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 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
 333                                                       int return_valid,
 334                                                       int return_code)
 335 {
 336         struct audit_context *context = tsk->audit_context;
 337
 338         if (likely(!context))
 339                 return NULL;
 340         context->return_valid = return_valid;
 341         context->return_code  = return_code;
 342
 343         if (context->in_syscall && !context->auditable) {
 344                 enum audit_state state;
 345                 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
 346                 if (state == AUDIT_RECORD_CONTEXT)
 347                         context->auditable = 1;
 348         }
 349
 350         context->pid = tsk->pid;
 351         context->uid = tsk->uid;
 352         context->gid = tsk->gid;
 353         context->euid = tsk->euid;
 354         context->suid = tsk->suid;
 355         context->fsuid = tsk->fsuid;
 356         context->egid = tsk->egid;
 357         context->sgid = tsk->sgid;
 358         context->fsgid = tsk->fsgid;
 359         context->personality = tsk->personality;
 360         tsk->audit_context = NULL;
 361         return context;
 362 }
 363
 364 static inline void audit_free_names(struct audit_context *context)
 365 {
 366         int i;
 367
 368 #if AUDIT_DEBUG == 2
 369         if (context->auditable
 370             ||context->put_count + context->ino_count != context->name_count) {
 371                 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
 372                        " name_count=%d put_count=%d"
 373                        " ino_count=%d [NOT freeing]\n",
 374                        __FILE__, __LINE__,
 375                        context->serial, context->major, context->in_syscall,
 376                        context->name_count, context->put_count,
 377                        context->ino_count);
 378                 for (i = 0; i < context->name_count; i++) {
 379                         printk(KERN_ERR "names[%d] = %p = %s\n", i,
 380                                context->names[i].name,
 381                                context->names[i].name ?: "(null)");
 382                 }
 383                 dump_stack();
 384                 return;
 385         }
 386 #endif
 387 #if AUDIT_DEBUG
 388         context->put_count  = 0;
 389         context->ino_count  = 0;
 390 #endif
 391
 392         for (i = 0; i < context->name_count; i++) {
 393                 char *p = context->names[i].ctx;
 394                 context->names[i].ctx = NULL;
 395                 kfree(p);
 396                 if (context->names[i].name)
 397                         __putname(context->names[i].name);
 398         }
 399         context->name_count = 0;
 400         if (context->pwd)
 401                 dput(context->pwd);
 402         if (context->pwdmnt)
 403                 mntput(context->pwdmnt);
 404         context->pwd = NULL;
 405         context->pwdmnt = NULL;
 406 }
 407
 408 static inline void audit_free_aux(struct audit_context *context)
 409 {
 410         struct audit_aux_data *aux;
 411
 412         while ((aux = context->aux)) {
 413                 if (aux->type == AUDIT_AVC_PATH) {
 414                         struct audit_aux_data_path *axi = (void *)aux;
 415                         dput(axi->dentry);
 416                         mntput(axi->mnt);
 417                 }
 418                 if ( aux->type == AUDIT_IPC ) {
 419                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 420                         if (axi->ctx)
 421                                 kfree(axi->ctx);
 422                 }
 423
 424                 context->aux = aux->next;
 425                 kfree(aux);
 426         }
 427 }
 428
 429 static inline void audit_zero_context(struct audit_context *context,
 430                                       enum audit_state state)
 431 {
 432         uid_t loginuid = context->loginuid;
 433
 434         memset(context, 0, sizeof(*context));
 435         context->state      = state;
 436         context->loginuid   = loginuid;
 437 }
 438
 439 static inline struct audit_context *audit_alloc_context(enum audit_state state)
 440 {
 441         struct audit_context *context;
 442
 443         if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
 444                 return NULL;
 445         audit_zero_context(context, state);
 446         return context;
 447 }
 448
 449 /**
 450  * audit_alloc - allocate an audit context block for a task
 451  * @tsk: task
 452  *
 453  * Filter on the task information and allocate a per-task audit context
 454  * if necessary.  Doing so turns on system call auditing for the
 455  * specified task.  This is called from copy_process, so no lock is
 456  * needed.
 457  */
 458 int audit_alloc(struct task_struct *tsk)
 459 {
 460         struct audit_context *context;
 461         enum audit_state     state;
 462
 463         if (likely(!audit_enabled))
 464                 return 0; /* Return if not auditing. */
 465
 466         state = audit_filter_task(tsk);
 467         if (likely(state == AUDIT_DISABLED))
 468                 return 0;
 469
 470         if (!(context = audit_alloc_context(state))) {
 471                 audit_log_lost("out of memory in audit_alloc");
 472                 return -ENOMEM;
 473         }
 474
 475                                 /* Preserve login uid */
 476         context->loginuid = -1;
 477         if (current->audit_context)
 478                 context->loginuid = current->audit_context->loginuid;
 479
 480         tsk->audit_context  = context;
 481         set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
 482         return 0;
 483 }
 484
 485 static inline void audit_free_context(struct audit_context *context)
 486 {
 487         struct audit_context *previous;
 488         int                  count = 0;
 489
 490         do {
 491                 previous = context->previous;
 492                 if (previous || (count &&  count < 10)) {
 493                         ++count;
 494                         printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
 495                                " freeing multiple contexts (%d)\n",
 496                                context->serial, context->major,
 497                                context->name_count, count);
 498                 }
 499                 audit_free_names(context);
 500                 audit_free_aux(context);
 501                 kfree(context);
 502                 context  = previous;
 503         } while (context);
 504         if (count >= 10)
 505                 printk(KERN_ERR "audit: freed %d contexts\n", count);
 506 }
 507
 508 static void audit_log_task_context(struct audit_buffer *ab)
 509 {
 510         char *ctx = NULL;
 511         ssize_t len = 0;
 512
 513         len = security_getprocattr(current, "current", NULL, 0);
 514         if (len < 0) {
 515                 if (len != -EINVAL)
 516                         goto error_path;
 517                 return;
 518         }
 519
 520         ctx = kmalloc(len, GFP_KERNEL);
 521         if (!ctx)
 522                 goto error_path;
 523
 524         len = security_getprocattr(current, "current", ctx, len);
 525         if (len < 0 )
 526                 goto error_path;
 527
 528         audit_log_format(ab, " subj=%s", ctx);
 529         return;
 530
 531 error_path:
 532         if (ctx)
 533                 kfree(ctx);
 534         audit_panic("error in audit_log_task_context");
 535         return;
 536 }
 537
 538 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
 539 {
 540         char name[sizeof(tsk->comm)];
 541         struct mm_struct *mm = tsk->mm;
 542         struct vm_area_struct *vma;
 543
 544         /* tsk == current */
 545
 546         get_task_comm(name, tsk);
 547         audit_log_format(ab, " comm=");
 548         audit_log_untrustedstring(ab, name);
 549
 550         if (mm) {
 551                 down_read(&mm->mmap_sem);
 552                 vma = mm->mmap;
 553                 while (vma) {
 554                         if ((vma->vm_flags & VM_EXECUTABLE) &&
 555                             vma->vm_file) {
 556                                 audit_log_d_path(ab, "exe=",
 557                                                  vma->vm_file->f_dentry,
 558                                                  vma->vm_file->f_vfsmnt);
 559                                 break;
 560                         }
 561                         vma = vma->vm_next;
 562                 }
 563                 up_read(&mm->mmap_sem);
 564         }
 565         audit_log_task_context(ab);
 566 }
 567
 568 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
 569 {
 570         int i;
 571         struct audit_buffer *ab;
 572         struct audit_aux_data *aux;
 573         const char *tty;
 574
 575         /* tsk == current */
 576
 577         ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
 578         if (!ab)
 579                 return;         /* audit_panic has been called */
 580         audit_log_format(ab, "arch=%x syscall=%d",
 581                          context->arch, context->major);
 582         if (context->personality != PER_LINUX)
 583                 audit_log_format(ab, " per=%lx", context->personality);
 584         if (context->return_valid)
 585                 audit_log_format(ab, " success=%s exit=%ld",
 586                                  (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
 587                                  context->return_code);
 588         if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
 589                 tty = tsk->signal->tty->name;
 590         else
 591                 tty = "(none)";
 592         audit_log_format(ab,
 593                   " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
 594                   " pid=%d auid=%u uid=%u gid=%u"
 595                   " euid=%u suid=%u fsuid=%u"
 596                   " egid=%u sgid=%u fsgid=%u tty=%s",
 597                   context->argv[0],
 598                   context->argv[1],
 599                   context->argv[2],
 600                   context->argv[3],
 601                   context->name_count,
 602                   context->pid,
 603                   context->loginuid,
 604                   context->uid,
 605                   context->gid,
 606                   context->euid, context->suid, context->fsuid,
 607                   context->egid, context->sgid, context->fsgid, tty);
 608         audit_log_task_info(ab, tsk);
 609         audit_log_end(ab);
 610
 611         for (aux = context->aux; aux; aux = aux->next) {
 612
 613                 ab = audit_log_start(context, GFP_KERNEL, aux->type);
 614                 if (!ab)
 615                         continue; /* audit_panic has been called */
 616
 617                 switch (aux->type) {
 618                 case AUDIT_IPC: {
 619                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 620                         audit_log_format(ab,
 621                                          " qbytes=%lx iuid=%u igid=%u mode=%x obj=%s",
 622                                          axi->qbytes, axi->uid, axi->gid, axi->mode, axi->ctx);
 623                         break; }
 624
 625                 case AUDIT_SOCKETCALL: {
 626                         int i;
 627                         struct audit_aux_data_socketcall *axs = (void *)aux;
 628                         audit_log_format(ab, "nargs=%d", axs->nargs);
 629                         for (i=0; i<axs->nargs; i++)
 630                                 audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
 631                         break; }
 632
 633                 case AUDIT_SOCKADDR: {
 634                         struct audit_aux_data_sockaddr *axs = (void *)aux;
 635
 636                         audit_log_format(ab, "saddr=");
 637                         audit_log_hex(ab, axs->a, axs->len);
 638                         break; }
 639
 640                 case AUDIT_AVC_PATH: {
 641                         struct audit_aux_data_path *axi = (void *)aux;
 642                         audit_log_d_path(ab, "path=", axi->dentry, axi->mnt);
 643                         break; }
 644
 645                 }
 646                 audit_log_end(ab);
 647         }
 648
 649         if (context->pwd && context->pwdmnt) {
 650                 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
 651                 if (ab) {
 652                         audit_log_d_path(ab, "cwd=", context->pwd, context->pwdmnt);
 653                         audit_log_end(ab);
 654                 }
 655         }
 656         for (i = 0; i < context->name_count; i++) {
 657                 unsigned long ino  = context->names[i].ino;
 658                 unsigned long pino = context->names[i].pino;
 659
 660                 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
 661                 if (!ab)
 662                         continue; /* audit_panic has been called */
 663
 664                 audit_log_format(ab, "item=%d", i);
 665
 666                 audit_log_format(ab, " name=");
 667                 if (context->names[i].name)
 668                         audit_log_untrustedstring(ab, context->names[i].name);
 669                 else
 670                         audit_log_format(ab, "(null)");
 671
 672                 if (pino != (unsigned long)-1)
 673                         audit_log_format(ab, " parent=%lu",  pino);
 674                 if (ino != (unsigned long)-1)
 675                         audit_log_format(ab, " inode=%lu",  ino);
 676                 if ((pino != (unsigned long)-1) || (ino != (unsigned long)-1))
 677                         audit_log_format(ab, " dev=%02x:%02x mode=%#o"
 678                                          " ouid=%u ogid=%u rdev=%02x:%02x",
 679                                          MAJOR(context->names[i].dev),
 680                                          MINOR(context->names[i].dev),
 681                                          context->names[i].mode,
 682                                          context->names[i].uid,
 683                                          context->names[i].gid,
 684                                          MAJOR(context->names[i].rdev),
 685                                          MINOR(context->names[i].rdev));
 686                 if (context->names[i].ctx) {
 687                         audit_log_format(ab, " obj=%s",
 688                                         context->names[i].ctx);
 689                 }
 690
 691                 audit_log_end(ab);
 692         }
 693 }
 694
 695 /**
 696  * audit_free - free a per-task audit context
 697  * @tsk: task whose audit context block to free
 698  *
 699  * Called from copy_process and do_exit
 700  */
 701 void audit_free(struct task_struct *tsk)
 702 {
 703         struct audit_context *context;
 704
 705         context = audit_get_context(tsk, 0, 0);
 706         if (likely(!context))
 707                 return;
 708
 709         /* Check for system calls that do not go through the exit
 710          * function (e.g., exit_group), then free context block.
 711          * We use GFP_ATOMIC here because we might be doing this
 712          * in the context of the idle thread */
 713         /* that can happen only if we are called from do_exit() */
 714         if (context->in_syscall && context->auditable)
 715                 audit_log_exit(context, tsk);
 716
 717         audit_free_context(context);
 718 }
 719
 720 /**
 721  * audit_syscall_entry - fill in an audit record at syscall entry
 722  * @tsk: task being audited
 723  * @arch: architecture type
 724  * @major: major syscall type (function)
 725  * @a1: additional syscall register 1
 726  * @a2: additional syscall register 2
 727  * @a3: additional syscall register 3
 728  * @a4: additional syscall register 4
 729  *
 730  * Fill in audit context at syscall entry.  This only happens if the
 731  * audit context was created when the task was created and the state or
 732  * filters demand the audit context be built.  If the state from the
 733  * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
 734  * then the record will be written at syscall exit time (otherwise, it
 735  * will only be written if another part of the kernel requests that it
 736  * be written).
 737  */
 738 void audit_syscall_entry(int arch, int major,
 739                          unsigned long a1, unsigned long a2,
 740                          unsigned long a3, unsigned long a4)
 741 {
 742         struct task_struct *tsk = current;
 743         struct audit_context *context = tsk->audit_context;
 744         enum audit_state     state;
 745
 746         BUG_ON(!context);
 747
 748         /*
 749          * This happens only on certain architectures that make system
 750          * calls in kernel_thread via the entry.S interface, instead of
 751          * with direct calls.  (If you are porting to a new
 752          * architecture, hitting this condition can indicate that you
 753          * got the _exit/_leave calls backward in entry.S.)
 754          *
 755          * i386     no
 756          * x86_64   no
 757          * ppc64    yes (see arch/powerpc/platforms/iseries/misc.S)
 758          *
 759          * This also happens with vm86 emulation in a non-nested manner
 760          * (entries without exits), so this case must be caught.
 761          */
 762         if (context->in_syscall) {
 763                 struct audit_context *newctx;
 764
 765 #if AUDIT_DEBUG
 766                 printk(KERN_ERR
 767                        "audit(:%d) pid=%d in syscall=%d;"
 768                        " entering syscall=%d\n",
 769                        context->serial, tsk->pid, context->major, major);
 770 #endif
 771                 newctx = audit_alloc_context(context->state);
 772                 if (newctx) {
 773                         newctx->previous   = context;
 774                         context            = newctx;
 775                         tsk->audit_context = newctx;
 776                 } else  {
 777                         /* If we can't alloc a new context, the best we
 778                          * can do is to leak memory (any pending putname
 779                          * will be lost).  The only other alternative is
 780                          * to abandon auditing. */
 781                         audit_zero_context(context, context->state);
 782                 }
 783         }
 784         BUG_ON(context->in_syscall || context->name_count);
 785
 786         if (!audit_enabled)
 787                 return;
 788
 789         context->arch       = arch;
 790         context->major      = major;
 791         context->argv[0]    = a1;
 792         context->argv[1]    = a2;
 793         context->argv[2]    = a3;
 794         context->argv[3]    = a4;
 795
 796         state = context->state;
 797         if (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT)
 798                 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
 799         if (likely(state == AUDIT_DISABLED))
 800                 return;
 801
 802         context->serial     = 0;
 803         context->ctime      = CURRENT_TIME;
 804         context->in_syscall = 1;
 805         context->auditable  = !!(state == AUDIT_RECORD_CONTEXT);
 806 }
 807
 808 /**
 809  * audit_syscall_exit - deallocate audit context after a system call
 810  * @tsk: task being audited
 811  * @valid: success/failure flag
 812  * @return_code: syscall return value
 813  *
 814  * Tear down after system call.  If the audit context has been marked as
 815  * auditable (either because of the AUDIT_RECORD_CONTEXT state from
 816  * filtering, or because some other part of the kernel write an audit
 817  * message), then write out the syscall information.  In call cases,
 818  * free the names stored from getname().
 819  */
 820 void audit_syscall_exit(int valid, long return_code)
 821 {
 822         struct task_struct *tsk = current;
 823         struct audit_context *context;
 824
 825         context = audit_get_context(tsk, valid, return_code);
 826
 827         if (likely(!context))
 828                 return;
 829
 830         if (context->in_syscall && context->auditable)
 831                 audit_log_exit(context, tsk);
 832
 833         context->in_syscall = 0;
 834         context->auditable  = 0;
 835
 836         if (context->previous) {
 837                 struct audit_context *new_context = context->previous;
 838                 context->previous  = NULL;
 839                 audit_free_context(context);
 840                 tsk->audit_context = new_context;
 841         } else {
 842                 audit_free_names(context);
 843                 audit_free_aux(context);
 844                 tsk->audit_context = context;
 845         }
 846 }
 847
 848 /**
 849  * audit_getname - add a name to the list
 850  * @name: name to add
 851  *
 852  * Add a name to the list of audit names for this context.
 853  * Called from fs/namei.c:getname().
 854  */
 855 void audit_getname(const char *name)
 856 {
 857         struct audit_context *context = current->audit_context;
 858
 859         if (!context || IS_ERR(name) || !name)
 860                 return;
 861
 862         if (!context->in_syscall) {
 863 #if AUDIT_DEBUG == 2
 864                 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
 865                        __FILE__, __LINE__, context->serial, name);
 866                 dump_stack();
 867 #endif
 868                 return;
 869         }
 870         BUG_ON(context->name_count >= AUDIT_NAMES);
 871         context->names[context->name_count].name = name;
 872         context->names[context->name_count].ino  = (unsigned long)-1;
 873         ++context->name_count;
 874         if (!context->pwd) {
 875                 read_lock(&current->fs->lock);
 876                 context->pwd = dget(current->fs->pwd);
 877                 context->pwdmnt = mntget(current->fs->pwdmnt);
 878                 read_unlock(&current->fs->lock);
 879         }
 880
 881 }
 882
 883 /* audit_putname - intercept a putname request
 884  * @name: name to intercept and delay for putname
 885  *
 886  * If we have stored the name from getname in the audit context,
 887  * then we delay the putname until syscall exit.
 888  * Called from include/linux/fs.h:putname().
 889  */
 890 void audit_putname(const char *name)
 891 {
 892         struct audit_context *context = current->audit_context;
 893
 894         BUG_ON(!context);
 895         if (!context->in_syscall) {
 896 #if AUDIT_DEBUG == 2
 897                 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
 898                        __FILE__, __LINE__, context->serial, name);
 899                 if (context->name_count) {
 900                         int i;
 901                         for (i = 0; i < context->name_count; i++)
 902                                 printk(KERN_ERR "name[%d] = %p = %s\n", i,
 903                                        context->names[i].name,
 904                                        context->names[i].name ?: "(null)");
 905                 }
 906 #endif
 907                 __putname(name);
 908         }
 909 #if AUDIT_DEBUG
 910         else {
 911                 ++context->put_count;
 912                 if (context->put_count > context->name_count) {
 913                         printk(KERN_ERR "%s:%d(:%d): major=%d"
 914                                " in_syscall=%d putname(%p) name_count=%d"
 915                                " put_count=%d\n",
 916                                __FILE__, __LINE__,
 917                                context->serial, context->major,
 918                                context->in_syscall, name, context->name_count,
 919                                context->put_count);
 920                         dump_stack();
 921                 }
 922         }
 923 #endif
 924 }
 925
 926 void audit_inode_context(int idx, const struct inode *inode)
 927 {
 928         struct audit_context *context = current->audit_context;
 929         const char *suffix = security_inode_xattr_getsuffix();
 930         char *ctx = NULL;
 931         int len = 0;
 932
 933         if (!suffix)
 934                 goto ret;
 935
 936         len = security_inode_getsecurity(inode, suffix, NULL, 0, 0);
 937         if (len == -EOPNOTSUPP)
 938                 goto ret;
 939         if (len < 0)
 940                 goto error_path;
 941
 942         ctx = kmalloc(len, GFP_KERNEL);
 943         if (!ctx)
 944                 goto error_path;
 945
 946         len = security_inode_getsecurity(inode, suffix, ctx, len, 0);
 947         if (len < 0)
 948                 goto error_path;
 949
 950         kfree(context->names[idx].ctx);
 951         context->names[idx].ctx = ctx;
 952         goto ret;
 953
 954 error_path:
 955         if (ctx)
 956                 kfree(ctx);
 957         audit_panic("error in audit_inode_context");
 958 ret:
 959         return;
 960 }
 961
 962
 963 /**
 964  * audit_inode - store the inode and device from a lookup
 965  * @name: name being audited
 966  * @inode: inode being audited
 967  * @flags: lookup flags (as used in path_lookup())
 968  *
 969  * Called from fs/namei.c:path_lookup().
 970  */
 971 void __audit_inode(const char *name, const struct inode *inode, unsigned flags)
 972 {
 973         int idx;
 974         struct audit_context *context = current->audit_context;
 975
 976         if (!context->in_syscall)
 977                 return;
 978         if (context->name_count
 979             && context->names[context->name_count-1].name
 980             && context->names[context->name_count-1].name == name)
 981                 idx = context->name_count - 1;
 982         else if (context->name_count > 1
 983                  && context->names[context->name_count-2].name
 984                  && context->names[context->name_count-2].name == name)
 985                 idx = context->name_count - 2;
 986         else {
 987                 /* FIXME: how much do we care about inodes that have no
 988                  * associated name? */
 989                 if (context->name_count >= AUDIT_NAMES - AUDIT_NAMES_RESERVED)
 990                         return;
 991                 idx = context->name_count++;
 992                 context->names[idx].name = NULL;
 993 #if AUDIT_DEBUG
 994                 ++context->ino_count;
 995 #endif
 996         }
 997         context->names[idx].dev   = inode->i_sb->s_dev;
 998         context->names[idx].mode  = inode->i_mode;
 999         context->names[idx].uid   = inode->i_uid;
1000         context->names[idx].gid   = inode->i_gid;
1001         context->names[idx].rdev  = inode->i_rdev;
1002         audit_inode_context(idx, inode);
1003         if ((flags & LOOKUP_PARENT) && (strcmp(name, "/") != 0) &&
1004             (strcmp(name, ".") != 0)) {
1005                 context->names[idx].ino   = (unsigned long)-1;
1006                 context->names[idx].pino  = inode->i_ino;
1007         } else {
1008                 context->names[idx].ino   = inode->i_ino;
1009                 context->names[idx].pino  = (unsigned long)-1;
1010         }
1011 }
1012
1013 /**
1014  * audit_inode_child - collect inode info for created/removed objects
1015  * @dname: inode's dentry name
1016  * @inode: inode being audited
1017  * @pino: inode number of dentry parent
1018  *
1019  * For syscalls that create or remove filesystem objects, audit_inode
1020  * can only collect information for the filesystem object's parent.
1021  * This call updates the audit context with the child's information.
1022  * Syscalls that create a new filesystem object must be hooked after
1023  * the object is created.  Syscalls that remove a filesystem object
1024  * must be hooked prior, in order to capture the target inode during
1025  * unsuccessful attempts.
1026  */
1027 void __audit_inode_child(const char *dname, const struct inode *inode,
1028                          unsigned long pino)
1029 {
1030         int idx;
1031         struct audit_context *context = current->audit_context;
1032
1033         if (!context->in_syscall)
1034                 return;
1035
1036         /* determine matching parent */
1037         if (dname)
1038                 for (idx = 0; idx < context->name_count; idx++)
1039                         if (context->names[idx].pino == pino) {
1040                                 const char *n;
1041                                 const char *name = context->names[idx].name;
1042                                 int dlen = strlen(dname);
1043                                 int nlen = name ? strlen(name) : 0;
1044
1045                                 if (nlen < dlen)
1046                                         continue;
1047
1048                                 /* disregard trailing slashes */
1049                                 n = name + nlen - 1;
1050                                 while ((*n == '/') && (n > name))
1051                                         n--;
1052
1053                                 /* find last path component */
1054                                 n = n - dlen + 1;
1055                                 if (n < name)
1056                                         continue;
1057                                 else if (n > name) {
1058                                         if (*--n != '/')
1059                                                 continue;
1060                                         else
1061                                                 n++;
1062                                 }
1063
1064                                 if (strncmp(n, dname, dlen) == 0)
1065                                         goto update_context;
1066                         }
1067
1068         /* catch-all in case match not found */
1069         idx = context->name_count++;
1070         context->names[idx].name  = NULL;
1071         context->names[idx].pino  = pino;
1072 #if AUDIT_DEBUG
1073         context->ino_count++;
1074 #endif
1075
1076 update_context:
1077         if (inode) {
1078                 context->names[idx].ino   = inode->i_ino;
1079                 context->names[idx].dev   = inode->i_sb->s_dev;
1080                 context->names[idx].mode  = inode->i_mode;
1081                 context->names[idx].uid   = inode->i_uid;
1082                 context->names[idx].gid   = inode->i_gid;
1083                 context->names[idx].rdev  = inode->i_rdev;
1084                 audit_inode_context(idx, inode);
1085         }
1086 }
1087
1088 /**
1089  * auditsc_get_stamp - get local copies of audit_context values
1090  * @ctx: audit_context for the task
1091  * @t: timespec to store time recorded in the audit_context
1092  * @serial: serial value that is recorded in the audit_context
1093  *
1094  * Also sets the context as auditable.
1095  */
1096 void auditsc_get_stamp(struct audit_context *ctx,
1097                        struct timespec *t, unsigned int *serial)
1098 {
1099         if (!ctx->serial)
1100                 ctx->serial = audit_serial();
1101         t->tv_sec  = ctx->ctime.tv_sec;
1102         t->tv_nsec = ctx->ctime.tv_nsec;
1103         *serial    = ctx->serial;
1104         ctx->auditable = 1;
1105 }
1106
1107 /**
1108  * audit_set_loginuid - set a task's audit_context loginuid
1109  * @task: task whose audit context is being modified
1110  * @loginuid: loginuid value
1111  *
1112  * Returns 0.
1113  *
1114  * Called (set) from fs/proc/base.c::proc_loginuid_write().
1115  */
1116 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
1117 {
1118         if (task->audit_context) {
1119                 struct audit_buffer *ab;
1120
1121                 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1122                 if (ab) {
1123                         audit_log_format(ab, "login pid=%d uid=%u "
1124                                 "old auid=%u new auid=%u",
1125                                 task->pid, task->uid,
1126                                 task->audit_context->loginuid, loginuid);
1127                         audit_log_end(ab);
1128                 }
1129                 task->audit_context->loginuid = loginuid;
1130         }
1131         return 0;
1132 }
1133
1134 /**
1135  * audit_get_loginuid - get the loginuid for an audit_context
1136  * @ctx: the audit_context
1137  *
1138  * Returns the context's loginuid or -1 if @ctx is NULL.
1139  */
1140 uid_t audit_get_loginuid(struct audit_context *ctx)
1141 {
1142         return ctx ? ctx->loginuid : -1;
1143 }
1144
1145 static char *audit_ipc_context(struct kern_ipc_perm *ipcp)
1146 {
1147         struct audit_context *context = current->audit_context;
1148         char *ctx = NULL;
1149         int len = 0;
1150
1151         if (likely(!context))
1152                 return NULL;
1153
1154         len = security_ipc_getsecurity(ipcp, NULL, 0);
1155         if (len == -EOPNOTSUPP)
1156                 goto ret;
1157         if (len < 0)
1158                 goto error_path;
1159
1160         ctx = kmalloc(len, GFP_ATOMIC);
1161         if (!ctx)
1162                 goto error_path;
1163
1164         len = security_ipc_getsecurity(ipcp, ctx, len);
1165         if (len < 0)
1166                 goto error_path;
1167
1168         return ctx;
1169
1170 error_path:
1171         kfree(ctx);
1172         audit_panic("error in audit_ipc_context");
1173 ret:
1174         return NULL;
1175 }
1176
1177 /**
1178  * audit_ipc_perms - record audit data for ipc
1179  * @qbytes: msgq bytes
1180  * @uid: msgq user id
1181  * @gid: msgq group id
1182  * @mode: msgq mode (permissions)
1183  *
1184  * Returns 0 for success or NULL context or < 0 on error.
1185  */
1186 int audit_ipc_perms(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode, struct kern_ipc_perm *ipcp)
1187 {
1188         struct audit_aux_data_ipcctl *ax;
1189         struct audit_context *context = current->audit_context;
1190
1191         if (likely(!context))
1192                 return 0;
1193
1194         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1195         if (!ax)
1196                 return -ENOMEM;
1197
1198         ax->qbytes = qbytes;
1199         ax->uid = uid;
1200         ax->gid = gid;
1201         ax->mode = mode;
1202         ax->ctx = audit_ipc_context(ipcp);
1203
1204         ax->d.type = AUDIT_IPC;
1205         ax->d.next = context->aux;
1206         context->aux = (void *)ax;
1207         return 0;
1208 }
1209
1210 /**
1211  * audit_socketcall - record audit data for sys_socketcall
1212  * @nargs: number of args
1213  * @args: args array
1214  *
1215  * Returns 0 for success or NULL context or < 0 on error.
1216  */
1217 int audit_socketcall(int nargs, unsigned long *args)
1218 {
1219         struct audit_aux_data_socketcall *ax;
1220         struct audit_context *context = current->audit_context;
1221
1222         if (likely(!context))
1223                 return 0;
1224
1225         ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
1226         if (!ax)
1227                 return -ENOMEM;
1228
1229         ax->nargs = nargs;
1230         memcpy(ax->args, args, nargs * sizeof(unsigned long));
1231
1232         ax->d.type = AUDIT_SOCKETCALL;
1233         ax->d.next = context->aux;
1234         context->aux = (void *)ax;
1235         return 0;
1236 }
1237
1238 /**
1239  * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
1240  * @len: data length in user space
1241  * @a: data address in kernel space
1242  *
1243  * Returns 0 for success or NULL context or < 0 on error.
1244  */
1245 int audit_sockaddr(int len, void *a)
1246 {
1247         struct audit_aux_data_sockaddr *ax;
1248         struct audit_context *context = current->audit_context;
1249
1250         if (likely(!context))
1251                 return 0;
1252
1253         ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
1254         if (!ax)
1255                 return -ENOMEM;
1256
1257         ax->len = len;
1258         memcpy(ax->a, a, len);
1259
1260         ax->d.type = AUDIT_SOCKADDR;
1261         ax->d.next = context->aux;
1262         context->aux = (void *)ax;
1263         return 0;
1264 }
1265
1266 /**
1267  * audit_avc_path - record the granting or denial of permissions
1268  * @dentry: dentry to record
1269  * @mnt: mnt to record
1270  *
1271  * Returns 0 for success or NULL context or < 0 on error.
1272  *
1273  * Called from security/selinux/avc.c::avc_audit()
1274  */
1275 int audit_avc_path(struct dentry *dentry, struct vfsmount *mnt)
1276 {
1277         struct audit_aux_data_path *ax;
1278         struct audit_context *context = current->audit_context;
1279
1280         if (likely(!context))
1281                 return 0;
1282
1283         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1284         if (!ax)
1285                 return -ENOMEM;
1286
1287         ax->dentry = dget(dentry);
1288         ax->mnt = mntget(mnt);
1289
1290         ax->d.type = AUDIT_AVC_PATH;
1291         ax->d.next = context->aux;
1292         context->aux = (void *)ax;
1293         return 0;
1294 }
1295
1296 /**
1297  * audit_signal_info - record signal info for shutting down audit subsystem
1298  * @sig: signal value
1299  * @t: task being signaled
1300  *
1301  * If the audit subsystem is being terminated, record the task (pid)
1302  * and uid that is doing that.
1303  */
1304 void audit_signal_info(int sig, struct task_struct *t)
1305 {
1306         extern pid_t audit_sig_pid;
1307         extern uid_t audit_sig_uid;
1308
1309         if (unlikely(audit_pid && t->tgid == audit_pid)) {
1310                 if (sig == SIGTERM || sig == SIGHUP) {
1311                         struct audit_context *ctx = current->audit_context;
1312                         audit_sig_pid = current->pid;
1313                         if (ctx)
1314                                 audit_sig_uid = ctx->loginuid;
1315                         else
1316                                 audit_sig_uid = current->uid;
1317                 }
1318         }
1319 }