)
 940 {
 941         struct audit_context *context = current->audit_context;
 942         const char *suffix = security_inode_xattr_getsuffix();
 943         char *ctx = NULL;
 944         int len = 0;
 945
 946         if (!suffix)
 947                 goto ret;
 948
 949         len = security_inode_getsecurity(inode, suffix, NULL, 0, 0);
 950         if (len == -EOPNOTSUPP)
 951                 goto ret;
 952         if (len < 0)
 953                 goto error_path;
 954
 955         ctx = kmalloc(len, GFP_KERNEL);
 956         if (!ctx)
 957                 goto error_path;
 958
 959         len = security_inode_getsecurity(inode, suffix, ctx, len, 0);
 960         if (len < 0)
 961                 goto error_path;
 962
 963         kfree(context->names[idx].ctx);
 964         context->names[idx].ctx = ctx;
 965         goto ret;
 966
 967 error_path:
 968         if (ctx)
 969                 kfree(ctx);
 970         audit_panic("error in audit_inode_context");
 971 ret:
 972         return;
 973 }
 974
 975
 976 /**
 977  * audit_inode - store the inode and device from a lookup
 978  * @name: name being audited
 979  * @inode: inode being audited
 980  * @flags: lookup flags (as used in path_lookup())
 981  *
 982  * Called from fs/namei.c:path_lookup().
 983  */
 984 void __audit_inode(const char *name, const struct inode *inode, unsigned flags)
 985 {
 986         int idx;
 987         struct audit_context *context = current->audit_context;
 988
 989         if (!context->in_syscall)
 990                 return;
 991         if (context->name_count
 992             && context->names[context->name_count-1].name
 993             && context->names[context->name_count-1].name == name)
 994                 idx = context->name_count - 1;
 995         else if (context->name_count > 1
 996                  && context->names[context->name_count-2].name
 997                  && context->names[context->name_count-2].name == name)
 998                 idx = context->name_count - 2;
 999         else {
1000                 /* FIXME: how much do we care about inodes that have no
1001                  * associated name? */
1002                 if (context->name_count >= AUDIT_NAMES - AUDIT_NAMES_RESERVED)
1003                         return;
1004                 idx = context->name_count++;
1005                 context->names[idx].name = NULL;
1006 #if AUDIT_DEBUG
1007                 ++context->ino_count;
1008 #endif
1009         }
1010         context->names[idx].dev   = inode->i_sb->s_dev;
1011         context->names[idx].mode  = inode->i_mode;
1012         context->names[idx].uid   = inode->i_uid;
1013         context->names[idx].gid   = inode->i_gid;
1014         context->names[idx].rdev  = inode->i_rdev;
1015         audit_inode_context(idx, inode);
1016         if ((flags & LOOKUP_PARENT) && (strcmp(name, "/") != 0) &&
1017             (strcmp(name, ".") != 0)) {
1018                 context->names[idx].ino   = (unsigned long)-1;
1019                 context->names[idx].pino  = inode->i_ino;
1020         } else {
1021                 context->names[idx].ino   = inode->i_ino;
1022                 context->names[idx].pino  = (unsigned long)-1;
1023         }
1024 }
1025
1026 /**
1027  * audit_inode_child - collect inode info for created/removed objects
1028  * @dname: inode's dentry name
1029  * @inode: inode being audited
1030  * @pino: inode number of dentry parent
1031  *
1032  * For syscalls that create or remove filesystem objects, audit_inode
1033  * can only collect information for the filesystem object's parent.
1034  * This call updates the audit context with the child's information.
1035  * Syscalls that create a new filesystem object must be hooked after
1036  * the object is created.  Syscalls that remove a filesystem object
1037  * must be hooked prior, in order to capture the target inode during
1038  * unsuccessful attempts.
1039  */
1040 void __audit_inode_child(const char *dname, const struct inode *inode,
1041                          unsigned long pino)
1042 {
1043         int idx;
1044         struct audit_context *context = current->audit_context;
1045
1046         if (!context->in_syscall)
1047                 return;
1048
1049         /* determine matching parent */
1050         if (dname)
1051                 for (idx = 0; idx < context->name_count; idx++)
1052                         if (context->names[idx].pino == pino) {
1053                                 const char *n;
1054                                 const char *name = context->names[idx].name;
1055                                 int dlen = strlen(dname);
1056                                 int nlen = name ? strlen(name) : 0;
1057
1058                                 if (nlen < dlen)
1059                                         continue;
1060
1061                                 /* disregard trailing slashes */
1062                                 n = name + nlen - 1;
1063                                 while ((*n == '/') && (n > name))
1064                                         n--;
1065
1066                                 /* find last path component */
1067                                 n = n - dlen + 1;
1068                                 if (n < name)
1069                                         continue;
1070                                 else if (n > name) {
1071                                         if (*--n != '/')
1072                                                 continue;
1073                                         else
1074                                                 n++;
1075                                 }
1076
1077                                 if (strncmp(n, dname, dlen) == 0)
1078                                         goto update_context;
1079                         }
1080
1081         /* catch-all in case match not found */
1082         idx = context->name_count++;
1083         context->names[idx].name  = NULL;
1084         context->names[idx].pino  = pino;
1085 #if AUDIT_DEBUG
1086         context->ino_count++;
1087 #endif
1088
1089 update_context:
1090         if (inode) {
1091                 context->names[idx].ino   = inode->i_ino;
1092                 context->names[idx].dev   = inode->i_sb->s_dev;
1093                 context->names[idx].mode  = inode->i_mode;
1094                 context->names[idx].uid   = inode->i_uid;
1095                 context->names[idx].gid   = inode->i_gid;
1096                 context->names[idx].rdev  = inode->i_rdev;
1097                 audit_inode_context(idx, inode);
1098         }
1099 }
1100
1101 /**
1102  * auditsc_get_stamp - get local copies of audit_context values
1103  * @ctx: audit_context for the task
1104  * @t: timespec to store time recorded in the audit_context
1105  * @serial: serial value that is recorded in the audit_context
1106  *
1107  * Also sets the context as auditable.
1108  */
1109 void auditsc_get_stamp(struct audit_context *ctx,
1110                        struct timespec *t, unsigned int *serial)
1111 {
1112         if (!ctx->serial)
1113                 ctx->serial = audit_serial();
1114         t->tv_sec  = ctx->ctime.tv_sec;
1115         t->tv_nsec = ctx->ctime.tv_nsec;
1116         *serial    = ctx->serial;
1117         ctx->auditable = 1;
1118 }
1119
1120 /**
1121  * audit_set_loginuid - set a task's audit_context loginuid
1122  * @task: task whose audit context is being modified
1123  * @loginuid: loginuid value
1124  *
1125  * Returns 0.
1126  *
1127  * Called (set) from fs/proc/base.c::proc_loginuid_write().
1128  */
1129 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
1130 {
1131         if (task->audit_context) {
1132                 struct audit_buffer *ab;
1133
1134                 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1135                 if (ab) {
1136                         audit_log_format(ab, "login pid=%d uid=%u "
1137                                 "old auid=%u new auid=%u",
1138                                 task->pid, task->uid,
1139                                 task->audit_context->loginuid, loginuid);
1140                         audit_log_end(ab);
1141                 }
1142                 task->audit_context->loginuid = loginuid;
1143         }
1144         return 0;
1145 }
1146
1147 /**
1148  * audit_get_loginuid - get the loginuid for an audit_context
1149  * @ctx: the audit_context
1150  *
1151  * Returns the context's loginuid or -1 if @ctx is NULL.
1152  */
1153 uid_t audit_get_loginuid(struct audit_context *ctx)
1154 {
1155         return ctx ? ctx->loginuid : -1;
1156 }
1157
1158 static char *audit_ipc_context(struct kern_ipc_perm *ipcp)
1159 {
1160         struct audit_context *context = current->audit_context;
1161         char *ctx = NULL;
1162         int len = 0;
1163
1164         if (likely(!context))
1165                 return NULL;
1166
1167         len = security_ipc_getsecurity(ipcp, NULL, 0);
1168         if (len == -EOPNOTSUPP)
1169                 goto ret;
1170         if (len < 0)
1171                 goto error_path;
1172
1173         ctx = kmalloc(len, GFP_ATOMIC);
1174         if (!ctx)
1175                 goto error_path;
1176
1177         len = security_ipc_getsecurity(ipcp, ctx, len);
1178         if (len < 0)
1179                 goto error_path;
1180
1181         return ctx;
1182
1183 error_path:
1184         kfree(ctx);
1185         audit_panic("error in audit_ipc_context");
1186 ret:
1187         return NULL;
1188 }
1189
1190 /**
1191  * audit_ipc_perms - record audit data for ipc
1192  * @qbytes: msgq bytes
1193  * @uid: msgq user id
1194  * @gid: msgq group id
1195  * @mode: msgq mode (permissions)
1196  *
1197  * Returns 0 for success or NULL context or < 0 on error.
1198  */
1199 int audit_ipc_perms(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode, struct kern_ipc_perm *ipcp)
1200 {
1201         struct audit_aux_data_ipcctl *ax;
1202         struct audit_context *context = current->audit_context;
1203
1204         if (likely(!context))
1205                 return 0;
1206
1207         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1208         if (!ax)
1209                 return -ENOMEM;
1210
1211         ax->qbytes = qbytes;
1212         ax->uid = uid;
1213         ax->gid = gid;
1214         ax->mode = mode;
1215         ax->ctx = audit_ipc_context(ipcp);
1216
1217         ax->d.type = AUDIT_IPC;
1218         ax->d.next = context->aux;
1219         context->aux = (void *)ax;
1220         return 0;
1221 }
1222
1223 /**
1224  * audit_socketcall - record audit data for sys_socketcall
1225  * @nargs: number of args
1226  * @args: args array
1227  *
1228  * Returns 0 for success or NULL context or < 0 on error.
1229  */
1230 int audit_socketcall(int nargs, unsigned long *args)
1231 {
1232         struct audit_aux_data_socketcall *ax;
1233         struct audit_context *context = current->audit_context;
1234
1235         if (likely(!context))
1236                 return 0;
1237
1238         ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
1239         if (!ax)
1240                 return -ENOMEM;
1241
1242         ax->nargs = nargs;
1243         memcpy(ax->args, args, nargs * sizeof(unsigned long));
1244
1245         ax->d.type = AUDIT_SOCKETCALL;
1246         ax->d.next = context->aux;
1247         context->aux = (void *)ax;
1248         return 0;
1249 }
1250
1251 /**
1252  * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
1253  * @len: data length in user space
1254  * @a: data address in kernel space
1255  *
1256  * Returns 0 for success or NULL context or < 0 on error.
1257  */
1258 int audit_sockaddr(int len, void *a)
1259 {
1260         struct audit_aux_data_sockaddr *ax;
1261         struct audit_context *context = current->audit_context;
1262
1263         if (likely(!context))
1264                 return 0;
1265
1266         ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
1267         if (!ax)
1268                 return -ENOMEM;
1269
1270         ax->len = len;
1271         memcpy(ax->a, a, len);
1272
1273         ax->d.type = AUDIT_SOCKADDR;
1274         ax->d.next = context->aux;
1275         context->aux = (void *)ax;
1276         return 0;
1277 }
1278
1279 /**
1280  * audit_avc_path - record the granting or denial of permissions
1281  * @dentry: dentry to record
1282  * @mnt: mnt to record
1283  *
1284  * Returns 0 for success or NULL context or < 0 on error.
1285  *
1286  * Called from security/selinux/avc.c::avc_audit()
1287  */
1288 int audit_avc_path(struct dentry *dentry, struct vfsmount *mnt)
1289 {
1290         struct audit_aux_data_path *ax;
1291         struct audit_context *context = current->audit_context;
1292
1293         if (likely(!context))
1294                 return 0;
1295
1296         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1297         if (!ax)
1298                 return -ENOMEM;
1299
1300         ax->dentry = dget(dentry);
1301         ax->mnt = mntget(mnt);
1302
1303         ax->d.type = AUDIT_AVC_PATH;
1304         ax->d.next = context->aux;
1305         context->aux = (void *)ax;
1306         return 0;
1307 }
1308
1309 /**
1310  * audit_signal_info - record signal info for shutting down audit subsystem
1311  * @sig: signal value
1312  * @t: task being signaled
1313  *
1314  * If the audit subsystem is being terminated, record the task (pid)
1315  * and uid that is doing that.
1316  */
1317 void audit_signal_info(int sig, struct task_struct *t)
1318 {
1319         extern pid_t audit_sig_pid;
1320         extern uid_t audit_sig_uid;
1321
1322         if (unlikely(audit_pid && t->tgid == audit_pid)) {
1323                 if (sig == SIGTERM || sig == SIGHUP) {
1324                         struct audit_context *ctx = current->audit_context;
1325                         audit_sig_pid = current->pid;
1326                         if (ctx)
1327                                 audit_sig_uid = ctx->loginuid;
1328                         else
1329                                 audit_sig_uid = current->uid;
1330                 }
1331         }
1332 }